Volume 2 Interfacial Kinetics and Mass Transport

3
1.1
The CurrentPotential Relationship
section of this volume covers kinetics and

mechanisms of selected electrochemical
processes.
Ernesto J. Calvo
Universidad de Buenos Aires, Buenos Aires,
Argentina
1.1.2
1.1.1
Scope
This rst chapter to Volume 2 Interfacial

Kinetics and Mass Transport introduces
the following sections, with particular focus on the distinctive feature of electrode
reactions, namely, the exponential currentpotential relationship, which reects
the strong effect of the interfacial electric
eld on the kinetics of chemical reactions
at electrode surfaces. We then analyze the
consequence of this accelerating effect on
the reaction kinetics upon the surface concentration of reactants and products and
the role played by mass transport on the
currentpotential curves. The theory of
electron-transfer reactions, migration, and
diffusion processes and digital simulation of convective-diffusion are analyzed
in the rst four chapters. New experimental evidence of mechanistic aspects in
electrode kinetics from different in-situ
spectroscopies and structural studies are
discussed in the second section. The last
Introduction
The currentpotential relationship at electrodes is set by a number of complex physical and chemical phenomena, depending on the experimental
conditions.
Electrode reactions are heterogeneous
chemical processes that may involve one
or more electron-transfer steps across the
electrochemical double layer [1, 2]. Electrode reactions provide a switch for charge
to ow between phases of different type of
electrical conductivity: electrodes and electrolyte [3]. Therefore, their response can
be analyzed either on the basis of electrical
or chemical models. The distinctive feature of reactions at electrodes is the strong
dependence of both the surface concentrations and the kinetics on the electrode
potential [410].
The following reactions at electried
interfaces are examples of electrode reactions:
Feaq 2+ Feaq 3+ + e(metal)
Cu Cuaq
2+
(1)
+ 2e(metal) (2)
Fundamentals
Xaq Xorg
(3)
The rst reaction is an electron transfer

across the double layer at the electrodeelectrolyte interface between redox
species in the electrolyte that exchange
electrons with a metal electrode, the second one is an ion-transfer reaction across
the double layer since the electron lost
by the Cu atom remains at the metal.
The third one is an ion-transfer process
across the waterorganic solvent interface
or ion transfer at immiscible electrolyte
solutions (ITIES) without the transfer of
electrons. In all cases the electrochemical reaction takes place at an electried interface and therefore the rate of
these reactions follow similar exponential dependence on the interfacial electrical
potential.
While electrodes can be metals, carbon or semiconductors, and so on and
the charge carriers are electrons or holes,
the charge carriers in liquid and solid
electrolytes are solvated or nonsolvated
ions in molten salts and solid electrolytes.
Therefore the ow of electricity is accomplished by the electrode reactions at the
interface between these electrode and electrolyte phases.
The relation between the total quantity of electricity, q, and the number of
moles of each chemical being transformed
during the electrolysis has been established by Michael Faraday in the eighteen
century
q = nF N
(4)
where n is the number of charges
transferred per molecule or ion, N the
number of moles and q the total quantity
of electricity passed through the interface
and thus through the external circuit.
Differentiating the amount of charge
with respect to time yields the current,
which is proportional to the rate of the

electrochemical reaction (in moles per
second):
I = nF
dN
dt
(5)
Since we are dealing with heterogeneous

reactions, we need to normalize the
total current dividing by the electrode
area, A and thus the current density
is i = I /A. It should be noticed that
the current measures an average value
of the reaction rate over the electrode
area, which is important to keep in
mind given our present knowledge of
heterogeneities at electrode surfaces in
micro- and nanoscale. (At present it is
possible to study electrode surfaces in
situ in contact with electrolytes and under
potential control using scanning probe
microscopes with spatial resolution in the
micro and nanometer.)
In 1905, Tafel [11] described the experimental relationship between current, I ,
and the overpotential the difference between the working electrode potential, E,
and the equilibrium potential Ee during
the electrochemical reduction of protons to
molecular hydrogen on different electrode
materials such as Hg, Sn, Bi, Au, Cu, Ni,
and so on:
= a + b log I
(6)
Figure 1 depicts a reproduction of the

original currentpotential curves reported
by Julius Tafel for different metals, which
demonstrates the electrocatalytic nature of
this reaction [11].
Electrode reactions are heterogeneous
chemical reactions in solution and may
include elementary electron-transfer steps,
ion transfer, potential-independent, or
chemical steps, and so on.
1.1 The CurrentPotential Relationship

Fig. 1 Steady state
overpotential versus log I curves
for the hydrogen evolution on
different metals. (Taken from
Ref. [11].)
Hg
2.0
Sn
1.8
Bi
Au
1.6
Cu
Ni
1.4
1.2
1.0
0.0
0.00
0.05
0.10
0.15
log I
1.1.3
Thermodynamics of Electrochemical Cells
electrons transferred from the electrode,

the standard free-energy change for the
overall reaction with hydrogen:
The electrode potential can be expressed by

E = (M S ) + Const
(7)
where M is the inner or Galvani potential

of the metal phase and S the inner or
Galvani potential of the electrolyte solution phase. In electrokinetic experiments,
the electrode potential is measured with
respect to a reference electrode and therefore an undetermined constant adds to the
electrode potential [10].
If we consider the generic electroreduction half-reaction:
O + ne
(8)
where O is the soluble oxidized and

R the reduced species present in the
electrolyte and n is the number of
O+
n
H2 = R + nH+
2
(9)
is given by
G = nF E = RT ln Keq
(10)
where E is the standard potential of

the O/R redox couple with respect to
the standard hydrogen reference electrode
(SHE) and Keq is the equilibrium constant
for Reaction (9).
The dependence of the electrode potential on the activities of O and R at
equilibrium can be described by the Nernst
equation:

RT
ao
(11)
ln
Ee = E +
nF
aR
Fundamentals
where aO and aR denote the activities of O

and R, respectively. The use of formal or

conditional equilibrium potentials, E , is
more convenient in electrode kinetics [2]
since concentrations instead of activities
can be used, but it must be kept in mind
that activity coefcients are included in the
rate coefcients.
The steady state currentpotential relationship under electrode kinetics conditions has been described by Butler [12, 13]
and Volmer [14].
1.1.4
The ButlerVolmer Equation
In this section, we shall derive an expression for the currentpotential relationship

in electrode reactions. Since multielectrontransfer reactions at interfaces are most
likely to occur in a series of one-electron
steps, we shall consider rst the general
one-electron electrode reaction:
kf
O+e
(12)
kb
the rate of which is considered to be

totally determined by the charge-transfer
process, and mass-transport limitations
or chemical step limitations will not be
considered here so that the concentrations
of reactant and product at the interface will
be the same as the analytical concentration
in the bulk of the electrolyte.
The net rate of Reaction (12) will be given
by the difference between the forward and
backward rates:
v = vf vb
(13)
which, for reaction order one for reactant

and product, can be expressed as
=
vf = kf CO
ic
F
(14)
and,
vb = kb CR =
ia
F
(15)
where kf and kb are the corresponding

rate coefcients and ic and ia the corresponding formal cathodic and anodic
partial current densities that ow at the
electrodeelectrolyte interface. These partial currents cannot be measured since
we cannot distinguish in the external circuit those electrons owing towards the
electrolyte from those owing towards the
electrode, and therefore we can only detect
the net current, I = ic ia . This difference
implies that we must adopt a convention
of signs for the reduction and oxidation
currents [2].
The net reaction rate is
v = vf vb =
i
F
(16)
According to the absolute reaction rate

theory of Eyring, rst applied to electrode
reactions, the rate coefcient of a chemical
reaction, k, can be expressed as

=
Gi
kB T
(17)
exp
ki =
h
RT
where Gi is the free energy of activation for the reaction, kB the Boltzmann
constant, h the Plank constant and the
transmission coefcient which for simplicity we shall consider here equal to one.
In the case of electrode reactions that occur at an electried interface and therefore
under the inuence of a strong electrical eld (c.a. 107 V cm1 ), the driving
force for the reaction of charged species
=
Gi has a term of electrostatic work
wel = F E as shown in the scheme of
Fig. 2. The negative sign corresponds to
the negatively charged electron crossing
the electrical double layer.
B
A
G f (E)
FE
aFE
(1 a)FE
A
G b (E)
Gf (Ee)
FE
O + e
Fig. 2 Schematic representation of the effect of electrode potential

on the free energy versus reaction coordinate curves for an electron
reactant at two electrode potentials: E = Ee (solid line) and E < Ee
(broken line).
For convenience, we divide the free

energy of activation into a potentialindependent and a potential-dependent
contribution:
Gf (E) = Gf (Ee ) + F E
(18)
and,
=
Gb (E)
It is convenient to dene

=
Gf (Ee )
kB T
kf ,o =
exp
h
RT
and
kb,o
=
Gb (Ee ) (1 )F E
(19)
where 0 0 is the fraction of the
applied potential that drives the reaction
and a measure of the symmetry of the
energy barrier.
Replacing in Eq. (17):

=
Gf (Ee )
kB T
kf =
exp
h
RT

F E
exp
(20)
RT

=
Gb (Ee )
kB T
exp
kb =
h
RT

(1 )F E
exp
(21)
RT
(22)

=
Gb (Ee )
kB T
=
exp
h
RT
And therefore,
kf (E) = kf ,o exp
F E
RT
(23)

and
(1 )F E
kb (E) = kb,o exp
RT

(24)
Thus from Eq. (16), the net current density

at the electrode interface is

F E
i = F kf ,o CO exp
kb,o CR
RT

(1 )F E
exp
(25)
RT
which has an exponential dependence on
the electrode potential, E, as observed in
experiments.
Fundamentals
We note that at equilibrium i = 0

and E = Ee and therefore from Eq. (25)
we obtain

kf ,o
RT
CO
ln
+ ln
(26)
Ee =
F
kb,o
CR
which is identical to the Nernst equation for the reaction between O and R
at the electrode surface. Note that the rst
term in Eq. (26) is the conditional or formal standard electrode potential for the

reaction, E , which contains the nonelectrical terms of the free energy of activation
including the activity coefcients for the
species O and R.
We can group all rate coefcients by
substituting E = Ee in Eqs. (24 and 26)
and multiplying by F and obtain the
exchange current density, io :

E
1
io = F kf ,o CO CR exp F
RT

E
1
= F kb,o CO CR exp (1 )F
RT
(27)
we can further dene a standard state both
for the electrode potential E = Ee and
the reactant and product concentrations
CO = CR = 1 M:

E
ko = kf ,o exp F
RT

E
= kb,o exp (1 )F
(28)
RT
1
CR . ko is the
and thus, io = F ko CO
standard rate constant for unit surface
concentration of reactants at the standard
equilibrium potential.
Substituting in Eq. (25) and introducing
the overpotential for the reaction as
= (E Ee ) and substituting io into
Eq. (27), we obtain the ButlerVolmer

equation (29):

F
(1 )F
i = io exp
exp
RT
RT
(29)
named after Butler and Volmer, the
pioneers in this eld. The validity of
the ButlerVolmer equation has been
veried by isotopic labeling [8]. It relates
the current density that ows through
the electrodeelectrolyte interface due to
the electrode reaction to the overpotential
in terms of two kinetic parameters: the
exchange current density, io , and the
transfer coefcient, .
The exchange current density introduced by Butler [12, 13] is a measure of
the exchange rate between O and R at
the electrodeelectrolyte interface at equilibrium and measures the height of the
activation barrier when both the reactant
and the product of an electrode reaction
are at the same free-energy level (as shown
in Fig. 2).
The transfer coefcient, , introduced
in electrode kinetics by ErdeyGruz and
Volmer [14] for the hydrogen evolution
reaction (HER), measures the symmetry of
the free-energy curves at their intersection
in the transition state. Figure 3 depicts a
plot of ln i versus according to Eq. (29).
Rewriting the ButlerVolmer equation,
one obtains an expression for the overpotential that explains the experimental
ndings of Tafel:
2.303aRT
F
i
+ log io

log
1 exp
RT
(30)
log ( j/jo)
0
1
2
150
100
50
50
100
150
E Eo
Fig. 3
Plots of log |i/io | versus calculated with Eq. (29).
The Tafel relationship is a linear freeenergy relation of the rate coefcient for a
net electrode reaction (neglecting the back
reaction) and the transfer coefcient:

=
1 Gf
(31)
=
F
E
Ee
and since E = G /F , the Tafel dependence is a linear free-energy relationship

between the kinetics and the thermodynamic driving force like the Bronsted
relationship between the rate of proton
transfer and the acid pKa in acidbase
catalysis [10].
At low overpotentials ( RT /nF ),
which fullls for 510 mV excursion
around the equilibrium potential, the
exponential expressions in Eq. (29) can be
expanded in series, and the rst series
term in the expansion gives
nF
i
= io
RT
(32)
a linear i relationship for the ow of

charge at the interface near equilibrium,
which resembles the ohmic current

potential behavior observed in conductor
phases (metals and electrolytes). Notice
that the slope of the current potential plot
near equilibrium is independent of .
By analogy with Ohms law, the chargetransfer resistance, Rt , can be dened
as a particular case of the polarization
resistance: Rp = (d/di)i0 :
Rt =
RT 1
nF io
(33)
which is a measure of io and is also

independent of .
It should be noted that in deriving
Eqs. (18 and 19), only the rst term of the
series expansion of the free-energy curves
has been taken into account (i.e. the reactants and products parabolas have been
considered linear at the crossing point) and
therefore results independent of the electrode potential; this is not the case when
microscopic theories like the Marcus theory based on a parabolic description are
introduced. Thus, a limitation of the ButlerVolmer currentpotential relation is
10
Fundamentals
that it cannot account for the curvature

of with the electrode potential far from
the equilibrium potential and does not
incorporate the temperature dependence
of k o explicitly.
1.1.5
Theory of Electron Transfer
Rudolf Marcus was Chemistry Nobel laureate in 1992 for his work on the theory
of homogeneous and heterogeneous electron transfer (ET) in such diverse processes
as the bioenergetic of photosynthesis and
respiration, homogeneous redox and heterogeneous electrode reactions, corrosion,
color photography, and so on. Marcus theory provides an insight on the physical
and chemical phenomena of the heavy
particle subsystem during electron transfer in redox reactions. Furthermore, it can
accommodate disparate rates over many
orders of magnitude for soluble redox
species, intramolecular electron transfer
in biological reactions of utmost importance, and a common framework for a
comparison of redox processes in solution
and at electrode surfaces.
In its simplest form the Marcus expression for the electron-transfer rate coefcient is given by [1520]:

G 2
1+
kET = Z exp
(34)
4RT
with the reorganization free energy
of the reactants to yield products and
G the standard free energy of the
reaction (thermodynamic driving force)
and Z a frequency factor taken by Marcus
as 104 cm1 for heterogeneous electrode
reactions [19]. The reorganization free
energy, comprises the solvation, o , and
the vibrational, i contributions:

= o + i
(35)
Assuming the solvent as a dielectric

continuum, the outer sphere component,
o , can be expressed in terms of the ionic
radii a1 and a2 of the inner coordination
shell (a2 for reactions at electrodes),
the center-to-center separation distance of
the reactants R, the optical (op ) and static
(s ) dielectric constants of the solvent and
the charge transferred (e) between the
reactants:

1
1
1
+
+
o = (e)2
2a1
2a2
R

1
1
(36)
op
s
The inner sphere component, i is
given by
i =
1
p
kj (qjr qj )2
2
(37)
where qjr and qj are the equilibrium

values for the j th normal mode coordinate
q and kj is a reduced force constant,
p
p
kj = 2kjr kj /(kjr + kj ).
For interfacial electrode reactions, the
free energy of activation is thus given by
=
Gi

F (E E ) 2
=
1+
4
(38)
since G = FE .
The Marcus free-energy relation predicts
a nonlinear dependence of the transfer
coefcient, on electrode potential:
1 F (E E )
1 Gi
= +
F E
2
2

w o wR
(39)
2
where wo and wR are the electrostatic

work terms due to the electrical work
involved in bringing charged reactants at
the reaction site or a charged reactant close
to the charged electrode surface (double
layer or Frumkin effects, see below).
Notice that while the ButlerVolmer
equation predicts = 0.5 independent
of the electrode potential, Marcus theory
predicts curvature dependent on electrode
potential and reorganization energy. Only
when |G /2| 1, the Marcus theory
predicts = 0.5, which corresponds to low
polarization and slow reactions.
The Marcus theory also allows to compare homogeneous electron-transfer reactions with the same reactions at electrode surfaces and cross-reaction rates
in terms of the individual self-exchange
reactions, that is, 1,2
= 1/2(1,1 + 2,2 ).
Experimental verication over 20 orders of
magnitude has been the greatest success
of this theory.
Another, very impressive prediction of
the Marcus theory is the inverted region,
coined by Marcus in 1960 [17]: For a series of related reactions of similar but
different driving force, G , as G becomes negative the activation free energy,
G
= , rst decreases as one would expect in a Bronsted free-energy plot for
acid-base catalyzed reactions, and reaches
G
= = 0 when G = . Then the activation energy increases again when G
becomes even more negative, leading to
an inversion in the plot of logarithm of
kET versus G . An experimental test of
this inverted region was provided in 1984
by Miller, Calcaterra, and Closs for electron transfer between a biphenyl group to
an acceptor center [21]. The difculty with
detecting the inverted region has been for
many years the mass transport limit in very
fast redox reactions unless the electron
donor and acceptor are positioned at xed
distance like in intramolecular electron

transfer. The rst evidence of the inverted
Marcus region at a polarized interface was
provided by the study of ET at liquidliquid
interfaces with an adsorbed C-10 lipid layer
by scanning electrochemical microscopy
(SECM) [2225] and is described in
Chapter 4.2.
Since the appearance of the Marcus
theory, for many years a clear potential
dependence of with potential has been
difcult to prove experimentally with
soluble redox species reacting at electrodes
because the diffusion limit is reached at
high ET transfer rates where one would
expect the quadratic term to become
important.
For heterogeneous electron transfer,
the use of ordered organic monolayers
(self-assembled monolayers or SAMs)
at electrode surfaces as blocking lms
either with electroactive species in the
electrolyte [26] or with electroactive groups
tethered at the opposite end of the blocking
molecule from the covalent attachment
end [27] has provided a method to study
the effect of electrode-redox center distance
and the effect of the electrode potential on
the electron-transfer rate.
The effect of tunneling distance in heterogeneous electron transfer at electrode
surfaces has been shown by using blocking organized thin lms (thickness less
than 1.5 nm) where the electrons can tunnel through the lms in the absence of
defects or pinholes that would otherwise
allow the access of electroactive species in
direct contact with the electrode surface.
For -hydroxy-alkylthiol monolayers of different number of methylene units Miller
and coworkers [26] showed that the ET
heterogeneous rate constant, kET , decays
exponentially with the number of methylene groups in the alkylthiol, n, which is
equivalent to the distance from the redox
11
Fundamentals
species to the electrode, x:

kET (x) = ko (x = 0)e
(40)
where ko is the preexponential rate coefcient and the tunneling coefcient,

assumed potential independent, and close
to 1 A 1 [28, 29] for saturated alkylthiols
in analogy to intramolecular through-bond
electron transfer, and 0.40.6 A 1 for conjugated molecular spacers. Electron
tunnel factors in ET through molecules
and molecular interfaces is considered
in Chapter 1.3 and a similar exponential
decay has been observed in tunneling
spectroscopy for tip-to-molecule distance
as described in Chapter 3.3. Ratner and
Whitesides [30] have recently introduced a
simple experimental procedure to measure
Fe
S
S
S
S
S
AuAu Au Au Au Au Au AuAu
(a)
105
rates of electron transport across organic

thin lms with a range of structures by using a Metal-SAM(1)SAM(2)-Metal junction
using a mercury drop covered with C-16
alklylthiol. The current measured across
these junctions obeys an exponential relation with a decay constant dependent
on the molecular structure. Similar studies have been reported by Schiffrin [31],
Lindsay [32], and Bard [33] with scanning
tunneling microscopy (STM) examination
of SAMs at electrode surfaces.
For ferrocene sites at the end of long
alkanethiols self-organized at gold electrodes and diluted with unsubstituted
thiols with the redox moiety in contact
with the electrolyte (Fig. 4a), Chidsey has
reported [34] curved Tafel plots (Fig. 4b),
which could be tted by equations derived from Marcus theory with values of
= 0.85 eV and Z = 6.73 104 s1 eV1

for a reaction rate of k = 2.5 s1 at E in
Fig. 4(b). Similar curvature in Tafel plots
has been reported by Faulkner and coworkers [35] for adsorbed osmium complexes at
ultramicro-electrodes (UME). The temperature dependence of the rate coefcient
could also be tted from Marcus equation
and electron states in the metal and coupling factors given by quantum mechanics.
At high driving forces in Fig. 4(b), a
plateau is reached instead of the inverted region predicted by Marcus theory
due to the continuum of states in the
104
kf + k b
[s1]
12
103
102
101
100
1.0
(b)
0.5
0.0
E Eo
[V]
0.5
1.0
Fig. 4 (a) Schematic representation of

thiolated gold with ferrocene covalently
attached to the alkylthiol end opposite
to the S-Au bond. (b) Log (kf + kb )
versus (E Eo ) for the
oxidationreduction of terminal
ferrocene sites at SAM modied gold
electrode (taken from Ref. [34]). The
dotted line corresponds to calculation
with the ButlerVolmer equation (29),
the dashed line calculated curves from
MarcusLevichDogonazde theory.

Fig. 5 Log of current density versus
potential curve of 0.1 M ZnSO4 at 0.6
atom % Zn amalgam electrode with
50 M [(C4 H9 )4 N]2 SO4 . Current densities were obtained from radiolabeled 65 Zn amalgam. (Taken from Ref. [8].)
1.1.6
Ion-Transfer Reactions
Electrode reactions also characterized by exponential currentpotential
log i
[a cm2]
metal electrode unlike the inversion found

for electron-transfer rates between two
molecular states. The electronic states
at the metal surface are considered in
the LevichDogonadze theory [58], which
takes into account the electronic states
overlapping between the electrode and redox electrolyte and also describes the temperature dependence (see Chapter 1.2).
Gerischer developed a quantum mechanics treatment for redox systems at metal
and semiconductor electrodes, where the
difference between metal and semiconductor electrodes is the integration over
the electronic states in the electrode as
in the LevichDogonadze theory [59]. In
Gerischer model the electronic levels in
the redox electrolyte are oxidized species
(unoccupied states) and reduced species
(occupied states) with a Gaussian distribution. The electronic overlap of these
redox states with the states in the semiconductor or metal electrode describe the
different contributions to the exchange
current density and the currentpotential
dependence [3]. The currentpotential relationship for electrode reactions at semiconductor electrodes present asymmetric
curves due to the overlapping of the
electronic structure of the semiconductor
(electron or hole density of states) and the
electronic states in the redox electrolyte.
Zn Hgx/Zn2+
i(calculated)
i (experimental)
6
i+ (from Zn55)
0.80
0.75
0.70 V
0.65
eh
relationship are interfacial ion processes

that can take place at
1. solidliquid interface like in metal
dissolution in liquid electrolyte
2. liquidliquid interfaces like the interfaces between two immiscible electrolyte solutions (ITIES).
During the dissolution of a metal for instance, metal cations belonging to the
metal crystallographic lattice can reach
the surface and become ad-atoms, a term
coined by Lorenz [38] to describe metal
atoms adsorbed on surfaces. These adsorbed cations can loose electrons into
the metal surface electronic plasmon and
cross the electrical double layer to become
solvated cations in the electrolyte. A very
important role is the solvation process
13
14
Fundamentals
that stabilizes the cation in solution. Note,

however, that there is no electron transfer across the electrical double layer in
the dissolution of a metal. Likewise, at
ITIES not only electrons can be transferred
across the liquidliquid electrolyte boundary from donor redox states in one liquid
phase to an acceptor redox states in the
other, and vice versa; but also ion transfer at ITIES is possible with resolvation of
the ion being transferred as described in
Chapter 4.2. The currentpotential curve
is also exponential in that case and can be
described in a similar way to ion transfer
at solidliquid interfaces with surmounting an activation barrier by a Boltzman
population of species with enough energy.
Gerischer [39] considered the process
described by
M(metal) M
z+
(solvated)
+ ze(metal)
(41)
where a metal ion at the surface metal

lattice loses an electron (which remains
in the metal) and crosses the double
layer to become a solvated cation in the
electrolyte. For this process an exponential
currentpotential curve is observed, which
can be derived by considering the anodic
(i+ ) and cathodic (i ) partial current
densities proportional to the concentration
of the reacting species and to the Boltzman
factor eEi/kT :
Ei

i+ = k+ cM cT exp
RT

= k+ cM cT exp
and
zF
E
RT
E

i
i = k co exp
RT

(42)

= k co exp
(1 )zF
E
RT

(43)
therefore the total current that ows at the

interface is given by

zF
i = k+ cM cT exp
E
RT

(1 )zF
k co exp
E (44)
RT
or dening an exchange current density,
io and overpotential, :

zF
i = io exp
RT

(1 )zF
exp
(45)
RT
a typical ion-transfer electrode reaction
is the dissolution of zinc amalgam into
solvated Zn2+ cations in the aqueous
electrolyte:
Zn(Hg) Zn2+ (aq) + 2e(Hg) (46)
For this reaction a Tafel plot is shown in
Fig. 5 as predicted by Eq. (45). A kinetic
analysis of this data yields for the trans
fer coefcient: = RT /zF ln io /o =
0.72 for the charge-transfer valence z = 2.
The free-energy barrier for the ow of
ionic charge across the oxideelectrolyte
interface has an electrical contribution and
consequently the currentpotential curve
can be described by a ButlerVolmer type
equation [8].
The ion-transfer reactions at the oxideelectrolyte interface:
Mz+ (oxide) Mz+ (electrolyte) (47)
and,
O2 (oxide) + 2H+ (electrolyte)
H2 O(electrolyte)
(48)
are also examples of interfacial ion-transfer

reactions. Under steady state conditions,
the net current at the interface is given
by the sum of cation and anion partial
ionic currents. Deposition and dissolution
of the oxide requires that both cations and
anions ow in the same direction while the
corresponding electrical currents are opposite [39]. The steady state potential of the
oxide electrode at a particular composition
MOn lies between the formal equilibrium
potential with respect to oxygen and metal
electrodes as required by the electroneutrality condition [40].
The transfer of ions across immiscible electrolyte interfaces (ITIES) can be
described by a similar ButlerVolmer formalism and involves resolvation of the
transferred ions. The ux of ions can be
measured by the electronic current in the
external circuit generated at symmetrical
reference electrodes reversible to one of
the ions.
1.1.7
Multiple Electron Transfer
In general, reactions at electrodes involving more than one electron-transfer events

occur in successive steps with participation
of intermediaries, which in most cases are
adsorbed at the electrode surface. For instance, consider
reorganization in the metal-ligand vibration scale, 1013 s or slower, the probability

of simultaneous electron transfer in a single step is extremely low. Therefore, two
one-electron steps with the stabilization
of an intermediate state are more likely
to occur than a single two-electron step.
Figure 6 depicts the free-energy prole for
a two-step electron-transfer electrode reaction, where the relative heights of the
activation barriers for the elementary steps
is differently inuenced by the electrode
potential. At E1 , the rate-determining step
corresponds to the rst electron transfer
while at E2 , the rds is the second electron
transfer and at E both transition states
have the same activation barrier (k2 = k1 )
and therefore there is a shift in the position
of the transition state from the rst to the
second electron-transfer process.
Note that both rate constants k1 and k2
may have different kinetics with io,1 and 1
and io,2 and 2 respectively and a general
expression for a sequential two-electron
reaction has been derived by Vetter [8]:

i = 2io,1 exp
1 F
RT
Cuaq 2+ + e Cuads + + e Cu

2F
1 exp
RT

io,1
(1 + 2 1 )F
1+
exp
io,2
RT
(51)

where = E E and E is the standard
electrode potential for the overall twoelectron process. In Fig. 7 we represent
the Tafel plot for the current potential
relationship described by Eq. (51).
We can consider two limiting cases:
(50)
Since the time scale for the electrontransfer event, ca. 1016 s is much smaller
than the time scale of the fastest chemical
1. The rst electron-transfer step is rds,

then k2 k1 in Eq. (51), and the
forward rate coefcient kf k1 and
k1
H+ (aq) + e Hads (metal)

k2
+ H+ (aq) + e H2
(49)
or,
k1
k2
15
Fundamentals
Standard free-energy diagram
for a two-step electron-transfer
electrode reaction with two potential
dependent transition states.
Fig. 6
=/ 1
E1
E*
=/ 2
E2
the backward rate coefcient kb

(k2 /k2 )/k1 .
2. Second electron-transfer step in
Eq. (51) being rds. k2 k1 then kf
(k1 /k1 )k2 and, kb k2 .
1
2
k1
k2
k2 = k1
k2
i0.2
Log i
16
k1
i0.1
i0
Eo2
E*E o
Eo1
The relative height of each individual

activation barrier changes with electrode
potential, as illustrated in Fig. 6.
As depicted in Fig. 7, at potentials more
positive than E , k2 > k1 and the rst
electron transfer is rds while for E < E ,
k1 > k2 and the second electron transfer
becomes rate-determining. The Tafel slope
in the rst case is 2RT/F and 3RT /2F for
the preequilibrium in the second case.
In multiple-electron electrode reactions
extrapolation of Tafel lines from high
overpotential regions to reversible potential for the overall reaction Eo , does not
yield the same apparent standard rate coefcient (exchange current density) and
may be misleading: io,2 is larger than
io,1 , the exchange current density for the
kinetics that change more rapidly with
potential. This feature distinguishes an
n-electron sequential multistep reaction
from a single n-electron charge-transfer
step [8].
The rate-determining step, rds, for
a given mechanism may change with
potential, but this does not violate the
principle of microscopic reversibility since
inspection of Fig. 7 shows that, at any
potential, the transition state is always
the same for the forward and backward
reactions in Eq. (51). Transition states at
different electrode potentials, on the other
hand, need not be the same (see Fig. 6).
Furthermore, there is no reason, why the
potential E , determined by kinetic factors,
Fig. 7 Schematic representation of the
variation of the individual rate
coefcients ki (broken lines) and the
overall rate coefcient (solid line) for a
two-electron electrode reaction as a
function of the electrode potential.
at which the transition of rds occurs,

should be the same as the thermodynamic
equilibrium potential Ee .
In Fig. 7, each transition state can be
identied and is characterized by io,i , Eo,j
and j ; io,i measures the barrier height
of each transition state in the respective
standard conditions, whereas at the formal

standard potential, E , for the overall
reaction, kf = kb and this potential value is

half way between E1 and E2 . The change
in transition states with electrode potential
is observed, for instance, in the complex
oxygen electroreduction (4-electron and 4proton) reaction, the mechanism of which
may also change with pH.
Stoichiometric Number
The concept of stoichiometric number
was introduced in electrochemistry by
Horiuti and Ikusima [42, 43] for the
hydrogen electroreduction reaction. We
need to introduce the stoichiometric number in complex multielectron electrode kinetics in order to distinguish
different possible mechanisms. The International Union of Pure and Applied Chemistry (IUPAC) denes the
stoichiometric number in electrochemistry as a positive integer that indicates the number of identical activated
complexes formed and destroyed in the
completion of the overall reaction as
formulated with the charge number,
n [44, 45]
A general expression can be derived for
the electrode reaction:
1.1.7.1
aA + bB + . . . .ne = cC + dD + . . . .
(52)
which takes place in several consecutive
intermediate steps, the rate of one of
them controlling the overall kinetics.
The reactants and products in the ratedetermining step (rds) will be identied by
R and P , respectively.
We shall assume that the completion of
the overall process represented by Eq. (52)
requires the formation and decomposition
of identical activated complexes.
According to Parsons the rate-determining step can be represented [46] by
a
b
n
c
d
A + B + .... e = C + D + ....
(53)
The stoichiometric number indicates the
number of times the rate-determining
step occurs in the overall stoichiometric
reaction. For example, the Tafel mechanism for the reduction of H+ on metals
occurs in two steps with Reaction (54) occurring twice for each time Reaction (56)
takes place.

(54)
2 H+ + e Had
2Hads H2
(55)
where Had indicates an adsorbed hydrogen

atom (ad-atom) on the metal electrode.
If the overall reaction is written in
the form:
2H+ + 2e = H2
(56)
= 2 and n = 2; but if the reaction is

represented as
H+ + e =
1
H2
2
(57)
then, = 1 and n = 1.
It is worthwhile noticing that, while
and n are arbitrary quantities that depend
on the way we write the stoichiometric
equation, the electron number n/ is
characteristic of the electrode reaction
kinetics. In the Tafel hydrogen reduction
mechanism, n/ = 1 and indicates the
17
18
Fundamentals
number of electrons transferred for the

completion of the rate-determining step.
The forward and backward reaction rate
coefcients can be expressed by Eq. (17)
with G
= , the standard free energy of
formation of one mole of activated complex
from reactants in Eq. (52), and should be
replaced by the electrochemical free energy
=
of activation, Gi , for charged particles.
Gf = R
(58)
=
Gb
(59)
= P
The standard electrochemical potentials of

reactants and products may be written
in terms of the chemical and electrical
contributions:
a
b
A + B
n n
+ e + F M
c
d
n
P = C + D + F S
R =
nF E
(64)
=
where Gi represents the nonelectrical
contribution to the standard free energy of
activation and E = (M S ) + const. It
should be noted that the right-hand side of
Eqs. (63 and 64) should be divided by the
activity coefcient of the activated complex,
which could be potential dependent [46].
Introducing the overpotential, = E
Ee , and replacing

nF
i = io exp
RT

(1 )n
(65)
exp
and the exchange current density, io
(60)
(61)
In order to evaluate the potential dependence of the free energy of activation,

without knowledge of the structure of the
activated complex, it is assumed that the
electrical contribution to the standard free
energy of the transition state lies between
that of the standard free energy of P and
that to the standard free energy of R in
the rds. The symmetry factor or transfer
coefcient, , for the rds in analogy to the
case for one-electron reactions is given by
= +

=
Gb
(1 ) nF E
kB T
kf =
exp
+
h
RT
RT
(62)
which after substitution in Eq. (17) with

the free energy of activation given by
Eqs. (58 and 59) is

=
Gf
kB T
nF E
(63)
exp
+
kb =
h
RT
RT
io =
1
1
n
F ko [A] [B] [C] [D]
(66)
for a rst-order reaction.

For || RT /nF , the Tafel approximation is valid and

RT
=
ln i T,P,ci
nF
and
ln i

=
T,P,ci
RT
(1 )nF
(67)
and therefore the stoichiometric number,

, can be obtained from anodic and
cathodic Tafel plots,

1
ln i
ln i
nF
=
RT
E a
E c
(68)
if the rate-determining step is the same
over the whole range of potentials from
which Tafel slopes are obtained. Note
that from a Tafel plot we can only
determine the product n/, not the
quantities separately.
For small overpotentials, || RT /

nF in the linear polarization region, an
explicit expression for the stoichiometric
number can be found:

nF io
nF io
=
=
RT ic =0
RT
ia =0
(69)
which is valid if the reversible potential is
far from the pzc, otherwise double-layer
effects are important (see Sect. 1.1.10).
1.1.8
Mass Transport at Electrodes
In electrode reactions, the reactant has to

nd the electrode surface where electrons
are taken or released, and therefore
the mass transport of reactants and
products becomes very important in the
description of electrode reactions where
we will be interested in currenttime and
currentpotential relations. The master
equation for mass transport to an electrode
surface is the NernstPlanck equation:

zi F
+ ci
Ji = Di ci + ci
RT
(70)
which accounts for the ux of species i
(mol cm2 s1 ) at a given distance from
the electrode surface, Di is the diffusion
coefcient of the diffusing electroactive
species (cm2 s1 ). The rst term on the
right hand side of Eq. (70) represents the
transport mechanism by diffusion in a
concentration gradient, ci , by random

thermal motion of the particles, which
tends towards uniform concentration of
mobile species. The second and third term
on the right hand side describe, respectively, migration under the inuence of the
electric eld, E = , and convection

under the inuence of a ow eld, with the
uid velocity
(cm s1 ) and the particles
acquire a component of their velocity along

the direction of the eld. These effects
will be treated in detail in Chapter 2.1:
Diffusion and migration. However, in
electrokinetic experiments, the migration
contribution can be reduced by using an inert supporting electrolyte at a much higher
concentration than that of the electroactive
species so that the transport number of the
electroactive species is negligible. Forced
convection combined with diffusion can be
controlled with hydrodynamic electrodes
with a good mathematical description of
the ow pattern and thus of the steady
state currentpotential relation. [9].
c
= D 2 c Vx c
t
(71)
The solution of this partial differential

second-order equation depends on the
initial and boundary conditions of the
particular experiment, giving rise to a
multitude of techniques. In Chapter 2.2,
the digital simulation of voltammetry under stagnant and hydrodynamic conditions
is described. By changing the electrode
potential one can modify the boundary
conditions and transient effects arise until
a new steady state is reached.
The steady state diffusion is reached
when
c
=0
(72)
t
This condition can be accomplished when
the concentration gradient close to the
electrode surface is constant (linear) or
its curvature is compensated by a term X:
c
= D 2 c X
t
(73)
where X can arise from the diffusion eld

geometry, that is, nonlinear terms like in
cylindrical or spherical electrodes:
Xspherical =
2 c
D
r r
(74)
19
20
Fundamentals
or,
Xcylindrical =
1 c
D
r r
(75)
or approximations to these terms in ultra

microelectrodes where the electrode size is
smaller or comparable to the characteristic
diffusion length at a given time, namely,
r (2Dt).
Two other sources of a term X that
can compensate for the curvature of the
concentration gradient at the electrode
,
surface are a convective term X = ci
where is the uid velocity or a chemical

reaction (not involving a charge-transfer
step) term X = k1 ci coupled to the chargetransfer reaction, that is,
k1
AH
A +H
(76)
k1
followed by the electron-transfer reaction:

1
H (aq) + e H2 (g)
2
+
is to be determined from the limiting

convective-diffusion current.
According to the classical treatment by
Randles [47] for a simple electroreduction
of O to R in solution and assuming,
for simplicity, that the mass transport
rate coefcients, kd,i for the oxidized
and reduced species are the same, the
net current density under steady state
conditions is
i = nF kd co cos = nF kd cRs cR
(79)
with the diffusion-limiting current densities for the anodic and cathodic processes
given by.
(80)
iL = nF kd ci
A diffusion layer of thickness , which
has a purely formal signicance, can be
dened as
=
(77)
The diffusion equation is

cAH
2 CH
k1 cHA + k1 cA cH+
= DH
t
x 2
(78)
which has a steady state at the electrode
characterized by a reaction layer, xR =
(DH /k1 cA ), a distance over which a

proton once formed from dissociation of
HA has a chance to diffuse before reacting
with A [4].
In most practical applications, where the
maximum yield of a product or electricity
in electrochemical energy conversion systems is desired, the rate of mass transport
should be fast enough in order not to limit
the overall rate of the process. Conversely,
in electroanalytical techniques, such as polarography or gas sensing, the reaction
is limited by the transport of the reactant since the analyte bulk concentration
D
(nF DAci )
=
Kd,i
i
(81)
In Eq. (25) we have considered the surface

concentration of the reacting species equal
to the analytical concentration in the bulk
electrolyte. If we replace in Eq. (25) the
surface concentrations from Eqs. (7981):

i
ci (0, t)
= 1
(82)
ci
iL
we obtain

i
nF
i
= 1
exp
io
iL , c
RT

i
(1 )nF
1
exp
iL , a
RT
(83)
Figure 8 depicts a plot of the currentpotential curve calculated with
Eq. (83), i/iL versus and the corresponding normalized concentration proles at
the limiting current and at equilibrium.
The mass transport free ButlerVolmer
C0/C*0
i/iL
ButlerVolmer
id,a
C0/C*0
C0/C*0
id, c
X
1
X
Plot of i/iL versus calculated with Eq. (83) (solid line) and
ButlerVolmer currentpotential curve calculated with Eq. (29) for
= 0.5 and n = 1. Inset: concentration proles of the oxidized species
normalized to the analytical concentration at limiting current and at
equilibrium.
Fig. 8
predicted currentpotential curve calculated with Eq. (29) (dashed line) is shown
for comparison.
We can reexamine Eq. (83), both in the
Tafel region, for || RT /nF :

i
nF
log i = log io + 1
exp
iL,c
RT
(84)
and for the linear polarization region, for
|| RT/nF:
nF
1

1
1
RT 1
+
+
io
iL,c
iL,a
(85)
which in the absence of concentration
polarization is coincident with Eq. (32).
Expressing Eq. (84) in terms of rate
coefcient rather than current densities
and rearranging terms, the rate coefcient obtained experimentally at constant
i=
potential, kobs :
kobs =
i
nF ci
(86)
can be separated into the electrode reaction

apparent rate coefcient for extrapolated
innitely fast mass transport conditions, k,
and the mass transport rate coefcient, kd :
1
1
1
= +
kobs
k
kd
(87)
Heterogeneous electrode reactions can be

compared with homogeneous kinetics in
solution, with regard to mass transport.
The second-order rate coefcient for a
fast homogeneous reactions in solution,
k(hom), which would be observed if diffusion were innitely fast, can be related
to the measured rate coefcient, kobs (hom)
by application of Ficks rst law in a spherical continuum diffusion eld around
the reacting molecule. At a collision distance rAB , this corresponds to the average
21
22
Fundamentals
concentration of molecules undergoing

encounters equal to the average analytical
concentration in the bulk solution [48].
kobs (hom) =
4rAB DAB
4rAB DAB
1+
k(hom)
(88)
where rAB is the encounter distance of

molecules A and B in solution and DAB is
the relative diffusion coefcient.
Rewriting Eq. (88) with kd (hom) =
4rAB DAB NA where NA is Avogadros
number:
1
1
1
=
+
(89)
kobs (hom)
k(hom) kd (hom)
The analogy in the mass transport effects
in electrode reaction and homogeneous
second-order fast reactions in solution
becomes clear. In electrode kinetics, however, the charge-transfer rate coefcient
can be externally varied over many orders
of magnitude through the electrode potential and kd can be controlled by means of
hydrodynamic electrodes. For instance the
mass transport rate coefcient, kd , for a rotating disc electrode at the maximum practical rotation speed of 10 000 per min1 is
approximately 2 102 cm s1 .
1.1.9
Electrochemical Reaction Order
The order of an electrochemical reaction is

an empirical factor widely used in chemical
kinetics, which may give an insight into the
events at the molecular level for kinetics of
complex reactions. It relates the reaction
rate to the concentration of a particular
species in the kinetic equation.
In electrode reactions, the reaction order
with respect to the species k can be
dened by

pk =
log ij
log ck

(90)
ci
=k ,E
where ij is a partial anodic or cathodic

current density and ck the concentration
of species k.
Because of the strong dependence of the
kinetics on potential, the determination
of the electrochemical reaction order
requires that the partial anodic or cathodic
current densities be measured at constant
potential in addition to the concentration
of the other species being kept constant.
Vetter used extensively the concept of
reaction order, for instance, as diagnostic
of the mechanism of the electrode reaction:
Mn(IV) + e Mn(III)
(91)
on platinum electrode [8], which looks

simple at rst sight. However the reaction
orders found experimentally are PMn(IV) =
0 and PMn(III) = 1, which suggests the
following mechanism:
Mn(III) + e Mn(II)
Mn(II) + Mn(IV) 2Mn(III)
(92)
with the rst electron transfer to Mn(III)
rate-determining step.
For the HER on mercury the reaction
order PH + = 1 at pH less than 8 and
PH + = 0 at pH larger than 9, allows
the following elementary steps to be
distinguished:
Haq + + e Hads
at pH < 8 (93)
and
H2 O + e Hads + HO
at pH > 9
(94)
1.1.10
Double-Layer Effects on the

Currentpotential Relationship
Since the electrode reactions occur

at charged interfaces the structure of
this electried interface inuences the
electrode kinetics of charged species.
Frumkin [49] pointed out the importance
of the double-layer structure on electrode
kinetics for the hydrogen reduction
reaction on mercury. The electrical
potential at the plane where the
reactant undergoes electron transfer
to become a product inuence the
concentration of charged species different
from the analytical concentration the
bulk electrolyte due to the electrostatic
work required for the charged reactant
(product) to reach the electrode surface.
Furthermore, the inner (Galvani) potential
difference with respect to the solution
(p S ) is less than the applied electrodeelectrolyte potential difference (M S ).
The electrostatic work necessary to
bring the electroactive ion to a charged
interface, where the electrode reaction
occurs, is equivalent to the electrostatic
contribution to the free energy of activation
in homogeneous reactions between ions
in solution. When two ions of the same
charge react, their kinetics are slower as
compared to the uncharged species and
conversely when two ions of different
charge react, the apparent kinetics are
faster than for the uncharged molecules.
When no specic adsorption occurs, the
closest the reactant can get to the electrode
surface is assumed to be the outer
Helmholtz plane (OHP) and the reaction
plane is identied with this OHP, p 2 .
The simplest case is when the reactant is
the only ionic component of the solution,
then the effects on electrode kinetics
due to the properties of the interface
are maximum. For the electroreduction

reaction:
OzOo + ne RzRR
(95)
two effects due to the double-layer structure can be described:

1. the concentration of the reacting ion,
p
CO in the outer Helmholtz plane, is different from the analytical concentration
S,
in the solution, CO

zi F2
p
S
CO = CO exp
(96)
RT
The electrical work to place a mole of
O from the bulk solution in the OHP is
zi F2 where 2 is the potential of the
OHP with respect to the potential of the
solution.
2. The driving force for the electrochemical reaction is the potential jump
between the metal and the reaction site
(in the absence of specic adsorption,
identied with the OHP). The apparent rate coefcient for the reduction
reaction involving n electrons is
kapp = kM =2 CO2

nF (M 2 )
(97)
exp
RT
By combining Eqs. (96 and 97) and replacing the concentration of O at the plane of
reaction, the apparent rate constant can be
expressed in terms of the applied electrode
potential.

nF (M S )
S
kapp = kM =2 CO
exp
RT

(n z)F (2 S )
exp
RT
(98)
The rst exponential term shows the
potential dependence of the apparent
23
24
Fundamentals
rate constant on the applied potential difference (metalelectrolyte solution),

(M S ). The second exponential term is
the double-layer correction to the apparent
rate constant due to concentration at the
reaction plane and potential 2 . It is also
shown that the same correction applies for
the reverse reaction for zr = zo n since
the transition state and hence the reaction
plane must be the same for the forward
and backward processes.
It should be noticed that the Galvani
potential difference, 2 , cannot be measured directly but its value can be derived
from the GouyChapman theory of the
diffuse double layer. If the excess charge
in the metal, q m , is determined experimentally, that is, from the integration of
differential capacity curves, then for a 1 : 1
electrolyte of concentration cs , then 2
can be calculated:
m

2RT
q (2)1/2
2 =
sinh1
zi F
(2RT cs )1/2
(99)
The double-layer correction of the observed
electrode kinetics in Eq. (98) is more
important at low ionic concentration, high
ionic charge and close to the potential of
zero charge (pzc). This correction can be
minimized by using a supporting or inert
electrolyte of high concentration so that
most of the potential drops operates in the
inner Helmholtz plane and 2 0.
The consequence of double-layer correction on the exchange current density
is

(n z)F2
io = io (app) exp
RT
(100)
and on the apparent charge-transfer coefcient,
app =
ln iapp
2
F
=
(n zo )
E
RT
E
(101)
Nonlinear Tafel plots are predicted by

Eq. (101) when 2 changes appreciably
with electrode potential: at low ionic
concentrations and close to the pzc.
The apparent electrochemical reaction
order is inuenced by double-layer effects:
app
PO

=
ln iapp
ln cO

= PO
E
+ (n zo )
F
RT
2
ln CO

E
(102)
the double-layer correction in the second
term of Eq. (102) due to interfacial potential distribution is zero at the pzc.
1.1.11
Adsorption and Electrocatalysis
Adsorption of reactants, intermediates and

products at electrode surfaces signicantly
inuence the currentpotential relationship for electrode reactions. In particular
species conned to electrode surfaces can
be oxidized and reduced without the need
for the reactants to diffuse towards the
electrode surface. The adsorption free energy plays a key role in electrode kinetics
and for charged species on the potential
dependence.
An electrocatalytic reaction is an electrode reaction sensitive to the properties
of the electrode surface. An electrocatalyst
participates in promoting or suppressing
an electrode reaction or reaction path without itself being transformed. For example,
oxygen reduction electrode kinetics are
enhanced by some ve orders of magnitude from iron to platinum in alkaline
solutions or from bare carbon to carbon
electrodes modied with adsorbed iron or
cobalt phthalocyanines or porphyrins and
when certain metals are under potential
deposited (upd).
As early as 1905, Tafel [11] showed

that the kinetics of the proton reduction
strongly depends on the nature of the
electrode material. Thirty years later,
Horiuti and lkusima used the word
catalyst instead of electrode [42], and
Butler pointed out the very important role
of adsorbed intermediates in the evolution
of hydrogen [12, 13].
The main catalytic inuence of the nature of the electrode material is through
the adsorption of intermediates of complex electrode reactions. Hortiuti and
Polanyi [50] suggested that the activation
energy of an electrode reaction should
be related to the heat of adsorption of
adsorbed intermediates by a relationship
of the form of the Bronsted rule in homogeneous solutions. This corresponds
to a vertical shift of the potential energy
curves by an amount Hads with a
symmetry factor. Correlation of the activation energy or overpotential at a given
current for a particular rds on different
substrates with relative measures of the
heat of chemisorption of intermediates
have been successfully made for hydrogen evolution [51], oxygen evolution [52],
and oxygen reduction [53], and so on. For
the hydrogen electrode reaction (HER) on
different metals, it is predicted that the
curve log io vs. Gads goes through a
maximum with decreasing log io both at
positive and negative values of Gads due
to the opposite effects of the free energy of
adsorption on the geometric and exponential factors in the electrokinetic equation
(volcano plots).
Several chapters in this volume describe
new experimental evidence of adsorbate
bonding with different in situ spectroscopic techniques, the effect of the interfacial structure on kinetics and mechanisms
of selected electrode reactions with bond
breaking and surface bonding like oxygen
reduction and CO and methanol oxidation. In particular, Chapter 5.1 analyzes

electrocatalysis in depth.
1.1.12
Mixed Potential
At the equilibrium potential, both anodic

and cathodic processes of a single electrontransfer reaction take place at the same
exchange rate (a measure of which is
io ) and the current through the external
circuit is zero. The open-circuit potential is
the thermodynamic equilibrium potential.
When more than one reaction can take
place at the electrode surface, on the
other hand, the open-circuit potential
is not a thermodynamic quantity but a
mixed potential set up by the kinetics
of both simultaneous electrode reactions
that exchange electrons with the metal.
At open circuit all the charge transferred
from the electrode to the solution must be
counterbalanced by the ow of charge in
the opposite direction (Ia = Ic ) with no net
ow of charge through the interface. Thus,
anodic and cathodic processes are coupled
through the charge exchanged and a net
chemical reaction proceeds.
In order to explain the corrosion process
of metals, Wagner and Traud [54] developed the mixed potential theory, which
assumes that the currentpotential relationship is given by
I (E) =
|Ia,i (E)|
|Ic,i (E)| = 0
(103)
The additive individual component currents are based on the simplifying assumption that the anodic and cathodic processes
are statistically independent and that the
electrode surface sites are indistinguishable for both reactions. Since the total
anodic and cathodic currents equal at the
25
26
Fundamentals
O R + n e
O R + n e
ja
Ee1
Ee2
EM
jc
R + n e O
R + n e O
Schematic representation of the establishment of mixed

potential, EM , at an electrode. Partial current densities (solid lines) and
total current density curve (broken line).
Fig. 9
open-circuit or mixed potential, EM

|ia,i |Aa,i =
|ic,i |Ac,i
(104)
where ik,i represents the partial anodic
and cathodic current densities and Ak,i
the respective reaction areas.
In Fig. 9 we consider two simultaneous electrode reactions (O/R and O /R )
with transfer of n and n electrons, respectively and formal equilibrium potentials

E1e and E2e . At open circuit, the mixed
potential, EM , adopts intermediate values
E1e < EM < E2e , closer to the equilibrium
potential of the faster partial reaction. The
overall currentpotential curve is indicated by the broken line. Notice that EM
is not determined by the thermodynamic
values of E1e and E2e but by the kinetics of the respective reactions, that is, by
the respective anodic and cathodic component curves in Fig. 9 with the condition of
Eq. (104). These curves may be altered by
mass transport, surface area and specic
properties; and consequently the mixed

potential EM may be susceptible to those
kinetic factors, unlike the equilibrium potential of each partial electrode reaction,
which is xed by thermodynamics and
the activities in the bulk solution. Unlike
consecutive electron-transfer reactions the
kinetics of which are determined by the
slowest process, mixed potentials are determined by the fastest of several possible
electrode reactions.
If one of the partial electrode reactions
is the dissolution of the electrode (i.e.
metal or semiconductor) the open-circuit
potential is a corrosion potential and the
system undergoes corrosion at a level
given by the corrosion current (icorr ),
which measures the dissolution rate. The
magnitude of icorr for a corroding system
may vary signicantly over several orders
of magnitude, like the exchange current
densities of different individual electrode
reactions span a range of orders of
magnitude.
In Fig. 9, the occurrence of two simultaneous redox reactions at an electrode

surface has been considered; however, in
most corrosion problems, more than two
reactions may take place and both forward and backward individual electrode
reactions may not take place at signicant rates in the potential range where the
mixed potential is observed, owing to the
slow kinetics of the participating reactions
under those conditions.
Figure 10 illustrates the corrosion of two
metals, M and M , in aqueous aerated (oxygenated) solutions. Three anodic partial
reactions are considered: the active dissolution of two metals M and M with
different kinetics in the absence of their
ions in the bulk electrolyte and decomposition of water with the evolution of oxygen.
The kinetics of the latter process is so
slow on most corroding metals that only
at very negative potentials can dissolved
oxygen be electroreduced and this eventually becomes limited by mass transport
due to the limited solubility of oxygen

in water. At even more negative potentials, hydrogen evolution takes place on
the electrode surface. The cathodic reduction of some dissolved metal ions in the
solution adjacent to the electrode as a consequence of corrosion is also considered
in Fig. 10 but the rate of oxygen reduction
is the controlling factor in the overall rate.
The measurable currentpotential curve
results from the balance of the anodic and
cathodic processes as indicated by the broken line and the corrosion potential, EM ,
is also shown. Metal M is more noble than
M and is not corroded because the oxygen
reduction reaction is far too slow to bring
M to a potential where it can be corroded.
Hydrogen evolution is a faster reaction but
cannot corrode M because its equilibrium
e and
potential is more negative than EM
therefore the reaction cannot proceed on
M under those conditions.
It is well known that the reduction of
organic molecules such as nitrobenzene
M MZ+ + z e
M MZ+ + z e
jcorr
Eo2
Eo1
Ecorr
MZ+
+z
jcorr
Eo4
Eo3
H2O O2 + 4H+ + 4e
2H+ + 2e H2
O2 + 4e + 4H+ 2H2O
Representation of the formation of the corrosion

potential, Ecorr , by simultaneous occurrence of metal
dissolution, hydrogen evolution, and oxygen reduction. Partial
current densities (solid lines) and total current density curve
(broken line).
Fig. 10
27
28
Fundamentals
with metals in acid leads to different nal

products depending on the nature of the
metal employed and the composition of
the electrolyte. The kinetics of hydrogen
evolution, either from proton or from
water, as well as the kinetics of dissolution
of some metals, particularly those like Zn
that can be complexed by HO , strongly
depend on the pH and metal surface; the
mixed potential is then xed by the kinetics
of hydrogen reduction and dissolution
of the metal, which are relatively fast
reactions. Organic electroreductions are
slower processes and not likely to be
potential-determining reactions, but they
proceed through different reaction paths
at different potentials depending on the
mixed potential of the metal surface.
The theory of mixed potentials is based
on the independence of partial anodic
and cathodic reactions. However, in real
corroding systems, coupling of these
reactions, not only through exchanged
charge at the interface, may sometimes
occur as a result of (1) geometric surface
limitations due to the blocking effect
of the intermediates of one reaction
on the kinetics of others taking place
simultaneously at the same electrode
surface and (2) changes in local pH
due to the fact that some corrosion
reactions such as hydrogen evolution may
promote changes in the kinetics of other
participating reactions.
Electroless deposition of metals like
nickel-phosphorus, introduced in 1946
by Brenner and Riddell [55], is of great
industrial importance and allows to deposit metals on nonconducting substrates
such as printed circuit boards (copper),
contacts (silver), and so on. Its mechanism is based on a particular case of
mixed-potential where the partial electrode
reactions are metal deposition at metal
nuclei and oxidation of organic and inorganic reductants. A quantitative treatment

has been derived by Spiro [56] and practical aspects have been recently reviewed by
OSullivan [57].
1.1.13
Electrocrystallization
The formation or dissolution of a new

phase during an electrode reaction such as
metal deposition, anodic oxide formation,
precipitation of an insoluble salt, and so
on involves surface processes other than
charge transfer. For example, the incorporation of a deposited metal atom into a
stable surface lattice site introduces an
extra hindrance to the ow of electric
charge at the electrodesolution interface
and therefore the kinetics of these electrocrystallization processes are reected in
the overall electrode reaction.
During the initial stages of phase formation at electrodes (nucleation) incorporation of an adsorbed species into the
surface lattice occurs only at active sites
(crystal plane edges or kinks) where interactions with other species already in
the lattice are possible. The process of
formation of such nuclei may become kinetically limiting and supersaturation or
overpotential is required for nuclei to be
formed and survive. Depending on the relative time scale of nuclei formation and
further growth or thickening of the deposit local potential distribution and mass
transport conditions determine the morphology.
If the electrocrystallization is controlled by the formation of two or threedimensional isolated nuclei, the currentpotential relationship has a stronger
overpotential dependence than predicted
by the ButlerVolmer equation [58],
that is,
B
i = A exp
n

(105)
with n = 2 for three-dimensional and n =

1 for two-dimensional nucleation. Growth
of isolated nuclei at an electrode surface
is eventually limited when they start to
coalesce due to their number and size and
the size of the electrode area.
In metal deposition, depending on the
binding energy of the metal to be deposited with respect to the foreign metal
substrate, two limiting cases have been
observed: (1) formation of structured underpotential deposited layers (upd) and
further growth of the bulk metal phase
on top of the two-dimension upd layer
or (2) deposition in the overpotential region involving relatively few nuclei by
three-dimensional nucleation and further
growth. The inuence of upd ad-adatoms
on electrode kinetics is treated in Chapter 4.3 and electrochemical nucleation and
growth is described in Chapter 5.3.
Acknowledgments
The author wishes to acknowledge nancial support during the preparation of

this chapter from Conicet, ANPCyT (Argentina), Fudetec, Motorola SPS, and the
Guggenheim Foundation for a fellowship
2000/2001.
References
1. E. Gileadi, M. Urbakh, (Eds.), Thermodynamics and Electried Interfaces, Vol. 1. EoE,
Wiley-VCH, Weinheim, Germany, 2002.
2. A. J. Bard, L. R. Faulkner, Electrochemical
Methods. Fundamentals and Applications, 2nd
ed., Wiley, New York, 2001.
3. H. Gerischer, D. M. Kolb, J. K. Saas, Adv.
Phys. 1978, 27, 437.
4. W. J. Albery, Electrode Kinetics, Oxford University Press, Clarendon Press, Oxford, 1975.
5. J. OM. Bockris, K. N. Reddy, Modern Electrochemistry, Plenum Press, New York, 1970.
6. B. E. Conway, Theory and Principles of Electrode Processes, Ronald Press, New York, 1964.
7. P. Delahay, Double Layer and Electrode Kinetics, 2nd edn., Interscience Publishers, New
York, 1966.
8. K. J. Vetter, Electrochemical Kinetics (English
Translation), Academic Press, New York,
1967.
9. C. M. A. Brett, A. M. Oliveira Brett, Electrochemistry: Principles, Methods and Applications, Oxford University Press, Oxford, 1993.
10. E. J. Calvo in Comprehensive Chemical Kinetics (Eds.: C. H. Bamford, R. G. Compton),
Elsevier, Amsterdam, 1986, Vol. 26.
11. J. Tafel, Z. Phys. Chem. 1905, 50A, 641.
12. J. A. V. Butler, Trans. Faraday Soc. 1932, 28,
379.
13. J. A. V. Butler, Proc. R. Soc. London, Ser. A
1936, 157, 423.
14. T. Erdey-Gruz, M. Volmer, Z. Phys. Chem.
Abt. A 1931, 157, 165.
15. R. A. Marcus (Nobel Lecuture), Angew.
Chem., Int. Ed. Engl. 1993, 32, 1111.
16. R. A. Marcus, J. Chem. Phys. 1956, 24, 966.
17. R. A. Marcus, Discuss. Faraday Soc. 1960, 29,
21.
18. R. A. Marcus, J. Phys. Chem. 1963, 67, 853.
19. R. Marcus, J. Chem. Phys., 1965, 43679.
20. R. Marcus, J. Phys. Chern. 1968, 72, 891.
21. J. R. Miller, L. T. Calcaterra, G. L. Closs, J.
Am. Chem. Soc. 1994, 106, 3047.
22. A. L. Barker, J. V. Macpherson, C. J. Slevin
et al., J. Phys. Chem. B 1998, 102, 1586.
23. A. L. Barker, P. R. Unwin, S. Amemiya et al.,
J. Phys. Chem. B 1999, 103, 7260.
24. M. Tsionsky, A. J. Bard, M. V. Mirkin, J.
Phys. Chem. 1996, 100, 17 881.
25. M. Tsionsky, A. J. Bard, M. V. Mirkin, J. Am.
Chem. Soc. 1997, 119, 10 785.
26. C. J. Miller in Physical Electrochemistry, Principles, Methods and Applications (Ed.: I. Rubinstein), Marcel Dekker, New York, 1995,
Chap. 2.
27. H. O. Finklea, Electroanal. Chem. 1996, 19,
109.
28. C. E. D. Chidsey, J. Am. Chem. Soc. 1997,
119, 10 563.
29. S. Creager, S. J. Yu, D. Bamdad et al., J. Am.
Chem. Soc. 1999, 121, 1059.
30. R. Holmlin, R. F. Ismagilov, R. Haag et al.,
Angew. Chem., Int. Ed. 2001, 40, 2316.
29
30
Fundamentals
31. D. I. Gittins, D. Bethell, D. J. Schiffrin et al.,
Nature 2000, 408, 67.
32. X. D. Cui, A. Primak, X. Zarate, J. Tomfohr,
O. F. Sankey, A. L. Moore, T. A. Moore,
D. Gust, G. Harris, S. M. Lindsay, Science
2001, 294, 571.
33. F. F. Fan, J. Y. S. M. Dirk, D. W. Price et al.,
J. Am. Chem. Soc. 2001, 123, 2454.
34. C. E. D. Chidsey, Science 1991, 251, 919.
35. R. J. Forster, L. R. Faulkner, J. Am. Chem.
Soc. 1994, 116, 5444.
36. V. G. Levich in Physical Chemistry. An Advanced Treatise (Eds.: H. Eyring, D. Henderson, W. Jost), Academic Press, New York,
1970, pp. 9851074, Vol. IXB.
37. H. Gerischer, Adv. Electrochem. Electrochem.
Eng. 1961, 1, 139.
38. W. Lorenz, Z. Elektrochem. 1953, 57, 382.
39. H. Gerischer, Z. Elektrochem. 1953, 57, 604.
40. K. Heusler, Electrochim. Acta 1983, 28, 439.
41. E. J. M. O Sullivan, E. J. Calvo in Reactions at
Metal Oxide Electrodes Comprehensive Chemical Kinetics (Ed.: R. G. Compton), Elsevier,
New York, 1987.
42. J. Horiuti, M. Ikusima, Proc. Imp. Acad.
(Tokyo) 1939, 15, 39.
43. J. Horiuti, J. Res. Inst. Catal. Hokkaido Univ.
1948, 1, 8.
44. R. Parsons, Pure Appl. Chem. 1974, 37, 501.
45. R. Parsons, Pure Appl. Chem. 1979, 52, 233;
reprinted in Electrochim. Acta 1981, 26, 1869.
46. R. Parsons, Trans. Faraday Soc. 1951, 47,

1332.
47. J. E. B. Randles, Can. J. Chem. 1959, 37, 238.
48. J. D. Clark, R. P. Wayne in Comprehensive
Chemical Kinetics (Eds.: C. H. Bamford,
C. F. H. Tipper), Elsevier, Amsterdam, 1969,
Vol. 2, Chap. 4, 377462.
49. A. N. Frumkin, Z. Phys. Chern. 1933, 164,
121.
50. J. Horiuti, M. Polanyi, Acta Physicochim.
URSS 1935, 2, 505.
51. J. OM. Bockris, B. E. Conway, J. Chern. Phys.
1956, 26, 532.
52. P. Ruetschi, P. Delahay, J. Chern. Phys. 1955,
23, 1167.
53. A. J. Appleby, Mod. Aspects Electrochem. 1974,
9, 369.
54. C. Wagner, M. Traud, Z. Elektrochem. 1938,
44, 391.
55. A. Brenner, G. Riddell, J. Res. Natl. Bur.
Stand. 1946, 37, 31.
56. M. Spiro, J. Chem. Soc., Faraday Trans. I
1979, 1507.
57. E. J. O Sullivan in Advances in Electrochemical Science and Engineering (Eds.: R. C. Alkire,
D. M. Kolb), Wiley-VCH, Weinheim, Germany, 2002.
Abt. A 1931, 157, 177.
1.2 Quantum Theory of Electrochemical Electron-Transfer Reactions
1.2
Quantum Theory of Electrochemical

Electron-Transfer Reactions
Wolfgang Schmickler and Stefan Frank
Abteilung Elektrochemie, University of Ulm,
Ulm, Germany
1.2.1
Introduction
Electron-transfer reactions are amongst

the most important processes in electrochemistry. Indeed, all electrochemical
reactions involve an exchange of electrons
between a particle and the electrode at
some stage. Therefore, electron transfer
has been the subject of intensive research
ever since the rise of electrochemical kinetics toward the middle of the last century.
Much of our present understanding
of electron-transfer reactions rests on
concepts that were established in the
pioneering papers by Marcus [1] and by
Hush [2] in the 1950s. In particular, these
works pointed out the important role
that the reorganization of the solvent
plays. Since that time the theory of
these reactions has developed in various
directions. A good review of the present
state of the theory and its applications
to physics, chemistry and biology can be
found in the book by Kuznetsov [3]. Here
we shall focus on a few topics relevant to
electrochemical electron transfer at metal
electrodes semiconductors are treated in
a recent monograph by Memming [4].
The formal development of the topic will
rest on a model Hamiltonian proposed by
one of us in the 1980s. From this we will
rst derive the Levich and Dogonadze theory [5], which was the rst quantum theory
for electron transfer in condensed media,
and then obtain the classical potentialenergy surfaces that are generalizations
of those familiar with Marcus and Hush.
These surfaces will also serve as the

basis for a discussion of solvent dynamics. The model Hamiltonian will then be
extended to describe bond-breaking reactions, which will conclude the main
part of this review. The last section is
devoted to computer simulations, which
can ll in some of the gaps that are
missing in proper theory, and give quantitative results.
In this chapter, we attempt to address
a broad audience that comprises both
experimentalists and theoreticians. We
therefore have to strike a compromise
between readability and rigor and will
not be able to cover some of the more
formal aspects and second-order effects,
for which the book by Kuznetsov [3] is
a good reference. As an introduction,
we start with a simple presentation of
the Marcus theory, which should provide
the necessary background for the more
formal sections.
1.2.2
A Simple Introduction to Marcus Theory
We consider electron transfer from a metal

electrode to a solvated species nearby
according to the scheme

red
ox + e (Me)
(1)
At rst, we limit ourselves to the so-called

outer sphere reactions in which chemical
bonds are neither formed nor broken.
During such a reaction, the charge on
the reactant, and hence the conguration
of its solvation sheath, changes. The
order in which electron transfer and
solvent reorganization occur is governed
by the FrankCondon principle: electron
transfer is fast and occurs practically at
a xed position of the solvent. Obviously,
it would be energetically unfavorable if
electron transfer were to take place before
31
1 Fundamentals
ox + e
Red
Energy
32
Fig. 1 Potential energy vs.

generalized solvent coordinate q
in Marcus theory.
Eact
G
qi
qf
Reaction coordinate
the solvent had started to reorganize, since

the reactant would be surrounded by a
solvation sheath pertaining to a different
charge. Likewise, a solvent reorganization
prior to electron transfer would require
high energy, and is equally unlikely.
Therefore, the solvation sheath must rst
take up a conguration between those
for the initial and the nal charge state,
then the electron is transferred, and the
system relaxes to its new equilibrium
conguration.
These ideas can be put into quantitative
terms within a simple, one-dimensional
model in which the (free) energy of
the system is plotted as a function of
a generalized solvent coordinate q characterizing the solvation of the reactant.
There are two such curves (see Fig. 1),
one for each side of Eq. (1). Each curve
attains its minimum at the equilibrium conguration for the corresponding
state. If we develop each curve into a
Taylor series about its minimum and
keep terms up to second order, we obtain the harmonic approximation familiar
through many branches of physics and
chemistry in which each curve is represented by a parabola. We make the
further assumption that the curvatures
of both parabolas are equal this can be
justied within an explicit model for the

solvation [6].
The minima for the two parabolas correspond to the equilibrium congurations
for the initial and the nal states, and
the difference between the values at the
minima gives the free energy G, of the
reaction. Changing the electrode potential affects the energy of the transferred
electron, and hence G. The free energy
of activation is the difference between the
energies of the crossing point and the minimum of the initial state; it can easily be
calculated within the harmonic approximation: Let the two parabolas for the initial (i)
and the nal (f ) states be
Ui (q) = 12 k(q qi )2 ,
Uf (q) = 12 k(q qf )2
(2)
where qi and qf denote the two minima.

A simple calculation gives, for the energy
of activation
( + G)2
4
(3)
= 12 k(qi qf )2
(4)
Eact =
where
is the so-called energy of reorganization for

the solvent.
The preexponential factor Z for the

rate constant can be calculated from the
activated complex or from Kramers theory
(vide infra); it is independent of G
or, in the electrochemical context, of the
overpotential . Thus, the rate constant can
be written as
k = Z exp
( + G)2
4kB T
(5)
where kB is the Boltzmanns constant.

This expression gives the rate for a
homogeneous electron-transfer reaction;
it is usually associated with the name
of Marcus, but the independent work
of Hush is also equivalent. The sign
convention in Eq. (3) is the usual one:
a negative sign denotes a lowering of
the free energy during the reaction; thus,
G measures the driving force for
the reaction. An unusual feature is the
quadratic dependence of the activation
energy on G. As long as + G > 0,
the reaction gets faster as the driving force
increases, but for very large driving forces,
where + G < 0, the reaction becomes
slower. This is the so-called inverted region,
whose existence has been established for
homogeneous reactions. For reasons that
will become apparent below, it cannot
be accessed in heterogeneous reactions at
metal electrodes.
In order to apply Eq. 5 to electrochemical
reactions, we must make two modications:
1. Equation 5 gives the rate for a reactant sitting at a given distance from
the electrode. In order to get the total rate, we must integrate over all
distances. Typically, the rate decreases
exponentially with the distance, with
a decay constant of 1 A1 . Integration, therefore, adds an additional
factor C = 1/ 108 cm to the preexponential factor.

2. For electrochemical reactions at metal
electrodes, the reaction-free energy G,
refers to the exchange with the Fermi
level, and can then be related to the
overpotential by G = e0 , where
the sign depends on the direction of the
current, anodic or cathodic. However,
in a reduction reaction the electron can
come from any occupied state, and in
an oxidation it can be transferred to
any empty level. Let us denote by
the energy of a metal electron with
respect to the Fermi level, and by (),
the density of electronic states at the
surface. To be specic, we consider
electron transfer from the metal to a
redox partner in the solution. The free
energy for the transfer of an electron
with energy is by an amount
lower than that for electron transfer
from the Fermi level. The probability
of nding an electron with that energy
is given by ()f (), where f () is the
FermiDirac distribution. Hence, the
rate of electron transfer from the metal
to the reactant is given by

kred = ZC
()f ()
exp
( + e0 )2
d
4kB T
(6)
where we have used the fact that a negative overpotential favors a reduction.
Equation (6) has also been derived by
Gerischer [7, 8], but with a different
interpretation. He denes
Wox (, ) = (4kB T )(1/2)
exp
( + e0 )2
4kB T
(7)
33
1 Fundamentals
as the normalized density of oxidized

or empty states in the solution. Since
()f () is the density of occupied states
on the metal, Eq. (6) can be interpreted in
the following way: The rate of exchange
of electrons of energy is proportional
to the density of occupied states on the
metal times the density of empty states in
the solution. Changing the overpotential
simply shifts the density of oxidized states
with respect to the Fermi level.
The form of Eq. (6), where the total rate
is written as an integral over the energy
of the transferring electron is found in
several versions of the electron-transfer
theory. For future reference, we write it as

(8)
kred = ZC ()f ()kr () d
where kr () is the energy-resolved rate.
The integration in Eq. (6) is to be
performed over the conduction band of the
electrode; in practice, it may be extended
to , since the major contribution
to the integral comes from the vicinity

of the Fermi level. If the density of
states () is taken to be a constant,
and the FermiDirac distribution replaced
by a step function, the integral can be
performed explicitly:

e0 +
kred = ZC kB T erfc
(4kB T )1/2
(9)
Here erfc denotes the complement of
the error function. Equations (6) and (9)
predict rate constants with a transfer coefcient of 1/2 for small overpotentials, and
a constant rate klim for high overpotentials
(see Fig. 2).
From the form of the current-potential
curve, it is clear that the Marcus-inverted
region cannot be accessed in a simple
redox reaction on a metal electrode. For
large overpotentials, the electrons are
simply transferred to energy levels above
the Fermi level so that the inverted part of
the energy-resolved rate plays no role.
0
= 0.5 eV
ln (kred/klim)
34
= 0.75 eV
= 1 eV
12
0.0
0.4
0.8
1.2
[V]
Logarithm of the normalized rate constant kred /klim vs.
overpotential for the reduction of a redox couple at a metal electrode
according to Marcus theory (Eq. 9).
Fig. 2
1.6
1.2.3
Energy of Reorganization
As outlined above, the energy of reorganization plays a major role in the electrontransfer theory. Generally it contains two
contributions: one from the surrounding
solvent and the other arising from changes
in the bond lengths and vibration frequencies of the reacting complex itself. They are
also referred to as outer-sphere and innersphere reorganization, respectively.
The reorganization of the solvent can
be expressed through its polarization,
which also contains two contributions: one
from the electronic polarizability of the
solvent molecules, and the other from the
librational and vibrational motion. Only
the latter are slower than the electron
exchange as such, and contribute to the
solvent reorganization energy out . This
takes a form that is reminiscent of the
Born formula for the energy of solvation:

1
1
1
out =
20
s

(Dox Dred )2 dV (10)
Here, and s are the optical and the
static dielectric constants of the solution;
the former appears because the contribution from the electronic polarizability has
been subtracted. Dox and Dred are the dielectric displacements when the reactant is
in the reduced and in the oxidized form, respectively. The integral is to be performed
over the space lled by the solution. When
the reactant is close to the electrode surface, image terms arise, which contribute
to the displacement.
The energy of reorganization is a molecular concept, and an equation such as (10),
which is based on macroscopic electrostatics, can only be a rough approximation.
Several other expressions, on the basis
of nonlocal electrostatics or involving the

whole frequency spectrum of the dielectric
constant, have been suggested, but the best
estimates nowadays come from computer
simulations.
The contribution from inner-sphere
modes are easily obtained within the
harmonic approximation. Let q be the
change in the bond length of a mode
with frequency and effective mass m;
then its contribution to the energy of
reorganization is
i = 12 m2 q 2
(11)
When electron transfer is accompanied

by a signicant change in frequency
from 1 to 2 , an effective frequency
21 2 /(1 + 2 ) can be used to a good
approximation.
1.2.4
Quantum-mechanical Theory
A Model for Electron Exchange
Between a Metal and a Solvated Reactant
Before we set up the model Hamiltonian
for electrochemical electron transfer, we
have to specify the models for the various parts of the system. For the electrons
in the metal, we use the quasi-free electron model in which the electronic states
are labeled by their quasi-momentum k.
For outer-sphere electron transfer on metals, it is usually permissible to ignore the
spin index keeping it would introduce an
additional factor of two, which can be incorporated into the interaction constants.
On the reactant, we consider a single orbital, labeled a, with which the electrons
are exchanged.
An important aspect of electron transfer
is the accompanying reorganization of the
solvent. The latter can be modeled as a
phonon bath, or a collection of harmonic
oscillators. Other models in terms of the
1.2.4.1
35
36
1 Fundamentals
solvent polarization [9] or the orientation

of the solvent dipoles [10] are equivalent.
The formulation in terms of harmonic oscillators has the additional advantage that
it is easy to include quantum modes: Most
reactions are coupled to a few quantum
modes, such as inner-sphere modes of the
reactant, or even solvent modes in the
rst solvation shell [11]. These modes can
be included in the collection of harmonic
oscillators; of course, in contrast to the
classical solvent modes they have to be
treated by quantum mechanics.
The solvent modes can be separated
into the slow librational and vibrational
modes, and the fast electronic modes. Only
the slow modes are represented in the
phonon bath. The fast modes are supposed
to be in electronic equilibrium with the
transferring electron so that they simply
renormalize the energy of the reactants
electronic states.
The Model Hamiltonian
The quantitative formulation of the model
outlined above is conveniently written in
terms of second quantization. For this
purpose, we introduce number operators
n and creation and annihilation operators
c+ and c for the various electronic
states. Then the Hamiltonian takes the
form

k nk +
[Vk ck+ ca
H = a na +
1.2.4.2
+ Vk ca+ ck ] +
na

1
2
h g q
h (p2 + q2 )

(12)
The rst term describes the electronic state

of the reactant with energy a , the next
term represents the electrons in the metal.
Electron exchange between the metal
and the reactant is incorporated in the

third term; the amplitudes Vk characterize
the strength of the electronic coupling
between the two partners. The fourth term
describes the bath of harmonic oscillators;
q and p are dimensionless coordinates
and momenta, and the frequencies. The
summation runs over both the classical
solvent modes and over the quantum
(inner-sphere) modes. The coupling of the
oscillators to the redox couple is assumed
to be linear and proportional to the charge,
as given in the last term.
This Hamiltonian, which was introduced by Schmickler [12], is equivalent to
earlier formulations by Levich and Dogonadze in terms of wave mechanics [5]; it is
also related to the spinboson model for
homogeneous electron exchange [13] and
to the AndersonNewns model for specic
adsorption [14].
Further treatment of this model depends
on the magnitude of the electron-transfer
term: if the amplitudes Vk are small,
the rst-order perturbation theory can be
applied; if they are large, it is better
to start from the adiabatic potentialenergy surfaces.
1.2.5
First-order Perturbation Theory

Semiclassical Systems
We next consider the case in which the
electron-transfer terms are small so that
the rst-order perturbation theory can be
applied and assuming that all the phonon
modes are classical quantum modes will
be considered below. In this case, the
coordinates q of the oscillators can be
treated as external parameters. It is then
convenient to include the electron-phonon
coupling the last term in Eq. (12) into
the electronic energy of the reactant,
and rewrite the rst and last term as
1.2.5.1
a na , where
a = a
h g q
(13)
To rst order in the coupling constants

Vk , the rate of electron transfer from a
state k on the metal to the reactant at a
given conguration q is given by Fermis
golden rule:
Wka (q ) =
|Vk
h
|2
(a k )
(14)
The full derivation of the rate constant

for this case can be found in Refs. [6, 15].
Here we list only the main steps involved:
The delta function in Eq. (14) is replaced
by its Fourier transform; this greatly
simplies the mathematics.
Next the thermal average over all
congurations q is performed. This
results in an energy-resolved rate for
the reduction:
1/2

|Vk |2
kr (k ) =
4kB T
h
exp
(a k )2
(15)
4kB T
for the rate at a given energy k , which

is to be compared with Eq. (5) of the
simple Marcus theory. Here
=
1
h g2
2
(16)
is the energy of reorganization of the

phonon bath. Thus, in the semiclassical
case the coupling to the bath is characterized by a single constant.
To get the total rate, we have to sum
over all occupied levels on the metal.
This sum is converted into an integral:

d f ()()
(17)
k,occ.
Finally, we relate a to the overpotential

by noting that for = 0 the forward
and backward rate must be equal;
this gives: a = + e0 . Further, we
introduce the factor C converting bulk to
surface concentrations (see Sect. 1.2.2),
and take the coupling elements Vk
as constant.
This procedure results in the following
expression for the electrochemical rate
constant for the reduction:

1/2
|Vk |2
kred = C
h
kB T

( + e0 )2
()f () exp
4kB T
(18)
which differs from the MarcusHush
expression in the preexponential factor,
which is here determined by the square
of the transition amplitude Vk . This is
characteristic for the nonadiabatic, weak
coupling case. Solvent dynamics, which
is so important for the adiabatic, strong
coupling case plays no role.
Quantum Modes
When quantum modes are reorganized
during the electron exchange, they have
to be treated separately from the classical
modes. We consider explicitly the case
in which one such mode is present;
the generalization to many modes is
straightforward, though it results in rather
cumbersome formulae.
The situation is simplest if the frequency
i of the quantum mode is so high that its
quantum of energy is much higher than
the thermal energy, hi kT , so that it
is not excited in the initial state. In this
case, we only have to consider transitions
from the ground state 0 to the nth excited
state. Let
1.2.5.2
37
38
1 Fundamentals
f
M0n = |0i |n |2 =

exp
1
n!
i
hi
i
hi
n
(19)
be the FrankCondon factors for these

f
transitions. Here, ni , n are the vibronic
wave functions for the initial and nal
states; i is the energy of reorganization
pertaining to the quantum mode. The
rate of electron transfer involving the
transition 0 n is then obtained by
multiplying the classical rate of Eq. (17) by
the corresponding FrankCondon factor,
and by noting that the energy nhi ,
which goes into the excitation of the
quantum mode, is no longer available for
the classical reorganization. The energyreduced rate for the reduction is then
1/2

|Vk |2
kr (0 n) = M0n
kB T
h
exp
( + nh e0 )2
4kB T
where . . . denotes a thermal average.

A good general treatment of reactions
involving quantum modes has been given
by Tanaka and Hsu [16].
Adiabatic Reactions
When the electronic interaction is large,
perturbation theory does not sufce. We
consider the adiabatic limit in which
the system is always in electronic equilibrium the general case of an interaction with an intermediate strength
has not yet been solved and construct the potential-energy surface for
the reaction.
The electronic interaction with the metal
induces a broadening of the electronic
level of the redox partner, which is
determined by the matrix elements Vk :
1.2.6
=
|Vk |2 ( k )
(23)
(20)
The total rate is obtained by summing over

all nal states n:

kr (0 n)
(21)
kr =
n
As long as e0 < h, transitions from

ground state to ground state dominate.
When the quantum of energy is of the
same order of magnitude as the thermal
energy, h kT , we have to consider
transitions from excited initial states m
to nal states n, and perform a thermal average over the initial states of the quantum
modes:

|Vk |2
kr = Mmn
(/kB T )1/2
h

(+(n m)he0 )2
exp
(22)
4kB T
In principle, is a function of the

electronic energy . When the reactant
interacts with a broad electronic band, the
width usually does not vary much in
the relevant region near the Fermi level
and can be taken as constant; this is the
so-called wide-band approximation, which
will be used here.
So far the adiabatic case has only been
fully worked out for the semiclassical
case in which quantum modes play no
role. In this case, it is again convenient
to use the q-dependent electronic energy a introduced in Eq. (13). In the
wide-band approximation, the broadening of this energy level takes the form
of a Lorentzian so that its density of
states becomes
() =
2
[ (q )]2 + 2
(24)
The reactant orbital is lled only partially

up to the Fermi level of the metal, and its
occupancy is given by
n(q ) na (q ) =
1
a
arccot
(25)
where the Fermi level has been taken as

the energy zero. The potential energy of the
system, as a function of the coordinates q ,
is then
1
E[q ] = a n(q ) +
h q2
2
+

ln(a2 + 2 )
2
(26)
where a constant term has been neglected.

With the aid of Eqs. (25 and 26) the
potential energy of the system can be
calculated as a function of the coordinates q . Of particular interest are the
stationary points of this curve, at which
the occupancy of the reactant obeys the
self-consistency equation:
n0 =
1
a na 2n0
arccot
(27)
which must be solved numerically. Depending on the values of the system

parameters, this equation has either one or
three solutions [17]. In the case in which
the system is not too far from equilibrium,
they can be stated in the following form:
When the electronic interaction is
large and the energy of reorganization
is small, there is only one stationary
state. This corresponds to an adsorbed
state, which sits in a single potential well
on the electrode surface.
When the electronic interaction is weak
and the energy of reorganization is large,
there are three stationary solutions: two
minima separated by a barrier. The
two minima correspond to two different
charge states of the reactant. In an

electron-transfer reaction, the system
moves from one minimum via the
barrier to the other minimum.
In mathematical terms, three stationary
points exist if both the conditions
2
>1
(28)

2
| |
|arccos( ) (1 2 )1/2 |
(29)
are fullled, where 2 = /2.
Within this formalism, it is in principle possible to calculate the potentialenergy surfaces for electron-transfer reactions [17]. The important system parameters are , and a ; they can be
estimated from quantum chemical calculations and computer simulations. At
present, such surfaces involve fairly rough
approximations; however, they are useful
visualizations for understanding the dynamics of electron-transfer reactions. It
is convenient to restrict the system to a
minimum set of coordinates using a generalized solvent coordinate that follows a
reaction path from one minimum of the
surface via a saddle point to the other
minimum. This coordinate can be normalized in such a way that its value at the
two minima corresponds to the charge on
the reactant in these congurations [18].
It, then, is identical to the reaction coordinate in the theory of Hush [2]. If quantum
modes are reorganized during the electron
transfer typically, these would be innersphere modes they have to be singled out
and treated separately.
Figure 3 shows an adiabatic potentialenergy surface for one single solvent
coordinate q. Two minima are separated by
a barrier. This barrier is the lower and the
atter the larger the electronic coupling
between the reactant and the metal. For
39
1 Fundamentals
15.0
12.5
10.0
Energy
[kB T ]
40
7.5
5.0
2.5
0.0
1
Solvent coordinate q
Adiabatic potential-energy curves for various values of the energy
broadening ; full line: = 0.01 eV; long dashes: = 0.05 eV; short dashes:
= 0.1 eV. The energy of reorganization was taken as 6 kB T.
Fig. 3
the lowest coupling shown, the surface

resembles two intersecting parabola. This
is the weakly adiabatic regime in which
the electronic coupling is strong enough
to establish electronic equilibrium, but too
weak to lower the barrier.
1.2.7
Solvent Dynamics
In the nonadiabatic case, electron transfer

is a rare event. Therefore, the equilibrium
distribution in the initial well is undisturbed, and the preexponential factor is
determined by the electronic coupling between the reactants. In adiabatic reactions,
the electronic system is in equilibrium at
all points of the trajectory; therefore, the
rate is not limited by electron transfer but
by solvent dynamics.
The main points can be explained within
the one-dimensional model outlined in the
preceding section in which the free energy
is taken to be a function of a generalized
solvent coordinate the two-dimensional
case involving an inner-sphere mode has

been treated, amongst others, by Sumi
and Marcus [19]. The simplest estimate
for the rate of an adiabatic reaction is
based on the transition-state theory. If the
potential-energy well of the initial state is
approximated by a harmonic oscillator of
frequency i , the rate in that theory is
given by
k=
i
Eact
exp
2
kB T
(30)
where Eact is the energy of activation. However, as was pointed out by Kramers [20],
the transition-state theory overestimates
the rate since the population at the barrier
may be depleted, and also because a system that has passed the saddle point can
return, a phenomenon denoted as barrier
recrossing. These effects can be described
in terms of the solvent friction, which encompasses the inuence of all the solvent
modes that are not explicitly considered in
the one-dimensional model. As a result,
the preexponential factor is modied by an
additional transition factor:

=
( 2 /4 + b2 )1/2
/2
(31)
where is the friction coefcient, and

b is the barrier frequency obtained
by approximating the barrier region by
an inverted parabola. In effect, Eq. (31)
describes a competition between the force
exerted by the barrier, which is determined
by b and the stochastic interaction with
the surrounding bath characterized by .
When the friction is weak, b , the
transition factor reduces to unity, and
the transition-state theory holds. In the
opposite limit, the case of strong friction,
the transition factor simplies to
=
2b
(32)
For electron transfer in solutions, it is

often assumed that the strong friction
limit holds. The friction coefcient has
been identied with the inverse of the
longitudinal relaxation time:
=
1
,
l
where
l =
D (33)
s
and D is the usual relaxation time.

1.2.8
Computer Simulations
We have already emphasized the important role of the solvent dynamics and structure in the kinetics of electron-transfer
reactions. During the last few years, a
number of classical molecular dynamic
(MD) simulations have been performed
to obtain free energy surfaces of the reaction [2127]. These simulations require
explicit interaction potentials between the
constituents of the system: the reactants,
the solvent, and the electrode. Again, a
generalized solvent coordinate is used to
represent the results; however, instead of

the coordinate q dened in Sect. 1.2.6
most researchers use an electronic shift
E = 2q. As electron transfer requires
relatively large uctuations of the ion solvation, umbrella sampling techniques must
be applied and in most cases it is not possible to obtain the electron-transfer rates
directly from the MD trajectories. Instead,
one applies transition-state theory or performs simulations near the saddle point
in order to explore solvent dynamics near
the barrier.
Rose and Benjamin [22] have computed
the diabatic free energy surface of the redox pair Fe2+ /Fe3+ at a xed distance
from a Pt(100) surface (Fig. 4). Differences to a parabolic representation of the
free energy are only signicant far from
equilibrium, which affects the reorganization energy , but hardly affects the
free activation energy. In contradiction to
a common assumption, the curvature at
the potential minimum for the oxidized
and the reduced states are slightly different. Both the electron-transfer rate k and
the transfer coefcient depend on the
applied overpotential . In the regime of
strong electronic coupling, they elucidate
a classical transmission coefcient in
agreement with GroteHynes theory [28],
which is a generalization of Kramers theory; in particular, increases as the barrier
becomes sharper.
Smith and Halley [23] performed simulations on a similar system for various
xed distances between the ion and the
metal surface. They found a weak dependence of the reorganization energy on the
separation, but strong variations of the
thermodynamic driving force; even the direction of the reaction changed when the
ion was moved away from the surface.
Since their model includes an external
electric eld, but no realistic screening,
41
1 Fundamentals
100
80
G
[kcal mol1]
42
60
40
20
Fe3+
0
4
Fe2+
(E E )
[eV]
Diabatic solvent reorganization free energy curves for Fe2+
and
[22]. The bold solid lines represent the molecular dynamics
result, the thin solid lines the best parabolic t of the region, near the
bottom of each well.
Fig. 4
Fe3+
it is difcult to assess the signicance of

their results. We note that their results
contradict an earlier work of Straus and
Voth [21].
In a series of publications, Voth and
coworkers [21, 24, 25, 26, 27] explored various aspects of electron-transfer reactions.
Their calculations are based on a version
of the AndersonNewns Hamiltonian presented in Sect. 1.2.4.2, which allows the
direct computation of adiabatic free energy surfaces. For a Fe2+ /Fe3+ couple
at a xed distance to a Pt(111) surface,
they compared a classical and a quantized water model [24]. The quantization
of the high-frequency degrees of freedom

of water leads to a major shift in the thermodynamic driving force, and hence, in
the solvation of the redox couple. Remarkably, the reorganization energy is almost
unaffected; this means that the activation
energy at equilibrium does not change,
but only the equilibrium potential (Fig. 5).
When the constraint of a xed ion position is released and the ion is allowed
to move [25], the adiabatic free energy
surface is still very similar to the earlier
case, but the barrier close to equilibrium
is slightly higher. The ion is on average
further away from the surface, weakening

40
30
G(E)
[kcal]
20
Classical
10
0
Quantum
10
250
300
350
400
450
500
E
[kcal]
Adiabatic free energy curves for the Fe2+ /Fe3+ couple from
molecular dynamics [25]. The solid line represents results for the
quantized high-frequency degrees of water, the dashed line for the
classical degrees of freedom only.
Fig. 5
the electronic coupling. When the ion is

moved away from the surface, the thermodynamic driving force can change for the
same reason. The presence of a counterion causes nonmonotonical shifts in the
barrier height and driving force depending on the separation of the two ions.
For short distances, the reorganization energy is markedly increased. The inuence
of the electronic polarizability has been
investigated for a Na0 /Na+1 couple by introducing a uctuating charge on the water
molecules [26]. The electronic polarizability causes a stronger solvation of the ion
and a lower reorganization energy . The
polarizable water can accommodate the
change of the ion charge to some extent
and less solvent reorganization is needed.
Also, changes in the thermodynamic driving force are observed. For two different
neutral redox species [27], the free energy surface shows four distinct wells, one
for the uncharged reactants, two for the

transfer of an electron to one of the reactants, the other remaining neutral, and
one for the transfer of one electron to each
of the reactants (Fig. 6). Agreement with
an analytic model for such a process [29]
is satisfying, showing similar trends in the
driving force of the elementary steps.
The solvent dynamics of a one-dimensional system was explored by Kuznetsov
and Schmickler [30]. They singled out
one important solvent coordinate and
represented all the others by a heat bath.
This is a similar concept to analytical
theories such as Kramers [20]. Adiabatic
potential-energy surfaces were calculated
from the AndersonNewns Hamiltonian
(see Fig. 3). The result is a symmetric
double well, the barrier being the lower
and the less sharp, the higher the electronic
coupling . MD-simulations of a motion
on this surface coupled to the heat bath
43
1 Fundamentals
I(2)
F(E1, E2)
[kcal]
44
30
25
20
15
10
5
P
200
150
50
0
100
E1
[kcal]
I(1)
50
150
0
50
200
100
50
E2
[kcal]
Adiabatic free energy surface for the electron transfer between two different redox
couples and a metal electrode from molecular dynamic simulations [27].
Fig. 6
were performed. The time that was needed

to cross the transition state starting from
the reactant well with thermal velocity
yields the reaction rate. The higher the
friction exerted by the heat bath, the more
the motion resembles a random walk,
and the more likely is barrier recrossing
(Fig. 7). The authors compare the reaction
rates to the predictions of transitionstate theory, and obtain a transmission
coefcient . The results are in qualitative,
but not in quantitative agreement with
Kramers theory [20].
1.2.9
Bond Breaking Reactions
The models presented in the preceding

section consider only elementary electrontransfer steps. In the same sense, many
theories such as those of Marcus and Hush
are restricted to outer-sphere reactions.
For homogeneous electron transfer, bond
breaking was considered in early papers
by German and Dogonadze [31], and
later by German and Kuznetsov [32]. For
electrochemical reactions, bond breaking

was introduced into the MarcusHush
framework by Saveant [33, 34]. Koper
and Voth [35, 36] reformulated these
ideas in terms of an extension of the
model Hamiltonian of Eq. (12). The model
applies to a reaction of the type:
R X + e R + X
(34)
Typically, R is an organic fragment, and

X a halide anion. The electron-transfer
step and bond breaking are assumed
to be concerted; an example for such
a reaction is the reductive cleavage of
alkyl halides. The authors present both
an analytical theory, [35] and molecular
dynamic computer simulations [36], and
calculate adiabatic energy surfaces. In both
cases, the bond breaking is introduced
in a similar way following Saveants
ideas. The R X bond is described by a
Morse potential, the interaction between
the fragments R and X is described
by the repulsive part of the same Morse
potential. The following term is added to
the AndersonNewns Hamiltonian:

0.80
g = 1.98
0.75
0.70
0.65
(a)
0.60
g = 1.98
Distance
0.72
0.70
0.68
0.66
(b)
0.64
g = 18.1
0.750
0.625
0.500
0.375
0.00
(c)
0.25
0.50
0.75
1.00
Time
Molecular dynamic trajectories of a particle coupled to a heat

bath on the potential-energy surface of Fig. 1. The saddle point is
indicated by the dashed line. (a) and (b) are for relatively high
friction and (c) low friction. = 0.05 eV. The unit of time has been
chosen such that the period of the harmonic oscillator is 2 , and the
friction coefcient is given in terms of this unit.
Fig. 7
45
46
1 Fundamentals
HRX = (1 na )D{1 exp[(r r0 )]}2

+ na D exp[2(r r0 )]
(35)
where D is the dissociation energy of the

bond, and r0 its equilibrium distance. The
parameter determines the vibration frequencies and the anharmonicity. Since
the bond is broken when the orbital a
is occupied, that orbital is antibonding. All
vibrations are assumed to be classical. The
potential-energy surface calculated from
analytical theory [35] for small electronic
coupling shows a distinct well for the
reactants and an exit channel for the products; both are separated by a saddle. These
qualitative features are found again on the
free energy surface calculated from MD
simulations [36]; however, there are qualitative differences (Fig. 8). The product exit
channel shows a shallow well for short distances between the fragments. This is a
consequence of the microscopic solvation
effects, which are not included in the analytical theory. However, the barrier that
has to be overcome to separate the products is much lower than the barrier for
the electron-transfer reaction itself, so fragment dissociation is never the rate-limiting
step. Comparison to a linear response
model on the basis of the results of the analytical theory, show quantitative deviations
for the minima and curvature of the surface, as well as for the overpotential. These
differences arise from nonlinear response
of the solvent to the change in the charge
on the redox center and the volume exclusion of the R fragment. Thus, the analytical
theory gives rather a rough qualitative picture, but serves as a helpful visualization
of the basic characteristics of the reaction.
The case of weak coupling, in which
the rst-order perturbation theory applies,
was explored by Schmickler [37]. For the
parameters chosen, the major contribution
to the reaction rate comes from the
ground state of the bonding mode, but

the energy of the nal state is much
larger than the thermal energy kB T .
This is energetically unfavorable, but the
overlap with higher energetic states is
better. As a consequence, the transfer
coefcient at equilibrium is larger
than 1/2. At higher overpotentials,
approaches 0 as familiar from outer-sphere
electron-transfer reactions. In comparison
to Saveants theory, the activation energy
is smaller owing to the inclusion of
tunneling effects.
1.2.10
Conclusion
After fty years of research, the theory

of electrochemical electron-transfer reactions is well developed, and has provided
us with a basic understanding of the essentials. However, a number of important
questions are still open, and remain the
subject of ongoing research. We list what
we believe to be the most important points:
At present, we have a theory for weak
coupling on the basis of rst-order perturbation, and for strong coupling, on
the basis of the adiabatic limit. However,
the intermediate case is unexplored; in
particular, we do not fully know how
the rate constant will behave as a function of the electronic coupling. A recent
paper by Mohr and Schmickler [38] investigates this behavior, but the effects
of solvent dynamics are not included in
this work so that it will apply to very
short timescales only.
Although computer simulations have
provided us with some insights into
deviations from the harmonic approximation, much more work needs to
be done in order to provide us with
accurate potential-energy surfaces, and
F (E, r )
[kcal mole1]
100
50
0
50
0
50
100
[kcal mole1] 150

200
250
300
(a)
r
[Angstroms]
F (E, r )
[kcal mole1]
100
50
0
50
0
50
100
E
150
[kcal mole1]
200
250
300
(b)
r
[Angstroms]
Adiabatic free energy surface for simultaneous electron transfer and

bond breaking from MD simulations (upper) and a linear response solvent
model (lower panel) [36].
Fig. 8
also to explore subtle effects such as the

variation of the electronic overlap with
electrode potential.
Solvent dynamics have been explored
mainly within a wholly classical context;
its effect on quantum modes remains unexplored.

During the last decade, long-range electron transfer along molecular chains
has become very interesting, not least
47
48
1 Fundamentals
because of its application to biological

systems. While the superexchange theory [39, 40] is a good rst approximation,
it leaves many questions, in particular,
the coupling to solvent and vibrational
modes, open.
With these and other problems still
open, the electron-transfer theory remains
a vibrant eld in physics, chemistry and
biology alike.
Acknowledgment
Financial support by the Volkswagenstiftung is gratefully acknowledged.

References
2. N. S. Hush, J. Chem. Phys. 1958, 28, 962.
3. A. N. Kuznetsov, Charge Transfer in Physics,
Chemistry and Biology, Gordon & Breach,
Longhorne, Pa., 1995.
4. R. Memming, Semiconductor Electrochemistry, Wiley-VCH, Weinheim, Germany,
2000.
5. V. G. Levich, Kinetics of reactions with
charge transport in Physical Chemistry,
An Advanced Treatise (Eds.: H. Eyring,
D. Henderson, W. Jost), Academic Press,
New York, 1970, Vol. IXb, pp. 9861014.
6. W. Schmickler, Interfacial Electrochemistry,
Oxford University Press, New York, 1996.
7. H. Gerischer, Z. Phys. Chem. N.F. 1960, 26,
223.
8. H. Gerischer, Z. Phys. Chem. N.F. 1961, 27,
40, 48.
9. R. R. Dogonadze, Theory of molecular electrode kinetics in Reactions of Molecules at Electrodes (Ed.: N. S. Hush), Wiley-Interscience,
London, 1971, pp. 135227.
10. W. Schmickler, Ber. Bunsen-Ges. Phys. Chem.
1976, 80, 834.
11. J. S. Bader, R. A. Kuharski, D. Chandler, J.
Chem. Phys. 1990, 93, 230.
12. W. Schmickler, J. Electroanal. Chem. 1986,
204, 31.
13. A. J. Legget, S. Chakravarty, A. T. Dorsey
et al., Rev. Mod. Phys. 1987, 59, 1.
14. D. M. Newns, Phys. Rev. 1969, 178, 1123.

15. P. P. Schmidt, Specialist Periodical Reports,
Electrochemistry, The Chemical Society, London, 1975, Vol. 5, pp. 21131.
16. S. Tanaka, C.-P. Hsu, J. Chem. Phys. 1999,
111, 11 117.
17. W. Schmickler, Chem. Phys. Lett. 1995, 237,
152.
18. W. Schmickler, Electrochim. Acta 1996, 41,
2329.
19. H. Sumi, R. Marcus, J. Chem. Phys. 1986, 84,
4894.
20. H. A. Kramers, Physica 1940, 7, 284.
21. J. B. Straus, G. A. Voth, J. Phys. Chem. 1993,
97, 7388.
22. D. A. Rose, I. Benjamin, J. Chem. Phys. 1994,
100, 3545.
23. B. B. Smith, J. W. Halley, J. Chem. Phys.
1994, 101, 10 915.
24. J. B. Straus, A. Calhoun, G. A. Voth, J. Chem.
Phys. 1995, 102, 529.
25. A. Calhoun, G. A. Voth, J. Phys. Chem. 1996,
100, 10 746.
26. A. Calhoun, G. A. Voth, J. Electroanal. Chem.
1998, 450, 253.
27. A. Calhoun, G. A. Voth, J. Chem. Phys. 1998,
109, 4569.
28. R. F. Grote, J. T. Hynes, J. Chem. Phys. 1980,
95, 2715.
29. Y. G. Boroda, A. Calhoun, G. A. Voth, J.
Chem. Phys. 1997, 107, 8940.
30. A. M. Kuznetsov, W. Schmickler, Chem.
Phys. Lett. 2000, 327, 314.
31. E. D. German, R. R. Dogonadze, Int. J.
Chem. Kinet. 1974, 6, 467.
32. E. D. German, A. M. Kuznetsov, J. Phys.
Chem. 1994, 98, 6120.
33. J. M. Saveant, J. Am. Chem. Soc. 1987, 109,
6788.
34. J. M. Saveant, Acc. Chem. Res. 1993, 26, 455.
35. M. T. M. Koper, G. A. Voth, Chem. Phys. Lett
1998, 282, 100.
36. A. Calhoun, M. T. M. Koper, G. A. Voth, J.
Phys. Chem. B 1999, 103, 3442.
37. W. Schmickler, Chem. Phys. Lett. 2000, 317,
458.
38. J. H. Mohr, W. Schmickler, Phys. Rev. Lett.
2000, 84, 1051.
39. H. Taube, Electron Transfer Reactions of
Complex ions in Solutions, Academic Press,
New York, 1970.
40. D. N. Beratan, J. J. Hopeld, J. Am. Chem.
Soc. 1984, 106, 1584.
1.3 Electronic Tunnel Factors in Molecular Electron Transfer and Molecular Conduction
1.3
1.3.1
Electronic Tunnel Factors in Molecular

Electron Transfer and Molecular
Conduction
Introduction
Abraham Nitzan
Tel Aviv University, Tel-Aviv, Israel
Abstract
Electron transmission through molecules

and molecular interfaces has been a subject of intensive research owing to recent
interest in electron transfer phenomena
underlying the operation of the scanning tunneling microscope (STM) on one
hand, and in the transmission properties
of molecular bridges between conducting
leads, on the other. In these processes,
the traditional molecular view of electron transfer between donor and acceptor
species give rise to a novel view of the
molecule as a current carrying conductor,
and observables such as electron transfer rates and yields are replaced by the
conductivities, or more generally by currentvoltage relationships, in molecular
junctions. Such investigations of electrical junctions in which single molecules
or small molecular assemblies operate as
conductors, constitute a major part of what
has become the active eld of molecular
electronics.
In this chapter, the author reviews
the current knowledge and understanding of this eld with particular emphasis on theoretical issues, and on the
relationship between this new look at
electron transfer phenomena and the traditional study of molecular electron transfer in solution. Different approaches to
computing the conduction properties of
molecules and molecular assemblies are
reviewed, and the relationships between
them is discussed.
Electron transfer, a fundamental chemical process underlying all redox reactions,

has been under experimental and theoretical study for many years [16]. Theoretical
studies of such processes seek to understand the ways in which their rate depends
on donor and acceptor properties, on the
solvent, and on the electronic coupling
between the states involved. The different roles played by these factors and the
way they affect qualitative and quantitative aspects of the electron transfer process
have been thoroughly discussed in the past
half-century. This kind of processes, which
dominate electron transitions in molecular
systems, is to be contrasted with electron
transport in the solid state, that is, in metals and semiconductors. Electrochemical
reactions that involve both molecular and
solid-state donor/acceptor systems, bridge
the gap between these phenomena [6].
Here, electron transfer takes place between
quasi-free electronic states on one side,
and bound molecular electronic states on
the other.
The focus of the present discussion is
another class of electron transfer phenomena: electron transmission between
two regions of free or quasi-free electrons
through molecules and molecular layers.
Examples for such processes are photoemission (PE) through molecular overlayers, the inverse process of low energy
electron transmission (LEET) into metals through adsorbed molecular layers and
electron transfer between metal, and/or
semiconductor contacts, through molecular spacers. Figure 1 depicts a schematic
view of such systems. The standard
electron transfer model in Fig. 1(a) shows
donor and acceptor sites connected by a
molecular bridge. In Fig. 1(b), the donor
49
50
1 Fundamentals
B
D
(a)
Fig. 1 Schematic views of typical

electron transmission systems. (a) A
standard electron transfer system
containing a donor, an acceptor, and a
molecular bridge connecting them (not
shown are nuclear motion baths that
must be coupled to the donor and
acceptor species). (b) A molecular
bridge connecting two electronic
continua, L and R, representing, e.g.,
two metal electrodes (c) same as
(b) with the bridge replaced by a
molecular layer.
(b)
(c)
and the acceptor are replaced by a continua

of electronic states representing free space
or metal electrodes. (This replacement can
occur on one side only, representing electron transfer between a molecular site and
an electrode). In Fig. 1(c), the molecular
bridge is replaced by a molecular layer.
A schematic view of the electronic states
involved is shown in Fig. 2. The middle
box represents the bridging molecule or
molecular layer and a set of levels {n}
represents the relevant molecular orbitals.
In a standard electron transfer system,
(Fig. 2a) this bridge connects the donor
and acceptor species, now represented
by potential surfaces associated with the
vibronic structure of the corresponding
intramolecular and solvent nuclear motions. When the bridge connects two metal
electrodes (or separates a metal substrate
from vacuum), these nuclear baths are replaced by manifolds of electronic states
{l} and {r} that represent the continua of
free or quasi-free electron states in the

substrates (or, depending on the process,
in vacuum). In addition, coupling to the
thermal environment (represented by the
box ) may affect transmission through
the bridge. The double arrows in the Figure
represent the couplings between these different subsystems.
The rst two of the examples given
above, PE and LEET, involve electrons
of positive energy (relative to zero kinetic
energy in vacuum), and as such are related to normal scattering processes. The
third example, transmission between two
conductors through a molecular layer, involves negative energy electrons, and as
such is closely related to regular electron
transfer phenomena. The latter type of
processes has drawn particular attention
in recent years owing to the growing interest in conduction properties of individual
molecules and of molecular assemblies.
Such processes have become the subjects
{n}
(a)
{l}
{n}
{r}
(b)
Fig. 2 A schematic view of the electronic and nuclear states involved in

typical electron transmission systems. See text for details.
of intensive research owing to recent

interest in electron transfer phenomena
underlying the operation of the STM
on one hand, and in the transmission
properties of molecular bridges between
conducting leads on the other. In the latter case, the traditional molecular view of
electron transfer between donor and acceptor species, gives rise to a novel view
of the molecule as a current carrying conductor, and observables such as electron
transfer rates and yields are replaced by
the conductivities, or more generally, by
currentvoltage relationships, in molecular junctions. Of primary importance is
the need to understand the interrelationship between the molecular structure of
such junctions and their function, that
is, their transmission and conduction
properties. Such investigations of electrical junctions in which single molecules or

small molecular assemblies operate as conductors connecting traditional electrical
components, such as metal or semiconductor contacts, constitute a major part
of what has become the active eld of
molecular electronics [717]. Their diversity, versatility and amenability to control and manipulation, make molecules
and molecular assemblies potentially important components in nano-electronic
devices. Indeed, basic properties pertaining to single electron transistor behavior
and to current rectication have already
been demonstrated. At the same time,
major difculties lie on the way to real
technological applications [18]. These difculties stem from problems associated
51
52
1 Fundamentals
with the need to construct, characterize,

control and manipulate small molecular
structures at conned interfaces with a
high degree of reliability and reproducibility, and from issues of stability of such
small junctions.
It should be obvious that while the
different processes outlined above correspond to different experimental setups,
fundamentally they are controlled by similar physical factors. Broadly speaking, we
may distinguish between processes for
which lifetimes or rates (more generally, the
time evolution) are the main observables,
and those that monitor uxes or currents.
In this review, we focus on the second
class, which may be further divided into
processes that measure currentvoltage
relationships, mostly near equilibrium,
and those that monitor the nonequilibrium
electron ux, for example, in photoemission experiments.
The focus of the following discussion is
electron transfer between two conducting
electrodes through a molecular medium.
Such processes bear strong similarity to
the more conventional systems that involve
at least one molecular species in the
donor/acceptor pair. However, important
conceptual issues arise from the fact that
such systems can be studied as part
of complete electrical circuits, providing
currentvoltage characteristics that can be
analyzed in terms of molecular resistance,
conductance, and capacitance.
1.3.2
Standard Electron Transfer Theory
To set the stage for our later discussion,

we rst briey review the rate expressions
for standard electron transfer processes
(Figs. 1a, 2a, 3a). We focus on the
particular limit of nonadiabatic electron
transfer, where the electron transfer rate is
given (under the Condon approximation)

by the golden rule based expression
ket =
2
|VDA |2 F
h
(1)
where VDA is the coupling between the

donor (D) and acceptor (A) electronic states
and where
F = F(EAD ) =

D
Pth (D (D ))
|D |A |2 (A (A ) D (D ) + EAD )

(2)
is the thermally averaged and Franck
Condon (FC) weighted density of nuclear
states. In Eq. (2), D and A denote donor
and acceptor nuclear states, Pth is the
Boltzmann distribution over donor states,
D (D ) and A (A ) are nuclear energies
above the corresponding electronic origin
and EAD = EA ED is the electronic
energy gap between the donor and acceptor
states. In the classical limit, is given by
e(+EAD ) /4kB
4kB
2
F(EAD ) =
(3)
where kB is the Boltzmann constant and

is the temperature, and where is the
reorganization energy, a measure of the
electronic energy that would be dissipated
after a sudden jump from the electronic
state describing an electron on the donor
to that associated with electron on the
acceptor. If the donor (say) is replaced
by an electrode [6, 19, 20], we have to sum
over all occupied electrode states
|VDA |2 F

=
f (k )F(k e)|VkA |2
df ()F( e)

k
( k )|VkA |2
(4)
(a)
E
N+1
=A
0=D
(b)
Fig. 3 Simple level structure models (a) for molecular electron transfer and (b) for electron
transmission. The molecular bridge is represented by a simple set of levels that represent local
orbitals of appropriately chosen bridge sites. This set of levels is coupled to the donor and
acceptor species (with their corresponding nuclear environments) in (a), and to electronic
continua representing metal leads (say) in (b). In the latter case, the physical meaning of states 0
and N + 1 depends on the particular physical problem: They can denote donor and acceptor
states coupled to the continua of environmental states (hence the notation 0 = D, N + 1 = A),
surface localized states in an metal-molecule-metal junction, or they can belong to the right and
left scattering continua.
where
1
f () =
1 + e/kB
(5)
is the FermiDirac distribution function

with measured relative to the electron
chemical potential in the electrode,
and , which determines the position
of the acceptor level relative to is the
overpotential. Dening

( k )|VkA |2 |V ()|2 (6)
k
the electron transfer rate takes the

form
ket =
2
h
e(e+) /4kB
4kB
2
|V ()|2 f ()
(7)
Note that the reorganization energy that

appears in Eq. (7) is associated with the
change in the redox state of the molecular species only. The nominal change in
the oxidation state of the macroscopic
electrode does not affect the polarization
state of the surrounding solvent because
the transferred electron or hole does not
stay localized.
53
54
1 Fundamentals
Much of the early work on electron transfer have used expressions such as (1 and
7) with the electronic coupling term VDA
being used as a tting parameter. More
recent work has focused on ways to characterize the dependence of this term on the
electronic structure of the donor/acceptor
pair and on the environment. In particular,
studies of bridge-mediated electron transfer, where the donor and acceptor species
are rigidly separated by molecular bridges
of well-dened structure and geometry,
have been very valuable for characterizing the interrelationship between structure
and functionality of the separating environment in electron transfer processes. As
expected for a tunneling process, the rate
is found to decrease exponentially with the
donor-acceptor distance

ket = k0 e RDA
(8)
where is the range parameter that characterizes the distance dependence of the
electron transfer rate. The smallest values for are found in highly conjugated
organic bridges for which is in the
range of 0.2 to 0.6 A 1 [2132]. In contrast, for free space, taking a characteristic
ionization
barrier UB = 5 eV, we nd =

2
8mUB /h 2.4 A 1 (m is the electron
mass). Lying between these two regimes
are many motifs, both synthetic and natural, including cytochromes and docked
proteins [3341], DNA [4250], and saturated organic molecules [5157]. Each
displays its own characteristic range of
values, and hence its own timescales and
distance dependencies of electron transfer. A direct measurement of along
a single molecular chain was recently
demonstrated [58].
In addition, to bridge assisted transfer
between donor and acceptor species, electron transfer has been studied in systems
in which the spacer is a well-characterized

LangmuirBlodgett lm [5961]. The
STM provides a natural apparatus for such
studies [58, 6276]. Other approaches include break junctions [7779] and mercury drop contacts [8084].
Simple theoretical modeling of VDA
usually relies on a single electron (or hole)
picture in which the donor-bridge-acceptor
(DBA) system is represented by a set of
levels: |D, |A, {|1, . . . |N } as depicted
in Fig. 3. In the absence of coupling
of these bridge states to the thermal
environment, and when the energies En
(n = 1, . . . , N ) are high relative to the
energy of the transmitted electron (the
donor/acceptor orbital energies in Figs. 1a,
2a and 3a, or the incident electron energy
in Figs. 1b to c, 2b and 3b), this is
the superexchange model for electron
transfer [85]. Of particular interest are
situations where {n} are localized in space,
so that the state index n corresponds to the
position in space between the donor and
acceptor sites (Fig. 3a) or between the two
electron reservoirs (Fig. 3b). These gures
depict generic tight binding models of this
type, where the states n = 1, . . . , N are the
bridge states, here taken as degenerate
in zero order. Their localized nature
makes it possible to assume only nearestneighbor coupling between them, that is,
Vn,n = Vn,n1 n ,n1 . We recall that the
appearance of VDA in Eq. (1) is a loworder perturbation theory result. A more
general expression is obtained by replacing
VDA by TDA where the T operator is
dened by T (E) = V + V G(E)V , with
G(E) = (E H + (1/2)i#)1 and where
# stands for the inverse lifetime matrix
of bridge levels. Assuming for simplicity
that the donor level |D is coupled only
to bridge state |1 and that the acceptor
level |A is coupled only to bridge level N ,
the effective coupling between donor and
acceptor is given by
TDA (E) = VDA + VD1 G1N (E)VN A
(9)
This naturally represents the transition

amplitude as the sum of a direct contribution, VDA , which is usually disregarded
for long bridges and a bridge mediated contribution. In using TDA instead of VDA in
Eq. (1), the energy parameter E in Eq. (9)
should be taken to be equal to ED = EA
at the point where the corresponding
potential surfaces cross (or go through
an avoided crossing). In the level structure of Fig. 3(a) that corresponds to the
DBA system in Fig. 1(a), making the tight
binding approximation and in the weak
coupling limit, max |V | min(EB E),
the Greens function element in (9) is
given by
G1N (E) =
N
1

1
Vn,n+1
E EN
E En
(10)
n=1
For a generalization of Eq. (10) that

does not assume weak coupling see
Refs. [8688]. For a model with identical
bridge segments, En and Vn,n+1 are
independent of n and will be denoted
En = EB and Vn,n+1 = VB . Using this in
Eq. (1) leads to
ket =
2
h

V1D VN A 2 VB 2N

F
%E
V
B
B
(11)
where %EB = EB E. Similarly, for
a bridge-assisted transfer between a
molecule and an electrode, Eq. (7) applies
with |V ()|2 given by

2
|V ()| =
VB
%EB
2N
( k )

V1k VN A 2

VB
(12)
These results imply a simple form for the

distance parameter of Eq. (8)

%EB
2

= ln
(13)
a
VB
where a measures the segment size, so that
the bridge length is Na. The exponential
dependence on the bridge length is a manifestation of the tunneling character of this
process. For typical values, for example,
Eq. (13) gives
%EB /VB = 10 and a = 5 A,
= 0.92 A 1 . More rigorous estimates of
the electronic coupling term in electron
transfer processes involve electronic structure calculation for the full DBA system.
Such calculations, in the context of molecular conduction, will be discussed later.
1.3.3
Transmission Between Conducting Leads
Equations (1, 7 and 11) are expressions

for the rate of electron transfer between
donor and acceptor molecules or between a
molecule and a metal electrode. As already
mentioned, for electron transfer in metalmolecule-metal (MMM) junctions, the
primary observable is the currentvoltage
characteristics of the system. Putting it in
another way, while the primary observable
in standard charge transfer processes involving molecular donors and/or acceptors
is a transient quantity, in MMM junctions
we focus on the steady state current through
the junction for a given voltage difference
between the two metal ends. Note that for
electron transfer processes, in addition to
rates, other observables are the yields of
different products of the electron transfer
reaction. Furthermore, for light induced
electron transfer processes, the steady state
current under a constant illumination can
be monitored.
Consider rst a simple model for a
metal-insulator-metal (MIM) system in
55
56
1 Fundamentals
which the insulator is represented by a

continuum characterized by a dielectric
constant [89]. For specicity, assume that
the electrode surfaces are innite parallel
planes perpendicular to the x direction.
In this case, the transmission problem is
essentially one-dimensional and depends
only on the incident particle
velocity in
the x direction, x = 2Ex /m. In the

WKB approximation, the transmission
probability is given by

4 s2
T (Ex ) = exp
[2m(UB (x)
h s1

(14)
Ex )]1/2 dx
where UB (x) is the barrier potential that
determine the turning points s1 and s2
and m is the mass of the tunneling
particle. Thetunneling ux is given by
T (Ex )n(Ex ) 2Ex /m, where n(Ex ) is the
density per unit volume of electrons of
energy Ex in the x direction. n(Ex ) is
obtained by integrating the FermiDirac
function with respect to Ey and Ez . When
a potential is applied so that the right
electrode (say) is positively biased, the net
current density is obtained in the form [89]

J =
dEx T (Ex )(Ex )
(15)
0
where
(Ex ) =
2m2 e
(2 h)3
dy
f (E + e)] =
dz [f (E)
4me
(2 h)3
[f (E) f (E + e)]
dEr
0
(16)
and where Er = E Ex = (1/2)m(y2 +

z2 ) is the energy in the direction perpendicular to x. In obtaining this result,
it is assumed that the electrodes are
chemically identical. At zero temperature,
and when 0, f (E) f (E + e) =

e(E EF ). Equations (15 and 16) then
lead to an expression for the conduction
per unit area, that is, the conductivity per
unit length

4me2 EF
dEx T (Ex )
(17)
x =
(2 h)3 0
For nite , these expressions provide a framework for predicting the
currentvoltage characteristics of the junction; explicit approximate expressions
were given by Simmons [89]. Here we only
emphasize [89], that the dependence on
arises partly from the structure of Eqs. (15)
and (16), for example, at zero temperature
EF e
4m2 e2
J =
dEx T (Ex )
e
(2 h3 )
0

EF
dEx (EF Ex )T (Ex )
+
EF e
(18)
but mainly from the voltage dependence of
T. The simplest model for a metal-vacuummetal barrier between identical electrodes
without an external eld is a rectangular
barrier of height above the Fermi energy
given by the metal work function. When a
uniform electric eld is imposed between
the two metals, a linear potential drop
from EF on one electrode to EF e on
the other is often assumed (see Fig. 4). In
addition, the image potential experienced
by the electron between the two metals will
considerably modify the potential barrier.
For a point charge e, located at position
x between two conducting parallel plates
that are a distance d apart, the image potential is

e2
VI =
4

nd
1
1
+
2x
[(nd)2 x 2 ] nd
n=1
(19)
EF
EF
EF
EF
Tunneling gap between two metal electrodes in an unbiased

(left) and a biased (right) situation. The bare gap, given by the work
function W, is modied by the image interaction the resulting
barriers are represented by the curved lines.
Fig. 4
where is the dielectric constant of the

spacer. For x = d/2, this becomes
VI =
e2 ln 2
2d
(20)
This negative contribution to the electrons energy reduces the potential barrier
(Fig. 4), and has been invoked to explain
the lower than expected barrier observed
in STM experiments [90, 91]. Some points
should however, be kept in mind. First,
the classical result (19) fails close to
the metal surface where quantum mechanical and atomic size effects change
both the position of the reference image plane and the functional form of the
image potential [9296]. Second, consideration of the dynamic nature of the image
response should be part of a complete theory [97100]. The timescale of electronic
response of metals can be roughly estimated from the plasma frequency to be
1016 s. This should be compared to the
time during which a tunneling particle can
respond to interactions localized in the barrier. For transmitted particles, this is the
traversal time for tunneling [101, 102] that,
for an electron traversing a 10 A wide 1 eV
barrier is of the order of 1 fs. This comparison would justify the use of the static
image approximation in this context, but
this approximation becomes questionable
for deeper tunneling or narrower barriers.
The planar geometry implied by the

assumption that transmission depends
only on the energy of the motion parallel
to the tunneling direction, as well as the
explicit form of Eq. (14) are not valid for a
typical STM conguration that involves a
tip on one side and a structured surface on
the other. To account for these structures,
Tersoff and Hamman [103] have applied
Bardeens formalism [104], which is a
perturbative approach to tunneling in
arbitrary geometries. Bardeens formula
for the tunneling current is
4e
[f (El )(1 f (Er + e))
I=
h
l,r
(1 f (El ))f (Er + e)]|Mlr |2

2e
(El Er ) =
[f (El )
h
l,r
f (Er + e)]|Mlr | (El Er ) (21)

where
h2
Mlr =
2m
l r l r ) (22)
dS(
is the transition matrix element for the

tunneling process. This is just the Golden
rule rate expression multiplied by the
electron charge e, with M playing the
role of coupling. In Ref. [103], only the
rst term in the square brackets of the
rst line appears. This gives the partial
current from the negative to the positive
57
58
1 Fundamentals
electrode. The net current is obtained by

subtracting the reverse current as shown
in Eq. (21). Also, compared with Ref. [103],
Eq. (21) contains an additional factor of 2
that accounts for the spin multiplicity of
the electronic states.
In Eqs. (21) and (22), 4 and r are
electronic eigenstates of the negatively
biased (left) and positively biased (right)
electrodes, respectively, is the bias
potential, and the integral is over any
surface separating the two electrodes
and lying entirely in the barrier region.
The wave functions appearing in Eq. (22)
are eigenfunctions of Hamiltonians that
describe each electrode in the absence of
the other, that is, interfaced with an innite
spacer medium. These functions therefore
decay exponentially in the space between
the two electrodes in a way that reects
the geometry and chemical nature of the
electrodes and the spacer. For 0,
Eq. (21) yields the conduction
g
I
4e2
|Mlr |2 (El EF )
=

h
l,r
(Er EF )
(23)
Tersoff and Hamman [103] have used

substrate wave functions that correspond
to a corrugated surface of a generic metal
while the tip is represented by a spherical
s orbital centered about the center r0 of the
tip curvature. In this case, they nd

| (r0 )|2 (E EF )
(24)
I
The rhs of Eq. (24) is the local density

of states of the metal. While this result is
useful for analysis of spatial variation of the
tunneling current on a given metal surface,
the contributions from the coupling matrix
elements in Eq. (23) cannot be disregarded
when comparing different metals and or
different adsorbates [20].
1.3.4
The Landauer Formula
The results (14)(17) and (21)(23) are

special cases of a more systematic representation of the conduction and the
currentvoltage characteristic of a given
junction due to Landauer [105, 106]. Landauers original result was obtained for a
system of two one-dimensional leads connecting two macroscopic electrodes (electron reservoirs) via a scattering object or
a barrier characterized by a transmission
function T (E). The zero-temperature conductance, measured as the limit 0 of
the ratio I / between the current and the
voltage drop between the reservoirs, was
found to be
g=
e2
T (EF )
h
(25)
(The corresponding resistance, g 1 , can

be represented as a sum of the intrinsic resistance of the scatterer itself,
[(e2 / h)(T/(1 T ))]1 and a contribution (e2 / h)1 from two contact resistances between the leads and the reservoirs. See Chapter 5 of Ref. [107] for a
discussion of this point.) This result is
obtained by computing the total unidirectional current carried in an ideal lead
by electrons in the energy range (0, E) =
2 /(2m)). In a one-dimensional sys(0, h2 kE
tem of length L, the density of electrons, including spin, with wave vectors in the range between k and k +
dk is n(k) dk = 2(1/L)(L/2)f (Ek ) dk =
f (Ek ) dk/. The corresponding velocity is
= hk/m. Thus

I (E) = e
kE
0

dk(k)n(k) = e
e
f (Ek )
=
E
0
kE
dk
0
dE f (E )

hk
m
(26)
At zero temperature, the net current

carried under bias is

e
dE(f (E) f (E + e))
I =
h 0
0
e2

h
(27)
Thus, the conductance of an ideal

one-dimensional lead is I / = e2 / h =
(12.9K8)1 . In the presence of the scatterer, this is replaced by

e
dE T (E)(f (E) f (E + e))
I =
h 0
0, 0
e2
T (EF )
h
e2
Tij (E)
h
e2
e2
Tij (EF ) =
Tr(SS )EF
h
h
i,j
(29)
where Tij = |Sij |2 is the probability that
a carrier coming from the left (say) of
the scatterer in transversal mode i will be
transmitted to the right into transversal
mode j (Sij , an element of the S matrix, is
the corresponding amplitude). The sum
in (29) is over all transversal modes
whose energy is smaller than EF . The
analog of Eq. (29) for the microcanonical
chemical reaction rate was rst written by
Miller [109]. More generally, the current
for a voltage difference between the
electrodes is given by

g(E)
dE[f (E) f (E + e)]
I=
e
0
(30)
(31)
i,j
As an example, consider the case of

a simple planar tunnel junction (see
Eqs. (14)(17)), where the scattering process does not couple different transversal
modes. In this case, the transmission function depends only on the energy in the
tunneling direction

Ly Lz
dky
Tij (E) =
Tii (E) =
(2)2
i,j
dkz T E
(28)
which leads to (25). This result is valid for

one-dimensional leads. When the leads
have nite size in the direction normal
to the propagation so that they support
traversal modes, a generalization of (25) to
this case yields [108]
g=
g(E) =
Ly Lz 2m
(2)2 h2

0

h2
2
2
(ky + kz )
2m
dEr T (E Er )
(32)
Er is dened below Eq. (16). Using this

in Eq. (29) yields the conductivity per
unit length

g
4me2 EF
=
dEx T (Ex )
Ly Lz
h3
0
(33)
in agreement with Eq. (17).
Similarly, Eqs. (21 and 23) are easily seen
to be equivalent to Eqs. (25 or 31) if we
identify Mlr with Tlr in Eq. (37) below. An
important difference between the results of
(29 and 31) and results based on Bardeens
formalism, that is Eqs. (21 to 23), is that the
former are valid for any set of transmission
probabilities, even close to 1, while the
latter yields a weak coupling result.
Another important conceptual difference
is the fact that the sums over l and r in
Eqs. (21 to 23) is over zero order states
dened in the initial and nal subspaces,
while the sums in Eqs. (29 to 31) is over
scattering states, that is, eigenstates of
the exact systems Hamiltonian. It is the
essence of Bardeens contribution [104]
that in the weak coupling limit (i.e.
59
60
1 Fundamentals
high/wide barrier), it is possible to write

the transmission coefcient Tij in terms
of a golden rule expression for the
transition probability between the zero
order standing wave states |l and |r
localized on the left and right electrodes,
thus establishing the link between the
two representations. (For an alternative
formulation of this link, see Ref. [110])
To explore this connection on a more
formal basis, we can replace the expression based on transmission coefcients
T by an equivalent expression based on
scattering amplitudes, or T matrix elements, between zero order states localized
on the electrodes. This can be derived
directly from Eqs. (29 or 31) by using
the identity
Tij (E) = 4 2
i,j
|Tlr |2

=
l,r
dE[f (E) f (E + e)]
g(E)
e
(36)
where
g(E)
4e2
|Tlr |2
h l,r
(E El )(E Er ) (37)
Note that Eqs. (34 and 37) imply again

Eq. (31). For 0, Eqs. (36 and 37) lead
to I = g with
l,r
(E El )(E Er )
4
[f (El )(1 f (Er + e))
h
l,r
f (Er + e)(1 f (El ))]|Tlr |2

4e
(El Er ) =
[f (El )
h
f (Er + e)]|Tlr | (El Er )
(38)
The analogy of this derivation of the result

(23) is evident.
1.3.5
Molecular Conduction
Equations (36 to 38) provide a convenient

starting point for most treatments of currents through molecular junctions where
the coupling between the two metal electrodes is weak. In this case, it is convenient
to write the systems Hamiltonian as the
sum, H = H0 + V , of a part H0 that represents the uncoupled electrodes and spacer
and the coupling V between them. In the
weak coupling, limit the T operator
T (E) = V + V G(E)V ;
G(E) = (E H + i)1
l,r
g = g(EF )
(34)
On the left side of Eq. (34), a pair of

indices (i, j ) denote an exact scattering
state of energy E, characterized by an
incoming state i on the left (say) electrode
and an outgoing state j on the right
electrode. On the right, l and r denote
zero order states conned to the left and
right electrodes, respectively. T is the
corresponding transition operator whose
particular form depends on the details
of this connement. Alternatively, we can
start from the golden-rule-like expression
I =e
(An additional factor of 2 on the rhs

accounts for the spin degeneracy). It is
convenient to recast this result in the form

4e
I=
dE[f (E) f (E + e)]
h 0

|Tlr |2 (E El )(E Er )
(35)
(39)
is usually replaced by its second term

only. The rst direct term V can be
disregarded if we assume that V couples

the states l and r only via states of the
molecular spacer. Consider now a simple
model where this spacer is an N -site
bridge connecting the two electrodes so
that Site 1 of the bridge is attached to the
left electrodes and Site N to the right
electrode, a variant of Fig. 3(b). In this
case, we have Tlr = Vl1 G1N VN r , so that at
zero temperature [111]

Tij (E) = |G1N (EF )|2
VB =
(L)
(R)
#1 (EF )#N (EF )
(40)
and (using Eqs. 36 and 37)

I () =
e
h
EF
(L)
EF e
(R)
#N (E
dE|G1N (E, )|2 #1 (E)
+ e)
(41)
with
(L)
#1 (E) = 2
|Vl1 |2 (El E);
(R)
#N (E)
= 2
|VN r |2 (EN E) (42)
The Greens function in Eq. (40) is itself reduced to the bridges subspace
by projecting out the metals degrees of
freedom. This results in a renormalization of the bridge Hamiltonian: in the
bridge subspace
(E HB ;B (E))1
(43)
where HB = HB0 + VB is the Hamiltonian

of the isolated bridge entity with
HB0 =
N

n=1
En |nn|;
(44)
and where, in the basis of eigenstates of

HB0
;nn (E) = n,n (n,1 + n,N )[<n (E)

12 i#n (E)]
(45)

|Vnj |2 (E Ej ) (46)
#n (E) = 2
j
PP
2
dE
#n (E )
(E E )
(47)
In Eq. (46), the sum is over both the right

and the left manifolds (that is, j goes over
all states {l} and {r} in these manifolds),
(L)
(R)
so that #n = #n + #n ; n = 1, N . The
transmission problem is thus reduced
to evaluating a Greens function matrix
element and two width parameters. The
rst calculation is a simple inversion of
a nite (order N ) matrix. The width #
and the associated shift < represent the
nite lifetime of an electron on a molecule
adsorbed on the metal surface, and
can be estimated, for example Ref. [111],
using the NewnsAnderson model of
chemisorption [112]. In the simple tightbinding model of the bridge and in the
weak coupling limit, G1N is given by
Eq. (10) modied by the inclusions of the
self-energy terms
G1N (E)
=
(E H + i)1
Vn,n |nn |
n=1 n =1
<n (E) =
i,j
N
N
V1,2
(E E1 ;1 (E))(E EN ;N (E))
N1

j =2
Vj,j +1
E Ej
(48)
Equations (40 to 48) thus provide a

complete simple model for molecular
conduction, equivalent to similar approximations used in theories of molecular
61
62
1 Fundamentals
electron transfer (e.g. [113, 114] and references therein). For applications of variants
of this formalism to electron transport in
specic systems, see Refs. [86, 87, 115,
116]. In the following text, we discuss more
general forms of this formulation.
1.3.6
Relation to Electron Transfer Rates
It is interesting to examine the relationship between the conduction of a

molecular species and the electron transfer properties of the same species [117].
We should keep in mind that because
of tunneling there is always an Ohmic
regime near zero bias, with conduction
given by the Landauer formula. Obviously, this conduction may be extremely
low, indicating in practice an insulating
behavior. Of particular interest, is to estimate the electron transfer rate in a given
DBA system that will translate into a
measurable conduction of the same system when used as a molecular conductor
between two metal leads. To this end,
consider a DBA system with a bridge
that consists of N identical segments (denoted 1,2, . . . , N ) with nearest neighbor
coupling VB . The electron transfer rate
is given by Eq. (11) that we rewrite in
the form
2
kDA =
|VD1 VN A |2 |G1N (ED )|2 F
h
(49)
where, in the weak coupling limit, |VB |
|EB E| (cf. Eq. 10)
G1N (E) =
|N 1
|VB
(EB E)N
(50)
and where F is the FranckCondonweighted density of nuclear states, given

in the classical limit by Eq. (3). The appearance of F in Eq. (49) indicates that
the process is dominated by the change

in the nuclear conguration between the
two localization states of the electron. Suppose now that the same DBA complex
is used to connect between two metal
contacts such that the donor and acceptor species are chemisorbed on the
two metals (denoted left and right,
respectively). We wish to calculate the conduction of this junction and its relation
to kDA . First note that the conduction
process does not involve localized states
of the electron on the donor or the acceptor, so the factor F will be absent.
(We will disregard energy loss arising
from transient distortions of the nuclear
conguration associated with transient
populations of electronic states of the
DBA complex). Assuming as before that
states of the molecular complex are coupled to the metal only via the D and A
orbitals, and that the latter are coupled
only to their adjacent metal contacts, the
conduction is given by an equation similar to Eq. (40), except that the bridge
(1, . . . , N ) is replaced by the complex
DBA = (D, 1, . . . , N , A)
e2
(L)
(R)
|GDA (E)|2 #D (E)#A (E)
h
(51)
where, in analogy to Eq. (48)
g(E) =
GDA (E)
=
VD1 VN A
(E ED ;D (E))(E EA ;A (E))
G1N (E)
(52)
Since the donor and acceptor species are

chemisorbed on their corresponding metal
contacts, their energies shift closer to the
Fermi energies. We assume that this shift
occurs uniformly in the DBA complex
without distorting its internal electronic
structure (strictly speaking this can happen
only in the symmetric case of identical
donor and acceptors and identical metal

electrodes, but the result of Eq. (53) below
is probably a good approximation for more
general cases because G1N (E) is often
not strongly dependent on E). Assuming
therefore that the denominator in Eq. (52)
is dominated by the imaginary parts of the
self-energies ;, we get
g = g(EF ) =
16e2
|VD1 VN A |2
h # (L) (EF )# (R) (EF )
D
|G1N (EF )| ;
EF = ED = EA (53)
Comparing to Eq. (49) we get

g=
e2 kDA
8h
h F # (L) # (R)
D
(54)
It has been argued that, Eq. (52) and

therefore Eq. (54) holds the energy spacing EB EF between the bridge levels
and the Fermi energy is large relative
to kB T , Eq. (54) holds also when the
electron transfer process involves thermal activation into the bridge states (and
not only for the bridge assisted tunneling implied by Eq. 53) [118]. Using
the classical expression for F, Eq. (3),
we
the present case F =
have in
1
( 4kB T ) exp(/4kB T ). For a typical value of the reorganization energy
0.5 eV, and at room temperature, this
(L)
(R)
is 0.02 (eV)1 . Taking also #D = #A
2
13
0.5 eV leads to g (e / h)(10 kDA
(s 1 ))
= [1017 kDA (s 1 )] 81 . This sets
a criterion for observing Ohmic behavior for small voltage bias in molecular
junctions: With a current detector sensitive to pico-amperes, kDA has to exceed
106 s1 (for the estimates of F and #
given above) before measurable current
can be observed at 0.1 V voltage across
such a junction.
1.3.7
Quantum Chemical Calculations
The simple models discussed above

are useful for qualitative understanding
of molecular conductivity; however, the
Landauer formula or equivalent formulations can be used as a basis for
more rigorous molecular calculations
using extended Huckel (EH) based
computations [7072, 79, 119133] or
HartreeFock (HF) [134137]. These approaches follow similar semiempirical and
ab initio calculations of electron transfer
rates in molecular systems [138]; however, instead of focusing on the computation of the electronic coupling VDA
needed in Eq. (1), the sum in Eq. (34)
is calculated directly. Structural stability
considerations suggest that useful MMM
bridges should involve strong chemisorption bonding between the molecule and
the metal substrate, implying large electronic coupling between them [139]. It
is, therefore, preferable to use a supermolecule approach, in which the
quantum chemical calculations are carried out for a species that comprises
the molecule and two clusters of metal
atoms, so that the reduction that introduces the self energy ; Eq. (43) is done
at some deeper metalmetal contact. Such
atomic level calculations usually start from
a (nonorthogonal) basis set of atomic orbitals, so the formalism described above
has to be generalized for this situation.
(Alternatively, it has been shown by Emberly and Kirczenow [140, 141] that one
can map the problem into a new Hilbert
space in which the basis states are orthogonal.) We also relax the assumption that
the moleculemetal contact is represented
by coupling to a single molecular orbital.
Dening the operator
H(E) = EZ H
with Zij = i|j (55)
63
64
1 Fundamentals
The Greens function is G(E) = H(E)1 .

In Eq. (55), i and j denote atomic orbitals
that may be assigned to the supermolecule
(M), the left metal (L) and the right metal
(R) subspaces. Denoting formally the
coupling between the subspace M and the
subspaces K = L, R by the corresponding
submatrices HMK , the Greens function for
the supermolecule subspace is
G(M) (E) = (H ; (L) ; (R) )1
(56)
with ; (K) is a matrix in the molecular

subspace and
; (K) = HMK (H1 )KK HKM
(57)
Eq. (57) is a compact notation for

(; (K) )n,n = k,k Hnk (H1 )kk Hk n where
k and k are states in the metal K subspace.
Using also

Hln Gnn Hn r
(58)
Tlr =
n,n
(l and r in the metal L and R subspaces,

respectively; n, n in the supermolecule
subspace) in Eq. (57) leads to
g(E) =

e2
T r G(M) (E)# (R) (E)
h

(59)
G(M) (E)# (L) (E)
where, for example, for the left metal

(L)
Hnl Hln (E El )
#n,n = 2
l
(n and n in the molecular subspace)

(60)
In practice, ; and # = 2Im(;) can
be computed by using closure relations
based on the symmetry of the metal
lattice [130]. The trace in Eq. (59) is over
all basis states in the (super)molecular
subspace. The evaluation of the Greens
function matrix elements and of this trace
is straightforward in semiempirical single
electron representations such as the EH

approximation, and can be similarly done
at the HF level using, after convergence,
the Fock rather then the Hamiltonian
matrix in Expressions (55 to 60). (Note that
the Fock operator depends on the ground
state electronic conguration. The latter
is taken in Refs. [134, 135] to be that of
the isolated supermolecule, assuming that
the contact with the bulk electrodes does
not affect it appreciably. In particular, the
supermolecule is usually assumed neutral
in these calculations.)
An important attribute of the approach
described above is that, within the approximation used, it provides the total current
carried by the system, both through the
unoccupied molecular levels (electron conduction) and the occupied ones (hole
conduction). This results from the fact that
the trace in Eq. (59) is over all the atomic orbitals that comprise the (super)molecular
basis set, that upon diagonalization in the
(super)molecular Hamiltonian will yield
both occupied and unoccupied molecular orbitals. In a one-electron theory such
as the EH approximation, both types of
orbitals contribute in the same way. For
example, the terms in Eq. (59) that describe an electron moving from the highest
occupied molecular orbital (HOMO) into
empty states of the anode, followed by an
electron moving from the cathode into the
HOMO (hole transport), and an electron
moving from the cathode to the lowest
unoccupied molecular orbital (LUMO),
then moving on into the cathode (electron transport), are similar (their values
depend on the energies of the molecular
orbitals involved with respect to the Fermi
energies), irrespective of whether the corresponding orbitals are occupied or not.
The same is true in the HF calculation
if the Koopmans theorem [142], stating
that the HF orbital energies represent
the actual energies involved in removing

an electron from an occupied orbital or
putting an electron into an unoccupied
one, holds. The Koopmans theorem is
accurate only for large systems, and the
approximation involved in applying it to
small systems is one reason why HF is not
necessarily superior to EH for calculating
the conduction properties of small molecular junctions. This is true particularly
for LUMO dominated conduction, because
the HF is notoriously inadequate for electron afnities [143146]. See Ref. [137] for
further discussion of this point. Another
potential (but in principle avoidable) problem in these calculations is associated
with the nite, relatively small basis of
atomic orbitals used. Close to resonance,
when the electrode electrochemical potential = EF + e approaches the HOMO
or LUMO energies, the corresponding
HOMO or LUMO orbitals dominate the
electron transfer and a small basis that
correctly describes these orbitals is sufcient. When EF is a distance %E away
from EHOMO or ELUMO , all molecular
orbitals in the range %E below EHOMO
and in a similar range above ELUMO can
contribute to the transmission probability and cannot be ignored, implying the
need for a larger molecular basis [110].
We note in passing that the recently discussed transmission antiresonances [140,
141] associated with the nonorthogonality
of the atomic orbital basis sets, have been
shown [110] to be sometimes artifacts of a
small basis calculation.
In spite of these limitations, EH- and
HF-based calculations have provided important insight into the conduction properties of molecular junctions. In view of other
unknowns, associated both with the uncertainty about the junction structure and
with the simplied computation, the main
value of the results obtained from such
calculations should be placed not in the

absolute numbers obtained, but rather in
highlighting the importance of the model
parameters in determining the junction
conduction behavior. Examples for qualitative issues that were investigated with
these types of calculations include the effect of the nature (length and conjugation)
of the molecular bridge [126, 127], the effect of the moleculeelectrode binding and
of the molecular binding site [130], the relation of conductance spectra to molecular
electronic structure [79] and the effect of
bonding molecular wires in parallel [128].
(See also Ref. [147])
1.3.8
Spatial-grid Based Pseudopotential

Approaches
Another way to evaluate the expressions

appearing in Eqs. (34 and 37) as well as
related partial sums is closely related to the
discrete variable representation of reaction
probabilities as formulated by Seideman
and Miller [148150]. We have already
seen that the sum

s(E)
|Tlr |2 (E El )(E Er )
l,r
(61)
which is related to the conduction by
g(E) = (4e2 /h)s(E) (c.f Eq. 37) can be
represented by (c.f. Eq. 59)
s(E) =

1
T r G(M) (E)# (R) (E)
4 2

(62)
G(M) (E)# (L) (E)
If, instead of considering transitions from

left to right electrode, we think
of Eq. (61) as expressing a sum over
transition probabilities from all initial
(i) states of energy E in the reactant space to all nal (f ) states of the
same energy in the product space, s(E)
65
66
1 Fundamentals
is also associated with the so called

cumulative reaction probability [148150],
which in terms of the reaction
S matrix

is dened by N (E) = i,f |Sif (E)|2 =

4 2 s(E), that is, N (E) = i,f Tif (E).
Equation (62) now expresses the important
observation that the cumulative reaction
probability for a reactive scattering process
can be expressed as a trace over states,
dened in a nite subspace that contains the
interaction region, of an expression that depends on the reduced Greens function
and the associated self energy dened in
that subspace. Following Seideman and
Miller we can use a spatial grid representation for the states in this subspace, so that
the trace in Eq. (62) becomes a sum over
grid points. Also, in this representation the
overlap matrix Z is zero. In general, any
subspace of position space that separate
reactants from products (i.e. that encompasses the entire interaction region; the
molecular bridge in our application) can
be used in Eq. (62), provided that the consequences of truncating the rest of the
universe, expressed by the corresponding
; and # can be computed. The absorbing boundary condition Greens function
(ABCGF) method of Seideman and Miller
is based on the recognition that if this
subspace is taken to be large enough so
that its boundaries are far from the interaction region, the detailed forms of ; and
# are not important; the only requirement
is that scattered waves that approach these
boundaries will be absorbed and not reected back into the interaction zone. In
the ABCGF method, this is accomplished
by taking ; = (1/2)i# = i(r), a local
function in position space, taken to be zero
in the interaction region and gradually increasing from zero when approaching the
subspace boundaries. Its particular form
is chosen to effect complete absorption
of waves approaching the boundary to
a good numerical accuracy. Equation (62)

then becomes

s(E) = 4T r GABC (E)R GABC (E)L
(63)
where GABC (E) = (E H + i)1 ; =
R + L and where R and L are different
from zero only on grid points near the right
side (more generally the product side),
and the left (reactant) side of the inner
subspace, respectively.
A similar development can be done for
the partial sum

sl (E)
|Tlr |2 (E Er )
(64)
r
which, provided that l is taken as an

eigenstate of the Hamiltonian describing the left electrode (or the reactant
subspace), is related the one to all
rate, kl (E) to go from an initial state
of energy E on the left electrode (or in
reactant space) to all possible states on
the right one (product space) according to
kl = (2/h)sl . (The microcanonical
rate

is dened by k(E) = L 1 (E) l kl (E
El ) = (2 hL (E))1 4 2 s(E) = ( hL
(E))1 N (E).) We use the same denition
of the coupling V between our subspace
(bridge) and the reactant and product (electrode) states. Putting T = V GV in Eq. (64)
we get
sl (E) =
1
l|V G(M) # (R) G(M) V |l (65)
2
Using again a position grid representation

of the intermediate states used to evaluate
this expression, and applying the same
methodology as above, Eq. (65) can be
recast in the form
1
sl (E) = l|V GABC (E)R GABC (E)V |l
1
= l|L GABC (E)R GABC (E)L |l
(66)
The second part of Eq. (66) is obtained by using the identity R |l =
0 to write R G V |l = R (1 + G V )|l =
R G (G1 + V )|l, which, together with
G1 = E H0 V + i, (E H0 )|l = 0
and |l = l |l, yields the desired result.
The results (63 and 66) are very useful
for computations of transmission probabilities in models in which the interaction
between the transmitted particle and the
molecular spacer is given as a position dependent pseudopotential. Applications to
electron transmission through water and
other molecular layers are discussed in
Refs. [151158].
1.3.9
Density Functional Calculations
Density functional methods provide a convenient framework for treating metallic interfaces [100]. Applications of this methodology to the problem of electron transport
through atomic and molecular bridges
have been advanced by several workers. In
particular, Langs approach [90, 159165]
is based on the density functional formalism [166, 167] in which the single electron
wave functions 0 (r) and the electron density n0 (r) for two bare metal (jellium)
electrodes is computed, then used in the
LippmanSchwinger equation

(r) = 0 (r) + dr dr G0 (r, r )V
(r , r )(r )
(67)
to get the full single electron scattering

wave functions (r) in the presence of
the additional bridge. (Langs earlier calculations [90, 168] use a related density
functional approach to calculate the tunneling current between an atomic tip and
a jellium electrode without an atomic or
molecular bridge). In Eq. (67), G0 is the
Greens function of the bare electrode system and V is the difference between
the potential of the full system containing an atomic or a molecular spacer and
that of the bare electrodes. In atomic units
(|e|, , m = 1), it is

V (r, r ) = Vps (r, r ) + Vxc (n(r)) Vxc

n(r )
n0 (r) + dr
|r r |
(68)
where Vps is the sum of nonlocal pseudopotentials representing the cores of

the spacer atoms and Vxc is the local density approximation (LDA) of the
exchange-correlation potential. n is the
electron number density for the full system
(electrodes and atoms) and n = n n0 .
Equation (67) yields scattering states that
can be labeled by their energy E, momentum k in the direction (yz) parallel
to the electrodes and spin. In addition,
Lang distinguishes between wave functions that in the electrode regions carry
positive (+) or negative () momentum
in the tunneling direction. Denoting by
L and R the electron electrochemical
potential in the left and right electrode,
respectively, the zero-temperature electrical current density from left to right (for
L > R ) is then
R
J (r) = 2
dE d2 K I m{+
+ }
L
(69)
The factor 2 accounts for the double occupancy of each orbital. This approach
was used recently [169] to calculate current through a molecular species, Benzene
1,4-dithiolate molecule (as used in the experiment described in Ref. [67]), between
two jellium surfaces. The result demonstrates the large sensitivity of the computed
current to the microscopic structure of the
moleculemetal contacts.
67
68
1 Fundamentals
A similar density functional approach,

using an atomic level description of the
electrodes, was described by Di ventra and Pantelides [170]. These authors
used density-functional based ground state
molecular dynamics [171] in order to get
the relaxed structure of the metal-atomic
system-metal junction; then evaluated the
current through the relaxed structure.
The density-functional based calculations described above were done for small
applied potential bias between the electrodes. In contrast, the density functional
approach of Hirose and Tsukada [172]
calculates the electronic structure of a
metal-insulator-metal system under strong
applied bias. The main difference from the
density functional approaches described
above comes in the way the effective
one-electron potential is calculated. The
potential used in this work contains the
usual contributions from the Coulomb
and the exchange-correlation interactions
as well as from the ionic cores. However, the Coulomb (Hartree) contribution
is obtained from the solution of a Poisson equation
2 VH (r) = 4[(r) + (r)]
(70)
in the presence of the applied potential

boundary conditions. + (r) is the xed
positive charge density, and the electron
density (r) is constructed by summing
the squares of the wave functions over
the occupied states. At the same time,
the exchange-correlation potential is calculated in the standard LDA, neglecting
the effect of the nite current that exists in
the steady state system. The resulting formalism thus accounts approximately for
nonequilibrium effects within the density
functional calculation. (A simplied version of the same methodology has recently
been presented by Mujica et al. [173].)
To end this brief overview of densityfunctional-based computations of molecular conduction, we should note that this
approach suffers in principle from problems similar to those encountered in using
the HF approximation, that is, the inherent inaccuracy of the computed LUMO
energy and wave functions. The errors
are different, for example HF overestimates the HOMOLUMO gap (since the
HF LUMO energy is too high [143146,
174176]) while density functional theory
(DFT) underestimates it [167, 177]. Common to both approaches is the observation
that processes dominated by the HOMO
level will be described considerably better
by these approaches than by processes controlled by coupling to the LUMO [137, 178].
1.3.10
Potential Proles
The theoretical and computational approaches described above are used to

compute both the Ohm-law conduction,
g(EF ) of a molecular bridge connecting two metals, (Eqs. 37 or 59), and
the currentvoltage characteristics of the
junction, also beyond the Ohmic regime
(Eq. 36). We should keep in mind that
these calculations usually disregard a potentially important factor the possible
effect of the imposed electrostatic eld on
the nuclear structure as well as on the electronic structure of the bridge. A change
in nuclear conguration under the imposed electrostatic eld is, in fact, not very
likely for stable chemisorbed molecular
bridges. On the other hand, the electronic
wave functions can be distorted by the
imposed eld, and this in turn may affect the electrostatic potential distribution
along the bridge, the electronic coupling
between bridge segments and the position
of the molecular energy levels vis-a-vis
the metals Fermi energies. (In a single

electron description, the local electrostatic
potential will be an input, associated with
the underlying many-electron response of
the molecular bridge, to the position dependent energies of the bridge electronic
states in the site representation.) These
effects were, in fact, taken approximately
into account by Hirose and Tsukada [172]
and by Mujica, Roitberg and Ratner [173]
by solving simultaneously, the coupled
Schrodinger and Poisson equations. The
latter yields the electrostatic potential for
the given electron density and under the
imposed potential boundary conditions.
The importance of the electrostatic potential prole on the molecular bridge
in determining the conduction properties of a metal-molecule-metal junction
was recently discussed by Tian et al [72]
in conjunction with the currentvoltage
characteristics of a junction comprised of
an STM tip, a gold substrate and a molecule
with two bonding sites (e.g. , -xylyl
dithiol) connecting the two.
1.3.11
Rectication
The possibility of constructing molecular

junctions with rectifying behavior has been
under discussion ever since Aviram and

Ratner [179] suggested that an asymmetric
donor-bridge-acceptor system connecting
two metal leads could rectify current. The
proposed mechanism of operation of such
a device is shown in Fig. 5. When the left
electrode is negatively biased, that is, the
corresponding electrochemical potentials
satisfy L > R as shown, electrons can
move from this electrode to the LUMO of
molecular segment A, as well as from the
HOMO of molecular segment D, to the
right electrode. Completion of the transfer
by moving an electron from A to D is assisted by the intermediate bridge segment
B. When the polarity of the bias is reversed,
the same channel is blocked. This simple
analysis is valid only if the molecular energy levels do not move together with the
metal electrochemical potentials, and if the
coupling through the intermediate bridge
is weak enough so that the orbitals on the D
and A species maintain their local nature.
Other models for rectication in molecular junctions have been proposed [180].
As discussed above, the expected rectifying behavior can be very sensitive to
the actual potential prole in the ABD
complex, which in turn depends on the
molecular response to the applied bias [72,
181]. This explains why rectication is
Fig. 5 A model for current rectication

in a molecular junction: Shown are the
mL
chemical potentials L and R in the
two electrodes, and the HOMO and
LUMO levels of the donor, acceptor,
and bridge. When the right electrode is
positively biased, (as shown) electrons
can hop from left to right as indicated by
the dotted arrows. If the opposite bias
can be set without affecting the
electronic structure of the DBA system
too much, the reverse current will
be blocked.
D
mR
A
B
69
70
1 Fundamentals
often not observed even in asymmetric

molecular junctions [181]. However, rectication has been observed in a number
of metal-molecule-metal junctions, as well
as in several STM experiments involving
adsorbed molecules [61, 65, 182185].
1.3.12
Carriercarrier Interactions
The models and calculations discussed

so far focus on processes for which the
probability that a charge carrier populates
the bridge is low so that carriercarrier
interactions can be disregarded. Electronelectron interactions were taken into
account only in so far that they affected
the single electron states, either in constructing the molecular spectrum (in the
ab initio HF or DFT calculations) or in
affecting the junction electrostatic potential through the electronic polarization
response of the molecule or the metal contacts. When the density of carriers in the
space between the metal contacts becomes
large, Coulomb interactions between them
have to be taken into account explicitly.
Here we briey discuss the effect of such
interactions.
In classical (hopping) transport of carriers through insulating lms separating
two metals, intercarrier interactions appear as suppression of current owing to

lm charging [186, 187]. In nano-junctions
involving double barrier structures, increased electron population in the intermediate well under resonance transmission
should affect the transport process for similar reasons. For example, consider a small
metal sphere of radius R in the space between two metal electrodes (Fig. 6), and
assume that both sphere and electrodes
are made of the same metal of work
function W. Neglecting possible proximity effect of these electrodes, the classical
energy for removing an electron from
the sphere to innity is W + e2 /2R and
the classical energy for the opposite process is W e2 /2R. (From experimental
and theoretical work on ionization potentials of small metal clusters [188] we know
that the actual energies are approximately
W + 0.4e2 /R and W 0.6e2 /2R, respectively; with the differences arising from
quantum size effects.) Here, the sphere
plays the role of a molecular bridge in
assisting electron tunneling between the
two electrodes, and these energies now
play the same role as the corresponding
HOMO and LUMO energies of the bridge.
This implies that a nite voltage difference is needed before current can ow in
this sphere-assisted mode between the two
EF + e2/(2R)
EF
EF
EF e2/(2R)
A nano-dot between two

conducting leads: A model for
Coulomb blockade phenomena.
Fig. 6
metals, a phenomenon known as Coulomb

blockade. For a larger potential bias, other
conduction channels, corresponding to
more highly charged states of the sphere
give rise to the phenomenon of Coulomb
steps [189]. For experimental manifestations of such and related phenomena,
see, for example, Refs. [190196]. The possibility to observe such phenomena in
electrochemical systems was discussed by
Kuznetsov and Ulstrup [197] and possibly
demonstrated by Fan and Bard [198].
When the junction consists of a
molecule or a few molecules connecting
two metal leads, such Coulomb blockade
phenomena are not expected to appear
so clearly. The rst Coulomb threshold is
replaced, as just described, by the gap associated with the position of the metals
Fermi energies relative to the molecular
HOMO and LUMO levels (modied by
appropriate electron correlations). However, the discreteness (in the sense that
%E kB T ) of the molecular spectrum
implies that for any given charging state of
the molecule, for example, a molecule with
one excess electron or one excess hole,
there will be several distinct conduction
channels that will appear as steps in the
current versus voltage plot. It will be hard
to distinguish between this structure and
between a genuine Coulomb blockade
structure. It should be emphasized that
for potential applications, such as using
the molecular junction in single electron
transistor devices, the distinction between
the origins of these conduction structures
is in principle, not important.
However, understanding the role played
by electronelectron interactions (in particular, correlation effects beyond the
HF approximation) remains an important
challenge in the study of molecular nanojunctions. Several recent theoretical works
have addressed this problem within the
Hubbard model with [199201]or without [202] the mean eld approximation.
In particular, Malysheva and Onipko have
derived a tight binding analog of the
model for negative differential resistance
originally proposed by Davydov and Ermakov [203] (see also Refs. [204206]).
Numerical simulations [207] can assist in
gauging the performance of the mean eld
approximations used in these calculations.
Such models may be relevant to the understanding of recent experimental observation of negative differential resistance in
a metal-self-assembled monolayermetal
junction with the SAM containing a nitroamine redox center [208].
We conclude this discussion by emphasizing again that understanding correlated
carrier transport in molecular junctions
continues to be an important experimental and theoretical challenge. Recent work
by Gurvitz et al [209211], using exactly
solvable models of electron transport in
two and three barrier structures, has indicated that new phenomenology may arise
from the interplay of inelastic transitions
and intercarrier interactions in the barrier.
In fact, dephasing transitions in the barrier
may prove instrumental in explaining the
charge quantization that gives rise to the
single electron transport behavior of such
junctions (Sect. 6.3 in Ref. [212]).
1.3.13
Some Open Issues
This section discusses some subtle difculties that are glossed over in most of
the treatments of electron transmission using the formalisms described above. These
should be regarded as open theoretical issues that should be addressed in future developments. The source of these problems
is our simplied treatment of what is actually a complex many-body open system.
71
72
1 Fundamentals
In particular, common ways of incorporating many-body-effect using single-body

effective potentials become questionable
in particular limits of timescales and interaction strengths.
One such issue, already mentioned, is
the use of static image to account for
the effect of metal polarizability (that
is, the response of the metal electrons)
on charge transfer processes at metal
surfaces. Timescale estimates for electron
tunneling in molecular systems are of the
same order as metal plasma frequencies
that measure the electronic response
time of metals. However, static image
theory has been used in the analysis
of Sect. 1.3.3, and in other treatments
of electron injection from metals into
insulating phases [213]. To what extent
dynamic image effects are important is not
known, though theories that incorporate
such effects have been developed [97100].
Assuming that image interactions at
metal surfaces should be accounted for
in the static limit, namely that the metal
responds instantaneously to the tunneling charge, opens other questions. Many
calculations of electronic processes near
metal surfaces [e.g. [89] (See Sect. 1.3.3
above)] assume that the metal electrons respond instantaneously to the position of the
tunneling electron. Other calculations use
atomic or molecular orbitals, or more general electronic charge distributions, and
computing these under the given potential
boundary conditions (see, e.g. Ref. [178])
implies that the corresponding orbitals or
charge distributions are well dened on
timescales shorter than the metal response
times. (Computing molecular orbitals selfconsistently with image interactions is the
common practice in quantum chemistry
calculations for solvated molecules using
reaction eld (cavity) models. Again we
have a choice: either imposing the reaction eld on the electronic Hamiltonian
in the position representation, thus modifying all Coulomb interaction terms, then
calculate the electronic wave functions under the new potential, or compute the
electronic wave functions with the original
Hamiltonian under the imposed dielectric
boundary conditions. The fact that the two
representations are not equivalent is associated with the approximate nature of
the approach, which replaces a detailed
treatment of the electronic structure of the
solvent by its electronic dielectric response
and with the fact that the Schrodinger
equations derived from them are nonlinear in the electronic wave functions).
Examination of the energies and timescales involved suggests that assuming
instantaneous metal response to the electron position is more suitable in most
situations than taking instantaneous response to the charge distribution dened
by a molecular orbital, but the corresponding timescales are not different enough to
make this a denite statement.
A similar issue appears in attempts to
account for the electronic polarizability
of a solvent in treating fast electronic
processes involving solute molecules or
excess electrons in this solvent. For
example, in treating electron transmission
in MIM junctions, the potential barrier that
enters into expressions such as Eq. (14)
depends on the electronic structure of the
insulating spacer. For vacuum tunneling,
a rectangular barrier, whose height above
the metal Fermi energy is the metal work
function, modied by image interactions
as discussed above and in Sect. 1.3.3,
seems appropriate. For a dielectric spacer,
the barrier should be further modied
by the fast (electronic) dielectric response
of this spacer in the same way that it
is modied by the electronic response of
the metal, raising issues similar to those

discussed above.
An interesting point of concern is related
to the way the Fermi distribution functions
enter into the current equations. For example, Bardeens transmission formula (21)
is based on weak coupling between states
localized on the two electrodes, the partial or unidirectional currents contain a
product, f (1 f ),that is, the probability
that the initial state is occupied multiplied by the probability that the nal state
is not. In this viewpoint, the transitions
occur between two weakly coupled systems, each of them in internal thermal
equilibrium, which are out of equilibrium
with each other because of the potential
bias. Alternatively, we could work in the
basis of exact eigenstates of the whole system comprising the two electrodes and
the spacer between them. This system
is in an internal nonequilibrium state in
which transmission can be described as
a scattering problem. The relevant eigenstates correspond to incident (incoming)
waves in one electrode and transmitted
waves in the other. The ux associated
with those scattering states arising from
an incident state in the negatively biased
electrode is proportional to f (E), while
that associated with incoming waves in the
positively biased electrode is proportional
to f (E + e). The net ux is therefore,
found again to be proportional to the difference f (E) f (E + e). This argument
cannot be made unless the process can
be described in terms of coherent scattering states dened over the whole system.
When inelastic scattering and dephasing
processes take place, the description in
terms of exact scattering states of the
whole system becomes complicated [212,
214], although kinetic equations for electron transport can be derived for relatively
simple situations [209211]. On the other
hand, it appears that for weakly coupled

contacts the perturbative approach that
leads to Eq. (21) is valid. This approach
describes the transmission in terms of
electron states localized on the two electrodes where unidirectional rates appear
with f (1 f ) factors, and can in principle
be carried over to the inelastic regime. The
exact correspondence between these different representations needs further study.
Finally, the theoretical and computational issues discussed in this chapter are
based on the assumption that the electron
tunnel factors do not depend on the interaction between the molecular system and
its thermal environment. This assumption
covers one important scenario of electron
transmission. On the other extreme are
systems in which electron transfer is thermally activated and long range electron
transfer is dominated by hopping. An important example is provided by the recent
results on long range electron transport
in DNA [215219]. The transition from
tunneling to hopping and other issues
involving thermal effects in long-range
electron transfer have recently been discussed by several workers [220].
Acknowledgments
This work was supported by the the

Israel Science Foundation, the U.S. Israel
Binational Science foundation and by the
Israel Ministry of Science.
References
1. J. Jortner, M. Bixon, (Eds.), Advances in
Chemical Physics: Electron Transfer from
Isolated Molecules to Biomolecules, John
Wiley & Sons, New York, 1999, Vol. 106.
2. A. M. Kuznetsov, Charge Transfer in Physics,
Chemistry and Biology, Gordon & Breach,
New York, 1995.
73
74
1 Fundamentals
3. A. M. Kuznetsov, J. Ulstrup, A. M. Kuzne,
T. Sov, Electron Transfer in Chemistry and
Biology: An Introduction to the Theory, John
Wiley & Sons, New York, 1998.
4. D. R. Lamb, Electrical Conduction Mechanisms in Thin Insulating Films, Methuen,
London, 1967.
5. R. J. Miller, G. McLendon, A. Nozik et al.,
Surface Electron Transfer Processes, VCH
Publishers, New York, 1995.
6. F. Schmickler, Interfacial Electrochemistry,
Oxford University Press, Oxford, 1996.
7. R. M. Metzger in Molecular Electronics Science and Technology (Ed.: A. Aviram), American Institute of Physics Conference Proceedings, New York, 1992, Vol. 262, p. 85.
8. C. A. Mirkin, M. A. Ratner, Annu. Rev.
Phys. Chem. 1992, 43, 719754.
9. K. Sienicki, (Ed.), Molecular Electronics and
Molecular Electronic Devices, CRC Press,
New York, 1994.
10. M. C. Petty, M. R. Bryce, D. Bloor, (Eds.),
An Introduction to Molecular Electronics,
11. C. Joachim, S. Roth, (Eds.), Atomic and
Molecular Wires, Kluwer, Dordrecht, The
Netherlands, 1997, Vol. 341.
12. J. Jortner, M. Ratner, (Eds.), Molecular Electronics, Blackwell Science, Oxford, 1997.
13. L. Kouwenhoven, Science 1997, 275, 1896,
1897.
14. A. Aviram, M. Ratner, (Eds.), Molecular
Electronics: Science and Technology, New
York Academy of Sciences, New York, 1998.
15. C. Dekker, Phys. Today 1999, 52, 2228.
16. M. A. Reed, Proc. IEEE 1999, 87, 652658.
17. A. G. Davies, Philos. Trans. R. Soc. London
Ser. A Phys. Sci. Eng. 2000, 358, 151172.
18. R. Landauer, IEEE Trans. Electron Devices
1996, 43, 16371639.
20. S. Gosavi, R. A. Marcus, J. Phys. Chem.
2000, 104, 20672072.
21. M. R. Wasielewski, M. P. Niemczyk, D. G.
Johnson et al., Tetrahedron 1989, 45, 4785.
22. S. B. Sachs, S. P. Dudek, L. R. Sita et al., J.
Am. Chem. Soc. 1997, 119, 10 563.
23. V. Grosshenny, A. Harriman, R. Ziessel,
Angew. Chem., Int. Ed. Engl. 1996, 34, 2705.
24. J. M. Tour, Chem. Rev. 1996, 96, 537.
25. A. Osuka, N. Tanade, S. Kawabata et al., J.
Org. Chem. 1995, 60, 7177.
26. A. C. Ribou, J. P. Launay, K. Takahashi
et al., Inorg. Chem 1994, 33, 1325.
27. S. Woitellier, J. P. Launay, C. W. Spangler,

Inorg. Chem. 1989, 28, 758.
28. A. Helms, D. Heiler, G. Mclendon, J. Am.
Chem. Soc. 1992, 114, 6227.
29. L. M. Tolbert, Acc. Chem. Res. 1992, 25, 561.
30. T. S. Arrhenius, M. Balnchard-Desce,
M. Dvolaitzky et al., Proc. Natl. Acad. Sci.
1986, 83, 5355.
31. P. Finkh, H. Heitele, M. Volk et al., J. Phys.
Chem. 1988, 92, 65846590.
32. R. F. Ziessel, Chem. Educ. 1997, 74,
673679.
33. M. J. Bjerrrum, D. R. Casimiro, I. J. Chang
et al., J. Bioenerg. Biomembr. 1995, 27, 295.
34. R. Langen, I. J. Chang, J. P. Germanas et al.,
Science 1995, 268, 17331735.
35. J. R. Winkler, H. B. Gray, Chem. Rev. 1992,
92, 369.
36. M. Y. Ogawa, J. F. Wishart, Z. Young et al.,
J. Phys. Chem. 1993, 97, 11 456.
37. S. S. Isied, M. Y. Ogawa, J. F. Wishart,
Chem. Rev. 1992, 92, 381.
38. K. S. Schanze, L. A. Cabana, J. Phys. Chem.
1990, 94, 2740.
39. J. M. Vanderkooi, S. W. Englander, S. Papp
et al., Proc. Natl. Acad. Sci. 1990, 87, 5099.
40. C. E. Moser, C. C. Page, R. Farid et al., J.
Bioenerg. Biomembr. 1995, 27, 263274.
41. M. W. Mutz, M. A. Case, J. F. Wishart et al.,
J. Am. Chem. Soc. 1999, 121, 858, 859.
42. F. D. Lewis, T. F. Wu, Y. F. Zhang et al.,
Science 1997, 277, 673676.
43. R. E. Holmlin, P. J. Dandliker, J. K. Barton,
Angew. Chem., Int. Ed. Engl. 1997, 36, 2714.
44. P. Lincoln, E. Tuite, B. Norden, J. Am.
Chem. Soc. 1997, 119, 1454, 1455.
45. A. M. Brun, A. Harriman, J. Am. Chem. Soc.
1994, 116, 10 383.
46. T. J. Meade, J. F. Kayyem, Angew. Chem.,
Int. Ed. Engl. 1995, 34, 352354.
47. P. F. Barbara, E. J. C. Olson, Experimental
electron transfer kinetics in a DNA environment in Advances in Chemical Physics:
Electron Transfer from Isolated Molecules
to Biomolecules, Pt 2 (Eds.: M. Bixon,
J. Jortner), Horizon Pubs & Distributors
Inc., 1999, pp. 647676, Vol. 107.
48. K. Fukui, K. Tanaka, Angew. Chem., Int. Ed.
Engl. 1998, 37, 158161.
49. E. Meggers, M. E. Michel-Beyerle, B. Giese,
J. Am. Chem. Soc. 1998, 120, 12 95012 955.
50. E. Meggers, D. Kusch, M. Spichty et al.,
Angew. Chem., Int. Ed. Engl. 1998, 37,
460462.
51. K. P. Ghiggino, A. H. A. Clayton, J. M.
Lawson et al., New J. Chem. 1996, 20, 853.
52. H. Oevering, M. N. Paddon-Row, M. Heppener et al., J. Am. Chem. Soc. 1987, 109,
3258.
53. M. T. Carter, G. K. Rowe, J. N. Richardson
et al., J. Am. Chem. Soc. 1995, 117, 2896.
54. H. L. Guo, J. S. Facci, G. Mclendon, J. Phys.
Chem. 1995, 99, 8458.
55. B. Paulson, K. Pramod, P. Eaton et al., Phys.
Chem. 1993, 97, 13 042.
56. M. D. Johnson, J. R. Miller, N. S. Green
et al., J. Phys. Chem. 1989, 93, 1173.
57. C. A. Stein, N. A. Lewis, G. Seitz, J. Am.
Chem. Soc. 1982, 104, 2596.
58. V. J. Langlais, R. R. Schlittler, H. Tang et al.,
Phys. Rev. Lett. 1999, 83, 28092812.
59. E. E. Polymerpopoulos, D. Mobius,
H. Kuhn, Thin Solid Films 1980, 68, 173.
60. M. Fujihira, K. Nishiyama, H. Yamada, Thin
Solid Films 1985, 132, 77.
61. S. Roth, M. Burghard, C. M. Fischer, Resonant tunneling and molecular rectication in Langmuir-Blodgett lms in
Molecular Electronics (Eds.: J. Jortner,
M. Ratner), Blackwell Science, Oxford,
1997, pp. 255280.
62. E. Delamarche, B. Michel, H. A. Biebuyck
et al., Adv. Mater. 1996, 8, 718.
63. L. A. Bumm, J. J. Arnold, M. T. Cygan et al.,
Science 1996, 271, 17051707.
64. U. Durig, O. Zuger, B. Michel et al., Phys.
Rev. B 1993, 48, 1711.
65. A. Dhirani, P. H. Lin, P. Guyot-Sionnest, J.
Chem. Phys. 1997, 106, 52495253.
66. L. Ottaviano, S. Santucci, S. D. Nardo et al.,
J. Vac. Sci. Technol., A 1997, 15, 1014.
67. M. A. Reed, C. Zhou, C. J. Muller et al.,
Science 1997, 278, 252254.
68. D. Porath, O. Millo, J. Appl. Phys. 1997, 81,
22412244.
69. D. Porath, Y. Levi, M. Tarabiah et al., Phys.
Rev. B 1997, 56, 98299833.
70. C. Joachim, J. K. Gimzewski, R. R. Schlittler
et al., Phys. Rev. Lett. 1995, 74, 2102.
71. S. Datta, W. D. Tian, S. H. Hong et al.,
Phys. Rev. Lett. 1997, 79, 25302533.
72. W. D. Tian, S. Datta, S. H. Hong et al., J.
Chem. Phys. 1998, 109, 28742882.
73. J. R. Hahn, Y. A. Hong, H. Kang, Appl.
Phys. A -Mater. Sci. Process. 1998, 66,
S467S472.
74. W. Han, E. N. Durantini, T. A. Moore et al.,
J. Phys. Chem. 1997, 101, 10 71910 725.
75. Y. Manassen, D. Shachal, Scanning tunneling microscopy (STM) on physisorbed

chemical groups of individual immobilized
molecules, Molecular Electronics: Science and
Technology, 1998, Vol. 852, pp. 277289.
76. G. Nagy, J. Electroanal. Chem. 1996, 409,
1923.
77. J. M. Tour, L. Jones, D. L. Pearson et al., J.
Am. Chem. Soc. 1995, 117, 95299534.
78. C. Zhou, C. J. Muller, M. A. Reed et al.,
Mesoscopic phenomena studied with mechanically controllable break junctions at
room temperature in Molecular Electronics
(Eds.: J. Jortner, M. Ratner), Blackwell Science, Oxford, 1997, pp. 191213.
79. C. Kergueris, J. P. Bourgoin, S. Palacin
et al., Phys. Rev. B: Condens. Matter 1999,
59, 12 50512 513.
80. J. D. Porter, A. S. Zinn, J. Phys. Chem. 1993,
97, 11901203.
81. K. Slowinski, R. V. Chamberlain, C. J.
Miller et al., J. Am. Chem. Soc. 1997, 119,
11 91011 919.
82. K. Slowinski, K. U. Slowinska, M. Majda, J.
Phys. Chem. 1999, 103, 85448551.
83. K. Slowinsky, H. K. Y. Fong, M. Majda, J.
Am. Chem. Soc. 1999, 121, 72577261.
84. R. E. Holmin, R. F. Ismagilov, R. Haag,
et al., Angewandte Chemie 2001, 40, 2316.
85. H. M. McConnel, J. Chem. Phys. 1961, 35,
508515.
86. A. Onipko, Y. Klymenko, L. Malysheva et al.,
Solid State Commun. 1998, 108, 555559.
87. A. Onipko, Y. Klymenko, J. Phys. Chem. A
1998, 102, 42464255.
88. B. L. Burrows, A. T. Amos, S. G. Davison,
Int. J. Quantum Chem. 1999, 72, 207220.
89. J. G. Simmons, J. Appl. Phys. 1963, 34,
17931803.
90. N. D. Lang, Phys. Rev. B 1987, 36, 8173.
91. N. D. Lang, Phys. Rev. B 1988, 37, 10 395.
92. N. Lang, W. Kohn, Phys. Rev. B 1973, 7,
3541.
93. R. Monnier, J. P. Perdew, Phys. Rev. B 1978,
17, 2595.
94. R. Monnier, J. P. Perdew, Phys. Rev. B 1980,
22, 1124(E).
95. A. Kiejna, Phys. Rev. B 1991, 43,
14 69514 698.
96. K. L. Jensen, J. Vac. Sci. Technol., B 1998,
17, 515519.
97. P. A. Serena, J. M. Soler, N. Garcia, Phys.
Rev. B 1986, 34, 6767.
98. A. Liebsch, Phys. Scr. 1986, 35, 354.
75
76
1 Fundamentals
99. B. G. Rudberg, M. Johnson, Phys. Rev. B
1991, 34, 9358.
100. A. Liebsch, Electronic Excitations at Metal
Surfaces, Plenum Press, New York, 1997.
101. M. Buttiker, R. Landauer, Phys. Rev. Lett.
1982, 49, 17391742.
102. M. Buttiker, Phys. Rev. B 1983, 27,
61786188.
103. J. Tersoff, D. R. Hamman, Phys. Rev. B
1985, 5031, 805.
104. J. Bardeen, Phys. Rev. Lett. 1961, 6, 57.
105. R. Landauer, IBM J. Res. Dev. 1957, 1, 223.
106. R. Landauer, Philos. Mag. 1970, 21,
863867.
107. Y. Imry, Physics of mesoscopic systems
in Directions in Condensed Matter Physics
(Eds.: G. Grinstein, G. Mazenko), World
Scientic, Singapore, 1986, p. 101.
108. Y. Imry, Introduction to Mesoscopic Physics,
109. W. H. Miller, S. D. Schwartz, J. W. Tromp,
J. Chem. Phys. 1983, 79, 48894898.
110. M. Galperin, D. Segal, A. Nitzan, J. Chem.
Phys. 1999, 111, 15691579.
111. V. Mujica, M. Kemp, M. A. Ratner, J. Chem.
Phys. 1994, 101, 68496864.
112. D. M. Newns, Phys. Rev. 1969, 178, 1123.
113. J. N. Onuchic, D. N. Beratan, J. Chem. Phys.
1990, 92, 722.
114. P. C. P. D. Andrade, J. N. Onuchik, J. Chem.
Phys. 1998, 108, 42924298.
115. A. Onipko, Phys. Rev. B 1999, 59,
999510 006.
116. A. Barraud, P. Millie, I. Yakimenko, J. Chem.
Phys. 1996, 105, 69726978.
117. A. Nitzan, J. Phys. Chem. A 2001, 105, 2677.
118. D. Segal, A. Nitzan, M. A. Ratner et al., J.
Phys. Chem. 2000, 104, 2790.
119. P. Sautet, C. Joachim, Chem. Phys. 1989,
135, 99.
120. P. Sautet, C. Joachim, Chem. Phys. Lett.
1991, 185, 23.
121. P. Sautet, J. C. Dunphy, D. F. Ogletree et al.,
Surf. Sci. 1994, 315, 127.
122. C. Chavy, C. Joachim, A. Altibeli, Chem.
Phys. Lett. 1993, 214, 569.
123. P. Doumergue, L. Pizzagalli, C. Joachim
et al., Phys. Rev. B 1999, 59, 15 91015 916.
124. C. Joachim, J. K. Gimzewski, Europhys. Lett.
1995, 30, 409.
125. C. Joachim, J. F. Vinuesa, Europhys. Lett.
1996, 33, 635640.
126. M. Magoga, C. Joachim, Phys. Rev. B 1997,
56, 47224729.

57, 18201823.
59, 16 01116 021.
129. M. P. Samanta, W. Tian, S. Datta et al.,
Phys. Rev. B 1996, 53, R7626R7629.
130. S. N. Yaliraki, M. Kemp, M. A. Ratner, J.
Am. Chem. Soc. 1999, 121, 34283434.
131. F. Biscarini, C. Bustamante, V. M. Kenkre,
Phys. Rev. B 1995, 51, 11 08911 101.
132. E. G. Emberly, G. Kirczenow, Phys. Rev. B
1999, 60, 60286033.
133. L. E. Hall, J. R. Reimers, N. S. Hush et al.,
J. Chem. Phys. 2000, 112, 15101521.
134. F. Faglioni, C. L. Claypool, N. S. Lewis et al.,
J. Phys. Chem. 1997, 101, 59966020.
135. C. L. Claypool, F. Faglioni, W. A. Goddard
et al., J. Phys. Chem. 1997, 101, 59785995.
136. S. Larsson, A. Klimkans, Theochem. J. Mol.
Struct. 1999, 464, 5965.
137. S. N. Yaliraki, A. E. Roitberg, C. Gonzalez
et al., J. Chem. Phys. 1999, 111, 69977002.
138. M. D. Newton, Chem. Rev. 1991, 91, 767.
139. T. Vondrak, C. J. Cramer, X. Y. Zhu, J.
Phys. Chem. B 1999, 103, 89158919.
140. E. Emberly, G. Kirczenow, Phys. Rev. Lett.
1998, 81, 52055208.
141. E. G. Emberly, G. Kirczenow, J. Phys. C:
Condens. Matter 1999, 11, 69116926.
142. T. Koopmans, Physica 1933, 1, 104.
143. M. F. Falcetta, K. D. Jordan, J. Phys. Chem.
1990, 94, 5666.
144. M. F. Falcetta, K. D. Jordan, J. Am. Chem.
Soc. 1991, 113, 2903.
145. C.-S. Chen, T. H. Feng, J. S.-Y. Chao, J.
Phys. Chem. 1995, 99, 8629.
146. P. D. Burrow, A. E. Howard, A. R. Johnston
et al., J. Phys. Chem. 1992, 96, 7570.
147. Z. I. Miskovic, R. A. English, S. G. Davison
et al., J. of Phys. Condensed Matter 1997, 9,
10 74910 760.
148. T. Seideman, W. H. Miller, J. Chem. Phys
1992, 97, 2499.
149. T. Seideman, W. H. Miller, J. Chem. Phys
1992, 96, 4412.
150. W. H. Miller, T. Seidman, Cumulative and
state to state reaction probabilities via a
discrete variable representation-absorbing
boundary condition greens function in
Time-Dependent Quantum Molecular Dynamics: Experiment and Theory (Ed.: J. Broeckhove), Plenum Press, New York, 1992.
151. D. Evans, I. Benjamin, T. Seidman et al.,
Abstr. Papers Am. Chem. Soc. 1996, 212,
194.
152. A. Haran, A. Kadyshevitch, H. Cohen et al.,
Chem. Phys. Lett. 1997, 268, 475480.
153. A. Mosyak, A. Nitzan, R. Kosloff, J. Chem.
Phys. 1996, 104, 15491559.
154. A. Mosyak, P. Graf, I. Benjamin et al., J.
Phys. Chem. A 1997, 101, 429433.
155. A. Nitzan, I. Benjamin, Acc. Chem. Res.
1999, 32, 854861.
156. I. Benjamin, D. Evans, A. Nitzan, J. Chem.
Phys. 1997, 106, 12911293, 66476654.
157. R. Naaman, A. Haran, A. Nitzan et al., J.
Phys. Chem. B 1998, 102, 36583668.
158. U. Peskin, A. Edlund, I. Bar-On et al., J.
Chem. Phys. 1999, 111, 75587566.
159. N. D. Lang, Phys. Rev. B 1988, 38, 10 395.
160. N. D. Lang, A. Yacoby, Y. Imry, Phys. Rev.
Lett. 1989, 63, 1499.
161. N. D. Lang, Phys. Rev. B 1992, 45, 13 599.
162. N. D. Lang, Phys. Rev. B 1995, 51, 2029(E).
163. N. D. Lang, Phys. Rev. B 1995, 52,
53355342.
164. N. D. Lang, P. Avouris, Phys. Rev. Lett. 1998,
81, 35153518.
165. N. D. Lang, P. Avouris, Phys. Rev. Lett. 2000,
84, 358361.
166. R. G. Parr, W. Yang, Density Functional Theory of Atoms and Molecules, Oxford University Press, Oxford, 1989.
167. O. Gunnarsson, R. O. Jones, Rev. Mod.
Phys. 1989, 61, 689.
168. N. Lang, Phys. Rev. Lett. 1986, 56, 1164.
169. M. Di Ventra, S. T. Pantelides, N. D. Lang,
Phys. Rev. Lett. 2000, 84, 979982.
170. M. Di Ventra, S. T. Pantelides, Phys. Rev. B
1999, 59, R5320R5323.
171. G. Galli, M. Parrinello in Computer Simulations in Material Science (Eds.: V. Pontikis,
M. Meyer), Kluwer, Dordrecht, The Netherlands, 1991.
172. K. Hirose, M. Tsukada, Phys. Rev. B 1995,
51, 52785290.
173. V. Mujica, A. E. Roitberg, M. A. Ratner, J.
Chem. Phys. 2000, 112, 68346839.
174. A. Szabo, N. S. Ostlund, Modern Quantum
Chemistry: Introduction to Advanced Electronic Structure Theory, McGraw-Hill, New
York, 1989.
175. H. J. Silverstone, M. L. Yin, J. Chem. Phys.
1968, 49, 2020.
176. S. Huzinaga, C. Arnan, Phys. Rev. A 1970,
1, 1285.
177. H. Burke, E. K. U. Gross in Density Functionals: Theory and Applications (Ed.: D. Joubert), Springer, Berlin, 1998.
178. D. Lamoen, P. Ballone, M. Parrinello,
Phys. Rev. B: Condens. Matter 1996, 54,
50975105.
179. A. Aviram, M. A. Ratner, Chem. Phys. Lett.
1974, 29, 277.
180. D. H. Waldeck, D. N. Beratan, Science 1993,
261, 576, 577.
181. R. A. Marcus, J. Chem. Soc., Faraday Trans.
1996, 92, 39053908.
182. M. Pomerantz, A. Aviram, R. A. McCorkle
et al., Science 1992, 255, 1115.
183. A. S. Martin, J. R. Sambles, Adv. Mater.
1993, 5, 580582.
184. A. S. Martin, J. R. Sambles, G. J. Ashwell,
Phys. Rev. Lett. 1993, 70, 218221.
185. C. M. Fischer, M. Burghard, S. Roth et al.,
Europhys. Lett. 1994, 28, 129134, 375.
186. K. Nagesha, J. Gamache, A. D. Bass et al.,
Rev. Sci. Instrum. 1997, 68, 38833889.
187. R. M. Morsolais, M. Deschenes, L. Sanche,
Rev. Sci. Instum. 1989, 60, 27242732.
188. G. Makov, A. Nitzan, L. E. Brus, J. Chem.
Phys. 1988, 88, 50765085.
189. D. K. Ferry, S. M. Goodnick, Transport
in Nanostructures, Cambridge University
Press, Cambridge, 1997, Vol. 6.
190. R. Wilkins, E. Ben-Jacob, R. C. Jaklevic,
Phys. Rev. Lett. 1989, 63, 801804.
191. M. A. Kastner, Rev. Mod. Phys. 1992, 64,
849.
192. S. H. M. Persson, L. Olofsson, L. Gunnarsson, Appl. Phys. Lett. 1999, 74, 25462548.
193. M. Dorogi, J. Gomez, R. Osifchin et al.,
Phys. Rev. B: Condens. Matter 1995, 52,
90719077.
194. R. P. Andres, S. Datta, M. Dorogi et al., J.
Vac. Sci. Technol., A 1996, 14, 11781183.
195. R. P. Andres, T. Bein, M. Dorogi et al., Science 1996, 272, 13231325.
196. A. N. Korotkov, Coulomb blockade and digital single electron devices in Molecular Electronics (Eds.: J. Jortner, M. Ratner), Blackwell Science, Oxford, 1997, pp. 157189.
197. A. M. Kuznetsov, J. Ulstrup, J. Electroanal.
Chem. 1993, 362, 147152.
198. F.-R. F. Fan, A. J. Bard, Science 1997, 277,
17911793.
199. L. I. Malysheva, A. I. Onipko, Phys. Rev. B
1992, 46, 39063915.
200. V. Mujica, M. Kemp, A. Roitberg et al., J.
Chem. Phys. 1996, 104, 72967305.
77
78
1 Fundamentals
201. Y. Kawahito, H. Kasai, H. Nakanishi et al.,
J. Appl. Phys. 1999, 85, 947952.
202. Y.-Q. Li, C. Gruber, Phys. Rev. Lett. 1998, 80,
10341037.
203. A. S. Davidov, V. M. Ermakov, Physica 1987,
28D, 168180.
204. H. L. Berkowitz, R. A. Lux, J. Vac. Sci.
Technol. 1987, B5, 967970.
205. F. W. Sheard, G. A. Toombs, Appl. Phys.
Lett. 1988, 52, 12281230.
206. N. C. Kluksdahl, A. M. Kriman, D. K. Ferry
et al., Phys. Rev. B 1989, 39, 77207735.
207. A. Nakano, R. K. Kalia, P. Vashishta, Phys.
Rev. B: Condens. Matter 1991, 44,
81218128.
208. J. Chen, M. A. Reed, A. M. Rawlett et al.,
Science 1999, 286, 15501552.
209. S. A. Gurvitz, H. J. Lipkin, Y. S. Prager,
Mod. Phys. Lett. B 1994, 8, 1377.
210. S. A. Gurvitz, H. J. Lipkin, Y. S. Prager,
Phys. Lett. A 1996, 212, 9196.
211. S. A. Gurvitz, Y. S. Prager, Phys. Rev. B
1996, 53, 15 93215 943.
212. S. Datta, Electric Transport in Mesoscopic

Systems, Cambridge University Press, Cambridge, 1995.
213. E. L. Wolf, Principles of Electron Tunneling
Spectroscopy, Oxford University Press, New
York, 1985.
214. E. Emberly, G. Kirczenow, Phys. Rev. B
2000, 61, 57405750.
215. J. Jortner, M. Bixon, T. Langenbacher et al.,
Proc. Natl. Acad. Sci. USA 1998, 95,
12 75912 765.
216. M. Bixon, B. Giese, S. Wessely et al.,
Proc. Natl. Acad. Sci. USA 1999, 96,
11 71311 716.
217. B. Giese, S. Wesswly, M. Spormann et al.,
Angew. Chem. Int., Ed. Engl. 1999, 38,
996998.
218. Y. A. Berlin, A. L. Burin, M. A. Ratner, J.
Phys. Chem. 2000, 104, 443445.
219. B. Giese, Acc. Chem. Res. 2000, 33, 631636.
220. A. Nitzan, Annu. Rev. Phys. Chem. 2001, 52,
681750.
81
2.1
Diffusion and Migration

Jose A. Manzanares
University of Valencia, Burjasot, Spain
..
Kyosti Kontturi
Helsinki University of Technology, Espoo,
Finland
Introduction
This chapter describes mass transport to

electrodes by diffusion and migration. It
is assumed throughout that there is no
convective motion of the solution, and
transport is described with respect to a
reference frame xed to the laboratory
[1, 2]. Thus, many of the equations derived
in this chapter cannot be directly applied
to systems with bulk motion.
The chapter emphasizes migrational effects and provides analytical expressions
for the electric potential drop in the
solution under different experimental conditions. The fundamental concepts are
discussed in detail, and a number of
important restrictions are introduced for
the sake of clarity. First, diluted solutions are considered throughout and the
ux equations incorporate neither cross
terms nor activity coefcients [25]. Second, one-dimensional systems are considered, except when presenting the transport
equations in Sect. 2.1.1. Third, except in
a couple of sections, it is assumed that

there are no homogeneous reactions. Insoluble species are also absent. Fourth,
since the chapter is mainly of interest
to cases in which mass transport is rate
controlling, reversible electrode reactions
are assumed and no kinetic effects are
considered. Fifth, most sections involve
steady state transport because the theoretical description of transient experiments
in the presence of migration is difcult
and quite often requires numerical solutions that bring much more mathematical
complexity than new physical insights.
Nevertheless, the effect of migration is analyzed in a simple transient case. Finally, the
local electroneutrality (LEN) assumption is
used. This is indeed a good approximation
for most experimental situations, although
deviations from LEN occur under some
conditions. These deviations are discussed
in the last section.
2.1.1
Fundamental Concepts
This section presents the basic equations

describing mass transport to electrodes:
the ux equations, the continuity equation, and the LEN assumption. The ux
equations are derived from the concept of
electrochemical potential, which is used as
a general term valid for both charged and
82
2 Transport Phenomena
neutral species. The connection between

electric conduction and the basic transport
mechanisms, diffusion and migration, is
discussed in detail, and the difference between ohmic and total electric potential
gradient is emphasized. The spatial region in which mass transport takes place
and the coupling between mass transport
and the electrode reaction are discussed
next. At this point, the one-dimensional
geometry to be used henceforth is introduced. The electric potential drops in the
system and the sign conventions are also
dened. The boundary conditions for the
transport equations are written in terms of
the integral transport numbers. For the
sake of simplicity, parallel reactions [6]
and adsorption phenomena [7] are not
considered.
2.1.1.1 Diffusionmigration Flux
Equations
The ux equations describing the behavior
of electrochemical systems are intimately
related to the Second Law of Thermodynamics. Consider two neighboring volume
elements V and V of a solution having the same temperature and pressure,
but different electrochemical potentials of
their constituents. The difference between
the electrochemical potential of species
i, i and i , in these volume elements
implies that there is no distribution equilibrium and this species tends to move
from one volume element to the other.
This motion of species i from one volume
element to its neighbor is generically called
transport of species i. Since the transport
of the different species in solution takes
place under thermal and mechanical equilibrium, the change in internal energy U
of these two volume elements is [1, 8, 9]
dU = T dS +

i
i dni
(1)
dU = T dS +
i dni
(2)
where T is the thermodynamic temperature, S the entropy, and ni the number of

moles of species i. Moreover, considering
that the exchange of matter between V and
V takes place without energy exchange
with their surroundings, it is satised that
dU + dU = 0 and dni + dni = 0. The
Second Law of Thermodynamics requires
that the entropy of the isolated system
formed by the two volume elements must
increase

T (dS + dS ) =
( i i )dni 0
i
(3)
When the transport of species i is not
coupled to that of other species, each
individual term of this sum is positive.
In this case, dni is determined only by
( i i ) and species i moves towards
the region in which its electrochemical
potential is lower, that is, dni < 0 when
i < i and vice versa. This happens, for
instance, in the case of ionic species in
diluted solutions. When the transport of
different species is coupled, one or more
terms in the sum can be negative, but the
sum is always positive.
The transport of species i can be
described in terms of either its velocity vi or
its ux density Ji , which are related to each
other by the molar concentration, Ji =
ci vi . If the area of the surface separating
the two volume elements is dA and its
orientation is given by the unit vector
n(from
V to V ), the number of moles

of species i crossing the surface in a time
dt is dni = Ji n dA dt.
When the difference between i and i
is not too large, it can be assumed [10]
that the rate of change of the amount of
species i, dni /dt, is proportional to the
difference i i or, more exactly, to
2.1 Diffusion and Migration
the gradient of this potential normal to

i n.
In this linear
the surface, i /n
approximation, the velocity of species i is
expressed as
i
vi = ui
(4)
where ui is its mobility. Similarly, the ux

density takes the form
i (5)
i = Di ci
Ji = ci vi = ui ci
RT
where R is the universal gas constant
and the Einstein relation between mobility
and diffusion coefcient, Di = ui RT , has
been used.
At constant temperature and pressure,
i originates from changes
the gradient
in composition and electric potential
so that
ln ci + zi F

i = RT
(6)
where F is the Faraday constant and zi

is the charge number of species i. Taking
Eq. (6) to Eq. (5), the NernstPlanck ux
equation [11]
i + zi ci f )

Ji = Di (c
(7)
is obtained, where f denotes the ratio

F /RT . The two terms in the righthand side of this equation represent the
transport mechanisms of diffusion and
migration, respectively. Diffusion is a consequence of the random thermal motion
of the particles, which tends to make the
concentration of all species uniform. Migration accounts for the inuence of the
= ,

on the random
electric eld, E
motion of the charged particles, and Eq. (7)
shows that the particles acquire a component of their velocity along the direction
of the electric eld as a result of this
inuence.
Poisson Equation and the LEN

Assumption
Equation (5) states that the ux density of
species i in a diluted solution depends on
i but not on other gradients
j =i . In
this sense, there is no correlation between

the motions of two different species, i
and j . The migrational term in Eq. (7),
however, imposes a strong constraint on
the motion of charged species by coupling
their ux densities Ji .
The electrical state of the system is so
sensitive to small changes in composi in Eq. (7)
tion [9] that the gradient
cannot be assumed, in general, to be a
simple constant [12], and an additional
equation is required to determine it. Given
the slowness of particle motion in solution,
it is justied to use the Poisson equation
from electrostatics to relate the changes in
electric potential to the local electric charge
density e
2.1.1.2
2 =
e
F
zi ci
=
(8)
where is the dielectric permittivity of

the solution.
Consider, for example, a 10 mM 1 : 1
electrolyte solution that contains a positive
charge density of the order of 106 C m3
counterbalanced by a negative charge of
106 C m3 . Since the dielectric permittivity of aqueous electrolyte solutions is of
the order of 109 C V1 m1 , the righthand side of Eq. (8) is roughly the difference between two terms whose order
of magnitude is 1015 V m2 . The typical
order of magnitude of the electric elds,
however, can be estimated as 10 V m1
and they are practically independent on
position (except at the interfacial regions),
so that 2 is many orders of magnitude smaller than 1015 V m2 . This means
that the imbalance between cation and
83
84
anion concentrations in the system that

is required to set up the typical electric
potential distributions is many orders of
magnitude smaller than the solution concentration. Therefore, the space charge
density can be neglected and use can be
made of the LEN assumption [13]

zi ci = 0
(9)
i
This equation replaces Eq. (8) in the

description of transport processes. Though
it is customary to use the equal symbol
in Eq. (9), the electric charge density is
not strictly zero, and the Poisson equation
cannot be reduced to the Laplace equation
2 = 0.
Continuity Equation
The spatial variation of the ux density of
species i is determined by the principle of
conservation of the amount of substance
for this species, which is represented by the
continuity equation. Consider an arbitrary
volume V in an electrolyte solution, in
which the molar concentration of species
i is ci (r , t). The amount of substance of
species i within V can change as a result
of either a net ow of particles through
the surface S enclosing the volume V , or
due to homogeneous chemical reactions
occurring within V . If the surface S is
divided into surface elements of area dA
perpendicular to the unit vector n (oriented
outwards from V ), the number of moles of
species i entering V through this surface
element in a time dt is dni = Ji ndAdt.

The homogeneous reactions generate or
consume this species at a net rate ri (r , t).
Since the (net) accumulation rate of
species i within V is given by the sum
of the (net) generation rate due to the
chemical reaction and the (net) incoming
ux through S, it is concluded that
d
dt

ci (r , t)dV =
ci
dV
t

ri dV Ji ndA
=
Ji )dV
(ri
(10)
where Gauss divergence theorem has

been used. Moreover, since no restrictions apply to the volume V , Eq. (10)
implies that
ci
Ji = ri
+
t
(11)
which is the continuity equation for

species i.
2.1.1.3
Ohms Law and Migrational

Transport Numbers
In a diluted electrolyte solution of uniform
composition, the ux density Ji = ci vi is

proportional to ,
and the conduction
electric current density
2.1.1.4
zi Ji
(12)
i = 0, i)
(c
(13)
I F
satises Ohms law

I = ,
where
F2 2
zi Di ci
RT
(14)
is the electrical conductivity of the solution.

The minus sign in Eq. (13) means that the
electric current ows in the direction of
decreasing electric potential. Equation (14)
evidences that the conductivity of the solution is determined by all ionic species in
solution, no matter whether they are active
or not. The migrational transport number

z 2 D i ci
ti i
zj2 Dj cj
(15)
represents the relative contribution of

species i to the conductivity of the solution.
Unfortunately, concentration gradients
develop under most experimental conditions, and the motion of charge carriers
takes place by both diffusion and migration
as described by the NernstPlanck equation, Eq. (7). In this case, Eqs. (14 and 15)
provide the local (i.e. position-dependent)
values of the conductivity and the transport
number of species i, respectively. By writing the NernstPlanck equation in terms
of ti as [1, 14]
ti
i
Ji = 2
zi F 2
(16)
Eq. (12) leads to the modied Ohms

law [5, 15]
ohm (17)

dif )
I = (
in which
ti
ohm = 1
i
F
zi
(18)
is the ohmic potential gradient, and

dif = F
i
zi Di c
RT ti
ln ci
F
zi
(19)
is the diffusion potential gradient.

That is, the concentration gradients
break the proportionality between the
current density and the electric eld; fur are not
thermore, in some cases, I and
even parallel vectors, a fact that is relevant
to the tertiary current distribution [16].
This is due to the diffusion potential gradient that originates from the diffusion
process, and therefore an expression similar to Ohms law can only be recovered
by subtracting the diffusion potential gradient from the total potential gradient, as
shown by Eq. (17). A theoretical simplication sometimes used in the literature
consists in neglecting the diffusion potential by assuming all ionic diffusivities to be
equal [17, 18].
Diffusion-conduction Flux
Equation
The introduction of the modied Ohms
law, Eq. (17), in the NernstPlanck equation, Eq. (7), leads to the following form of
the ux equation
2.1.1.5

i + zi ci f
dif ) + ti I
Ji = Di (c
zi F
(20)
For the sake of simplicity, consider the case
of a binary solution obtained by complete
dissociation of a strong electrolyte A1 B2
into 1 ions Az1 of charge number z1
and 2 ions B z2 of charge number z2 ;
obviously, the relation z1 1 + z2 2 = 0 is
satised. The ionic concentrations are then
related to the stoichiometric electrolyte
concentration c12 by the relations c1 =
1 c12 and c2 = 2 c12 which allows Eq. (20)
to be rewritten as

ti I
12 + ti I
Ji = i J12 +
= i D12 c
zi F
zi F
(21)
where J12 is the electrolyte ux density, and
D12 =
(1 + 2 )D1 D2
2 D1 + 1 D2
(22)
is the NernstHartley expression for the

diffusion coefcient of the dissociated
electrolyte [3].
85
86
Equations (20 and 21) describe ionic

transport as a combination of electrolyte
diffusion and ionic conduction, and hence
the name diffusionconduction ux equation is suggested. Its comparison with
the diffusionmigration ux equation,
Eq. (7), evidences the difference between
migration and conduction. Migration is
due to electric elds, either external or
internal, and does not require a nonzero
current density. Conduction is the ionic
motion associated to the part of the electric eld that is controlled externally. Note
that the electric current density is a measurable magnitude but not the local electric
eld. Conduction requires a nonzero electric current density.
Note, nally, that the dissociated electrolyte satises Ficks law, J12 = D12
12 , because both ions move at the same
c
velocity in the absence of current (I = 0),
and the dissociated electrolyte is the only
independently moving substance [1]. This
is not in contradiction with the fact that the
ionic diffusion coefcients are different to
each other, D1 = D2 , because the diffusion potential gradient affects the ionic
motion even when I = 0. In conclusion,
although the NernstPlanck equation establishes that the ux of species i in a
diluted solution is only proportional to the
i , the contribution of the elecgradient
i couples the
tric potential gradient to
transport of different species.
Diffusion Boundary Layer
In general, the electrode reaction makes
the concentration of species i to vary with
the distance to the electrode surface. These
concentration changes occur not only for
the species participating in the electrode
reaction, but also to other charged species
because of their electrostatic coupling.
The progress of the electrode reaction
is accompanied by a propagation of
2.1.1.6
the concentration variations toward the

interior of the solution, which affects
a region that grows in thickness. It
is possible, however, to conne the
concentration variations to a thin region
near the electrode by vigorous stirring,
thin here means of the order of tenths
of a millimeter. The solution can then
be divided in two regions: the diffusion
boundary layer (DBL) adjacent to the
electrode surface, in which concentration
varies with position and diffusion affects
mass transport; and the bulk solution, in
which concentration is made uniform by
stirring and transport is due to migration
and convection [5, 19, 20]. Although the
transition between these two regions
is actually smooth, it is practical to
assume that the transition is abrupt and
to assign a thickness to the DBL,
which can be empirically determined from
the stirring rate or other hydrodynamic
conditions [21, 22]. The thickness is
considered to be nite when describing
steady state transport (indeed, this is a
necessary condition for reaching the steady
state); innite DBL thickness, however, is
assumed in Sect. 2.1.5.
In the following sections, the onedimensional geometry described by Fig. 1
is considered. The electrode is located at
x = 0 and the DBL in the region 0 < x < .
The electric current density I is positive for
anodic oxidations and negative for cathodic
reductions, in agreement with the IUPAC
convention. The electric potential drop in
the DBL is dened as the potential in the
bulk solution (x = ) relative to that at the
electrode surface (x = 0), % b s .
Faradays Law and Integral
Transport Numbers
The charge transfer between an electrolyte
solution and an electrode (quite frequently,
a metal phase) can be represented by the
2.1.1.7
DBL
Electrode
Bulk solution
I
fb
f(x)
Potential drop in DBL
f fb fs
fs
Electrode potential
E fe fs
fe
Fig. 1 Schematic drawing of the system considered. The electric

potential distribution has been shown for a cathodic process (I < 0).
Superscripts b, s, and e denote, respectively, bulk solution (x = ),
solution adjacent to the electrode surface (x = 0), and electrode phase.
general electrode reaction equation

ir Bizi + ne = 0
(23)
where Bi and ir are the molecular formula

and the stoichiometric number of species
i, respectively. The value of ir is positive
for products and negative for reactants.
The stoichiometric number
of the electron
in the reaction, n = i zi ir , is positive
for (anodic) oxidations and negative for
(cathodic) reductions. The electrons in
Eq. (23) are considered to be in the metal
phase, and species Bi can be either in
solution or in the electrode phase; it
is obviously necessary that at least one
active species Bi is in solution because
Eq. (23) describes charge transfer between
two phases.
The evolution of the electrode reaction
needs the transport of reactants from the
solution to its surface and products vice
versa. Faradays law establishes that the
reaction rate d/dt is proportional to the
conduction electric current density and,

therefore, that the ux density of species i
at the electrode surface is given by
r d
1 dni
I
= i
= ir
(24)
A dt
A dt
nF
where A is the electrode surface area.
Equation (24) shows clearly that, at the
electrode surface, the fraction of electric
current density transported by species i is
given by the magnitude
Jis =
zi ir
zi F Jis
=
(25)
I
n
which is called the integral transport
number of species i [23]; this magnitude
is zero for the species that are not
involved in the electrode reaction (i.e. the
electroinactive species).
It is important

to observe that i Tis = 1 but Tis is not
bounded between 0 and 1. For example, in
the electrode reaction
Tis
Fe(CN)6 3 (aq) Fe(CN)6 4 (aq) + e = 0

(26)
87
88
the transport numbers at the elec= 3 and

trode surface are T s
Fe(CN)6 3
s
= 4.
T
Fe(CN)6 4
Some electrode reactions, such as the
anodic dissolution of copper
Cu+ (aq) Cu(s) + e = 0
(27)
the hydrogen evolution at a platinum

electrode
H2 (g) 2H+ (aq) + 2e = 0
(28)
or the oxidation of silver at a Ag|AgCl

electrode
AgCl(s) Ag(s) Cl (aq) + e = 0
(29)
involve only one electroactive species
(Cu+ , H+ , and Cl , respectively), which is
denoted by subscript 1 henceforth. Since
the current density in these processes is
due to the single electroactive species,
n = z1 1r and T1s = 1, while Tis = 0 for
i = 1. In these simple cases, the DBL
becomes depleted of this species when
J1 < 0 (i.e. the species is consumed at the
electrode surface), and the potential drop
satises z1 % > 0. On the contrary, when
J1 > 0, the DBL becomes concentrated in
species 1 and z1 % < 0.
Note also that Tis is zero for neutral
species but this does not imply that
they are not transported towards or
from the electrode. For example, in the
case of nitrous acid in the electrode
reaction
3H+ (aq) + NO3 (aq) HNO2 (aq)
H2 O(aq) + 2e = 0
(30)
s
s
THNO
= 0 but JHNO
= I /2F , as can
2
2
be easily deduced from Eq. (24). (For the
neutral species, the transference number
i Ti /zi constitutes a good alternative to
Ti [24].)
The integral transport number of species

i can also be dened at locations different from the electrode surface as Ti
zi F Ji /I [23]. Under steady state conditions and in the absence of homogeneous
reactions, both the current density I and
the ux density Ji do not vary with position,
and Ti takes the value specied by Eq. (25)
throughout the whole system even when
concentration gradients are present. Otherwise, Ti is position-dependent. It is
important to observe that the integral
transport numbers Ti are equal to the
migrational transport numbers ti in the
absence of concentration gradients only;
indeed, concentration gradients originate
due to the fact that Ti = ti .
Nernst Equation and
Concentration Overpotential
In order to drive an electric current density across the system depicted in Fig. 1,
an electric potential difference must be applied between the electrode and the bulk
solution. This potential difference can be
expressed in terms of three contributions:
the potential drop in the DBL, the equilibrium interfacial or Nernst potential drop,
and the activation overpotential. For large
enough kinetic currents, the latter contribution can be neglected [7, 25] and the
potential difference required for the electrode process is Eeq %, where Eeq is
the equilibrium electrode potential (i.e. the
electric potential of the electrode relative
to that at the adjacent electrolyte solution).
The value of Eeq is given by the Nernst
equation. For the electrode Reaction (23),
the Nernst equation takes the form
2.1.1.8
Eeq = E +
RT r cis
i ln
nF
c
(31)
where E is the standard (or, more exactly,

the formal) electrode potential and c is
the standard concentration. The concentrations cis of the active species in Eq. (31)
are those at the electrode surface, which
are evaluated by extrapolation to the electrode surface of the concentration prole
within the DBL [19]. In other words, the
electrode surface at position x = 0 should
be understood as the neutral solution just
outside the electrical double layer (EDL).
Similarly, % refers to the potential difference between the bulk solution and
that point in the solution (that is, between
the bulk solution|DBL and the EDL|DBL
boundaries). Hence, no details of the EDL
structure are considered when evaluating
the equilibrium interfacial potential drop
and the surface concentrations of the active species.
Since the interfacial concentrations of
the active species vary with the electric
current density, Eeq also does and its
variation is often described in terms of
the concentration overpotential [19, 26]
b
c Eeq Eeq
RT r cis
i ln b (32)
nF
ci
i
which represents the potential of the electrode in equilibrium with the interfacial
concentrations that are established when
a current density I is driven through the
system, Eeq , relative to the (constant) potential of a similar electrode equilibrated
b . The potential
with the bulk solution, Eeq
difference between the electrode and the
b + %.
bulk solution is then Eeq
c
Note, nally, that when the supporting
electrolyte is used, the ionic strength
of the DBL is not signicantly affected
by the concentration polarization of the
active species, and therefore, their activity
coefcients are practically independent of
the current density [26]. This justies the
use of concentrations instead of activities
in the above equations.
2.1.2
Steady State Currentvoltage Curves of

Systems with One Active Species
The nite transport rate of the electroactive species toward the electrode surface
implies the existence of a limiting current
density, that is the most important consequence of mass transport. This limiting
current density is dened by the condition
of maximum concentration gradient of the
electroactive species, which is achieved
when the concentration of this species
tends to zero at the electrode | depleted
DBL interface (or, more accurately, at the
EDL | DBL boundary).
In this section, the mass transport effects
on the steady state currentvoltage curves
are described for the (relatively common)
case of systems with only one electroactive
species in solution. This species is denoted
by subscript 1. The integration procedure
of the transport equations in multi-ionic
systems is presented rst and the general
currentvoltage curve is derived. Except
in Sect. 2.1.2.5, homogeneous reactions
are assumed to be absent. The interesting case of excess supporting electrolyte,
in which the active species is present in
very small concentration and transport
takes place mostly by diffusion, is discussed. Finally, the simple cases of binary
and ternary solutions are considered to illustrate the conclusions drawn from the
general solution.
Integration of the Transport
Equations
The fact that the ux density of the
electroinactive species, i = 1, is zero
simplies signicantly the solution of the
transport equations. These species are
in electrochemical equilibrium within the
DBL, and hence their concentrations ci
are related to the electric potential by the
2.1.2.1
89
90
where %ci cib cis . Introducing the limiting diffusion current density
Boltzmann equation
d
dci
= zi ci f
dx
dx
(i = 1, Ji = 0) (33)
Thus, the ionic concentrations at the

electrode surface, x = 0, are related to
those in the bulk solution, x = , by
cis = cib ezi f %
(i = 1, Ji = 0)
(34)
where % b s is the electric potential drop across the DBL (see Fig. 1).
Furthermore, the LEN assumption

zi ci = z1 c1 +
zi ci = 0
(35)
i =1
implies that not only ci (i = 1) but also c1

is determined by . Thus, for instance, the
surface concentration of the electroactive
species is
c1s =
1 s
1 b zi f %
zi ci =
zi ci e
z1
z1
i =1
i =1
(36)
The currentvoltage relation can be obtained from the ux equation of the
electroactive species

I
d
dc1
J1 =
= D1
+ z1 c1 f
z1 F
dx
dx
(37)
by transforming the migration term with
the help of Eq. (35) to
z1 c1 f

dci
d
d
zi ci f
=
=
dx
dx
dx
i =1
i =1
(38)
Integration of Eq. (37) over the DBL
then gives

I
= %c1 +
%ci
z1 F D1
=

i =1
i =1
zi
1
z1

%ci (39)
ILd,1
z1 F D1 c1b
(40)
and using Eqs. (34 and 36), the currentvoltage relation takes the form

cb
zi
i
1
(1 ezi f % )
b
z
c
1
i =1 1
(41)
Note that the denition of ILd,1 has not
been deduced from the condition c1s = 0,
but it rather corresponds to a situation
in which the transport of species 1 takes
place mostly by diffusion, and hence its
name. Moreover, Eq. (40) corresponds to
the particular case in which there is only
one active species; see Eq. (78) below for
the general denition.
It is interesting to observe that species
1 is the only one that moves, while the
inactive species are standing still over
the system. In a sense, this is similar to
the conduction of electrons in a metal,
in which electrons move and the ions
are standing on the crystal lattice and,
therefore, Ohms law is expected to be
satised. Indeed, Eq. (37) shows that the
electric current density is proportional
to the electric eld, I eff (d/dx),
in which
I = ILd,1
eff (x) =

F2
zi2 ci (x)
D1
RT
(42)
is the local effective conductivity and

Eqs. (33 and 35) have been used. There is,
however, a difference between these two
systems. The static distribution of the inactive species is determined by the electric
potential distribution and hence, although
I is only due to species 1, all species are
relevant to the conduction process.
The local ohmic behavior is not in

contradiction with the fact that Eq. (41)
shows a nonlinear variation of I with
the potential drop %. At low potentials,
f % 1, the DBL is practically nonpolarized (i.e. the ionic concentrations take
their bulk values throughout the DBL) and
the slope of the currentvoltage curve is
b . When increasing %, the slope of the
eff
currentvoltage curve decreases owing to
the development of concentration polarization. The effective conductivity varies then
with position and this makes the overall
system behavior to be nonohmic.
Solutions of Homovalent Ions,
|zi | = z
The theoretical analysis of ion transport
requires the grouping of ions according
to their valencies [2731]. All ions of
the same valency (a term used here as
equivalent to charge number) are grouped
together to constitute a class and the
sum of their concentrations is termed the
class concentration [32], denoted by C.
The simplest transport problem involves
only two classes corresponding to valencies
z and z. This is sometimes referred to as
the homovalent case [33]. Some interesting
conclusions are drawn from Eqs. (36 and
41) in this situation.
Denote the class of the electroactive
species by subscript 1 and the other class by
subscript 2. The LEN assumption requires
the two class concentrations to be equal to
each other

C1
ci =
ci C2 C (43)
2.1.2.2
zi =z1
zi =z1
Equation (36) then simplies to

c1s = C b ez1 f % + (c1b C b )ez1 f %
(44)
and the maximum potential drop in the
DBL, which is attained under limiting
conditions c1s = 0, satises

1/2

c1b
z1 f %L
= 1 b
e
C
(45)
Thus, when the ionic strength of the

solution is much larger than the concentration of the electroactive species (C b c1b ),
that is, in the presence of excess supporting electrolyte case, z1 %L tends to zero
and the electric potential drop is then negligible [5, 26, 34]. On the contrary, in the
absence of a supporting electrolyte, the
electroactive species is the only species of
its class, C b = c1b , and z1 %L diverges.
The conclusion that migration is absent
when the ionic strength of the solution
is much larger than the concentration of
the electroactive species is not restricted to
the homovalent case; note that the equation I eff (d/dx) implies that d/dx
decreases as eff increases.
In homovalent systems, the currentvoltage curve, Eq. (41), takes the form
I = IL
where
IL = ILd,1
1 ez1 f %
1 ez1 f %L
2C b
c1b
1 1
c1b
Cb
(46)
1/2
(47)
is the limiting current density (i.e. that
corresponding to c1s = 0 and % = %L ).
The value of IL varies between 2ILd,1
in the absence of supporting electrolyte
(C b = c1b ) and ILd,1 in the presence
of excess supporting electrolyte (C b
c1b ). Migrational effects in homovalent
solutions can therefore account for up to a
factor 2 in IL .
Binary Electrolyte Solutions
When the solution contains only two
ionic species, the concentration proles
2.1.2.3
91
includes the case of both cathodic and

anodic processes. Consider rst that the
electroactive species is an anion and,
consequently, that IL > 0; note that the
charge numbers have been dened with
sign. In the depleted DBL adjacent to
the anode, the current density is positive
and tends to the limiting value when
the (negative) potential drop increases
in absolute value. In the concentrated
DBL adjacent to the anode (for the
same geometry described in Fig. 1, that
is, electrode located at x = 0 and DBL
extending from x = 0 to ), the current
is negative, and the potential drop is
positive. Increasing the potential drop
now leads to an increase of the electrical
conductivity of the DBL, and hence of the
slope of the currentvoltage curve. When
the electroactive species is a cation, these
comments must be changed accordingly,
but Fig. 2 remains valid.
Figure 3 shows the concentration, ci (x),
and electric potential, (x), proles
given by
c2
I
c1
x
z2 f ( b )
=
=
e
=
1
IL
c1b
c2b
(50)
which can be obtained by integration of
Eq. (37) from x to , after use of Eq. (38).
In the depleted DBL, the interfacial
are linear and the diffusion and migration

terms in the ux equations are independent on position within the DBL. The electroinactive species can be thought to be in
equilibrium as a result of the balance of diffusional and migrational terms in its ux
equation. The migrational contribution to
the ux of the electroactive species is equal
(due to the LEN assumption) to that of the
inactive species, except for a sign reversal, and hence the ux of the electroactive
species can be described as the sum of two
diffusional contributions, Eq. (38). This
means that the slopes of the concentration
proles are proportional to the electric current, Eq. (39), and the occurrence of a limiting current is clearly associated to the maximum concentration gradient that can be
established over the DBL, that is, to vanishing concentrations at the electrode surface.
The currentvoltage curve of this system
is a particular case of Eq. (41), which takes
the simple form [34]
I = IL (1 ez2 f % )
(48)
1.0
0.5
Concentrated
DBL
where

(z1 z2 )F D1 c2b
z1
IL = 1
ILd,1 =
z2
(49)
is the limiting current density. This curve
has been plotted in Fig. 2. The gure
I/IL
92
Depleted DBL
0.0
0.5
1.0
1
Currentvoltage curves of a DBL

containing a strong binary electrolyte.
Magnitudes z2 , %, I, and IL are dened
with sign; subscript 2 denotes the
electroinactive species. Negative values
of I/IL correspond to electrode
5 dissolution (anodic or cathodic), and
positive values to single ion discharge at
the electrode.
Fig. 2
z2 ff

Concentrated DBL
2.0
1.0
1.5
0.5
ci(x)/cbi
ci(x)/cbi
Depleted DBL
0.0
0.4
0
0.0
0.5
1.0
0.0
x/d
0.5
1.0
0.8
1.5
z2 f [fb f(x)]
z2 f [fb f(x)]
1.0
0.0
4
x/d
Fig. 3 Concentration and electric potential proles for depleted

(I/IL > 0) and concentrated DBL (I/IL < 0). The continuous lines
correspond to |I/IL | = 0.5 and the dashed lines to |I/IL | = 1.0.
concentration cis decreases with increasing

%, and the limiting current IL is
reached when z2 % and cis 0.
In the concentrated DBL, the interfacial
concentration cis increases with increasing
% (in absolute value).
It is important to stress that % is
the potential drop in the DBL only.
There is also a potential drop at the
electrode|DBL interface that changes with
concentration polarization. The variation
of the latter is given by the concentration
overpotential that, in this case, takes
the form

cs
1
1
I
ln 1b =
ln 1
z1 f
z1 f
IL
c1
(51)
Combining Eqs. (48 and 51), the currentvoltage curve can be expressed as [34]
c =

z2 z1
f (c %)
I = IL 1 exp
z1 z2
(52)
where c % is the change in the
potential difference between the electrode
and the bulk solution due to the passage
of the electric current.
Ternary Electrolyte Solutions. The

Supporting Electrolyte
Consider a ternary system in which the
common ion, i = 3, is an inactive species.
(If the electroactive species were the
common ion, the system behavior would
be similar to that of a binary solution.)
A key feature of systems with only one
electroactive species is that there is a
background concentration of indifferent
electrolytes that maintain a fairly uniform
conductivity of the DBL. This implies
that the electric potential drop is small
and is quite uniformly distributed over
the DBL. The concentration prole of the
electroactive species is (slightly) nonlinear,
and the same is true in general for the
inactive ions, but the system behavior
is still quite simple. In the particular
case z1 = z2 , the concentration prole
of the common ion is linear and is
given by
2.1.2.4
z1 c1b
I
x
1
z1 z3 ILd,1
(53)
which has been obtained by integration of
Eq. (37) and making use of Eq. (38). The
c3 (x) = c3b
93
distributions of the electric potential and

concentration of the other inactive ion can
be obtained from Eq. (33) as
1
c3 (x)
1
c2 (x)
ln b =
ln b
z3
z2
c3
c2
(54)
Finally, the concentration prole of the
electroactive species is determined from
the LEN assumption, Eq. (35). Figure 4
shows these proles for charge numbers z1 = z2 = 1, z3 = 2, a concentration
ratio c1b /c2b = 0.2, and current densities
I /ILd,1 = 0.5 and 1.0. The most significant characteristic shown by this gure
is that the electric potential drop is very
small, even for the case I /ILd,1 = 1.0. As
a consequence, ion transport takes place
mostly by diffusion and the concentration proles are practically linear. As could
be expected from the ratio c1b /c2b = 0.2,
the concentration changes associated to
current transport are more important in
relative terms for the electroactive species
than for the inactive ones. A detail that can
(hardly) be observed in Fig. 4 is that c1s = 0
f [ b (x)] =
when I /ILd,1 = 1.0, which is due to the

fact that there is a small migrational contribution to the transport of this species. In
other words, the limiting current is slightly
larger than ILd,1 under these conditions.
For the discussion of the currentvoltage
relation, the homovalent case of Eq. (46) is
considered. In a ternary case, this equation
becomes
I =
2z1 F D1 c3b
(1 ez1 f % )
%L =
(57)
Equations (56 and 57) have been represented in Fig. 5 as a function of the electroactive/supporting electrolyte concentration ratio. When the electroactive species
0.02
f [fb f(x)]
0.04
1
f
0
0.0
0.5
1.0
0.00
1.5
x/d
Electric potential and ionic concentration proles
(marked with the respective subscripts 1, 2, 3) in a ternary
system with z1 = z2 = 1, z3 = 2, for a concentration ratio
electroactive to inactive cation c1b /c2b = 0.2, and two values of the
electric current density: I/ILd,1 = 0.5 (continuous line) and 1.0
(dashed line).
Fig. 4
(56)

1/2
2z1 F D1 c3b
c1b
1 1
IL =
c3b
3
2
RT
ln(1 + c1b /c2b )
2z1 F
and
0.06
(55)
while Eqs. (45 and 47) simplify to
ci(x)/cb1
94
3
2.0
1.5
1.0
z1 ffL
IL/ILd, 1
102
100
102
104
cb1/cb2
Dependence of the limiting current density
(continuous line) and the electric potential drop (dashed
line) in the DBL under limiting conditions on the
electroactive/supporting electrolyte concentration ratio.
Fig. 5
is a tracer ion, c1b /c2b 1, %L 0 and

c1b
z1 F D1 c1b
1 + b ILd,1
IL
4c2
(58)
On the contrary, in the absence of
supporting electrolyte, c1b /c2b 1, IL
2ILd,1 and %L diverges; the ratio IL /ILd,1
tends to (1 z1 /z3 ) when c1b /c2b 1 in
systems with other charge numbers, as
can be deduced from Eq. (49).
The currentvoltage curves given by
Eq. (55) have been represented in Fig. 6.
b ,
The initial slope of these curves is eff
which obviously increases with decreasing
ratio c1b /c2b , when the value of ILd,1 is xed

(i.e. c1b is xed). For very low values of this
ratio, % is very small. It is then more
relevant to consider also the variation of
the electrode potential. Figure 7 shows the
same curves as in Fig. 6 though presented
in terms of % c , where
c
1000
50
5
I/ILd, 1
1.5
Currentvoltage curves of a
depleted DBL for different values of the
electroactive/supporting electrolyte
concentration ratio, c1b /c2b , shown on the
curves. % is the electric potential drop
in the DBL.
(59)
is the concentration overpotential. The potential drop % c is the change in electrode potential (relative to bulk solution)
due to the concentration polarization.
2.0
Fig. 6
cs
1
ln 1b
z1 f
c1
1
0.1
1.0
0.5
0.0
z1 ff
95
I/ILd, 1
96
2.0
1000
50
1.5
5
1
1.0
Fig. 7 Currentoverall voltage

drop for different values of the
electroactive/supporting
electrolyte concentration
ratio, c1b /c2b .
0.1
0.5
0.0
z1 f(f hc)
A validity test for the excess supporting

electrolyte assumption (in the homovalent
case) can be obtained by calculating the
migrational ux from Eqs. (37 and 38) as
J1,mig z1 D1 c1 f
z 2 c1
d
J1
= 1
dx
zi2 ci
i
c1
=
J1
2c3
(60)
and comparing it to the diffusional ux

c1
J1,dif = J1 J1,mig = J1 1
2c3
(61)
Equations (60 and 61) show clearly that
the migrational contribution to J1 is
negligible compared to the diffusional one
in as much as c1 is negligible compared
to c3 . Alternatively, the importance of
migration can be estimated from Eq. (58),
since the term c1b /4c2b accounts for the
migrational contribution to IL .
When the supporting electrolyte is electroinactive (and chemically inert), its presence can be totally ignored in the description of transport processes. The absence
of migration and the zero ux condition
for the inactive species implies that these
species have an approximately uniform
concentration throughout the DBL, which
is not affected by the passage of the electric current. It must be stressed, however,
that this statement refers to the concentration gradients in relative terms to the bulk
concentration of the respective species. In
absolute terms, the concentration gradients of the inactive species are similar to
that of the active species, and by no means
the concentration gradient of the active
species can be neglected. In fact, in the
homovalent case, it can be easily shown
that dc1 /dx 2dc3 /dx 2dc2 /dx when
migration is negligible. A straightforward
implication of these comments is that
the LEN assumption cannot be used in
conjunction with the excess supporting
electrolyte assumption.
Weak Binary Electrolyte
Consider an electrolyte A1 B2 dissociating into 1 ions Az1 of charge number z1
and 2 ions B z2 of charge number z2
2.1.2.5
z1
z2
A1 B2
1 A + 2 B
(62)
where z1 1 + z2 2 = 0. Denote by c12 =

c12,T and c12,u = (1 )c12,T the concentrations of dissociated and undissociated electrolyte fractions, respectively,
being the degree of dissociation and
c12,T = c12 + c12,u the total electrolyte
concentration. When the dissociation Reaction (62) is fast compared with the transport process, it can be assumed that the
dissociated and undissociated electrolyte
fractions are in equilibrium and the mass
action law
K=
1 2 12
c11 c22
12 1 1 2
= c12,T
c12,u
1
dc12,u
dx
(64)
vary with position. The same applies to the

ux density of the dissociated electrolyte,
J12 [35]. The total electrolyte ux
dc12,u
dx
dc12
dc12,T
D12
= D12,T
(65)
dx
dx
J12,T = J12,u + J12 = D12,u
however, is constant throughout the system as required by the continuity equation,

Eq. (11). In spite of this fact, the total electrolyte concentration c12,T does not vary
linearly with position because the effective
diffusion coefcient of the electrolyte
D12,T
12 (1 )D12,u + D12
12 (1 ) +
J12,T =
(63)
allows then to determine as a function

of c12,T . In deriving Eq. (63), the following
relations have been used, c1 = 1 c12 , c2 =
2 c12 , and 12 1 + 2 .
Because of the electrolyte dissociation,
the ux densities Ji of the ions and that of
the undissociated electrolyte
J12,u = D12,u
If only species 1 is electroactive, the

boundary conditions at the electrode
s
surface are J12,u
= 0, J2s = 0, and J1s =
I /z1 F , and therefore the ux density
J12,T can be evaluated, from Eq. (21)
and J2s = 0, as
(66)
is a function of , and is a function

of c12,T . Exception is made for the
special cases of very weak electrolyte (
0, K 0, D12,T D12,u ), very strong
electrolyte ( 1, K , D12,T
D12 ), and D12,u = D12 , when D12,T is a
constant and c12,T varies linearly.
t2 I
z2 2 F
(67)
The concentration prole of the electrolyte

c12,T could then be determined from
Eqs. (63, 65, and 67), and the boundary
b
value c12,T
. Note that the transport numbers are constant in a binary system.
For the determination of the currentvoltage characteristics of this system,
the integration of Eq. (65) over the DBL
J12,T =
t2 I
z2 2 F
= D12,u %c12,u + D12 %c12 (68)

can be conveniently rewritten in the form

s
s
+ D12 c12
D12,u c12,u
I = IL,T 1
b
b
D12,u c12,u
+ D12 c12
(69)
where
z2 2 F
b
b
+ D12 c12
)
(D12,u c12,u
t2
(70)
is the limiting current density dened
from the condition that the electrolyte
concentration (both dissociated and undissociated) becomes zero at the electrode
surface when I IL,T . This limiting current density reduces to
IL,T =
IL,u
b
z2 2 F D12,u c12,T
(71)
t2
and
IL
b
z2 2 F D12 c12,T
t2
b
z1 12 F D1 c12,T
(72)
97
in the limits of very weak ( 0) and very

strong electrolyte ( 1), respectively. Indeed, Eq. (70) can be easily transformed to

D12,u
(73)
(1 b ) + b
IL,T = IL
D12
constant K when D12,u = D12 , for in this

case, IL,T = IL,u = IL .
The paradoxical results mentioned above
can be explained in terms of the electric potential drop across the DBL (see Appendix
A for its derivation)
where b is the degree of dissociation at

x = .
Figure 8 shows the dependence of the
limiting current on the dissociation degree for the cases of 1 : 1 and 2 : 1
electrolytes and different values of the ratio
D12,u /D12 . At very low electrolyte concentration or very high dissociation constant,
the electrolyte is completely dissociated
and IL,T = IL . At very high electrolyte
concentration or very low dissociation
rate constant, the electrolyte is practically
undissociated and IL,T = IL,u . The ratio IL,T /IL is then equal to D12,u /D12 .
Surprisingly, in this limit of negligible
dissociation when there are no ions in
solution, the system conducts electric current and, moreover, the limiting current
can be greater than for a fully dissociated
electrolyte if D12,u > D12 . Note also from
Fig. 8 that the limiting current density
becomes independent of the dissociation
cb
RT
RT
12 D12,u
ln 12
s F z ( 1)D
z2 F
c12
2 12
2

b
s
c12,u
c12,u
(74)
s
b
c12
c12
% =
This potential drop increases when K

decreases and there are less ions to
conduct the electric current. In particular,
%dif vanishes and %ohm diverges when
K 0, so that the sentence in the
previous paragraph should have been
written as follows: the system is able to
conduct an electric current in the limiting
case 0, provided that an innite
potential difference is applied across the
depleted DBL.
The currentvoltage curves of this
system can be obtained by combining
Eqs. (69 and 74). Figure 9 shows these
curves for a 1 : 1 electrolyte, with an
electroactive cation, and different values of
(cb12)2/K
102
100
102
104
2
IL, T /IL
98
Limiting current density in a

depleted DBL containing a weak
electrolyte solution relative to the value
corresponding to complete dissociation.
The lower abcisa scale and the
continuous lines correspond to a 1 : 1
electrolyte. The upper abcisa and the
dashed lines correspond to a 2 : 1
electrolyte. The ratio D12,u /D12 has been
given the values 0.5, 1, and 2, as shown
on the curves.
Fig. 8
0.5
0
102
100
102
cb12/K
104

depleted DBL containing a weak 1 : 1
electrolyte for the values of the ratio
b
c12,T
/K, which are shown on the curves.
Both the ratio D12,u /D12 and the
transport number t1 have been set equal
to 0.5.
Fig. 9
0.01
1.0
IL, T /IL
0.1
1
10
0.5
100
1000
0.0
10
ff
the bulk electrolyte concentration relative

b
to the dissociation constant, c12,T
/K.
When the dissociation degree is close
to one, that is at low values of the
b
/K, the curves are practically
ratio c12,T
identical to that in Fig. 2. However,
b
when the ratio c12,T
/K is high and
the electrolyte is mostly undissociated,
the electric conduction becomes difcult.
b
/K =
Thus, for instance, in the case c12,T
b
100 where = 0.031 and IL,T /IL =
0.516, the electric potential drop in the
DBL, %, must be of the order of 15 times
RT/F for the current density to be close to
this limiting value. No need to mention
b
/K
that large values of the ratio c12,T
also imply lower concentration of the
electroactive species close to the electrode
and, therefore, larger electrode|solution
potential drops (i.e. Nernstian potential
drops in the case of reversible electrode
reactions). However, the currentvoltage
curves do not change qualitatively when
the electrode potential is taken into
account in the voltage drop.
Note, nally, that the boundary condis
= 0 requires (except in the trivial
tion J12,u
case I = 0, which lacks interest) that the
dissociation degree is unity in the vicinity
of the electrode, independent of the values
of I and K. This is equivalent to stating
s
that the value of s is independent of c12,T
,
which is inconsistent with Eq. (63). It is
then concluded that Eq. (63) is not valid in
the vicinity of the electrode. The study of
the limiting current without the assumption that the dissociation reaction is fast
compared with the electrodiffusion process has been carried out by Kharkats and
Sokirko [36, 37]. The validity of the dissociation equilibrium assumption and the
drastic changes in concentrations that occur when I IL have been considered by
Vorotyntsev [38].
2.1.3
Steady State Currentoverpotential Curves

in the Presence of Supporting Electrolyte
The use of an excess amount of inactive

electrolyte makes negligible the migrational contribution to mass transport and
therefore simplies the theoretical modeling of experimental observations. In
particular, the absence of potential drop
in the DBL eliminates the coupling of
ionic uxes. The currentvoltage curves
are then presented in terms of the concentration overpotential. Mass transport in the
presence of homogeneous reactions can be
described with relative simplicity in these
systems, as shown in the last section.
99
100
2.1.3.1 Systems with One Electroactive

Species
In the absence of migration, the ux
density of the electroactive species is
cb c1s
I
dc1
D1
= D1 1
J1 =
z1 F
dx
(75)
and the interfacial concentration can be
expressed as

I
s
b
(76)
c1 = c1 1
ILd,1
where ILd,1 is the limiting diffusion
current density dened by Eq (40). Equation (76) can be rewritten in terms of the
concentration overpotential, Eq. (32), as
I = ILd,1 (1 ez1 f c )
(77)
This equation is valid both for a depleted

DBL, in which I /ILd,1 > 0 and z1 c < 0,
and for a concentrated DBL, in which
I /ILd,1 < 0 and z1 c > 0.
Systems with Several Electroactive
Species
When several species are involved in the
electrode reaction described by Eq. (23),
the reaction rate is determined by the
reactant (ir < 0) with slower diffusion
rate. Thus, the limiting current exhibited
by the system is the smallest of the limiting
diffusion current densities [26]
The concentration overpotential in Eq. (32)

then takes the form

1 r
I
c =
i ln 1
(80)
nf
ILd,i
i
Consider, for example, the anodic oxidation of iron(II)

Fe3+ (aq) Fe2+ (aq) + e = 0
(81)
in a situation in which the bulk (aqueous)

solution is equimolar in ferrous and ferric
ions. Since the ferrous ion is consumed
at the electrode, the limiting current
exhibited by the system is the diffusion
limiting current of this ion and the
currentoverpotential curve is
c =
=
1 I /ILd,Fe3+
1
ln
f
1 I /ILd,Fe2+
1 + (DFe2+ /DFe3+ )I /ILd,Fe2+
1
ln
f
1 I /ILd,Fe2+
(82)
which has been represented in Fig. 10.
2.1.3.2
ILd,i
nF Di cib
ir
(78)
of the reactants. Note that I can be either

positive or negative, but it is always of the
same sign as ILd,i for the reactants.
From Eqs. (25 and 78), the interfacial
concentration of species i is given by

Ji
I
cis = cib +
= cib 1
(79)
Di
ILd,i
Diffusion-reaction Processes
The general strategy for solving transport
problems in which the electrode reaction,
Eq. (23), is coupled to a homogeneous
reaction in the DBL

i Cizi = 0
(83)
2.1.3.3
is outlined in this section. The stoichiometric numbers are dened to be positive

for products and negative for reactants.
Chemical equilibrium is assumed for both
reactions, which is justied from the fact
that chemical kinetics is usually much
faster than diffusion processes.
The homogeneous reaction is considered to be reversible.
The chemical
equilibrium
assumption,
i 0 or
i i

K i cii , applied to transport problems

Current density versus
concentration overpotential for the
anodic oxidation of iron(II) from an
equimolar solution of ferrous and ferric
ions. The diffusion coefcients
DFe3+ = 0.604 105 cm2 s1 and
DFe2+ = 0.719 105 cm2 s1 [39]
have been used.
Fig. 10
I/ILd, Fe2+
1.0
0.5
0.0
0.0
2.0
4.0
6.0
fhc
means that the forward and backward reaction rates are much larger than the
mass transport by diffusion, and hence
their difference can be neglected in a
rst approximation. Actually, there is a
small difference between the forward and
backward reaction rates, the net reaction rate
r
1 dJi
ri
=
i
i dx
(84)
which is responsible for the spatial

changes of the ux densities of the
reactive species. In the description of
transport processes, this equation can be
used to set up relations among the ux
density gradients of the reactive species
and to evaluate the reaction rate. If
the reaction rate constants are known,
the validity of the fast chemical equilibrium assumption should also be checked
a posteriori.
The continuity equation under steady
state conditions, Eq. (84), states that the
ux densities of the reactive species change
with position because they are either consumed or generated in the homogeneous
reaction. The reaction involves the transformation of chemical species by the
exchange of atoms or groups of atoms
among them, but the total amount of a
given chemical element is not modied.

It is then correct to state that the sum
of the ux densities, multiplied by a
convenient stoichiometric coefcient, of
all species that contain a given chemical element involved in Reaction (83) is
independent of position. By consideration of all the chemical elements, a large
number of equations are obtained. The
elimination of those that are simply linear combinations of the others leads to
a much smaller number of independent
equations. Alternatively, those groups of
atoms that are not modied by the reaction can be considered as a unique
entity. This is the so-called formalism
of constituents, which leads faster to the
same number of independent equations.
Consider, for instance, that the chemical
reaction involves several species with a
cyanide group. The reaction cannot make
the cyanide group disappear but it can
simply bind it to one or another chemical
species. Thus, the ux of cyanide constituent (dened as the group of chemical
species that contain cyanide) is not affected
by the chemical reaction and is constant
throughout the DBL (under steady state
conditions).
These conservation equations (one for
each constituent) are generically written in
101
102
the form

dJj
dJi
j,i
=
=0
dx
dx
(85)
where the sum extends over all species i

that contains the constituent (or chemical
element) j , and j,i is the stoichiometric
number of constituent j in species i. Taking into account that the ux densities of
the different species at the electrode surface are given by Eq. (24), and neglecting
migration owing to the presence of excess supporting electrolyte, Eq. (85) can be
written as
j,i Ji =
j,i Di
dci
dx
I
=
j,i ir
nF
(86)
Note that the equation for the conduction

current density, Eq. (12), is implicit in the
system of Eqs. (86).
As an example, consider the dissolution
of a cadmium metal electrode
Cd2+ (aq) Cd(metal) + 2e = 0 (87)
in a solution containing iodide ion,
which can complex to cadmium to form
tetraiodocadmiate
[CdI4 ]2 (aq) Cd2+ (aq) 4I (aq) = 0
(88)
Denoting the species Cd2+ , I , and
[CdI4 ]2 by subscripts 1 to 3, respectively,
and considering cadmium Cd and iodine
I as two constituents, Eq. (86) takes
the form
dc1
dc3
I
+ D3
=
dx
dx
2F
dc2
dc3
D2
+ 4D3
=0
dx
dx
D1
(89)
(90)
where 2r = 3r = I,1 = Cd,2 = 0, 1r =

Cd,1 = Cd,3 = I,2 = 1, n = 2, and I,3 =
4. These two equations are to be solved
together with the chemical equilibrium
condition
c3
K
(91)
c1 c24
using the bulk concentration values as boundary conditions. The diffusion coefcients and complexation
equilibrium constants are D1 = 0.7
105 cm2 s1 , D2 = 2.0 105 cm2 s1 ,
D3 = 0.5 105 cm2 s1 , and K = 2
106 M4 [39, 40].
A simple way to solve this equation
system is to consider c3 as the independent variable and evaluate c2 and c1 from
Eqs. (90 and 91), respectively. Finally, integration of Eq. (89) yields the position
variable. Figure 11 shows the concentration proles thus obtained for the cases
of anodic dissolution (I > 0) and cathodic
discharge of cadmium (I < 0). In the rst
case, the cadmium and complex ion concentration gradients are negative (i.e. surface concentration larger than bulk value)
and the iodide concentration gradient is
positive. The DBL is then concentrated in
the cadmium constituent. In the second
case, the cadmium ion is consumed by
the electrode reaction and the DBL is depleted in the cadmium constituent. The
reaction rate evaluated from Eq. (84) has
also been represented. It is thus evidenced
that the rate of the complexation reaction
is large in the case of the anodic dissolution and that the reaction front shifts
towards the bulk solution with increasing
current. On the contrary, the reaction rate
is negligible in the case of the cathodic discharge, in which the concentration proles
are practically linear. The electric current
values have been presented relative to the
limiting diffusion current density of cadmium ion, ILd,1 = 2F D1 c1b /. It can be
ci(x)
[mM]
50
1.0
I
[CdI4]2
104
[CdI4]2
102
Cd2+
0.5
ci(x)/cbi
100
Cd2+
0.0
I
ci(x)
[mM]
100
[CdI4]2
2 104
2 104
50
0
0.0
0.5
1
104
102
Cd2+
1.0
x/d
0.0
0.5
1.0
r
[mol L1 s1]
0
1.5
x/d
Concentration proles in a cadmium iodide solution adjacent to a

cadmium metal electrode and the rate of formation of tetraiodocadmiate
ion. The bulk concentrations are c1b = 5 M, c2b = 100 mM, c3b = 1 mM, and
the three values of the electric current density have been considered:
I/ILd,1 = 102 (cathodic cadmium discharge), 104 and 2 104 (anodic
cadmium dissolution).
Fig. 11
deduced from Eq. (89) that the limiting

current density (corresponding to c1s =
c3s = 0) is IL = 2F [D1 c1b + D3 c3b ]/. For
the bulk concentrations considered in
Fig. 11, c1b = 5 M, c2b = 100 mM, c3b =
1 mM, and a DBL thickness = 103 cm,
these currents are ILd,1 = 0.67 and IL =
97.2 mA cm2 .
2.1.4
Steady State Currentvoltage Curves

of Systems with Several Active Species
There are electrochemical processes in

which little or no supporting electrolyte
is used. This is the case, for instance,
of the study of overlimiting currents [34,
41, 42], and of microelectrode [4345], and
different voltammetric techniques [17, 18,
4648], in which the absence of supporting
electrolyte increases the sensitivity of the
detection of the redox species. In this
section, a general procedure for solving the
electrodiffusion equations is presented.
The procedure is based on determining
rst the electric potential distribution and

then the currentvoltage curves or the
ionic concentration distributions in the
DBL. The generality of the method, and
the inherent complexity of the nonlinear
migration term in the ux equations
could make it not very attractive for the
nonexpert, and therefore simpler solutions
based on the Goldman constant eld
(GCF) assumption are also presented. For
the sake of simplicity, no homogeneous
reactions are considered.
Kramers Integration Method
The mathematical difculty of the electrodiffusion equations, Eq. (7), arises from
the nonlinear migration term, which
couples the NernstPlanck equations of
the different ionic species. By assuming
that the electric eld is constant (GCF
assumption), this difculty is removed
and analytical solutions are easy to nd.
Unfortunately, the validity conditions for
this approximation are not always satised
in electrochemical systems [12, 49], and it
2.1.4.1
103
104
is then needed either to derive approximate expressions for the electric potential
distribution in some other way [50] or
to calculate the exact distribution as described below. In any case, if the electric
eld distribution E(x) or, equivalently, the
electric potential distribution
Edx
is considered to be known, the transport

equations can be formally integrated by
Kramers transformation [51]. This consists of multiplying both terms of Eq. (7)
by ezi f and integrating from x to to
obtain [8, 52, 53]

Ji

b
ci (x) = cib +
ezi f [(x ) ] dx
Di x
(x)]
(93)
The surface concentration cis is obtained

by setting x = 0 in this equation.
When the GCF approximation (x) =
b + ( x)E is valid, the integral in
Eq. (93) can be evaluated and this equation becomes

Ji
ci (x) = cib
ezi f E(x)
zi Di f E
+
Ji
zi Di f E
(94)
The Equation for the Electric Field

under the LEN Assumption
Nikonenko and Urtenov [54, 55] have
presented an algorithm for the solution of the steady state transport equations in multicomponent systems, which
is based on calculating rst the electric eld E and then using Eqs. (92
and 93). Their method is outlined here
for the particular case when the LEN
assumption is used. Unlike Schlogls
2.1.4.2
N
zik ci ,
k = 0, 2, 3, . . . , N (95)
i=1
(92)
Sk
(x) = b +
ezi f [
method, that uses the concept of valency classes [32], these authors proposed
to rewrite the NernstPlanck equation
system for N species in terms of the
magnitudes
Gk
N
zik Ji /Di ,
i=1
k = 0, 1, 2, . . . , N 1
(96)
instead of the N concentrations ci and ux

densities Ji . The LEN assumption
in this
notation takes the form S1 i zi ci = 0,
and this justies the absence of index
k = 1 in Eq. (95). The aim of the method
is obtaining an equation for the electric
eld in which the unknown concentrations
are eliminated by using their relation to
the ux densities, which are known from
Eq. (24).
Multiplying the NernstPlanck equation, Eq. (7), by 1/Di and summing
over species i, the electric eld is
eliminated
and the concentration sum

S0 i ci is shown to follow the linear prole
S0 (x) = S0b + ( x)G0
(97)
Similarly, multiplying the NernstPlanck

equation by zi /Di and summing over
species i, it is obtained that
G1 = f ES2
(98)
which could be used to evaluate the electric

eld if the relation between S2 and S0 were
known. Note that S2 is equal to double
the ionic strength of the solution, and
the advantages of using it as a variable
in the formulation of transport problems
have also been emphasized by other

authors [17, 18]. Some simple cases are
considered next and the general approach
is presented later.
In the case of a binary solution, S2 =
z1 z2 S0 the electric potential can be
obtained from Eq. (98) as
3 S0 (x)
ln
f
S0b
(x) = b +
where
3
1 G1
z1 z2 G0
(99)
(100)
Moreover, introducing the limiting current density

IL
S0b I
G0
(101)
cis
IL
(1 + zi 3)ILd,i

I 1+zi 3
I zi 3
1
1
1
IL
IL
cib
(104)
which reduces to cis = cib (1 I /IL ) in the
binary case. In deriving Eq. (104), note that
Ji /Di cib = I /ILd,i .
In the case of nonhomovalent multiionic systems, the solution procedure
is necessarily more complicated because
there is no simple relation such as S2 =
z2 S0 , a problem that was rst encountered
and solved in different ways by Schlogl [59]
and Brady and Turner [60]. If the system
contains N different ionic species, the N
NernstPlanck equations for the N ux
densities Ji are replaced by
the steady state currentvoltage curve can

be deduced from Eq. (99) as
cs
3
ln bi
f
ci

3
I
= ln 1
f
IL
dSk
+ f ESk+1 ,
dx
k = 0, 1, 2, . . . , N 1
Gk =
% =
(102)
In a multi-ionic, homovalent solution,

|zi | = z, S2 = z2 S0 and Eq. (99) is again
obtained [5658], though 3 is now dened
as 3 G1 /z2 G0 . Since the integral in
Eq. (93) can be evaluated by using Eqs. (97
and 99) as

zi f ( b )
dx =
(103)
the interfacial concentration is nally
given by
(105)
In order to solve the system of Eqs. (105),

a closure relation between SN and other
sums is needed. Such a relation is
SN = q1 SN 1 q2 SN 2 + q3 SN 3
+ (1)N 1 qN S0
(106)
where
q1 = z1 + z2 + + zN
(107)
q2 = z1 z2 + z1 z3 + + z1 zN + z2 z3 +
+ z2 zN + z3 z4 + + zN 1 zN (108)
S0zi 3 dS0
G0 (S0b )zi 3

1+zi 3
S0b
G0
1+
=
1
G0 (1 + zi 3)
S0b
0
1+
q3 = z1 z2 z3 + + z1 z2 zN + z1 z3 z4 +
+ z1 z3 zN + + zN 2 zN 1 zN (109)
..
.
qk =
+2k
N
+1k N
i1 =1 i2 =i1 +1
N
zi1 zi2 . . . zik
ik =ik1 +1
(110)
105
106
Some particular cases are considered in

the next sections.
Ternary Electrolyte Solutions
In a ternary system (N = 3), Eqs. (97 and
98) must be solved together with
2.1.4.3
dS2
+ f E S3
dx
S 3 = q1 S 2 + q3 S 0
G2 =
(111)
(112)
After some algebra, Eq. (111) can be

transformed into the following equation
for the electric eld
G1 dE
= (G2 q1 G1 )f E q3 (f E)2 S0 (x)
E dx
(113)
which has to be integrated numerically
from x = to x using the boundary
condition Eb = G1 /f S2b .
As an example, consider the case of a
redox system such as FeCl2 + FeCl3 in
the absence of supporting electrolyte in a
potential range in which the only possible
electrode process is the anodic oxidation
of ferrous ion
Fe3+ (aq) Fe2+ (aq) + e = 0
(114)
Denoting the ions Fe2+ , Fe3+ , and Cl

by subscripts 1, 2, and 3, respectively, the
ion uxes are J1 = J2 = I /F , and the
parameters Gk take the values

1
1
I
I
= 0.1904
G0 =
D2
D1 F
F D1
(115)

z2
z1 I
I
G1 =
= 1.5712
D2
D1 F
F D1
(116)

z22
z12 I
I
= 6.7136
G2 =
D2
D1 F
F D1
(117)
where D1 = 0.719 105 cm2 s1 and
D2 = 0.604 105 cm2 s1 [39]. Equation (113) can be rewritten in terms of the
dimensionless electric eld e f E as

x
de
2
= e 0.2729 + e 0.7271 1
d(x/)

4cb ILd,1
(118)
+ 11.456 1 + 2b
I
3c1
where ILd,1 = F D1 c1b / is the limiting
diffusion current density of ferrous ion.
Equation (118) can be integrated by
standard numerical methods, such as
fourth-order RungeKutta method [61],
starting from position x = , in which
the boundary condition eb = G1 /S2b is
applied as

x
0.2619
I
(119)
=1 =
e
ILd,1 1 + 2c2b /c1b

The potential drop % (and its distribution, if needed) can then be calculated by
numerical integration of the electric eld
using, for example, a fourth-order NewtonCotes method [61]. This procedure
yields the currentvoltage curves shown
in Fig. 12 in the case of a DBL depleted
of ferrous ions, that is, for I /ILd,1 > 0
and % < 0. These curves end when they
meet the dashed line that corresponds to
the limiting current density, that is, to
vanishing surface concentration of ferrous
ion. The limiting current density is a bit
smaller than the limiting diffusion current
density ILd,1 because the ferrous cation is
driven towards the anode by its concentration gradient against an electric eld that
pushes it away from the electrode. The
strength of this electric eld is reduced
when the solution conductivity increases,
that is, when the bulk concentration ratio
c2b /c1b increases for a xed value of c1b , and
the limiting current density tends to ILd,1
when this occurs.
The ionic concentration distributions
in the DBL are represented in Fig. 13.

ternary system FeCl2 + FeCl3 in the
absence of supporting electrolyte for
different values of the bulk
concentration ratio c2b /c1b shown near
the curves. The dashed line represents
IL /ILd,1 obtained from the (exact)
numerical solution, and the dotted line
represents IL /ILd,1 obtained from the
GCF assumption.
Fig. 12
1.0
I/ILd, 1
10
0.5
0.1
0.5
0.0
0.00
0.05
0.10
0.15
ff
The ferrous ion concentration is obtained from Eq. (93) and using the relation J1 /D1 c1b = I /ILd,1 . The chloride
ion follows Boltzmanns equation c3 =
b
c3b ef ( ) , and the ferric ion concentration is obtained from the LEN assumption
c2 = (c3 2c1 )/3.
Figures 12 and 13 evidence that the
potential drop in the DBL is small and
is distributed quite uniformly. It is then
expected that the GCF assumption, E =
%/, provides reasonable estimates.
Indeed, by setting c1s = 0 in Eq. (94),
the limiting current density can be
estimated as
IL
2f %
=
ILd,1
1 e2f %
(120)
which has been represented in Fig. 12

and agrees very well with the exact
numerical solution.
Quaternary Electrolyte Solutions
Equations (97, 98 and 111), and
2.1.4.4
dS3
+ f E S4
dx
S 4 = q1 S 3 q2 S 2 q4 S 0
G3 =
(121)
(122)
0.09
ci(x)/c1b
0.06
f
0.03
Fe3+
f[f(x) fb]
Cl
Fe2+
0
0.0
0.00
0.5
1.0
1.5
x/d
Electric potential and concentration proles of Fe2+ ion
(i = 1), Fe3+ ion (i = 2), and Cl ion (i = 3) for a bulk
concentration ratio c2b /c1b = 1.0 and the limiting current density
I = IL = 0.9261ILd,1 .
Fig. 13
107
108
lead in this case to the following equation

for the electric eld

G1 d2 E 3G1 dE 2
+ (G2 q1 G1 )
E dx 2
dx
E2
f
dE
(G3 q1 G2 + q2 G1
dx
+ q4 S0 f E)(f E)2 = 0
(123)
which has to be integrated numerically

from x = to x using the boundary
conditions Eb = G1 /f S2b and

S3b
G1
dE
G2 G1 b (124)
=
dx x=
f (S2b )2
S2
Although the mathematical difculty of
this equation system is not very high,
the GCF assumption, with the electric
eld estimated, for example, as E Eb =
G1 /f S2b , is a much simpler alternative to
be considered when the ionic strength is
roughly uniform over the DBL.
2.1.5
Ion Transport under Transient Conditions
Needless to say, transient transport conditions involve a higher level of mathematical complexity to the description of
ion transport in electrochemical systems.
The only case that can be worked out with
ease is that of binary solutions. Ternary
and multi-ionic solutions necessarily involve systems of partial differential equations coupled through the migration term
that require advanced numerical methods.
Given this situation, it is by no means
surprising that most theoretical studies of
transient problems neglect migration and
use the diffusion equation (or Ficks second law) to describe the transport of ionic
species. (The analysis of the role of supporting electrolytes in Sects. 2.1.2.2 and
2.1.2.4 was restricted to steady state conditions, but it is expected that a large excess
of supporting electrolyte makes migration
negligible under transient conditions too.)
This section describes diffusion and migration under transient conditions in some
simple illustrative cases. For the sake
of simplicity, an innite DBL thickness
is considered, which is justied in the absence of stirring. Current
and voltage steps in semi-innite planar geometry (and in the absence of
homogeneous reactions) are described.
Both Laplace transforms and Boltzmanns
change of variables are used. The inuence of migration is discussed analytically
(whenever possible) by using the latter
technique, which is introduced in Appendix B. Other solution techniques of
the diffusion equation [6264], and the
electrodiffusion equations are available in
the literature [13, 6572]. Further information on electrochemical techniques can
be found in Ref. [73].
Ficks Second Law for a Strong
Binary Electrolyte
The diffusionconduction equation,
Eq. (21), expresses the ux density of
an ionic species from a strong binary
electrolyte in terms of the electrolyte
ux J12 and the current density I .
Under transient conditions, both Ji
and J12 are position-dependent, but
I is not. Thus, taking into account
Eq. (21) and the relation ci = i c12 , the
continuity equation, Eq. (11), leads to
Ficks second law
2.1.5.1
c12
2 c12
= D12
t
x 2
(125)
The transient ionic electrodiffusion can

then be described as electrolyte diffusion and solved by standard techniques [6264].
Systems with One Active Species

and Supporting Electrolyte
The transport of the active species in the
presence of excess supporting electrolyte
under transient conditions is described by
the equation
2.1.5.2
c1
2 c1
= D1 2
t
x
(126)
which is obtained from the continuity

equation, Eq. (11), by neglecting the migrational contribution to the ux density J1 .
Although Eq. (126) resembles Ficks second law, it should be observed that Ficks
laws are valid for (neutral) electrolytes. Ion
transport takes place by electrodiffusion
and satises pseudo-Ficks laws only when
migration is negligible.
For a simple electrode process, like
cathodic deposition or anodic dissolution
(without contribution of nucleation), the
initial value is c1 (x, t = 0) = c1b and the
boundary value at the bulk solution is
c1b (t) = c1b . The boundary condition at the
electrode surface can be either Eq. (24), in
the form

c1
I (t)
=
(127)
x x=0
z1 F D1
or the Nernst equation, Eq. (31), for the
case of a single active species
cs (t)
RT
Eeq (t) = E +
ln 1
z1 F
c
Voltage Step in the Presence of

The boundary condition, Eq. (128), for a
voltage step perturbation reads
2.1.5.3
c1s (t) = c1b exp[z1 f %EH (t)]
where H (t) is the step function [H (t) = 0

if t < 0, H (t) = 1 if t > 0], and %E is the
width of the voltage step.
Using Laplace transformation with respect to time, the general solution of
Eq. (126) in the Laplace domain is [62]
c1 (x, s) =
c1b
1/2
+ Aex(s/D1 )
s
+ Bex(s/D1 )
1/2
When the system is perturbed by applying a known function, I (t), Eq. (127) is
used to evaluate the chronopotentiometric response of the system. Analogously,
when the system is perturbed by applying a
known function Eeq (t), Eq. (128) describes
the chronoamperometric response of the
system.
(130)
where c1 (x, s) is the Laplace transformed

concentration and s is the Laplace variable.
Since the concentration must be bounded,
the integration constant A vanishes. The
boundary condition, Eq. (129), gives the
value for B, resulting in
c1 (x, s) =
c1b
1 + (ez1 f %E 1)
s

1/2
ex(s/D1 )
(131)
By inverse Laplace transformation, the

concentration prole is obtained as
c1 (x, t) = c1b [1 + (ez1 f %E 1)erfc( )]
(t > 0)
(128)
(129)
1/2
(132)
where x/(2D1 t 1/2 ) is the Boltzmann

variable (see Appendix B) and erfc stands
for the complementary error function.
The electric current response is obtained
from Eqs. (127 and 132) as

c1
I (t) = z1 F D1
x x=0
1/2
D1
= z1 F c1b (ez1 f %E 1)
(133)
t
109
110
If the potential step width is large enough,

z1 f %E 1, the surface concentration
of the electroactive species is negligible
and Eq. (133) reduces to the Cottrell
equation [73]

I (t) =
z1 F c1b
D1
t
1/2
(134)
Current Step in the Presence of

Consider that the function I (t) in Eq. (127)
is a current step I (t) = I0 H (t), where I0
is the constant current applied from t = 0.
Equation (130) reduces now to
2.1.5.4
c1b
I0
1/2
s 3/2 ex(s/D1 )
+
1/2
s
z1 F D1
(135)
whose inverse transform provides the
concentration distribution
c1 (x, s) =
2I0 t 1/2
c1 (x, t) = c1b +
1/2
z1 F D1
[ 1/2 e erfc( )]
2
(136)
The potential response is given by Eq. (128)

with the surface concentration determined
from Eq. (136) at = 0, that is,
Eeq (t > 0) = E +

RT
z1 F
cb
2I0
ln 1 +
c
z1 F c
t
D1
1/2
(137)
In cathodic processes (I0 < 0), the surface

concentration becomes zero at the transition time

2
z1 F c1b
(138)
= D1
2I0
This relationship is called the Sand equation [73].
Voltage Step in the Absence of

Although it is important for practical
purposes to attempt the treatment of
systems which are not stripped of so
many features that they no longer bear
resemblance to reality, the simultaneous
treatment of a reasonable fraction of
these features makes the mathematical
problem so complex that there is a danger
of obscuring the real problem [74]. It is
then suggested to simplify the present
analysis of a large voltage step in a
ternary solution (where the common ion
is inactive) by making equal all diffusion
coefcients, Di = D, for i = 1,2,3, and
hence eliminating the diffusion potential.
This allows for a better understanding of
the effect of migration on the transient
response of the system.
The treatment in Sect 2.1.4.3 is followed
and magnitudes Sk and Gk are used instead of ionic concentrations and ux
densities, respectively. However, the relations between c1 , c2 , c3 , and S0 , S2 , are
still needed because the Nernst equation,
Eq. (31), involves ionic species.
The basic equations describing this
problem are the continuity equations,
Eq. (11), which owing to the simplifying
assumption Di = D reads as
2.1.5.5
Sk
Gk
= D
t
x
(139)
and Eq. (105) for k = 0, 1, and 2,

S0 (x, t)
x
(x, t)
G1 = f
S2 (x, t)
x
S2 (x, t)
G2 =
+ q1 G1
x
S0 (x, t)
+ q3 G1
S2 (x, t)
G0 =
(140)
(141)
(142)
where the closure relation, Eq. (112),

and Eq. (141) have been used to obtain
Eq. (142). Equation (141) is Ohms law,
Eq. (13), for this system.
At the electrode surface, Eq. (24) requires that
Gs0 (t) =
I (t) r
i
nF D
(143)
Gs2 (t) =
I (t) 2 r
I (t)
(z1 + z2 )
zi i =
nF D
FD
i
(144)
which are used as boundary conditions
in chronopotentiometric techniques or to
evaluate the system response in chronoamperometric techniques. In the second
equality of Eq. (144), the common ion
is considered to be inactive, 3r = 0. The
parameter G1 = I (t)/FD is only a function of t.
Combining Eq. (139) for k = 0 and
Eq. (140), the partial differential equation
for G0 is
G0
2 S0
S0
= D
=D 2
t
x
x
(145)
which is subject to the initial and boundary

conditions S0 (x, t = 0) = S0b and S0 (x
, t) = S0b , in addition to Eq. (143). The
DBL thickness is considered to be innite
due to the absence of stirring. The solution
of Eq. (145) can be obtained by following
the method explained in Sect. 2.1.5.3 as
S0 (x, t) = S0 ( ) = S0b + (S0s S0b )erfc( )
(146)
Combining Eq. (139) for k = 2 and
Eq. (142), the partial differential equation
for S2 is
S2
2 S2 (x, t)
=D
t
x 2

S0 (x, t)
q3 DG1
(147)
x S2 (x, t)
The difculty introduced by the second

term in the right-hand side of Eq. (147)
makes it very convenient to introduce a
further simplifying assumption.
When the number of electroactive
species is conserved by the electrode reac
tion, i ir = 0, it is satised that Gs0 = 0,
and the only solution to Eq. (145) is the
trivial result S0 = S0b . This is the principle of unchanging total concentration
established by Oldham and Feldberg [17].
Equation (147) then reduces to
q3 I S0b 1 S2
S2
2 S2
=D 2 +
t
x
F S22 x
(148)
The initial and boundary conditions for

this equation are Eq. (31), S2 (x, t = 0) =
S2b and S2 (x ) = S2b .
In the case of a large voltage step,
f |%E| 1, Eq. (31) is approximated by
the condition of zero surface concentration of the reactant, c1s (t) = 0 (or S2s =
z2 z3 S0b ), and both boundary conditions
for S2 turn out to be independent of time.
This implies that S2 is a function of x and t
1/2
through variable x/(2D1 t 1/2 ). Moreover, since the electric current density is
obtained from Eqs. (142) and (144) as

S2
I (t)
=
(z3 z1 )

x x=0
FD
(149)
it is concluded that I (t) is proportional to

t 1/2 .
Equation (148) can then be transformed
into the ordinary differential equation for
S2 ( ) (see Appendix B)
0=
(S b )2 dS2
d2 S2
dS2
+ 2
+ 22
(150)
2
d
d
S2 d
Parameter in Eq. (150) is related to the

current density by
111
Current density in a voltage
step of large width relative to that in
absence of migration, given by Cottrell
Eq. (152), as a function of the bulk
concentration ratio product/reactant.
The charge numbers z1 , z2 , z3 are
shown close to the curves.
Fig. 14
4, 1, 1
2.0
4, 1, 2
I/ICottrell
112
1.5
Reduction
2, 1, 1
3, 2, 1
1.0
2, 3, 1
1, 2, 1
1, 4, 1
0.5
3
Oxidation
Log10(cb2/cb1)
I =
=
F (S2b )2
2q3 S0b
D
t
1/2
(z3 z1 )(S2b )2 1/2

2q3 S0b (S2b + z2 z3 S0b )
ICottrell (151)
where

ICottrell = (z2 z1 )F c1b
D
t
1/2
(152)
is the classical result obtained by Cottrell

in the absence of migration. The value of
must be obtained from the numerical solution of Eq. (150) by the shooting method
starting from S2 (0) = S2s = z2 z3 S0b and

(z1 z3 )(S2b )2
dS2
=
d =0
q3 S0b
(153)
and modifying it by iteration until

S2 () = S2b is satised.
Myland and Oldham [18] have compiled
the results for the case in which the
product of the electrode reaction is absent
in bulk solution, c2b = 0. Figure 14 shows
the current density as a function of the
bulk product/reactant concentration ratio
for several values of the charge numbers
in the case of cationic redox couples. The
current is then larger than in the absence
of migration in the case of reductions and
smaller in the case of oxidations. It is

observed that, for a given stoichiometric
number of the electron n = z2 z1 in the
electrode reaction, increasing the charge
numbers of either the inactive anion
or the active cations reduces the effect
of migration. Moreover, the effect of
migration is larger when |n| increases.
The results of these numerical simulations can be understood by the following
intuitive arguments. Consider, for example, the oxidation of a cationic species
[I > 0, n > 0, z1 > 0, z2 = (n + z1 ) > 0,
z3 < 0]. According to Ohms law, the electric potential gradient in solution pushes
the cations away from the anode, so that
their gradients are smaller (in absolute
value) than in absence of migration. Since
the voltage step makes the surface concentration of the reduced species zero, c1s = 0,
the ux density of this species at the electrode surface is only due to diffusion,

I
dc1
=
(154)
dx x=0
(z1 z2 )F D
and a smaller concentration gradient implies then smaller current density. Contrarily, in the case of reduction of cations,
migration pushes the cations towards the
cathode, so that the concentration gradients are steeper. Since Eq. (154) still
applies, this implies larger current densities than in absence of migration.

Consider, nally, the oxidation of an anion [I > 0, n > 0, z1 < 0, z2 = (n + z1 ) <
0, z3 > 0]. The electric potential gradient
in solution pushes the anions towards the
anode, so that the concentration gradients
are steeper and the current density is larger
than in the absence of migration. The opposite effect is observed in the reductions
of anions.
2.1.6
Deviations from Local Electroneutrality
This last section is devoted to the space

charge density and its variation with time
and position. The dielectric permittivity
of the solution is assumed to be constant, which is a good approximation
for moderate concentrations and electric
elds [7577]. It is further assumed that
the relaxation of the dielectric (i.e. the reorientation of water molecules) is much
faster than the redistribution of the space
charge density [78].
Space Charge Density
The Poisson equation, Eq. (8), identies
the space charge
density in an electrolyte
solution, e = F i zi ci , as the source of
the electric eld. Thus, whenever the electric eld changes with position, there is a
space charge distribution over the system.
In general, e is a function of both
position and time. Its time derivative is
2.1.6.1
e
3
=
t
tx 2

Id
=
=
(155)
x t
x
x
where Id is the displacement current,
that is, the time derivative of the electric displacement D = E. Alternatively,
e /t can be evaluated from the continuity equation, Eq. (11) with ri = 0, as

ci
e
zi
=F
t
t
i
= F
zi
Ji
I
=
x
x
(156)
where I is the conduction electric current density. Equation (156) represents the
conservation of electric charge and its combination with Eq. (155) leads to
I
Id
IT
=
+
=0
x
x
x
(157)
which states that the total electric current

density IT that ows through an electrochemical cell is independent of position,
and is made up of two contributions, the
conduction and the displacement current
densities [6669, 79]
2
tx
i
(158)
The current density IT is equal to that
due to the electrons owing through the
external circuit connected to the electrodes.
In the absence of concentration gradients, no space charge density can exist in
the solution under steady state conditions.
Moreover, if by any means some electrical
charge density is generated at any position
within the system, the time required for
this charge to disappear (in fact, to migrate
to the system boundaries) is of the order
of nanoseconds. This time is the so-called
electrical relaxation time of the system, e ,
and can be obtained from Eqs. (8, 13, and
156) as
IT = I + Id = F
zi Ji
e
I
2
=
= 2
t
x
x
e
e
=
(159)
113
114
so that e / 109 s for a 100 mM

aqueous solution.
Concentration gradients, however, are
present in most situations of electrochemical interest, and the time evolution of e
is then determined by the time variation
of the conditions imposed on the system,
which always involve times much larger
than e .
Usually, space charge regions exist
close to the system boundaries, such
as the EDL at a planar electrode. In
the presence of concentration gradients,
Eq. (159) becomes

ohm
x

dif
=
(160)
x
x
x
x
=
t
x
which is much smaller, in magnitude,

than e /e . Under equilibrium conditions,
I = 0, the two terms on the right-hand side
cancel out. In the presence of conduction
current, these terms are very similar to
each other and much larger in magnitude
than e /t. An order of magnitude analysis of these two terms allows us to estimate
the characteristic length for the space
charge distribution, LD . Because of the
large electric elds existing in space charge
regions, the ions have a strong tendency
to migrate. If the ions were indeed migrating, the space charge distribution would
disappear, so that the necessary condition
for an equilibrium space charge distribution is that this tendency to migrate has
to be compensated by a strong tendency
to diffuse in the opposite direction [68].
The diffusional ux associated to concentration changes is of the order of Dc/LD ,
where c is a typical ion concentration. The
magnitude of the charge density can be
estimated as e F c, and the Poisson
equation, Eq. (8) tells that the eld associated to this charge density over a region
of thickness LD is of the order of e LD /.
The migrational ux density associated to
this eld is of the order of F Dce LD /RT .
The condition for the existence of an equilibrium space charge density is Dc/LD
F Dce LD /RT , which yields the expression of the Debye length

LD
RT
Fe
1/2
RT
F 2c
1/2
(161)
Space charge distributions are then expected to exist over regions of thickness
LD , which is of the order of 107 cm
for a 100 mM aqueous solution. A practical consequence of this comment is that
the behavior of electrochemical systems
comprising microgeometries is affected by
space charge layers [8082].
Note, nally, that ionic motions associated to space charge redistribution involve
distances of the order of LD , which suggests that e can be interpreted as the time
e L2D /D required for the ions to diffuse
over LD [79, 8385].
Deviations from Local
Electroneutrality
Planck suggested that the electric potential calculated from the LEN assumption,
(0) , could be used to check the validity
of this assumption [27, 31]. When this approximate electric potential is introduced
in the Poisson equation, the space charge
density obtained
2.1.6.2
E(0)
2 (0)
=
x
x 2
(162)
provides an idea of the deviations from

LEN, which would be required to create
the distribution (0) (x, t). To make clear
the difference between the true e in the
Poisson equation, Eq. (8), and r dened
by Eq. (162), the latter is often named

residual space charge density [13, 86].
Consider, for instance, the steady state
electric eld in a binary solution under
current ow conditions with species 2
being inactive. From Sect. 2.1.4.2, E(0) =
S2 /f G1 = z2 S0 /f G0 and the residual
space charge density required to establish
this approximate electric eld is

b 2
c2
RT
I 2
r =
z2 F IL
c2 (x)
2

RT
x
IL
=
(163)
+ 1
z2 F 2 I
where Eqs. (97 and 101) have been used;

note also that G1 = z1 G0 = I /F D1 . This
shows that signicant deviations from the
LEN only occur at the limiting current
and close to the electrode surface x = 0,
in which the ionic concentrations tend to
zero and r diverges [34, 42, 87].
The actual space charge density in this
binary solution can be obtained from the
(exact) electric eld E by using the same
formalism and the Poisson equation
dE
e
F S1
=
=
dx
(164)
First, integrating Eq. (105) for k = 0

G0 =
dS0
f E S1
dx
(165)
from x to , and using Eq. (164), Eq. (97)

is modied to
S0 (x) = S0b + ( x)G0 +
E2
2RT
(166)
where the electric eld in bulk solution

has been neglected. Then, substituting
Eqs. (164 and 166) into Eq. (105) for k = 1
G1 =
dS1
dS1
f E S2 =
+ z1 z2 f ES0
dx
dx
(167)
and introducing the limiting current density IL S0b I /(G0 ) dened in Eq. (101),
the equation for E becomes

L2D d2 E
x
I
1
IL
z
+ 1
=
z
1
2
E dx 2
IL f E
I

IL
+ (f ELD )2
(168)
2I
where LD (RT /F 2 S0b )1/2 is the Debye
length.
Analytical studies of Eq. (168) have
used different changes of variables to
write the electric eld in terms of either
Painleve transcendents [68] or Jacobian elliptic functions [88]. Alternatively, asymptotic expansions have also been used [68,
87, 89, 90]. The approximate solution
methods have neglected different terms
of Eq. (168). Thus, while Urtenov and
Nikonenko [55, 91, 92] have considered
that the space charge density is quasiuniform, dS1 /dx 0, Bass [93] assumed
that E2 RT S0b and I /IL /LD to express the electric eld as the sum of the
electroneutral electric eld, which can be
derived from Eq. (99), and a modied Gouy
distribution corrected for the presence of
current. In any case, both analytical and
numerical [94, 95] solutions have shown
that in the range of underlimiting currents,
the deviations from LEN are conned to
the EDL and that the latter is not signicantly perturbed by the electric current.
In conclusion, r in Eq. (163) is a
physically meaningful magnitude that is
very similar to the actual e when LEN is
a good approximation. Contrarily, when
deviations from LEN occur, the actual
electric potential surely differs from that
obtained by using the LEN, and then the
actual e is likely to show no resemblance
to r .
Although the equations presented in
this chapter cannot account for this fact,
115
116
overlimiting currents can be observed in

practice [41]. The study of these currents
requires not only to abandon the LEN assumption but also to take into account
the electroconvective motions that originate due to the electrical force e E acting
on the electrolyte solution in space charge
regions [68, 90, 91, 9597].
Validity of the LEN Assumption
MacGillivray [98] was the rst to provide a
mathematical justication for the LEN assumption on the basis of the perturbation
theory. The idea behind such a justication
is that an order of magnitude analysis of
the Poisson equation, Eq. (8), shows that

d2
RT /F
=
Fc
F c dx 2
F c L2

LD 2
=
1
(169)
L
2.1.6.3
where L is a typical length for variation of

the electric potential, which is considered
to be of the order of the size of the electrochemical cell and hence much larger
than the Debye length LD . This suggests
that the solution of the NernstPlanck and
Poisson equations can be found by writing
the concentrations and the electric potential as an asymptotic expansion in terms of
the small parameter (LD /L)2 ,
(0)
ci = ci
(1)
+ ci
(2)
+ 2 ci
(170)
= (0) + (1) + 2 (2) + (171)

Thus, the zeroth-order approximation (in
(0)
) of the Poisson equation is e = 0.
In other words, the LEN assumption is a
consequence of the Poisson equation when
1. Moreover, the rst-order term of
the Poisson equation is
e(1) =
d2 (0)
dx 2
(172)
Since (0) is the electric potential calculated from the transport equations at zero
order in (that is, making use of the LEN
assumption), the residual space charge
density dened in Eq. (162) is identied as
(1)
e . The use of higher-order terms in the
Expansions (170 and 171) is only needed
(1)
when e is of the order of F c [13, 27, 31,
67]. A different validity test for the LEN as(2)
(1)
sumption, namely e e , has been
recently proposed by Feldberg (with a different notation) [99].
Note that it is possible to use either the
full Poisson equation or the terms of its
asymptotic expansion, such as the LEN as(0)
sumption or Eq. (172), but e can never
be interpreted as the full e in the Poisson
equation. This would lead to the wrong
conclusion that the LEN assumption implies a constant electric eld.
Appendix A
The total potential drop across the DBL in

Eq. (74) can be obtained as the sum of the
diffusion and ohmic potential drops. The
former is given by Eq. (19) as

RT t1
t2
%dif =
% ln c12
+
F
z1
z2
(173)
To integrate the ohmic potential gradient,
it is convenient to introduce rst the molar
conductivity of the electrolyte as
F2
= z1 z2
(2 D1 + 1 D2 )
c12
RT
(174)
where the last equality is known as the
NernstEintein relation [3]. Equation (17)
then implies
;12
I
I dx
dohm = dx =
;12 c12
(175)
Eliminating the position variable from

Eqs. (65) and (67) as
z1 1 F
(D12,u dc12,u + D12 dc12 )
t2 I
(176)
and using Eq. (63) to deduce that d lnc12,u=
12 d ln c12 , Eq. (175) is transformed to
dx =
z1 1 F 1
(D12,u dc12,u
t2 ;12 c12

RT 12 t1 dc12
+ D12 dc12 ) =
F
z1 1 c12

D12,u dc12,u
RT 12 t1
d ln c12
=
z2 D2 c12
F
z1 1

12 D12,u
c12,u
d
(177)
z2 (12 1)D2
c12
dohm =
Straightforward integration over the DBL

and addition of %dif yields Eq. (74).
Appendix B
The perturbation of the concentration eld

introduced by the electrode propagates at
a nite rate determined by the diffusion
coefcients and only affects a region close
to the electrode whose thickness is of the
1/2
order of the diffusion length 2D1 t 1/2 .
Boltzmann showed that the diffusion
equation, Eq. (126), can be reduced to
an ordinary differential equation under
certain conditions if expressed in terms of
1/2
the variable x/(2D1 t 1/2 ) [62, 100].
To obtain the diffusion equation in terms
of the variables ( , t), the elementary concentration variation is rst expressed as

c1
c1
d +
dt (178)
dc1 =
t
t
The partial derivatives of c1 with respect to
x and t (at constant x) are then given by

2
c1
c1
=
2
x t
x t x t t t
2
1
c1
=
(179)
4D1 t 2 t
c1
t

c1
c1
+
t t x
t

c1
c1
=
+
(180)
2t t
t
=
x
and the diffusion equation becomes

c1
2 c1
c1
+ 2
= 4t
2
(181)
When the boundary conditions imposed by

the experiment allow for the separation of
variables and t (linear sweep voltammetry, for instance, does not), a solution of the
form c1 (, t) = c1b + Z( )T (t) transforms
Eq. (181) into

4t dT
dZ
1 d2 Z
=
+ 2
= (182)
Z d 2
d
T dt
where is a constant.
In the chronoamperometric case, the
boundary condition, Eq. (129), does not
introduce any dependence on t, and = 0
in Eq. (182). Then, the solution of this
equation is
Z( ) = c1 (, t) c1b = Aerfc( ) + B
(183)
where erfc( ) is the complementary error
function. The boundary condition c1 (x
, t) = c1b imposes B = 0, and Eq. (132)
is obtained.
In the chronopotentiometric case, the
boundary condition, Eq. (127), becomes

dZ
2I0
c1
=
T
(t)
=
t 1/2

1/2
=0
d =0
z1 F D1
(184)
which makes = 2 in Eq. (182). The solution of the ordinary differential equation
for the function Z( ) are the rst integrals
of the complementary error function [61]
Z( ) = A[ 1/2 e erfc( )] (185)
2
where the boundary condition c1 (x

, t) = c1b has been taken into account.
117
118
position (m)
charge number, dened with sign
Finally, the use of Eq. (184) to determine

A yields Eq. (136).
x
zi
List of important symbols
Greek symbols
degree of dissociation, Eq. (63)
DBL thickness (m)

%
difference dened as bulk minus
surface value
dielectric permittivity (C V1 m1 ),
Eq. (8)
electric potential (V)

3
G1 /z1 z2 G0
auxiliary variable, Eq. (100)
concentration
c
overpotential (V), Eq. (32)
conductivity (=1 m1 ), Eq. (14)

eff I /(d/dx) effective
conductivity (=1 m1 ), Eq. (42)
;12 molar conductivity
(m2 =1 mol1 ), Eq. (174)
i
electrochemical potential (J mol1 )
12 1 + 2 , Eq. (63)
i
stoichiometric
number, Eqs. (62, 83)
stoichiometric number
ir
in electrode reaction, Eq. (23)
charge density (C m3 ), Eq. (8)

e
electrical relaxation
time (s), Eq. (159)
reaction coordinate (mol), Eq. (24)
x/(2D 1/2 t 1/2 )

Boltzmann variable, Eq. (132)
Latin symbols
C
class concentration (mol m3 ),
Eq. (43)
ci
molar concentration (mol m3 )
D
diffusion coefcient (m2 s1 )
Eeq
equilibrium
electrode potential (V), Eq. (31)
,E
E
electric eld (V m1 )
e
f E dimensionless
electric eld, Eq. (118)
F
Faraday constant (C mol1 )
f
F /RT (V1 )

Gk
i zik Ji /Di (mol m4 ), Eq. (96)
I
electric current
density (A m2 ), Eqs. (12, 158)
Ji
ionic ux density (mol m2 s1 ),
Eq. (5)
K
equilibrium
constant
3
[(mol m ) i ], Eq. (63)
LD
Debye length (m), Eq. (161)
N
number of ionic species, Eq. (95)

n
= i zi ir stoichiometric
number of the electron, Eq. (23)
ni
number of moles of species
i (mol)
qk
auxiliary variable, Eq. (110)
R
gas constant (J mol1 K1 )
r
chemical reaction
rate (mol m3 s1 ), Eqs. (11, 84)

Sk
i zik ci (mol m3 ), Eq. (95)
s
Laplace variable (s1 ), Eq. (130)
T
temperature (K)
Ti
integral transport number, Eq. (25)
t
time (s)
ti
migrational transport number,
Eq. (15)
ui
ionic mobility (m2 V1 s1 ),
Eq. (4)
vi
ionic velocity (m s1 ), Eq. (4)
Subscripts and superscripts

d
displacement, Eq. (155)
dif
diffusion, Eqs. (19, 61)
e
electrical, Eq. (8)
i
ionic species
ij
dissociated electrolyte, Eq. (63)
ij, T
total electrolyte, Eq. (63)
ij, u undissociated electrolyte, Eq. (63)
j
constituent, Eq. (85)
k
power of charge number, Eq. (95)
L
limiting, Eq. (47)
Ld, i
mig
ohm
T
(i)
b
r
s
limiting diffusion
(current density), Eq. (78)
migration, Eq. (60)
ohmic, Eq. (18)
total, Eq. (158)
perturbation order, Eq. (170)
bulk solution (x = )
standard state, Eq. (31)
electrode reaction, Eq. (23)
electrode surface (x = 0)
17.
18.
19.
20.
21.
References
1. R. Haase, Thermodynamics of Irreversible
Processes, Addison-Wesley, New York, 1969.
2. K. Kontturi, Acta Polytech. Scand. 1983, 152,
140.
3. R. A. Robinson, R. H. Stokes, Electrolyte Solutions, Butterworths, London, 1955.
4. N. Ibl in Comprehensive Treatise of Electrochemistry (Ed.: E. Yeager, J. OM. Bockris,
B. E. Conway et al.), Plenum Press, New
York, 1983, Vol. 6, Chap. 1, pp. 163.
5. J. S. Newman, Electrochemical Systems,
Prentice-Hall, Englewood Cliffs, N.J., 1991.
6. A. V. Sokirko, J. Electroanal. Chem. 1994,
364, 5162.
7. E. J. Calvo in Comprehensive Chemical Kinetics (Eds.: C. H. Bamford, R. G. Compton),
Elsevier, Amsterdam, 1986, Vol. 26,
Chap. 1, pp. 178.
8. R. Schlogl, Stofftransport durch Membranen, Steinkopff-Verlag, Darmstadt, 1964,
Chap. 1, pp. 615, and Sect. 6.5, pp. 7993.
9. E. A. Guggenheim, Thermodynamics, North
Holland, Amsterdam, 1967, Chap. 8, pp.
298302.
10. D. Kondepundi, I. Prigogine, Modern Thermodynamics, John Wiley & Sons, Chichester, 1998.
11. R. P. Buck, J. Membr. Sci. 1984, 17, 162.
12. J. Pellicer, S. Mafe, V. M. Aguilella, Ber.
Bunsen-Ges. Phys. Chem. 1986, 90, 867872.
13. S. Mafe, J. Pellicer, V. M. Aguilella, J. Phys.
Chem. 1986, 90, 60456059.
14. D. G. Miller, Chem. Rev. 1960, 60, 1537.
15. D. G. Miller, J. Phys. Chem. 1967, 71,
35883592.
16. N. Ibl in Comprehensive Treatise of Electrochemistry (Eds.: E. Yeager, J. OM. Bockris,
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
B. E. Conway et al.), Plenum Press, New

York, 1983, Vol. 6, Chap. 4, pp. 239315.
K. B. Oldham, S. W. Feldberg, J. Phys.
Chem. B 1999, 103, 16991704.
J. C. Myland, K. B. Oldham, Electrochem.
Commun. 1999, 1, 467471.
N. Ibl, Pure Appl. Chem. 1981, 53,
18271840.
E. L. Cussler, Diffusion, Mass Transfer in
Fluid Systems, 2nd ed., Cambridge University Press, Cambridge, 1997, Chap. 13,
pp. 331355.
J. N. Agar, Discuss. Faraday Soc. 1947, 1,
2637.
L. L. Bircumshaw, A. C. Riddiford, Q. Rev.
Chem. Soc. London 1952, 6, 157185.
J. A. Manzanares, K. Kontturi in Surface
Chemistry and Electrochemistry of Membranes
(Ed.: T. S. Srensen), Marcel Dekker, New
York, 1999, Chap. 11, pp. 399435.
K. Hattenbach, Terminology for Electrodialysis, European Society of Membrane Science
and Technology, Geesthacht, Germany,
1988.
P. Delahay, Double Layer and Electrode Kinetics, Interscience Publishers, New York,
1965.
K. J. Vetter, Electrochemical Kinetics, Academic Press, New York, 1967, Chap. 2,
pp. 157184.
M. Planck, Ann. Phys. 1890, 39, 161186.
K. R. Johnson, Ann. Phys. (Leipzig) 1904,
14, 9951003.
H. Pleijel, Z. Phys. Chem. 1910, 72, 137.
D. A. MacInnes, The Principles of Electrochemistry, Reinhold, New York, 1939,
Chap. 13, pp. 220245.
M. Planck, Ann. Phys. 1890, 40, 561576.
R. Schlogl, Ber. Bunsen-Ges. Phys. Chem.
1978, 82, 225232.
K. B. Oldham, Z. G. Zoski in Comprehensive
Chemical Kinetics (Eds.: C. H. Bamford,
R. G. Compton), Elsevier, Amsterdam,
1986, Vol. 26, Chap. 2, pp. 79143.
V. G. Levich, Physicochemical Hydrodynamics, Prentice-Hall, Englewood Cliffs, N.J.,
1962, Chap. 6, pp. 231371.
Yu. I. Kharkats, Sov. Electrochem. 1988, 25,
503506.
A. V. Sokirko, Yu. I. Kharkats, Sov. Electrochem. 1989, 25, 287292.
Yu. I. Kharkats, A. V. Sokirko, J. Electroanal.
Chem. 1991, 303, 1725.
119
120
38. M. A. Vorotyntsev, Sov. Electrochem. 1988,
24, 11501154.
39. Y. Marcus, Ion Properties, Marcel Dekker,
New York, 1997.
40. J. G. Stark, H. G. Wallace, Chemistry Data
Book, John Murray, London, 1982.
41. V. K. Indusekhar, P. Meares in Physicochemical Hydrodynamics (Eds.: D. B. Spalding), Building & Sons, Guilford, 1977,
pp. 13011313, Vol. II.
42. K. Aoki, J. Electroanal. Chem. 2000, 488,
2531.
43. K. B. Oldham, T. J. Cardwell, J. H. Santos
et al., J. Electroanal. Chem. 1997, 430, 2537,
3946.
44. M. F. Bento, L. Thouin, C. Amatore et al., J.
Electroanal. Chem. 1998, 443, 137148.
45. M. F. Bento, L. Thouin, C. Amatore, J. Electroanal. Chem. 1998, 446, 91105.
46. A. M. Bond, S. W. Feldberg, J. Phys. Chem.
B 1998, 102, 99669974.
47. C. Amatore, L. Thouin, M. F. Bento, J. Electroanal. Chem. 1999, 463, 4552.
48. K. B. Oldham, J. Phys. Chem. B 2000, 88,
47034706.
49. A. D. MacGillivray, D. Hare, J. Theor. Biol.
1969, 25, 113126.
50. A. V. Sokirko, J. A. Manzanares, J. Pellicer,
J. Colloid Interface Sci. 1994, 168, 3239.
52. W. E. Morf, Anal. Chem. 1977, 49, 810813.
53. U. Behn, Ann. Phys. Chem. N.F. 1897, 62,
5467.
54. M. Kh. Urtenov, Methods of Solution of the
Nernst-Planck-Poisson Equation System [in
russian], Kuban State University, Krasnodar,
Russia, 1998.
55. V. V. Nikonenko, M. K. Urtenov, Russ. J.
Electrochem. 1996, 32, 187194.
56. A. Guirao, S. Mafe, J. A. Manzanares et al.,
J. Phys. Chem. 1995, 99, 33873393.
57. P. Ramrez, S. Mafe, A. Tanioka et al., Polymer 1997, 38, 49314934.
58. J. A. Manzanares, G. Vergara, S. Mafe et al.,
J. Phys. Chem. B 1998, 102, 13011307.
59. R. Schlogl, Z. Phys. Chem. N. F. 1954, 1,
305339.
60. J. F. Brady, J. C. R. Turner, J. Chem. Soc.,
Faraday Trans. 1 1978, 74, 28392849.
61. M. Abramowitz, I. A. Stegun, Handbook of
Mathematical Functions, Dover Publications, New York, 1965.
62. J. Crank, The Mathematics of Diffusion, 2nd
ed., Oxford University Press, Oxford, 1977.
63. S. L. Marchiano, A. J. Arvia in Comprehensive Treatise of Electrochemistry (Eds.:

E. Yeager, J. OM. Bockris, B. E. Conway
et al.), Plenum Press, New York, 1983,
Vol. 6, Chap. 2, pp. 65132.
64. D. Britz, Digital Simulation in Electrochemistry, Springer-Verlag, Berlin, Germany,
1988.
65. H. Cohen, J. W. Cooley, Biophys. J. 1965, 5,
145163.
66. J. R. Sandifer, R. P. Buck, J. Electroanal.
Chem. 1975, 79, 384391.
67. J. R. Sandifer in Ion-Transfer Kinetics (Ed.:
J. R. Sandifer), VCH Publishers, New York,
1995, Chap. 6, pp. 115138.
68. I. Rubinstein, Electro-diffusion of Ions, SIAM,
Philadelphia, Pa., 1990, Chap. 1, pp. 121.
69. W. D. Murphy, J. A. Manzanares, S. Mafe
et al., J. Phys. Chem. 1992, 96, 99839991.
70. G. R. Engelhardt, A. D. Davydov, Russ. J.
Electrochem. 1994, 30, 953, 954.
71. A. A. Moya, J. Castilla, J. Horno,
Ber. Bunsen-Ges. Phys. Chem. 1995, 99,
10371042.
72. A. A. Moya, J. Horno, Electrochim. Acta
1996, 99, 10371042.
Methods, John Wiley & Sons, New York,
1980.
74. I. C. Bassignana, H. Reiss, J. Phys. Chem.
1983, 87, 136149.
75. D. C. Grahame, J. Chem. Phys. 1950, 18,
903909.
76. F. Booth, J. Chem. Phys. 1951, 19, 391394,
13271328, 1615.
77. I. C. Yeh, M. L. Berkowitz, J. Chem. Phys.
1999, 110, 79357942.
78. J. M. G. Barthel, H. Krienke, W. Kunz, Physical Chemistry of Electrolyte Solutions. Modern
Aspects, (Ed.: Deutsche Bunsen-Gesellschaft
fur Physikalische Chemie e.V.), Steinkopff,
Darmstadt, 1998, Sect. 2.6.
79. S. Mafe, J. A. Manzanares, J. Pellicer, J.
80. J. D. Norton, H. S. White, S. W. Feldberg, J.
Phys. Chem. 1990, 94, 67726780.
81. C. P. Smith, H. S. White, Anal. Chem. 1993,
65, 33433353.
82. D. R. Baker, M. W. Verbrugge, Proc. R. Soc.
London, A 1998, 454, 18051829.
83. R. P. Buck, J. Electroanal. Chem. 1969, 23,
219240.
84. A. D. MacGillivray, J. Chem. Phys. 1970, 52,
31263132.

85. J. L. Jackson, J. Phys. Chem. 1974, 78,
20602064.
86. V. M. Aguilella, S. Mafe, J. Pellicer, Electrochim. Acta 1987, 32, 483488.
87. W. H. Smyrl, J. Newman, Trans. Faraday
Soc. 1967, 63, 207216.
88. R. H. Tredgold, Biochim. Biophys. Acta
1972, 274, 563574.
89. J. Newman, Trans. Faraday Soc. 1965, 61,
22292237.
90. A. V. Listovnichii, Russ. J. Electrochem. 1989,
25, 14791482.
91. M. Kh. Urtenov, V. V. Nikonenko, Russ. J.
Electrochem. 1993, 29, 314324.
92. V. V. Nikonenko, M. Kh. Urtenov, Russ. J.
Electrochem. 1996, 32, 195198.
93. L. Bass, Trans. Faraday Soc. 1964, 60,

16561663.
94. I. Rubinstein, L. Shtilman, J. Chem. Soc.,
Faraday Trans. 2 1979, 75, 231246.
95. J. A. Manzanares, W. D. Murphy, S. Mafe
et al., J. Phys. Chem. 1993, 97, 85248530.
96. I. Rubinstein, React. Polym. 1984, 2,
117131.
97. V. V. Nikonenko, V. I. Zabolotskii, N. P.
Gnusin, Sov. Electrochem. 1989, 25,
262266.
98. A. D. MacGillivray, J. Chem. Phys. 1968, 48,
29032907.
99. S. W. Feldberg, Electrochem. Comm. 2000, 2,
453456.
100. L. Boltzmann, Ann. Phys. 1894, 53, 959964.
121
122
2.2
The Digital Simulation of Voltammetry

under Stagnant and Hydrodynamic
Conditions
Kerry A. Gooch, Fulian L. Qiu, and Adrian
C. Fisher
University of Bath, Bath, United Kingdom
2.2.1
Introduction
Digital Simulations
The advent of the personal computer has
led to the rapid development of digital
simulations in the eld of electrochemistry. The motivation for these advances
has been driven by many factors including: the desire to gain quantitative insights
from electroanalytical measurements, extract kinetic and mechanistic information
regarding the pathways of electrolysis reactions, and assist in the design of new
quantitative electrochemical methods.
Quantication of the related phenomena proceeds by the denition of a model
that describes the physical processes (e.g.
mass transport, chemical reactivity, etc.)
in terms of a set of mathematical expressions. Traditionally, workers have attempted to solve these relationships directly via standard mathematical methods;
however, in many cases the complexity of the problem does not permit this
analytical-solution approach. Digital simulation breaks down (discretizes) the problem into a series of steps that can be
solved sequentially by the composition of
a suitable computer program. This discretization process gives rise to a variety
of different digital strategies with which
electrochemical or related problems can
be solved.
In this overview, three digital simulation
approaches (the nite difference method
(FDM), the nite element method (FEM),
2.2.1.1
and the boundary element method (BEM))

are discussed. Each has been used in varying degrees for the solution of the partial
differential equations relating to the mass
transport of material during electrolysis. At
least one formulation strategy is outlined
in detail for each method, to give workers, who are new to the eld, a jump-off
point to the extensive literature available.
Although not included within this review, sample codes for the FDM can be
found in a variety of publications, including, Feldberg [1], Bard and Faulkner [2],
Britz [3], and Gosser [4]. Sample codes for
the FEM and BEM are available on our
website [5]. In each case, the general formulation stages are followed by a review
of the digital simulation of mass transport using the specic technique. Special
emphasis is placed on the literature associated with hydrodynamic techniques (e.g.
the rotating disc, channel ow cell, or wall
jet electrode), since there are already excellent overviews of diffusional applications
using FDM (e.g. Ref. [3]).
Simulation Methods. The
Fundamentals
Irrespective of the simulation technique
chosen, the key stages in the development
of a numerical model and its validation are
shared and these are indicated in Fig. 1.
The procedure begins with a proposed,
realistic, or preliminary working physical model. A mathematical description
is formulated, which in the case of
transport-related problems proceeds via
conservation of momentum, mass, charge,
and energy.
For an electrolysis reaction, in which an
electrode is sited in an electrolyte solution
containing an electroactive species, the
species distribution (in one spatial dimension) may be described by an expression of
2.2.1.2
2.2 The Digital Simulation of Voltammetry under Stagnant and Hydrodynamic Conditions
O
Reactant
transport
Physical model
R
Product
transport
O+ e
R
Electrode
2c
2c
c0
= D 20 + D 20
y
t
x
Mathematical description
y
Numerical discretisation
Grid points
x
Electrode
c0
Computational prediction
Analytical solution
Computational predictions/
model validation
Distance
Fig. 1
Simulation procedure.
the form
c
= D 2 c Vx c k
t
(1)
In Eq. (1), c is the reagent concentration, t

the time, Vx the velocity, and k any kinetic
terms (e.g. rst-, second-order reactions,
etc.). For more complex geometries, such
as disc or band electrodes, the form of the
transport equation becomes more complex
but still retains terms of the same form as
Eq. (1).
Next, digital solution of the model
is required and this is achieved by
discretization of Eq. (1). This process
results in a set of algebraic equations that
can then be employed to solve for the
unknown variables.
To give a more general solution to

the expressions, it is common to cast
the equation variables in a dimensionless
form. For example, the concentration, c, is
cast dimensionless using
c =
c
cBULK
(2)
where cBULK is the concentration of a

specic starting material. Similarly, the
distance parameter x can be cast into a
dimensionless (X) form
X=
(3)
with representing a xed distance. The

time variable, t, can be normalized by the
123
124
introduction of a xed observation time, T ,

t
=
T
(4)
Additionally, dimensionless current, potential, reaction kinetic, and scan rate

parameters may also be employed. The
choice of , T , and so forth depends upon
the specic problem tackled and a good
overview may be found in the text by
Britz [3].
Solution of Eq. (1) in the dimensionless
form leads to the prediction of the temporal
and spatial variation of c. These may be
easily returned to dimensional variables
by reverse transformation.
Once the predictions have been obtained, they need to be validated and this
is often the most challenging aspect of
the simulation procedure. Workers often
congure codes so that they initially tackle
documented problems with analytical solutions, providing a quantitative check on
the simulations before extension to more
complex problems.
The Methods. An Overview
Before proceeding to the detailed development and numerical procedures for each
method, it is of interest to note the relationship between the FDM, FEM, and
BEM formulations. Although fundamentally different in their implementation,
they are linked mathematically (Fig. 2).
By way of example, consider the numerical procedure for the solution of Ficks
second law of diffusion in two dimensions
2.2.1.3
c
2c
2c
=D 2 +D 2
t
x
y
(5)
The FDM uses a truncated Taylor series

(Sect. 2.2.2.2) to approximate the differential equation. This approach uses the
original formulation of Ficks law and
the variable of interest (concentration)
appears as the second derivative. The solution procedure requires discretization

of the domain into a two-dimensional
grid of points, at which the concentration
is computed.
Implementation of the FEM requires
integration of the differential equations
once (by parts), to produce what is
often referred to as the weak formulation
(Sect. 2.2.3.2). In contrast to the FDM, the
integral expression contains the variable
and its rst derivative, thus reducing the
order of the differential by one with respect
to the variable. Domain discretization uses
a mesh of interconnecting elements and
the variable approximated using a chosen
interpolation function.
The BEM involves integration of the differential equation twice; this yields an expression that includes the variable and its
derivative. Discretization employs a mesh
of elements with the variable approximated
using an interpolation function. In the
BEM, the formulation permits (in some
cases) the variable distribution within the
domain to be related to the values only at
the boundary. The integration procedure
therefore effectively reduces the order of
the digital simulation from a two- to a onedimensional problem. This reduces the
complexity of the grid and number of elements required in comparison to the FDM
and the FEM.
The choice of the optimal method depends on the problem of interest. The
FDM has found widespread application in
the electrochemical eld due to the relative
ease of application, the routine modication for a variety of transport-related
problems, and the simple modication
for chemical reactivity issues. Apart from
its popularity in the engineering elds,
the FEM (Workers have used a variety of
approaches based around the method of
weighted residuals of which further detail
Finite
difference
Original
formulation
Electrode
Series of grid
points
Integration
Finite element
Weak formulation
Electrode
Mesh of interconnecting
triangular elements
Integration
Boundary
element
Inverse
formulation
Electrode
Mesh of elements around
boundary of domain
Fig. 2
The relationship between the FDM, FEM, and BEM formulations.
can be found in Sect. 2.2.3.2.) has, more

recently, also gained popularity in electrochemical applications. Merits of the FEM
include a considerably reduced number
of discretization steps within the domain
and the ability to address complex or irregular geometries.
Perhaps the most neglected of the three
techniques within electrochemistry is the
BEM. This in part has been due to the

perceived difculty [6] of implementation
with some relatively involved mathematical steps required before numerical solution. In addition, the application of the
method to problems of complex chemical reactivity is challenging. However, the
method does possess some benets over
both the FEM and FDM. In particular,
125
126
the reduction in dimensionality brought

about by the formulation procedure
can give rise to signicant time-savings
and reduced grid-generation requirements
when complex geometries and/or threedimensional analysis of transport problems are required.
FDM along with the application to a simple

electrochemical simulation.
The Explicit Finite Difference (EFD)
Method
Consider Ficks second law of diffusion (in
one dimension):
2.2.2.2
c
2c
=D 2
t
x
2.2.2
Finite Difference Methods

Introduction
The basis of the FDM was described as
early as 1911 [7] and underwent signicant
development throughout the rst half
of the 1900s. Courant and coworkers [8]
reported a detailed implementation and
later Emmons [9] applied the method to
a range of problems. In 1948, Randles
simulated an electrochemical linear-sweep
experiment [10] (with all of the calculations
performed by hand). However, it was
not until the 1960s with the work of
Feldberg [1] that the digital simulation of
electrode processes became a focused area
of electrochemical research (Feldberg used
a ux balance of material into and out
of a box to derive the discretized form
of the difference equations required for
solution.). Many formulations using the
FDM have since been reported and a
number of the key methods have been
highlighted below. The review begins
with a detailed overview of the key steps
required for the formulation of the explicit
2.2.2.1
cti1
cti
This expression may be cast into a nite

difference form via the following. The
simulation domain (x = 0 to x = ) is
divided into a series of equally spaced
(x) points as indicated in Fig. 3. The
t
t ) to an
concentration close (e.g. ci1
or ci+1
t
individual point (ci ) is approximated using
the Taylor expansion
f (x x) = f (x) xf (x) +
(x)3
f (x)
3!
(7)
where f , f , and f are the rst, second,
and third derivatives of f respectively.
If x is small, then the terms from
(x)3 /3! may reasonably be neglected and
t
expressions for the concentration at (ci1
t
and ci+1 ) predicted

t
ci+1
cit
+ x
c
x
(x)2
+
2!
2c
x 2

(8)
cti+1
Bulk
solution
x
X=d
The discretization of a domain into a series of sample points,

where the cit represents the concentration at the point i at a specic time t.
Fig. 3
(x)2
2!
f (x)
Electrode
X=0
(6)

t
ci1
cit
x
c
x
(x)2
+
2!
2c
x 2

(9)
summing Eqs. (8 and 9) yields

2c
1
t
=
(ct 2cit + ci1
) (10)
x 2
(x)2 i+1
Similarly the variation of concentration at
a point i during a small increment in time
(t) can be approximated as
ct+t cit
c
= i
t
t
(11)
cit+t and cit are the concentrations at the

point i at the new and old time intervals,
respectively. Substituting Eqs. (10 and 11)
into Eq. (6) followed by rearrangement
gives the nite difference form of Ficks
second law
t
t
cit+t = (ci1
+ (1 2)cit + ci+1
)
(12)
where
Dt
=
(13)
(x)2
The above expression provides a direct

method of obtaining the concentrations at
an individual point (i) via the values at the

previous time-step; hence this approach
is termed an explicit formulation. Note
that Eqs. (10 to 12) have been written in
terms of the concentrations at the time
interval t, this is referred to as a backwards
difference approximation. Application of
Eq. (12) permits the simulation of a variety
of voltammetric measurements under
diffusional transport control.
Consider a potential step experiment
performed under conditions in which
the analytical Cottrell equation is valid.
Assuming that initially the solution contains a single electroactive reagent of
concentration cBULK . At t = 0 the applied
potential is insufcient to drive the electrolysis reaction, following the potential
step at t > 0, the reagent is destroyed
at the electrode surface at a transport
limited rate. Digital simulation proceeds
by dividing the domain into N I nite
difference points (Fig. 4a) and applying
boundary conditions
t=0
cit = cBULK
t>0
cit
cit
i=1
(14)
=0
i=0
(15)
= cBULK
i = NI
(16)
i = NI
Electrode
Bulk
solution
X=d
X=0
(a)
i = 0, N I
cb
(b)
(a) Grid and (b) concentration distribution calculated using

an EFD simulation.
Fig. 4
127
128
The boundary condition (14) permits the

calculation of new concentrations at all
points (i) for the time increment t + t,
subject to the conditions specied in
Eqs. (15 and 16). The current may be
predicted by taking the concentration
gradient at x = 0

c
i = nFAD
(17)
x x=0
The choice of should ensure that
concentration changes from their bulk
values occur within the simulation region.
The simplest value for x would be
given by /N I , however, since the greatest
change in concentration is usually found
at or near the electrode surface, a grid
that concentrates mesh points close to the
surface would be desirable (Fig. 4a).
For one-dimensional simulations, Joslin
and Pletcher [11] and Feldberg [12] have
proposed alternative strategies based on
a smooth spacing function and exponentially expanding grid, respectively. Recently Bieniasz [13] has reported a more
sophisticated adaptation. The approach exploits a set of steps that recalculate the
mesh point positions following a concentration distribution calculation. Irrespective of the choice of function, the overall
goal is the accurate calculation of the concentration distribution too few sample
points yield an inaccurate (nonconverged)
approximation.
Selection of t in electrochemical simulations has generally remained a xed
value, although, a few groups have reported a range of time-dependent variations [1416]. However, with all of these
approaches, one must be aware of an
operating restriction of the EFD, which
arises from the parameter Eq. (13). If
its value exceeds 1/2, then the simulation
will not give a stable solution. This condition arises from the formulation procedure
and is known as the von Neumann stability criterion [17], and it has been converted
into a complete proof by John [18]. The
stability criterion can prove problematic
when a mesh contains small increments
of x. For a xed diffusion coefcient,
this requires a small value of t, thus
considerably increasing the real simulation time.
Extension of the EFD method to tackle
convection and chemical reactions provides no conceptual problems. The convective term from Eq. (1) can be cast into
the EFD form

t
cit ci1
c
= Vx
(18)
t
x
and summed with Eq. (10) to yield the EFD
expression for the convectivediffusion
equation. Similarly, kinetic complications
arising from homogeneous chemical reactions can be introduced (e.g. for a
rst-order chemical reaction)
c
= kcit
t
(19)
The ease of introduction of chemical reactivity into digital simulation methods

provides a key advantage over analytical procedures that are difcult to solve
for complex chemical problems. However, introduction of chemical reactivity
needs to be tackled with care (especially
when the rate constants are large). The
rapid chemical depletion of one or more
reagents within the simulation can result
in ner mesh requirements or shorter
time increments to retain stability and
accuracy within the simulation. Readers
are directed to a review by Speiser [19]
in which these issues are highlighted in
some detail.
The EFD method can be routinely applied to multidimensional spatial problems,
however the restriction imposed by has

led workers to look for alternative formulation strategies to overcome these difculties. In the next sections, some popular
alternative formulations are examined.
The Crank & Nicolson (CN)
Method
The CN method [20, 21] was used by
Randles [10] in his work carried out before
the advent of modern computers. A central
difference scheme is applied in which the
diffusional term (10) is centered at the
midpoint of the time-step
2.2.2.3
2 c(t+t/2)
1 t
t
=
2cit + ci+1
c
2
x
2(x)2 i1
t+t
t+t
+ ci1
2cit+t + ci+1
(20)
The concentration at (t + t/2) is evaluated via linear interpolation using
(t+t/2)
ci
(cit + cit+t )
2
(21)
and leads to an equivalent expression to

that of Eq. (12) for the explicit method
2( + 1) t+t
t+t
+ ci+1
ci
2( 1) t
t
t
= ci1
+
ci ci+1
t+t
ci1
(22)
where is given by Eq. (13). Unlike

Eq. (11), the new concentrations cannot
be directly obtained from Eq. (20) and
hence the method is generally referred
to as implicit (Some workers refer to the
CN scheme as a semi-implicit method
since the resulting expression contains
both concentration terms centered at t
and t + t.). Solution of Eq. (20) can be
achieved, since for a general problem of
N I points, a set of simultaneous equations
can be written in the tridiagonal form

2( + 1) t+t
+ c2t+t
c1
2( 1) t
= c0t +
(23)
c1 c1t
2( + 1) t+t
t+t
t+t
+ ci+1
ci
ci1
2( 1) t
t
t
= ci1
+
(24)
ci ci+1
2( + 1) t+t
t+t
t+t
cN
cN I + cN
I 1
I +1
2( 1) t
t
t
cN I cN
= cN
I 1 +
I +1
(25)
Application of appropriate boundary conditions yields a set of simultaneous equations that can be solved using a variety of
methods including Gaussian elimination,
of which the Thomas Algorithm [22] is a
simplied version. Although formulation
and implementation of the CN scheme is
considerably more involved than the EFD,
it does have advantages. In principle, the
stability of the method is not restricted
by the value of , therefore permitting
the use of larger time increments within
simulations. It has been noted that the CN
method does possess some limitations that
can be overcome by modication and the
reader is directed to the text by Britz [3] for
further details. The CN technique has been
applied to simulate one-dimensional diffusion to planar and spherical electrodes and
diffusion in redox polymers [23].
c0t+t
The Alternating Direction Implicit

(ADI) Method
The ADI method was reported by Peaceman and Rachford [24] in 1955 and introduced into electrochemical applications by
Heinze [2527]. A fully implicit procedure
is used to solve multidimensional diffusion or convective-diffusion problems.
2.2.2.4
129
One reported strategy [28] breaks each individual time increment into two half steps
(in the case of a two-dimensional problem). In the rst half-time increment,
one of the dimensions is solved explicitly whilst the other is retained in implicit
form, for example,

t+t/2
t
ci,j
ci,j

t+t/2
t+t/2
t+t/2
+ ci,j 1
= x ci,j +1 2ci,j

t
t
t
2ci,j
+ ci1,j
(26)
+ y ci+1,j
where
x =
Dt
(x)2
y =
Dt
(y)2
(27)
and i, j , x, and y are dened in Fig. 5.

In the following half-time increment, the
implicit/explicit directions are swapped

t+t/2
t+t
ci,j
ci,j

t+t/2
t+t/2
t+t/2
= x ci,j +1 2ci,j
+ ci,j 1

t+t
t+t
t+t
+ y ci+1,j
2ci,j
+ ci1,j
(28)
Values at t + t/2 are known from the
previous time-step and the implicit terms
are calculated at the interval t + t. The set
of tridiagonal matrix equations obtained
in each half-time increment may be solved
in an identical manner to those for the

CN scheme.
The ADI method was used by Heinze [29] to assess the voltammetric properties at microdisc electrodes. Gavaghan [30]
reported a thorough examination of optimum mesh generation and errors present
in the FD analysis of the disc electrodes
using the ADI method. More recently,
the technique has been widely adopted
in the eld of scanning electrochemical
microscopy [3137] for the simulation of
approach curve response and interfacial
chemical reactivity.
The ADI method has also been applied
to convective-diffusion problems to assess
the inuence of convective transport on the
voltammetric response of microdisc electrodes [38] in channel cells. The convective
component may be routinely introduced
to Eq. (28) in an analogous manner to
that noted for the EFD Eq. (18). The
three-dimensional transport problem was
addressed in an analogous manner to the
two-dimensional problem by splitting each
full time increment into three time-steps of
one third, to allow the x, y, and z directions
to be solved implicitly.
The Hopscotch (HS) Method
The HS method [39] was reported as an
alternative strategy to ADI by Shoup and
2.2.2.5
ci, j+1
ci1, j
130
ci, j
ci+1, j
ci, j1
Fig. 5 Two-dimensional mesh

for diffusional calculations.
Szabo [4042]. An implicit and explicit

scheme is employed at alternating points
within the two-dimensional mesh (Fig. 5),
which allows the new concentrations to be
calculated explicitly, but with the advantage
of an implicit stability criteria [22]. A
counter (a sum of the number of time
increments plus the i and j pointers) is
used to determine whether the calculation
is performed explicitly (if the counter
is odd)

t+t
t
ci,j
ci,j
t

t
t
= x ci,j
+1 2ci,j + ci,j 1
t

t
t
+ y ci+1,j
2ci,j
+ ci1,k
(29)
or implicitly (if the counter is even) at a
specic grid point.

t+t
t
ci,j
ci,j

t+t
t+t
t+t
= x ci,j
+ ci,j
+1 2ci,j
1

t+t
t+t
t+t
+ y ci+1,j
2ci,j
+ ci1,j
(30)
The values of x and y are given by
Eq. (27) and i, j , x, y are dened
in Fig. 5. When an implicit calculation
is performed at the point i, j , the
surrounding mesh points will have been
estimated explicitly at the new time
increment using Eq. (29). The counter
system ensures that at the next timeincrement, the points that have previously
been calculated explicitly will now be
evaluated implicitly and vice versa. The
approach allows unrestricted values of
to be used and was demonstrated
by Shoup and Szabo to give a good
level of accuracy for the solution of
transport limited current measurements
at the microdisc electrode. Despite its
relative ease of application and proven
stability, the method has not been widely
adopted by workers in the electrochemical

eld. Modication of Eqs. (29 and 30)
to include convective transport has been
reported for hydrodynamic applications at
a band ultramicroelectrode in a channel
ow cell [43], and at a tubular band
microelectrode in which the conditions for
the current to be independent of ow rate
are established [44].
The Backwards Implicit (BI)
Method
The fully implicit BI method was used by
Anderson [45, 46] to simulate the steady
state current response to ow rate within
a channel ow cell. Subsequently, this was
implemented by Fisher and Compton [47]
to evaluate the time-dependent convectivediffusion problem
2.2.2.6
c
c
2c
= D 2 Vx
t
y
x
(31)
Cast into fully implicit form, the FD form

of Eq. (31) is given by

t+t
t
ci,j
ci,j

t+t
t+t
t+t
= y ci+1,j
2ci,j
+ ci1,j

t+t
t+t
(32)
c ci,j
+1 ci,j
where i and j are pointers as shown in
Fig. 5, y is given by Eq. (27) and
c =
Vx t
(x)
(33)
Following application of the boundary conditions, the set of tridiagonal simultaneous

equations may be solved routinely as noted
above. Compton and coworkers have used
the technique extensively in mechanistic
analysis including ECE reactions [4853]
and coupled kinetics reactions [5459].
The technique has also been used to assess the inuence of hydrodynamics on the
131
132
current density at microband, microstrip,

and microdisc electrodes operating in
channel-based cells [6063]. In addition,
the technique has been used to simulate
voltammetric responses at wall jet and wall
jet ring disc electrodes [6467].
The Fast Implicit Finite Difference
(FIFD) Method
The Fast Implicit Finite Difference method,
implemented by Rudolph [68, 69] has
marginally higher computational requirements but has higher potential accuracy
when tackling demanding problems. The
FIFD formulation expresses the diffusional transport equation on an exponentially expanding spatial grid via

t+t
cit+t cit = D2i

cit+t
ci+1

t+t
cit+t ci1
D1i
2.2.2.7
(34)
where

D2i
= D exp 2
i
3

= D exp 2
i
D1i
4
i > 1 (35)
i 2 (36)
with , the grid expansion factor and

D =
Dt
(x)2
(37)
The approach has undergone extensive

development and interested readers are directed to the recent review by Speiser [19].
Feldberg and Rudolph [70] have adopted
one of the developments to create a commercial electrochemical software package
DigiSim [71, 72].
The Strongly Implicit Procedure
(SIP)
Compton and coworkers recently proposed the SIP [73] for a variety of
2.2.2.8
transport-related problems. The SIP was

originally reported by Stone [74] in 1968
as an efcient approach to the solution of two- and three-dimensional transport problems. The SIP has been developed to tackle a range of steady state
and time-dependent problems related to
voltammetric measurements. In this approach, the fully implicit form of the
two-dimensional transport equation is
solved directly:
t+t
t+t
x ci,j
y ci+1,j
+1 + (2x + 2y + 1)
t+t
t+t
t+t
t
x ci,j
ci,j
1 y ci1,j = ci,j
(38)
where y and x are given by Eq. (27).
Using this approach, the system equations
were cast into matrix form to yield
[A]{p} = {q}
(39)
where p is the vector of unknown concentrations, q the vector of known values,

and [A] contains the coefcients from the
right-hand side of Eq. (38). To overcome
the traditional difculties associated with
solution of the pentadiagonal matrices, the
matrix [A] can be modied by the addition of a further matrix [B] so that the
summed matrices [A] + [B] can be factorized into upper and lower triangular
matrices ([L] + [U ]).
[A][B] = [L][U ]
(40)
This approach permits the efcient iterative solution of the matrix equations using
a standard NAG routine (NAG FORTRAN
Library (D03EBF)). The approach has been
compared to the ADI and HS methods with
the authors concluding that the SIP provides a highly efcient competitor to these
strategies in both diffusion and convectivediffusion problems [75].
2.2.3
Formulation
The key steps involved in the formulation
of a one-dimensional diffusional problem
under steady state conditions now follow:
2.2.3.2
Finite Element Methods

Introduction
The FEM is part of a larger group of
techniques that exploit the method of
weighted residuals (MWR) [76]. These use
a set of weighting functions to allow
the approximation of a variable over a
domain. The choice of weighting function
leads to a number of different alternative
formulations, including the collocation
method, subdomain method, method of
moments, and the Galerkin method.
The FEM traces its lineage back to classical structural analysis [77] and Argyris [78],
who introduced the concept underlying the
FEM in 1954. In 1956, Turner and coworkers [79] made a key contribution to the
area when they reported the use of pinjointed bars and triangular plates for the
analysis of aircraft structures. The term,
nite element, however, was not widely
adopted until Clough [80] used it in the
title of a paper in 1960. The method
has subsequently been applied to a vast
range of problems, including, soil mechanics [81], aerodynamics [82], heat transfer [83], electric circuit analysis [84], and
has also been described in a large number of texts [8588]. Outlined below is
one formulation strategy for the FEM
analysis of one-dimensional diffusional
transport.
2.2.3.1
2c
=0
x 2
(41)
The FEM uses a mesh of elements to cover

the domain, as shown in Fig. 6
The variation of concentration c(x)
along the element length l is dened
by an interpolation function with the
concentrations related to the values at the
nodal points (Fig. 6) ca and cb
c(x) = Na ca + Nb cb
(42)
where the linear interpolation functions

are given by
1
(xb x)
l
1
Nb = (x xa )
l
Na =
(43)
(44)
and xa and xb are the bounds of each

element on Fig. 6. For an individual element, these relationships may be written
in matrix form
c = N e ce
where
ce =
ca
cb
N e = [ Na
(45)
Nb ]
Node
numbers
1
Electrode
2
1
3
2
4
3
ca
Fig. 6
5
4
One-dimensional element mesh.
(46)
6
5
b
cb
Element
numbers
133
134
Summing the contribution from ne elements

ne

N e ce
(47)
c=
and this is now integrated (by parts)

to yield

e=1
A weighted residual formulation [89, 90]

is used in order to solve Eq. (41). This approach seeks to approximate the variation
of concentration over a region in space
using a residual R(x) such that
2c
R(x) 2 = 0
x
w(x)R(x) dx = 0
From Eq. (47)

N e
C
C
=
x
x
(48)
c e b
N e N e
dx C D
(N ) = 0
a
x x
x
a
(54)
Noting the second term of Eq. (54) relates
to the boundary of the element and may
be applied later; evaluation of Eq. (54)
now requires the remaining integral. From
Eq. (45), we can calculate
b
(50)
where C is now an approximate solution

for the concentration. Recasting Eq. (41)
in weighted residual form gives

a
Ne
2C
dx = 0
x 2
(51)
1
l
1
N
=
x
l
(49)
The weighting function effectively distributes any residual error over the space
interval dened by the bounds a and b.
As noted, a number of possible routes to
apply Eq. (49) exist, including: orthogonal
collocation, the subdomain, and Galerkin
methods [89 pg. 78, 91]. In the FEM, the
Galerkin method exploits weighting functions that take the same form as the
interpolation functions Eqs. (43, 44). The
residual takes the same form as the original differential
(53)
2C
=0
R(x)
x 2

b

e C

N
=0
x a
(52)
and Eq. (52) can be rewritten
The aim of the numerical procedure is

to ensure that the residual is minimized
within the domain (ideally it will be zero).
Mathematically, this is achieved by forcing
R(x) to be zero (on average) over a given
nite region in space by multiplying by a
weighting function w(x)

N e C
dx D
x x
it follows from Eq. (54)

D
1 1
C=0
1
l 1
(55)
(56)
written in matrix form for an individual element

[K]e C = P
where
1
[K] =
1
e
1
1
(57)
(58)
([K]e is usually referred to as the characteristic matrix of the element).

For a mesh of ne elements (Fig. 6),
a global matrix (Fig. 7) is now constructed with contributions from each
node summed into the relevant matrix position.
At this stage, any boundary conditions
for the problem are introduced (e.g. ux
Node
numbers
1
1
2
3
4
11
c1
k1
1 + 1 1
c2
k2
c3
k3
c4
k4
1+1
1+1
1+1
c5
k5
c6
k6
5
6
Contributions
from each
element
Fig. 7
Unknown
concentration
vector
Known
concentration
vector
Global matrix for a problem in which ne = 5.
or xed values of concentration). For the

internal nodes, it is assumed that the
ux across the boundaries between each
element are in balance leaving only the
two values of ux at the external points of
the mesh to be dened (if required). Left
undened, the values correspond to the
Neumann boundary condition of no-ux.
However, a xed concentration, Dirichlet
boundary condition, may be routinely
introduced and further details can be
found in [86 pg. 184]. It is apparent that the
matrix formed is symmetric and banded
around the center diagonal and this
permits standard Gaussian elimination
routines to be employed for solution of
the matrices [9294].
Formulation to introduce convective
transport into electrochemical calculations
provides no conceptual difculties and
readers are referred to the article by
Stevens [95]. Similar procedures may be
used to formulate two-dimensional analogies and these have been described previously [96, 97].
Applications
Early applications of the MWR/FEM in
electrochemistry were reported by Whiting and coworkers [98] and Speiser and
coworkers [99105], who exploited the
method for electroanalytical purposes. In
1984, Penczek and coworkers [91, 106]
demonstrated the application of the FEM
to nite and innite one-dimensional
diffusion problems relating to the voltammetry at mercury amalgam lms.
Early two-dimensional simulations focused on the evaluation of the current
distribution at microdisc electrodes [107,
108] and simulations of a variety of
electrode geometries [109111] including
the inuence of recessed microelectrode
congurations [112]. Work has been also
extended to cases involving coupled homogeneous kinetics, adsorption [113], and
time-dependent redox polymer electrochemistry [114].
More recently, workers have reported
the use of automatic mesh generation
routines [115] and strategies that permit
2.2.3.3
135
136
the solutions to be obtained to a specied

accuracy [116]. The FEM has been utilized
to simulate hydrodynamic voltammetry at
the rotating disc, channel, wall jet, and
related electrode geometries [95, 117119]
and as a computer-aided design tool
by Fisher and coworkers [120124] for
the development of new electroanalytical
devices and to quantify voltammetry under
microuidic control.
Workers have also used a number
of commercially available FEM packages to allow the simulation of immiscible liquidliquid interfacial measurements [125127] and approach curves for
scanning electrochemical microscope applications [128]. Compton and coworkers
have also used a commercial FEM package
to model coupled heat and mass transport
at a wire [129], and dissolution kinetics in
ow-through devices [130].
that approximate the (nonlinear) variation

of the variable at the boundary. Below, one
formulation strategy for a one-dimensional
diffusion-controlled problem is outlined
Eq. (41).
Formulation
The formulation procedure begins in an
analogous manner to that of the FEM.
For the case of the diffusional operator
Eq. (41), the weighted residual form is
given by Eq. (51) as before, integration (by
parts) yields
2.2.4.2

w C
C b
D
=0
dx D w(x)
x a
a x x
(59)
A second integration gives the inverse
formulation

2.2.4
Boundary Element Method

Introduction
The BEM forms part of a group of
integral methods [131142] that have been
used sporadically for the simulation of
electrochemically related problems. The
BEM was described by Banerjee and
coworkers and Brebbia and coworkers and
has been outlined by them and others in
detail in a number of engineering texts [6,
143146]. A wide range of engineeringrelated problems have been tackled using
the BEM including: heat transfer, uid
mechanics, and structural analysis.
The BEM uses an integral approach
to solve the differential equations related
to the transport of material within a
domain. The differential equations are
transformed into an equivalent integral
equation at the boundary of the domain
and discretized using a series of elements
2.2.4.1
C b
2w
C
dx
+
D
w(x)
x 2
x a

w b
D C
= 0 (60)
x a
The procedure yields an integral in which

the differential operator acts on the
weighting function and not C, plus two
terms that are evaluated at the boundary.
Careful selection of w such that
2w
=0
x 2
(61)
reduces Eq. (60) to an integral expression

relating to the boundary values of concentration, the weighting function, and the
gradient of concentration. This procedure
is referred to as the homogeneous solution
method and has been applied to a range of
one-dimensional problems. An alternative
strategy usually adopted for solving twoand three-dimensional transport problems
is the use of the Dirac delta function
(x ) such that
2w
= (x )
x 2
(62)
when applied at the point x = , further

details may be found in [6 pg. 30]. The
choice of w is discussed elsewhere [144]
but is dependent upon the specic differential to be solved. A general solution that
satises Eq. (61) is
w = Ax + B
Applications
Despite the potential of the BEM to reduce
the dimensionality of the numerical solution and provide a direct measure of the
interfacial ux, it has been poorly exploited
by workers in the electrochemical eld in
comparison with the FDM and the FEM.
By comparison, heat and mass transfer
have been widely treated using the BEM
in the engineering literature [148151]. In
1984, the BEM was employed to calculate
the primary current distribution during an
electropolymerization reaction [152], the
potential of the BEM for applications to irregular geometries was also noted. Hume
and coworkers [153] used the approach to
analyze mass transport effects of electrodeposition through polymeric masks.
Of related interest, Ramachandran
and coworkers have reported a range
of papers on the inuence of mass
transport [154156], including diffusionreaction problems [157]. Further work was
reported on a variety of current distribution
problems [158161] in the early 1990s,
with a comparison of the FEM and the
BEM efciency reported by Matlosz and
coworkers [162]. A two-dimensional study
of coplanar auxiliary electrodes was reported by Mehdizadeh and coworkers [163]
and was used to assess the inuence of the
electrode conguration on uniform growth
over the cathode electrode. Electroplating
and corrosion protection in industrial cell
congurations have also been addressed
by Druesne and coworkers [164, 165].
In 1994, Barbero and coworkers [166]
extended BEM applications in the electrochemical eld to investigate convective electrodiffusion problems in charged
membranes. Qiu, Wrobel, and Power have
also outlined the application of the BEM to
transport-related problems including detailed procedures for assessing current
distribution effects controlled by combined
2.2.4.3
(63)
where A and B are arbitrary constants. Selecting two solutions of Eq. (63)
for example,
w1 = x
(64)
w2 = B
(65)
and then substituting these into Eq. (60)

yields two simultaneous equations

Ca
Cb
a
+ Ca + b
Cb = 0
x
x
(66)

Ca
Cb
+
=0
(67)
x
x
where Ca , Cb (Ca /x) and (Cb /x) correspond to the concentrations and uxes
at the boundary points a and b respectively. Application of boundary conditions
(dening the ux or concentration at each
boundary) permits the solution for the unknown values. Of particular interest to the
electrochemist is the ability to gain a direct
measure of the ux from the simulation
unlike the FEM and the FDM that rely on
a ne mesh normal to the surface.
Extension of the approach to convective
and indeed migratory transport has been
achieved in electrochemical applications
and readers are referred to the article
by Qiu, Wrobel, and Power [147] for
further details.
137
138
diffusion, convection, migration [147], examining binary and three electrolyte ion
systems [167], and in the prediction of electrode topography [168].
The BEM has been applied in scanning
electrochemical microscopy applications
by Fisher and Denuault [169] to examine
the inuence of probe and substrate
surface topography. In addition, timedependent phenomena have been assessed
in oil droplets [170, 171] and a range of
microelectrode geometries using the dual
reciprocity method (DRM) [172] closely
related to the BEM [173].
2.2.5
Alternative Simulation Strategies
In addition to the topics reviewed above,

which form the vast majority of the articles
published to date in the eld of electrochemical simulation, there are a number
of other alternative methods that have
been exploited by workers. These include,
statistical techniques such as the Monte
Carlo method [174179], which has been
exploited to examine the fractal nature
of electrode surfaces and electrodeposited
polymer lm growth. The nite volume
method, which has found signicant application in the engineering literature [180,
181], remains poorly exploited in the
electrochemical eld [182, 183] as does
the multidimensional upwinding method,
which has been applied by Van Den Bossche and coworkers [184, 185] to multi-ion
systems at the rotating disc electrode. For
recent advances, readers are referred to the
review of Speiser [19].
Acknowledgments
FLQ (GR/N00548) and KAG (studentship

No. 9930027X) would like to thank the
EPSRC for their support.
References
1. S. W. Feldberg in Electroanalytical Chemistry
(Ed.: A. J. Bard), Marcel Dekker, New York,
1969, pp. 199296.
1980.
3. D. Britz, Digital Simulation in Electrochemistry, 2nd ed., Springer-Verlag, Berlin, 1988.
4. D. K. Gosser, Cyclic Voltammetry. Simulation and Analysis of Reaction Mechanisms,
Wiley-VCH, Weinheim, Germany, 1993.
5. http://www.bath.ac.uk/chsacf.
6. P. A. Ramachandran, Boundary Element
Methods in Transport Phenomena, 1st ed.,
Elsevier, Southampton, UK, 1994.
7. L. F. Richardson, Philos. Trans. 1911, 210,
307357.
8. R. Courant, K. Friedrichs, H. Lewy, Math.
Ann. 1928, 100, 32.
9. H. W. Emmons, Q. Appl. Math. 1944, 2,
173195.
10. J. E. B. Randles, Trans. Faraday Soc. 1948,
44, 327338.
11. T. Joslin, D. Pletcher, J. Electroanal. Chem.
1974, 49, 171.
12. S. W. Feldberg, J. Electroanal. Chem. 1981,
127, 1.
13. L. K. Bieniasz, J. Electroanal. Chem. 1993,
360, 119138.
14. J. B. Flanagan, K. Takahashi, F. Anson, J.
15. J. W. Dillard, J. A. Turner, R. A. Osteryoung, Anal. Chem. 1977, 49, 1246.
16. R. Seeber, S. Stefani, Anal. Chem. 1981, 53,
10111016.
17. J. von Neumann, R. D. Richtmyer, J. Appl.
Phys. 1950, 53, 10211099.
18. F. John, Common Pure Appl. Math. 1952, 5,
155211.
19. B. Speiser in Electroanalytical Chemistry
(Eds.: A. J. Bard, I. Rubinstein), Marcel
Dekker, New York, 1996, pp. 1108.
20. J. Crank, P. Nicolson, Proc. Cambridge Philos. Soc. 1947, 43, 5067.
21. J. Crank, The Mathematics of Diffusion,
Clarendon Press, Oxford, 1956.
22. L. Lapidus, G. F. Pinder, Numerical Solution
of Partial Differential Equations in Science and
Engineering, John Wiley & Sons, New York,
1982.
23. M. Storzbach, J. Heinze, J. Electroanal.
Chem. 1993, 346, 127.
24. D. W. Peaceman, H. H. Rachford, J. Soc.
Ind. Appl. Math. 1955, 3, 28.
25. J. Heinze, M. Storzbach, Ber. Bunsen-Ges.
Phys. Chem. 1986, 90, 10431048.
26. J. Heinze, M. Storzbach, J. Mortensen, J.
27. J. Heinze, M. Storzbach, J. Mortensen, J.
28. M. K. Jain, Numerical Solution of Differential
Equations, Wiley Eastern Limited, New
Delhi, 1979.
29. J. Heinze, J. Electroanal. Chem. 1981, 124,
73.
30. D. J. Gavaghan, J. Electroanal. Chem. 1998,
456, 135.
31. A. J. Bard, F. R. F. Fan, F. F. Fu-Ren et al.,
Anal. Chem. 1989, 61, 132.
32. C. Demaille, P. R. Unwin, A. J. Bard, J.
Phys. Chem. 1996, 100, 14 13714 143.
33. J. L. Amphlett, G. Denuault, J. Phys. Chem.
B 1998, 102, 99469951.
34. A. J. Bard, M. V. Mirkin, P. R. Unwin et al.,
J. Phys. Chem. 1992, 96, 18611868.
35. P. R. Unwin, A. J. Bard, J. Phys. Chem. 1991,
95, 78147824.
36. J. V. Macpherson, P. R. Unwin, J. Phys.
Chem. 1996, 100, 19 47519 483.
37. J. V. Macpherson, P. R. Unwin, J. Phys.
Chem. 1994, 98, 17041713.
38. J. Booth, R. G. Compton, J. A. Cooper et al.,
J. Phys. Chem. 1995, 99, 10 94210 947.
39. A. R. Gourlay, J. Inst. Math. Appl. 1970, 6,
375.
40. D. Shoup, A. Szabo, J. Electroanal. Chem.
1982, 140, 237245.
1984, 160, 117.
1986, 199, 437441.
43. P. Pastore, F. Magno, J. Lavagnini et al., J.
44. H. R. Corti, D. L. Goldfarb, A. S. Ortiz et al.,
Electroanalysis 1995, 7, 569573.
45. S. Moldoveanu, J. L. Anderson, J. Electroanal. Chem. 1984, 175, 67.
46. J. L. Anderson, S. Moldoveanu, J. Electroanal. Chem. 1984, 179, 107119.
47. A. C. Fisher, R. G. Compton, J. Phys. Chem.
1991, 95, 75387542.
48. A. C. Fisher, R. G. Compton, J. Appl. Electrochem. 1992, 22, 3842.
49. R. G. Compton, B. A. Coles, A. C. Fisher, J.
Phys. Chem. 1994, 98, 24412445.
50. J. A. Alden, R. G. Compton, J. Phys. Chem.

B 1997, 101, 96069616.
51. J. A. Alden, M. A. Feldman, E. Hill et al.,
Anal. Chem. 1998, 70, 17071720.
52. A. B. Miles, R. G. Compton, J. Phys. Chem.
B 2000, 104, 53315342.
53. A. B. Miles, R. G. Compton, J. Electroanal.
Chem. 2001, 499, 116.
54. R. G. Compton, M. B. G. Pilkington, G. M.
Stearn, J. Chem. Soc. 1988, 84, 21552171.
55. J. A. Alden, R. G. Compton, R. A. W. Dryfe,
J. Electroanal. Chem. 1995, 397, 1117.
56. B. A. Coles, R. A. W. Dryfe, N. V. Rees et al.,
57. J. A. Alden, R. G. Compton, J. Phys. Chem.
B 1997, 101, 89418954.
58. F. Prieto, W. J. Aixill, J. A. Alden et al., J.
Phys. Chem. B 1997, 101, 55405544.
59. W. J. Aixill, J. A. Alden, F. Prieto et al., J.
Phys. Chem. B 1998, 102, 15151521.
60. R. G. Compton, A. C. Fisher, R. G. Wellington et al., J. Phys. Chem. 1993, 97,
10 41010 415.
61. C. W. Davies, M. K. Walters, A. C. Fisher
et al., J. Phys. Chem. 1995, 99, 10 942.
62. W. J. Aixill, A. C. Fisher, Q. Fulian, J. Phys.
Chem. 1996, 100, 14 06714 073.
63. C. W. Davies, Q. Fulian, M. K. Walters et al.,
Electroanalysis 1996, 11, 849.
64. R. G. Compton, A. C. Fisher, G. P. Tyley, J.
Appl. Electrochem. 1991, 21, 25.
65. A. C. Fisher, R. G. Compton, J. Appl. Electrochem. 1991, 21, 208212.
66. R. G. Compton, A. C. Fisher, M. H. Latham
et al., J. Phys. Chem. 1992, 96, 83638367.
67. J. C. Ball, R. G. Compton, C. M. A. Brett, J.
Phys. Chem. B 1998, 102, 162166.
68. M. Rudolph, J. Electroanal. Chem. 1991, 314,
1322.
8598.
70. S. W. Feldberg, C. I. Goldstein, M. Rudolph,
71. J. Mocak, S. W. Feldberg, J. Electroanal.
Chem. 1994, 378, 1729.
72. M. Rudolph in Physical Electrochemistry (Ed.:
I. Rubinstein), Marcel Dekker, New York,
1995, pp. 81129.
73. R. G. Compton, R. A. W. Dryfe, R. G. Wellington et al., J. Electroanal. Chem. 1995, 383,
1319.
74. H. L. Stone, Siam J. Numer. Anal. 1968, 5,
530.
139
140
75. J. A. Alden, R. G. Compton, J. Electroanal.
Chem. 1996, 402, 110.
76. B. A. Finlayson, Br. Chem. Eng. 1969, 14,
5357.
77. R. Courant, Bull. Am. Math. Soc. 1943, 49,
123.
78. J. H. Argyris, Aircraft Eng. 1954, 26, 347.
79. M. J. Turner, R. W. Clough, H. C. Martin
et al., J. Aerosol Sci. 1956, 23, 805.
80. R. W. Clough, Proceedings of 2nd ASCE
Conference on Electronic Computation,
Pittsburg, Pa., 1960.
81. C. V. Girijavallabhan, L. C. Reese, J. Soil
Mech. Proc. ASCE 1968, 94, 473496.
82. S. S. Rao, Proc. Int. Symp. Disc. Methods Eng.
Milan, 1974, pp. 512525.
83. S. G. Ravikumaur, K. N. Seetharamu, P. A.
Aswathanarayana, Proc. Int. Conf. On Finite
Elements in Computational Mechanics 1985,
861, Vol. 2.
84. P. P. Silvester, R. L. Ferrari, Finite Elements
for Electrical Engineers, 3rd ed., Cambridge
University Press, Cambridge, 1996.
85. O. C. Zienkiewicz, R. L. Taylor, The Finite
Element Method, 5th ed., ButterworthHeinemann, Oxford, 2000.
86. S. S. Rao, The Finite Element Method in
Engineering, Pergamon Press, Oxford, 1982.
87. G. Dhatt, G. Touzot, The Finite Element
Method Displayed, John Wiley & Sons, New
York, 1984.
88. C. Cuvelier, A. Segal, A. A. van Steenhoven,
Finite Element Methods and Navier-Stokes
Equations, D. Reidel, Dordrecht, The
Netherlands, 1986.
89. J. Villadsen, M. L. Michelsen, Solution of
Differential Equations by Polynomial Approximation, Prentice Hall, Englewood Cliffs, NJ,
1978.
90. G. Fairweather, Finite Element Galerkin
Methods for Differential Equations, John
Wiley & Sons, New York, 1978.
91. M. Penczek, Z. Stojek, J. Osteryoung, J.
92. E. Isaacson, H. B. Keller, Analysis of Numerical Methods, John Wiley & Sons, New York,
1966.
93. L. W. Johnson, R. D. Riess, Numerical Analysis, 2nd ed., Addison-Wesley, Reading,
Mass., 1982.
94. W. H. Press, S. A. Teukolsky, W. T. Vetterling et al., Numerical Recipes in Fortran 77:
The Art of Scientic Computing, 2nd ed.,
95.
96.
97.
98.
99.
100.
101.
102.
103.
104.
105.
106.
107.
108.
109.
110.
111.
112.
113.
114.
115.
116.
117.
118.
Cambridge University Press, Cambridge,

1996.
N. P. C. Stevens, A. C. Fisher, J. Phys.
Chem. B 1997, 101, 82598263.
A. J. Chapman, Heat Transfer, 4th ed.,
Macmillan Publishing, New York, 1984.
J. P. Holman, Heat Transfer, 8th ed.,
McGraw-Hill, New York, 1997.
L. F. Whiting, P. W. Carr, J. Electroanal.
Chem. 1977, 81, 120.
B. Speiser, A. Rieker, J. Electroanal. Chem.
1979, 102, 120.
B. Speiser, J. Electroanal. Chem. 1980, 110,
6977.
S. Pons, B. Speiser, J. F. McAleer, Electrochim. Acta 1982, 27, 11771179.
B. Speiser, S. Pons, A. Rieker, Electrochim.
Acta 1982, 27, 11711176.
B. Speiser, J. Electroanal. Chem. 1984, 171,
95109.
P. Hertl, B. Speiser, J. Electroanal. Chem.
1987, 217, 225238.
P. Hertl, B. Speiser, J. Electroanal. Chem.
1988, 250, 237256.
M. Penczek, Z. Stojek, J. Electroanal. Chem.
1984, 181, 8391.
J. Galceran, D. J. Gavaghan, J. S. Rollet, J.
M. Penczek, Z. Stojek, J. Electroanal. Chem.
1987, 227, 271274.
M. M. Stephens, E. D. Moorhead, J. Electroanal. Chem. 1987, 220, 130.
N. P. C. Stevens, S. J. Hickey, A. C. Fisher,
An. Quim. 1997, 93, 225232.
N. P. C. Stevens, F. L. Qiu, K. A. Gooch
et al., J. Phys. Chem. B 2000, 104,
71107114.
P. N. Bartlett, S. L. Taylor, J. Electroanal.
Chem. 1998, 453, 4960.
E. D. Moorhead, M. M. Stephens, J. Electroanal. Chem. 1990, 282, 126.
E. Deiss, O. Haas, C. Daul, J. Electroanal.
Chem. 1992, 337, 299324.
T. Nann, J. Heinze, Electrochem. Commun.
1999, 1, 289294.
K. Harriman, D. Gavaghan, P. Houston
et al., Electrochem. Commun. 2000, 2,
150170, 567585.
R. Ferrigno, P. F. Brevet, H. H. Girault, J.
N. P. C. Stevens, A. C. Fisher, Electroanalysis 1998, 10, 1620.
119. B. A. Coles, R. G. Compton, C. M. A. Brett
et al., J. Electroanal. Chem. 1995, 381,
99104.
120. Q. Fulian, N. P. C. Stevens, A. C. Fisher, J.
Phys. Chem. B 1998, 102, 37793783.
121. K. A. Gooch, N. A. Williams, A. C. Fisher,
Electrochem. Commun. 2000, 2, 5155.
122. Q. Fulian, A. C. Fisher, D. J. Riley, Electroanalysis 2000, 12, 503508.
123. F. L. Qiu, K. A. Gooch, A. C. Fisher et al.,
Anal. Chem. 2000, 72, 34803485.
124. N. P. C. Stevens, K. A. Gooch, A. C. Fisher,
J. Phys. Chem. B 2000, 104, 12411248.
125. J. Josserand, J. Morandini, H. J. Lee et al., J.
126. R. Ferrigno, H. H. Girault, J. Electroanal.
Chem. 2000, 492, 16.
127. H. J. Lee, C. Beriet, R. Ferrigno et al., J.
128. Y. Lee, S. Amemiya, A. J. Bard, Anal. Chem.
2001, 73, 22612267.
129. A. Beckmann, B. A. Coles, R. G. Compton
764769.
130. B. A. Coles, R. G. Compton, M. Suarez
et al., Langmuir 1998, 14, 218225.
131. S. Coen, D. K. Cope, D. E. Tallman, J. Electroanal. Chem. 1986, 215, 2948.
132. M. Fleischmann, J. Daschbach, S. Pons, J.
133. J. Daschbach, S. Pons, M. Fleischmann, J.
17.
135. U. Kalapathy, D. E. Tallman, D. K. Cope, J.
136. D. K. Cope, C. H. Scott, D. E. Tallman, J.
137. D. K. Cope, D. E. Tallman, J. Electroanal.
Chem. 1990, 285, 7992.
138. D. R. Baker, M. W. Verbrugge, J. Newman,
139. M. V. Mirkin, A. J. Bard, J. Electroanal.
Chem. 1992, 323, 127, 2951.
140. L. K. Bieniasz, Comput. Chem. 1992, 16,
311317.
141. L. K. Bieniasz, J. Electroanal. Chem. 1993,
347, 1530.
142. B. Pillay, J. Newman, J. Electrochem. Soc.
1993, 140, 414420.
143. P. K. Banerjee, R. Buttereld, Boundary Element in Engineering Science, McGraw-Hill,
New York, 1981.
144. C. A. Brebbia, J. C. F. Telles, L. C. Wrobel,

Boundary Element Techniques. Theory and
Application in Engineering, Springer-Verlag,
Berlin, Germany, 1984.
145. C. A. Brebbia, Boundary Element Method for
Engineers, Pentech, London, 1984.
146. B. E. Beskos, Boundary Element Methods
in Mechanics, Computational Methods in
Mechanics, North Holland, Amsterdam,
1987, 125, Vol. 3.
147. Z. H. Qiu, L. C. Wrobel, H. Power, Eng.
Anal. Boundary Elements 1995, 15, 299.
148. P. H. L. Groenenboom, Appl. Math. Modell.
1982, 6, 35.
149. C. A. Brebbia, P. Skerget, J. Appl. Phys.
1982, 53, 8366.
150. W. L. Wendland, J. Zhu, Math. Comput.
Modell. 1991, 15, 19.
151. S. Y. Long, X. C. Kuai, J. Chen et al., Eng.
152. R. Bialecki, R. Nahlik, M. Lapkowski, Electrochim. Acta 1984, 29, 905.
153. E. C. Hume, W. M. Deen, R. A. Brown, J.
Electrochem. Soc. 1984, 131, 1251.
154. P. L. Mills, S. Lai, M. P. Dudukovic et al.,
Siam J. Sci. Stat. Comput. 1988, 9, 271.
155. P. A. Ramachandran, Int. J. Numer. Methods
Eng. 1990, 29, 1021.
156. P. A. Ramachandran, Chem. Eng. J. 1990,
45, 49.
157. P. A. Ramachandran, J. Comput. Phys. 1992,
102, 63.
158. J. F. Yan, S. N. R. Pakalapati, T. V. Nguyen
et al., J. Electrochem. Soc. 1992, 139, 1932.
159. P. Cicognani, F. Gasparoni, B. Mazza et al.,
J. Electrochem. Soc. 1990, 137, 1689.
160. J. Horkins, L. T. Romankiw, J. Electrochem.
Soc. 1989, 136, 756.
161. E. K. Yung, L. T. Romankiw, J. Electrochem.
Soc. 1989, 136, 764.
162. M. Matlosz, C. Creton, C. Clerc et al., J.
Electrochem. Soc. 1987, 134, 3015.
163. S. Mehdizadeh, J. Dukovic, P. C. Andricacos
et al., J. Electrochem. Soc. 1990, 137, 110.
164. F. Druesne, P. Paumelle, Corros. Prevention
Control 1998, 45, 118.
165. F. Druesne, P. Paumelle, P. Villon, Eng.
166. A. J. Barbero, S. Mafe, P. Ramirez, Electrochim. Acta 1994, 39, 2031.
167. Z. H. Qiu, L. C. Wrobel, H. Power, J. Appl.
Electrochem. 1997, 27, 1333.
168. Z. H. Qiu, H. Power, J. Appl. Electrochem.
2000, 30, 575.
141
142
169. F. Qiu, A. C. Fisher, G. Denuault, J. Phys.
Chem. B 1999, 103, 4387, 4393.
170. F. L. Qiu, J. C. Ball, F. Marken et al., Electroanalysis 2000, 12, 10121016.
171. J. C. Ball, F. Marken, F. L. Qiu et al., Electroanalysis 2000, 12, 10171025.
172. F. L. Qiu, A. C. Fisher, Electrochem. Commun. 2000, 2, 738742.
173. D. Nardini, C. A. Brebbia in Boundary
Element Methods in Engineering (Ed.:
C. A. Brebbia), Springer-Verlag, Berlin,
1982, pp. 312326.
174. B. Marner, W. Schmickler, J. Electroanal.
Chem. 1986, 214, 589596.
175. B. Aurian-Blajeni, M. Kramer, M. Tomkiewicz, J. Phys. Chem. 1987, 91, 600605.
176. T. Pajkossy, L. Nyikos, Electrochim. Acta
1989, 34, 171179.
177. F. Sagues, J. M. Costa, J. Chem. Educ. 1989,
66, 502506.
178. S. Zalis, N. Fanelli, L. Posps il, J. Electroanal. Chem. 1991, 314, 111.
179. G. Nagy, Y. Sugimoto, G. Denuault, J. Electroanal. Chem. 1997, 433, 167173.

180. H. K. Versteeg, W. Malalasekera, An Introduction to Computational Fluid Dynamics:
The Finite Volume Method, Longman Group,
Harlow, UK, 1995.
181. L. Ronghua, W. Wei, C. Zhongying, Generalised Difference Methods for Differential
Equations: Numerical Analysis of Finite Volume Methods, Marcel Dekker, New York,
2001.
182. S. Bacha, A. Bergel, M. Comtat, J. Electroanal. Chem. 1993, 359, 2138.
183. Yu. V. Benderskii, V. G. Mairanosvkii, Elektrokhimiya 1992, 28, 835841.
184. B. Van Den Bossche, L. Bortels, J. Deconinck et al., J. Electroanal. Chem. 1995, 397,
3544.
185. B. Van Den Bossche, L. Bortels, J. Deconinck et al., J. Electroanal. Chem. 1996, 411,
129143.
147
3.1
NMR Spectroscopy in Electrochemistry

YuYe Tong
Georgetown University, Washington, Washington DC
Eric Oldeld and Andrzej Wieckowski
University of Illinois at Urbana-Champaign,
Urbana, Illinois
3.1.1
Introduction
The fundamental physical origins of the

different chemical reactivities of catalytic
or functional metal surfaces be they used
as heterogeneous catalysts, electrocatalysts, substrates for materials engineering,
or simply as templates for corrosion studies are electronic [1]. Most surface processes can be considered to involve the
optimization of the electronic interaction
energy of the system through suitable geometric rearrangements. These surfaces
processes necessarily involve the activation of one or more chemical ingredients
together with the concomitant excitation
of the many-electron metal surface. They
are, therefore, in general rather complex
processes.
In the gas (dry) phase, ultrahigh vacuum
(UHV)-based surface science characterization techniques have been developed
for many years for scrutinizing complex

surface processes [2]. Unfortunately, few
of these techniques are readily applicable
to investigating problems at the electrochemical interface, because of the ubiquitous presence of the condensed-phase
electrolyte. In electrocatalysis/fuel cell applications, for instance, specic chemical
and physical properties of the electrode become of paramount importance in terms
of understanding the fundamental aspects of such interfacial electrochemical
phenomena [3]. It is in this eld of interfacial electrochemistry that the newly
developed electrochemical nuclear magnetic resonance (EC-NMR) spectroscopy
technique [4] is mainly applicable. This
development is now beginning to bring
results that illuminate both electronic and
dynamic aspects of processes in real-world
catalytic systems [410].
Technically, however, NMR probably
possesses the lowest mass-detection sensitivity of any spectroscopic technique [11]
and is still quite a challenging prospect.
To appreciate the difculties, note that a
typical high-eld NMR instrument needs
1018 1019 NMR-active atoms, for example 13 C spins, to make a signal detectable
within a reasonable time period. However,
1 cm2 of a single-crystal metal surface contains only about 1015 atoms. Therefore,
at least 1 m2 surface area of NMR-active
148
3 New Experimental Evidences
atoms is needed to meet the sensitivity

requirement, which is the primary reason
that most surface NMR investigations reported so far have used high-surface-area
materials. Nevertheless, there are several
advantages to using NMR to investigate
(electrochemical) surface processes. For
example, NMR is extremely sensitive to
local static (geometric), electronic, and
dynamic structure [12]. Also, NMR can
characterize samples in most types of catalysts metal powders, oxide-supported
metal catalysts, carbon-supported catalysts, and both metal and adsorbate structures can be probed. In addition, it is
a nondestructive technique and can be
applied under technically demanding environmental conditions close to real-world
operating conditions. For example, it is
possible to investigate graphite-supported
metal catalysts in an electrochemical environment under potential control. Given
adequate mass sensitivity then, NMR has
a unique ability to provide a wealth of
electronic and structural information on
the atomic level, and in addition permits one to access motional information
on adsorbates over a time-span unattainable by any other spectroscopic technique.
As such, NMR of oxide-supported metal
catalysts has therefore played an important historical role in the development of
many fundamental aspects of gas-phase
heterogeneous catalysis at transition metal
surfaces [1316], and the techniques developed, primarily by Slichter and coworkers [17], can be readily adapted to electrochemical systems.
EC-NMR was pioneered by Wieckowskis group [18] at the University of
Illinois at the beginning of 1990s and,
as noted above, has its roots in metal
NMR of oxide-supported metal catalysts.
However, interfacing solid-state NMR with
interfacial electrochemistry is somewhat
more challenging than the more standard

gas-phase techniques because of the
basic incompatibility between NMR detection and the conducting requirements
for proper electrode function. That is, the
presence of conducting material in the
NMR-detection coil greatly degrades the
so-called quality factor of the NMR probe,
which substantially decreases sensitivity.
Nevertheless, the potential advantages of
interfacing in situ solid-state NMR with
electrochemistry for investigating the electrochemical interface are many. For example: NMR gives a direct probe of the
Fermi level local densities of states for
both adsorbate and substrate; structure
and bonding in an electrochemical environment can be investigated using a
variety of NMR interactions; surface dynamics can be readily probed; and, of
course, the effects of potential control can
be investigated in working electrochemical systems even where the adsorbates
(such at CN on Pt) would not be accessible in conventional gas-phase or
dry systems. Furthermore, by using high
magnetic elds and high-surface-area electrode materials, the current state-of-the-art
signal-detection sensitivity is in fact quite
reasonable, and permits the routine acquisition of NMR spectra of surface species.
As an example, we show in Fig. 1 13 C
NMR spectra of 13 CO chemisorbed onto
a 10-nm Pt sample having a CO coverage
of about 0.8, recorded at room temperature and in, respectively, Fig. 1(a), 14.1 and
Fig. 1(b), 8.5 T magnetic elds. The numbers of scans were 1600 and 2000, and the
signal-to-noise (S/N) ratios (after using a
1-kHz line broadening) were 50 and 10,
respectively, showing a signicant S/N enhancement in the higher magnetic eld.
Sensitivity is also greatly enhanced at lower
temperatures at which 13 C NMR spectra
can now be recorded in as little as 10 min.
3.1 NMR Spectroscopy in Electrochemistry
3.1.2
Principles of Solid-state Surface EC-NMR

Spectroscopy
Metal NMR Basics
In electrochemistry, electrodes are indispensable parts of an electrical circuit, so the
EC-NMR of electrode surfaces in general
has to deal with metallic systems. In NMR,
in general, a specic resonance position (a
frequency shift) with respect to a welldened reference corresponds to a specic
chemical environment (a structure) [12].
This shift is the chemical shift (CS ) in a
molecule or the Knight shift (K) [19] in a
metal, and can be considered a ngerprint
of the corresponding site or structure [20].
While the familiar chemical or orbital shift
is ubiquitous, arising from the shielding
of the external magnetic eld at the nucleus by the otherwise quenched orbital
angular momentum of the surrounding
electrons, the Knight shift is metal-specic,
being produced by nonzero magnetic interactions between nuclear and electronic
spins since only electron spins at the
Fermi energy can be polarized and thus
create a nonzero spin density in an external magnetic eld (the so-called Pauli
paramagnetism). In addition, the Korringa
relationship [21] (vide infra), which governs the nuclear spinlattice relaxation in
metals, is also metal-specic. Together,
the Knight shift and the Korringa relation
represent the two primary NMR probes
of electronic structure in metals, and of
surface molecular bonding in general.
3.1.2.1
NMR amplitude
Fig. 1 13 C NMR spectra of CO

adsorbed on a 10-nm Pt electrocatalyst,
recorded at room temperature and at
(a) 14.1 T with 1600 scans and (b) 8.5 T
with 2000 scans. The S/N ratios are
ca. 50 and 10, respectively.
(a)
(b)
800
600
400
200
d
[ppm]
Theoretically, the Hamiltonian of the nuclearelectron hyperne interaction can

be written for a nuclear spin I = 1/2 as

8
H = n hI B
S(r)(r)
3

3r(Sr)
S
3
(1)
3
r
r
r5
where B is the Bohr magneton and n is
the nuclear gyromagnetic ratio. I , S, and l
are the nuclear spin, electron spin, and
electron orbital moments, respectively.
The symbol r is the radius vector of an
electron with respect to the nucleus at the
origin. The rst term in the brace in Eq. (1)
is the Fermi contact term, the second
(square bracket) is the dipole term, and the
third (l/r 3 ) is due to the orbital movement
of electrons. In a diamagnetic molecule,
149
150
since all electron spins are paired up, the

expectation values of the rst and second
terms become zero, leaving only the third
term, which produces the usual so-called
chemical or orbital shift, CS . In the case
of a metal, the total line shift is the sum
of the rst term (K) and the third term
(CS ) in Eq. (1). In many cases, K, which
arises from the hyperne interaction of
the observed nucleus with electron spins,
dominates the purely chemical shift term,
CS . The dipolar term normally contributes
to relaxation, but not to the shift, since the
trace of the operator is zero.
The Korringa relationship [21] indicates
that 1/T1 T , where T1 is the spinlattice
relaxation time and T is the absolute
temperature of the sample. This unique
temperature dependence of 1/T1 is the
NMR ngerprint of a metallic state. It
results from the fact that only conduction electrons around the Fermi level
can satisfy energy conservation for the
electron-nuclear spin ip-op relaxation
process, and the fraction of these electrons
is proportional to kB T . When all relaxation
mechanisms other than the rst term in
Eq. (1) can be neglected, and there are only
s-like electrons at the Fermi level (such as
in the alkali metals), the Korringa relationship takes its simplest form:

2
K T1 T =
e
n
2
h
4kB

=S
(2)
Here, e and n are the electronic and

nuclear gyromagnetic ratios, respectively,
and kB is the Boltzmann constant. The
Knight shift, K, also has a simple
relationship with the electronic density of
states at the Fermi level through the Pauli
susceptibility, Pauli :
K = Pauli
Hhf
= B Hhf D(EF )
B
(3)
where Hhf is the hyperne eld produced

by an electron at the site of the nucleus,
and D(EF ) is the (local) density of
states at the Fermi level. Despite their
simplicity, Eqs. (2 and 3) already give the
simple physical picture that the NMR of
metallic species can measure, in principle,
D(EF ). For systems having more than
one band, which cuts the Fermi level,
the deduction of multiple D(Ef )s becomes
more involved, though the principles are
the same. More specically, for 195 Pt [22]
or chemisorbed 13 CO [10], one can express
the two primary metal NMR observables,
the Knight shift K and the Korringa
constant S(T1 T )1 , in terms of D(EF )s.
This is the so-called two-band model, the
bands corresponding to metal s and d
bands or to ligand (13 CO) 5 and 2
bands. By experimentally measuring K
and S(T1 T )1 (for the metal or ligand)
and by solving two equations for two
unknowns, s (5 )- and d (2 )-D(EF ) can
be deduced. For more details, the reader is
referred to Refs. [10, 20, 22].
It is important to note that many metallic properties, such as the Knight shift
and the Korringa relationship, are determined by the nite and quasi-continuous
nature of the Fermi level local density
of states (EF -LDOS). In the approximation most familiar to chemists, what this
means is that the highest occupied molecular orbitallowest unoccupied molecular
orbital (HOMOLUMO) gap in metals is
much smaller than the thermal energy
kB T , and the value of the EF -LDOS reects the frontier orbital contributions in
a metallic system [23]. The EF -LDOS also
represents a crucial metal surface attribute
that can serve as an important conceptual bridge between the delocalized band
structure (physics) picture and the localized chemical bonding (chemical) picture
of metaladsorbate interactions.
In addition to the determination of

the chemical and/or Knight shift and
the spinlattice relaxation time in the
laboratory frame (T1 ), there is another important NMR observable the spinspin
relaxation time (T2 ). While the chemical
and/or Knight shift contains essentially
static structural information, the temperature and/or magnetic eld dependence of
the relaxation times, both T1 and T2 , are
related to the dynamics of the observed
nucleus. T1 measures the rate at which
the spin system returns to thermal equilibrium with its environment (the lattice)
after a perturbation, while T2 measures
the rate of achieving a common spin temperature within the spin system. Both T1
and T2 provide exceptionally important information on motions, and can cover the
timescale from 109 to 102 s. Moreover,
the temperature dependence of these motions provides important thermodynamic
information in the form of activation energies for ligand motion on the catalyst
surface.
EC-NMR Instrumentation
Two types of EC-NMR experiments
have been carried out: electrode potentialdependent studies of an adsorbate
(13 CO, 13 CN) at room temperature [6, 8,
24], and temperature-dependent studies
of an adsorbate [5, 10] and of platinum
electrocatalysts [7, 25], down 10 K. For
all of the EC-NMR measurements, the
electrode materials, either polycrystalline
platinum black or carbon-supported commercial fuel cell grade platinum electrodes,
were immersed in a supporting electrolyte,
typically 0.5 M H2 SO4 . For temperaturedependence studies, EC-NMR samples are
prepared in a conventional three-electrode
ow cell with oxygen-free N2 or Ar as a
protecting gas, then together with supporting electrolyte and under the protection of
3.1.2.2
oxygen-free N2 or Ar, are transferred into

a precleaned glass ampoule and amesealed. The potential drift of a sample in
a conventional three-electrode cell, in an
oxygen-free environment, is only a few
millivolts over 12 h. The same observation
holds after transferring the electrode material from the NMR ampoule back into
the cell. These observations suggest that
the surface potential does not change signicantly once a sample is sealed (and
does not change at all when the sample
is frozen).
For room temperature electrode potentialdependence studies, we have incorporated an electrochemical cell inside
the NMR probe [6]. This permits running
NMR measurements while an (variable)
external electrode potential is applied.
Figure 2 shows a picture of a real EC-NMR
probe as well as a schematic diagram of
the setup. The working electrode material was loosely packed in the cell that
was placed inside an NMR coil and electrical connection to an external circuit
(potentiostat) made by inserting a Pt wire
into the sample. (A very similar setup
has also been developed by the Lausanne
group [8, 24].) Our cell design permits both
electrochemical sample preparation and
characterization (voltammetry) and NMR
data acquisition under active potential control, avoiding sample transfer from the
preparative electrochemical cell. A very important technical issue here is that while
the potentiostat is on, there is considerable noise injected into the NMR probe,
since the potentiostat leads act as excellent antennas right inside the NMR coil!
This noise has to be removed by using
extensive electronic ltering of the electrochemical leads entering the probe, in order
to eliminate extensive radio frequency (rf)
pickup. The low-pass lter circuit elements
Lcounter Ccounter , Lref . Cref . , and Lwk. Cwk.
151
152

NMR spectrometer
Cmatching
Ctuning
Counter
electrode
NMR coil
Reference
Lcounterelectrode
Ccounter
Lref.
Working
electrode
Lwk.
Cwk.
Cref.
Potentiostat
Schematic diagram (left) and picture (right) of an EC-NMR probe and its circuitry,
showing the interface between NMR and electrochemistry.
Fig. 2
in Fig. 2 were designed for this purpose,

plus they serve a second purpose of acting
as a high impedance to the rf pulses from
the NMR spectrometer. These are typically
1 kW and may be incompatible with long
potentiostat life.
Figure 3 shows another type of setup for
an EC-NMR cell, developed by the Berkeley
group to investigate 13 CO adsorption from
the gas phase onto a carbon-supported
platinum electrocatalyst [26, 27]. In this
system, a platinum electrocatalyst was
pressed into a graphitized carbon cloth
that was then cut into pieces and stacked
into a cylindrical form that ts snugly
into the cells compartment. In order to
reduce degradation in the quality factor
of the NMR coil due to the presence
of bulk electronic or ionic conductivity,
the electrode layers were also alternately
sandwiched with nonconductive berglass
cloth. All the electrode layers were individually stitched with gold wire and
were then connected to each other and

to the external potentiostat. Cyclic voltammograms of high quality are obtained by
both types of EC-NMR cell.
3.1.3
Applications of Solid-state Surface EC-NMR

Spectroscopy
195 Pt
NMR of Carbon-supported
Electrocatalysts and a Potential-scangenerated Sintering Effect
At the beginning of the 1980s, Slichter
and coworkers discovered several unique
features of the 195 Pt NMR of oxidesupported small platinum particles [28].
They found that the overall 195 Pt NMR
lineshape was extremely broad, extending
downeld some 4 kG from the position
of bulk platinum (1.138 G kHz1 ), and
contained a feature on the low-eld side
(1.089 G kHz1 ), which arose from the
oxidized Pt surface atoms. Later on, van
3.1.3.1
F
E
A
D
B
C
Fig. 3 Schematic representation of the EC-NMR probe cell

developed by the Berkeley group [27]: (A) reference electrode
compartment; (B) NMR coil; (C) working electrode compartment;
(D) counterelectrode compartment; (E) electrolyte inlet;
(F) electrolyte outlet; (G) carbon-supported Pt electrocatalysts; and
(H) berglass separator. (Reproduced with permission from
Ref. [27], Copyright by 2001 Electrochemical Society.)
der Klink and coworkers conrmed these

observations, nding that the signal from
clean-surface Pt atoms was centered at
1.100 G kHz1 [29], a position clearly very
different to the 1.138 G kHz1 bulk position (the frequency difference is magnetic
eld dependent and is about 2.5 MHz in a
eld of 8.5 T). Ab initio theoretical calculations on a ve-layer Pt (001) cluster [40]
then demonstrated that the surface shift
must be due to a gradual drop in the dlike Fermi level LDOS on moving from
the inside of the particle to the surface. It
is this distinguishable surface signal that

makes 195 Pt NMR unique in investigating the surface physics and chemistry of
nanoscale platinum particles.
We have observed that carbon-supported
nanoscale platinum electrodes retain quite
closely the 195 Pt NMR spectral characteristics of isolated small platinum particles
supported on oxides [4, 7, 25, 30], with the
clean-surface platinum atoms also resonating at 1.100 G kHz1 with respect to the
value for bulk atoms, 1.138 G kHz1 . We
show in Fig. 4(ac), 195 Pt NMR spectra
153
NMR Amplitude
154
1.08
1.10
1.12
1.14 1.08
1.10
1.12
1.14
Field/frequency
[G kHz1]
Fig. 4 Typical point-by-point 195 Pt NMR spectra showing electrochemical
cleaning, sintering by potential cycling, and a layer-model analysis of the
2.5-nm sample: (a) as-received catalyst; (b) electrochemically cleaned in
0.5 M H2 SO4 by holding electrode potential at 0.45 V versus reversible
hydrogen electrode (RHE); (c) cleaned by extensive potential cycling; and
(d) layer-model deconvolution of spectrum (b). The solid line in (b) is the
result of the simulation.
of a 2.5 nm carbon-supported commercial

Pt electrocatalyst after different surface
treatments. Figure 4(a) shows the spectrum of an as-received sample, while
Fig. 4(b) shows a sample that has been
electrochemically cleaned by holding the
electrode potential within the electrochemical double-layer region with no potential
cycling until the reduction current could
no longer be measured. Figure 4(c) shows
the spectrum obtained after extensive potential cycling (which is still widely used
to clean electrode surfaces), and in which
there are clearly discernable spectral differences, as discussed below.
In order to further analyze these results,
we use a layer-model [31] (vide infra)
deconvolution of the lineshape, and results
for the clean-surface spectrum (Fig. 4b) are

shown in Fig. 4(d). The NMR layer model
assumes that the nanoscale platinum
particles can be represented by ideal
cubo-octahedral particles built up layerby-layer from a central atom, that NMR
signals from atoms within a given layer
can be approximated by a Gaussian, and
that the average Knight shift of the nth
layer, Kn , which is the center of the
corresponding Gaussian, heals back
exponentially towards the bulk platinum
position when moving inwards. That is,

n
Kn = K + (K0 K ) exp
(4)
m
where n is the layer number, counting
inwards from the surface layer (where
n = 0), m is the characteristic number

(of layers) dening the healing length
for the Knight shift, and K0 and K
(= 3.34%) are the Knight shifts of the
surface layer and the bulk, respectively.
The relative contribution of each Gaussian
is dictated by the fraction of the atoms
within the corresponding layer, which can
be determined from the size distribution
of the sample. The healing length, dened as m times the distance between two
consecutive layers (0.229 nm for Pt), is
0.46 nm (
=2 Pt layers) from the deconvolution shown in Fig. 4(d). In addition, the
shape of a 195 Pt NMR spectrum conveys
information on the size of the electrocatalyst. That is, the stronger is the surface
peak, the higher is the dispersion, or the
smaller is the particle size.
The surfaces of the as-received materials are covered by O and/or OH
species, as indicated by the peak at
1.098 G kHz1 . Remarkably, these surfaces were effectively cleaned simply by
holding the electrode potential within the
electrochemical double-layer region. This
mild surface-reduction procedure should
be contrasted to the rigorous and often
technically demanding methods employed
at the solidgas interface, which usually
involve several cycles of high-temperature
calcination and reduction, limiting the
choice of catalyst support to, typically, nonconductive oxides such as alumina, titania,
and silica.
However, extensive potential cycling,
which is a widely used in electrode-surface
cleaning procedure, sinters the nanoscale
electrocatalyst. Although the sample in
Fig. 4(c) started with the same as-received
electrocatalyst as that shown in Fig. 4(b),
the dispersion estimated from the corresponding 195 Pt NMR spectrum is only
about 27% [30], far lower than the 50%
obtained from Fig. 4(b). This dispersion
lowering provides clear evidence for sintering produced by potential cycling, an

effect that clearly needs to be borne in
mind when analyzing the results of both
ligand and metal EC-NMR experiments.
Correlation between the Knight
Shift of a Platinum Surface and the
Electronegativity of the Adsorbate
Understanding how adsorbates modify the
chemical and physical properties of metal
surfaces has long been one of the central themes of surface science, and is of
importance in relation to the ability to
scientically engineer metal surface properties for targeted applications, such as
enhancing the CO tolerance of electrocatalysts used in fuel cells, or more generally
in promoting metal surface catalysis. As
demonstrated in Sect. 3.1.3.1, interfacial
electrochemistry offers an elegant way to
modify electrode surfaces while 195 Pt NMR
provides a powerful probe to follow the
variations in the physical properties of
metal surfaces caused by such modications. By using this strategy, the inuence
of a series of different ligands (H, O, S,
CN , CO, and Ru), electrochemically adsorbed onto carbon-supported nanoscale
Pt particles from the same starting batch,
has been investigated by 195 Pt NMR, in an
electrochemical environment [10].
The starting electrode material used for
this series of experiments was the same
as that used in the previous section. As
shown there, the frequency difference between the resonance positions of surface
and bulk platinum atoms at 8.5 T is about
2.5 MHz, large enough to provide a very
convenient spectral visualization of how
deep the inuence of an adsorbate can
go, as illustrated by the point-by-point
195 Pt NMR spectra (recorded at 80 K)
presented in Fig. 5. These spectra are normalized by equalizing their amplitudes
3.1.3.2
155
Pt/clean
Pt/Ru
Pt/CO
Pt/O
Pt/H
Pt/CN
Pt/S
NMR amplitude
156
1.08
1.10
1.12
1.14
Field/frequency
[G kHz1]
Superimposition of point-by-point, 8.47 T 195 Pt NMR spectra of
a 2.5 nm, carbon-supported Pt electrocatalyst without and with
different chemisorbed ligand: Ru, CO, O, H, CN , and S. The spectra
were normalized by equalizing the amplitude at 1.131 G kHz1
(indicated by the arrow). The invariance of signals beyond
1.131 G kHz1 provides experimental conrmation of the
FriedelHeine invariance theorem. The surface peaks range over
11 000 ppm. (Reproduced with permission from Ref. [10], Copyright
by 2000 American Chemical Society.)
Fig. 5
at 1.131 G kHz1 , which is indicated by

the arrow. The interesting observation
here is that the position of the high-eld
signal intensity (above 1.131 G kHz1 ),
which is due to Pt atoms within the
three innermost layers, is independent of
adsorbate type. This independence provides direct experimental evidence for
the validity of the FriedelHeine invariance theorem [32, 33], which states that
the integral electronic properties at an
atom, such as its LDOS, are determined
primarily by the surrounding medium,
within a few electronic wavelengths. This
is exactly what is observed although the
chemical identity of the adsorbates is
quite varied, their inuences, to a very
good approximation, do not go beyond

their next-nearest neighbors. That is, the
high-eld signals are invariant towards
boundary changes.
In complete contrast to the Friedel
Heine invariance of the electronic properties observed for the innermost platinum
particle layers, the surface and subsurface NMR signals undergo major frequency shifts as different chemical species
are adsorbed, as can be seen in Fig. 5.
The NMR layer-model spectral deconvolution [31] technique was applied to these
spectra in order to obtain the variations
in the surface and subsurface Knight
shifts. Because of the FriedelHeine
invariance, the NMR parameters (peak
NMR amplitude
position, width, relative intensity) for the

three central layers were xed to the
clean-surface values, while the surface and
subsurface peak positions were allowed to
vary. For Pt/O, the position of the third
layer was also varied, as a result of the
long healing length in this system, plus
the surface peak areas were also varied in
some simulations, as a result of saturation
effects caused by longer T1 values. The
results of these simulations are shown in
Fig. 6 and the surface and subsurface peak
shift values obtained are plotted in Fig. 7
as a function of the AllredRochow electronegativity (dened as the electrostatic
force exerted by the nucleus of the atom on
its valence electrons) of the bonded adsorbate atom. Remarkably, both the surface
and subsurface peak positions map almost
linearly the electronegativity of the adsorbate. The reasons for this are as follows:
According to the layer model, Eq. (4), one
has K1 = K + (K0 K ) exp(1/m)
for the subsurface layer, n = 1. A simple mathematical conversion gives m =
1/ ln[(K0 K )/(K1 K )]. By replacing K1 and K0 by the correlation lines

shown in Fig. 7,
K0 (ppm) = (10.2 + 5.3) 103 (5)
K1 (ppm) = (21.5 + 5.8) 103 (6)
m (the healing length) can then be
directly expressed as a function of the
electronegativity :

1 (23.2 + 5.3)
(7)
m=
ln (11.9 + 5.8)
In Fig. 8, the healing length (= 0.229 m,
where 0.229 nm is the layer thickness for
Pt particles) is plotted as a function of
the electronegativity, Eq. (7). This result
shows that the larger the electronegativity
of the adsorbate, the longer the healing
length, that is, the deeper the inuence of
the adsorbate goes. It is also interesting to
note that the value of the healing length
m for hydrogen ( = 2.2) adsorption
obtained via Eq. (7) is about 2.9, which
is very close to the m = 2.6 value found
Pt/clean
Simulation
Pt/CO
Pt/CN
Pt/Ru
Pt/S
Pt/H
Pt/O
1.08 1.10 1.12 1.14 1.08 1.10 1.12 1.14 1.08 1.10 1.12 1.14 1.08 1.10 1.12 1.14
Field/frequency
[G kHz1]
The NMR layer-model simulations of the
spectra shown in Fig. 5 and the small dots
are the variances between the experimental and
tted values. (a) clean-surface Pt; (b) Simulation
of (a); (c) Pt with adsorbed Ru; (d) Pt with
Fig. 6
195 Pt
adsorbed H; (e) Pt with adsorbed CO; (f) Pt with

adsorbed cyanide; (g) Pt with adsorbed S (from
Na2 S solution); (h) Pt with adsorbed O.
(Reproduced with permission from Ref. [10],
Copyright by American Chemical Society 2000.)
157

Surface
0
d ( 103)
[ppm]
158
10
Subsurface
20
Pt bulk shift
30
1.5
2.0
2.5
3.0
3.5
AllredRochow electronegativity
Correlation between surface/subsurface frequency shifts
(with respect to the Pt NMR reference H2 PtCl6 ) and the
AllredRochow electronegativity. The dashed horizontal line
indicates the Knight shift of bulk platinum atoms. The solid
straight lines are linear ts to the surface and subsurface shifts
as a function of the electronegativity. Both have R2 values of
ca. 0.92. (Reproduced with permission from Ref. [10], Copyright
by American Chemical Society 2000.)
Fig. 7
0.9
0.8
0.7
0.6
H
Li
0.5
K
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
Electronegativity
Solid curve showing the relationship between the
healing length and electronegativity as determined from
Eq. (7). For comparison, we also show the three experimental
points based on previous observations. See the text for
details. (Reproduced with permission from Ref. [10],
Copyright by American Chemical Society 2000.)
Fig. 8
by a detailed layer-model analysis [31].

In addition, for the adsorption of the
alkali metal elements ( 1), Eq. (7)
predicts m 2, virtually the same m as
obtained for clean-surface Pt particles in an
electrochemical environment. One would,
therefore, expect that alkali adsorption
would not signicantly change the 195 Pt
NMR spectrum of small Pt particles after
adsorption. This is exactly what has been
observed experimentally (in a gas-phase
system) [34]. For comparison, these three
experimentally estimated points are also
plotted in Fig. 8.
The invariance of the frequency shifts
of the more buried (deeper than the
third layer) Pt atoms to surface adsorbate electronegativity provides strong evidence demonstrating the applicability of
FriedelHeine invariance of the EF -LDOS
in these nanoscale electrocatalyst systems.
This is a rather interesting extension
of the original FriedelHeine invariance
theorem to nanoscale systems and may
have utility in describing the electronic
surface structure of metal catalysts in
different chemical environments. These
results may also be expected to lead to
useful general correlations between electronic properties and more conventional
chemical descriptors (such as ligand electronegativity), which could be helpful in
understanding the electronic structure of
metaladsorbate interfaces by providing
guidelines for engineering new electrode
surfaces.
Correlation between the
Clean-surface EF -LDOS of Metals and the
Knight Shift of Adsorbates
Because NMR has elemental specicity
and is also nondestructive, it offers unique
opportunities for investigating electrochemical problems from both sides of
the electrochemical interface. This strategy
3.1.3.3
has been applied to investigate the

metalCO interactions in a number of
recent studies [25].
The discussions in the two previous
sections demonstrate that there is an unambiguous NMR discrimination between
the surface peak (at 1.100 G kHz1 for
nanoscale platinum particles) and the rest
of the platinum sites. By carrying out
detailed temperature-dependent T1 relaxation measurements on a given Pt sample,
a value for S(T1 T )1 can be obtained,
and by using the two-band model (see
Sect. 3.1.2.1), a pair of surface Pt s- and
d-D(EF )s can therefore be deduced. Similarly, S(T1 T )1 and K measurements on
13 CO adsorbed onto the metal surfaces
can give the corresponding 13 C 5 and
2 (EF )s (see Sect. 3.1.3.4).
It is of fundamental interest to see if
there exists any type of correlation between the clean-surface EF -LDOS of the
metal and the 13 C Knight shift of the
chemisorbed CO. The former is an important surface attribute dening the ability
of the metal surface to donate (to act as
a HOMO) as well as to accept (to act
as a LUMO) electrons, while the latter
measures, at least to a certain extent, the
degree of metallization of chemisorbed CO
(see Sect. 3.1.3.4). We show in Table 1, a
set of data for atop CO on Pt and Pd,
shown graphically in Fig. 9. The purely
orbital contributions (chemical shifts) for
atop CO on Pt and on Pd were calculated
by using density functional theory (DFT)
calculations on model COPt7 and COPd7
clusters in which CO sits atop the central
metal atom, which is coordinated by the
remaining six metal atoms. For COPt7 ,
a value of 160 20 ppm from tetramethylsilane (TMS) was obtained, while
for COPd7 , the value was 203 20 ppm.
These calculated values are in good agreement with available experimental data
159

Correlation between clean-surface EF -LDOS and 13 C shift of chemisorbed
CO [25]. The Knight shift = total shift chemical shift
Tab. 1
Sample
Clean-surface EF -LDOS
(Ry1 atom1 )
Total 13 C shift of COchemisorbed

(ppm w.r.t TMS)
0
14.8
17.7a
18.6
0
24
160
330
351a
383
203
500
PtCO7
COPt/oxides (dry)
COPt/carbon (wet)
COPt/carbon (wet)
PdCO7
COPd/oxides (dry)
a For
an 8.8-nm sample.
500
400
Shift
[ppm]
160
300
200
10
15
20
25
30
Clean surface LDOS (Ry1 atom1)

Correlation between the Fermi level total electronic
densities of states at transition metal (Pt and Pd) surfaces (in
Ry1 atom1 ) and the corresponding 13 C Knight shift of
chemisorbed CO. (Partially reproduced with permission from
Ref. [25], Copyright by American Chemical Society.)
Fig. 9
while the standard deviations account for

uncertainties caused by the use of different functionals and basis sets. The
clean-surface EF -LDOS for Pd was estimated on the basis of the changes in
magnetic susceptibility of small Pd particles with respect to their bulk value,
by application of the exponential healing model.
As shown in Fig. 9, the 13 C NMR
shift of CO responds linearly to the
clean-surface EF -LDOS before adsorption.

The straight line is a linear t to the data,
giving a slope of 12 ppm/Ry1 atom1 .
This linear relationship can be readily
related to the frontier orbital interpretation
of the Blyholder model [35] a higher
clean-surface EF -LDOS means more metal
electrons and holes are available to engage
in 5 -forward and 2 -back donation.
As the EF -LDOS increases, CO becomes
more metallic, which results in a larger
Knight shift. A correlation very similar

in nature to this has also been found to
exist in the gas phase, in which the
CO vibrational stretch frequency (after
chemisorption) was shown to correlate
linearly with the Pt clean-surface EF -LDOS
before chemisorption [36]. The higher the
surface EF -LDOS, the lower the CO
stretching frequency.
The linear relationship shown in Fig. 9
is important because it demonstrates the
validity of the frontier orbital interaction picture of metal surface chemistry,
in which the importance of the cleansurface EF -LDOS is highlighted. In addition, it puts 13 C NMR spectroscopy
of chemisorbed CO on a rmer footing, by probing the electronic properties
of transition metal surfaces before CO
chemisorption.
Coupling EC-NMR with In Situ
Infrared Spectroelectrochemistry
It is now clear that CO acquires metallic
properties upon adsorption onto transition
metal surfaces, as indicated by the presence of a Knight shift and a Korringa
relationship, measured via 13 C NMR.
Furthermore, by using the experimentally measured 13 C Knight shift and the
Korringa constant, a phenomenological
two-band model [10] permits a quantitative partitioning between the 5 - and
2 -D(EF ) at 13 C to be obtained, providing insight into metalCO bonding at
the electronic level. These types of 13 C
NMR study have been carried out on 13 CO
chemisorbed (ex MeOH) onto a series of
carbon-supported, fuel cell grade, commercial platinum electrocatalysts having
different average particle size: 2.0, 2.5,
3.2, 3.9, and 8.8 nm, in which both metal
and adsorbation EF -LDOS might be expected to vary with particle size. Shown
3.1.3.4
in Fig. 10 is a typical 13 C NMR spectrum and temperature-dependent T1 data

for the 8.8-nm sample [10]. The straight
line through the origin is characteristic of
the Korringa relationship, indicating the
metallic state of adsorbed CO in this electrocatalyst system.
Table 2 collects all of the 13 C NMR
results together with values of the 5 and 2 -D(EF ) deduced from these results, in addition to the corresponding
infrared data [37] (vide infra). The results
shown in Table 2 indicate that the major part of the total EF -LDOS at 13 C is
from the 2 -like electrons, with 2 D(EF ) being about 10 times larger than
the 5 D(EF ). These values are in good
agreement with theoretical band structure calculations. The contribution to the
spinlattice relaxation rate (1/T1 ) is dominated by orbital and dipolar interactions,
consistent with the dominance of -like
electrons at the Fermi level. Notice that
while 5 D(EF ) is almost a constant,
the 2 -D(EF ) varies noticeably from sample to sample. However, since NMR only
measures the electron density at a given
energy level, that is, the Fermi level, it does
not provide complete information about
the total electron densities for forwardThe 5 - and 2 -D(EF )s deduced by the
two-band model from 13 C NMR data of CO
chemisorbed on Pt electrocatalysts [10] and the
corresponding SNIFTIRS CO vibrational
stretching frequencies [37]
Tab. 2
Size
(nm)
5 -D(EF )
(Ry
molecule)1
2 -D(EF )
(Ry
molecule)1
IR
frequency
(cm)1
8.8
3.9
3.2
2.5
2.0
0.6
0.6
0.6
0.7
0.6
6.4
6.6
6.5
6.8
7.3
2044
2043
2044
2038
2028
161
1/T1
(s1)
162
Pt/13CO
Carbon support
0
0
200
Temperature
[K]
1400 1200 1000
800
600
400
200
200
d (ppm from TMS)

A typical 13 C NMR spectrum and temperature-dependent T1 data
for an 8.8-nm Pt/C sample. The straight line through the origin is
characteristic of the Korringa relationship, indicating the metallic state of
adsorbed CO in this electrocatalyst system. (Adapted from Ref. [10].)
Fig. 10
and backdonation, which are the integrals of the LDOS from the bottom of
the conduction band to the Fermi level.
Nevertheless, as can be seen in Table 2,
the variation in the total EF -LDOS between different samples are mainly due to
changes in the 2 -D(EF ). This suggests
that the changes in chemical properties
of adsorbed CO are mainly determined
by variations in backbonding. In particular, if the Fermi level cuts the tail of the
2* band as it rises, then the increase in
the 2 -D(EF ) would indicate an enhancement in backdonation, and a decrease in
the corresponding vibrational CO stretch
frequency.
In situ subtractively normalized interfacial Fourier transform infrared reectance
spectroelectrochemistry (SNIFTIRS) studies conrm this prediction [37]. They also
yield a linear correlation between CO and

the 2 -D(EF ), Table 2, with a slope of
19 2 cm1 /(Ry molecule)1 and an
intercept of 2167 20 cm1 , as shown
in Fig. 11. The latter value is quite close
to the value expected for free CO, =
2143 cm1 . On the basis of these results,
the following overall physical picture of
carbon monoxide adsorbed onto a platinum electrode can be proposed: Since the
5 orbital of CO is essentially nonbonding,
with the lone pair concentrated on carbon
and the 2 orbital is a much less polarized antibonding orbital, the change in CO
stretch frequency is mainly governed by
changes in 2 backdonation: the higher
the backdonation, the lower the CO stretch
frequency.
The above example demonstrates the
complementary nature of in situ infrared

Correlations between CO and
2 -D(EF ). The straight line is the linear
t that gives a slope of
19 cm1 /Ry1 molecule1 and an
intercept of 2167 cm1 , with a
R2 = 0.98. (Part of the results shown in
this gure has been published in
Refs. [10, 37].)
Fig. 11
2050
(cm1)
2040
2030
2020
6.0
6.5
7.0
7.5
8.0
D2p*(Ef)
[Ry1 molecule1]
spectroscopy and solid-state NMR, in providing insights into the details of surface
chemistry of electro-chemisorbed CO on
platinum. While SNIFTIRS a major research technique of the interfacial electrochemical community provides important qualitative information regarding the
bond strengths of adsorbed CO, it offers
somewhat less insight into the electronic
relocations that occur within the bond, the
electron distribution as a function of sample composition and morphological detail.
For instance, how the bond changes as a
function of particle size is difcult to learn
using infrared alone. However, a 13 C NMR
analysis of the 5 - and 2 -EF -LDOS at
the carbon atom offers a quantitative description of metalCO bonding in terms
of changes in the EF -LDOS. The joint
SNIFTIRS/NMR approach provides new
information on the electronic structure
of the metalsolution interface, which
is relevant to the characterization of industrial fuel cell catalysts, as well as
other interfacial electrochemical systems
of practical importance, by combining
bond strength (IR) and quantitative EF LDOS results (NMR).
Surface Diffusion
The rationale behind using NMR to study
surface diffusion is that the correlation
time for diffusion, , can be related to the
temperature dependence of the NMR relaxation rates observed for diffusing atoms
or molecules [12]. Under certain circumstances, such a relation can be expressed
analytically, permitting a quantitative analysis of the data in dynamic and thermodynamic terms.
When using the temperature dependence of the nuclear spinspin or nuclear
spinlattice relaxation rate to study molecular motion, as is the case with the
surface diffusion we are dealing with here,
there exist so-called strong and weak
collision limits. Different mathematical relationships are needed to describe these
limits. Consider the nuclear spinspin
relaxation rate (1/T2 ) as measured by a conventional Hahn-echo pulse sequence, and
suppose that is the amplitude of the
local eld uctuation responsible for relaxation. Also assume that is the correlation
time for the motion, say a jump, which
causes the local eld to uctuate. The
strong collision limit is dened such that
3.1.3.5
163
164

1/. At t = 0, that is, just after the
rst /2 pulse, which ips all nuclear spins
into the xy plane, all spins are in phase, and
they retain their phase memory until t = ,
when the collision changes the direction of
the local eld. Since is sufciently long
to dephase the spins, they lose all phase
memory before the next collision. Therefore, 1/T2 = 1/ . In contrast, in the weak
collision limit, 1/. Since is relatively short, insufcient dephasing builds
up between two consecutive jumps. When
the spin jumps randomly, many jumps are
needed in order to accumulate sufcient
dephasing, and one has 1/T2 = ()2 .
In summary then, we have for these two
limiting cases:
1
1
1
= for
T2
(8)
1
1
= ()2 for
T2

(9)
and
Fortunately, the function

()2
1
=
T2
(1 + ()2 2 )
(10)
satises both limiting cases. In fact,

Eq. (10) is similar to the typical spectral
density function used for a randomly
uctuating local eld at a frequency ,
and as we show below, the use of these
simple relations leads to information on
both the rates of surface diffusion and the
associated energetics.
The temperature dependence of 1/T2
for 13 CN (a) [6] and for 13 CO (b) [5] on
polycrystalline platinum black at saturation coverage are shown in Fig. 12.
For CO or CN adsorbed onto small
platinum particles, 1/T2 can be expressed as

1
1
1
1
1
=
+
+
+
T2 T2 RL T2 Pt T2 dip T2 diff
()2
1
+
T2
(1+()2 2 ) (N )
(11)
where 1/T2 RL is the temperature-independent rigid lattice contribution, and
1/T2 Pt (= aT , where a is a constant
and T the absolute temperature) accounts for the effects of platinum spins.
1/T2 dip = ()2 /(1 + ()2 2 ), where
is the dipolar eld in Hertz and
= 0 exp(E/kB T ), with 0 a preexponential factor, chosen to be 1013 s, and E
is the activation energy for surface diffusion), arises from dipolar interactions
between neighboring carbon- 13 spins.
1/T2 diff (= 1/N , where N is the number
of jumps needed to dephase the carbon
spins) comes from the eld difference
(positive or negative) between adjacent adsorption sites, due to the presence of a
signicant magnetic eld inhomogeneity
on the particle surface.
The solid lines in Fig. 12 are the t to
Eq. (11) with 1/T2 RL , a, , E, and N as
tting parameters. The results of the ts
(in which 1/T2 RL are 293 and 930 s1 for
CO and CN, respectively) are also shown
in Fig. 12. Clearly, this simple model describes the experimental data very well, for
both 13 CO and 13 CN. It is interesting to
note that the activation energy found for
CO (7.8 kcal mol1 ) is close to that found
for CO on dry alumina-supported Pt
clusters, as well as that found in singlecrystal studies. The results of the ts also
indicate that the activation energy for surface diffusion is higher for CN than for
CO, while the difference in the constant a
suggests, paradoxically, that platinum has
more of an electronic inuence on 13 C
=
RL
+aT +

Temperature dependence of
spinspin relaxation rates of
chemisorbed and (panel a) 13 CN and
(panel b) 13 CO on 10-nm Pt
electrocatalyst surfaces. The solid lines
are the ts to Eq. (11) and the numbers
are the corresponding results of the ts.
(Adapted from Refs. [5, 6].)
Fig. 12
1600
Pt/13CN
E = 10.3 kcal/mol
1400
a = 0.8 s1 K1
w = 966 Hz
N = 1222
1200
1/T2
(s1)
1000
2500
Pt/13CO
E = 7.8 kcal/mol
2000
a = 5.4 s1 K1
w = 1053 Hz
N = 977
1500
1000
500
100
200
300
Temperature
[K]
spinspin relaxation of CO than on that

of CN. The latter may, however, be related
to the fact that the J -coupling is mediated by orbital electrons between 13 C and
195 Pt and is expected to be primarily inuenced via -bonding rather than through
-bonding, for ||2 = 0 at C for the 2
orbital. If the 195 Pt 13 C dipolar interaction
were the only mechanism for 1/T2 Pt , the
ratio in a values would yield rPtCN /rPtCO
equal to 1.9, which is unrealistic. Thus,
while the activation values may indicate an
overall stronger metal surface bonding for
PtCN than for PtCO, the a values suggest that -bonding is stronger in the latter
case. Finally, for the 13 C 13 C dipolar
interaction is essentially identical in each
of the systems investigated, consistent
with a similar adsorbate coverage for

both samples.
Isotopic Labeling for Site
Selectivity
Isotopic labeling is a very useful and
geometrically specic NMR technique
when the natural abundance of the nucleus
of interest is too low to be detected
under normal conditions, such as with
13 C, 15 N, and 17 O. As an example,
we consider here the bonding of a
larger species, acetonitrile, bonded to Pt
in an electrochemical environment [40].
13 C 1/T versus T data for CH 13 CN
1
3
and 13 CH3 CN adsorbed on platinum are
shown in Fig. 13. A linear 1/T1 versus
T plot, reecting Korringa behavior with
3.1.3.6
165

Pt-CN13CH3
2.0
Pt-13CNCH3
1.5
1/T1
(s1)
166
1.0
0.5
0.0
0
50
100
150
200
Temperature
[K]
250
300
Temperature dependence of spinlattice relaxation rates

for chemisorbed CH3 13 CN and 13 CH3 CN. The clear absence of a
Korringa relationship for 13 CH3 CN (solid circles) indicates that the
carbon atom in the methyl group does not acquire any metallic
character. (Adapted from Ref. [38].)
Fig. 13
T1 T = 170 s K, is only obtained with the

CH3 13 CN adsorbate. This proves that at
least the carbon of the CN group is
directly attached to platinum, and that
the 13 C has acquired metallic character.
Note that the value of T1 T = 170 s K for
CH3 13 CN is quite close to that of 13 CN
(135 s K). The T1 T product for 15 N in
CH3 C15 N at room temperature is 224 s K,
showing that the spinlattice relaxation
of 15 N in chemisorbed CH3 C15 N is
much more efcient than in chemisorbed
C15 N, where T1 T varied in the range
30004000 s K. Indeed, after taking into
account the 13C /15N 2.5 factor in
Eq. (3), the relaxation of 15 N appears even
more efcient than that of the metallic
13 C in CH 13 CN. This suggests that both
3
C and N in the CN group are directly
attached to the platinum surface, forming a
side-on parallel orientation with respect to
the PtPt surface bonds. The clear absence
of a Korringa relationship for 13 CH3 CN
(solid circles) indicates that the carbon
atom in the methyl group does not acquire
any metallic character. This implies that

neither CC bonding nor antibonding
orbitals are involved with bonding to the
platinum surface, while both 13 C and
15 N atoms in the CN group are strongly
involved, in contrast to the situation with
the CN itself, in which cyanide has a Cdown on-top bonding.
Potential-dependent EC-NMR
Understanding the nature of the electric
eld (electrode potential) effects on the
electronic structure at the solidliquid
interface is an outstanding issue in
electrocatalysis and in the theory of
the electrical double layer. To illustrate
such effects via NMR, we show in
Fig. 14, the electrode potentialinduced
13 C line shifts for CO (circles) [8] and
CN (squares) [6] on polycrystalline Pt.
These results were obtained under active
external potentiostatic control, and at room
temperature, and the inset shows typical
13 C NMR spectra of 13 CN, recorded at
3.1.3.7

Electrostatic potential
dependence of 13 C shifts for
13 CO (circles) and 13 CN
(squares) chemisorbed on a
10-nm Pt electrocatalyst.
(Adapted from Ref. [4].)
Fig. 14
500
400
[ppm]
13CO, 13CN
Shift
450
13
CO
13CN
350
200
150
0.8
0.4
0.0
0.4
Potential
[V]
different applied potentials. Apparently,

for both adsorbates, there is a linear
relation between the electrode potential
and the line shift the more negative
the potential, the larger the frequency
shift. 13 C becomes more deshielded as the
potential goes to more negative values, and
the slope is about 71 ppm V1 for 13 CO
and 50 ppm V1 for 13 CN. For CO on
Pd, the slope was found to be even larger,
that is, about 136 ppm V1 [8].
A similar response in CO vibrational
frequency, CO , to variations in the electrode potential has been reported [39].
There, CO decreased when the electrode
potential became more cathodic. However,
the origin of this so-called vibrational Stark
effect has been controversial for some
time, since a clear picture of how the
short-range (electronic) bonding and the
longer-range (electrostatic) eld effects at
surfaces are controlled by varying either
the electrode or the surface potential, has
not been available.
EC-NMR provides an additional
technique with which to probe these
questions, for several reasons. First, the
response of 13 C NMR to the electrode
potential has been found to be due

essentially to the Knight shift [8]. The
13 C chemical shift response of CO
to electric elds (in biomolecules) has
been estimated to be at most about
15 ppm V1 , much smaller than the
values found at the electrochemical
interface (see above). In addition, for
CO on Pd, a corresponding change in
the nuclear spinlattice relaxation rate
as a function of applied eld has also
been observed [8], indicating the electronic
nature of the change in NMR shift.
Second, the magnitude of the cleansurface EF -LDOS of the adsorbent is
responsible for the extent of the Knight
shift as well as the vibrational stretch
frequency of chemisorbed CO [25, 36].
That is to say, the higher the cleansurface EF -LDOS, the larger (lower) will
be the 13 C Knight shift [25] (CO stretch
frequency [36]) of CO, after chemisorption.
Both relationships are linear, with
slopes of about 12 ppm/Ry1 atom1
and 4 cm1 /Ry1 atom1 , respectively,
giving a ratio of 3 ppm/cm1 . This
is to be compared with the ratio of
2.8 ppm/cm1 , obtained independently
167
168
from the slopes (71 ppm V1 and

25 cm1 V1 , respectively) of (13 C)
and CO versus electrode potential
relationships for CO on Pt black. Third,
as shown in Sect. 3.1.3.4, variations in CO
can be directly correlated with changes in
2 backdonation that is, the higher the
backdonation, the lower the CO stretch
frequency. When taken together, all these
observations indicate that the potential
dependence of both the 13 C NMR shift
and the vibrational stretch frequency of
adsorbed CO are primarily electronic in
nature, and originate from changes in
the EF -LDOS at the metal surface and
at the adsorbate, induced by electrode
polarization.
3.1.4
Future Perspectives
EC-NMR has made considerable progress

during the past few years. It is now possible to investigate in detail metalliquid
interfaces under potential control, to deduce electronic properties of electrodes
(platinum) and of adsorbates (CO), and
to study the surface diffusion of adsorbates. The method can also provide
information on the dispersion of commercial carbon-supported platinum fuel
cell electrocatalysts and on electrochemically generated sintering effects. Such
progress has opened up many new research opportunities since we are now
in the position to harness the wealth of
electronic, EF -LDOS as well as dynamic
and thermodynamic information that can
be obtained from NMR experiments. As
such, it is to be expected that EC-NMR
will continue to thrive and may eventually become a major characterization
technique in the eld of interfacial electrochemistry.
Acknowledgments
This work was supported by the United

States National Science Foundation (grant
CTS 97-26419), by an equipment grant
from the United States Defense Advanced
Research Projects Agency (grant DAAH
04-95-1-0581), and by the US Department of Energy under grant DEFG0296ER45439.
References
1. A. Zangwill, Physics at Surfaces, Cambridge
University Press, Cambridge, 1988.
2. G. A. Somorjai, Introduction to Surface Chemistry and Catalysis, John Wiley & Sons, New
York, 1994.
3. A. Wieckowski, (Ed.), Interfacial Electrochemistry Theory, Experiment, and Applications,
Marcel Dekker, New York, 1999.
4. Y. Y. Tong, E. Oldeld, A. Wieckowski, Anal.
Chem. 1998, 70, 518A527A.
5. J. B. Day, P.-A. Vuissoz, E. Oldeld et al., J.
Am. Chem. Soc. 1996, 118, 13 04613 050.
6. J. J. Wu, J. B. Day, K. Franaszczuk et al.,
J. Chem. Soc., Faraday Trans. 1997, 93,
10171026.
7. Y. Y. Tong, C. Belrose, A. Wieckowski et al.,
J. Am. Chem. Soc. 1997, 119, 11 709, 11 710.
8. P.-A. Vuissoz, J.-P. Ansermet, A. Wieckowski, Phys. Rev. Lett. 1999, 83, 24572460.
9. Y. Y. Tong, C. Rice, E. Oldeld et al. in Theoretical Modeling of the Solid/Liquid Interface:
Electronic Perspective and Comparison with Experiment (Ed.: W. Halley), American Chemical Society, Washington, DC, 2001.
10. Y. Y. Tong, C. Rice, A. Wieckowski et al., J.
Am. Chem. Soc. 2000, 122, 11231129,
11 92111 924.
11. M. E. Lacey, R. Subramanian, D. L. Olson
et al., Chem. Rev. 1999, 99, 31333152.
12. C. P. Slichter, Principles of Magnetic Resonance, 3rd Enlarged and Updated Edition,
Springer-Verlag, Berlin, Germany, 1990.
13. J.-P. Ansermet, C. P. Slichter, J. H. Sinfelt,
Prog. NMR Spectrosc. 1990, 22, 401421.
14. T. M. Duncan, Colloids Surf. 1990, 49, 1131.
15. M. Pruski in Encyclopedia of Magnetic Resonance (Eds.: D. M. Grant, R. K. Harris),
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
John Wiley & Sons, New York, 1996, pp.

46384638, Vol. 7.
J. J. van der Klink, Adv. Catal. 2000, 44,
1117.
P.-K. Wang, J.-P. Ansermet, S. L. Rudaz
et al., Science 1986, 234, 3541.
P. J. Slezak, A. Wieckowski, J. Magn. Reson.,
Ser. A 1993, 102, 166172.
W. D. Knight in Solid State Physics (Eds.:
F. Seitz, D. Turnbull), Academic Press, New
York, 1956, pp. 93136, Vol. 2.
J. J. van der Klink, H. B. Brom, Prog. NMR
Spectrosc. 2000, 36, 89201.
J. Korringa, Physica 1950, 16, 601610.
J.-P. Bucher, J. J. van der Klink, Phys. Rev. B
1988, 38, 11 03811 047.
R. Hoffmann, Rev. Mod. Phys. 1988, 60,
601628.
P.-A. Vuissoz, J.-P. Ansermet, A. Wieckowski, Electrochim. Acta 1998, 44, 24572460.
Y. Y. Tong, C. Rice, N. Godbout et al., J. Am.
Chem. Soc. 1999, 121, 29963003.
M. S. Yahnke, B. M. Rush, J. A. Reimer et al.,
J. Am. Chem. Soc. 1996, 118, 12 250, 12 251.
B. M. Rush, J. A. Reimer, E. J. Cairns, J. Electrochem. Soc. 2001, 148, A137A148.
H. E. Rhodes, P.-K. Wang, H. T. Stokes et al.,
Phys. Rev. B 1982, 26, 35593568.
29. J. J. van der Klink, J. Buttet, M. Graetzel,

Phys. Rev. B 1984, 29, 63526355.
30. C. Rice, Y. Y. Tong, E. Oldeld et al., Electrochim. Acta 1998, 43, 28252830.
31. J. P. Bucher, J. Buttet, J. J. van der Klink
et al., Surf. Sci. 1989, 214, 347357.
32. J. Friedel, Adv. Phys. 1954, 3, 446507.
33. V. Heine in Solid State Physics (Eds.:
H. Ehrenreich, F. Seitz, D. Turnbull), Academic Press, New York, 1980, pp. 1127,
Vol. 35.
34. Y. Y. Tong, G. A. Martin, J. J. van der Klink,
J. Phys. Condens. Matter 1994, 6, L533L538.
35. G. Blyholder, J. Phys. Chem. 1964, 68,
27722778.
36. Y. Y. Tong, P. Meriaudeau, A. J. Renouprez
10 15510 158.
37. C. Rice, Y. Y. Tong, E. Oldeld et al., J. Phys.
Chem. B 2000, 104, 58035807.
38. J. Wu, C. Coretsopoulos, A. Wieckowski in
Electrochemical Society Proceedings, The Electrochemical Society, Pennington, NJ, 1997,
pp. 426442, Vol. 97-17.
39. D. K. Lambert, Electrochim. Acta 1996, 41,
623630.
40. M. Weinert, A. J. Freeman, Phys. Rev. B 1983,
28, 62626269.
169
170
3.2
EPR Spectroscopy in Electrochemistry

Jay D. Wadhawan, Richard G. Compton,
Oxford University, Oxford, United Kingdom
3.2.1
Introduction
Electron paramagnetic resonance (EPR),

or its equivalent term electron spin resonance (ESR), is an ideal spectroscopic
tool to study the formation of paramagnetic species (such as radicals) during the course of electrolysis, as species
with concentrations as low as 108 M
may be detected [1]. Further, electrolysis has consistently proved to be one of
the most convenient and attractive techniques for the production of interesting
radicals [25]. Additional advantages of
using a spectroscopic approach in the interrogation of complex electrode reaction
mechanisms are that it does not require
the destruction (nor activation) of the sample, and that it permits an inference of
the identity of paramagnetic species that
are formed as intermediates in a plethora
of electron-transfer processes. An example
that illustrates this is the electrochemical reduction of 2,4-dichloronitrobenzene,
studied by Lawless and Hawley [6]. In this
reaction, the 2,4-dichloronitrobenzene radical anion undergoes a thermally-induced
CCl bond cleavage to form the neutral chloronitrobenzene radical, which,
after H atom abstraction from the solvent/electrolytic system, is subsequently
further reduced. The question that arises
is: which of the two CCl bonds is
cleaved? Analysis of the EPR spectrum
of the intermediates revealed that the
ortho-chlorine is removed, leaving the parachloronitrobenzene radical anion that is
detected.
EPR spectroscopy was rst applied to

electrochemistry, albeit at low temperature, in 1958 by Ingram and coworkers [3],
who demonstrated the formation of radical
ion species from the potentiostatic electrolysis of various aromatic compounds in
N ,N -dimethylformamide (DMF). In these
experiments, aliquots of sample were withdrawn during the course of electrolysis,
frozen in liquid nitrogen, and subsequently examined by EPR. It was Maki and
Geske [4, 5, 7, 8] who developed a complementary concept of electrochemical EPR
by studying the solution-phase spectra of
radicals that had been electrogenerated
within the EPR cavity, thereby paving the
way for two forms of electrochemical EPR
methodology, via in situ techniques and
ex situ techniques. In the former, paramagnetic species are generated within the
EPR cavity, whilst in the latter, radicals
that are externally generated are subsequently moved into the EPR cavity. It will
be appreciated that the former technique
is more advantageous than the latter, in
that it may be used in the study of radicals with a shorter lifetime range. Whist
the ex situ technique has many merits,
notably in Alberys use of a tube electrode
with a mathematically well-dened ow of
electrolytes in the study of the kinetics
and mechanisms of a variety of electrode processes [914], this overview will
be exclusively concerned with the more
sensitive in situ approach to electrochemical EPR; the interested reader is referred to
two earlier reviews by Waller and Compton [15, 16] that describe the historical
development of electrochemical EPR.
The ability of the in situ electrochemical EPR technique to provide information
on kinetics and mechanisms of electrode processes is the major reason for
its adoption by electrochemists although
its major limitation is that it requires
3.2 EPR Spectroscopy in Electrochemistry
the presence of a paramagnetic reaction intermediate. Following Maki and

Geske, Adams [17], Cauquis [1822], Kastening [23, 24], and Dohrmann [2528] all
developed cells based upon a variety of cell
geometries, but these did not provide advances in the study of reaction kinetics;
Sect. 3.2.3 covers the in situ electrochemical EPR cell designs that aim to remedy
this, and Sect. 3.2.4 gives examples of the
use of such cells in the analysis of kinetics
and mechanisms. First, however, the underlying theoretical principles of electron
paramagnetic resonance are covered, and
from which, the similarity between this
and its sister technique, nuclear magnetic
resonance (NMR, see Chapter 3.1 of this
volume), may be noted.
3.2.2
Principles of EPR Spectroscopy
This section outlines the theoretical concepts of EPR. The theory is similar to
the related technique of NMR described
in Volume 2 Chapter 3.1. In the following, a basic knowledge of quantum
mechanics will be assumed. Throughout
the following, the symbol B refers to the
magnetic ux density (or magnetic induction eld); the symbol H being reserved
for the magnetic eld strength. In SI units,
H = 4 107 B.
EPR Theory
The origin of EPR spectroscopy lies in the
fact that electrons have both electrostatic
charge and spin angular momentum. The
former is observed by electron-deection
from negatively charged surfaces, in for
example, a cathode ray oscilloscope; the
latter is easily veriable in the classic SternGerlach experiment [29, 30] in
which a beam of S-state silver atoms is
observed to split into two separate beams
3.2.2.1
upon passage through an inhomogeneous

magnetic eld.
The electronic charge (e) together with
the angular momentum (s) confer a magnetic moment () upon the electron, according to Eq. (1),
= ge e s
(1)
in which e is the magnetogyric ratio of

the electron, given by the ratio e/2me ,
me representing the mass of the electron.
The constant ge is referred to as the
electron g-factor and takes the value
2.002319314 for a free electron. The origin
of this factor may be inferred from Dirac
theory; the deviation from exactly 2 reects
the existence of a zero-point energy in
the harmonic oscillating electromagnetic
eld. The interested reader is referred
to Ref. [31] for further details regarding
these phenomena.
The magnitude of the spin angular
momentum vector (s) is given by

(2)
|s| = h s(s + 1)
in which s refers to the spin angular momentum quantum number. Experiment
shows that s = 1/2, and the spin angular
momentum vector is perceived to precess around the direction of the applied
magnetic eld (conventionally labeled the
z-direction); the component of the spin
angular momentum in this direction being h/2 (Fig. 1). It is useful to introduce
the quantum number Ms to label the allowed eigenvalues of sz ,
sz = Ms h
(3)
Evidently, Ms = 1/2. This space quantization of s and will necessarily cause

the two possible electron energy states to
split under the application of a magnetic
eld (the Zeeman Effect); the energy, E
171
172

The precession of the
spin angular momentum vector
about a magnetic eld applied
in the z-direction.
Fig. 1
1h
2
Ms = + 1
2
Ms = 1
2
of a magnetic dipole of moment in the

presence of a magnetic induction, B, is
given by
E = B
(4)
The negative sign indicates that if the
magnetic dipole is aligned parallel to
the applied eld, the interaction energy
will be at a minimum; an antiparallel
alignment corresponds to a maximum
energy interaction. If the eld is applied in
the z-direction, Eq. (4) becomes
E=
ge e Ms Bz
h
Since Ms = 1/2, only two energy states

are possible as given above. In the absence
of a magnetic eld, the two energy states
are necessarily degenerate; application of
a magnetic eld causes the spin states
to adopt different energies, illustrated in
Fig. 2. Note that the lower state (E1 ) has
Ms = 1/2, whilst that antiparallel, upper
state (E2 ), has Ms = +1/2. If the energy
Ms = + 1
2
2phn = gegeBZ
B=0
E
Ms = 1
2
BZ
B
dA
dB
Bpp
(5)
Fig. 2 Splitting of the spin levels of a

free electron in a magnetic eld, of
induction Bz . Also shown is the rst
derivative of the absorption intensity, A,
with respect to B, as a function of B.
of the interaction is large when compared

to kT, the alignment of the great majority
of dipoles would correspond to the lower
energy state. However, the energy of
interaction in the EPR experiment is
considerably smaller than kT.
Transitions between the two spin states
may be induced by the application of electromagnetic radiation of the appropriate
frequency, , such that the applied energy,
2 h exactly matches that of the Zeemanlevel separation, ge e Bz , that is
=
ge e Bz
2 h
(6)
This phenomenon is known as magnetic

resonance; the frequency at which these
transitions occur is called the resonant
frequency. Resonance can be observed in
two different ways; the frequency may be
varied at a xed magnetic eld strength,
or mutatis mutandis the converse. In practice, it is more benecial to utilize the
latter of these two methods. Equation (6)
suggests that the resonant frequency of
a free electron is 9.4 GHz for an applied
magnetic eld of 3400 G (1 G = 104 T).
These approximate conditions are employed to experimentally observe EPR
transitions; the frequency of 9.4 GHz is
in the microwave region, corresponding to
a wavelength 3 cm (so-called X-band).
Occasionally, measurements are undertaken under other conditions, such as
Q-band (corresponding to = 35 GHz,
Bz = 12 500 G, = 0.86 cm) or S-band
( = 3.2 GHz, Bz = 1140 G, = 9.4 cm).
In contrast to NMR spectra, EPR spectra
are recorded as the rst derivative of the intensity, A, of the absorption with respect to
the eld induction, B, and as a function of
B, that is dA/dB versus B, as indicated in
Fig. 2. The reasons for doing this lie in the
phase-sensitive methods of detection employed in EPR, and consequently this way
of recording the spectrum depends upon

the modulation used to amplify the absorption signal (vide infra). The peak-to-peak
linewidth, Bpp , is generally taken as the
abscissa distance between the maximum
and the minimum of the dA/dB curve.
The g-value
Until now only free electrons have been
discussed. In reality, the electron responsible for the resonant absorption is normally an unpaired one within a molecule,
thereby experiencing a local magnetic eld
different from the applied eld, since the
applied eld will induce currents in the
electrons of the molecule. The resonance
condition is consequently modied to
3.2.2.2
ge e (1 )Bz
2 h
(7)
where represents a shielding constant

similar to that in NMR spectroscopy. The
g-factor of the molecule is then given by
the quantity g = ge (1 ).
In addition to the magnetism arising
from the electron spin, there may also
be a contribution from the electronic
orbital motion since atomic orbitals have
associated
angular momenta of magnitude
h ( + 1), in which = 0,1, and 2 for

s-, p-, and d-orbitals, respectively. Hence,
occupied orbitals that are nondegenerate
lead to nonzero angular momenta. An
orbital angular momentum, l, has an
associated magnetic moment
L = e l
(8)
For light atoms and certain linear radicals,

s and l combine via the RussellSaunders
Coupling scheme to give a total angular momentum, j. EPR spectroscopy then
interrogates transitions between the different j levels. Hence, since the Hamiltonian
operator for an electron spin in a molecule
contains a term representing the effect of
173
174
the magnetic eld on the electron spin

magnetic moment, another for spin-orbit
coupling, and another for the interaction
of the applied eld with the orbital angular
momentum, the g-factor can absorb the
effects of the orbital angular momentum
operators, giving rise to an effective Spin
Hamiltonian.
In most molecular radicals, the orbital
contribution is lost (quenched), as the
existence of covalent bonds lifts the
degeneracy of the orbitals, and providing
the electron is in an orbital sufciently
removed in energy from other levels, the
effective g-value will be close to the spinonly value, and the resonance condition
is given by
ge Bz
=
(9)
2 h
in which g is now a unique property of the
whole molecule. The value of the g-factor
will be independent of the magnetic eld
direction only in isotropic systems such as
organic radicals in low-viscosity solutions
in which there is rapid and random
tumbling of the molecules so as to permit
an average g-value to be observed. For the
purposes of understanding liquid phase
EPR spectra generated in electrochemical
EPR experiments, this average value is
sufcient. However, in a large number of
other systems, the g-factor is anisotropic,
varying markedly with the orientation of
the sample (this is a consequence of its
dependence upon the spin-orbit coupling).
In these cases, especially for solid-state
spectra, the g-factor must be regarded as
a tensor [32]. Furthermore, the anisotropy
of the g-tensor may contribute in some
cases to the linewidth of the spectra from
high-viscosity solution-phase radicals.
Linewidths
The linewidth of an EPR spectrum depends upon two instrumental factors, the
3.2.2.3
homogeneity of the magnetic eld and

the frequency of the magnetic induction
modulation. Several other processes affect
the linewidth and these are considered in
turn. It should be noted that the two most
frequently encountered lineshapes observed are Gaussian and Lorentzian. For a
line centered at frequency , a Lorentzian
lineshape is given by the equation
f () =
2T2
1 + 4 2 T22 ( )2
(10)
and a Gaussian lineshape by the equation
f () = T2 2 exp[2 2 T22 ( )2 ]
(11)
where the symbol T2 is dened later. The
Gaussian lineshape is observed when the
line is the superposition of a large number of unresolved individual components,
such a line being referred to as inhomogeneously broadened, and is commonly
encountered in solid-state EPR spectra.
More important is the Lorenztian lineshape, which is often observed for radicals
in solution, when complications due to unresolved hyperne splittings do not occur.
3.2.2.3.1 Spin-lattice Relaxation The transitions between eigenstates induced by
the electromagnetic radiation necessarily
involves a perturbation of the Boltzmann
Distribution Law,

(E2 E1 )
n2
= exp
(12)
n1
kT
In this equation, k and T refer to the Boltzmann constant (1.380662 1023 J K1 )

and the absolute temperature, respectively,
and ni refers to the electron population of
the ith non-degenerate energy level of energy Ei , as given in Fig. 2. At equilibrium,
the lower state is more lled, and resonance initially induces more transitions
from E1 to state E2 than the reverse. It is
expected that the populations will rapidly

equalize, after which, there would be no
resultant energy absorption, even though
resonant transitions would still be occurring. This phenomenon is termed saturation and is observed when high microwave
powers are applied. When low-power microwaves are employed, magnetic resonance transitions are observed and nonradiative relaxation processes occur.
These processes oppose saturation and
help restore the natural order of the thermal Boltzmann Distribution, and they are
termed spin-lattice relaxation phenomena.
The mechanism for spin-lattice relaxation is as follows. All paramagnetic
species in the sample have an associated magnetic eld surrounding them with
which each of the other paramagnetic
species may interact. In liquids, the random molecular collisions that constitute
Brownian motion permit these local magnetic elds to uctuate; a uctuation that
occurs at the resonant frequency will induce a radiationless transition. The spinlattice relaxation is characterized by a spinlattice relaxation time, T1 , which thus effectively controls the degree of saturation.
Since the relaxation determines the lifetime, t, of a spin state, the Heisenberg
Uncertainty Principle relates it to the
uncertainty of the Zeeman eigenvalues,
E1 and E2 , thereby allowing this phenomenon to affect the linewidth of EPR
signals, as these depend inversely on T1 .
E t h
(13)
Clearly, if the spinlattice interaction is

strong, T1 is small, t is small, and E
must be large, leading to a spread in
the energy states between which transitions occur. Hence, absorption effectively
occurs over a small range of magnetic
eld strengths. Mutatis mutandis, if the
spinlattice interaction is weak, narrow
absorption lines occur. The former situation is often encountered in the EPR
spectra of transition metal ions, whilst the
latter is usually found for organic radicals
in solution.
Another form of spinlattice relaxation
occurs when there is a quadrupolar nucleus (vide infra) present in the paramagnetic species. Associated with this nucleus
is an electric quadrupolar eld, which
will interact with gradients in the electric
component of the microwave radiation.
However, as this electric quadrupolar interaction is generally weak, it does not
contribute much towards the relaxation.
Spinspin Relaxation The second major nonradiative phenomenon is
termed spinspin relaxation and is characterized by a relaxation time T2 . The
processes that contribute to T1 also inuence T2 , which controls the linewidth
(vide supra). However, the random forces
that modulate the electron spin energy
levels at low frequencies (below 1010 Hz)
contribute to T2 , whereas they have no
inuence upon T1 . Again, the source of
the uctuating magnetic eld lies in the
anisotropic magnetic interaction within
the molecule, mainly due to the anisotropy
in the g-tensor and to the dipolar hyperne coupling with magnetic nuclei. This
has assumed that the radical (or parent
species) concentration is not too high;
otherwise, spin exchange occurs as a result of radicalradical collisions, thereby
causing a broadening of the EPR lines.
The spinspin process involves the interchange of the spins of the colliding radicals. At very high radical concentrations,
the spin sees an average environment as
it moves from radical to radical, and so the
hyperne splitting is lost and a single line
results. This is the reason solid state EPR
spectra are generally of a single line, as
3.2.2.3.2
175
176

Typical EPR signal
obtained upon reduction of
poly(nitrostyrene) immobilized
on an electrode.
Fig. 3
5G
shown in Fig. 3 for the poly(nitrostyrene)

radical anion.
At high parent concentration, electron
exchange may occur between radical and
parent species, thereby inducing a line
broadening; for example, the extent of such
broadening has been used to calculate a
rate constant for the chemical exchange
between benzonitrile and the benzonitrile
radical anion generated electrochemically
in situ [33] yielding a rate constant of
2 108 dm3 mol1 s1 .
Hyperne Structure
The most useful characteristic of EPR
spectra in the identication of radical
species is the splitting due to hyperne
coupling. This arises from the electron
spin interacting with any magnetic nuclei
in the molecule. Magnetic nuclei, with
nuclear spin I, behave analogously to
electrons when placed into a magnetic
eld. Hence the nuclear magnetic moment
may be written in the form of Eq. 14,
3.2.2.4
N = gN N I
(14)
The gN values are dimensionless and

characteristic of the type of nucleus. N
is the nuclear gyromagnetic ratio. Some
values are listed in Table 1, together with
the corresponding quantum numbers I
and MI , in which the quantum number
MI allows for the space-quantization of
Tab. 1
Characteristics of some nuclei
Electron
1H
2D
6 Li
7 Li
12 C
13 C
14 N
15 N
19 F
23 Na
31 P
Natural
abundance
%
Spin
angular
momentum
99.98
0.02
7.43
92.5
98.9
1.1
99.63
0.37
100
100
100
1/2
1/2
1
1
3/2
0
1/2
1
1/2
1/2
3/2
1/2
g-value
2.0023
5.585
0.857
0.8219
2.1707
0
1.405
0.404
0.567
5.257
1.477
2.263
the magnetic eld and is integral or

half-odd integral in value. This range of
(2I + 1) values can be expressed as the
ClebschGordon series, MI = I , (I 1),
(I 2), . . . (I 1), I .
It is instructive to consider the mechanism by which the hyperne coupling
arises. In a strong magnetic eld, the interaction between an unpaired electron and
a magnetic nucleus acts as a small perturbation E to the Zeeman levels E1 and E2
of the electron spin. This perturbation is
composed of two components,
E = (E)isotropic + (E)anisotropic
(15)
The anisotropic term, (E)anisotropic , represents the dipolar interaction that depends
upon the relative positions of the magnetic moments of the unpaired electron
and the nucleus (e and N ). In single crystals, this gives helpful information
on the crystal geometry; in polycrystalline
substances, it causes line broadening.
In liquid samples, however, since the
molecules are constantly tumbling, the
magnetic dipoledipole interactions average to zero, except for a small amount
that is dependent upon the liquid viscosity
and hence gives rise to line broadening.
The hyperne structure due to radicals
in solution, therefore, results from an
isotropic term, (E)isotropic , which is a
consequence of a nonclassical interaction,
known as the Fermi-Contact interaction.
This interaction arises because of the presence of a small, but nevertheless nite
spin density at the position of the nucleus.
In a strong magnetic eld applied in the
z-direction, this term, referred to as the
coupling constant, a, is proportional to the
square of the electronic wave function at
the nucleus,
8
ge gN e N |(0)|2
(16)
a=
3
Since only s-orbitals have a nite probability density at the nucleus (other orbitals all have nodes in their wave functions there), the contact interaction can
only occur when the electron occupies a
molecular orbital in which there is some
s-orbital character.
Thus, for each nuclear spin state, a separate transition in the EPR spectrum will
I = 1/2
I=1
MI
MI
+1
0
1
+1/2
1/2
MS
+1/2
MS
+1/2
2phn
1/2
1/2
+1
0
1
1/2
+1/2
a
B
(a)
MI +1/2 1/2
(b)
MI +1
Fig. 4 Splitting of an EPR signal by hyperne interaction

between the unpaired electron and a nucleus with spin
quantum number (a) I = 1/2 or (b) I = 1. Also shown are the
EPR spectra anticipated for both cases.
177
178
be observed. This is illustrated in Fig. 4

in which the unpaired electron interacts
with only one nucleus having a spin quantum number, I , of 1/2 or 1, corresponding
to a 1 H or a 14 N nucleus, respectively.
The hyperne coupling constant is the
separation between the two lines, and is
eld-independent, and a characteristic of
the nucleuselectron interaction. During
the electronic transitions, the nuclear spin
quantum number, MI , does not change,
and we may formulate the rst-order
selection rules,
Ms = 1;
MI = 0
The above rules are valid provided the

magnitude of the hyperne coupling is not
too large; otherwise, the spins are partially
coupled and the behavior is more complicated [32]. Hence, two hyperne lines
(a doublet) are detected in the spectrum
resulting from the interaction of the unpaired electron with a proton, and three
hyperne lines (a triplet) when the electron
interacts with a 14 N nucleus. Quantitatively, under rst-order conditions, the
electronic energy levels of a one-electron,
one-nucleus system are given by
E = Ms (ge e hBz + aMI )
(17)
The above can be extended for radicals

containing more than one nucleus, by
splitting the levels E1 and E2 in succession
depending upon interaction with the
nucleus in question. For a radical coupling
to two spin-1/2 nuclei, the energy states
are given by
E = Ms (ge e hBz + a1 M1 + a2 M2 )
(18)
in which a1 and a2 are the two coupling
constants, and M1 and M2 are the nuclear spin quantum numbers. The EPR
spectrum has four lines of equal intensity, and is described as being a doublet
of doublets. When, however, a1 = a2 , the

nuclei 1 and 2 are referred to as being equivalent. Typical EPR spectra are illustrated in
Fig. 5 for the case of two equivalent spin1/2 nuclei and equivalent spin-1 nuclei. In
these cases, since three are certain spin
congurations that are degenerate, some
lines coincide.
In general, n equivalent nuclei with spin
quantum number I give rise to (2nI + 1)
equidistant hyperne lines in the EPR
spectrum. In particular, for an electron
interacting with n equivalent protons (I =
1/2), (n + 1) lines are produced whose
relative intensities may be predicted from
the coefcients of the binomial expansion
of (1 + x)n . These can be recalled most
readily with the help of Pascals triangle,
as shown in Fig. 6. Also shown in Fig. 6
are the relative line intensities for the
coupling of n equivalent spin-1 nuclei
with an electron (giving rise to (2n + 1)
hyperne lines). To illustrate this, Fig. 7
shows the EPR spectrum of the benzene
anion. This species has six equivalent
protons, and the EPR spectrum is clearly
a 1 : 6 : 15 : 20 : 15 : 6 : 1 septet, as would be
anticipated from the above ideas.
It is not immediately obvious how
hyperne splitting arises in the case of
the benzene anion in which the electron
occupies a -orbital that has a nodal
plane coincident with the plane of the
aromatic ring. Similarly, it is not obvious
as to why in the ethyl radical (CH3 CH2 )
the methyl (or ) protons show a larger
coupling constant that the -protons. This
is rationalized below.
Consider and isolated >C H unit with
one electron in the -orbital on the C-atom.
This orbital is perpendicular to the plane
of the three trigonal -bonds. Electrons
in the CH bond will keep apart, as
they will mutually repel each other. This
is represented in Fig. 8 in which two
structures are possible, depending upon

whether the electron spin nearest the
C-atom is parallel or antiparallel to the
electron spin in the -orbital. Exchange
forces [32] cause the structure with the
parallel spins at the C-atom to have slightly
lower energy. Hence, the spins in the
CH bond are slightly polarized so that
I = 1/2
MS
the spin of the electron in the sigma

orbital on the C-atom takes on slightly
more of the character of the -orbital. This
means that the electron at the H atom
must have a small excess of the opposite
spin and this resides in the 1s orbital
used to form the CH bond. The result
is a nite spin density at the H-nucleus
M1
M2
M 1
+1/2
+1/2
1/2
1/2
+1/2
1/2
+1/2
1/2
+1
0
1
M 1
I=1
+2
+1
0
1
2
MS
+1/2
+1/2
2phn
1/2
1/2
1/2
1/2
+1/2
+1/2
1/2
+1/2
1/2
+1/2
1
0
+1
+2
+1
0
+1
+2
M1 +1 0
M1 +2 +1 0 1 2
Splitting of an EPR signal by the hyperne interaction between the unpaired electron
and two equivalent nuclei.
Fig. 5
I = 1/2
1
1
1
1
1
1
2
3
4
5
6
n=0
1
1
10
15
1
3
1
1
3
1
5
15
4
1
10
16
19
16
10
2
1
10
20
I = 1:
5
1
Fig. 6 EPR line intensities for the coupling between two equivalent nuclei when I = 1/2
(Pascals Triangle) and I = 1.
179
180

5G
The EPR spectrum for the benzene anion in

1,2-dimethoxyethane at 80 C.
Fig. 7
Canonical structures of
the CH fragment.
Fig. 8
and, hence, a hyperne splitting is seen.

Quantitatively [32], the hyperne coupling
constant should be proportional to the
-electron density, , on the carbon
atom. This is expressed mathematically
in McConnells equation [34],
a(H) = Q
(19)
where Q is a proportionality constant and

takes a value of about 23 G. The negative
coupling constant that is thus anticipated
is because the unpaired spin density at the
H atom is of opposite sign to that in the
-orbital on the C-atom.
The above can be easily extended to
account for EPR signals at -protons,
since exchange forces are capable of

transmitting spin density through more
than one bond. A more dominant effect, however, is due to hyperconjugation. This mechanism involves the direct overlap of the orbital containing
the unpaired electron with the CH
bonds formed by the -hydrogens,
as illustrated in Fig. 9. The magnitude
of this interaction depends upon the
relative orientation of the CH bond
to the orbital containing the unpaired
electron. If the angle between the two
is , then
R
R
q
C
Fig. 9 The magnitude of the

hyperne coupling between an
R electron and a -proton
depends on the dihedral
R q H
angle, , between the CH
R
R
C
bond and the p-orbital
containing the
R
unpaired electron.
a(H) = A + B cos2
(20)
where A is close to zero and B is about

46 G. In the ethyl radical, the methyl group
can freely rotate and so an average value
is seen. Since the averaged value of cos2
is 1/2, a value of ca. 23 G would be predicted for the coupling to these protons, in
excellent agreement with the experimentally determined value (22.38 G).
EPR Instrumentation
Figure 10 shows a schematic illustration
of an EPR spectrometer. An electromagnet
supplies the adjustable applied magnetic
eld. Between the poles, the eld is homogeneous to one part in 105 106 . The
stability of this uniform eld is achieved
by using a regulated power supply, any
variation in the eld being detected by a
Hall effect device. Klystron oscillators generate microwaves, the frequency of which
is determined by the voltage applied to the
klystron. The stability of the microwave
frequency is achieved using an automatic
frequency control system. The microwaves
travel through waveguides hollow metal
pipes of rectangular cross section with
inside surfaces that are gold plated and
whose dimensions are critically dependent
3.2.2.5
upon the wavelength of the radiation that

is reected off them to the sample in the
EPR cavity. The resonant cavity is located
in the homogeneous part of the magnetic
eld. Within this, standing waves are set
up by reections off the end walls when the
radiation is fed in through a small coupling
hole from the waveguide; Fig. 11 shows the
type of cavity most frequently encountered,
a rectangular TE102 cavity. The standing
wave permits the spatial separation of the
electric and magnetic components of the
microwave radiation. It can be seen that the
magnetic component, which is responsible
for inducing EPR transitions, has maximum intensity at the cavity center and
is therefore concentrated upon the sample, when placed there. The distribution
of the microwave eld within the cavity
is such that the sensitivity prole along
the vertical axis of the cavity obeys a sin2
relationship, with zero sensitivity at the
cavity edges and maximum sensitivity in
the cavity center [35]. The electric component has a node at the cavity center. This
is advantageous, since aqueous samples
in particular exhibit appreciable dielectric loss at microwave frequencies, which
Detector
crystal
Waveguide
Klystron
Sample
tube
Phase-sensitive
detection
Modulation coils
Signal out
Electromagnet
Fig. 10
The basic features of an EPR spectrometer.
Magnetic field
modulation (100 kHz)
181
182

Fig. 11
Sample tube
A TE102 EPR cavity.
Magnetic vector
of radiation
Electric vector
prevents the establishment of a standing

wave; by containing these samples within
a thin at cell in which the solution is
held on the nodal plane, thereby escaping
interaction with the electric component of
the microwave eld, this problem is overcome. The cells used are fabricated from
silica in preference to glass, since the latter
contains paramagnetic impurities observable in the EPR spectrum. As well as the
rectangular cavity described above, cylindrical cavities and microwave helices [36]
have found application in electrochemical EPR.
The quality of the resonant EPR cavity is
measured by its efciency for integrating
microwave energy. A measure of this
efciency is described by the quality factor
(Q-factor),
Energy Stored
Energy Dissipated Per Cycle
(21)
A higher Q implies a greater capacity
for energy storage, and only a narrower
bandwidth of frequencies will be admitted
into the cavity.
An attenuator and a ferrite insulator are
placed between the klystron and the waveguide. The former allows regulation of
Q = 2
the power input, whilst the latter protects

the klystron from reected radiation.
This radiation reaches the detector, a
crystal diode via a magic tee bridge, as
illustrated in Fig. 10. The bridge can be
adjusted (balanced) so that no radiation
reaches the detector if no absorption of
microwaves occurs. At resonance, energy
absorption by the bridge unbalances the
bridge and microwave radiation is detected
by the crystal diode detector.
One nal point to note about the experimental EPR arrangement is that a
second magnetic eld is set up around
the cavity using Helmholtz coils through
which an alternating current is applied,
thereby producing a sinusoidally varying magnetic eld with time. This is
known as eld modulation, and, with a
feedback mechanism acts to prevent saturation from arising. Further advantages
include improvements in the signal-tonoise ratio made by increasing the modulation amplitude, although caution must
be duly exercised, since if the modulation amplitude is approximately equal
to the linewidth, the resolution deteriorates. Hence, modulation amplitude
should generally not exceed about half the
linewidth.
The ideas outlined in this section are

treated in depth in the numerous textbooks on EPR [1, 32, 3741]. In particular,
the book by Symons [37] provides an excellent and stimulating introduction for
chemists. All aspects of EPR instrumentation are thoroughly covered in the work by
Poole [35].
3.2.3
In Situ Electrochemical EPR: Cell Design
In this section, contemporary methods for

in situ electrochemical EPR are described,
with particular emphasis given to the use
of such techniques for the investigation
of electrogeneration kinetics of short-lived
intermediates and the mechanisms of
their decay. The EPR-observable is the
EPR signal intensity. Since there are
many lines in each EPR spectrum, the
EPR signal intensity is strictly the double
integral of the EPR signal with respect
to the magnetic eld [42]. However, this
can often be approximated by the peakto-peak amplitude of the largest line in
the spectrum, provided the peak-to-peak
linewidth (Bpp ) is sufciently small.
A full discussion of approximations to the

double integral may be found in the review
by Goldberg [42].
3.2.3.1
The BardGoldberg Cell
3.2.3.1.1 Cell Design This in situ stationary solution cell developed by Bard
and coworkers [4345], is illustrated in
Fig. 12. It is similar to the in situ at cell
pioneered by Adams and coworkers [17].
It consists of a platinum mesh working
electrode (1 mm grid size) placed in a
at cell, with a U-shaped tungsten rod
counter electrode along the cell edges, and
a silver wire quasi-reference electrode located in the at cell above the working
electrode. Unlike the Adams cell design,
ohmic drop problems were removed by
the location of all three electrodes within
the EPR cavity for electrolyte solutions in
DMF. However, a careful choice of supporting electrolyte was necessary; ClO2
radicals may be produced at the counter
electrode when reductions occurred at the
working electrode in the presence of perchlorate background electrolytes. The use
Pt wire
Reference
Ag wire
Teflon sleeve
Counter electrode
Working
electrode
Glass wall
Schematic diagram of
the BardGoldberg cell.
Fig. 12
183
184
of tetrabutlyammonium iodide as supporting electrolyte mitigated this problem,

since nonparamagnetic iodine is formed
at the counter electrode during reductions [43]. It was concluded that where
radical formation is possible in this way,
it might be necessary to add a nonradical
producing electroactive substance so that
the production of secondary radicals is prevented. Later work [45] conrmed these
problems did indeed arise from radicals
produced at the counter electrode.
Typical
Applications Experiments performed with stable radical ions
in DMF, such as those produced from
the reduction of nitrobenzene, azobenzene, and 9,10-anthraquinone, indicated
that the observed signal was stable for
2030 s after generation by a current step,
and after which natural convection set in,
suggesting that kinetic measurements are
possible from the magnitude of this constant steady state EPR signal, within the
20 s time limit [43].
It was further shown that the homogeneous kinetics of the radical ions
produced electrochemically could be inferred from a current pulse experiment in which the EPR signal is measured during and following the current pulse. The signal transients were
then analyzed in terms of working
curves for rst-order radical decomposition, second-order radical ion dimerization, and second-order radical ion-parent
dimerization [44]. In this manner, the
mechanism of the reduction of olens was
deduced, illustrated best by the dimerization of dimethyl fumarate radical anions.
Analysis of the EPR signal transients
generated by a current pulse yielded
rate constants of 160 26 dm3 mol1 s1
for radical-anion dimerization and 50
30 dm3 mol1 s1 for radical anion-parent
3.2.3.1.2
dimerization [45]. Comparison of these

data with the rate constants deduced
from conventional cyclic voltammograms,
160 40 dm3 mol1 s1 [45], suggesting
that the mechanism is, in fact, a radicalanion dimerization,
R + e R
R + R R2 2
3.2.3.2
The Allendoerfer Cells
3.2.3.2.1 Cell Design Allendoerfer and

coworkers [4649] pioneered the electrochemical EPR cell design to exhibit maximum sensitivity to unstable radicals.
These state-of-the-art in situ cells are probably the most sensitive for the detection
of short-lived (105 s) paramagnetic intermediates. Two designs have emerged,
one using a TE011 cylindrical EPR cavity [4648], the basis of which is illustrated
in Fig. 13, and a cell based on a loopgap resonator, Fig. 14 [49]. These designs
and their applications to the study of radicals with small lifetimes are discussed in
turn.
The Coaxial Cell The cylindrical TE011 cavity is modied by centering a

metallic rod along the axis of the cavity. A
quartz cell contains this cylindrical metallic conductor, and the liquid sample is
contained within the unlled volume of the
cell. For conventional EPR spectroscopy,
the sensitivity of such a coaxial cavity was
shown to be at least as good as that of a
rectangular TE102 cavity for lossy samples;
but as an in situ cell for electrochemistry, a
cell based upon the coaxial design exhibits
greater sensitivity compared with in situ
cells based upon the conventional EPR at
cell due to the signicantly larger electrode
area that can be accommodated when the
3.2.3.2.2

A schematic diagram of the
Allendoerfer coaxial cell cavity design. A
cross section of the TE011 cylindrical
cavity (A) is shown. A metal cylinder
(B) is located along the axis of the cavity
within a silica sample tube (C).
Fig. 13
central conductor of the coaxial cavity acts

as the working electrode.
Initial experiments [47] involved a central conductor of 6 mm diameter and
50 mm length. One problem that arose was
that the resonant frequency of the coaxial
cavity was greater that the 9.09.8 GHz
range of the employed klystron. To
overcome this, the resonant frequency was
lowered by surrounding the central metal
rod with a material of suitable relative permittivity, such as Teon(R) . It was shown
that for a solid central conductor of good
conductivity, the cavity Q-factor (a direct
measurement of the EPR sensitivity) was
not signicantly different from that of the
TE011 cavity employed.
To overcome some problems associated
with using a solid central conductor,
the inner part of the coaxial cavity was
constructed from a nely wound shallow
pitched helix, which stands freely against
the inner wall of a quartz test tube (internal
diameter 6 mm). In this form, the standing

microwave eld sees only the portion
of solution between the helix and the
inner wall of the quartz tube and does not
penetrate the inside of the helix, as shown
0.4 mm od
0.3 mm id
5 mm
Working
electrode
Loop-gap
resonator
0.25 mm
0.8 mm od
0.6 mm id
A schematic illustration
of the active region of
Allendoerfers electrochemical
loop-gap resonator geometry.
Fig. 14
1 mm
185
186

The EPR-visible part of the
Allendoerfer coaxial cell is that between
the wire helix and the quartz sample
tube (shaded region). The inner part of
the helix is free to house the reference
and counter electrodes without
interference with the EPR.
Fig. 15
Quartz
Wire
Inactive volume
1 mm
in Fig. 15. Since there is no penetration of

microwaves within the helix, any material
located here does not affect the EPR signal
(or sensitivity), allowing the counter and
reference electrodes to be placed here. The
counter electrode is a platinum cylinder
placed centrally along the cavity axis such
that the current ow is radial, and so
ensuring that there is a uniform current
density over the whole cylindrical workingelectrode surface. The uniform current
density meant that potential control of
the whole electrode can be achieved
by monitoring a single point with a
Luggin capillary.
3.2.3.2.3 Application The surface area of
the working electrode was estimated to
be about 22 cm2 , signicantly larger than
in previous electrochemical EPR cell designs. The cell resistance was found to
be 13 - when lled with 0.1 M tetrabutylammonium perchlorate in DMF [47], a
value that compares favorably with the
510 k- estimated by Bard and Goldberg [43] for an in situ cell based upon
a conventional EPR at cell. Electrochemical measurements made with this
cell design were shown to be free from
ohmic distortions, even when used with
poorly conducting electrolytes, and that
the cyclic voltammograms observed from
the in situ cell were indistinguishable

from those obtained from normal polarographic cells, except for the large
increase in current due to the large
surface area. It was estimated that the
uncompensated solution resistance was
less than 1 -, even when highly resistive
nonaqueous solvents were employed [47].
Radicals as short-lived as 105 s were
anticipated to be amenable to study,
providing a high concentration (approximately 0.1 M) of electroactive material
was used; in agreement with previous estimates [26] of the shortest-lived radical,
it would be possible to study using in
situ techniques.
A modication of the Allendoerfer cell
has been described by OhyaNishiguchi
and coworkers [5052], whereby lowtemperature studies were undertaken on
the electrochemistry of twenty aromatic
compounds. The low temperature was
achieved by placing the electrolysis cell
within a temperature control Dewar. However, it should be noted that this lowtemperature in situ cell did not use a
reference electrode.
In the study of short-lived radicals, the
presence of efcient hydrodynamic ow
is essential to sustain a constant supply
of electroactive material to the workingelectrode surface, and hence ensures a

The Allendoerfer
Carroll cell as modied to
provide the capability for
solution ow.
Fig. 16
Reference electrode
Luggin capillary
Quartz tube
Auxiliary electrode
Working electrode
Teflon flow
baffles
Flow
steady ux of radicals. Carroll adapted the

Alendoerfer cell to this end [46, 48]. The
changes to the helical arrangement are
illustrated in Fig. 16. Solution is prevented
from owing into the central EPR inactive
volume of the cell by a complex series
of bafes. The value of ow coupled
to the Allendoerfer cell was shown by
the electroreduction of nitromethane in
aqueous solutions; the radical anion of
nitromethane was observed and this was
shown to have a lifetime in the order of
10 ms. To date, this is the shortest radical
lifetime observed by electrochemical EPR,
although it has been calculated [50] that
such a cell should be capable of observing
radicals with lifetimes of 105 s.
The Loop-gap Resonator Cell
The loop-gap resonator was introduced
into EPR spectroscopy by Froncisz and
Hyde [53], who used an arrangement
shown in Fig. 17. Its design and applications have been reviewed [54]. The loop
behaves as an inductive element, whilst
the gaps act as capacitative elements. It
has been found [53] that increasing the
number of gaps permits EPR spectroscopy
to be undertaken on larger samples and at
3.2.3.2.4
Flow
higher frequencies. Loop-gap resonators

have a number of advantages over conventional resonant cavities [53], notably
their smaller size and greater observed
signal intensity. For electrochemical purposes, these are benecial; the former lead
to smaller electrolysis currents and concomitantly less potential control problems,
whilst the latter allows enhanced sensitivity on the basis of the total number of
spins (that is spin concentration multiplied
by sample volume), and is illustrated in
Fig. 17 for the case of a small sample of
, -diphenyl--picryl hydrazyl (DPPH).
Allendoerfers loop-gap resonator cell
design [49] is illustrated in Fig. 13. The
working electrode comprises a gold wire
(although a variety of other materials
were also found to work equally as well)
placed inside the active volume of the
resonator; insertion into the resonator in
this manner was found to have no deleterious effect, and little change in the
resonator Q-factor. However, since the
EPR active volume decreases, the energy
density increases; hence, to avoid saturation, the microwave power must decrease.
A platinum needle acted as the counterelectrode. For the purpose of generating
187
188
iii
iv
ii
R
iii
Z
i
ii
ii
ii
(a)
TE102
X40
(b)
Loop gap
4G
(c)
(a) The loop-gap resonator as used by Froncisz and Hyde. The

principle components, (i) loop, (ii) gaps, (iii) shield, and (iv) inductive
coupler are shown. The critical dimensions, Z, resonator length, r, resonator
radius, R, shield radius, t, gap separation, and W, gap width are also
indicated. The sample is inserted into the loop (i) through the coupler (iv).
The microwave magnetic eld in the loop is parallel to the axis of the loop
and (b, c) Comparison of EPR spectra obtained of a point sample of DPPH
measured at X-band in a rectangular TE102 cavity and in a loop-gap resonator
of dimensions r = 0.6 mm, Z = 5 mm. The incident power was held
constant at a low, nonsaturating level. The loop-gap resonator yields 37 times
greater signal than the TE102 cavity.
Fig. 17
radicals for EPR spectroscopy, this twoelectrode arrangement was adequate, with
the platinum electrode acting as a quasireference electrode in aqueous chloride.
However, for the purposes of in situ electrochemical EPR, this arrangement was
rened to accommodate a three-electrode
conguration employing a Ag|AgCl reference electrode.
3.2.3.2.5 Application Allendoerfer and

coworkers [49] observed the time-dependent formation and post electrolysis decay

Schematic illustration of the in
situ channel ow cell.
Fig. 18
of the dianion radical of para-nitrobenzoic

acid in water at pH 11. It was found that
the radical decay kinetics were not a simple
function of pH or time.
3.2.3.3
The Channel Electrode Cell
3.2.3.3.1 Cell Design Compton and

Coles designed a channel ow cell that
could be used for in situ electrochemical
EPR [55]. The design of this demountable
ow cell, constructed in synthetic silica, is
given in Fig. 18. A rectangular metal foil
(such as platinum), cemented onto a silica plate, comprises the working electrode,
with a ne lead out wire at the rear making
the connection to the electrode. The whole
cell is sealed together with a low-melting
wax, which was found to be non-lossy. The
cell is connected to Polytetrauoroethane
(PTFE) tubing for connection into a ow
system, and is held within a TE102 resonant cavity, accomplished by placing it
inside a silica tube that runs right through
the cavity and xing its position with
PTFE spacers. Movement of this supporting tube, which also serves to protect the
cavity from any leakage, may nely adjust
the cell position within the EPR cavity. Optimally, the cell is positioned at the point
of maximum EPR sensitivity, which corresponds to the center of the EPR cavity,
with the plane of the electrode parallel to
the electric component of the microwaves
standing wave. For an empty cell without electrode, the cavity Q-factor is 6400,
which is lowered to 2500 when lled with
an aqueous 0.1 M electrolytic solution. An
empty cell with an electrode present has
a Q-value of 4600, conrming that microwave losses due to the lead out wire are
acceptably low, whilst the Q-value when

this cell is lled as before is 1200 a value
that is typically observed for conventional
aqueous sample EPR methods.
The ow system used in conjunction
with the channel unit is capable of
a variable laminar ow, established by
having a Reynolds Number, Re < 102
(see Volume 2, chapter 2.2 (by Fisher), or
equally Volume 3, chapter 3.5 (by Mount).),
employing ow rates in the range 104 to
101 cm3 s1 , achieved using a gravity-fed
system (so as to avoid variation in pumping
pressure [56]) together with capillaries
of different diameters and changes in
reservoir height. A platinum gauze counter
electrode is placed downstream of the
working electrode and a saturated calomel
or Ag|AgCl reference electrode upstream,
both outside the cavity. The ow within
the channel cell is parabolic, provided
that there is a sufciently long lead-in,
189
190
that is if the lead-in length is given by

0.1 h Re, where h is the half-height
of the channel [57]. This was veried
experimentally by the observation of plug
ow using a colored solution [55]. Edge
effects were found to be negligible if the
electrode size is less than 4 mm wide
in a 6-mm channel. Further, because of
the electrolyte resistance in the narrow
gap through which solution ows [58], the
potential may be slightly non-uniform [23,
59]. This effect may be anticipated to
perturb the current density unless the
distance over which radicals are formed
is small compared with the total distance
over which radicals are detected, thereby
encouraging the use of small electrodes.

x = 0
(h y)2
1
h2
1/3
ilim = 0.925nF cbulk D 2/3 Vf

2/3
2 c(x, y)
c(x, y)
x
=0
y 2
x
(22)
where D is the diffusion coefcient of the

electroactive species, c(x, y) its concentration, x, the distance along the channel
starting from the upstream edge of the
electrode, y the distance normal to the electrode starting from the electrode surface,
and x , the velocity of the solution in the xdirection. Equation (22) reects the effects
of convection in the direction of ow and
diffusion in a direction (y) perpendicular
to this. Diffusion in the direction of ow
can be shown [62] to be negligible for the
case of macroelectrodes, although this is
not the case for microelectrodes [63]. Since
the velocity prole is parabolic,
(23)
where 0 is the mean solution velocity

at the center of the channel. The velocity prole may be simplied for the
case of macroelectrodes by employing the
Leveque Approximation [64], (y/2h) 1,
yielding,
y
x 20
(24)
h
Using this linearized form of the velocity
parabola, the Levich equation [57] can be
deduced for the case of a chemically
and kinetically uncomplicated electrontransfer process,
(h2 d)1/3 wxe

3.2.3.3.2 Application Channel electrodes have well-dened hydrodynamic properties [57], and the laminar ow convectivediffusion equation describing the mass
transport regime (see Chapter 2.2 of this
volume) within this cell at steady state for
large electrodes is given by [60, 61]
(25)
where d is the channel width, w is

the electrode width, xe is the electrode
length, and Vf (= 40 dh/3) is the volume
ow rate.
Two electrochemical EPR methodologies may be envisaged using a channel
ow cell for the determination of kinetic
and mechanism of radical decay. First is
the measurement of the steady state EPR
signal as a function of the electrode current and ow rate, and using variable ow
rates to probe the radical lifetimes. Alternatively, if the electrogenerated radical is
unstable, transient EPR signals may be
recorded (at constant magnetic eld induction) when the working electrode is
open-circuited, after a steady state has
been achieved. These are considered in
turn for the case of a radical reacting
via rst-order kinetics. Physical application of these methods is reserved until
later, Sect. 3.2.4.
For the steady state case, the EPR
signal intensity, S, can be calculated by
convoluting the sin2 sensitivity of the
cavity with the number of spins. Figure 19

The coordinate system
dening the location of the
channel electrode in the
EPR cavity.
Fig. 19
Cavity length
y=0
xu
shows the coordinate system dening the

position of the electrode relative to the
EPR cavity. xc denotes the center of the
cavity and x = 0 is the position of the
upstream edge of the electrode. The cavity
length, l = xu + xd , where xu and xd
denote the x-coordinates of the upstream
and downstream edges of the cavity,
respectively. Thus, xc = xd l, where l is
typically 24 mm and xc is 15 mm. The
sin2 sensitivity prole allows maximum
sensitivity at the center of the cavity
(x = xc ), whilst the sensitivity at the cavity
edges (xu and xd ) is effectively zero.
Thus, integrating the spins throughout
the cavity,

xd
2 (x xc )
S = S0
sin

xu

2h
c(x, y)dy dx
(26)
where S0 is the signal due to one mole of

the EPR active species located at the center
of the cavity, and the concentration of
the paramagnetic species can be deduced
from Eq. (22). Since the radical species is
generated at the electrode, the integral can
be shortened to

xd
(x xc )
S = S0
sin2

0
2h

c(x, y)dy dx
(27)
0
This equation can be solved either analytically [55, 58] or numerically [6568] using
the Backward Implicit approach [6971]
for the case of macroelectrodes at steady
y = 2h
Flow
xc xe
xd
state to yield a current|EPR signal|ow rate

relationship for two cases. First, where
there is uniform concentration of radical at the electrode surface, and second,
where there is a uniform current density at the electrode. The rst case holds
at all points on a reversible voltammetric wave and for an irreversible wave if
the current is diffusion-limited; the second
case holds near the foot of an irreversible
wave. In both cases, it was found that the
steady state EPR signal for a stable radical
followed the following dependence upon
the mass transport-limiting current and
ow rate,
i
S 2/3
(28)
Vf
This dependence was experimentally veried for the case of the reduction of
uoroscein to the semiuorescein radical
anion [55]. For electrogenerated radicals
that subsequently decay by rst-order kinetics, the dependence of S upon the rate
constant of the decay, k, is most conveniently expressed in terms of an EPR
detection efciency, MK , given by
2/3
MK =
SVf
(29)
i
where K is the normalized rate constant
for the radical decay,
1/3

4h4 2 d 2
(30)
K=k
9Vf2 D
For a stable radical, M0 is constant.
Figure 20 shows how the ratio MK /M0
191
1.0
MK
M0
192
0.5
0.0
0.0
2.5
5.0
7.5
10.0
K
The variation of the EPR detection efciency, MK , with the
normalized rate constant, K, for different sized electrodes (+, 2 mm;
, 3 mm; O, 4 mm) located centrally in the cavity.
Fig. 20
varies with K for an electrode positioned at

the center of the EPR cavity for electrodes
of practical dimensions. In these cases,
the EPR signal and the Faradaic current
are measured experimentally, for a range
of ow rates, permitting the inference of
K from the appropriate value of MK in the
working curve (Fig. 20). It follows from the
denition of the normalized rate constant
2/3
that a plot of K against Vf
allows the
deduction of the rate constant for the decay
process, k.
In Fig. 21, the quantity MK is normalref obtained
ized with respect to the value MK
when the electrode is located at the cavity
center. This gure describes the EPR signal dependence upon location of a 5 mm
electrode within the EPR cavity. For high
rate constants (K 5.0), the signal follows the sin2 sensitivity prole of the
EPR cavity, since all the radicals are found

adjacent to the electrode surface and decay before they can be transported away
from the electrode; in contrast, at lower
rates of decay, the curve deviates from
sin2 behavior, and a maximum is observed
upstream of the cavity center. Hence, although the highest radical concentration
is at the electrode surface, by generating
radicals above the maximum sensitivity, a
greater number of radicals occupy the EPR
active area.
Attention is now turned to the use of
EPR signal transients in the deduction
of radical decay kinetics. The transient
signals may be experimentally observed
(at constant magnetic eld induction) by
either open-circuiting an electrode that
had been previously held at a potential
at which Faradaic processes occur, or

1.5
M 0ref
MKref
1.0
0.5
0.0
0.0
0.5
1.0
xu/I
ref
The variation of Mref
k /M0 for a 5 mm electrode with position of the
electrode in the cavity for different rate constants (K = 0.0, ; K = 1.0, O;
K = 5.0, ; K = 7.5, ).
Fig. 21
by stepping the potential between two

dened values. The former method is normally adopted, since in the latter method,
contributions to the radical decay may occur from, for example, the reoxidation
of a radical-anion if it is generated by
an electrochemically reversible reduction.
Theory describing the transient EPR signals associated with the growth (when the
current is stepped up from zero) or decay (when the current is stepped down to
zero) of the electrogenerated radical upon
a galvanostatic step at the working electrode, for rst-order kinetics is given elsewhere [72]. For macroscopic electrodes,
the relevant convective-diffusion equation is
C(x, y)
c(x, y)
2 c(x, y)
x
=D
t
y 2
x
kc(x, y)
(31)
where k is the rst-order rate constant. The

temporal EPR signal for a growth transient
can be shown to be,
S( ) = S( )
1 bn (K + p)n Jn

2n 5
p
n=0 ?
+
3
3
L1
bn K n J n

2n 5
n=0 ?
+
3
3
(32)
in which is the normalized time variable,
1/3
2/3
9Vf D
= t 4 2 2
(33)
4h d
L1 represents the inverse Laplace Transformation (variable p), and ?(x) is the
Gamma function [73]. The coefcients bn
are given by

1
1
?
n
+
3(n2)/3
3
3
bn =
n!
n=0

2
(n + 1)
(34)
sin
3
193
and Jn are given by the expression

xd
x xu
sin2
Jn =

0

x 2(n+1)/3
x xe 2(n+1)/3

x xe
x
U
d
(35)

where U(x) is the unit step function.
Theoretical growth transients for various
values of normalized rate constant are
shown in Fig. 22 for an electrode located at
the cavity center. For fast mass transport
(or sluggish kinetics), the transient is
governed by both the mass transport of
material throughout the cavity and by
the kinetic decay or the growth of the
radical. As the normalized rate constant,
K, increases (or as the mass transport
decreases), the time taken for the transient
to reach a steady state value decreases,
and the transient changes shape until they
attain a simple exponential form, which
for a decay transient is

S(t) = S(t = 0) exp(kt)
0.5
0.0
0.5
1.0
1.5
t
Calculated transients for a 4 mm electrode located at the cavity center
for different values of K (K = 0.0, O; K = 0.1, ; K = 1.0, ; K = 10.0, ;
K = 20.0, ).
Fig. 22
(36)
This can be rationalized on the basis of the

decay within a convection-free reaction
layer immediately adjacent to the electrode surface: the paramagnetic species
are assumed to decay before they can diffuse sufciently far from the electrode for
them to experience a signicant amount
of transport by convection. Experimental
transients obtained from the reduction
of uorescein to the semiuorescein radical anion, and from the reduction of
2-nitropropane veried these results [58,
72, 74].
This cell has been applied successfully
for in situ electrochemical EPR, deducing
both kinetics and reaction mechanisms,
for radicals with lifetimes typically in the
range 10 to 100 ms: Sect. 3.2.4 will examine its application to ECE [58], DISP1 [74],
EC [68], and comproportionation reaction
mechanisms [7577]. Other applications
include the study of an ECEEE mechanism
1.0
S(t)/S(t )
194
for the reduction of nitromethane in aqueous alkaline solution [78]. One last point to
note is that since the cell is constructed of a
transparent substance, silica, illumination
of the electrode is possible, permitting photoelectrochemical EPR [55, 6567, 7981].
3.2.3.4
The Albery In Situ Tube Electrode
Cell
3.2.3.4.1 Cell Design Albery and coworkers [914] used tubular electrodes for ex
situ electrochemical EPR experiments.
The tubular electrode is equivalent to the
channel electrode in all respects, except
that the cross section is circular rather
than rectangular [82, 137]. Like the laterdeveloped channel ow cell, this setup
(shown in Fig. 23) permits the interrogation of electrode reaction mechanisms of
relatively long-lived radical species, [914]
since the convective-diffusion equations
are mathematically well dened, which at
steady state are given by Eq. (37)
2

2 c 1 c
0 1
+
r 2
r r
(37)
where 0 is the velocity of ow at the center
of the tube, r0 is the radius of the tube, r is
the radial distance from the center of the
r
r0
c
=D
x
tube, x is the distance down the tube, D

is the diffusion coefcient of the species
generated on the electrode, and c is its
concentration. However, recognizing that
an ideal electrochemical cell for both
kinetic and mechanistic studies is of an
in situ design with the electrode located
within the EPR cavity for maximum sensitivity, with well-dened and calculable
ow, experiments were undertaken using
an in situ tubular electrode [83]. Figure 24
shows how the sensitivity of a standard
rectangular TE102 cavity to radicals varies
with distance through the EPR cavity ().
With the tube electrode outside the cavity
(Curve a), the expected sin2 dependence
is observed [35]; moving the tube electrode into the cavity (Curve b) effects not
only a lower sensitivity, but also a grossly
distorted sin2 curve. Furthermore, the sensitivity is negligible in the vicinity of the
electrode, irrespective of the placement of
the latter within the cavity. It was found
that the sensitivity of the EPR cavity is not
distorted if the tube electrode is replaced
by a semiannular tube, as shown in Fig. 24
(Curve c). With the semiannular tube electrode positioned centrally within the EPR
cavity, the expected sin2 sensitivity prole is observed with the electrode centered
at the maximum in the sensitivity curve,
r0
The semiannular tube

electrode as used by Albery and
coworkers. The ne-hatched
segment is made of platinum
and the course-hatched
segment of Teon . The
parabolic velocity prole is
shown. Typical dimensions are
radius, r0 = 0.5 mm and the
electrode length, xe = 2.0 mm.
Fig. 23
xe
195
d
50
S/%Mn2+
196
25
a
c
b
0
0.00
0.25
0.50
0.75
1.00
x/l
Variation of the sensitivity of the cavity with distance through
the cavity. Curve a is obtained with no electrode inside the cavity;
Curve b is obtained with a tube electrode consisting of a complete
annulus at the cavity center. Curves c and d are obtained with an
electrode at the center of the cavity consisting of a half annulus and an
annulus with a missing sector, respectively.
Fig. 24
albeit reduced by a factor of two, but nevertheless adequate for EPR measurements.
Curve d in Fig. 24 shows the sensitivity
when an annular electrode with a sector
subtending 10 has been removed. The
sin2 function is now distorted, but the sensitivity at the electrode is enhanced by a
factor of three when compared to Curve a.
The difference between a complete annular electrode and a part-annular one is
that in the former, the magnetic component of the standing microwave eld
induces eddy currents in the complete circle and therefore is unable to penetrate to
the electrolyte solution. These eddy currents are minimized using an incomplete
annulus. In the case of the semiannular
electrode, there is a greater loss of sensitivity, but the standing microwave eld is
preserved.
Application For a semiannular
tubular electrode, the limiting current is
given by the appropriate form of the Levich
3.2.3.4.2
equation [10],
2/3
1/3
ilim = 2.75nF D 2/3 xe Vf cbulk
(38)
in which xe is the electrode length and Vf

the volume ow rate. Equation (38) was
veried by experimental observations on
the oxidation of N ,N ,N ,N -tetramethylpara-phenylene diamine (TMPD) in aqueous sulfate solution. The diffusion coefcient calculated from these data was in
agreement with literature values, showing
the semiannular tube electrode to be a
satisfactory hydrodynamic electrode.
As with the channel electrode cell,
and following from the previous work
undertaken using the ex situ tube electrode
arrangement, the EPR signal (S) is a
function of the diffusion-limited current
(i) and the volume ow rate,
S = S0 Ix/ ( xe )2/3 r02
i
2/3
nF D 1/3 Vf
(39)
where S0 is the EPR signal from one

mole of spins at the center of the cavity,
and Ix/ is a numerical factor that arises
from convoluting the concentration prole
with the sensitivity of the cavity; for the
electrode just outside the cavity, I0 = 0.53,
while when the electrode is at the cavity
center, I1/2 = 0.47. Experimental results,
for the study of radicals in the coats
of modied electrodes [83, 84], veried
2/3
the predicted S i/Vf
dependence,
suggesting that like the channel electrode
EPR ow cell, this in situ cell is easily
amenable to the study of electrode reaction
mechanisms. However, it suffers from
several disadvantages when compared to
the channel cell. First, the Albery cell is not
readily demounted, making examination
and polishing of the electrode difcult.
Second, it is uncertain whether eddy
currents are induced in the electrode
by the 100 kHz modulation, rendering
areas of the electrode EPR-insensitive.
Third, the channel electrode does not
suffer as large an ohmic drop. Fourth,
irradiation of the electrode is not possible,
rendering this in situ cell inadequate
in the study of photoelectrochemical
phenomena.
3.2.3.5
The In Situ Coaxial Flow Cell
3.2.3.5.1 Cell Design Waller and Compton [85] effectively mimicked the Allendoerfer coaxial design (vide supra), whilst
simultaneously maintaining the mathematically well-dened laminar ow of the
channel ow cell. This improved cell for
electrochemical EPR [85] allowed an improvement in the channel cell regarding
lifetimes of radicals amenable to study,
whilst retaining the hydrodynamic ow
that is essential for the investigation of
electrode reaction mechanisms.
The cell itself (see Fig. 25) comprises a

TE011 cylindrical cavity that is converted
into a coaxial cavity by the addition of
a smooth, polished copper rod that is
positioned centrally within the cavity. The
rod itself (diameter 9 mm) is located within
a precision-bore silica tube, so that there
is an annulus (of size 2h) around the rod,
constrained to be of uniform thickness by
a ne nylon thread running the length
of the cavity. Typically 2h 100 m, but
this is adjustable by polishing the rod.
The central 4 mm of the copper rod is
insulated from the rest of the rod, as this
part is the working electrode. To ensure
a large negative potential window, this
central part of the rod was plated with
mercury. As in the case of the channel cell,
this cell was connected to a reservoir-fed
gravity ow system, with a platinum gauze
counter electrode located outside the cavity
and sufciently away downstream so as to
preclude counter electrode products from
diffusing into the cavity, and an upstream
reference electrode. Electrolyte solutions
ow through the annular gap between the
copper rod and the silica tube. Flow rates
within the ranges 104 to 101 cm3 s1
were obtained in this manner. Since the
ow pathway is narrow and resistive,
the potentiostat was modied so as to
provide high voltages to drive the current
through the channel. Electrolyte resistance
in the narrow gap through which the
solution ows may lead to a non-uniform
potential difference (vide supra), affecting
the voltammetry only in poorly conducting
nonaqueous solutions using low currents.
Like the Allendoerfer coaxial cells (vide
supra), it was found that the resonant
frequency of the cavity shifted above
that of the empty cavity and outside
the klystron tuning range as a result of
the reduced effective cavity size produced
by the addition of the copper rod. As
197
198
35
A
B
115
40
D
E
40
35
15
55
F
G
11
The in situ coaxial electrochemical EPR cell. A, Teon

annulus, B, Teon insulator, C, mercury-plated copper
electrode, D, copper, E, TE011 cylindrical cavity, F, Teon
sheath, G, precision-bore silica tubing. The numbers shown
represent the dimensions in mm.
Fig. 25
above, this problem was eliminated by

partially lling the cavity with a material of
appropriate relative permittivity (such as a
PTFE shielding around the silica tubing),
shifting the resonant frequency back to
typically 9.4 GHz.
The sensitivity of this modied cavity, as
observed with a small crystal of DPPH, was
found to follow a sin2 function along the
length of the cavity and at a xed distance
from the copper rod, as anticipated theoretically [35]. However, the EPR sensitivity
was observed to vary in a cos2 manner
around the copper rod at a xed distance

from the cavity. Since the cylindrical cavity
sets up cylindrically symmetric standing
microwave eld patterns, this behavior can
only be rationalized by a perturbation of
the component of the magnetic eld that
is modulated at 100 kHz, caused by eddy
currents induced in the copper rod by the
100 kHz component. Lenzs Law requires
the induced magnetic eld to oppose the
applied eld, and this effect is maximally
observed when the applied eld is perpendicular to the copper surface, and zero
when it is parallel to the copper surface.

This cos2 variation in sensitivity has also
been observed in the Allendoerfer coaxial
cell [46, 48]. This effect inevitably means
that paramagnetic species will experience
different amplitudes of the modulated eld
depending upon their location within the
cell. However, it was shown that this will
reduce the EPR signal by only a factor
of one half from that which would be
otherwise seen.
Application Laminar ow was
established using ow rates such that the
Reynolds Number, Re < 10. Thus, after
sufcient lead-in (specically 0.1 Re
h), which in this case is negligible, a
parabolic velocity prole develops across
the ow-path. In this manner, the hydrodynamics of this electrochemical setup are
equivalent to that of the channel electrode
ow system; the mass transport-limiting
current is therefore given by the Levich
equation [86],
3.2.3.5.2
ilim = 0.7767nF cbulk D 2/3 0 h1/3 xe

(40)
in which is the circumference of the
electrode, 0 is the mean solution velocity,
1/3
2/3
xe is the electrode length, D is the diffusion coefcient of the electrochemically

active species, and n is the number of electrons transferred per mole reactant during
the electrode reaction. The applicability of
this equation was tested using the known
one-electron reduction of uorescein in
aqueous alkali, to yield the semiuorescein radical anion, and it was found that
the well-dened current|potential curves
observed gave rise to linear Tafel plots,
with the expected room temperature value
of 59 mV decade1 slope. A linear Levich
plot allowed inference of a diffusion coefcient of 3.0 106 cm2 s1 , in agreement
with literature values [87]. Thus, the hydrodynamics of this particular ow cell are
such that there is the expected parabolic,
laminar ow.
Since the hydrodynamics are analogous
to those experienced in the channel ow
cell, it follows that steady state EPR signals
should behave analogously, that is,
S
(41)
2/3
Vf
This equation should hold, providing the

diffusion-layer thickness at the electrode
is small when compared to the thickness
of the ow-path. Figure 26 shows data
log10 s/i
[a.u.]
0.0
0.5
1.0
Steady state EPR

signal|current|ow rate data.
Fig. 26
3.0
2.0
log10(Vf /cm3 s1)
199
obtained from the uorescein system

plotted according to Eq. (41). A straight line
is observed for a wide range of ow rates;
the line drawn has a gradient of 2/3,
as predicted by the assumed theory. The
deviation at low ow rates is entirely due to
the diffusion layer becoming comparable
in size to h; less radical is produced than is
predicted by assuming a wide channel [70,
89]. Hence at sufciently slow ow rates,
there is near-exhaustive electrolysis, and
the cell behaves as a thin-layer cell [88], in
which the current is given by
i=
4
nF 0 h cbulk
3
(42)
The full transition between the two

limits has been calculated [70, 89] and
corresponds to a ow rate of 103 cm3 s1
for xe = 4 mm and 2h = 100 m. The EPR
signal will be more sensitive to these
effects than the observed current, since
the EPR signal reects the concentration
of radicals throughout the channel, not
merely the concentration gradient at the
electrode surface.
3.2.3.6
The Bond In Situ Microelectrode
Cell
3.2.3.6.1 Cell Design Bond and coworkers have designed various cells suitable
for in situ electrochemical EPR [9094].
These cells generally have a small volume and permit the recording of variable
temperature EPR. In this section, only
the most recently designed cell [94] will
be discussed.
Figure 27 illustrates the all-glass cell
used for electrochemical EPR over a wide
temperature range. The reference electrode (either an Ag|AgCl or platinum)
is xed in the center of a Pt workingelectrode coil, the latter extending to the
bottom of the quartz tube. These electrodes have diameters of less than 70 m,
so as to minimize the amount of metal
present in the EPR cavity; the counter electrode is situated 15 mm above the working
electrode outside the cavity, also permitting interference-free measurements from
species diffusing from the counter electrode during electrolysis. With this design,
Cu rods
Soldered connections
Nylon support
PVC sleeves
149
125
150
B7 joint (quartz socket,

pyrex cone)
180
200
O ring
Average filling level
(approx. 0.25 ml)
Counter electrode
2
4
Sensitive region of cavity

Reference electrode
(70-m Pt coated with Ag/Agcl) Fig. 27 Schematic diagram of
Working electrode
the in situ Bond cell for
(70-m Pt)
microelectrodes. All lengths are
given in mm.
only a small volume of solution may be

placed into the bottom of the cell, typically
0.2 cm3 , allowing even lossy solvents to be
used in electrochemical EPR work. Furthermore, by allowing complete sample
handling under an inert atmosphere, the
cell facilitates the study of electrochemically generated radicals that are highly
oxygen and/or moisture sensitive. Furthermore, it was observed that this cell has a
low RC time constant, making fast scan
cyclic voltammetry viable [94].
3.2.3.6.2 Application This cell has been
used for a variety of studies such as the oxidation of Cr(CO)2 (Ph2 PCH2 CH2 PPh2 )2
[94], the reduction of [Cp2 Ti(acac)]ClO4
(Cp = 5 C5 H5 ,
acac = acetylacetone)
[94],
dipropylpyridine-3,4-dicarboxylate
[95], and S,S-dipropylbenzene-1,3-dicarbothioate [95]. Generally, good voltammetric
curves were obtained at room temperature,
but these were often distorted at low temperatures (typically 180 K), possibly due to
iR drop or due to the onset of thin-layer
electrochemistry [94].
This cell has been recently compared
with the channel ow cell [96]. Although
the Bond cell has many advantages, its major limitation is that it does not permit the
direct interrogation of electrode reaction
mechanisms through the EPR signal in
an analogous way to that by Compton and
Coles (vide supra). Furthermore, the channel electrode cell is much more sensitive in
detecting paramagnetic species than this
cell by some orders of magnitude [96].
3.2.3.7
The In Situ Wall-jet Electrode Cell
3.2.3.7.1 Cell Design As described in

Chapter 2.2 of this volume, a wall-jet
electrode is a well-dened hydrodynamic
electrode in which the convective mass
transport is the ow due to a jet of

electrolyte striking a planar electrode
surface at right angles, and rapidly spreads
out over the electrode surface, the ow
outside the jet being at rest [97]. Like the
channel electrode, the wall-jet electrode
is a non-uniformly accessible electrode,
that is, the diffusion layer is not constant
over the working electrode surface, making
this electrode geometry successful in
mechanistic discrimination [56].
The wall-jet electrode for in situ electrochemical EPR [98] is shown in Fig. 28.
This arrangement is in fact a renement
to an in situ cell designed by Scholz and
coworkers Fig. 29 [93], in which a jet of
electrolyte impinges on a wire or mercury drop electrode. The modications by
Compton and coworkers [98] allow for controlled convection and enhancement of its
sensitivity towards unstable radicals. The
cell itself consists of a silica tube with
a mercury working electrode placed at
one end. Solution ows into the cell via
the steel jet so that it impinges normally
and centrally on the electrode surface.
The system was connected to a gravityfed ow system, with a silver or platinum
pseudo-reference electrode positioned upstream of the mercury electrode and a
platinum gauze counterelectrode located
downstream of the cell.
The sensitivity prole of the cavity to radicals with the wall-jet cell placed along the
vertical axis of the cavity, corresponding to
a node in the electric component of the
standing microwave eld, with the working electrode surface at the cavity center,
exhibited little distortion from the theoretically predicted sin2 behavior, even though
large amounts of metal (mercury and copper) are present. Measurement of the cavity
Q-factor when the wall-jet cell containing
acetonitrile and 0.5 M tetrabutylammonium perchlorate supporting electrolyte
201
202

Schematic illustration
of the in situ wall-jet EPR cell.
Fig. 28
Solution from
reservior
Reference
electrode
Altex T-piece
To
counter electrode
Glass
tubing
od = 0.4 cm
id = 0.2 cm
Jet
17.5 cm
Mercury working
electrode
Copper
wire
were present in the EPR cavity, revealed

Q = 1000, which is a comparable value to
that for the channel cell (vide supra).
Solution
intlet
RE
3.2.3.7.2 Application Albery [99] and Matsuda [100] have derived an approximate
equation for the mass transport-limited
current for an n-electron transfer reaction
at a wall-jet electrode,
AE
Solution
outlet
ilim = 1.59kc nF D 2/3 5/12 Vf
3/4
a 1/2 R 3/4 cbulk
(43)
in which kc is a constant determined by

experiment to be close to 0.90 [99101],
is the kinematic viscosity, a is the
diameter of the jet, and R is the radius
of the electrode. Analysis of the welldened voltammograms for the reduction
Schematic of the Scholz design
electrochemical EPR cell for use with
owing solutions.
Fig. 29
WE
of 1,4-benzoquinone in acetonitrile in
terms of Eq. (43), by plotting ilim against
3/4
Vf revealed a straight line, and gave a
diffusion coefcient in good agreement
with literature values, indicating that the
hydrodynamics of this cell are well dened
and understood.
It has been seen that the other in
situ hydrodynamic cells discussed above,
the channel electrode cell and the tube
electrode cell, both exhibit EPR signal (S)
characteristics dependent upon the ratio
2/3
i/Vf . However, in the case of the walljet electrode under steady state conditions,
in which the entire cell downstream of the
working electrode is lled with radicals,
the rate of loss of radicals from the cell
is proportional to Vf , while the rate of
their formation is proportional to i, thereby
anticipating
i
S
(44)
Vf
This relationship was veried experimentally for the benzoquinone/acetonitrile system. The EPR signals were observed with
excellent signal-to-noise ratio.
Analogous experiments in aqueous solutions gave similar responses. These
present a more stringent test in terms
of EPR compatibility since they are notoriously lossy; in order to reduce dielectric
loss to acceptable levels, the internal diameter of the glass tube was reduced
to 0.15 cm. Although satisfactory hydrodynamic behavior was observed, when the
cell dimensions were reduced for EPR
compatibility and using a cell of the exact
type to that described by Scholz [93], this
was no longer the case signicant deviations from wall-jet behavior were found
whatever the separation between the working electrode and the jet. Reproducibility
problems were additionally encountered,
in part due to the trapping of air bubbles
within the cell, and due to a greater sensitivity of the cell to the exact location of the
jet, pointing to loss of the wall-jet behavior and to the existence of edge effects
caused by the walls of the cell.
In the Scholz design, the separation between the mercury working electrode and
the jet is minimized so as to give rise to
a thin-layer cell as shown in Fig. 30.
Near-exhaustive electrolysis of the electroactive material is thus likely to occur,
leading to large currents and a high sensitivity towards the detection of radicals that
are stable on the experimental timescale;
the shape of the mercury electrode is
such that it may screen the workingelectrode|electrolyte interface from the
100 kHz magnetic eld modulation used
in conventional EPR spectrometers, so that
only radicals downstream of the shielded
area will be EPR active; the wall-jet approach may be more suited in not only
observing mechanistic pathways, but also
Flow
The ow pattern
between the jet and the mercury
working electrode under
thin-layer conditions.
Fig. 30
Hg
203
204
in the determination of EPR spectra from

short-lived radicals. Furthermore, since
the hydrodynamics of the Scholz owelectrode design are not mathematically
well described, it limits the use of its design for the study of electrode reaction
mechanisms and kinetics.
The DunschPetrNeudeck Cell
for In Situ UV-VIS EPR Spectroelectrochemistry
In situ electrochemical EPR is a versatile spectroelectrochemical technique as
it allows quantitative inference of both
the structure and concentration of paramagnetic species formed during electrochemical processes. However, this it is
often difcult to distinguish between paramagnetic species that have been formed
by the heterogeneous transfer of an
electron and those formed subsequently
in homogeneous follow-on reactions. Furthermore, EPR alone cannot give structural information concerning diamagnetic reaction participants. Neudeck and
coworkers [102108] combined UV-VIS
and EPR spectroscopies to give insights
into whether primary or non-primary radicals are formed during electrolysis.
3.2.3.8
Working
electrode
Second
counter electrode
Adjustable
cell holder
To photodiode
array detector
3.2.3.8.1 Cell Design The in situ electrochemical UV-VIS-EPR cell is shown

schematically in Fig. 31. A TE102 cavity
was modied to permit optical transmission. Fiber-optic light guides were used
for the illumination of the EPR cavity as
well as to connect the cavity to a photodiode array detector, the latter so as
to improve UV-VIS measurements. The
light guides were arranged near the walls
of the cavity so that there was no loss
in EPR sensitivity. Further, since the
Reference
electrode
Gas inlet
ESR-cavity
From light source
Modulation
coils
Schematic of the cell

used for simultaneous in situ
electrochemical EPR and
UV-VIS measurements.
Fig. 31
First
counter electrode
distance between the two light guides is

just the cavity width, the sensitivity of the
UV-VIS measurement is maximal. In order to minimize photochemical reactions
from taking place, the light intensity used
was kept to a minimum.
The electrochemical cell itself consisted
of essentially a at cell, xed inside in
the optimal position within the EPR cavity
so as to ensure maximum sensitivity.
The working electrode employed was
either a platinum mesh [102] or later,
a partly laminated gold mesh [103, 104].
An Ag|AgCl electrode was used as the
reference, whilst two counter electrodes
were employed to minimize oscillation of
the cell voltage. The rst counter electrode,
a palladium sheet, is situated below the
at part of the cell, whilst the second,
connected by a resistor to the potentiostat,
is located above it.
3.2.3.8.2 Application The cell was tested with two electroactive compounds,
N ,N ,N ,N -tetramethyl-para-phenylenediamine (TMPD) and para-aminodiphenylamine (PAD) in acetonitrile solutions.
The former compound undergoes a oneelectron oxidation to yield a dark blue
cation radical, TMPD+ whilst the latter is thought to give rise to nonprimary radicals upon oxidation, by comproportionation of the oxidation product, N -phenylquinonediimine with paraaminodiphenylamine. In this manner,
absorbance|potential data (at a xed wavelength) can be obtained together with the
EPR signal|potential and cyclic voltammetric (current|potential) data.
For long-lived primary radicals, the
EPR signal|potential curves should directly
overlay a suitable normalized integrated
charge|potential curve, since the former
reects the number of spins in the cavity
and the latter, the number of radicals
formed. This was found to be so in the case

of TMPD. The absorbance is also related
to the concentration of the absorbing
species. Measurements were conducted
using 560-nm light, corresponding to the
absorption of light by TMPD+ . Since
the concentration of this species is not
uniform, the BeerLambert law is replaced
by a differential equation relating the
intensity of light transmitted (I ) to the
concentration of radicals (c),
I
= c(x)I
x
(45)
where is the extinction coefcient at

a xed wavelength, and x is the thickness of the solution layer. Since the
radical species is formed in the diffusion layer at the electrode, changing the
scan rate of the cyclic voltammetric experiment perturbs the concentration gradient.
It was found that in the range 20 to
200 mV s1 , the EPR signal|potential and
the UV-VIS absorption|potential curves
match up exactly, thereby allowing a
semi-quantitative measurement of radicals, or UV-VIS absorbing compounds.
At high scan rates (500 mV s1 ), the
EPR signal|potential curve matches the
integrated charge|potential curve, but deviates from the absorption curve, reecting a greater concentration gradient
at the electrode surface. At low scan
rates (10 mV s1 ), further deviation due
to diffusive processes occurs, permitting
distinction between primary and nonprimary radicals; at these scan rates, if
non-primary radicals are formed, neither
the EPR signal|potential curve nor the
absorption|potential curve coincides with
the charge|potential curve.
This cell design has been used to study
conducting polymer lms at electrode
surfaces [104108].
205
206
The DryfeWebster Cell for Charge

Transfer Across a Liquid|Liquid Phase
Boundary
3.2.3.9
3.2.3.9.1 Cell Design In situ EPR spectroscopy is an ideal method to detect

and identify paramagnetic participants in
liquid|liquid processes, but to our knowledge, has only been undertaken by Dryfe
and coworkers [109, 110]. The cell used
to perform the interfacial polarizationEPR experiments is shown in Fig. 32,
(b)
(h)
(c)
(a)
5 cm
(g)
and is essentially a conventional silica

at cell [17] modied so that counter and
reference electrodes can be inserted into
both ends. It is assembled by rst lling the higher density solvent through
the capillary at the base of the cell using silicone tubing and a syringe, so
that the solvent reaches the midpoint of
the at portion of the cell, and then
sealing the silicone tubing using a Hoffmann clamp. The upper, less dense phase
is then introduced into the cell from
the top using a Pasteur pipette. Using
this lling method, a liquid|liquid interface can be generated and maintained
in the midportion of the at part of the
cell.
The EPR cell has two coiled platinum
wire counter electrodes located outside the
thin-layer section, and two silver wire reference electrodes attened at one end to
a thickness that enables their insertion
into the at part of the cell, positioned
within 5 mm of the liquid|liquid interface. To obtain optimal control of the
potential difference between the liquid
phases, the portions of the reference
electrodes outside the at cell were jacketed in PTFE tubing and sealed so that
they are only in contact with the solutions inside the at part of the cell. The
(e)
(h)
(i)
(d)
(j)
(f)
Silica cell for performing

liquid|liquid electrochemical EPR
experiments: (a) interface between two
immiscible liquids, (b), platinum wire
counter electrode 1, (c) silver wire
reference electrode 1, (d) platinum wire
counter electrode 2, (e) silver wire
reference electrode 2, (f) capillary to ll
lower portion of cell, (g) thin-layer
portion of cell, (h) Teon /silicone
rubber sleeves surrounding lower
portions of reference electrodes,
(i) electrical contact to reference
electrode 2, and (j) electrical contact to
counterelectrode 2.
Fig. 32
inherent high resistivity of a at cell combined with low dielectric solvents (such
as 1,2-dichloroethane, DCE) can be partially overcome using this arrangement,
and reasonable linear sweep voltammetry
was observed using a four-electrode potentiostat. All experiments were undertaken
in a TE102 cavity.
3.2.3.9.2 Typical Applications EPR spectra were obtained for both the interfacial
reduction of 7,7,8,8,-tetracyanoquinodimethane (TCNQ) and oxidation of tetrathiafulvalene (TTF), when dissolved in DCE
by the aqueous phase ferri/ferrocyanide
redox couple, following the application
of a potential difference directly to the
liquid|liquid interface. Previous work [111,
112] suggested that a charge-transfer process occurs at the liquid|liquid interface,
due to the heterogeneous reduction of
TCNQ by the aqueous couple,
TCNQ(DCE) + Fe(CN)6 4 (aq)
C
D
TCNQ (DCE) + Fe(CN)6 (aq)

(46)
A similar mechanism can be written for
the oxidation of TTF,
TTF(DCE) + Fe(CN)6 3 (aq)

+
4
C
D
TTF (DCE) + Fe(CN)6 (aq) (47)
The presence of the organic ion radicals,

and hence the occurrence of an electrontransfer process, was veried using EPR
spectroscopy. The EPR spectra for TCNQ
and TTF+ in DCE at room temperature
consisted of single broad lines with
Bpp of 5.2 and 3.0 G, respectively. The
broad lines were thought not to arise
by interaction of the radical ions with
the aqueous phase iron complexes, since
broad EPR spectra were also observed from
TCNQ and TTF+ electrochemically
generated in the appropriate singlephase three-electrode cell [17] using

DCE and the same supporting electrolyte, bis(triphenylphosphoranylidene)ammonium tetrakis(4-chlorophenyl)borate
(BTPPA+ TCPB ). The broad EPR spectra
were thus explained by the strong ionpairs formed between the product radical
ions and the organic phase cation
or anion, respectively undergoing fast
electron exchange with neutral TCNQ or
TTF [113].
EPR time-sweep experiments (see
Fig. 33) for the oxidation of TTF,
in which the magnetic eld is held
constant, corresponding to the maximum
adsorption, by applying a potential
difference (1 ) between the silver quasireference electrodes located in each
phase sufcient to cause an electron
to be transferred across the water|DCE
interface from TTF to Fe(CN)6 3 , showed
increases in the EPR signal intensity,
corresponding to the formation of
TTF+ , exhibiting a square-root temporal
dependence as expected for a potentialstep occurring under planar diffusion
(Volume 3 Chapter 2.3). Reversing, to a
potential sufciently negative to drive the
back reaction (2 ), immediately causes
a decrease in the EPR signal intensity,
since the radical cation is reduced to
neutral TTF. Similar observations were
made with TCNQ. Figure 33 also shows
that when the potentiostatic circuit is
broken after the formation of TTF+ , the
EPR signal decreases, as TTF+ , a stable
cation radical in DCE, diffuses away from
the central, most sensitive part of the
EPR cavity. Further application of this
work to benzoquinones [110] permitted
the deduction that at the aqueous|DCE
interface, reduction of the parent
compounds to yield semiquinones occurs
without their immediate protonation.
207
208
Open
circuit
Time-sweep rst derivative EPR

spectrum of TTF+ recorded at constant
Signal
magnetic induction using liquid|liquid
intensity electrochemical EPR cell. The organic
[a.u.] phase (DCE) contained 5 mM TTF,
36 mM BTPPA+ TCPB . The aqueous
phase contained 25 mM K4 [Fe(CN)6 ],
500
25 mM K3 [Fe(CN)6 ], and 0.1 M Li2 SO4 .
EPR sweep time = 500 s, EPR sweep
width = 10 G, modulation
amplitude = 2.0 G, and time
constant = 10 102 ms.
1 = oxidizing potential,
2 = reducing potential. The potential
difference imposed to induce oxidation of
the TTF was +0.7 V, whilst this process
was reversed using a potential difference
of 0.1 V (versus the silver
pseudo-reference electrodes used).
Fig. 33
f1 f2 f1 f2 f1
Open
circuit
t
[s]
3.2.4
Illustrative Examples of EPR in

Electrochemistry
Applications of in situ electrochemical

EPR are presented in this section.
Radical and Radical Ion
Identication
3.2.4.1
3.2.4.1.1 The Anodic Oxidation of paraAminodiphenylamine (PAD) Allendoerfer

and coworker [114] studied the oxidation
of PAD at platinum electrodes in aqueous
solution, in the pH range 1.24.8. Cyclic
voltammetry in buffered solution produces
a single peak, with Emid (= 0.5(Epa +
Epc ) at 0.40 V versus SCE. This potential
parameter was found to linearly shift with
pH in the range pH 13, with a gradient
corresponding to a proton : electron ratio
of 1 : 1. Further, from the variation in
peak potential separation with pH, this
process was estimated to correspond to
two electrons.
In unbuffered solutions at pH 5.0, the
cyclic voltammograms appeared to look
like two one-electron processes, suggesting the presence of a radical intermediate.

An EPR study followed, conrming the
presence of a radical, identied as being
essentially PAD+ . A typical EPR spectrum
of this radical is illustrated in Fig. 34. Since
the pKa s of PAD are 4.4 and 0.072,
a mechanism consistent with these observations suggested by Allendoerfer and
Male is
C
PADH
D
NPQH + 2e + 2H (48)
+
C
PADH + NPQHD
2PAD
(49)
where NPQ is to N -phenylquinonediimine, and the extra H refers to the fact

that the single protonated states are involved. The comproportionation reaction
is postulated as the EPR spectrum corresponds to the unprotonated form of the
radical. Further, since only a single reversible two-electron process is observed
electrochemically, the comproportionation
constant must be less than one.
Compton and coworkers [13] observed
a similar comproportionation reaction for
the case of para-phenylenediamine.

Experimental EPR
spectrum of PAD+ .
Fig. 34
The Reduction of Dicyanobenzene

Gennaro and coworkers [115] studied the
reduction of 1,2-dicyanobenzene in acetonitrile at mercury electrodes. Two reduction signals were observed; the rst
(at 1.32 V vs. SCE) was assigned to the
1,2-dicyanobenzene radical anion, the second (at 2.35 V vs. SCE) attributed to
further reduction of the radical anion, producing the 1,2-dicyanobenzene dianion,
which was thought to immediately react
with the solvent, producing benzonitrile,
the reduction of which is thus observed.
This reaction pathway was later conrmed
by Compton and Waller [16] using in situ
electrochemical EPR: Fig. 35 shows the
EPR spectra observed by potentiostatting
at the appropriate voltages.
3.2.4.1.2
The Reduction of C60 This compound undergoes a series of reductions [116, 117]. Electrochemical EPR has
been undertaken in toluene|acetonitrile
mixtures [118], but it is notoriously difcult to unambiguously interpret these
spectra [119, 120].
3.2.4.1.3
3.2.4.1.4 Spin Trapping If the radical to

be investigated is extremely short-lived, it
may not be possible to observe it directly. In
these cases, the radical is spin trapped
by chemically reacting it with a diamagnetic compound to produce a relatively
10G
stable paramagnetic species. This technique was rst introduced by Janzen and
Blackburn [121, 122], using N-tert-butyl-phenylnitrone (PBN) to spin-trap alkyl
and alkoxy radicals. In a similar manner,
Wadhawan and coworkers [123] were able
to propose the existence of alkyl peroxides
during sono-emulsion Kolbe electrosyntheses. It must be stressed that case should
be taken to avoid electroactive spin traps;
Bard and coworkers [124] showed that
spin trapping is possible during electrochemical experiments, and that PBN is
electroinactive in the potential range 1.5
to 2.5 V versus SCE in acetonitrile. The
interested reader is referred to Kemps
review on spin trapping [125] for further
information.
Electrode Reaction Kinetics and
Mechanism
This section shows the versatility of the
channel ow EPR cell for mechanistic
discrimination.
3.2.4.2
EC Versus ECE These reactions

can be characterized by the general reaction scheme [126]
3.2.4.2.1
Electrode
C
A e D
Solution
B C
Electrode
C
C e D
products (iii)
(i)
(ii)
209
210

CN
CN
b b
5G
(a)
CN
5G
(b)
EPR spectrum obtained by in situ electrolysis of 1,2-dicyanobenzene at

(a) 1.32 V versus SCE, where the spectrum was shown to be that of
1,2-dicyanobenzene radical anion and (b) 2.35 V versus SCE, where the spectrum
corresponds to the benzonitrile anion radical.
Fig. 35
Steps (i) and (ii) dene a pure EC mechanism, while Steps (i), (ii), and (iii) correspond to an ECE mechanism. In effect, the
mechanisms differ in terms of the electroactivity of the product, C, of reaction
(ii) so that if C is more readily oxidized
or reduced than A, the process is of the
ECE type, whereas if C is electrochemically

inert, the process is designated EC. Analytical theory describing these processes in
channel electrodes is established [127].
Compton and coworkers [58] deduced
that the EPR signal for an ECE reaction follows that for a simple E reaction (Eq. 27),

2.0
1.8
Neff
1.6
1.4
1.2
0.0
3
log10KECE
Working curve showing the relationship of Neff and K for an
ECE process.
Fig. 36
provided the current, i, is replaced by

that giving rise to unreduced B , namely,
(iA iC ). Further, a working curve, shown
in Fig. 36 was described showing how a
dimensionless rate constant, K, dened
by
1/3

4h4 xe2 d 2
(50)
K=k
9Vf2 D
varies with the effective number of
electrons, Neff , passed,
Total Limiting Current
Observed
Neff =
Limiting Current For A Single
Electron Transfer Process
(51)
The electrochemical reduction of 2-nitropropane at mercury-plated copper electrodes was investigated in aqueous solution, using Triton-X as a stabilizer. Two
waves were observed, the rst corresponding to the [(CH3 )2 CHNO2 ] radical
anion, as evidenced by electrochemical
EPR. Upon removing the applied potential
from the electrode, EPR signal transients
were obtained, analysis of which in terms
of simple rst-order kinetics (Eq. 36) permitted the inference of a rate constant,
2/3
k. Plotting this parameter against Vf ,
gave a straight line graph that allowed
extrapolation to zero ow rate to give
a value of 0.36 0.02 s1 for the rstorder decay of the radical anion. The
steady state EPR signal was found to
be directly proportional to the generating current, until after the onset of the
second voltammetric wave, when a dramatic decrease of EPR signal occurred.
These features are both consistent with
EC and ECE mechanisms. Conclusive evidence of the mechanistic discrimination
between EC and ECE reactions came from
analysis of the reduction current|ow rate
data, shown in Fig. 37. It can be seen
that at fast ow rates, the data tend towards one-electron behavior, whilst at slow
ow rates, they tend towards two-electron
behavior, indicative of an ECE reaction.
Further analysis of the EPR signal|ow
rates data permitted the determination of
EPR detection efciencies, from which a
rate constant of 0.28 s1 was obtained for
radical decay.
211
1500
1000
Iobs
[A]
212
500
0
0.0
0.3
0.2
(Volume flow rate)1/3

[cm s1/3]
The limiting current|ow rate behavior for the reduction of
2-nitropropane in aqueous solution. The solid lines are calculated
behavior for one- and two-electron processes.
Fig. 37
ECE Versus DISP1 DISP1 reactions are described by the following

general kinetic scheme.
3.2.4.2.2
Electrode
A e B
Solution
Solution
C
F + e D
S + H+ SH+
(i)
C
SH+ + S
D
F + L
B C
(ii)
C
B + CD
A + D
(iii)
where it was assumed that the protonation

of S is rate determining and irreversible,
as it is accompanied by a change in
hybridization from sp2 to sp3 at the central
carbon atom. In a similar manner to that
above, the current|ow rate data showed
one-electron behavior at fast ow rates,
but two-electron behavior at slow ow
rates, indicative of either an ECE or a
DISP process.
Analysis of EPR signal transients, obtained by open-circuiting the electrode,
using simple rst-order kinetics, was
found to give an apparent rate constant of
1.05 s1 , ruling out a DISP2 mechanism.
Using experimentally determined values
for the effective number of electrons transferred together with the working curves,
values of K, were obtained for different
ow rates. This is plotted in Fig. 39, where
In this scheme, Step (ii) is rate limiting.

Compton and coworkers [74] deduced the
analytical theory for the solution of the
relevant convective-diffusion equation at
steady state for the above homogeneous
kinetic scheme, in terms of a working
curve, see Fig. 38.
The reduction of uorescein (F) to give
the semiuorescein (S ) radical anion in
aqueous solution at a silver channel electrode was studied [74]. Cyclic voltammetric
experiments gave rise to linear Levich
plots, but the size of the reduction wave
increased with decreasing pH, suggesting
that S becomes unstable with respect to
disproportionation to yield F and leucouorescein (L),
1.8
Neff
1.4
1.4
1.2
0.0
3
log10 KDISP1
Working curve showing the relationship of Neff and K for a
DISP1 process.
Fig. 38
Analysis of the limiting

current|ow rate data at
pH 9.68 assuming an ECE
mechanism (O) or a DISP1
mechanism ().
Fig. 39
100
200
(Volume flow rate)2/3

[cm2 s2/3]
it is apparent that a rather better straight

line is obtained assuming a DISP1 rather
than ECE mechanism. Furthermore, the
variation in the decadic rate constant observed from the extrapolated EPR transient
data (kapp ) with pH using the relationship
k=
1
kapp
2 Vlim
f 0
(52)
which arises since two S radicals are

lost for each time that S is protonated, is linear with a slope of 1,
which would be that anticipated for the

pseudo-rst-order rate constant for the
protonation of S . Clearly, the analysis in terms of a DISP1 mechanism is
in better agreement with experimental
observations.
EC Waller and coworkers employed the channel ow cell for the study
of a catalytic mechanism [68]. The (preequilibrium) EC mechanism can be dened by the following kinetic scheme.
3.2.4.2.3
213
214
C
A e D
(i)
B+P
A+Q
k
Q Products
(ii)
(iii)
Solution of the appropriate convectivediffusion equations at steady state permits the behavior of this mechanism to
be described in terms of a normalized
rate constant:

1/3
[P ]bulk h2 xe2
K = kK
(53)
[A]bulk 402
where [P ]bulk and [A]bulk are the bulk concentrations of P and A. Under conditions
in which the Leveque approximation is
valid [63, 71], the current is a unique function of K . The observed behavior is best
expressed in terms of the effective number of electrons transferred, Neff . The EPR
signal behavior may be deduced from numerical simulation of the concentration of
the ion-radical species, B, as indicated in
Sect. 3.2.3.
Comproportionation Reactions
Comproportionation reactions are of
the form Y + Y2 2Y . Spackman
and coworkers [75] studied the effect
of a homogeneous comproportionation
reaction on an electrode reaction that
proceeds stepwise via two singleelectron transfers,
3.2.4.2.4
Electrode
C
A + e D
(i)
Solution
C
B + e D
(ii)
Electrode
A+C
(iii)
2B
0 is more positive than E 0 . For

where EA|B
B|C
the ComptonColes channel cell, the EPR
signal from the regions adjacent to and
downstream from the electrode is given
by the integral concentration of B (an

anion radical) over the whole volume of the
EPR cavity weighted by a sin2 sensitivity
prole,

2h xoffset +xe /2
x
b sin2
S
2xL
0
xoffset xe /2

xL
x
b sin2
dxdz
+
2xL
xoffset +xe /2
(54)
where b is the concentration of B, found
by numerical analysis of the pertinent
convective-diffusion equation, 2xL is the
length of the EPR cavity, and xoffset is the
distance (offset) from the center of the
electrode to the center of the cavity.
The behavior of the steady state EPR
signal (S1 ) of B at a potential corresponding
to steady state limiting current for the
reduction of A to B is related to the
generating current (provided B is stable),
as indicated in Sect. 3.2.3,
S1
i
2/3
Vf
(55)
Stepping the potential so that the steady

state limiting current for the reduction
of B to C is established, gives rise to a
new EPR signal (S2 ). Figures 40 and 41
show the dependence of the dimensionless
parameter S2 /S1 upon the rate constant
for step (iii) for the respective cases of
equal and mixed diffusion coefcients. As
kIII is initially increased from zero, the
EPR signal ratio rises rapidly, but this
tails off when the kinetics become mass
transport-controlled. S2 /S1 never reaches
the limiting value of two, as this is what
would be expected if all C produced by
the electroreduction of A is converted into
B by comproportionation. Rather, for innitely fast comproportionation kinetics,
two separate spatial zones are established,
one close to the electrode containing C (but

Working curve showing the
relationship between S2 /S1 and log(kIII )
for the channel electrode ow cell
geometry of d = 0.591 cm,
2h = 0.036 cm, w = 0.402 cm and
xe = 0.425 cm, and
DA = DB = DC = 1.75 105 cm2 s1
and a volume ow rate of 102 cm3 s1 .
Fig. 40
1.4
1.2
S2/S1
1.0
0.8
0.6
0.4
0.2
50
60
70
80
90
log10(kIII/ mol1 cm3 s1)
DB = 1.75 105 cm2 s1
1.4
DB = 1.20 105 cm2 s1
1.2
DB = 0.875 105 cm2 s1
S2/S1
1.0
0.8
0.6
0.4
0.2
50
60
70
80
90
log10(kIII/ mol1 cm3 s1)

Working curve showing the relationship between S2 /S1
and log(kIII ) for the channel electrode ow cell geometry of
d = 0.591 cm, 2h = 0.036 cm, w = 0.402 cm and xe = 0.425 cm,
and DA = 1.75 105 cm2 s1 , DC = 0.875 105 cm2 s1 , ow
rate = 102 cm3 s1 with DB = 0.875 105 , DB = 1.20 105 ,
and DB = 1.75 105 cm2 s1 .
Fig. 41
not A), and the other extending into the

center of the channel containing A (but
not C), with a sharp interface between the
two zones.
The ratio of the diffusion coefcients of
A and B has an effect upon the EPR signal
ratio S2 /S1 . If DB is decreased relative
to DA , assuming an initial condition
DA = DB , both S1 and S2 increase but at
different rates, so that S2 /S1 decreases. The
increased values S1 and S2 can be explained

by noting that decreasing DB narrows the
diffusion layer of B, consequently reducing
the ow rate of B out of the channel,
leading to an accumulation of radicals in
the channel ow cell.
Spackman and coworkers [75] looked
at the case in which A is p-chloranil.
First, with the electrode positioned at the
center of the cavity, it was found that
215
216
at a potential (E1 ) corresponding to the

formation of the anion radical (B), a plot
of log10 S1 /ilim versus log10 Vf gave a
straight line with slope 2/3, as would
be expected for a stable radical. EPR signal
(S2 )|ow rate data were then obtained by
potentiostatting the platinum foil working
electrode at a potential (E2 ) corresponding
to the reduction of the anion radical.
Repositioning the electrode so that it was a
few millimeters downstream of the cavity
center, allowed a second set of S|Vf
data at both potentials. In both cases,
good agreement was obtained between
experiment and theory.
The dependence of S2 /S1 on the location
of the electrode within the cavity may be
rationalized as follows. When the electrode
is at E1 , B is generated by the electroreduction of A and then ows unperturbed out of
the channel, so that a concentration prole
of B within the channel is established, and
so that the concentration of B decreases in
both x and z directions. However, when
the electrode potential is stepped to E2 , the
concentration of B no longer varies monotonically with the x and z coordinates, but
the precise distribution depends is determined by the electroreduction of B to C
and the comproportionation reaction of A
with C. In particular, at the downstream
edge of the electrode, [B] rises rapidly with
x and z, since B is no longer consumed
electrochemically, but then passes through
a maximum when C fully reacts. It is the
position of this maximum within the EPR
cavity that determines the ratio S2 /S1 .
In situ electrochemical EPR has been
employed in the study of the comproportionation of methyl viologen [76, 77] and
cyanophenyl paraquat [128].
3.2.4.2.5 Photoelectrochemical EPR The
channel ow cell is ideally suited for
the interrogation of photoelectrochemical
phenomena as it is constructed using silica a transparent material. Further, the

electrode material can be made to be
a semiconducting material. Irradiation is
simply achieved by illuminating the working electrode through the irradiation port
machined on one face of the cavity, allowing the transmission of a high proportion
of the interrogating light beam to the
working electrode and uniform irradiation of the latter so that photochemical
channel ow experiments can be routinely undertaken in an in situ mode [55,
6567, 7981]. In this manner, a variety of mechanisms have been studied,
notably photo-ECE [67, 79], photo-CE [65],
and photo-DISP2 [66].
Modied Electrodes
widely applied in the study of radical intermediates within polymer lms
on electrode surfaces. Examples include
the electrochemical doping in polypyrrole [129, 130] and the anodic oxidation
of poly(N -vinylcarbazole) lms [131].
3.2.4.3
Electrochemical EPR in Ionic

Liquids
Ionic liquids are ideal solvents for conducting electrochemical experiments, notably
due to their (1) high thermal stability
and wide liquid range, (2) ability to act
as solvents for many organic and inorganic materials, whilst still being highly
polar, noncoordinating media, (3) high
conductivity, and (4) negligible vapor pressure in contrast with conventional organic solvents, permitting their study
under conditions of high vacuum. Surprisingly, little work on electrochemical
EPR has been undertaken in these media.
Allendoerfer and Osteryoung have pioneered this area [132], primarily focusing
3.2.4.4
upon measurements on polymers (such as

polypyrrole [133] and polyaniline [134]) in
1-methyl-3-ethylimidazolium chloride and
aluminum chloride.
Recently, Marken and coworkers [135]
reported a novel method for the electrogeneration of ionic liquids in which microdroplets of N ,N ,N ,N -tetrahexyl-paraphenylene diamine (THPD) are deposited
upon the surface of a basal plane pyrolytic graphite (or gold) electrode and are
immersed into an aqueous electrolyte solution. The ionic liquid is formed as a consequence of the requirement to maintain
electroneutrality upon oxidizing THPD,
THPD(oil) + X (aq)
[THPD+ X ](oil) + e
(56)
It has been shown that this process occurs at the three-phase boundary
of THPD|aqueous electrolyte|solid electrode [135, 136]. In situ EPR transients for
the oxidation and rereduction of THPD
were seen to be consistent with the anion
insertion process occurring at the triple
phase junction [135].
3.2.5
Conclusion

shown to be a useful technique for probing
both the kinetics and mechanisms of electrode processes that produce or consume
paramagnetic species. A variety of cell
designs have been presented, illustrating
the scope of this spectroelectrochemical
technique.
Acknowledgments
We are gratefully indebted to all those who

have worked with us in the pursuit of
electrochemical EPR, notably Barry Coles,
Andy Waller, Rob Dryfe, Richard Webster,

Nathan Lawrence, Frank Marken, and Andreas Neudeck. We thank John Freeman
for the preparation of all of the gures in
this manuscript.
References
1. A. Carrington, A. D. McLachlan, Introduction to Magnetic Resonance, Chapman &
Hall, London, 1979.
2. G. K. Frenkel, P. H. Rieger, J. Chem. Phys.
1963, 39, 309318.
3. D. E. G. Austin, M. F. Peover, M. H. Ingram
et al., Nature 1958, 181, 17841785.
4. A. H. Maki, D. H. Geske, J. Chem. Phys.
1959, 30, 13561357.
5. A. H. Maki, D. H. Geske, J. Am. Chem. Soc.
1960, 82, 267269.
6. J. G. Lawless, M. D. Hawley, J. Electroanal.
Chem. 1969, 23, App. 15.
7. A. H. Maki, D. H. Geske, J. Chem. Phys.
1960, 33, 825830.
8. A. H. Maki, D. H. Geske, J. Am. Chem. Soc.
1961, 83, 18521856.
9. W. J. Albery, B. A. Coles, A. M. Couper
et al., J. Chem. Soc., Chem. Commun. 1974,
198199.
10. W. J. Albery, B. A. Coles, A. M. Couper, J.
11. W. J. Albery, A. T. Chadwick, B. A. Coles
229235.
12. W. J. Albery, R. G. Compton, A. T. Chadwick et al., J. Chem. Soc., Faraday Trans. 1
1980, 76, 13911398.
13. W. J. Albery, R. G. Compton, I. S. Kerr, J.
Chem. Soc., Perkin Trans. 2 1981, 823826.
14. W. J. Albery, R. G. Compton, J. Chem. Soc.,
Faraday Trans. 1 1982, 78, 15611568.
15. R. G. Compton, A. M. Waller in Spectroelectrochemistry (Ed.: R. J. Gale), Plenum Press,
New York, 1988, pp. 319366.
16. A. M. Waller, R. G. Compton, Comprehensive Chemical Kinetics, Elsevier, Amsterdam,
1989, pp. 297348, Vol. 29.
17. L. H. Piette, R. Ludwig, R. N. Adams, Anal.
Chem. 1962, 34, 916919, 15871592.
18. G. Cauquis, J. P. Billon, J. Cambrisson, Bull.
Soc. Chim. Fr. 1960, 20622067.
19. G. Cauquis, J. P. Billon, J. Raisson, Bull.
Soc. Chim. Fr. 1967, 199205.
217
218

20. G. Cauquis, M. Genies, Bull. Soc. Chim. Fr.
1967, 32203227.
21. G. Cauquis, M. Genies, H. Lemaire et al., J.
Chem. Phys. 1967, 42, 46424646.
22. G. Cauquis, C. Berry, M. Maivey, Bull. Soc.
Chim. Fr. 1968, 25102515.
23. B. Kastening, J. Divisek, B. CostissaMihelcic, Faraday Discuss. Chem. Soc. 1973,
56, 341348.
24. B. Kastening, J. Divisek, B. CostissaMihelcic et al., Z. Phys. Chem. 1973, 87,
125132.
25. J. K. Dohrmann, F. Galluser, H. Wittchen,
Faraday Discuss. Chem. Soc. 1973, 56,
350358.
26. J. K. Dohrmann, K. J. Vetter, J. Electroanal.
Chem. 1969, 20, 2329.
27. J. K. Dohrmann, F. Galluser, Ber. BunsenGes. Phys. Chem. 1971, 75, 432437.
28. J. K. Dohrmann, Ber. Bunsen-Ges. Phys.
Chem. 1970, 74, 575580.
29. O. Stern, Z. Phys. 1921, 7, 249252.
30. W. Gerlach, O. Stern, Ann. Phys. (Leipzig)
1924, 74, 673675.
31. A. I. Akhiezer, V. B. Berestetski, Quantum
Electrodynamics, Monographs in Physics
and Astronomy, Wiley Interscience, New
York, 1965, Vol. XI, p. 150.
32. N. M. Atherton, Electron Spin Resonance,
Ellis Horwood, Chichester, UK, 1973.
33. P. Ludwig, R. N. Adams, J. Chem. Phys.
1962, 37, 828831.
34. H. M. McConnell, D. B. Chestnut, J. Chem.
Phys. 1958, 28, 778781.
35. C. P. Poole, Electron Spin Resonance: A Comprehensive Treatise on Experimental Techniques, 2nd ed., Wiley, New York, 1983.
36. P. Boyer, J. Dericbourg, C. R. Acad. Sci.
Paris 1967, 429437.
37. M. C. R. Symons, Electron Spin Resonance
Spectroscopy, Van Nostrand Reinhold, London, 1978.
38. F. Gerson, High Resolution E.S.R. Spectroscopy, John Wiley & Sons, New York,
1970.
39. K. A. McLauchlan, Magnetic Resonance, Oxford University Press, Oxford, 1972.
40. P. W. Atkins, Molecular Quantum Mechanics, 3rd ed., Oxford University Press, Oxford,
1997.
41. B. J. Tabner in Spectroscopy (Eds.: B. P.
Straughan, S. Walker), Chapman & Hall,
London, 1976, pp. 209260, Vol. 1.
42. I. B. Goldberg, Electron Spin Resonance:

Royal Society of Chemistry Specialist Periodical Reports, The Royal Society of Chemistry,
London, 1981, pp. 120, Vol. 6.
43. I. B. Goldberg, A. J. Bard, J. Phys. Chem.
1971, 75, 32813285.
44. I. B. Goldberg, A. J. Bard, J. Phys. Chem.
1974, 78, 290294.
45. I. B. Goldberg, D. Boyd, R. Hirasawa et al.,
J. Phys. Chem. 1974, 78, 295301.
46. J. B. Carroll, Ph. D. thesis, State University
of New York at Buffalo, 1983.
47. R. D. Allendoerfer, G. A. Martinchek, S.
Bruckenstein, Anal. Chem. 1975, 47,
890894.
48. R. D. Allendoerfer, J. B. Carroll, J. Magn.
Reson. 1980, 37, 497501.
49. R. D. Allendoerfer, W. Froncisz, C. C. Felix
et al., J. Magn. Reson. 1988, 76, 100106.
50. H. Ohya-Nishiguchi, Bull. Chem. Soc. Jpn.
1979, 52, 20642068.
51. F. Gerson, H. Ohya-Nishiguchi, G. Wydler,
Angew. Chem., Int. Ed. Engl. 1976, 15,
552556.
52. J. Bruken, F. Gerson, H. Ohya-Nishiguchi,
Helv. Chim. Acta 1977, 60, 12201227.
53. W. Froncisz, J. S. Hyde, J. Magn. Reson.
1982, 47, 515519.
54. J. S. Hyde, W. Froncisz, Electron Spin Resonance: Specialist Periodical Reports, The
Royal Society of Chemistry, London, 1986,
pp. 175201, Vol. 10.
55. B. A. Coles, R. G. Compton, J. Electroanal.
Chem. 1983, 144, 8794.
56. C. M. A. Brett, A. M. O. Brett, Comprehensive Chemical Kinetics, Elsevier, Amsterdam,
1988, pp. 355376, Vol. 26.
57. V. Levich, Physicochemical Hydrodynamics,
Prentice Hall, Englewood Cliffs, N.J., 1962.
58. R. G. Compton, D. J. Page, G. R. Sealy, J.
59. I. B. Goldberg, A. J. Bard, S. W. Feldberg, J.
Phys. Chem. 1972, 76, 25502559.
60. P. R. Unwin, R. G. Compton, Comprehensive Chemical Kinetics, Elsevier, Amsterdam,
1989, pp. 190296, Vol. 29.
61. J. A. Cooper, R. G. Compton, Electroanalysis
1998, 10, 141152.
62. J. B. Flanagan, L. Marcoux, J. Phys. Chem.
1974, 78, 718723.
63. J. A. Alden, R. G. Compton, J. Electroanal.
Chem. 1996, 404, 2734.
64. M. A. Leveque, Ann. Mines Mem. Ser. 1928,
12, 201208.

65. R. G. Compton, B. A. Coles, G. M. Stearn
et al., J. Chem. Soc., Faraday Trans. I 1988,
84, 23572363.
66. R. G. Compton, B. A. Coles, M. B. G.
Pilkington, J. Chem. Soc., Faraday Trans.
I 1988, 84, 43474355.
67. R. G. Compton, B. A. Coles, M. B. G.
Pilkington et al., J. Chem. Soc., Faraday
Trans. 1990, 86, 663671.
68. A. M. Waller, R. J. Northing, R. G. Compton, J. Chem. Soc., Faraday Trans. 1990, 86,
335340.
69. S. Moldoveanu, J. L. Anderson, J. Electroanal. Chem. 1984, 178, 4549.
70. J. L. Anderson, S. Moldoveanu, J. Electroanal. Chem. 1984, 179, 107118, 119128.
71. R. G. Compton, M. B. G. Pilkington, G. M.
Stearn, J. Chem. Soc., Faraday Trans. I 1988,
84, 21552164.
72. R. G. Compton, D. J. Page, G. R. Sealy, J.
Mathematical Functions, Dover Publications, New York, 1970, p. 283285.
74. R. G. Compton, P. J. Daly, P. R. Unwin
et al., J. Electroanal. Chem. 1985, 191, 1521.
75. R. G. Compton, B. A. Coles, R. A. Spackman, J. Phys. Chem. 1991, 95, 47414748.
76. R. D. Webster, R. A. W. Dryfe, J. C. Eklund
167172.
77. C.-W. Lee, J. C. Eklund, R. A. W. Dryfe
et al., Bull. Korean Chem. Soc. 1996, 17,
162167.
78. F. Prieto, R. D. Webster, J. A. Alden et al., J.
79. R. G. Compton, R. G. Harland, M. B. G.
Pilkington et al., Port. Electrochim. Acta
1987, 5, 271275.
80. R. G. Compton, R. A. W. Dryfe, J. C. Eklund, Research in Chemical Kinetics (Eds.:
R. G. Compton, G. Hancock), Elsevier, Amsterdam, 1993, pp. 239315, Vol. 1.
81. R. G. Compton, R. A. W. Dryfe, Prog. React.
Kinet. 1995, 20, 245297.
82. R. G. Compton, P. R. Unwin, J. Electroanal.
Chem. 1986, 205, 120.
83. W. J. Albery, R. G. Compton, C. C. Jones, J.
Am. Chem. Soc. 1984, 106, 469475.
84. W. J. Albery, C. C. Jones, Faraday Discuss.
Chem. Soc. 1984, 78, 193199.
85. R. G. Compton, A. M. Waller, J. Electroanal.
Chem. 1985, 195, 289296.
86. H. Matsuda, J. Electroanal. Chem. 1967, 15,

325328.
87. P. J. Daly, D. J. Page, R. G. Compton, Anal.
Chem. 1983, 55, 11911197.
88. B. H. Vassos, G. W. Ewing, Electroanalytical
Chemistry, Wiley, New York, 1983, p. 130.
89. S. G. Weber, W. C. Purdy, Anal. Chim. Acta
1976, 100, 531535.
90. R. N. Bagchi, A. M. Bond, R. Colton, J.
91. R. N. Bagchi, A. M. Bond, C. L. Heggie
et al., Inorg. Chem. 1983, 22, 30073013.
92. R. N. Bagchi, A. M. Bond, F. Scholz et al., J.
93. R. N. Bagchi, A. M. Bond, F. Scholz, J. Electroanal. Chem. 1988, 252, 259264.
94. D. A. Fieldler, M. Koppenol, A. M. Bond, J.
Electrochem. Soc. 1995, 142, 862868.
95. R. D. Webster, A. M. Bond, R. G. Compton,
J. Phys. Chem. 1996, 100, 10 28810 296.
96. R. D. Webster, A. M. Bond, B. A. Coles
303311.
97. M. B. Glauert, J. Fluid Mech. 1956, 1,
625627.
98. R. G. Compton, C. R. Greaves, A. M.
Waller, J. Electroanal. Chem. 1990, 277,
8388.
99. W. J. Albery, C. M. A. Brett, J. Electroanal.
Chem. 1983, 148, 201207.
100. J. Yamada, H. Matsuda, J. Electroanal.
Chem. 1973, 44, 189194.
101. R. G. Compton, C. R. Greaves, A. M.
Waller, J. Appl. Electrochem. 1990, 20,
575585.
102. A. Petr, L. Dunsch, A. Neudeck, J. Electroanal. Chem. 1996, 412, 153159.
103. A. Neudeck, L. Kress, J. Electroanal. Chem.
1997, 437, 141148.
104. A. Neudeck, A. Petr, L. Dunsch, J. Phys.
Chem. B 1999, 103, 912922.
105. P.-H. Aubert, A. Neudeck, L. Dunsch et al.,
106. A. Neudeck, A. Petr, L. Dunsch, Synth. Met.
1999, 107, 143158.
107. P. Rapta, R. Faber, L. Dunsch et al., Spectrochim. Acta A 2000, 56, 357362.
108. P. Rapta, A. Neudeck, A. Petr et al., J. Chem.
Soc., Faraday Trans. 1998, 94, 36253630.
109. R. A. W. Dryfe, R. D. Webster, B. A. Coles
et al., Chem. Commun. 1997, 779780.
110. R. D. Webster, R. A. W. Dryfe, B. A. Coles
et al., Anal. Chem. 1998, 70, 792800.
219
220

111. Y. Cheng, D. J. Schiffrin, J. Chem. Soc.,
Faraday Trans. 1994, 90, 25172521.
112. T. Solomon, A. J. Bard, J. Phys. Chem. 1995,
99, 17 48717 492.
113. N. Haren, Z. Luz, M. Shporer, J. Am. Chem.
Soc. 1974, 96, 47884793.
114. R. Male, R. D. Allendoerfer, J. Phys. Chem.
1988, 92, 62376243.
115. A. Gennaro, F. Marron, A. Maye et al., J.
116. R. G. Compton, R. A. Spackman, R. G.
Wellington et al., J. Electroanal. Chem. 1992,
327, 337341.
117. R. G. Compton, R. A. Spackman, R. G.
Wellington et al., J. Electroanal. Chem. 1993,
344, 235247.
118. S. A. Olsen, A. M. Bond, R. G. Compton
et al., J. Phys. Chem. A 1998, 102,
26412652.
119. C. A. Reed, R. D. Bolskar, Chem. Rev. 2000,
1000, 10751189.
120. I. Noviandri, R. D. Bolskar, P. A. Lay et al.,
J. Phys. Chem. B 1997, 101, 63506359.
121. E. G. Janzen, B. J. Blackburn, J. Am. Chem.
Soc. 1968, 90, 5909, 5910.
122. E. G. Janzen, B. J. Blackburn, J. Am. Chem.
Soc. 1969, 91, 44814484.
123. J. D. Wadhawan, F. J. Del Campo, F. Marken et al., J. Electroanal. Chem. 2001, 507,
135143.
124. A. J. Bard, J. C. Gilbert, R. D. Godwin, J.
Am. Chem. Soc. 1974, 96, 620621.
125. T. J. Kemp, Prog. React. Kinet. 1999, 24,

287345.
126. A. C. Testa, W. H. Reinmuth, Anal. Chem.
1961, 19, 13201322.
127. P. D. Moorland, R. G. Compton, J. Phys.
Chem. B 1999, 103, 89518959.
128. R. G. Compton, A. M. Waller, P. M. S.
Monk et al., J. Chem. Soc., Faraday Trans.
1990, 86, 25832587.
129. A. M. Waller, A. N. S. Hampton, R. G.
Compton, J. Chem. Soc., Faraday Trans. I
1989, 85, 773778.
130. A. M. Waller, R. G. Compton, J. Chem. Soc.,
Faraday Trans. I 1989, 85, 977982.
131. R. G. Compton, F. J. Davis, S. C. Grant, J.
132. M. A. M. Noel, R. D. Allendoerfer, R. A.
Osteryoung, J. Phys. Chem. 1992, 96,
239245.
133. J. F. Oudard, R. D. Allendoerfer, R. A.
Osteryoung, J. Electroanal. Chem. 1988, 241,
231238.
134. J. Tang, R. D. Allendoerfer, R. A. Osteryoung, J. Phys. Chem. 1992, 96, 35313537.
135. F. Marken, R. D. Webster, S. D. Bull
209218.
136. J. C. Ball, F. Marken, Q. Fulian et al., Electroanalysis 2000, 12, 10171025.
137. B. A. Coles, R. G. Compton, J. Electroanal.
Chem. 1981, 127, 3743.
3.3 Spectroscopic Applications of STM and AFM in Electrochemistry
3.3
Spectroscopic Applications of STM and

AFM in Electrochemistry
Nongjian Tao
Arizona State University, Tempe, AZ
3.3.1
Introduction
Scanning Tunneling Microscopy (STM) [1]

and the related Atomic Force Microscopy
(AFM) [2] are dream techniques that allow
us to visualize atomic-scale phenomena
on surfaces. Shortly after the inventions,
the capability of the techniques for imaging electrode surfaces in solutions was
recognized and applied to electrochemistry [36]. Since the pioneering works,
various electrochemical phenomena have
been imaged and studied with atomic or
molecular resolution. Examples of such
studies can be found in Volume 3 of this
series and in many reviews [712]. In addition to high-resolution images, STM and
AFM can provide local spectroscopy information of the surfaces and adsorbates.
This chapter provides an introduction to
the spectroscopy applications in electrochemistry. We start with a brief outline
of STM and AFM fundamentals, and then
illustrate the spectroscopy applications using selective examples.
3.3.2
Electrochemical STM Elementary

Concepts
STM consists of a sharp metal tip that is

placed over a conductive sample surface
(Fig. 1). The conduction electrons are
free to propagate within the tip and the
sample, but classically forbidden in the gap
between the tip and the sample in which
the electron total energy is smaller than
the potential energy. However, according
to quantum mechanics, the electrons can

tunnel across the gap and give rise to
a measurable tunneling current when a
bias voltage is applied between the tip and
the sample. The tunneling probability is
proportional to the overlap of the electron
clouds (wave functions, sub and tip ) that
spill out of the tip and the sample surfaces.
The overlap decreases exponentially as
does the gap width, which leads to a simple
relation between the tunneling current (I )
and the sample-tip distance (s) by

(1)
I exp(1.025 s)
where is the work function or the
tunneling barrier height in eV and s
The work function is typically
is in A.
4 eV, so the tunneling current decays
according to
about e2 7.4 times per A,
Eq. (1). The rapid decay means that only
the atom in the closest proximity to the
sample surface dominates the measured
tunneling current, which is the basis of the
atomic resolution imaging of STM. When
scanning the STM tip across the sample
surface with a piezoelectric transducer, the
local tunneling current can be mapped.
A more common approach is to keep a
constant current by adjusting the vertical
position of the tip with a feedback loop.
The vertical position of the tip recorded
as a function of the lateral position
provides a three-dimensional image of the
sample surface.
The simple exponential decay given by
Eq. (1) was veried experimentally in vacuum over a wide range of tunneling
current despite that the simple model
ignores factors, such as imaging force experienced by the tunneling electrons [13].
Replacing the vacuum gap with an electrolyte, Eq. (1) holds approximately for a
limited range of the gap with an effective
barrier height () somewhat smaller than
221
222
Substrate
electrode
f
Ef + eVbias
ysub
Fig. 1 Electron tunneling

between an STM tip and a
conductive substrate.
Tip
electrode
ytip
Ef
the vacuum value [1420]. However, when

measuring the tunneling current over a
wide range of the gap, Lindsay and coworkers [18, 19] observed a signicant deviation
from the simple exponential dependence.
Hong and coworkers [20] reported that the
tunneling barrier for the electron tunneling from tip to substrate was remarkably
different from the reverse direction. Both
the nonexponential decay and asymmetric tunneling demand a better understanding of electron tunneling across an
electrolyte gap. Schmickler and Henderson [21] treated the water molecules in
the tunneling gap as a dielectric medium
and predicted a smaller tunneling barrier
than in vacuum. Kuznetsov and Ulstrups
considered the effects of solvent uctuations on the tunneling barrier height
and tunneling rate [22]. They found that
the solvent uctuations could enhance
the tunneling rate, thus leading also to
a lower effective barrier height. More elaborate models involving resonant tunneling
through solvent states [18, 23] and ab initio
simulations [24, 25] were also developed.
The STM image cannot always be literally interpreted as a geometrical topography of the sample surface because
the tunneling current depends on the
electronic states of the tip and the sample surface. The tunneling current across
a vacuum or insulation gap between the
tip and the sample surface is given by [26]

4e +
I =
[f (EF eVbias + )
h
f (EF )]s (EF eVbias + )
T (EF + )|M|2 d
(2)
where f (E) is the FermiDirac distribution function, EF is the Fermi energy of

the tip, Vbias is the bias voltage, T () and
S () are the local density of states (DOS)
of the tip and sample, respectively, and
|M|2 is related to the overlap between the
tip and the sample wave functions that
decays according to Eq. (1). When the temperature is low, (kB T EF ) and |M|2 is
independent of energy, the tunneling current is a convolution of the tip and the
sample DOS over an energy range eVbias ,
eVbias
S (EF eVbias + )
I
0
T (EF + ) d
(3)
In the presence of a molecular adsorbate,

one may treat the combined adsorbate and
the substrate as the sample. However,
many models have been proposed to

treat the adsorbate more explicitly both
in vacuum [2735] and in aqueous solutions [21, 3640]. For example, Kuznetsov
and Ulstrup [36, 41] pointed out that a
strong coupling between the electronic
states of a molecule and the environmental uctuations could affect the tunneling
current by lowering the effective tunneling barrier and by momentarily bringing
a molecular state to align with the substrate/tip Fermi levels. In addition to
the adsorbate-solvent coupling, Schmickler and coworkers [38, 42] took into account the adsorbate-substrate coupling.
They considered explicitly a resonant tunneling process when the molecular state
was brought to align with the substrate/tip
Fermi levels. We will return to these theories later.
3.3.3
Electrochemical AFM Elementary

Concepts
AFM is similar to STM in many aspects,

but a fundamental difference between
them is that STM measures the tunneling current, while AFM probes the force.
So unlike STM, which requires a conductive surface or small adsorbates on
a conductive surface, AFM can provide
high-resolution images of both conductive
and insulating surfaces. This is important
for imaging large biological molecules or
poorly conductive materials on electrodes.

The tip-sample force in AFM is typically
detected by measuring the bending of a
exible cantilever on which the tip is attached (Fig. 2) [43]. As the tip approaches
the sample surface, it rst experiences
long-range forces, such as the van der
Waals and electrostatic interactions, then
specic chemical bonds between the tip
and the sample surface, followed by a
rapid increase in the hard-core repulsive
force [44]. In the electrochemical environment, additional forces such as ionic and
hydration forces arise [45]. These forces
will be discussed in greater details later.
AFM can be operated in either the
attractive or repulsive-force modes. In the
repulsive mode, or the so-called contactmode AFM, the force is highly stable
and images that show the periodic lattice
structures of the samples are routinely
obtainable. However, most of the reported
images show either perfectly ordered
atomic structures or defects that are
much greater than an atom. Ohnesorge
and Binnig [46] investigated this issue by
imaging calcite in water. They observed
atomic-scale defects when the net repulsive
force applied on the sample surface by
the AFM tip was 0.1 nN or smaller.
Increasing the repulsive force, they found
that sharp monoatomic steps were slowly
wiped away because the large contact force
resulted in a contact area between the
tip and the sample much greater than
F = kz
z
(a)
Fig. 2
(b)
Tipsample interactions in AFM.
(c)
223
224
an atom. The subsequent AFM image,

nevertheless, still showed well-ordered
two-dimensional lattice structures owing
to the periodic modulation of the force by
the periodic lattice of the sample surface.
The experiment shows that most AFM
experiments operated with a contact force
much greater than 0.1 nN are probably not
truly atomic resolution.
The contact force can be avoided by
operating in the attractive force mode or
the noncontact mode. For high-sensitivity
measurement of the attractive force, an AC
technique was introduced. The technique
measures the change in the resonance
frequency of the AFM cantilever. When
the tip is far away from the sample surface,
the resonant frequency f0 is given by,

k
f0 = 2
(4)
m
where k is the spring constant of the
cantilever and m is the effective mass.
When the tip approaches the sample,
the tipsample interaction changes the
spring constant by k = Ftip sample /z.
Consequently the resonance frequency
changes by an amount of
f
k
f0
2k
(5)
assuming that both the changes in the

spring constant and oscillation amplitude
of the cantilever are small [47]. One way to
implement the idea for AFM imaging is
to drive the AFM cantilever at or near the
resonance frequency and measure the oscillation amplitude as the feedback signal.
When the tipsample interaction changes,
the amplitude changes because of the
change of the resonance frequency. This
method has a low response time when the
resonance is very sharp. The response time
is usually acceptable for imaging in solutions because the cantilever experiences
much greater damping in solutions, which

broadens the resonance. Faster response
time can be achieved by directly following
the resonance frequency [48]. This later
technique has been used to obtain atomic
resolution images of Si(111) 7 7 structure [49, 50]. We note that even if tipsample contact is involved, the AC mode
is still helpful for imaging soft materials,
such as biological molecules adsorbed on
electrodes, because the contact is intermittent. Commercial AFM, such as Tapping
Mode AFM [5153] from Digital Instruments and MAC Mode AFM [5456] from
Molecular Imaging Inc. are two examples.
3.3.4
Selective Examples of Applications
STM and AFM applications in electrochemistry as imaging tools are numerous. Novel nonimaging applications are
also abundant in the literature [57, 58].
Because of space limitations, here we focus on tunneling spectroscopy and force
spectroscopy applications using selective
examples that are directly related to electrochemistry. Other important nonimaging
applications that are relevant to electrochemistry include fabricating nanostructures [5968] and probing fast kinetics [6975].
Tunneling Spectroscopy
The electronic states of a sample can
be determined with STM by performing
Scanning Tunneling Spectroscopy (STS).
The concept of tunneling spectroscopy
was developed as early as in the sixties
using metal-oxide-metal junctions [26, 76].
The original idea of building the STM by
Binnig and Rohrer was actually to perform
tunneling spectroscopy locally on a small
area. According to Eq. (3), the tunneling
current is a convolution of the tip and
3.3.4.1
the sample DOS. In order to determine

the sample DOS, one has to know the
tip DOS, which is not always possible.
However, for a given voltage range, the tip
DOS is often weakly dependent on energy,
, then Eq. (3) becomes,
dI
S (EF eVbias + )
dV
(6)
Because the tunneling current is very

sensitive to the tip-sample separation,
the measured spectroscopic data are often
normalized by I /V and given by [(V /I )
(dI /dV )] = [(d ln(I ))/(d ln(V ))] [77].
In vacuum, STS is usually obtained by
xing the tip over the region of interest
and then measuring tunneling current by
sweeping the bias voltage. The method is,
however, difcult to apply to STM in the
electrochemical environment because the
sweeping bias voltage induces a large polarization current that can overwhelm the
tunneling current. Furthermore, faradic
current at large bias voltage limits the
spectroscopy to a relatively small range
of bias voltage. Despite the difculties,
useful electronic information has been
obtained by performing tunneling spectroscopy outside the electrochemical cell
or by indirect methods.
Redox States An early attempt
to extract DOS of redox molecules from
tunneling current was made by Moriaki and coworkers [78]. They used thin
SiO2 tunnel barriers on Pt silicide substrates for tunneling spectroscopy of complex ions, such as Fe(CN)6 3/4 and
Fe(EDTA)1/2 . The energy dependent
DOS extracted from the tunneling spectra was in good agreement with the
Classical MarcusGerischer model [79].
The experiment was macroscopic because
it measured a large number of molecules.
In contrast, STS has the capability of
3.3.4.1.1
probing the electronic states of a single

molecule or even a particular group within
a molecule.
Schmickler and coworkers [38, 42] suggested that lowest unoccupied molecular
orbital (LUMO) and highest occupied
molecular orbital (HOMO) of molecules
adsorbed on an electrode could be probed
by STM. The LUMO (HOMO) is broadened and shifted as a result of the
coupling of the molecule to the environment, which is described as density
of oxidized (reduced) states. They have
considered a coherent one-step electron
transfer from the tip electrode to the substrate electrode or vice versa via a resonant
transition through the oxidized states. An
electron transfers from one electrode to
the molecule and then to the second electrode before the molecule relaxes to the
reduced state. This implies that the electron tunneling time is much faster than
the vibration time of the molecule, which
is on the order of 1014 s. If assuming the
tunneling matrices, the tip, and the substrate DOS are independent of energy, ,
the model gives,

I
eVbias
Dox ( e + eVbias ) d
(7)
where is the overpotential and Dox is the
density of the oxidized states. When the
molecule couples weakly to the substrate
and the tip, Dox takes the Gaussian form

Dox () =

( )2
exp
kB T
4kB T
(8)
where is the reorganization energy. The
tunneling spectrum, dI /dVbias versus Vbias
gives approximately the density of states
(Dox ). In electrochemical STM, the overpotential () can be varied independent of
the tip-substrate bias voltage (Vbias ). So by
225
226
measuring the tunneling current as a function of the overpotential with a xed Vbias ,
one has a new way to perform tunneling
spectroscopy. It follows from Eqs. (7 and
8) that I versus gives also the density of
states for a small Vbias .
Kuznetsov and Ulstrup [22, 36, 41] have
considered another limit, a sequential
two-step process, in which the electron
transfers from the tip (substrate) to
the molecule and reduces the molecule.
The reduced states of the molecule are
located below the Fermi level of the
substrate, but a thermal uctuation can
shift the states up and then allow a second
electron transfer from the molecule to the
substrate (tip), returning the molecule to
the oxidized state. Neglecting backward
owing current, the current caused by this
two-step process is
I
Rsm Rmt
Rsm + Rmt
(9)
where Rsm and Rms are proportional to the

electron transfer rates from the substrate
to the redox molecule, and from the redox
molecule to the tip, respectively. Again, if
we assume energy independent DOS for
both the tip and the substrate, then Rsm
and Rms are given by,
+
Rsm
f ()

( E0 + eVbias )2
exp
d
4kB T
(10)
and

Rmt
[1 f ( + eVbias )]

( E0 + + eVbias )2
d
exp
4kB T
(11)
where eVbias is the energy shift in the

redox levels caused by the tip-substrate bias
voltage, and is between 0 and 1, reecting
the distribution of the bias voltage across
the tunneling gap. This two-step electron
transfer can also give rise to a maximum
at a potential depending on the relative
contributions from the two steps.
Friis and coworkers [80] considered an
intermediate mechanism, in which an
electron rst transfers to the molecule and
the molecule begins to relax towards the
reduced state. However, before it is fully
relaxed to the reduced state, an electron
transfer from the molecule to the second
electrode occurs when the temporarily occupied level passes the Fermi level of the
second electrode. They called this process coherent two-step electron transfer.
More recently, Kuznetsov and Ulstrup [81]
have developed a systematic theory for
the sequential two-step process. The electron transfers from one electrode to the
molecule and reduces the molecule, and
then transfers to the second electrode and
reoxidizes the molecule, so the process is
reviewed as a cycle of consecutive molecular reduction and reoxidation. A particular
interesting feature in the theory is that each
reduction-reoxidation cycle is composed
of a large number of individual electron
transfer events between the molecule and
both the tip and the substrate. This signicantly enhances the tunneling current
compared to a single electron transfer.
Experimentally extracting redox state
information from STS has been also
pursued. Snyder and White measured
STS of Fe-protoporphyrin IX multilayer lms adsorbed on graphite electrode in air [82]. They observed an enhancement that was identical for both
signs of the tip-substrate bias voltage.
The data were interpreted in terms of
tunneling via the adsorbate metal centers.
Tao studied electroactive molecule, Feprotoporphyrins IX, coadsorbed with

nonelectroactive protoporphyrins IX on
graphite substrate in electrolyte [83]. Because of the difculty in obtaining reliable
STS in electrolyte, the apparent height
of Fe-protoporphyrins IX relative to that
of protoporphyrins IX was measured as
a function of the electrochemical potential. A large increase in the tunneling
current occurred when the potential was
tuned to the redox potential. This observation was originally interpreted as a
resonant tunneling via the oxidized state
of Fe-protoporphyrins IX in terms of the
coherent one-step model [83, 84], but the
coherent one-step model predicts a maximum shifted away from the redox potential
by , the reorganization energy. This,
although, could be due to a small reorganization energy or experimental uncertainty
in the potential, the sequential two-step
electron transfer offers an alternative explanation [81].
Han and coworkers [85] studied STS of
a series of metal-porphyrin compounds
adsorbed onto a Au(111) electrode in an organic solvent (mesitylene). The molecules
were modied with thiol groups at one

end so that they were tethered to the
surface via the strong S-Au bonds. The
derivative of these curves (dI /dV ) had a
Gaussian-shaped peak at a voltage characteristic of the compound (Fig. 3). These
bias-dependent features were not observed in the less electroactive molecules,
so the STM was shown to be capable
of distinguishing electroactive molecules
from nonelectroactive molecules. They attempted to t their data with both the
coherent one-step and sequential twostep models outlined above. Both models yielded Gaussian-like peaks in the
dI /dVbias versus Vbias plot. The Gaussian
predicted by the rst model (Eqs. 7 and
8) was located at a lower potential than
that by the second model, but they could
not distinguish the two models within the
experimental error.
Hipps
and
coworkers
obtained
STS metal-phthalocyanines adsorbed
on Au(111) surfaces with STM
in conjunction with the tunneling
spectroscopy obtained with the metalinsulator junctions [8688]. STM images
showed that the molecules lay at on
0.03
FeBr
0.025
FeCl
0.02
Mn
dI/dV
0.015
2H
0.01
0.005
0
0.005
1000
Zn
Bare electrode
800
600
400
200
Bias
[mV]
Derivatives (dI/dVbias ) of the tunneling current versus bias voltage
curves of Fe, Mn, Zn, and H embedded in tetraphenol porphyrin (TPP).
(From Ref. [85] with permission.)
Fig. 3
227
228
the surface, and the brightness of the

central part of the molecule depended
on the metal ion located there [86, 89].
The measured resonant tunneling via the
molecular orbital shown in the STS was
in good agreement with the macroscopic
tunneling spectroscopy. Although the
experiments were measured in vacuum,
where reliable tunneling spectroscopy
can be more easily obtained than in
air or in water, they demonstrated a
correlation between the orbital-mediated
tunneling spectroscopy data and the redox
potentials of the metal-phthalocyanines in
solution [87, 90].
In addition to the relative simple redox
molecules, electrochemical STM has been
applied to study redox proteins, such as cytochrome c [9193], blue copper protein
Pseudomonas aeruginosa azurin [9496],
and PS I system [97]. An interesting observation by Ulstrup and coworkers is a
spike localized near the center of each
azurin molecule, which means a much
higher electron tunneling through the region of the spike [94]. The spike was
assigned to the electron transfer via the
redox center, copper, in azurin, which
enhanced the tunneling current. The tunneling current, however, depended only
weakly on the bias voltage and the electrochemical overpotential, which could be because of the relative large bias voltage used
in the experiment. This study provides
the rst evidence of local spectroscopic
information of redox proteins with STM.
identication of other species has also

been demonstrated. Schott and White [98]
studied a mixed monolayer of halogen
atoms (F, Cl, Br, and 1) on Ag(111),
in addition to the different radii and
they observed differences in the tunneling
spectra of the halogen species. Another
interesting molecular identication with
STM was carried in monolayer
lms of several primary substitute
hydrocarbons CH3 (CH2 )n CH2 X (X=CH3 ,
OH, NH2 , SH, Cl, Br, and I) on
graphitesolution interfaces by Flynn
and coworkers [99, 100] The hydrocarbons
formed well-ordered arrays on the surface.
The NH2 , SH, Br, and I groups appeared as
bright spots in the STM images while OH
and Cl groups were not distinguishable
from the alkyl chain, so it is possible
to identify the molecules with different
substitutes according to their brightness
in the STM images. Claypool and
coworkers imaged a series of substituted
alkanes and alkanols on graphite [101].
They found that all the functional
groups (halides, amines, alcohols, nitriles,
alkenes, alkynes, ethers, thioethers, and
disuldes) studied in their experiments
could be distinguished from each other
except for Cl and OH groups. Their results
were consistent with quantum chemistry
calculations [101]. Recently, Claypool and
coworkers extended the study to MoS2
surface and the images of tetradecanol
on MoS2 were nearly identical to those
obtained on graphite [102].
3.3.4.1.2 Molecular Identications Because STM probes the electronic

properties of the molecules, it can identify
structurally similar molecules on the
basis of the difference in the electronic
states of the molecules. The works on
redox molecules described above are good
examples of such applications. Molecular
3.3.4.1.3 Molecular Electronics As silicon-based microelectronics is heading towards the increasingly difcult road of
nanometer scale, building electronic devices with individual molecules (molecular
electronics) becomes a promising alternative [103, 104]. Indeed, many molecules
possess wonderful electronic properties
that could be used as basic elements in

a functional electronic device. A major
challenge is to nd a practical method
to wire individual molecules to the outside
world. STM can wire a single molecule
with the sharp tip and probe electron
transfer through individual molecules. A
large number of works have already been
performed [86, 88, 105112]. Gimzewski
and coworkers have published an excellent
review recently [113], we describe here an
interesting example that is based on the
redox properties of molecules.
Gittins and coworkers [114] explored
molecular electronics applications using
STS and electrochemistry (Fig. 4). Their
system consisted of gold nanoparticles
attached to a gold electrode surface via
the redox-active bipyridinum moiety. They
controlled the electron transport between
the nanoparticles and the electrode by controlling the redox state of the bipyridinum
moiety. They measured the decay constant
(k in Eq. 1) that dropped sharply on reduc
tion of bipy2+ to bipy + , from 16 nm1
to 7 nm1 . They attributed the observed
change to the injection of charge into the
redox group, and thus showed that the
electronic properties of the junction could
be switched by the redox state. They also
measured STS in mesitylene and observed
a sharp peak in the dI /dVbias versus Vbias
plot (Fig. 4). Because of the stability of
the bipy + state, the peak was interpreted

as a two-step electron transfer through
a temporary population of the molecular
redox state.
Force Spectroscopy
AFM has been applied to study various
surface forces [44, 115117]. The principle
is similar to the surface force apparatus (SFA) that was developed some years
ago [45]. SFA measures the force between
two curved atomically smooth surfaces.
3.3.4.2
The separation between the two surfaces

is controlled by a piezoelectric transducer
and measured by an optical interferometric technique. Because SFA uses surfaces
with known geometries, precise measurements of surface forces and energies are
possible. However, SFA measures the
force of a relative large surface area and
is limited to atomically at surfaces. That
is why most of the SFA studies have been
carried out on mica surfaces. AFM measure the force between a tip and a sample
surface. A drawback is that the AFM tip
geometry is not always precisely known,
although attempts to control the tip geometry have been made by attaching an object
with known geometry onto the regular
AFM tip. Examples include colloidal particle [118, 119] and carbon nanotube [120].
However, being a local probe, AFM does
not require atomically at sample surfaces.
Furthermore, AFM allows one to image the
sample rst and then place the tip over the
region of interest.
As mentioned before, AFM can measure
surface forces using two different operation modes, DC and AC. In the DC mode,
one measures the deection (Z) of a
cantilever as a function of the tip-sample
distance (D) that is varied usually by a
piezoelectric transducer (Fig. 3). The tipsample force is given by Hookes law in
terms of Z, F = kZ, where k is the
spring constant of the cantilever. This force
clearly depends on the tip-sample distance,
D, which is given by D = Z Z, in the
absence of sample deformation. Z in the
expression is the displacement of the piezo
and the one that can be controlled in the
experiment. When the tip is far away from
the sample (large D), the force is zero.
When the tip approaches the sample, it experiences various forces, electrostatic, van
der Waals, double-layer, solvation forces,
and so on. This is the regime of interest in
229

Molecule (1) used in redox gate
STM tip
N
S
8
2 Br
Au nanoparticle
6 nm
Counter
electrode
Redox
gate
3 nm
Au substrate
Current
[nA]
30
dI/dV
[nA V1]
230
20
5
0
2
Substrate bias
[V]
10
0
2.0
1.5
1.0
0.5
0.0
Substrate bias
[V]
(a) Schematic representation of a nanoscopic device. Electrons
can be injected into the redox gate by applying a suitable potential
between the substrate and the counterelectrode. (b) A currentvoltage
curve (inset) and its derivative (dI/dVbias ) in mesitylene. The dotted line
corresponds to the control experiment, in which gold nanoparticles were
deposited on a hexanedithiol monolayer. (From Ref. [114] with
permission.)
Fig. 4
terms of measuring surface forces. Finally,

when D approaches zero, the tip is in contact with the sample surface. The plots
of Z versus Z provide the tip-sample
force, F (D), and are often referred to as
force-distance curves.
As we have already mentioned, the AC

mode involves an oscillating AFM tip.
When the tip is driven towards a sample
surface, the oscillation is altered by the
tipsample interactions. If the cantilever
is driven to oscillate at a xed frequency,
the changes in the oscillation amplitude

can be used to obtain the tip-sample force
information. An alternative approach is
to use a feedback to constantly monitor
the resonance frequency that is changed
by the tipsample interactions. The AC
force-distance curves are often referred
to as dynamic force-distance curves. We
note that for small oscillation amplitude,
the change in the resonance frequency is
proportional to the derivative of the force
with respect to the tip-sample distance or
differential stiffness (Eq. 5).
Double-layer Forces Doublelayer forces arise when two charged
surfaces separated with an electrolyte
approach each other. The charge on each
surface is balanced by an equal but
oppositely charged region of counterions
in the electrolyte. Some of the counterions
are adsorbed on the surface, either
transiently or permanently, within the
so-called Helmholtz layer. Beyond the
Helmholtz layer is a so-called diffuse
electric double layer in which the ions
are in rapid thermal motion. When
bringing two such charged surfaces
together, a double-layer force appears that
can be traced to an electrostatic and
entropic contribution. The electrostatic
contribution favors an attraction between
the surfaces because they are pulled
together by the oppositely charged
counterions. The entropic contribution
is, however, repulsive because the
congurational entropy favors more space
between the surfaces. Because the van der
Waals force always presents between two
surfaces regardless the surface charge, it
needs to be considered together with the
double-layer force. Such a treatment was
given by the well-known DLVO theory [45].
For a spherical tip geometry with a radius
R and a at substrate, the van der Waals
force is given by Ha R/6D 2 [45], where

sign reects attractive force, Ha is
Hamaker constant.
Assuming two semi-innite surfaces
with xed surface charge density, Parsegian and Gingell [121] obtained an analytical expression for the double-layer
force in the limit of small surface potentials (<25 mV). Butt [122] calculated the
double-layer force between a at sample
and a spherical tip with radius R using the
Derjaguin approximation. If KD R 1, his
results can be simplied to
3.3.4.2.1
Fdl =
2R
[( 2 + S2 )e2KD D
0 KD T
+ 2T S eKD D ]
(12)
where T , S are the surface charge

densities of the tip and sample, and
KD is the inverse of the Debye length
(1/KD ). The Debye length depends on the
concentration and the valence of ions and
is given by
1
c
= nm
KD
[X]
(13)
where [X] is the electrolyte concentration

in moles, c = 0.304, 0.176, and 0.152 for
1 : 1, 1 : 2 (or 2 : 1), and 2 : 2 electrolytes,
respectively.
If T < S , that is, the tip is neutral,
the rst term in Eq. (6) dominates and
the double-layer force decays exponentially
with a decay length, 1/2KD . This term is
always positive and thus gives a repulsive
force, while the second term can be either
repulsive or attractive. The second term decays with 1/KD , slower than the rst term,
so the second term dominates the solvation
force at a large tip-sample separation.
231
232
The above models assume constant

surface charge densities. Assuming constant surface potentials, Lin and coworkers [123] obtained
Fdl = 40 RKD T S eKD D
(14)
where T and S are surface potentials

of the tip and the sample, respectively.
In most practical cases, it is likely that
neither the surface charge nor the potential
stays constant.
AFM has been used to measure doublelayer forces between many insulator surfaces, including Si3 N4 tips on mica, diamond tips on mica, glass tips on glass, ZnS
colloidal particle on mica, and polystyrene
colloidal particle on polystyrene colloidal
particle [119, 124, 125]. In these papers,
the experimental data were t with the
second term of Eq. (6) or Eq. (8). The
tting allowed Weisenhorn and coworkers [124] to extract the Debye length, which
was found to agree nicely with Eq. (7). In
addition to ionic concentration, other experimental parameters, such as pH, were
also varied, which allows the determination of the point of zero charge (isoelectric
point) [126].
The double-layer forces between a metal
and an insulator as well as between two
metal surfaces have also been studied. For
example, Biggs and coworkers [127, 128]
have measured forces between a gold
colloidal sphere and a at gold substrate
as a function of ionic concentration and
pH. Several groups have carried out
measurements of the double-layer forces
with molecular adsorptions taking place
on the metal surface [129133].
Most of the studies were focused on relative low surface potentials because of the
difculty of making clean surfaces with
high potentials. In the case of metal surfaces, one can easily obtain a high surface
potential by applying an external potential
to the metal electrode relative to a reference electrode in the electrolyte [134137].

Hillier and coworkers [137] studied the
force between a silica probe and a gold
electrode (Fig. 5). The force was attractive
at open circuit, but it depended strongly
on the potential applied to the gold and the
type of the electrolyte. It was attractive at
positive potentials and became repulsive
at negative potentials. This was because
the silica tip was negatively charged at a
pH of 5.5. They carried out a systematic
study of the force as function of the applied potential and found that the repulsive
force decreased as the potential increased
and passed through a minimum, which
they referred to as the potential of zero
force (pzf). They measured the pzf values
in solutions of NaF, KCl, and KI, which
showed a close correspondence between
pzf and pzc (potential of zero charge).
At large tip-surface separations, theoretical ts of the measured force to solutions
of the PoissonBoltzmann equation were
excellent. At small separations, however,
neither constant surface charge (Eq. 6) nor
constant surface potential boundary conditions (Eq. 8) provided good ts. The poor
ts at small separations were attributed
to an overestimate of the Hamaker constant, an ill-dened position for the plane
of surface charge, and the presence of
short-range solvation forces.
Raiteri and coworkers [135] carried out a
similar measurement of the force between
a silicon nitride tip and a platinum or
gold substrate and between two conductive
surfaces, a gold sphere glued to an AFM
cantilever and a gold electrode in 0.1 M
KCl. They observed a large decrease in a
repulsive force around the pzc of the gold
electrode, which agreed qualitatively with
the PoissonBoltzmann theory. Arai and
Fujihira [138] observed a similar repulsive

1.0
V
[mV]
F/R
[mN m1]
0.6
1.0
0.8
F/R
[mN m1]
0.8
700
500
400
300
200
100
0
+100
0.4
0.6
V
[mV]
0.4
0.2
700
400
200
100
0
+100
0.0
0.2
0.4
0.2
8 10 12 14 16 18 20
d
[nm]
0.0
0.2
0.4
0
10
20
30
40
50
60
d
[nm]
Fig. 5 Force between silica sphere and gold electrode in an aqueous solution of 1 mM KCl at
25 C and pH = 5.5 as a function of the applied potential at gold electrode. The curves
correspond to, from top to bottom, electrode potentials of 700, 500, 400, 300, 200,
100, 0, and 100 mV (vs SCE). Electrostatic repulsion decreases as the electrode potential
increases from 700 to 100 mV. Inset: force data for silica sphere and gold substrate in
10 mM KCl solution. (From Ref. [137] with permission.)
force between a gold-coated AFM tip and a

gold substrate at pH = 10.9.
3.3.4.2.2 Solvation Forces When the tipsample distance is smaller than a few nm,
the DLVO theory, that is, an attractive van
der Waals force and a repulsive doublelayer force, is often not adequate. One
possible reason is the break down of
the approximation that both the tip and
the substrate are treated as continuous
media. Another reason is because other
forces arise at the small distances. One
of such forces is the so-called solvation
force that arises when solvent molecules
between the tip and substrate are ordered
or form hydrogen-bonded network under
the inuence of the tip and the substrate.
When the tip approaches the substrate, it

has to break the order or hydrogen-bonded
network, so the measurement of solvation
forces provides ordering information and
discrete nature of solvent molecules near
the solid surfaces. The most general type
of solvation force is oscillatory, arising
from the ordering of the solvent molecules
into quasi-discrete layers near the surface.
This solvation force can be approximately
described by a cosine function with an
exponentially decaying amplitude [45],

2D
cos
d
eD/d
(15)
where d is the diameter of the solvent

molecule. Both the oscillation period and
233
234
the characteristic decay length are close

to d.
AFM has been used to study the solvation forces due to interfacial water
and organic liquids [139142]. Hoh and
coworkers [140] observed a discrete (oscillating) adhesive interaction between an
AFM tip and a solid surface. They attributed the oscillations to either individual
hydrogen bonds between the tip and surface, or ordered water layers near the
surface. OShea and coworkers used an
AFM in the DC mode to measure the
force on the AFM tip in ocatamethylcyclotetrasiloxane (OMCTS), n-dodecanol,
and water during the tip approaching the
graphite surface [141]. For both OMCTS
and n-dodecanol, they observed an oscillatory force that was related to the layered
structure of the solvents. But the oscillatory
force did not appear in the case of water.
Later they used a magnetically oscillated
tip microscope to directly measure the local compliance of ordered liquid layers of
OMCTS and n-dodecanol on graphite and
on mica [141]. However, the spacing between layers of OMCTS was substantially
smaller than the known smallest dimension of the molecule so that an orientation
of the molecules in the layers could not
be determined.
Using an AFM operated in the magnetic AC mode, Han and Lindsay [142]
determined the layered structure of solvents at interfaces with high accuracy
and reproducibility (Fig. 6). They found
that the AFM cantilever amplitude oscillated in OMCTS and mesitylene as the tip
approached the surface graphite. Similar
oscillations were also found when the tip
retracted away from the surface. The oscillations became smeared out gradually
when increasing the tip-surface distance,
showing a transition from highly ordered
solidlike structure to liquid structure. The
oscillations reect the layered structure

of the solvents near the surface, and the
oscillation periods give the spacing between two adjacent layers. For OMCTS,
the period was found to be 0.82 nm,
in good agreement with the oblate shape
of OMCTS with a major diameter of
1.0 to 1.1 nm and a minor diameter of
0.7 to 0.8 nm. The data for mesitylene
was similar to that of OMCTS, except
that the average period of the layers was
measured to be 0.45 nm. This spacing is consistent in that the mesitylene
rings lie at with respect to the graphite
surface.
3.3.4.2.3 Surface Stress Surface stress is
closely related to another quantity, surface
tension ( ), the energy cost per unit area
to create a surface, according to

(16)
ij = ij +
ij T
where ij , and ij are components of

the surface stress and strain tensors, and
ij = 1 when i = j and ij = 0 when
i = j . So, in general, the surface tension
and the surface stress are not identical.
However, the two quantities are identical
when the surface tension is independent of small strains (the second term
in Eq. 9 vanishes). This is true only for
liquid or liquidlike surfaces that are free
to rearrange in response to small perturbations. Surface stress is an important
thermodynamic quantity in electrochemistry and often known as electrocapillarity. The name originates from the
measurements of the interfacial tension
between a mercury electrode and an electrolyte in the earliest days of electrochemistry [143]. Since the surface tension of
a liquid electrode is identical to the surface stress, the Lippmann equation can

(a)
Oscillation amplitude of cantilever

[]
(b)
10
Z position
[nm]
Fig. 6 Two independent measurements of force spectra of the MAC mode
AFM at OMCTS-graphite interface. The oscillation amplitude of the magnetic
cantilever driven by an external magnetic eld oscillates in both approaching
(solid line) and retracting (dotted line) curves in the region of a few nanometers
away from the surface due to ordered layers of OMCTS molecules at the
interface. The arrows on this and subsequent plots correspond to
repulsive-force maxima. (From Ref. [142] with permission.)
be derived for the mercury electrode that

relates the electrode potential, the surface charge, and the interfacial stress.
The Lippman equation predicts that the
interfacial tension goes through a maximum at pzc, which has been used to
determine pzc. Surface tension continues
to be a useful quantity to measure in

todays research. For example, an electrochemical/thermodynamic procedure to
determine surface tension, excess, and
pressure has been developed and applied
to study molecular adsorptions on welldened crystals [144].
235
236
Surface stress can be measured by

a number of experiments, including
piezoelectric elements [145], ribbon electrodes [146], and more recently developed micromechanical sensors [147149].
It can be also measured with AFM
and STM, as demonstrated by several
groups [150, 151]. The AFM and STMbased surface stress measurements usually detect the bending of a thin plate
(cantilever) as a result of a surface stress
change at one of the two surfaces. If the
plate is isotropic and freely standing, the
change in the surface stress can be calculated from the amount of bending by [152]
=
Et 2
z
4L2 (1 )
(17)
where E, t, L, and are the thickness, length, Youngs modulus, and Poisson ratio of the cantilever, respectively.
This equation has been widely used to
determine changes in surface stress although one end of the cantilever is often
clamped in experiments. Dahmen and
coworkers [153] recently performed a nite
element analysis of single crystal plates
(anisotropic) with one end clamped. They
concluded that Eq. (10) derived for the unconstrained bending could still be used in
most cases with appropriate parameters.
It is clear from Eq. (10) that the bending
of a cantilever because of a surface stress
is (L/t)2 , which requires long and
thin cantilevers for high sensitivity. When
trying to improve the sensitivity, one has to
consider also the noise caused by environmental vibration in the measurement. In
order to minimize the environmental vibration, one requires a high-resonance fre
quency, f = 0.163(t/L2 ) E/, in which
is the density of the cantilever material.
In AFM, microfabricated silicon or silicon nitride cantilevers coated with various
metals on one of the two faces are used to

satisfy both requirements.
The microfabricated cantilever-based
surface stress measurements have recently found many applications: changes
in mass [154], heat dissipation [154, 155],
magnetic eld [156], infrared light [157],
and polarized light [158], to name a few.
There is a recent surge of interest in using
the cantilevers for chemical and biological
sensors in ambient air or aqueous environments [159161]. For example, if one side
of the cantilever with receptor molecules
is coated and the other side is left inert,
then a specic binding of chemical or biological species onto the receptor molecules
results in a surface stress change in the
cantilever. Fritz and coworkers [162] studied DNA hybridization and receptor-ligand
binding using microfabricated cantilevers
that were functionalized with a selection
of biomolecules, and they were able to detect a single base mismatch between two
12-base pair DNA oligomers.
The microfabricated cantilevers have
been used to study surface stress induced
by electrochemical processes [163166].
Brunt and coworkers [165, 166] monitored
surface stress changes due to the electrochemical deposition and dissolution of a
Pb and Ag on an Au(111)-coated AFM cantilevers. The changes corresponded nicely
with the peaks in the cyclic voltammograms. Raiteri and coworkers [164] deposited the two sides of a gold-coated
cantilever with different thiols, octadecanethiol, and 3-mercaptoprionic acid, respectively. The cantilever was exposed to
a constant ow of aqueous electrolyte solution with varying pH and the surface
stress-induced deection in the cantilever
was constantly measured. They found that
the change in the surface stress had a
minimum around pH 4 to 5. Increasing pH above 5 or decreasing it below 4
resulted in a bending of the cantilever

toward the mercaptoprionic acid side. They
attributed the observation to the proton dissociation of the carboxyl group,
which lowered the free enthalpy of the
carboxylate group. However, the simple
mechanism was not adequate to explain
the small increase in the surface stress
when reducing the solution pH below 4.
When replacing the mercaptoprionic acid
monolayer with 2-aminoethanethiol and
2-mercaptoethanol, the surface stress increased monotonically with increasing pH.
Another difference was that the surface
stress change was much smaller than
that of mercaptoprionic acid, in consistence with the smaller degrees of proton
dissociation of 2-aminoethanethiol and 2mercaptoethanol.
Measuring the surface stress of the AFM
cantilever has a number of advantages over
other approaches, but the approach is not
suitable if one wants to study the surface stress of a well-dened single crystal.
This task can be performed with an STM
setup [151, 167169]. The sample in STM
can be either a single crystal plate or thermally evaporated metal lms on glass plate
with one end clamped. One side of the single crystal plate must be covered with, for
example, nail polish. An STM tip is placed
over the other end to detect the surface
stress-induced bending in the plate. The
bending causes a change in the tunneling gap between the tip and sample and
is reected in a change of the tunneling
current. The STM feedback maintains a
constant current by adjusting the vertical
position of the STM via a z-piezoelectric
transducer. The amount of the adjustment
gives the deection, z, from which the
surface stress is extracted using Eq. (10).
A nice feature is that the same STM tip
allows one also to image the surface.
3.2
Au(100) 1 1
4.4
2.8
Au(111) 1 1
4.2
2.4
Au(111) rec
4.0
2.0
(111) (1 1)
3.8
(111) (rec)
1.6
400
200
200
400
600
800
Potential vs. SCE

[mV]
Surface stress versus the potential for the initially reconstructed and
unreconstructed surfaces of Au(100) and Au(111). All measurements are
from negative to positive potentials. (From Ref. [151] with permission.)
Fig. 7
1000
3.6
Surface stress (100) t (s)

[N m1]
(100) (hex)
Au(100) hex
Surface stress (111) t (s)

[N m1]
4.6
(100) (1 1)
237
238
Using STM, Haiss and coworkers [168]

studied the surface stress in the presence
of anion adsorption on Au(111) as a function of the electrode potential. They found a
linear dependence of the surface stress on
the surface charge, with the slope depending on the type of electrolyte. However,
the nding does not appear to be universal since the surface stress of Au(100)
deviates signicantly from the linear behavior at negative potentials [170]. Bach
and coworkers [151] measured the change
of the surface stress in the reconstruction
of the Au(111) and the Au(100) surfaces
in 0.1 M HClO4 (Fig. 7). By controlling the
electrochemical potential of the metal electrodes, the reconstruction was reversibly
lifted. For both surfaces, the reconstruction relaxed the tensile stress by 22% and
5%, respectively, which can be understood
since the average bond distances between
the surface atoms are shorter in the reconstructed phases. In the case of the Au(111)
surface, the energy gain due to the relaxation of the surface stress as estimated
from the FrenkelKontorova model was
close but not quite large enough to make
the reconstruction energetically favored
without the formation of the secondary
herringbone structure of the solitons. For
the Au(100) surface, the energy gain was
found to be too small to cause the formation of the reconstruction.
3.3.5
Conclusion
Since the rst applications of STM and

AFM in electrochemistry more than a
decade ago, both techniques have matured
into powerful tools for imaging various
electrochemical phenomena with high resolution. In addition to structural studies,
many nonimaging applications based on
STM and AFM have been demonstrated.
One important example is scanning tunneling spectroscopy that allows extraction of electronic states of the probed
molecules. The electronic information can
be used to identify structurally similar
molecules, to obtain chemical reactivity
of the molecules, and is directly relevant to
the current effort of developing molecular
electronics. However, in order to reliably
extract the information, improvements in
both the experimental method for measuring tunneling spectroscopy and theoretical
description of the STM imaging mechanism in the electrochemical environment
are needed. Another important example is
the local force spectroscopy with AFM,
which reveals not only various surface
forces but also mechanical properties of
individual molecules. The microfabricated
cantilevers used as force sensors in AFM
have found broad applications in chemical and biological sensors, in detecting
infrared light, magnetic eld, and other
physical quantities and properties. Other
nonimaging applications of STM and AFM
in conjunction with the electrochemical
techniques include nanofabrication and
nanomanipulation, which will continue to
play important roles in the emerging eld
of nanoscience and nanotechnology.
Acknowledgments
I am grateful to Shaopeng Wang, Salah

Boussaad, Huixin He, and Yuwen Zhao
for their comments and discussions. The
lab has been supported by grants from
NSF (CHE-9818073), PRF (33516-AC5,
administered by the ACS), and AFOSR
(F49620-99-1-0112).
References
1. G. Binnig, H. Rohrer, C. Gerber et al., Phys.
Rev. Lett. 1982, 49, 5761.

2. G. Binnig, C. F. Quate, C. Gerber, Phys.
Rev. Lett. 1986, 56, 930933.
3. R. Sonnenfeld, P. K. Hansma, Science 1986,
232, 211213.
4. H. Y. Liu, F. R. F. Fan, C. W. Lin et al., J.
Am. Chem. Soc. 1986, 108, 3838, 3839.
5. S. Manne, H. J. Butt, S. A. C. Gould et al.,
Appl. Phys. Lett. 1990, 56, 1758, 1759.
6. S. Manne, P. K. Hansma, J. Massie et al.,
Science 1991, 251, 183186.
7. J. Lipkowski, P. N. Ross, (Eds.), Imaging of
Surfaces and Interfaces, Wiley-VCH, New
York, 1999.
8. A. A. Gewirth, B. K. Niece, Chem. Rev. 1997,
97, 11291162.
9. A. J. Bard, H. D. Abruna, C. E. Chidsey
et al., J. Phys. Chem. 1993, 97, 71477173.
10. T. P. Moffat, Electroanal. Chem. 1999, 21,
211316.
11. D. M. Kolb, Electrochem. Acta 2000, 45,
23872402.
12. J. E. T. Andersen, J.-D. Zhang, Q. Chi et al.,
Trend Anal. Chem. 1999, 18, 665674.
13. C. J. Chen, Introduction to Scanning Tunneling Microscopy, Oxford University Press,
New York, 1993.
14. J. H. Coombs, J. B. Pethica, IBM J. Res. Dev.
1986, 20, 455459.
15. R. Christoph, H. Siegenthaler, H. Rohrer
et al., Electrochim. Acta 1989, 34, 1011.
16. J. Wiechers, T. Twomey, D. M. Kolb et al.,
17. M. Bingelli, D. Carnal, R. Nyffenegger et al.,
J. Vac. Sci. Technol., B 1991, 9, 19851992.
18. J. Pan, T. W. Jing, S. M. Lindsay, J. Phys.
Chem. 1993, 98, 42054208.
19. A. Vaught, T. W. Jing, S. M. Lindsay, Chem.
Phys. Lett. 1995, 236, 306310.
20. Y. A. Hong, J. R. Hahn, H. Kang, J. Chem.
Phys. 1998, 108, 43674370.
21. W. Schmickler, D. Henderson, J. Electroanal. Chem. 1990, 290, 283291.
22. M. V. Vigdorovich, A. M. Kuznetsov, J. Ulstrup, Chem. Phys. 1993, 176, 539554.
23. G. Repphun, J. Halbritter, J. Vac. Sci.
Technol., A 1995, 13, 16931698.
24. W. Schmickler, Surf. Sci. 1995, 335,
416421.
25. I. Benjamin, D. Evans, A. Nitzan, J. Chem.
Phys. 1997, 106, 12911293.
26. J. Bardeen, Phys. Rev. Lett. 1960, 6, 5759.
27. J. Tersoff, D. R. Hamann, Phys. Rev. B 1985,
31, 805814.
28. S. M. Lindsay, O. F. Sankey, Y. Li et al., J.

Phys. Chem. 1990, 94, 46554660.
29. N. D. Lang, Phys. Rev. Lett. 1986, 56,
11641167.
30. V. Mujica, M. Kemp, M. A. Ratner, J. Chem.
Phys. 1994, 101, 68496864.
31. H. Ou-Yang, B. Kallebring, R. A. Marcus, J.
Chem. Phys. 1993, 98, 75657824.
32. A. J. Fisher, P. E. Blochl, Phys. Rev. Lett.
1993, 70, 32633266.
33. D. N. Futaba, S. Chiang, J. Vac. Sci. Technol., A 1997, 15, 12951298.
34. V. M. Hallmark, S. Chiang, Surf. Sci. 1995,
329, 255268.
35. X. W. Wang, N. J. Tao, F. Cunha, J. Chem.
Phys. 1996, 105, 37473752.
36. A. M. Kuznetsov, P. Sommer-Larsen, J. Ulstrup, Surf. Sci. 1992, 275, 5264.
37. J. E. T. Andersen, A. A. Kornyshev, A. M.
Kuzenetsov et al., Electrochim. Acta 1997,
42, 819831.
38. W. Schmickler, C. Widrig, J. Electroanal.
Chem. 1992, 336, 213221.
39. H. Sumi, J. Phys. Chem. B 1998, 102,
18331844.
40. A. K. Mishra, S. K. Rangarajan, Theochem:
J. Mol. Struct. 1996, 361, 101109.
41. A. M. Kuznetsov, J. Ulstrup, Mol. Phys.
1992, 87, 1189.
296, 283289.
43. M. Tortonese, R. C. Barrett, C. F. Quate,
Appl. Phys. Lett. 1993, 62, 834836.
44. S. M. Lindsay in Scanning Tunneling Microscopy and Related Techniques (Ed.:
D. Bonnell), John Wiley & Sons, New York,
2000.
45. J. N. Israelachvili, Intermolecular and Surface Forces, Academic Press, London, 1992.
46. F. Ohnesorge, G. Binnig, Science 1993, 260,
14511456.
47. F. J. Giessibl, Phys. Rev. B 1997, 56,
16 01016 015.
48. T. R. Albrecht, P. Grutter, D. Horne et al.,
J. Appl. Phys. 1991, 69, 668673.
49. F. J. Giessibl, Science 1995, 267, 6871.
50. F. J. Giessibl, Appl. Phys. Lett. 2000, 76,
14701472.
51. Q. Zhong, D. Inniss, K. Kjoller et al., Surf.
Sci. 290, L688L690.
52. P. K. Hansma, J. P. Cleveland, M. Radmacher et al., Appl. Phys. Lett. 1994,
17381741.
239
240

53. C. A. J. Putman, K. O. Vanderwerf, B. G.
Degrooth et al., Appl. Phys. Lett. 1994, 64,
24542456.
54. S. M. Lindsay, Y. I. Lyubchenko, N. J. Tao
et al., J. Vac. Sci. Technol. 1993, 11, 808815.
55. E. I. Florin, M. Radmacher, B. Fleck et al.,
Rev. Sci. Instrum. 1994, 65, 639643.
56. W. Han, S. M. Lindsay, T. Jing, Appl. Phys.
Lett. 1996, 69, 41114113.
57. A. A. Kornyshev, M. Sumetskii in Electrochemical Nanotechnology: In situ Local Probe
Techniques of Electrochemical Interfaces (Ed.:
W. J. L. A. W. Plieth), Wiley-VCH, Weinheim, Germany, 1998, pp. 4555.
58. N. J. Tao, C. Z. Li, H. X. He, J. Electroanal.
Chem. 2000, 492, 8193.
59. R. M. Nyffenegger, R. M. Penner, Chem.
Rev. 1997, 97, 11951230.
60. L. A. Nagahara, T. Thundat, S. M. Lindsay,
Appl. Phys. Lett. 1990, 57, 270272.
61. D. M. Kolb, R. Ullmann, T. Will, Science
1997, 275, 10971099.
62. C. B. Ross, L. Sun, R. M. Crooks, Langmuir
1993, 9, 632636.
63. W. J. Li, J. A. Nirtanen, R. M. Penner, Appl.
Phys. Lett. 1992, 60, 11811183.
64. R. T. Potzshke, G. Staikov, W. J. Lorentz
et al., J. Electrochem. Soc. 1999, 146,
141149.
65. U. Stimming, R. Vogel, D. M. Kolb et al., J.
Power Sources 1993, 43, 169180.
66. D. Hofmann, W. Schindler, J. Kirchner,
Appl. Phys. Lett. 1998, 73, 32793281.
67. R. Schuster, V. Kirchner, X. H. Xia et al.,
Phys. Rev. Lett. 1998, 80, 55995602.
68. C. Z. Li, N. J. Tao, Appl. Phys. Lett. 1998, 72,
894897.
69. G. Binnig, H. Fuchs, E. Stoll, Surf. Sci.
1986, 169, L295L298.
70. M. L. Lozano, M. C. Tringides, Europhys.
Lett. 1995, 30, 537542.
71. M. Sumetskii, A. A. Kornyshev, Phys. Rev.
B 1993, 48, 17 49317 506.
72. M. Sumetskij, A. A. Kornyshev, U. Stimming, Surf. Sci. 1994, 307, 2327.
73. O. M. Magnussen, M. R. Vogt, Phys. Rev.
Lett. 2000, 85, 357360.
74. M. Giesen-Seibert, R. Jentjens, M. Poensgen et al., Phys. Rev. Lett. 1993, 71,
35213524.
75. M. Giesen, R. Randler, S. Baier et al., Electrochim. Acta 1999, 45, 527536.
76. I. Giaever, Phys. Rev. Lett. 1960, 5, 147, 148.
77. R. M. Feenstra, J. A. Stroscio, A. P. Fein,

Surf. Sci. 1987, 181, 295306.
78. H. Morisaki, H. Nishkawa, H. Ono et al., J.
Electrochem. Soc. 1990, 137, 27592763.
79. H. Gerischer, Z. Phys. Chem. N. F. 1961, 27,
48.
80. E. P. Friis, Y. I. Kharkats, A. M. Kuznetsov
et al., J. Phys. Chem. A 1998, 102,
78517859.
81. A. M. Kuznetsov, J. Ulstrup, J. Phys. Chem.
A 2000, 104, 11 53111 540.
82. S. R. Snyder, H. S. White, J. Electroanal.
Chem. 1995, 394, 177185.
83. N. J. Tao, Phys. Rev. Lett. 1996, 76,
40664069.
84. W. Schmickler, N. J. Tao, Electrochim. Acta
1997, 432, 28092815.
85. W. Han, E. N. Durantini, T. A. Moore et al.,
J. Phys. Chem. 1997, 10, 10 71910 725.
86. X. Lu, K. W. Hipps, X. D. Wang et al., J.
Am. Chem. Soc. 1996, 118, 71977202.
87. U. Mazur, K. W. Hipps, J. Phys. Chem. B
1999, 103, 97219727.
88. L. Scudiero, D. E. Barlow, K. W. Hipps, J.
Phys. Chem. B 2000, 104, 11 89911 905.
89. X. Lu, K. W. Hipps, J. Phys. Chem. B 1997,
101, 53915396.
90. U. Mazur, K. W. Hipps, J. Phys. Chem. 1995,
99, 66846688.
91. J. E. T. Andersen, P. P. Moller, M. V. Pedersen et al., Surf. Sci. 1995, 325, 193205.
92. B. Zhang, E. Wang, Probe Microsc. 1997, 1,
5760.
93. J. D. Zhang, Q. Chi, S. J. Dong et al., J.
94. E. P. Friis, J. E. T. Andersen, Y. I. Kharkats
et al., Proc. Natl. Acad. Sci. USA 1999, 96,
13791384.
95. E. P. Friis, J. E. T. Andersen, P. Moller et al.,
96. J. J. Davis, C. M. Halliswell, H. O. A. Hill
et al., New J. Chem. 1998, 22, 11191123.
97. I. Lee, J. W. Lee, R. J. Warmack et al., Proc.
Natl. Acad. Sci. USA 1995, 92, 19651969.
98. J. H. Schott, H. S. White, Langmuir 1994,
10, 486491.
99. D. M. Cyr, B. Venkataraman, G. W.
Flynn et al., J. Phys. Chem. 1996, 100,
13 74813 759.
100. H. B. Fang, L. C. Giancarlo, G. W. Flynn, J.
Phys. Chem. B 1999, 103, 57125715.
59785995.

70777080.
103. C. Joachim, J. K. Gimzewski, A. Aviram,
Nature 2000, 408, 541548.
104. P. Ball, Nature 2000, 406, 118120.
105. L. A. Bumm, J. J. Arnold, M. T. Cygan et al.,
Science 1996, 271, 17051707.
106. G. S. McCarty, P. S. Weiss, Chem. Rev.
1999, 99, 19831990.
107. G. Leatherman, E. N. Durantini, D. Gust
40064010.
108. Y. Q. Xue, S. Datta, S. Hong et al., Phys.
Rev. B 1999, 59, R7852R7855.
109. D. J. Wold, C. D. Frisbie, J. Am. Chem. Soc.
2000, 122, 2970, 2971.
110. S. Datta, W. D. Tian, S. H. Hong et al.,
Phys. Rev. Lett. 1997, 79, 25302533.
111. C. Joachim, J. K. Gimzewski, Chem. Phys.
Lett. 1997, 265, 353357.
112. V. J. Langlais, R. R. Schlittler, H. Tang et al.,
Phys. Rev. Lett. 1999, 83, 28092812.
113. J. K. Gimzewski, C. Joachim, Science 1999,
283, 16831688.
114. D. I. Gittins, D. Bethell, D. J. Schiffrin et al.,
Nature 2000, 408, 6769.
115. E. Meyer, H. Heinzelmann, P. Grutter et al.,
J. Microsc. 1988, 152, 269.
116. B. Cappela, G. Dietler, Surf. Sci. Rep. 1999,
34, 1104.
117. S. Manne, H. E. Gaub, Curr. Opin. Colloid
Interface Sci. 1997, 2, 145152.
118. W. A. Ducker, T. J. Senden, R. M. Pashley,
Nature 1991, 353, 239241.
119. Y. Q. Li, N. J. Tao, J. Pan et al., Langmuir
1993, 9, 637641.
120. H. J. Dai, J. H. Hafner, A. G. Rinzler et al.,
Nature 1996, 384, 147151.
121. V. A. Parsegian, D. Gingell, Biophys. J.
1972, 12, 1192.
122. H.-J. Butt, Biophys. J. 1991, 60, 14381444.
123. X. Y. Li, F. Creuzef, H. Arribart, J. Phys.
Chem. 1993, 97, 72727276.
124. A. L. Weisenhorn, P. Maivald, H.-J. Butt
et al., Phys. Rev. B 1992, 45, 11 22611 232.
125. D. T. Atkins, R. M. Pashley, Langmuir 1993,
9, 22322236.
126. R. Raiteri, S. Martinoia, M. Grattarola, Biosens. Bioelectron. 1996, 11, 10091017.
127. S. Biggs, P. Mulvaney, J. Chem. Phys. 1994,
100, 85018505.
128. S. Biggs, P. Mulvaney, C. Zukoski et al., J.
Am. Chem. Soc. 1994, 116, 91509157.
129. V. Kane, P. Mulvaney, Langmuir 1998, 14,

33033311.
130. B. A. Hu K, Langmuir 1997, 13, 54185425.
131. C. Z. Hu K, J. K. Whitesell, A. J. Bard, Langmuir 1999, 15, 33433347.
132. S. Manne, J. P. Cleveland, H. E. Gaub et al.,
Langmuir 1994, 10, 44094413.
133. I. Larson, D. Y. C. Chan, C. J. Drummond
et al., Langmuir 1997, 13, 24292431.
134. R. Raiteri, M. Preuss, M. Grattarola et al.,
Colloids Surf., A 1998, 136, 191197.
135. R. Raiteri, M. Grattarola, H.-J. Butt, J. Phys.
Chem. 1996, 100, 16 70016 705.
136. T. Ishino, H. Hieda, K. Tanaka et al., Jpn. J.
Phys. 1994, 33, L1552L1554.
137. A. C. Hillier, S. Kim, A. J. Bard, J. Phys.
Chem. 1996, 100, 18 80818 817.
138. T. Arai, M. Fujihira, J. Vac. Sci. Technol., B
1996, 14, 13781382.
139. J. B. Pathica, W. C. Oliver, Phys. Scr. 1987,
T19, 6166.
140. J. H. Hoh, J. P. Cleveland, C. B. Prater et al.,
J. Am. Chem. Soc. 1992, 114, 4917, 4918.
141. S. J. OShea, M. E. Welland, T. Rayment,
Appl. Phys. Lett. 1992, 60, 23562358.
142. W. Han, S. M. Lindsay, Appl. Phys. Lett.
1998, 72, 16561658.
1980.
144. J. Lipkowski, L. Stolberg in Adsorption
of Molecules at Metal Electrodes (Eds.:
J. Lipkowski, P. N. Ross), VCH Publishers,
New York, 1992, pp. 171237.
145. R. E. Malpas, R. A. Fredlein, A. J. Bard, J.
146. K. F. Lin, T. R. Beck, J. Electrochem. Soc.
1976, 123, 1145.
147. R. E. Martinez, W. M. Augustyniak, J. A.
Golovchenko, Phys. Rev. Lett. 1990, 64,
10351038.
148. A. J. Schell-Sorokin, R. M. Tromp, Phys.
Rev. Lett. 1990, 64, 10391042.
149. D. Sader, H. Ibach, Phys. Rev. B 1991, 43,
42634267.
150. R. Berger, E. Delamarche, H. P. Lang et al.,
Science 1997, 276, 20212023.
151. C. E. Bach, M. Giesen, H. Ibach et al., Phys.
Rev. Lett. 1997, 78, 42254228.
152. G. G. Stoney, Proc. R. Soc. London, Ser. A
1909, 82, 172183.
153. K. Dahmen, S. Lehwald, H. Ibach, Surf. Sci.
2000, 446, 161173.
241
242

154. R. Berger, H. P. Lang, C. Gerber et al.,
Chem. Phys. Lett. 1998, 294, 363369.
155. J. R. Barnes, R. J. Stephenson, M. E. Welland et al., Nature 1994, 372, 7981.
156. R. P. Cowburn, A. M. Moulin, M. E. Welland, Appl. Phys. Lett. 1997, 71, 22022204.
157. P. I. Oden, P. G. Datskos, T. Thundat et al.,
Appl. Phys. Lett. 1996, 69, 32773279.
158. P. Kremmer, A. M. Moulin, R. J. Stephenson et al., Science 1997, 277, 17991802.
159. H. P. Lang, M. K. Baller, R. Berger et al.,
Anal. Chem. Acta 1999, 393, 5965.
160. T. Thundat, E. Finot, Z. Hu et al., Appl.
Phys. Lett. 2000, 77, 40614063.
161. H. F. Ji, E. Finot, R. Dabestani et al., Chem.
Commun. 2000, 6, 457, 458.
162. J. Fritz, M. K. Baller, H. P. Lang et al., Science 2000, 288, 316318.
163. H.-J. Butt, J. Colloid Interface Sci. 1996, 180,

251260.
164. R. Raiteri, H.-J. Butt, M. Grattarola, Electrochem. Acta 2000, 46, 157163.
165. T. A. Brunt, T. Rayment, S. J. OShea et al.,
Langmuir 1996, 12, 59425946.
166. T. A. Brunt, E. D. Chabala, T. Rayment
et al., J. Chem. Soc., Faraday Trans. 1996,
92, 38073812.
167. W. Haiss, J. K. Sass, J. Electroanal. Chem.
1995, 386, 267270.
168. W. Haiss, R. J. Nichols, J. K. Sass et al., J.
169. H. Ibach, C. E. Bach, M. Giesen et al., Surf.
Sci. 1997, 375, 107119.
170. H. Ibach, Electrochem. Acta 1999, 45,
575581.
3.4 In situ FTIR as a tool for mechanistic studies. Fundamentals and applications
3.4
In situ FTIR as a tool for mechanistic

studies. Fundamentals and applications
Francisco C. Nart and Teresa Iwasita
Universidade de Sao Paulo, Sao Paulo, Brazil
3.4.1
Historical Overview
Electrochemistry experienced in the recent

years the introduction of many in situ
techniques for the purpose of improving
the understanding of the electrode process
at a molecular level. Among these techniques, the infrared spectroscopy plays
a relevant role as a tool to study, in
situ, the electrode surface. The molecular
specicity permits the identication of adsorbates and reaction products and allows
the study of physicochemical properties of
adsorbed molecules.
The infrared spectroscopy has been used
largely in Surface Science to study adsorbed molecules on surfaces in vacuum.
The infrared beam is reected at the crystal surface and the adsorbed molecules
interact with the incident light [1, 2].
Greenler proposed the basic principles
of the reection-absorption spectroscopy
in 1966 [3]. From his work, it becomes
clear that the light interacting with the
adsorbed molecules on a metal surface is
only the parallel mode of the electric eld
of the incident radiation. This leads to the
so-called surface selection rule, a central concept in reection-absorption spectroscopy
(see Sect. 3.4.2.1.2).
In situ infrared spectroscopy has been
used in electrochemistry as early as
1970 [4]. Obviously, to apply the infrared
spectroscopy to the study of the electrode
surface, the strong absorption of light
by the solvent must be minimized. One
of the rst attempts to use infrared
spectroscopy as an in situ tool was
made using the so-called attenuated total

reection (ATR), a technique largely used in
infrared spectroscopy for liquid samples.
In this technique, the infrared beam is
introduced into a high refraction index
infrared transparent crystal and the crystal
put in contact with the sample. When
the light hits the edge of the crystal in
contact with a medium of lower refraction
index, the light is totally reected. If
there is some absorbing species in the
sample, then the refraction index will have
an imaginary component that changes
the refraction indexes at the wavelength
where absorption exists. The light is then
transmitted to the liquid sample and it is
therefore attenuated at the detector.
The mostly used crystal as an ATR element and substrate for the electrode
has been germanium [4, 5]. Germanium
crystals have good transparency and good
electrical conductivity, but are easily dissolved at positive potentials in aqueous
solutions. The electrode can be deposited
onto the crystal or, in some cases, the germanium can be used as an electrode for
some processes in which the nature of the
electrode does not affect the reaction, such
as in outer sphere reactions.
The theory of internal reection applied
to electrochemical studies has been discussed in detail by Hansen [6] for both the
infrared and the visible region of the electromagnetic spectrum. However, the ATR
for electrochemical studies has been abandoned rapidly, because of the difculties
in obtaining an electrode with good optical transparency in the infrared region and
good electrical conductivity. In general, the
ATR crystal must be a semiconductor or
insulating material. In cases in which the
electrode is deposited on the surface of
the element, the condition of optical transparency requires a very thin lm in which
the conductivity is very low. Moreover,
243
244
these very thin lms are not enough to

protect the crystal from corrosion in the
usual electrochemical solutions.
Although ATR infrared spectroscopy
for in situ studies of electrochemical
systems presents many limitations, one
interesting property of this technique is the
amplication of the IR signal by excitation
of surface plasmons. Osawa and coworkers
used this technique to study the adsorption
of water [7] and other molecules [8] on
thin-lm gold electrodes. However, the
amplication of the IR signal through
plasmon excitation is limited to metals
such as gold, silver, and other sp metals.
Another important application of the
ATR infrared spectroscopy is for the
study of conducting polymers [9]. Most of,
these studies are conducted in nonaqueous
media in which the corrosion of the ATR
element is not important.
Though the internal reection mode
(ATR) encounters many applications, the
external reection mode is much more
versatile to study metallic electrode surfaces. The external reection mode has
been introduced to study metallic electrode
surfaces by Bewick and coworkers [10]. To
minimize the strong attenuation of the
radiation by the presence of a solvent in
electrochemical systems, the electrode was
placed very close to the infrared transparent window. Practically, the electrode is
slightly pressed against the at window
surface leaving a very thin electrolyte layer
between the electrode and the window.
At the beginning, a dispersive instrument was used. The signal-to-noise ratio
was very poor, since the amount of species
sampled was very low and the dispersive
instrument has a low throughput. The low
signal-to-noise ratio was then improved
by applying to the electrode, a potential
modulation of low frequency (typically between 7 and 10 Hz) and using a phase
detector (lock-in amplier). This technique

has been called Electrochemically Modulated
Infrared Spectroscopy (EMIRS) [11]. The potential modulation, which is a square wave,
and the spectra were collected at the two
potentials of the step modulation. The
spectra of each potential are added and
then ratioed to obtain the R/R.
EMIRS has been successfully applied to
many systems. Briey it can be mentioned
the study of adsorbates at the electrode
surface [10], the detection of adsorbed reaction intermediates for the oxidation of
small organic molecules [12], and the determination of the water structure in the
double layer [13]. However, the potential
modulation in EMIRS is its drawback,
since it prevents the study of irreversible
processes as the system must return to the
same conditions each time the potential
is changed. Other important limitations of
EMIRS are related to both the electrical and
chemical relaxation effects caused by the
potential modulation at 12 Hz. The electrical relaxation is due to the high ohmic
drop of the electrolyte conned in the thin
solution layer required for the in situ measurements. The chemical relaxation is due
to ion migration induced by the change in
solution composition caused by the electrode potential change. These aspects have
been discussed in detail in the following
text [1416] (see Sect. 3.4.2.3).
The introduction of Fourier Transformed Infrared Spectrometers to the
electrochemical experiments eliminated
the need of potential modulation to increase the signal-to-noise ratio. Fourier
Transform Spectrometers present two advantages compared to the dispersive equipment [17], which have improved the possibilities of the in situ measurements. One
advantage originates from the fact that
the beam hits the sample without passing through a monochromator (Jacquinot
advantage), thus avoiding the losses caused

by the monochromator components like
slit and gratings. The net result is a better
signal-to-noise ratio, eliminating the need
of a phase detector. The other advantage
is known as multiplex (or Felget) advantage, in which all the wave numbers can
be sampled simultaneously, increasing the
speed of spectral acquisition. However, the
spectral acquisition at different potentials
is necessary, since the Fourier transform
equipments are single-beam devices and a
reference spectrum is always needed. The
use of Fourier Transform IR spectroscopy
in in situ studies was introduced by Pons
and coworkers [18, 19]. The technique was
called Subtractively Normalized Interfacial
Fourier Transform-Infrared Spectroscopy
(SNIFTIRS) due to the way the spectra
were acquired.
The use of Fourier Transform instruments eliminates much of the limitations
of the EMIRS, since no more potential
modulation is needed. The signal-to-noise
ratio is far less than the dispersive instrument and can be improved statistically by
adding more scans, since the spectral acquisition time is much lower. With the
Fourier Transform equipment, also the irreversible processes can be studied, since
it is no longer required to return to the
same potentials as for modulation (see
Sect. 3.4.4). This not only allows the acquisition of derivative or bipolar bands but
also the acquisition of integral bands, as in
the case of the adsorbed CO on platinum
electrodes [2025], which was impossible
with EMIRS. The speed of spectral acquisition of Fourier Transform Infrared (FTIR)
instruments allows also the follow-up of
a reaction during a dynamic polarization
curve [26, 27].
To avoid the spectral acquisition at two
different potentials in order to have a reference spectrum, polarization modulation
has been used associated to the FTIRS.

In theory this technique allows the acquisition of only integral bands, since this
is based on the surface selection rule
(see below Sect. 3.4.2.1.2), in which the spolarized light is only active away from the
electrode surface. This technique has been
called Fourier transform-infrared reectionabsorption spectroscopy FT-IRRAS [28], although this acronym has been used also
for the integral bands obtained without polarization modulation. The advantages and
limitations of this technique have been
analyzed in detail by Seki et al. [29]. The
main limitation is that the reectivity is
different for the p- and s-polarized radiation, and it changes with the radiation
energy, the incident angle, and thickness
of the thin electrolyte layer. Thus the
presence of water and ions in the solution can cause absorption bands even
for species not adsorbed on the electrode.
Therefore, compensating mechanisms are
needed and even so, it is very difcult to
prevent artifacts. In the case in which the
bands of the adsorbed species are very
close to the bands in solution, it is possible to mix the signals coming from each
one [16, 30]. The polarization modulation
has been used to study adsorbed CO [28]
and CN [28, 29], since the bands of these
species are far away from the bands of the
water and ions used for the electrochemical
experiments.
3.4.2
The External Reectance Spectroscopy

Fundamentals
Let us consider the reection on a at
mirror-nished surface as depicted in
Fig. 1. The complex electric eld of the
standing wave above the surface is the result of the vectorial sum of the incident
and the reected beam [1]. The magnitude
3.4.2.1
245
Reection of the s-() and p-() polarized light on a at mirror surface. Note the
direction of the electric eld upon reection.
Fig. 1
of the eld is a function of the incidence

angle and of the light polarization. This
is a consequence of the phase shift upon
reection. The phase shift for the electric eld perpendicular to the incident
plane (s-polarized radiation) is almost 180
for all the incident angles. For the electric eld parallel to the incident plane
(p-polarization), the phase shift is angle dependent and changes drastically between
60 and 90 , as can be seen in Fig. 2.
These results have two practical consequences for the reection-absorption
spectroscopy: (1) the dependence of the
band intensity on the incident angle and
(2) the introduction of the surface selection

rule.
3.4.2.1.1 The Band Intensity The absorption of infrared radiation by a submonolayer of an adsorbed species is the result
of the interaction of the electric eld of
the radiation with the electric dipoles of
the adsorbed species and with the electrons of the reecting metallic surface.
Therefore, the intensity of the absorption
is determined by the electric properties of
the metal. The absorption is the normalized difference between the reectivity
of the surface free from the absorbing
Angle
0
Phase shift
246
45
90
180
Phase shift for the reected s-( ) and p-( ) polarized light upon
reection on a metallic surface. (Data from Ref. [3] with permission.)
Fig. 2
90
species (R0 ) and the surface covered with

an absorbing layer (R):
=
R R0
R0
(1)
The intensity of a band in the spectrum

is the result of the dissipation of the
electromagnetic radiation in the absorbing
media at a given energy. Under these
circumstances, the change in the energy
ux is proportional to the mean quadratic
eld (see Appendix A). Therefore, it is
important to understand the behavior of
the electric eld on a metallic surface
to rationalize the factors affecting the
intensity of light absorption.
The mean quadratic eld, referred to the
incident mean quadratic eld in the same
direction for the s-polarized radiation, is [6]
2
ES1
02
ES1

1/2
= (1 + RS ) + 2RS

z
cos Sr + 4
1 (2)
where RS is the s-polarized light reectivity, Sr the phase angle of the reected
radiation, is the wavelength of the radiation, z is the position above the reecting
surface, and = n cos , where n is the
refraction index and the incidence angle.
On the surface plane (z = 0), the electric
eld is given by [6]
2
ES1
02
ES1

1/2
= (1 + RS ) + 2RS cos Sr (3)
since for a reection on a metallic surface,

the phase shift is almost , cos Sr 1
and the mean quadratic eld for this case is
2
ES1
02
ES1

1/2 2
= 1 RS
(4)
Considering that for a metallic surface

RS 1, the electric eld at the surface
is practically zero.
For the p-polarized light, the mean
quadratic eld is given by [6]
2
Ep1
02
Ep1

1/2
= cos2 1 (1 + Rp ) 2Rp

z
cos pr + 4
1

1/2
+ sin2 1 (1 + Rp ) + 2Rp

z
cos pr + 4
(5)
1
The change in the intensity of the electric

eld for the p-polarized radiation is shown
in Fig. 3. The intensity of the eld
increases for higher incidence angle up
to ca. 90 . The best angle is around 85 ,
known as the Brewster angle.
From this discussion, two important
consequences can be derived: (1) the spolarized light does not interact with the
species very close to the metallic surface
(see Sect. 3.4.2.1.2). The immediate consequence is that adsorbed species cannot
be sampled with s-polarized light but only
with p-polarized light. (2) The intensity for
the absorption of radiation is angle dependent and maximum for large angles in the
case of metal/air or metal/vacuum surface.
For electrochemical applications, a very
thin liquid electrolyte on the surface is
needed, and for this a transparent window
must be used to keep the electrolyte lm
in front of the surface. This makes the
whole situation more complicated. The
introduction of a very thin liquid lm
along with the transparent window (see
below) changes the optimum incidence
angle, as can be seen in Fig. 4, in which it is
shown how the different media change the
direction of the beam. It is worth noting the
247
2.0
E z20
n3 = 3.0
k3 = 30.0
n1 = 1.0
E z2
248
1.0
45
90
Angle
Change of the mean square electric eld for the air/metallic surface with
the incidence angle. The simulation parameters are shown in the gure.
Fig. 3
qualitative differences if a at or prismatic

window is used.
The equations for a system consisting of metallic surface/thin-lm electrolyte/prismatic window were developed
by Seki et al. [29]. Briey, the equations are
2
ES2
02
ES2
+ n1 sin 1
+ in1 sin 1 cos 1
02
EP2
(6)
z

C = cos 22

z
S = sin 22
(8)
(9)
and
= cos 1 (1 rP )C

+ i2
2
(1 rP )
S
2
(7)
and
2
EP2

n1 (1 rP )
S
n22

where
= [(1 + rS )C + i(1 rS )S]2
2
n1 (1 + rP )
S
n22
rS =
rS12 + rS23 e2i

1 + rS12 rS23 e2i
(10)
rP =
rP12 + rP23 e2i

1 + rP12 rP23 e2i
(11)
Electrode
Absorbate
Electrolyte
Window
q2
Air
q1
(a)
Electrode
Absorbate
Electrolyte
Prismatic window
Air
q
Air
(b)
The reection of the light beam for an electrochemical interface, including

the window [(a) at or (b) prismatic], showing the different internal reections of
the beam.
Fig. 4
and r12 and r23 are the Fresnel coefcients for the window/electrolyte and
electrolyte/metal interfaces, respectively.
is the energy attenuation in the solution
phase and is given by

d
(12)
2
= 2
where d is the thickness of the thin

electrolyte lm.
So far the intensity of the electric eld which will be related to the
dispersion of the energy was described

in detail, but not the band intensity itself.
The band intensity obviously depends on
the magnitude of the electric eld (see
Appendix A), but the intensity of light absorption depends on the dynamic electric
dipole moment of the species interacting
with the electric eld. For the band intensity in reection-absorption spectra, one
elegant and relatively simple equation has
been presented by Persson [31]. According to this model, the adsorbed adlayer
249
is constituted by an ensemble of electric

dipoles and the integrated band intensity
is given by

16 2 N 2
G() (13)
d() =
hc A
where N/A is the density of adsorbed
particles, is the dynamic dipole moment
operator for the dipole perpendicular to the
surface, is the vibrational frequency, and
is the frequency of the incident radiation.
G() is a function of the incident angle and
of the dielectric constant of the metal and
is dened [31] as

sin2
cos
G() =
(14)
cos cos 1
Equation (13) is useful in understanding the different components affecting
the intensity in the reection-absorption

infrared spectroscopy. More specically,
both the dynamic dipole moment, which
will depend on a specic vibration, combines with the surface properties through
the G() function. Therefore, quantitative
studies from vibrational intensities must
take into account the incidence angle at
the electrode surface.
Surface Selection Rule The
most important consequence of the behavior of the electric eld upon reection on a
metal surface is the limitation of the dipole
that can interact with the electric eld at the
surface. The phase shift of the electric eld
upon reection for the s-polarized light on
a metal surface leads to a vanishing electric
eld and therefore there is no interaction
with the dipoles of the adsorbed molecules.
3.4.2.1.2
2.0
E 20
n3 = 3.0
k3 = 30.0
n1 = 1.0
E 2
250
1.0
5.0
10.0
Z
[m]
Change of the intensity of the reected mean square electric eld
with the distance from the reecting surface for the s- and
p-polarized radiation.
Fig. 5
The s-polarized radiation becomes active

only for distances of ca. 5 nm above the
surface. On the other hand, the electric
eld for the p-polarized light is a maximum at the surface and decreases when
going away from the surface (see Fig. 5).
The electric eld for the p-polarized light
is almost the same as the s-polarized light
at ca. 2.5 m for the metal/air interface
for 60 of incidence angle at 2000 cm1 .
Therefore, the p-polarized light interacts
with the dipoles at the surface and away
from it. Since in order to interact with
the electric eld the dipole must present
the same direction, it turns out that only
the species with a nonzero dipole moment
perpendicular to the surface will interact
with the infrared beam. This is the surface
selection rule and is an important tool in
reection-absorption spectroscopy applied
to electrochemical systems in order to distinguish adsorbed species from species in
solution and to assign specic adsorption
geometry for the adsorbed species.
In Fig. 6, the s- and p-polarized spectra for adsorbed sulfate species on
Au(111) [32] are presented to illustrate the
use of the surface selection rule to differentiate adsorbed species from species in
solution. In these spectra, the up band at
1100 cm1 is visible in both spectra and
this band is assigned to the dissolved sulfate. The down bands are seen only in the
p-polarized spectrum, since the s-polarized
light cannot sample the surface.
Another application of the surface selection rule is related to the fact that the
p-polarized light has a specic direction
on the surface, that is, the electric eld of
the p-polarized light will be perpendicular
to the surface. Therefore, the assignment
of the bonding of adsorbed species using the surface selection rule demands the
information on the directions of the different dipole displacements of the adsorbed
molecule. This will tell which dipoles will

be active for a given adsorption geometry thus affording for the ascertion of
the bonding of the adsorbed species to
the surface. Obviously very simple linear
molecules, like CO and CN will have
practically only two bonding possibilities
and the surface selection rule is not useful
to determine the kind of bonding. However, for more complex species, the surface
selection rule is of great help.
One interesting example is the application of the surface selection rule to assign
the kind of bonding to the surface of the
nitrate ion (NO3 ). This ion when isolated
presents a D3h symmetry (see Fig. 7).
In principle, the NO3 ions could
adsorb with the plane of the molecule
either parallel or perpendicular to the
surface. For a parallel adsorption, the
surface selection rule will allow only
the out-of-plane bending modes, since
the stretching modes present always the
displacement parallel to the molecular
plane. The perpendicular adsorption has
two possibilities, that is, the nitrate
ions can be adsorbed by coordination of
either one or two oxygen to the surface.
Experimentally, the stretching mode of the
adsorbed nitrate ions [33, 34] is detected.
Therefore it can be concluded that the
ion is adsorbed with the molecular plane
perpendicular to the surface, giving rise to
a C2v symmetry (see Fig. 8). The breakup
of symmetry occurs because the oxygen
in contact with the surface is no longer
equivalent to the noncoordinating oxygen.
The stretching vibrational modes of the
adsorbed NO3 are presented in Fig. 8,
in which the symmetry and direction
of displacement of each mode are indicated according to the character table
for the C2v point group (see Table 1).
The symbols in the character table have
the usual meaning [35]. Briey, C is the
251
252
s-pol. light
1100 cm1
R/R0 = 0.05%
p-pol. light
950 cm1
1196
1500
1400
1300
1200
1100
1000
900
800
Wave number
[cm1]
s- and p-polarized light spectra for the adsorbed sulfate species on
Au(111) in 0.69 mol L1 HF, 0.5 mol L1 KF, and 102 mol L1 K2 SO4
solution. Ref. Pd/H2 . The s-polarized spectrum is extracted. (From Ref. [32]
with permission.)
Fig. 6
symmetry axis, the subscript 2 stands

for an order two symmetry, and v is
for vertical. is the reection symmetry plane, I is the identity operation, T
is the translator vector of the molecule
in the indicated direction (x, y of z), and
R is the rotation vector of the molecule
in the indicated direction and stands for

the infrared activity. ij are the components of the polarizability and accounts
for the Raman activity. A and B are the
nondegenerated species (one-dimensional
representation). A represents a symmetric species with respect to the rotation
C3
sv
C2
sv
sv
sh
C2
C2
Fig. 7
Sketch of the D3h symmetry elements for the free nitrate ion.
Proposed NO
vibrations for the different
modes of the adsorbed NO3
ions under a C2v symmetry.
Fig. 8
N
O
O
A1
A1
O
N
O
O
B1
about the symmetry axis (represented by

the character +1). The meaning of A is
that upon the rotation operation the vibrational mode remains unchanged. B
represents the antisymmetric species with

respect to the rotation about the symmetry axis (represented by the character
1), meaning that upon the symmetry
253
254

Tab. 1
Character table for the C2v point group
C2v
C2 (z)
(xz)
(yz)
A1
A2
B1
B2
+1
+1
+1
+1
+1
+1
1
1
+1
1
+1
1
+1
1
1
+1
operation the vibrational mode has the

direction inverted.
The character table allows identifying
the active infrared modes. These modes
are those with nonzero translation represented by T in the character table. For the
C2v point group, the A and B modes are
active in the infrared spectroscopy. Adding
to this analysis the surface selection rule,
only those modes with translation in the direction perpendicular to the surface will be
active. Considering that the surface is included in the xy plane, only the A1 modes
can be active, since only the A1 have translation in the z direction. The A1 mode is
the symmetric species with respect to the
rotation about the symmetry axis, as described above. Therefore, for the adsorbed
NO3 ion, only the symmetric modes will
be active in the infrared spectroscopy.
A detailed assignment of vibration
to a given molecule requires careful
calculations, but for practical purposes, in
the case of adsorbed species, it is possible
to use the character table to assign the
vibrations observed, once the vibrational
modes of the free molecule (or ion) are
known. In Fig. 8, the possible allowed
vibrations are proposed for the adsorbed
NO3 ion.
The in situ reection-absorption spectrum of nitrate ions on a Au(100) electrode
is shown in Fig. 9, in which two bands at
1436 and 1024 cm1 can be observed. The
1436 cm1 band has been assigned to the
A1 symmetric mode for the nitrate ions
Tz
Rz
Tx , Ry
Ty , Rx
xx , yy , zz
xy
xz
yz
adsorbed forming a C2v symmetry and the

1024 cm1 is the other A1 mode, corresponding to the activation of the totally
symmetric NO3 stretching vibration [34].
This 1436 cm1 mode would be the 3
stretching vibration and the 1024 cm1
would be the 1 stretching vibration depicted in Fig. 8.
In the case of in situ infrared in electrochemical systems, strong limitations
in the spectral range, due to the use of
chemically inert windows, impede the observation of modes usually in the range
of stretching vibrations for the inorganic
molecules. Organic molecules usually display also bending modes in the same
spectral range, but the kind of analysis
is practically the same.
Experimental Setup
The experimental setup for in situ infrared
for electrochemical studies requires the
construction of an electrochemical cell
equipped with a window transparent to the
infrared radiation and stable in aqueous
electrolytes as used for the electrochemical
studies. The window can be either at
or prismatic and is usually placed at the
bottom of the cell. The prismatic window
allows a more favorable angle of incidence.
The most commonly used window
materials are CaF2 (90077 000 cm1 )
and ZnSe (50020 000 cm1 ). The BaF2
(77066 000 cm1 ) can be used only
in some solutions, since barium forms
3.4.2.2
1.2 V
1.1 V
1V
1024
R/R0 = 0.025%
1436
1600
1400
1200
1000
Wave number
[cm1]
In situ FTIR p-polarized light spectra of adsorbed NO3 ions on
Au(100) in 1 mol L1 and 0.05 mol L1 HNO3 . Ref. Pd/H2 . (From Ref. [33]
with permission.)
Fig. 9
insoluble salts with sulfate, for example,

and interferes with the spectral data. With
prismatic windows with angles of ca. 65 ,
it is possible to obtain an incidence angle
up to 77 . For a at window, even for an
incidence angle of 60 at the window/air
interface, the maximum angle of incidence
attainable is of ca. 47 [36].
A sketch of a typical electrochemical
cell for in situ infrared spectroscopic
measurements is shown in Fig. 10. In

the upper part is the entrance for the
electrodes and gases. This kind of cell is
to be placed vertically, which makes the
optical setup slightly more complicated,
but with clear advantages over the cell
placed horizontally, since in the vertical
cell solution leaks are rare, while in the
horizontal cell the leaks are more common.
Placing the cell vertically allows also
255
256
Screw for
fixation of the electrode
Connection
Guide for the electrode
PTFE support
Gas entrance
Position for voltammetry
Reference electrode
Auxiliary electrode
Position for the in situ
FTIR
Window
support
PTFE ring
Window
Sketch of an electrochemical cell used for in situ measurements specially

designed to operate vertically.
Fig. 10
the study of organic reactions, in which

replacement of solution is needed in some
cases. Other cells with distinct design can
be seen in the literature [3537].
The cell contains the three electrodes
as any usual electrochemical cell for the
control of potential. The working electrode
is a reecting surface and for this it must
be polished to a mirror nish.
An important detail is the difference
between the reectivity using a at window or a prismatic window for the in
situ spectra. Simulating a change in absorbance in the thin-lm layer, that is, a
change in the solution composition, data
of reectivity are shown in Fig. 11(a, b)
for the setup using a prismatic and at

window, respectively [38]. The simulation
shows a dramatic decrease in the reectivity when a at window is used. The
differences in s- and p-reectance are not
as big as in the case of a prismatic window.
Other important effects are the interference fringes for the at window, which
are more pronounced for the s-polarized
light. The fringes are caused mainly by the
at window and not by the thin electrolyte
layer, since the wavelength of the infrared
radiation is larger than the thickness of the
thin electrolyte lm. Finally, it is important
to stress that the reectance is almost independent of the incidence angle up to 40 .
There is a maximum for 70 of incidence

angle at the window. When a prismatic
window is used, the radiation reectivity
is strongly dependent on the incidence
angle between 40 and 90 and there is a

pronounced maximum around 70 .
The consequence of this behavior is that
when a prismatic window is used, the
1.02
1.00
0.98
R/R
0.96
0.94
0.92
0.90
0.88
0
(a)
20
40
60
80
Angle
[degree]
Simulated change in reectivity for the system: (a) at window/electrolyte/metal for a

change in the absorption in the electrolyte layer (note the interference fringes caused by the
internal reections) and (b) prismatic window/electrolyte/metal. The parameters used in the
simulation are: n1 = 1.0 (air); n2 = 1.4 (CaF2 window); n3 = 1.32 + i0.015 (electrolyte); and
n4 = 10.9 + 15.6 (Pt). Electrolyte lm thickness: 2.0 106 m; at window thickness:
2.0 103 m. The change in absorbance in the electrolyte has been simulated by changing k
by 0.01.
Fig. 11
257
1.0
0.9
0.8
R /R 0
258
0.7
0.6
0.5
0.4
0
40
60
80
Angle
[degree]
(b)
Fig. 11
20
(Continued)
large difference in reectivity between sand p-polarized light makes it difcult to

distinguish between adsorbed species and
species in solution, since the s-polarized
signal will be much lower than the ppolarized signal. In summary, even if
the s-polarized light spectral signal is too
weak, this does not mean that the spectral
feature is only due to adsorbed species.
On the other hand, the use of a prismatic
window will maximize the spectral signal
from the surface and it will be of great
advantage for the in situ studies, since

the overall signal-to-noise ratio will be very
high compared with that obtained with a
at window.
Transport Effects in the Thin
Electrolyte Film
In one typical in situ FTIR experiment, it is
necessary to measure a reference spectrum
at a potential in which the electrochemical
process does not take place and to measure
a sample spectrum in which the desired
3.4.2.3
process takes place. The two spectra are

ratioed to obtain the nal spectrum as
in Eq. (1). This kind of experiment was
originally called Subtractively Normalized
Interfacial Fourier Transform-Infrared Spectroscopy (SNIFTIRS).
In the process, of changing the potential
to acquire the two spectra, the components
of the solution contained in the thin
electrolyte lm between the electrode and
the window can diffuse either to or from
the thin layer. The diffusion inside the
thin electrolyte layer is relatively slow [5].
Therefore the compensation of species that
are consumed by reaction or adsorption
is also slow. This is the reason that
in SNIFTIRS spectra usually there are
bands in both up and down direction
with respect to the baseline. Taking R0
as the reference spectrum and R as the
sample spectrum, down bands (R0 > R)
correspond to the formation of species and
up bands (R0 < R) to the consumption of
new species (see Eq. 1).
These effects are particularly dramatic
for platinum electrodes. Usually, a potential of 0.05 V in acidic solutions is used
as reference both in studies of ion adsorption and organic molecules reaction.
At this potential, the platinum electrode
is covered with a monolayer of hydrogen. Changing the potential to higher
values will cause hydrogen desorption leaving H+ ions in the thin layer causing
pH changes.
Different groups [1416] have addressed
this problem in detail. It has been shown
that two different kinds of problems can
occur. Both double-layer charging and
consumption of species can cause migration or diffusion of species into or from
the thin solution layer, or pH changes
can shift acid-base equilibria that can be
sampled by the in situ FTIR spectra.
In both cases, these effects can mask
the observation of the electrochemical

process in study. The pH change effect
was observed incidentally by Christensen
et al. [39] during CO2 reduction in a phosphate buffer solution. The phosphate ions
equilibria are very well-known and are easily observed in the in situ FTIR spectra.
Corrigan and Weaver [15] demonstrated
that the double-layer charging during the
potential step used is already enough
to produce ion migration inside the solution lm as shown for the study of
N3 ions adsorption on a polycrystalline
gold electrode.
One example of ion equilibrium dislocation in a phosphate buffer solution
caused by pH changes [40] is shown in
Fig. 12. The spectra were obtained using a potential step between 0.25 V
versus a reversible hydrogen electrode
(RHE) and the potentials indicated in
the gure. The spectra were taken with
s- and p-polarized radiation to conrm
that the changes in the spectra occur
in the solution and not at the electrode surface.
The up bands are then those corresponding to the consumption of species
after the potential step and the down
bands are those corresponding to the
species formed after the potential step.
The bands are the 1162 and 1075 cm1 ,
corresponding to the antisymmetric and
symmetric modes of the PO2 group, and
the band at 942 cm1 , corresponding to
the symmetric stretching of the P(OH)2
group of the H2 PO4 . On the other hand,
the up bands at 1091 and 991 cm1 are
assigned to the doubly degenerated stretching mode of the PO3 and the POH
stretching vibration, respectively. Actually
the HPO4 2 species present the doubly degenerated mode at 1072 cm1 , but
the down band at 1075 cm1 are overlapped and disturb the band at 1072
259

s-pol. light
p-pol. light
0
50
400
600
600
R/R0 0.2%
E
[mV]
400
E
[mV]
260
R/R0 0.5%
1000
1000
1200
1100
1000
1200
1100
1000
Wave number
[cm1]
Fig. 12
In situ FTIR spectra of the changes in

phosphate protonation in solution for different
applied potential step between 0.25 V and the
potentials indicated on a platinum electrode in a
phosphate buffer pH = 7.4; ionic strength 0.2.
Up bands at 1162 cm1 , 1018 cm1 , and

942 cm1 are assigned to H2 PO4 and the
down bands at 1091 cm1 and 991 cm1 to
HPO4 2 . (Data reproduced from Ref. [41] with
permission.)
giving a peak at 1091 cm1 . In summary,

from the spectral signals observed in the
spectrum of Fig. 12, a phosphate buffer
equilibrium has been sampled showing
the reaction:
these effects, an appropriate supporting

electrolyte must be used to compensate
the acid-base equilibrium changes. Other
more complicated approaches, like a ow
cell with an electrode with a small hole in
the center to allow electrolyte ow, have
been proposed [37]. The other version is a
window with a hole that allows electrolyte
ow [39]. However, the operation of such
electrodes or windows is not easy and
the uctuations in the electrode-window
distance preclude the compensation of
the full spectrum of the optical path. So
far, the use of a supporting electrolyte
permitted the distinction of the solution
HPO4 2 + H3 O+ H2 PO4 + H2 O
(15)
This is an example of how pH changes
caused by H+ desorption during the
potential step used for the measurement
cause equilibrium changes in solution,
and any possible band due to adsorbed
phosphate is completely buried by the
solution spectral signals. To minimize
bands from the bands arising from the

adsorbates [41].
3.4.3
Infrared Spectroscopy under Equilibrium

Conditions. Study of Adsorbed Sulfate Ions
on Single-crystal Electrode Surfaces
In this section, some selected applications

of in situ infrared spectroscopy to study
the reversible adsorption of anions on the
electrode surface will be discussed. The
relevant contribution of the in situ FTIR
spectroscopy to this eld was to permit
the study of the structure of the adsorbed
anions in the double-layer of the electrochemical interface for solid electrodes
without a very well-dened double-layer
region, as in the case of platinum electrodes. Moreover, the identication of the
nature of adsorbed species using in situ infrared spectroscopy has been the object of
intense debate and relevant contributions
have been given.
The Adsorption of Anions (The
Electrical Double Layer)
The use of in situ FTIR spectroscopy to
study ions adsorption is related to the need
of experimental tools to access the doublelayer structure with molecular specicity.
This is especially true for the solid metals,
where the study of the double layer using
the classical capacitance studies is not easy.
Spectroscopy combined with in situ scanning tunneling microscopy (STM) [42, 43]
constitutes a very powerful approach, allowing the determination of the molecular
identity and organization of the doublelayer components. However, it is not the
purpose of this chapter to review in detail the double-layer structure, but rather,
to show how in situ infrared spectroscopy
can contribute to the understanding of the
double layer at molecular specicity.
3.4.3.1
The simplest anion studied is the

CN . The most recent studies on CN
adsorption have been done by Tadjeddine
and coworkers [44] using sum frequency
generation, SFG a spectroscopic technique
based on nonlinear optics. This technique
is surface specic, since it is based on the
property that when two laser beams meet
at a nonsymmetric place, the frequency
of the two beams is summed. Then,
with a laser in the visible region and
an infrared tunable laser, the vibrational
bands can be sampled. According to
the SFG measurements for the CN
adsorption, two bands are observed and
have been assigned to the C- and N-down
cyanide ion adsorbed on the gold surface.
Although many interesting studies on the
CN and other anions adsorption can be
found, in this chapter only the adsorption
of sulfate ions will be addressed in detail
to show the potentialities of in situ FTIR
in the study of electrochemical systems.
The importance of sulfate ions is related
to the fact that it is a largely used
anion in supporting electrolytes both for
fundamental studies and applications.
The Adsorption of Sulfate on
M(111) Single-crystal Electrodes One common feature of adsorbed sulfate ions on
the M(111) faces of different fcc metals is
that the vibrational features observed in
the in situ FTIR spectra do not depend on
the degree of dissociation of the species in
solution. When sulfate ions are adsorbed
from strongly acid solutions containing
mainly bisulfate ions or from mildly acid
solutions containing mainly sulfate ions,
the vibrational feature observed is basically
the same (see Fig. 13) [42, 4547]. In the
spectra of Fig. 13, it is possible to see the
up band at 1100 cm1 for the solution with
pH = 2.8. This band is the F mode of the
free sulfate ions in solution. The absence of
3.4.3.1.1
261
1150
1050
Wave number
[cm1]
1250
1050
R/R0 = 0.1%
1100
(i)
(ii)
(b)
950 1300
1100
1100
(i)
952
(ii)
1000
(c)
1300
(ii)
960
(i)
1100
1000
R/R0 = 0.05%
1150
1100
Wave number
[cm1]
1200
1247
1240
Rh(111)
900 1400
R/R0 = 0.01%
1050
Wave number
[cm1]
1200
1192
1185
Au(111)
Fig. 13
In situ FTIR spectra of adsorbed sulfate species on (a) Pt(111). (Data extracted from Ref. [45] with permission.) (b) Au(111). (Data extracted
from Ref. [32] with permission.) (c) Rh(111) in (i) 0.69 mol L1 HF, 0.5 mol L1 KF, and 102 mol L1 K2 SO4 and (ii) 7.3 mol L1 HF and 102 mol L1
K2 SO4 . (Data extracted from Ref. [46] with permission.)
(a)
1350
1272
1254
Pt(111)
262
up bands at 950 cm1 reveals that sulfate

ions are unperturbed when dissolved, retaining the Td symmetry. The up band,
as discussed above (see Sect. 3.4.2.3), is
due to the consumption of sulfate ions
from the thin electrolyte lm upon adsorption. For the same spectrum (pH = 2.8), a
down band can be observed between 1200
and 1250 cm1 for saturation coverage for
the three metals, corresponding to the adsorbed ions. This band was not observed
with s-polarized light, which conrms it
is due to the adsorbed sulfate species (see
Fig. 6 for the Au(111) case). Comparing the
spectra obtained from the different solutions, it is clear that this feature is the same,
independent of the species present in solution, indicating that either bisulfate is dissociated when adsorbed, or sulfate is protonated. Similar results are observed by other
authors for Au(111) electrodes even for
neutral SO4 2 solutions [42] (see Fig. 14).

On the Au(111) electrode, a weak band
at 958 cm1 , corresponding to the totally
symmetric vibration has been observed
(see Figs. 13(b) and 14). The activation of
this band reveals the breakup of symmetry
caused by the anion adsorption. This band
has been assigned either to the activation
of the totally symmetric mode of the SO4 2
ion due to the breakup of symmetry or to
the SOH stretching vibration of adsorbed
bisulfate [48]. Measuring the vibrational
features of the adsorbed ion from a deuterated solution (D2 SO4 in D2 O), Shingaya
and Ito observed two bands, depending
on the applied potential. For potentials
between 0.7 and 0.8 V, only one band at
1153 cm1 is found. At 0.9 V, this band
shifts to 1170 cm1 and a new band at ca.
1200 cm1 appears with a very weak intensity. The 1170 cm1 band looses intensity
1193
4 104 a.u.
958
1187
1103
In situ FTIR spectra of

adsorbed sulfate species on
Au(111) in 0.5 mol L1 H2 SO4
(lower) and 0.1 mol L1
Na2 SO4 (upper). (Reproduced
from Ref. [42] with permission.)
Fig. 14
900
1200
n
[cm1]
263
for more positive potentials and the band

at 1200 cm1 starts to increase intensity
and shifts strongly with the potential up to
1230 cm1 for 1.5 V.
An important observation is the intensity
decrease of the low-energy band followed
by the increase of the higher-energy band.
There is also a concurrent change in
the intensity of the 950 cm1 band. The
authors [48] attributed this behavior to a
mixture of sulfate and bisulfate on the
surface in a given range of potential.
The band at 1153 cm1 was assigned to
the asymmetric stretching of the sulfate
ions, while the 1220 cm1 band is due
to the bisulfate. The 950 cm1 band was
assigned to the SOD stretching. Probably
the 1153 cm1 band is not due to an
asymmetric mode, since this would have

a dipole parallel to the surface, but when
sulfate is adsorbed a C3v symmetry could
produce a second symmetric mode that
can account for this band. The main
argument used is that calculations (no
detail has been given) show that the
SO3 symmetric mode couples with the
SOH bending, but not with the SOD
and this would be the reason for the
band splits at a certain potential. The
main problem in this interpretation is
the 956 cm1 band (compare Figs. 14
and 15) that has been observed also for
nondeuterated solutions. It is therefore
difcult to explain the absence of isotopic
shift upon changing H for D in the
SOH stretching.
E
vs SHE
[mV]
1500
1234
0.5 M D2SO4 in D2O
1300
0.5 M H2SO4 in H2O
956
1100
R/R
1250
900
800
1170
700
957
Wave numbers
[cm1]
264
1225
1200
1175
0.001
1153
1400
(a)
1300
1200
1100
Wave numbers
[cm1]
1000
1150
600 800 1000 1200 1400 1600
900
(b)
E vs SHE
[mV]
(a) In situ FTIR spectra of adsorbed sulfate species on Au(111) in 0.5 mol L1 D2 SO4 in D2 O
and (b) band center energy as a function of the applied potential. (From Ref. [48] with permission.)
Fig. 15
More detailed results from the calculations are necessary to account for these
differences. The controversy on the nature
of the adsorbed sulfate species (sulfate or
bisulfate) can be attributed to the limitations of the in situ FTIR spectroscopy in
this case. Associated with the difculty in
identifying the nature of the adsorbed anion (sulfate or bisulfate) is the difculty
in determining the symmetry of the adsorbed species, which would give the clue
in determining the coordination to the surface. The problem of the adsorbed sulfate
is that only the symmetric modes will be
active, according to the surface selection
rule (see Sect. 3.4.2.1.2). It is assumed that
the Td symmetry is broken upon adsorption, since the oxygen atoms are no longer
equivalent. The nal symmetry will depend on the coordination to the surface.
Coordinating one or three oxygen to the
surface will lead to a C3v symmetry, while
coordinating two oxygen will lead to a C2v
symmetry. The C2v symmetry has no degenerated modes, while the C3v has two
degenerated modes, one for the stretching
modes. Therefore, for the free species in
solution, the C2v symmetry must present
one more stretching mode than the C3v .
Because of the fact that the surface selection rule limits the active modes to those
having the dynamic dipole moment perpendicular to the surface, two different
coordinations will present the same number of bands. Considering only the number
of bands, it is not possible to identify the
symmetry of the adsorbed sulfate species;
and consequently, it is not possible to assign unequivocally either the coordination
or the nature of the adsorbed species.
Other strategies have been used to clear
these points.
Shingaya and coworkers have simulated
the electrochemical environment in ultra
high vacuum (UHV) by adding water and
SO3 to a Pt(111) surface to compare it with

in situ spectra [49]. They concluded that
depending on the amount of water, the two
species can be identied. Bands at 1230
and 953 cm1 were (arbitrarily) assigned
to the bisulfate. This species would be
converted into sulfuric acid (bands at
1276, 1043, and 941 cm1 ) when water
evaporates. On the basis of these results,
they concluded that there are two adsorbed
species on Pt(111) electrodes, depending
on the applied potential. Bisulfate species
are adsorbed for potentials between 0.4
and 0.75 V (identied by the band at
1230 cm1 ) and sulfuric acid for potentials
between 0.75 and 1.10 V (identied by
the bands at 1276 cm1 ). Although this
is an interesting approach to identify the
ngerprints of adsorbed sulfate species, a
denitive proof for bisulfate and sulfuric
acid could be easily obtained in UHV by
using deuterated water to conrm the
presence of sulfuric acid. A mixture of
different species at a given potential or
surface condition cannot be ruled out
and this is a further complication to the
interpretation of in situ spectra.
Other explanations for the adsorbed
species have also been given. Faguy and
coworkers [50] suggested that the adsorbed
species forms ion pairs with the hydronium ions close to the surface, and they
even suggested that the ion is coordinated by three protonated oxygen [51]. This
model tries to rationalize the discrepancy
in the assignment of vibrational bands of
the in situ FTIR spectra to sulfate or bisulfate. Actually, the close proximity of the
coadsorbed hydronium ions can mimic
the bisulfate situation and this could be
also an explanation of the Shingaya and
Ito results. A change in the hydronium ion
concentration with potential may happen,
producing a different number of hydrogen
265
266
bonding between the sulfate ions and the

coadsorbed water.
Other strategy to identify the coordination and therefore to deduce the symmetry
of adsorbed sulfate was presented by
Lennartz and coworkers [43], using simultaneously in situ infrared spectroscopy and
in situ STM. In this study, the authors used
in situ STM to obtain the registry or the
sulfate adlayer to the Cu(111) substrate by
changing the bias current and obtaining an
image containing the contrast due to both
the adsorbed sulfate and the underlying
Cu(111). Using Fourier Transform ltration, they were able to separate both signals
and, since these were obtained from the
same experiment, the overlap of the two
images gave the registry between the adsorbate and the underlying surface. They
observed for Cu(111) a bridging type adsorption of the sulfate using two oxygen
resulting in a C2v symmetry. Knowing the
coordination, it is easier to obtain the symmetry and therefore the identity of the
adsorbed species. For example, two-fold
coordinated adsorbed bisulfate will have a
Cs symmetry. Under this symmetry, the
noncoordinated SO and SOH will produce two vibrations, other than the SO2
symmetric vibrations, with the component
in the direction perpendicular to the surface. In the particular case of sulfate ions
adsorbed on Cu(111), it is clear that the
adsorbed ion is the dissociated sulfate and
not the bisulfate ions, since only the strong
symmetric band at 1205 to 1220 cm1 has
been observed [43].
Another important result from in situ
STM imaging of adsorbed sulfate on
M(111), single-crystal metal electrodes is
the observation of the adsorbate structure.
It is found for Au(111) [42], Pt(111) [52],
Rh(111)
[53], Ir(111) [54], and Cu(111) [43]
a 3 7 structure. This kind of information can be used in determining lateral
interactions of the adsorbed ions, since the

ionion distance can be measured. However, there are evidences that hydronium
ions or water molecules are adsorbed
between two neighboring ions in the 7
direction. This obviously can break the
chemical interaction, but probably not the
dipoledipole coupling since water and
sulfate frequencies are well separated.
The
two ions adsorbed in the 3 direction

do not have water molecules in between
and therefore chemical interactions can be
operative as well.
Although the nal symmetry cannot
be deduced only from the spectroscopic
data, it is important to stress that the
spectroscopic data supply information on
the strength of the chemical bonding [48]
and on the effect of the applied electric
eld [55, 56]. Moreover, in situ STM
data have been reported only for a full
monolayer, but not for submonolayer
coverage. The in situ FTIR is sensitive
also in the submonolayer regime. By
appropriate calibration (through the full
monolayer as determined by in situ STM),
it is possible to determine the adsorption
isotherm, even for metals for which it is
very difcult to obtain capacitance data.
The Sulfate Ion Adsorbed on the
M(100) and (110) Single-crystal Electrodes
The adsorbed sulfate species on other
low-index planes of platinum and gold
present a more complex spectroscopic
structure [32, 46, 50, 57, 58]. Fig. 16
shows the spectra of adsorbed sulfate at
saturation coverage for the three low-index
planes of the platinum electrode in a
solution of pH = 2.8. Clearly, a second
band at 1008 cm1 can be easily identied
for Pt(100) (weak) [57] and for Pt(110)
(strong) [58]. These two planes do not
match a C3v symmetry for the adsorbed
sulfate ions, and therefore we should
3.4.3.1.2
Pt(110)
1106
1234
Pt(100)
1108
1208
1100
Pt(111)
R/R0 = 0.1%
1254
1300
1200
1100
1000
Wave number
[cm1]
In situ FTIR spectra for the adsorbed sulfate species
on platinum single-crystal electrodes in 0.69 mol L1 HF,
0.5 mol L1 KF, and 102 mol L1 K2 SO4 . Ref. Pd/H2 . (Data
extracted from Ref. [45] (Pt(111)), Ref. [57] (Pt(100)), and
Ref. [58] (Pt(110)) with permission.)
Fig. 16
expect a C2v symmetry if undissociated

bisulfate ions are adsorbed. However, a C2v
symmetry has only two allowed bands, that
is, the symmetric modes with respect to the
symmetry axis (the A1 modes). A second
band indicates that either the adsorbed

sulfate ion is tilted on the surface or that
a second species is also adsorbed. A tilted
adsorbed sulfate through a single oxygen
will permit other vibrations to be active,
267
268
since the symmetry is no longer the C2v

and practically all modes will have some
component of the dynamic dipole moment
perpendicular to the surface.
The other possibility is that on Pt(100)
and Pt(110), sulfate and bisulfate are
coadsorbed on the surface. Comparing
the bands for sulfate species adsorbed
on polycrystalline platinum electrodes in
strongly acid solutions, it is clear that
when bisulfate is the major species in

solution, a band at ca. 1100 cm1 is
observed [59]. A similar situation has been
observed for adsorbed sulfate on Au(100)
as shown in Fig. 17, in which two strong
bands are observed at 1165 and 1091 cm1
for an applied potential of 0.8 V. The
1165 cm1 band shifts to 1185 cm1 ,
but the 1091 cm1 band is practically
insensitive to the applied potential. A weak
R/R0 = 0.033%
1.2 V
956
1091
1.0 V
1181
0.8 V
1165
1091
0.7 V
1300
1200
1100
1000
900
Wave number
[cm1]
In situ FTIR spectra for the adsorbed sulfate species on Au(100)
single-crystal electrode in 7.3 mol L1 HF and 102 mol L1 K2 SO4 .
Ref. Pd/H2 . (Data reproduced from Ref. [34] with permission.)
Fig. 17
band at 956 cm1 can be viewed in some

spectra, which has been observed also
for the sulfate adsorbed on Au(111). The
presence of two strong bands suggests
that probably at this solution pH, there
is a mixture of sulfate and bisulfate on
the surface. Indeed, the sulfate species
adsorbed on Au(100) surfaces in solutions
in which the undissociated sulfate is the
major species, reveals only one strong
band at 1161 to 1190 cm1 , indicating that
when only sulfate is present in solution,
the 1091 cm1 band is absent for the
adsorbed species.
Interestingly the relative intensities of
the 1165 to 1185 cm1 and 1091 cm1
features change with the applied potential.
The 1165 to 1185 cm1 band is stronger
than the 1091 cm1 at more positive
applied potential, which could suggest
that bisulfate can be further dissociated,
since the applied potential induces a OH
adsorption and therefore the acidity on the
surface is changed favoring the bisulfate
dissociation.
Summarizing, the adsorption of sulfate
is of major interest in electrochemistry,
since sulfuric acid solutions are largely
used as a supporting electrolyte for many
applications, including kinetic studies. So,
the molecular structure of the double layer
is an important piece of information in
order to understand the effect of the double layer on the reaction kinetics. The
infrared spectroscopy is an interesting tool
to access this kind of data, providing information on the nature of the species,
and with the help of in situ STM, the
structure and coordination of the adsorbed
ions. It is likely that this structure will
not survive intact when faradaic reactions
take place on the electrode surface, but
certainly the structure formed will affect
the threshold of the reaction, since different structures can account for different
barriers for surface reactions to start on

the electrode surfaces.
3.4.4
Infrared Spectroscopy under

Nonequilibrium Conditions. Study of
Systems Relevant in Electrocatalysis
One of the major successes of in situ

FTIR spectroscopy concerns the study
of complex electrode reactions. In this
chapter, we present typical applications
related to research in electrocatalysis of
alcohol oxidation. With the aim to study
electrocatalysis of reactions involving candidate substances for fuel cells applications, aliphatic alcohols of low molecular
weight have been intensively investigated.
In particular, remarkable progress in this
direction was achieved in the last years in
the case of methanol oxidation. In fact,
the interest in establishing the nature of
the adsorbed species of methanol was the
motivation of one of the rst papers using
in situ infrared spectroscopy [60]. Because
of the fact that adsorbed carbon monoxide is typically formed when small organic
molecules interact with platinum, we start
this section presenting data on the electrocatalysis of CO oxidation.
As explained in Sects. 3.4.2.1 and 3.4.2.2,
application of the external reectance technique requires the use of a thin layer
of solution between the working electrode and the IR window. Because of this
experimental approach, there has been
some reticence to use the IR method
under conditions of current ow. Criticism was centered on IR-drop effects and
reactant-depletion problems in the solution thin layer. However, although IR
drop and reactant depletion indeed occur,
these problems can be experimentally minimized and should not be overestimated.
On one side, usually employed solutions
269
270
have a good conductivity and, for instance,

for a potential step of 100 mV within the
H-region of platinum, charge compensation via ion migration can occur within
tens of a second [30]. On the other side,
since the Fourier transform method allows a fast rate of data acquisition, reactant
depletion can be reduced to a negligible
extent by using a reasonable experiment
design, minimizing polarization time at
each potential. Very valuable information
can then be obtained under conditions of
current ow.
Carbon Monoxide
A signicant number of studies using in
situ FTIR spectroscopy have been devoted
to adsorbed carbon monoxide. This is not
surprising, in view of the supporting experimental and theoretical data on the CO
system in UHV and the interest on carbon
3.4.4.1
monoxide as a model molecule in both

electrochemical and UHV environments.
UHV data have been used to compare
the behavior of adsorbates at the electrochemical interface with that observed
at metal/gas interfaces [23]. Also, spectroelectrochemical data on adsorbed CO are
interesting in the study of the Stark effect,
which can produce changes in band center frequency [55, 61] and integrated band
intensity [56]. Spectroscopic studies of CO
within this context have been reported and
discussed elsewhere [30, 41]. In electrocatalysis, carbon monoxide can be considered as an intermediate (or poison) during
the oxidation of small organic molecules
(methanol, formic acid, ethanol, among
others). In the present section, we have
selected some studies concerning mainly
this issue, using spectroscopic data to compare the CO behavior on different catalysts:
2075
Pt
2061
PtRu
2010
Ru
R /R0 = 0.005
2250
2100
1950
Wave number
[cm1]
In situ FTIRS of a saturated CO

monolayer adsorbed at 300 mV on Pt,
1800 Ru, and PtRu (50 : 50) electrodes in
0.1 M HClO4 . Sample spectra measured
at 300 mV, reference spectra obtained at
800 mV.
Fig. 18
polycrystalline Pt, Ru, and (50 : 50) PtRu

alloy [62], and single-crystal electrodes as
Pt(111)Ru [63, 64] and (0001)Ru [65, 66].
Adsorption and Oxidation of CO
on Polycrystalline Substrates In situ FTIR
spectra in Fig. 18 show spectra for adsorbed CO at polycrystalline Pt, PtRu,
and Ru, respectively. The surfaces were
saturated with CO at 0.3 V in CO-saturated
0.1 M HClO4 solutions. After elimination
of bulk CO, sample spectra were collected
at 0.3 V and corresponding reference
spectra were measured at 0.8 V, a potential
at which CO is completely oxidized.
Only linear-bonded CO is observed on
the three surfaces. The ranges of observed
frequencies are 2065 to 2075 cm1 for Pt,
2055 to 2065 cm1 for PtRu(50 : 50), and
2000 to 2020 cm1 for Ru.
Of major importance for the electrocatalysis, is the comparison of CO2 production
on the different materials as a function
of potential. For this purpose, series of
spectra were taken using the three catalysts in the potential range between 0.1
and 0.6 V versus RHE [62], as shown in
Fig. 19 for the PtRu (50 : 50) alloy. The
left side of Fig. 19, showing the bands
for CO2 , was calculated against a reference spectrum at 0.1 V. The increasing
negative-going band represents the formation of CO2 as the potential is stepped
to higher values. The right side of the
gure was calculated taking as a reference a spectrum at 0.8 V, a potential at
which CO is completely oxidized. Here,
the intensity of the negative-going bands
is proportional to the CO coverage at
the respective potentials. The integrated
band intensities for CO2 , at 2341 cm1
are plotted in Fig. 20. Accordingly, the
catalytic activity for COad oxidation decreases in the order PtRu(50 : 50) >
Ru > Pt.
3.4.4.1.1
Potential dependence of the band frequency

As the potential is stepped in the positive
direction, the band center for adsorbed
CO shifts to higher wave numbers for all
electrode materials (Fig. 21). The following
characteristics in the E relationship
are noteworthy:
For pure platinum, the band center

exhibits rst a linear dependence on
potential with d/dE = 28 cm1 V1 .
As CO oxidation starts, CO remains
approximately constant in the potential
interval between 0.45 and 0.55 V and
then decreases as the oxidation to
CO2 proceeds.
For pure ruthenium, the band center frequency shifts linearly up to a potential
of 0.3 V with d/dE = 52 cm1 V1 .
Then a more pronounced shift (d/dE
77 cm1 V1 ) is observed up to 0.45 V.
The onset of oxidation can be extrapolated to 0.4 V, as judged from the CO2
band intensity.
For the PtRu alloy, the frequency shifts
linearly with d/dE = 32 cm1 V1 at
low potentials and, as oxidation begins,
the frequency drops immediately to
lower values.
At constant (saturation) coverage, the
marked effect of potential in pure Ru
(highest d/dE value) can be related
to a stronger backbonding effect. This
is also reected in the lower frequency
values of the COL band at this metal
as compared to the other two. Figure 21
gives also information on the properties
of the adlayer as oxidation sets on. Thus,
the faster red shift (to lower frequencies)
of the band center frequency for PtRu
beginning at the onset of CO oxidation
indicates a rapid decrease of the lateral
interactions, that is, oxidation of the
CO adlayer probably occurs within a
relatively loose structure. Contrasting, the
271
272
0.60 V
0.55 V
0.50 V
0.55 V
0.45 V
CO(L)
0.50 V
0.40 V
0.45 V
0.30 V
CO2
0.20 V
0.40 V
0.10 V
0.35 V
R/R = 5 103
2400
2300
2200
2100
2000
1900
Wavenumber
[cm1]
In situ FTIR spectra for saturated CO adsorbates on a
PtRu(50 : 50) alloy electrode in 0.1 M HClO4 from 100 mV to
600 mV at 50 mV interval. CO was adsorbed at 300 mV. For each
spectrum, 100 interferograms were collected at 8 cm1 resolution,
acquisition time ca. 44 s. Bands on the left side, corresponding to
CO2 produced during COads oxidation, were calculated against a
reference spectrum taken at 100 mV (CO2 was absent). Bands on
the right side, corresponding to adsorbed CO, were calculated
against a reference spectrum taken at 800 mV after complete
oxidation of CO. (From Ref. [62] with permission.)
Fig. 19
red shift observed on the pure metals

is somewhat slower indicating that CO
oxidation probably occurs only at the
border of islands, the local CO coverage
remaining relatively high.
Two different slopes observed for pure
Ru at potentials prior to CO oxidation have
been related to the possible interaction
with adsorbed oxygen (Thermal desorption
data from coadsorption of CO and O at
Ru(001) under UHV conditions [67] show

a decrease of CO desorption energy from
Ed = 140 kJ mol1 for the pure metal
to Ed = 107 kJ mol1 in the presence of
coadsorbed oxygen (O = 0.5)) [62, 67].
Such a behavior is expected in the presence
of electronegative coadsorbates, which
cause a withdrawal of metal electrons
and diminish the metal-* back donation.
The consequences of this effect are
1.0
Integrated band intensity

[a.u.]
CO2
0.8
0.6
0.4
0.2
PtRu
Ru
Pt
0.0
0.2
0.4
0.6
Potential (RHE)
[V]
Integrated band intensity for CO2 as a function of potential.
The intensities were calculated from series of spectra obtained at each
catalyst as shown in Fig. 20 for the PtRu alloy. (From Ref. [62] with
permission.)
Fig. 20
a lowering of the metal-CO binding

energy as observed in UHV [67] and a
stronger CO bond, which is reected
in the increase of CO shift between 0.35
and 0.45 V.
The relative absorption coefcients of CO
on Pt, Ru, and PtRu alloys Spectra for CO
submonolayers on the three materials are
shown in Fig. 22. The degree of coverage
CO indicated in the spectra was calculated using the ratio of the band intensity
for CO2 to the maximum CO2 intensity,
obtained after complete oxidation of a
CO-saturated layer. Figure 22 shows that
for comparable band intensities of CO2 ,
the corresponding intensity of the COad
feature largely differs for the three electrodes, the largest signal being observed at
Pt and the smallest one at Ru.
A comparison of the absorption coefcients for CO on different materials

requires a normalization, taking into account differences in electrode roughness.
With the aim of normalization, one can
use, again, the band intensity of CO2
produced upon total oxidation of the
respective adlayer. The (COads )/(CO2 ) intensity ratios are 0.87 (for Pt), 0.60 (for
PtRu), and 0.25 (for Ru) with an estimated error of ca. 8% [62]. These normalized intensities of COad reect, in a
rst approximation, the relative values of
the absorption coefcient of CO on the three
substrates.
On the basis of the above discussion,
the lower intensity of the CO band at Ru
may really reect a lower value for the
dynamic dipole moment of the transition
for CO on this metal as compared to the
273

Dependence of the band center
frequency on potential for saturated CO
adlayers on Pt, Ru, and PtRu(50 : 50)
alloy. The frequencies were taken from
series of spectra measured at each
catalyst as shown in Fig. 20 for the PtRu
alloy. (From Ref. [62] with permission.)
Fig. 21
2085
Pt
2070
2055
Wave number
[cm1]
274
PtRu
2040
2025
Ru
2010
1995
0.0
0.2
0.4
0.6
0.8
E vs RHE
[V]
other materials. In a study of the effects of

coverage on the band intensity and band
center of COad at Ru(001), H. Pfnur and
coworkers [68] have suggested that strong
dipoledipole coupling can reduce the dynamic dipole moment. Furthermore, these
authors show that this effect markedly
depends on the size of the CO islands. Indeed, the band intensity should be affected
by lateral interactions to a larger extent,
when adsorption takes place in the form
of large islands, since in such a case
most of CO molecules are surrounded by
neighbors of an identical nature. Using
a different theoretical approach, Persson
and Rydberg [69] conrmed the predominance of dipoledipole coupling in the
coverage effects observed by Pfnur and
coworkers [68], and explained the reduction of the band intensity as being due to
a screening effect caused by the electronic

polarizability of the adsorbed molecules.
In such a case, the reduced band intensity for CO could be caused by strong
dipoledipole coupling and/or large island
domains; this would be an alternative (or
additional) explanation for the lower rate
of CO oxidation observed during oxidative
stripping at pure ruthenium as compared
to the alloy.
The dependence of CO on CO for Pt,
PtRu, and Ru is shown in Fig. 23. A linear
dependence of CO values is observed
for all three materials, the respective
dCO /dCO values being 58 cm1 (for Pt),
64 cm1 (for Ru), and 50 cm1 (for PtRu).
It can be stated that, from the spectroscopic
point of view, the PtRu (50 : 50) alloy
approaches more closely the behavior of
pure Pt than that of pure Ru.
q = 0.12
q = 0.41
q = 0.43
q = 0.45
q = 0.69
q = 0.69
q = 0.75
q=1
q=1
q=1
R/R0 = 5 103
R/R0 = 5 103
Pt
2400
2200
2000
R/R0 = 5 103
PtRu (50:50)
2400
2200
2000
Ru
2400
2200
2000
Wave number
[cm1]
Comparison of in situ FTIR spectra for submonolayer CO adsorbates on Pt,
PtRu(50 : 50), and Ru electrodes in 0.1 M HClO4 . CO was adsorbed at 300 mV. Sample
spectra were acquired at 300 mV and computed against a reference spectrum taken at
800 mV; 100 interferograms were collected at 8 cm1 resolution, acquisition time ca. 44 s.
The CO coverages co indicated on each spectrum were calculated from the ratio of the
band intensity for CO2 at each coverage to the CO2 band obtained at saturation coverage.
(From Ref. [62] with permission.)
Fig. 22
Adsorption and Oxidation of CO

on Single-crystal Substrates
3.4.4.1.2
CO adsorbed on Pt(111)/Ru Attempts

to study the inuence of the surface
structure on the rate of oxidation of
CO, methanol, and ethanol motivated the
preparation of catalysts using adsorbed
Ru on well-dened Pt(111) surfaces [63].
However, the resulting structures, when
investigated via STM, show that Ru
segregates on the surface of Pt(111)
forming islands [70, 71]. The catalytic

action is thus expected to occur at the
border of Ru domains.
In situ FTIR spectra for adsorbed CO
on this material are shown in Fig. 24(a, b)
for two Ru coverages. At low Ru coverage
(Ru = 0.2), two bands related to CO
bonded to Pt atoms in on-top and bridge
positions are observed near 2058 cm1
and 1816 cm1 , respectively. A third band
at 1982 cm1 corresponds to CO linearly
bonded to Ru atoms. This feature is better
275
2080
Pt
PtRu
2040
Wave number
[cm1]
276
Ru
2000
1960
0.0
0.2
0.4
0.6
0.8
1.0
qCO
Dependence of vCO on CO for submonolayer CO adsorbates
on Pt, PtRu(50 : 50), and Ru electrodes in 0.1 M HClO4 at a constant
potential of 300 mV: CO adsorption at 300 mV. Other details as in
Fig. 23. (From Ref. [62] with permission.)
Fig. 23
developed at the electrode with higher

concentration of Ru, appearing at near
2008 cm1 . This and the band for COL at Pt
sites (2047 cm1 ) are the only prominent
features observed.
It is noteworthy that for both surface
coverages with Ru, Pt(111)/Ru exhibits
features corresponding to CO frequencies
near those of pure ruthenium and pure Pt
as well. This is not the case for the PtRu
(50 : 50) alloy, which exhibits only one band
at frequencies close to those for pure Pt
(Figs. 18 and 23). It can be thus stated that
opposite to the Pt(111)/Ru electrode, the
behavior of the alloy surface is that of a
more homogeneous surface, with respect
to the distribution of individual Pt and
Ru atoms.
In Fig. 25, the band center frequencies
for COL on different materials are plotted
as a function of the electrode potential;

pure polycrystalline Ru is included for
comparison. The Pt(111)/Ru samples
exhibit values near to pure Pt(111) (at low
Ru concentration) or pure Ru (at high Ru
concentration).
Comparing the binary catalyst with the
pure metals, it can be stated that the presence of Ru shifts the COPt feature to
lower wave numbers. Conversely, the presence of Pt shifts the CORu feature to
higher wave numbers. This observation
can be related to the degree of backbonding. Since the work function of Pt(111)
(5.93 eV) [72] is substantially higher than
that of Ru (4.71 eV) [73], a signicant electron transfer from Ru to neighboring Pt
atoms is expected. Such an electronic
effect should weaken the RuCO bond
(because of decreased back donation) and
600
600
550
550
500
500
550
550
450
450
400
400
500
500
350
350
300
300
450
250
400
450
250
400
200
200
150
150
R/R0 = 0.01
1816
2058
qRu = 0.2
2400
2200
(a)
Wave number
[cm1]
2000
qRu = 0.75
1800
1982
2047
2008
R/R0 = 0.01
1982
2400
2200
(b)
Wave number
[cm1]
2000
1800
In situ FTIR spectra for adsorbed CO and for the CO2 produced during the CO
stripping in 0.1 M HClO4 on Pt(111) modied with Ru. Ru coverage: (a) 0.2 mL and
(b) 0.75 mL. After the formation of the CO adlayer at 0.1 V, CO was eliminated from the
solution by N2 bubbling. The potential was then changed from 0.1 V onwards in 50 mV
steps. The sample potentials are indicated in the corresponding spectra. The CO2 region
(band at 2341 cm1 ) was calculated with a reference spectrum taken at 0.1 V (a potential
in which CO2 is not formed). For the CO region, the reference spectrum was one taken at
0.8 V, that is, a potential in which COads was completely oxidized. (From Ref. [63] with
permission.)
Fig. 24
strengthen the PtCO bond (increased

back donation). Thus, the relative values
of the IR frequencies can be taken as a
measure of the COMetal bond strength.
The rate of CO oxidation at 0.45 V on
different PtRu materials was monitored
during several minutes via the changes

in both CO and CO2 signals. The result
is plotted in Fig. 26. Surprisingly, the
Pt(111)/Ru electrode containing 75% Ru
presents the highest rate of CO oxidative
stripping. The result can be rationalized in
277
Pt(111)
Wave number
[cm1]
2060
2040
Pt(111)/Ru
2020
Ru
2000
0.0
0.2
0.4
0.6
Potential vs RHE
[V]
Comparison of the CO stretch wave number for the COL on the
Pt(111), Ru-modied Pt (111), and pure polycrystalline Ru as a function
of potential. Data were derived from IR spectra as those in the gure.
Experimental conditions as in Fig. 25. (From Ref. [63] with permission.)
Fig. 25
1.0
0.8
0.6
(a)
Ru
PtRu alloy
0.4
0.2
0.0
Pt(111)/Ru
0
0.6
Pt(111)/Ru
0.4
PtRu alloy
0.2
Ru
Pt(111)
0.0
Time
[min]
Change of the IR band intensity for

adsorbed: (a) CO and (b) for CO2 during CO
stripping at a constant potential of 0.45 V on
Pt(111), Pt(111)/Ru with Ru = 0.75,
PtRu(50 : 50) alloy, and pure polycrystalline Ru.
Fig. 26
0.8
Pt(111)
Band intensity for CO2

[a.u.]
Band intensity for COL

[a.u.]
278
0
(b)
Time
[min]
The respective surface was saturated with CO at

0.1 V after which, CO was eliminated from the
electrolyte by nitrogen bubbling. (From Ref. [63]
with permission.)
terms of a reaction mechanism as follows:

H2 O + * OHad + H+ + e
(16)
OHad + COad 2* + CO2 + H+ + e
(17)
Where * represents a free site on
the surface.
On pure Pt, the rst step is rate
determining, that is, the overall process
starts as soon as OH adsorption sets in.
The oxidation of the CO adsorbed layer
is considered to initiate from sites that
allow OH adsorption. On pure Ru, on
the other hand, the second step seems
to be rate determining. The adsorption of
O-containing species has been reported
already for potentials down to 0.2 V
versus RHE [74], that is, well below the
CO oxidation peak. This means that
OH and CO can coexist on the surface
within a certain potential region. The
second step sets in only at a higher
potential. In this model, the faster CO
oxidation on the Pt(111)/Ru electrode can
be explained by a decrease in the bond
energy CORu, reected in the decrease
of the backbonding effect.
CO adsorbed on Ru(0001) In situ FTIRS
was used to study CO adlayers on
Ru(0001) [65, 66]. Contrasting the behavior at polycrystalline Ru, where only the
band for linear-bonded CO is observed
(Fig. 27a), a saturated adlayer of CO adsorbed on Ru(0001) exhibits two different
geometries. The spectra in Fig. 27(a) show
two bands assigned to CO in linear
and hollow position, COL and COH ,
respectively. This assignment is supported
by STM data [66] showing this adsorption
geometry in a c(2 2)2CO structure.
The potential dependence of the band
intensity for both forms of adsorbed CO
and for the oxidation product, CO2 , is

plotted in Fig. 27(b). The initial values of
the band intensity for COL and COH are
constant and start decreasing concomitant
with the appearance of CO2 at ca. 200 mV
versus Ag/AgCl. The intensity of the CO2
feature passes through a maximum at
ca. 600 mV. From this potential onwards,
the signal decreases, in spite of the fact
that oxidation of COL continues up to a
potential of 1100 mV. The latter process,
however, is relatively slow and CO2
diffuses out of the thin layer, causing the
observed decrease of the band intensity.
When analyzing the potential-dependent
behavior of the band frequency, additional
information on the surface structure and
composition is useful. Thus, data from
Auger and low-energy electron diffraction
Spectroscopy (LEEDS) experiments were
performed on emersed Ru(0001) electrodes [65]. In the range of potentials
between 80 mV and +200 mV versus
Ag/AgCl, a (2 2) phase of oxygen exists
on the surface [65]. Furthermore, oxygen
was observed up to an emersion potential of 1100 mV, though forming another
phase, which was suggested to be (1 1).
Four different potential regions are distinguished in Fig. 27(c) for the potentialdependent frequency of linear-bonded
CO. The initial slope of the plot up to
175 mV, (dCO /dE) = 37 cm1 V1 , was
interpreted as due to either a Stark effect or
to electron back donation from the metal
into the 2 orbital of CO [30]. As the potential increases, between 200 and 425 mV
versus Ag/AgCl, the slow oxidation of CO
commences and the slope (dCO /dE) decreases slightly to 29 cm1 V1 . The third
region, with (dCO /dE) = 4 cm1 V1 , is
accompanied with a signicant increase
of CO oxidation. At potentials between
625 mV and 1100 mV versus Ag/AgCl,
the frequency changes again with a slope
279

1100 mV
(a)
Ru(0001)
550 mV
500 mV
(b)
16
Band intensity
[a.u.]
5
1000 mV
800 mV
650 mV
12
COL
8
4
CO2
COH
450 mV
0
2040
400 mV
(c)
250 mV
2030
200 mV
2020
150 mV
2010
nCO
[cm1]
280
0 mV
COL
(d)
1800
100 mV
1790
200 mV
COH
R/R0 =
0.005 COL
2100
2000
1780
1900
1800
1770
1760
Wave number
[cm1]
COH
0.2 0.0 0.2 0.4 0.6 0.8 1.0
E vs Ag/AgCl
[mV]
(a) In situ FTIR spectra for CO adsorbed at Ru(0001) at different potentials (vs
Ag/AgCl/Cl ) as indicated. CO was adsorbed at 100 mV during 5 min, then the solution was
replaced with pure base electrolyte (0.1 M HClO4 ). Spectra were normalized versus a
spectrum taken at 1100 mV, after waiting for 2 min. in order to completely oxidize CO.
(b) Integrated band intensities for COL , COH from spectra is shown in (a). CO2 band intensity
from the same experiments. (c), (d) Potential dependence of the band center frequency for
COL and COH , respectively, from spectra shown in (a). (From Ref. [65] with permission.)
Fig. 27
similar to the initial value for the saturated

layer. Observing that CO oxidation slowly
continues, it is obvious that the effect of
potential on the band center frequency
prevails over the lowering of the total CO
coverage. That means that compact island
domains of CO on the surface of Ru(0001)
remain on the surface even at very high
potentials [65].
At submonolayer coverage, a splitting of

the COL feature was reported [65], which
indicates different CO-adlayer structures
that can be tuned by potential. It has been
suggested that coadsorbed O, forming different structures depending on potential,
plays an important role in the compression
processes of the COL adlayer [65]. Conrming this interpretation, no splitting
of the COL adlayer was observed for

electrodes pretreated by annealing at
1200 K in a hydrogen stream [66]. Under
these conditions, the Ru oxide layer seems
to be completely eliminated. The band for
COH , Fig. 27(d), shows several transitions
whose origin needs to be investigated. A series of spectra with different CO coverages
exhibits this band only for the saturated
adlayer [66].
Methanol
The electrooxidation of methanol occupies,
since many years, the center of interest
in electrocatalysis of fuel cell reactions.
Some of the most important ndings on
this issue are described in Chapter 5.2.
We do not intend to review here all of
the spectroscopic work done on methanol
but to illustrate some highlights of the
FTIRS method when applied to study the
electrocatalysis of methanol oxidation.
It is well established that the rate of
methanol electrooxidation at platinum can
be enhanced in the presence of Ru as
a cocatalyst. In Fig. 28 are represented
in situ FTIR spectra obtained during
oxidation of methanol on three different
materials: Pt(111), Pt(111)/Ru(39%), and a
PtRu alloy (85 : 15); the second electrode
was prepared by adsorbing Ru from a
RuCl3 solution onto a well prepared
Pt(111) electrode [70]. For each material
sample, spectra (R) were obtained at the
indicated potential and calculated as the
ratio R/R0 , with R0 being a respective
reference spectrum obtained at 0.05 V, a
potential in which methanol adsorption
is minimized.
For all electrodes, the band at 2341 cm1
is due to CO2 formed in the thin-layer
cavity. Another feature due to a soluble
species is the band at ca. 1720 cm1 due to
a carbonyl group of the product formic acid
c/o methyl formate (Methyl formate can
3.4.4.2
originate through a reaction of HCOOH

with excess CH3 OH.). Other bands result
from adsorbed species produced during
methanol adsorption. Thus, the bands at
ca. 2050 cm1 and 1820 cm1 are due,
respectively, to linear and bridge adsorbed
CO [60].
The spectra reveal that from all three
materials, the alloy shows up (a) the largest
production of CO2 and (b) the largest
intensities for linear-bonded CO at low
potentials. Thus, the gained information
via IR spectroscopy is, not only the better
performance of the alloy in relation to CO2
production but also its ability to adsorb
methanol at lower potentials (observe
the band intensity for CO at 0.15 mV).
Besides this, for both PtRu materials,
the CO band reaches at 0.30 to 0.35 V
a stationary value and, in spite of a
strong production of CO2 at 0.5 V, the
band intensity for CO practically does
not diminish. This indicates a high-rate
methanol of readsorption at PtRu catalysts.
The catalytic mechanism of PtRu has
been interpreted in terms of a so-called
bifunctional effect of the surface in
which Pt sites adsorb and dissociate
methanol-forming CO and Ru atoms
adsorb and dissociate water molecules,
thus providing, at low potentials, oxygen
atoms needed to complete the oxidation of
adsorbed CO to CO2 [75]. The facts above,
showing an increased rate of adsorption of
methanol in the presence of Ru, indicate
that the bifunctional mechanism alone
does not fully describe the catalytic action
of ruthenium.
Ethanol
The electrocatalysis of ethanol oxidation
using in situ FTIR spectroscopy was studied in several papers [7680]. Some facts
concerning the soluble reaction products
at the low-index faces of Pt single-crystal
3.4.4.3
281
282

0.20 V
0.15 V
0.20 V
0.30 V
0.30 V
0.35 V
0.40 V
0.45 V
0.50 V
0.40 V
0.50 V
0.55 V
0.50 V
0.55 V
2056
2046
0.55 V
1721
2059
1826
1719
R/R0 = 0.02
R/R0 = 0.03
R/R0 = 0.01
2341
2500
2341
Pt(111)
2000
2500
Pt(111)/Ruads
2000
2500
2341
PtRu alloy
2000
Wavenumber
[cm1]
In situ FTIR spectra for Pt(111), Pt(111)/Ru 39%, and PtRu alloy
(85 : 15) in 0.5 M CH3 OH + 0.1 M HClO4 . Potentials as indicated on each
spectrum; reference spectrum taken at 0.05 V. (From Ref. [70] with permission.)
Fig. 28
electrodes have been discussed by Chang

et al. [77]. In a more or less quantitative
manner, these authors used FTIR spectra to estimate the amount of adsorbed
and soluble species formed during a potential scan. Some disagreement with later
results in Ref. [80] concerning the intensity of CO bands may be explained in
terms of a difference in the experimental
approach.
As in the case of methanol, when
Pt electrodes are contacted at 0.05 V
versus RHE, with the solution containing
C2 H5 OH + 0.1M HClO4 , ethanol adsorption is minimized. Therefore, this procedure allows the monitoring of the complete
process of ethanol adsorption and oxidation as the potential is increased in the
positive direction. Because of the wellknown problem of surface poisoning by
strongly adsorbed organic residues, data
obtained under these conditions may not
be observable in the second and following
potential scans or when the electrode
comes in contact with ethanol without potential control.
These anions migrate into the thin layer

of electrolyte to keep electric neutrality
(note, that on positively increasing the
potential, in addition to the double-layer
charge changes, the hydrogen ions concentration in the thin layer of electrolyte
increases. Both processes requiring the
migration of negatively charged ions into
the thin layer).
Adsorbed species observed are linearbonded CO at ca. 2050 cm1 and acetate ions at 1404 to 1412 cm1 [82, 83].
Adsorbed acetate ions originate from
Spectral sequences during the electrochemical adsorption and oxidation of

0.1 M ethanol in 0.1 M HClO4 on Pt(111)
and Pt(100) are displayed in Fig. 29. In
Table 2, the band assignments for the
bands marked in the spectra are given. Acetaldehyde, acetic acid, and CO2 have been
earlier reported from in situ FTIR results,
as products of ethanol oxidation [76]. In
addition to these soluble products, a
gain of bulk perchlorate ions (band at
1110 cm1 ) can be observed in the spectra with increasing positive potentials.
0.15
0.15
0.35
0.25
0.40
0.30
0.50
0.50
0.60
0.60
E
[V]
E
[V]
0.55
1866
0.65
0.70
3000
2500
2000
Wave number
[cm1]
1500
2341
R/R0 = 1.5%
2625
Pt(111)
3000
1407
1282
1712
1370
R/R0 = 1.5%
0.80
2055
1282
2343
2632
1412
2061
1713
2983
2906
0.70
Pt(100)
2500
2000
1500
Wave number
[cm1]
In situ FTIR spectra (256 scans, 8 cm1 resolution) at Pt(111) (left side) and Pt(100)
(right side) in 0.1 M C2 H5 OH + 0.1M HClO4 . Reference spectrum collected at 50 mV, and
sample spectra collected at the indicated potential in 50 mV steps (some spectra were
omitted for clarity). (From Ref. [80] with permission.)
Fig. 29
283

Tab. 2
Assignment of some of the fundamental bands in the spectra of Fig. 29
Wave number
[cm1 ]
Functional group or
chemical species
Mode, comments,
references
2983, 2906
2632 (broad)
2341
20552060
1713
1402
1370/1281
1100
CH3 , CH2
COOH
CO2
Adsorbed CO
COOH or CHO
Adsorbed CH3 COO
COOH
ClO4
CH str., 81
OH str., 81
CO asym. str., 81
Linearly bonded, 60
C=O str., carbonyl, 81
CO sym. str., 82
Coupl. CO str. OH def., 81
ClO str. (F), 35
dissociation of acetic acid in the adsorbed state. Bridge-bonded CO, at about

1860 cm1 is observed at Pt(100) and
to a minor extent at Pt(111). In the
presence of bulk ethanol, as in the experiments in Fig. 29, positive-going features
as those near 3000 cm1 (CH stretching
region [83]) are due to consumption of bulk
ethanol.
Potential Dependence of the Band
Intensities The dependence of the band
intensities for COad and CO2 plotted
3.4.4.3.1
in Fig. 30(a, b) attest for the surface

sensitivity of the reaction. At Pt(100) both
linear and bridge-bonded CO start being
formed short above 0.2 V. In the lowpotential region, before oxidation begins,
the site interconversion, CO-bridge
CO-linear, is observed. This phenomenon
is well-known for CO formed at Pt(100)
either from dissolved CO [23, 25] or from
the adsorption of other fuels [84, 85]. At
around 0.5 V, oxidation to CO2 begins;
maximum production of CO2 is observed
at 0.8 V. At this potential, all adsorbed
Pt(111)
Pt(100)
20
45
CO2
15
Band intensity
[a.u.]
Band intensity
[a.u.]
284
10
COL
COL
CO2
30
COB
15
5
0
0
0.2
(a)
0.4
0.6
0.8
1.0
Potential vs RHE
[V]
1.2
0.2
(b)
0.4
Band intensity for CO and CO2 from spectra collected during the
oxidation of 0.1 M ethanol in 0.1 M HClO4 at (a) Pt(111) and (b) Pt(100).
Experimental conditions as in Fig. 29. (From Ref. [80] with permission.)
Fig. 30
0.6
0.8
Potential vs RHE
[V]
1.0
CO is consumed and the CO2 signal

decays as CO2 diffuses out of the thinlayer cavity.
At Pt (111), the formation of linearbonded CO starts around 0.15 V. The
integrated band intensity in Fig. 30(a), indicates a maximum for CO formation
at 0.55 V. The arrest observed at intermediate potentials coincides with the
beginning of the formation of acetaldehyde (band at 1713 in Fig. 29), thus
indicating that this reaction competes with
that producing CO (i.e. cleavage of the
CC bond).
A general phenomenon of small organic
molecules is the presence of anodic
currents during the negative-going sweep
of the voltammogram as shown in Fig. 31.
For the interpretation of this reactivation
phenomenon, the formation of reaction
products during a cyclic scan as shown
in Fig. 32 should be considered. For
these experiments, a 60 CaF2 prism was
used to monitor the bands for adsorbed
CO and CO2 . Formation of acetaldehyde

can be followed without interference
via a feature at 933 cm1 [77]. However,
because of the cut off of CaF2 (around
1000 cm1 ), a at ZnSe window had to
be used to monitor the band intensity
for acetaldehyde (933 cm1 ). Data for
acetic acid (1280 cm1 ) in Fig. 32 were
taken from the spectra obtained with the
ZnSe window.
The changes for acetaldehyde and acetic
acid resemble those reported by Chang
et al. [77]. The onset of acetaldehyde formation, at about 0.4 V, is accompanied by
a sluggish appearance of acetic acid. Both
signals exhibit a pronounced increase in
intensity at 0.6 V, coinciding with the onset of CO oxidation, that is, with a higher
availability of Pt sites. It is interesting to
note, during the positive-going polarization, that CO2 grows up to 1.0 V, that
is, beyond the potential at which the surface becomes free from CO. This being
an indication of the presence of other
1.0
Current density
[mA cm2]
0.8
0.6
0.4
0.2
0.0
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
Potential vs RHE
[V]
Cyclic voltammogram for a Pt(111) electrode in 1 M C2 H5 OH + 0.1 M
HClO4 solution. Sweep rate 0.05 V/s.
Fig. 31
285
(a)
8
9
COB
2
0
CH3COOH
Band intensity
[a.u.]
Band intensity
[a.u.]
286
COL
1.5
5
4
CH3CHO 3
3
1.0
2
CO2
0.5
0
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
(b)
Potential vs RHE
[V]
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Potential vs RHE
[V]
(c)
Potential dependence of the band intensities for CO, CO2 , acetaldehyde, and acetic
acid, from spectra obtained at Pt(111) during the oxidation of ethanol in 1 M C2 H5 OH + 0.1 M
HClO4 . After taking a reference spectrum at 0.05 V, potential steps were applied towards higher
positive potentials and then back to the initial value. All spectra are referred to the rst one taken
at 0.05 V. A 60 prismatic CaF2 window was used for CO and CO2 and a at ZnSe window for
acetaldehyde and acetic acid. (From Ref. [80] with permission.)
Fig. 32
strongly adsorbed species not observed in

the spectra. As the applied polarization
is reversed, both acetaldehyde and acetic
acid increase slowly. However, at ca. 0.9 V,
which is the potential at which the current in the voltammogram increases again,
a pronounced increase of acetic acid is
observed. Also, CO2 and CO grow during
the reverse scan from 0.8 V downwards.
The recovering in activity has been related to a competition between water and
the organic substance for Pt sites [86, 87].
Infrared spectroscopy data show that the
strength of water adsorption increases as

the applied potential grows in the positive direction. During the negative-going
scan, a decrease of waterplatinum interaction favors the access of ethanol to the
surface as indicated by the increase of the
CO feature.
3.4.5
Final remarks
The examples presented in this chapter show the power of in situ FTIR
spectroscopy to study electrochemical systems. Under equilibrium conditions the

infrared spectroscopy provides molecular
level information on the double layer,
contributing to the understanding of the
physics and chemistry of the charged interface. Thus, the potential and structure
dependence of ion adsorption and the geometry of the adsorbed ions as well, can
be inferred from the inspection of the IR
spectra.
On the other hand, under conditions
of current ow, IR spectra can be used
as a tool to obtain analytical information on electrochemical processes. This
approach has provided valuable information on the electrocatalysis of the oxidation of organic substances, which is
currently used for mechanistic interpretations and for the evaluation of catalytic
materials.
Acknowledgments
FAPESP and CNPq are gratefully acknowledged for their nancial support.
References
1. L. H. Little, Infrared Spectra of Adsorbed
Species, Academic Press, London, 1966.
2. S. A. Francis, A. H. Ellison, J. Opt. Soc. Am.
1959, 49, 131138.
3. R. G. Greenler, J. Chem. Phys. 1966, 50,
310315.
4. A. H. Reed, E. Yeager, Electrochim. Acta
1970, 15, 13451354.
5. H. Neff, P. Lange, D. K. Roe et al., J. Electroanal. Chem. 1983, 150, 513519.
6. W. N. Hansen, J. Opt. Soc. Am. 1968, 58, 380.
7. K. Ataka, M. Osawa, Langmuir 1998, 14,
951959.
8. M. Osawa, K. Ataka, K. Yoshii et al., J. Electron. Spectrosc. 1993, 64, 371379.
9. Z. Ping, G. E. Nauer, H. Neugebauer et al.,
Electrochim. Acta 1997, 42, 16931700.
10. A. Bewick, K. Kunimatsu, B. S. Pons, Electrochim. Acta 1980, 25, 465480.
11. A. Bewick, K. Kunimatsu, B. S. Pons et al., J.

12. B. Beden, A. Bewick, C. Lamy, J. Electroanal.
Chem. 1983, 148, 147160.
13. A. Bewick, J. Electroanal. Chem. 1983, 150,
481493.
14. T. Bae, D. A. Scherson, E. Yeager, Anal.
Chem. 1990, 62, 4549.
15. D. S. Corrigan, M. J. Weaver, J. Electroanal.
Chem. 1988, 239, 5566.
16. T. Iwasita, F. C. Nart, J. Electroanal. Chem.
1990, 295, 215224.
17. P. R. Grifths, J. A. de Haseth in Chemical Analysis: A Series of Monographs on
Analytical Chemistry and Its Applications
(Eds.: P. J. Elving, J. D. Winefordner), John
Wiley & Sons, New York, 1986, Vol. 83,
pp. 274283.
18. S. Pons, J. Electroanal. Chem. 1983, 150,
495504.
19. S. Pons, T. Davison, A. Bewick, J. Electroanal.
Chem. 1984, 160, 6371.
20. S. C. Chang, M. J. Weaver, J. Chem. Phys.
1990, 92, 45824594.
21. S. C. Chang, M. J. Weaver, J. Phys. Chem.
1990, 94, 50955102.
22. S. C. Chang, M. J. Weaver, Surf. Sci. 1990,
230, 222236.
23. S. C. Chang, M. J. Weaver, Surf. Sci. 1990,
238, 142162.
24. F. Kitamura, M. Takahashi, M. Ito, Surf. Sci.
1989, 223, 493508.
25. F. Kitamura, M. Takahashi, M. Ito, J. Phys.
Chem. 1988, 92, 33203323.
26. D. S. Corrigan, M. J. Weaver, J. Electroanal.
Chem. 1988, 241, 143162.
27. L. H. Leung, M. J. Weaver, J. Phys. Chem.
1988, 92, 40194022.
28. K. Kunimatsu, H. Seki, W. G. Golden et al.,
Surf. Sci. 1985, 158, 596608.
29. H. Seki, K. Kunimatsu, W. G. Golden, Appl.
Spectrosc. 1985, 39, 437443.
30. T. Iwasita, F. C. Nart in Advances in Electrochemical Science and Engineering (Eds.:
H. Gerischer, C. W. Tobias), Verlag Chemie,
Weinheim, 1995, Vol. 4, pp. 123216.
31. B. N. J. Persson, Solid State Commun. 1979,
30, 163166.
32. I. R. Moraes, F. C. Nart, J. Electroanal. Chem.
1998, 461, 110120.
33. M. C. P. M. da Cunha, M. Weber, F. C. Nart,
34. I. R. Moraes, M. C. P. M. da Cunha, F. C.
Nart, J. Bras. Chem. Soc. 1996, 7, 453460.
287
288

35. K. Nakamoto, Infrared and Raman Spectra of
Inorganic and Coordination Compounds, 4th
edition, John Wiley & Sons, New York, 1986.
36. P. Christensen, A. Hamnett in Comprehensive Chemical Kinetics (Eds.: R. G. Compton,
A. Hamnett), Elsevier, Amsterdam, 1989,
Vol. 29, pp. 1104.
37. R. J. Nichols, A. Bewick, Electrochim. Acta
1988, 33, 16911694.
38. E. L. de Sa, Ph. D. Thesis, Universidad de
Sao Paulo, Sao Carlos, 1997.
39. P. Christensen, A. Hamnett, A. V. G. Muir,
40. A. C. Chapman, L. E. Thirwell, Spectrochim.
Acta 1964, 20, 937.
41. T. Iwasita, F. C. Nart, Prog. Surf. Sci. 1997,
55, 271340.
42. G. J. Edens, X. Gao, M. J. Weaver, J. Electroanal. Chem. 1994, 375, 357366.
43. M. Lennartz, P. Broekmann, M. Arenz et al.,
Surf. Sci. 1999, 442, 215222.
44. A. Tadjeddine, A. Le Rille, Electrochim. Acta
1999, 45, 601609.
45. F. C. Nart, T. Iwasita, M. Weber, Electrochim.
Acta 1994, 39, 20932096.
46. I. R. Moraes, F. C. Nart, J. Braz. Chem. Soc.;
2001, 12, 138143.
47. P. W. Faguy, N. Markovic, R. R. Adzic et al.,
48. Y. Shingaya, M. Ito, J. Electroanal. Chem.
1999, 467, 299306.
49. Y. Shingaya, M. Ito, Surf. Sci. 1997, 386,
3447.
50. P. W. Faguy, N. S. Marinkovic, R. R. Adzic,
Langmuir, 1996, 12, 243247.
51. P. W. Faguy, N. S. Marinkovic, R. R. Adzic,
52. M. Funtikov, U. Stimming, R. Vogel, J. Electroanal. Chem. 1997, 428, 147153.
53. L. J. Wan, S. L. Yau, K. Itaya, J. Phys. Chem.
1995, 99, 95079513.
54. L. J. Wan, M. Hara, J. Inukai et al., J. Phys.
Chem. B 1999, 103, 69786983.
55. F. C. Nart, T. Iwasita, Electrochim. Acta 1992,
37, 21792184.
56. F. C. Nart, T. Iwasita, Electrochim. Acta 1996,
41, 631636.
57. F. C. Nart, T. Iwasita, M. Weber, Electrochim.
Acta 1994, 39, 20932096.
58. T. Iwasita, F. C. Nart, A. Rodes et al., Electrochim. Acta 1995, 40, 5359.
59. F. C. Nart, T. Iwasita, J. Electroanal. Chem.
1992, 322, 289300.
60. Beden, C. Lamy, A. Bewick et al., J. Electroanal. Chem. 1981, 121, 343.
61. D. K. Lambert, Electrochim. Acta 1996, 41,
623630.
62. W. F. Lin, T. Iwasita, W. Vielstich, J. Phys.
Chem. B 1999, 103, 32503257.
63. W. F. Lin, M. S. Zei, M. Eiswirth et al., J.
Phys. ChemB 1999, 103, 69686977.
64. S. Cram, K. A. Friedrich, K. P. Geyzers et al.,
Fresenius J. Anal. Chem. 1997, 358, 189192.
65. W. F. Lin, P. A. Christensen, A. Hamnett
et al., J. Phys. Chem. 2000, 104, 66426652.
66. N. Ikemiya, T. Senna, M. Ito, Surf. Sci. 2000,
464, L681L685.
67. T. E. Madey, C. Benndorf, Surf. Sci. 1985,
164, 602624.
68. H. Pfnur, D. Menzel, F. M. Hoffmann et al.,
Surf. Sci. 1980, 93, 431452.
69. B. N. J. Persson, R. Rydberg, Phys. Rev. B
1981, 24, 69546970.
70. T. Iwasita, H. Hoster, A. John-Anacker et al.,
Langmuir 2000, 16, 522529.
71. H. Hoster, T. Iwasita, H. Baumgartner et al.,
PCCP 2001, 3, 337346.
72. B. E. Nieuwenhuys, W. H. M. Sachtler, Surf.
Sci. 1973, 34, 225.
73. B. E. Nieuwenhuys, R. Bouwman, W. H. M.
Sachtler, Thin Solid Films 1973, 21, 5.
74. E. Ticianelli, J. G. Beery, M. T. Paffett et al.,
75. M. Watanabe, S. Motoo, J. Electroanal. Chem.
1975, 60, 267273.
76. T. Iwasita, B. Rasch, E. Cattaneo et al., Electrochim. Acta 1989, 34, 10731079.
77. S.-C. Chang, L.-W. H. Leung, M. J. Weaver,
J. Phys. Chem. 1990, 94, 60136021.
78. F. Cases, M. Lopez-Atalaya, J. L. Vazquez
433440.
79. T. Iwasita, E. Pastor, Electrochim. Acta 1994,
39, 531537.
80. X. H. Xia, H.-D. Liess, T. Iwasita, J. Electroanal. Chem. 1997, 437, 233240.
81. G. Socrates, Infrared Characteristic Group
Frequencies, Wiley, New York, 1966.
82. A. Rodes, E. Pastor, T. Iwasita, J. Electroanal.
Chem. 1994, 376, 109118.
83. J. Shin, W. J. Tornquist, C. Korzeniewski
et al., Surf. Sci. 1996, 364, 122130.
84. X. H. Xia, T. Iwasita, F. Y. Ge et al., Electrochim. Acta 1996, 41, 711718.
85. T. Iwasita, X. H. Xia, E. Herrero et al., Langmuir 1996, 12, 42604265.
86. T. Iwasita, X. H. Xia, J. Electroanal. Chem.
1996, 411, 95102.
87. T. Iwasita, X. H. Xia, H.-D. Liess et al., J.
Phys. Chem. B 1997, 101, 75427547.
Appendix
If S is the vector representing the electromagnetic energy ux intensity, the

conservation equation will be
dw
+ div S = 0
dt
(A1)
where dw/dt is the speed of energy

dissipation. S is the Poynting vector and is
dened by Eq. (4)
S=EH
(A2)
Where E is the electric eld and H is

the magnetic eld of the electromagnetic
radiation. Giving that the electromagnetic
radiation is a high-frequency radiation, the
observed signal will be a time average of
the radiation dissipation. The time average
of the Poynting vector is Eq. (4)

1
S =
Re(E H )
2
(A3)
Therefore, the rate of the radiation dissipation will be Eq. (2)

1
(A4)
divS =
E E
2
where is the effective conductivity of the
absorbing medium at the eld frequency
and the quantity E E is the time average
of the E2 , therefore

1
divS =
(A5)
E2
2
Hence the rate of energy dissipation
is proportional to the square of the
mean electric eld of the incident radiation and it is the result of the
spectroscopic measurement. The intensity of the E2 is directly related to the
amount of dissipated energy by the absorbing medium.
289
295
4.1
Structure Relationships in Electrochemical

Reactions
C. A. Lucas
University of Liverpool, Liverpool, United
Kingdom
N. M. Markovic
University of California, Berkeley, California
4.1.1
Introduction
In this report, we review recent studies

of single-crystal transition metal electrode
surfaces by combining in situ surface X-ray
diffraction measurements with more traditional electrochemical techniques in order
to probe the inuence of the surface structure on the electrochemical reactivity. This
functionality is generally termed structure
sensitivity. In the last decade, the in situ surface X-ray diffraction technique has been
a critical tool for determining the potential
stability of specic surface structures in
electrolyte under reaction conditions. On
the other hand, the rotating ring disk electrode (RRDE) has been routinely used for
determining the kinetics of electrochemical reactions on single-crystal surfaces and
evaluating the potential-dependent surface
coverage by an adsorbed species. In combination, the X-ray diffraction and RRDE
methods have provided remarkable insight

into the surface electrochemistry and the
structure sensitivity of many important
electrochemical processes.
In this chapter, we review the progress in
the surface science of electrochemical reactions. The methods of the surface X-ray
diffraction and rotating ring disk experiments are described in Sect. 4.1.2, for surface X-ray diffraction the focus being on
the application of the technique to singlecrystal electrodes in the electrochemical
environment. Section 4.1.3 then describes
the nature of the surface structures that are
formed on bare single-crystal transition
metal surfaces in electrolyte. The phenomena of surface reconstruction and surface
relaxation are illustrated by detailed examples, namely Au(001) and Pt(110), in
which the dependence of the surface structure on the applied electrode potential is
described. Section 4.1.4 rst describes the
surface structures and energetics of anion adsorption onto different single-crystal
electrode surfaces, that is, Pt and Au, and
then the underpotential deposition (UPD)
of metals onto the same substrates. Results from the different metal surfaces
are presented in an attempt to understand the inuence of the substrate on
the structure of the overlayer and the effect
of anions on the UPD process. The results highlight the strength of combining
296
2 Interfacial Structure and Kinetics
RRDE measurements with X-ray diffraction results to investigate the detailed

nature of the surface structures that are
formed, particularly in coadsorption studies, for example, the inuence of anion
species on the UPD process. In Sect. 4.1.5,
the oxygen reduction reaction (ORR) is
used as a model electrochemical reaction
to demonstrate the relation between the
metalO2 energetics and reaction pathway/kinetics as well as the importance of
the local symmetry of surface atoms in determining the electrocatalytic properties of
metal surfaces.
4.1.2
Methods
4.1.2.1
X-ray Diffraction
4.1.2.1.1 Theory It is beyond the scope

of this review article to provide a comprehensive description of basic X-ray diffraction from surfaces. Instead, readers are
referred to the excellent reviews by Feidenhansl [1], Fuoss and Brennan [2], and
Robinson and Tweet [3] for explicit details.
In this section, we will focus on some basic ideas that pertain to X-ray diffraction
studies of surfaces in an electrochemical environment, in particular with regard
to crystal-truncation rod (CTR) measurements.
Since the interaction of hard X-rays
with matter is weak, provided that the
exact Bragg conditions for a perfect crystal
are not met, the kinematical scattering
approximation, in which scattering is
treated as a single event, can be used. For
a crystal volume dened by N1 , N2 , and
N3 unit cells along the crystal axes dened
by the vectors a1 , a2 , and a3 , the scattered
intensity can be written as the product of
two scattering amplitudes, one from the
unit cell and one from the lattice of unit
cells. This gives

I (Q) F (Q)F (Q)
sin2 (N1 Qa1 )

sin2 (Qa1 )
sin2 (N2 Qa2 ) sin2 (N3 Qa3 )

sin2 (Qa2 )
sin2 (Qa3 )
(1)
where Q is the momentum transfer,

dened by the scattering angle 2 and
the unit cell structure factor, F , is given by
F (Q) =
n
fj exp(iQrj )
(2)
j =1
where rj is the relative atomic position

in the unit cell and the sum is over all
of the atoms in the unit cell each having
an atomic scattering factor fj (fj is Qdependent and proportional to the atomic
number of the relevant atom). When the
crystal is large, Eq. (1) produces a series
of very sharp peaks in which Qai = . If
the scattering is from a single monolayer
(ML), that is, N3 = 1 (the a3 vector is along
the surface normal), then the scattering
normal to the crystal surface becomes
completely diffuse.
Considering the case of a perfectly
terminated crystal surface and assuming
that the Bragg conditions are met in the
plane of the crystal (i.e. Qx a1 = Qy a2 =
n), then for a semi-innite crystal, Eq. (1)
reduces to

N 2 N 2 F (Q)F (Q)
Qz a3
I (Q) 1 2
sin2
2
2
(3)
The Bragg points of the nite crystal are
connected by streaks of intensity along the
surface normal direction in the reciprocal
lattice. At the minimum position, midway
between Bragg reections, the scattered
intensity is similar to the scattering
expected from a single monolayer. The
streaks of intensity are known as crystaltruncation rods (CTRs) as they arise from
4.1 Structure Relationships in Electrochemical Reactions
the truncation of the crystal lattice at the

surface [3, 4].
Figure 1 illustrates some of the information that can be obtained from the analysis
of CTR data. The intensity variation along
a CTR as a function of the surface normal phase shift, Qa3 = 2l, is shown for
the case of a perfect surface (solid line),
a surface with a reduced density in the
topmost atomic layer (dotted line), and
a surface in which the topmost atomic
layer is expanded away from the bulk crystal (dashed line). When l is an integer,
the Bragg conditions are satised and the
kinematical approximation breaks down.
The minimum of the scattering is observed at the anti-Bragg positions when
l has half-integer values. Surface roughness causes a decrease in the scattering
at the anti-Bragg positions as shown by
the dotted line in the gure, which is a
calculation for a surface with only half of
the atoms in the surface layer present.

Similar effects are observed when adsorbates are formed on crystal surfaces as a
result of the difference in the scattering
factors between the adsorbed atoms and
the substrate atoms. Hence for a monolayer of adsorbed atoms, which have half
of the scattering power of the substrate
atoms, a similar result to the dotted line
in Fig. 1 would be observed. Similarly, an
adlayer with an increased surface density
would cause an increase in the intensity
at the anti-Bragg conditions. For CTRs
measured with an in-plane contribution
to the momentum transfer, the changes
to the CTR prole are dependent on the
relative phase of the scattering from the
substrate and the adlayer, which can cause
either an increase or decrease in the intensity at the anti-Bragg positions. Finally,
a change in the lattice spacing at the surface causes asymmetry in the intensity
Calculated intensity
103
102
10
2p
3p
4p
The intensity variation along a CTR as a function of the

surface normal phase shift, calculated for a perfect surface (solid
line), a surface with a lower density topmost atomic layer (dotted
line) and a surface in which the topmost atomic layer is
expanded by 10% of the layer spacing (dashed line).
Fig. 1
297
298
around the Bragg reections, as shown

by the dashed line in Fig. 1, which is
for a 10% expansion of the surface layer
relative to the bulk. Again, similar effects are produced by adsorbates, as the
adsorbate-substrate layer spacing is very
often different from that of the bulk substrate layer spacing.
By careful modeling of CTR data, it
is possible to extract structural information, such as surface coverage, surface roughness, and layer spacings (both
adsorbatesubstrate distances and the
expansion/contraction of the substrate
surface atoms themselves). By combining

specular CTR results (in which Q is entirely along the surface normal direction)
with nonspecular CTR results (in which Q
has an additional in-plane contribution), it
is possible to build up a three-dimensional
picture of the atomic structure at the electrode surface. If the surface or adlayer
adopts a different symmetry from that of
the underlying bulk crystal lattice, then
the scattering from the surface becomes
separate from that of the bulk in reciprocal space and it is possible to measure
the surface scattering independently. This
(0 1 l)
(1 1 l)
(0 1 l)
(015)
(114)
(0 0 l)
(013)
(1 0 l)
(0 0 l)
(012)
(003)
(0 k 0)
(104)
(004)
(0 k 0)
(1 0 l)
(112)
(011)
(103)
(110)
(002)
60
(a)
(101)
(101)
(h 0 0)
(b)
(h 0 0)
The real-space surface structures and corresponding reciprocal space lattices for:
(a) (111); (b) (001); and (c) (110) surface terminations of an fcc crystal. The unit cells are shown
in the real-space structures and the corresponding Bragg reections indicated in the reciprocal
space lattices.
Fig. 2

Fig. 2
(Continued)
(1 1 l)
(0 1 l)
(0 0 l)
(1 0 l)
(112)
(002)
(011 )
(0 k 0)
(101)
(110)
(c)
(h 0 0)
independent structural information can

be combined with CTR analysis to give
the registry of the surface layers with respect to the bulk lattice [13]. CTR and
surface structure analysis are presented in
Sects. 4.1.3 and 4.1.4 of this report for a
variety of electrochemical surfaces.
Reciprocal Lattices for Low-index
fcc Metal Surfaces This report describes
results from the three low-index surfaces
of metal crystals with the face centered
cubic (fcc) crystal structure. As such, it
is instructive to describe the reciprocal
space lattice for each surface orientation so
that the results from the X-ray diffraction
experiments presented in Sects. 4.1.3 and
4.1.2.1.2
4.1.4 can be quoted without reference

to the experimental details in each case.
Figure 2 shows the reciprocal space lattices
for the (111), (001), and (110) crystal
faces with the corresponding real-space
surface structures shown schematically in
each case. In surface X-ray diffraction
experiments, it is always desirable for the
surface normal direction to be dened as
the (0, 0, l) reciprocal space direction.
This is also the convention adopted for
low-energy electron diffraction (LEED)
experiments.
The close-packed (111) surface is shown
in Fig. 2(a). For this surface, a hexagonal
unit cell is dened, such that the surface
normal is along the (0, 0, l)hex direction and
299
300
the (h, 0, 0)hex and (0, k, 0)hex vectors lie in

the plane of the surface and subtend 60 .
The units for h, k, and l are a = b =
4/ 3aNN and c = 2/ 6aNN where

aNN is the nearest-neighbor distance in the
crystal. Because of the ABC stacking along
the surface normal direction, the unit cell
contains three monolayers and the Bragg
reections are spaced apart by multiples
of three in l. The basic scattering equation
for an unrelaxed, unmodied (111) surface
along the CTRs shown in Fig. 2(a) is then
given by
I (h, k, l)
2

f

1 exp[2i(h/3 k/3 l/3)] (4)
where f is the atomic form factor
(including the DebyeWaller temperature
factor [5]).
The (001) surface (Fig. 2b) is more open
than the (111) surface and is indexed to a
surface tetragonal unit cell. This is related
to the conventional cubic unit cell by the
transformations (1, 0, 0)t = 1/2(2, 2, 0)c ,
(0, 1, 0)t = 1/2(2, 2, 0)c , and (0, 0, 1)t =
(0, 0, 1)c . The units for h, k, and l are
a = b = 2/aNN and c = 4/ 2aNN ,

where aNN is the nearest-neighbor distance
in the crystal. The CTR scattering, by
analogy with Eq. (4), is given by

I (h, k, l)
2

f

1 exp[i(h + k + l)]
(5)
Finally, the (110) reciprocal surface unit
cell is rectangular as shown in Fig. 2(c) and
the reciprocal lattice notation is such that
h is along [1 1 0], k along [0 1 0], and l is
along the [1 1 0] surface normal. The units
for h, k, and l are a = c = 2/aNN and
b = 4/ 2aNN , where aNN is the nearestneighbor distance in the crystal. For this
surface, the CTR scattering reduces to
the form
I (h, k, l)
2

f

1 exp[2i(h/2 + k/2 + l/2)] (6)
Equations (46) are by no means rigorous
but they essentially form the basis for
analyzing CTR data from the three lowindex surfaces of fcc transition metals.
The equations are easily modied to
include surface relaxation effects, surface
roughness, or adsorbed adlayer species
and are used in the analysis of the results
presented in Sects. 4.1.3 and 4.1.4 of
this report.
4.1.2.1.3 Experimental Details The general experimental procedures used in X-ray
diffraction measurements of electrochemical systems have been described in detail
in previous articles [68]. After surface
preparation of the electrode to be studied,
the crystal is transferred (preferably whilst
being protected from the atmosphere by a
drop of electrolyte on the surface) into an
X-ray electrochemical cell. The design of
the electrochemical cell that we have used
in our experiments is shown schematically
in Fig. 3. In this design, the crystal is held
in place by tightening down the central column of the cell, which is screw-threaded
into the main body (all parts are made of
Kel-F material, which can be chemically
cleaned and is relatively easy to machine).
The crystal surface then forms the highest
part of the cell and so is easily accessible to the incident X-ray beam. Electrical
contact to the working electrode and to a
counter and reference electrode are made
through feedthrough connections on the
side of the cell. In our experiments, a palladium/hydrogen reference electrode was
used, which was in direct contact with
the electrolyte in the main body of the

Polypropylene film
Sample
Reference electrode
Electrolyte out
Electrolyte in
Counterelectrode
Working electrode
Fig. 3
A schematic diagram of the electrochemical X-ray scattering cell.
cell. Care has to be taken in this situation as, depending on the nature of the
electrochemical reactions that are being
studied, there can be shifts in the reference potential. Two further connections
form electrolyte feedthroughs to the cell
and the liquid is contained within the cell
by a thin (10 m) polypropylene window
sealed with an o ring. During the X-ray
experiment, the electrochemical cell can be
deated to trap a thin layer of electrolyte
(10 m) between the electrode surface
and the X-ray window to shorten the solution pathlength of the X-ray beam. Note
that this cell design allows the sample to be
oriented in any direction as the electrolyte
is contained, that is, there is no need for
any specialized diffractometer equipment
in order to maintain a horizontal surface
geometry. In order to minimize contamination from the atmosphere, the cell is
surrounded by an outer shell containing a
nitrogen atmosphere.
Most of the X-ray measurements described in this report were carried out
at room temperature on beamline 7-2
at the Stanford Synchrotron Radiation
Laboratory (SSRL) utilizing a focused
monochromatic X-ray beam. Additionally,
some results by other authors are used,

which were obtained on beamlines at the
Brookhaven National Synchrotron Light
Source (NSLS). For experimental details,
readers are referred to the referenced work
in each case. For a general description of
the procedures used in synchrotron surface X-ray diffraction experiments, readers
are referred to the excellent review by Feidenhansl [1]. A typical procedure for any
given electrochemical system used in our
experiments is roughly as follows:
(1) In the X-ray cell, immediately after
mounting in the X-ray goniometer and
being put under potential control, the
cyclic voltammetry (CV) was measured
for comparison with literature results.
After the initial cycling, the electrode
potential was transferred to computer
control for the duration of the X-ray
experiment. Although with time the
features in the CV became distorted
as the electrolyte made contact with
the polycrystalline back and sides of
the crystal, the general features were
always observable for the duration
of the experiment (typically up to a
maximum of 24 h).
301
302
(2) The X-ray spectrometer would be

aligned with two or more Bragg
reections to relate the spectrometer
angles to the reciprocal space lattice
of the sample. Subsequently all data
were measured in units of the sample
reciprocal space (see previous section).
(3) For a particular electrochemical system, the dependence of the X-ray
scattering intensity at various key reciprocal space positions (usually at
points on the CTRs) would be monitored as a function of the applied electrode potential over the potential range
of interest. This potentiodynamic technique, which we have termed as X-ray
voltammetry (XRV) [9], is described in
more detail in Sect. 4.1.2.1.4.
(4) The XRV measurement is used to
identify structural changes that are
occurring at different electrode potentials. These would then be probed with
the electrode potential held at the relevant value. Typically, in searching for
in-plane structures, scans along the
(h, 0, 0), (0, k, 0), and (h, k, 0) inplane reciprocal space directions at
small out-of-plane momentum transfer (l = 0.10.15) were performed.
From these scans, it would be hoped
to locate scattering that had a different symmetry to that of the bulk
CTR scattering. Data for CTR analysis would be obtained by performing
rocking scans at sequential l values
along a particular CTR. The rocking
scans can then be integrated (either
numerically or by tting a functional
form to the lineshape), which results
in the CTR prole plotted as l versus
integrated intensity. This data is suitable for analysis using the equations
given in the previous section, modied by the appropriate instrumental
resolution corrections [10].
4.1.2.1.4 New Techniques Surface X-ray

diffraction is now a well-established technique for probing the atomic structure at
the electrochemical interface and, since
the rst in situ synchrotron X-ray study in
1988 [6], several groups have used the technique to probe a variety of electrochemical
systems. Most analysis has followed the
methodology outlined in Sects. 4.1.2.1.1 to
4.1.2.1.3, whereby structural information
at a xed electrode potential is obtained
by detailed measurement of the scattering from surface structures and modeling
of the CTR proles. In this section, a
couple of recent applications of X-ray scattering are described that are particularly
useful in studies of the electrochemical
interface.
(1) X-ray Voltammetry (XRV): XRV is a
potentiodynamic measurement whereby
the scattered X-ray intensity at a particular
reciprocal lattice point is monitored as the
potential is cycled. As can be seen from
Fig. 1, positions on the CTRs are sensitive
to surface expansion or to adsorption
effects. By measuring the XRV at a number
of different CTR positions, an insight
into the nature of the structural changes
at the surface can be obtained without
recourse to the detailed measurement
of the CTR proles [11, 12]. If scattering
as a result of an ordered surface layer
with a different symmetry to that of
the underlying bulk crystal is obtained,
then the measurement of the potential
dependence of this scattering directly
indicates the potential range of stability
of the structure. This information is
particularly useful for comparison with CV
results as features in the CV can be directly
correlated with structural changes at the
electrode surface. Several examples of this
measurement are presented in Sects. 4.1.3
and 4.1.4 of this report.
(2) Anomalous scattering techniques: In

Sect. 4.1.2.1.1, the atomic form factor, f ,
gives the scattering power of each atom
within a particular unit cell structure.
When the incident X-ray energy is away
from atomic adsorption edges, f is Qdependent and related to the number of
electrons that the atom has. However,
if the X-ray energy is tuned near to an
adsorption edge then f must be modied
by the anomalous dispersion corrections
according to
f (Q, E) = f0 (Q) f (E) if (E) (7)
Close to an adsorption edge f (E) can increase to 10 electrons and so can have
a dramatic effect on the atomic form factor and hence the scattered X-ray signal.
In the complex electrochemical environment, this effect can be extremely helpful
in identifying structural models, especially
in coadsorbate systems. For example, the
contribution to the X-ray scattering from
a particular atom in a structure may be
effectively isolated by tuning the energy
of the incident X-ray beam through an
atomic adsorption edge. This can help
the interpretation of X-ray diffraction data
(examples are given in Sect. 4.1.4.2 for anion/metal coadsorption) and, additionally,
may be used to probe oxidation states
at surfaces and in adsorbates [13]. The
combination of diffraction measurements
with traditional extended X-ray absorption
ne-structure (EXAFS) techniques (socalled diffraction anomalous ne-structure
(DAFS) [14]) may lead to studies in which
the local environment of atoms selected by
the diffraction conditions is probed spectroscopically.
Rotating Ring Disk Methods
A large array of experimental techniques
has been developed to probe the behavior
4.1.2.2
of electrochemical reactions. Although

several experimental methods are now
routinely used in the surface electrochemistry, the rotating disk electrode (RDE)
and rotating ring disk electrode (RRDE)
techniques have become standard tools in
determining kinetic parameters, reaction
products, and potential-dependent surface
coverage by adsorbates [15, 16]. The RDE
is a solid disk embedded in an isolating
material attached to the shaft that rotates about the electrode surface normal.
The NavierStokes equation can be solved
exactly for this system, providing exact
correction for reactant mass transport limitations. As the concentration of dissolved
gasses, viz. H2 , O2 , CO, and so on, in aqueous solutions is relatively low, for example,
105 106 mol cm2 , the mass-transfer
limitation in the hydrogen oxidation reaction (HOR), oxygen-reduction reaction
(ORR), and CO oxidation reaction are very
signicant, and the use of the RDE is
absolutely essential for denitive kinetic
studies. The most useful equation to study
electrode kinetics that are under mixed
control, that is, the case when the activation (iac ) and mass transportcontrolled
(iL = B1/2 ) current densities combine to
yield the total current density as the sum
of reciprocal, is
1
1
1
+
=
i
iac
iL
(8)
where the diffusion limiting current is proportional to angular velocity 1/2 [15, 16].
The success of the RDE method has
stimulated the development of several
other rotating congurations. The RRDE
is perhaps the most useful extension of
the idea of the RDE. The RRDE was rst
developed by Frumkin and Nekrasov to
detect unstable intermediates in electrode
reactions. The RRDE, Fig. 4, consists of
a central disk electrode surrounded by
303
304
The RRDE consisting of a central disk electrode surrounded

by a concentric insulating annulus and a concentric ring electrode.
Fig. 4
a concentric insulating annulus (Teon)

and then a concentric ring electrode.
To operate an RRDE, a bipotentiostat is
needed, which can control the potential
of the disk and the ring independently,
in order to measure the current going
through each of them separately. The
reaction intermediates formed on the disk
electrode can be monitored using the
ring-collection properties. In the ringcollection mode, intermediates produced
on the disk electrode can be collected
by the ring electrode [16]. In the RRDE
conguration, however, only a fraction
of intermediates produced on the disk
electrode will reach the ring electrode. This
fraction is determined by the collection
efciency, N , which is dened as the
ratio between the ring and disk currents,
N = ir /id . In Sect. 4.1.5, the collection
mode of the RRDE is used to monitor
the formation of intermediates during the
ORR on the disk electrode.
The RRDE can also be useful to obtain
accurate thermodynamic data without
interference from kinetic and doublelayer effects. The best example is UPD
of copper in solution containing Br , in
which the amount of Cu UPD and/or
bromide anion deposited/adsorbed on the
disk electrode can be estimated using
the two ring-electrode properties ring
shielding and ring collection [16]. For both
modes, the UPD species is reduced to
metal at the ring, and electroactive species
in solution must undergo convectivediffusion controlled reduction at the ring
electrode. The ring-shielding property of
the RRDE is illustrated in Sect. 4.1.4.2.3 in
which it is used to assess the mass ux of
both the Cu2+ and the Br from and to the
Pt(111) disk electrode. If either Br or Cu
is adsorbed at the Pt(111) disk, the surface
coverages Br,Cu can be assessed from the
ring currents in either potentiodynamic or
potentiostatic experiments:

Br,Cu
1
=
Q
(1/v)
(ir ir ) dE
AnN
(9)

Br,Cu
1
=
Q
(ir(t) ir ) d
AnN
(10)
where ir and ir are unshielded (no

UPD process on the disk electrode) and
shielded ring currents, respectively, N is
the collection efciency, v refers to the
sweep rate, n is the number of electrons
(n = 2 for Cu2+ reduction to Cu0 and
n = 1 for Br oxidation to 1/2 Br2 on
the ring electrode), and Q is the charge
corresponding to monolayer formation
of ad-atoms on Pt(111) based on the
surface atomic density of Pt(111)-(1
1) (1.53 1015 atoms cm2 ), assuming
one completely discharged ad-atom per
platinum atom. (e.g. 208 C cm2 for
Brad and 480 C cm2 for Cuupd . By
combining the powerful in situ X-ray
scattering technique with the elegant
RRDE method, remarkable insight into the
structural nature of the Pt(hkl)-UPD-anion
electrochemical interface can be obtained
as is demonstrated in Sect. 4.1.4.
Surface Preparation
A key aspect in the study of single-crystal
metal electrodes is the preparation of the
surface prior to the experiment and the
transfer of the crystal into the electrochemical environment. One approach is to
prepare the surface in an ultrahigh vacuum
(UHV) environment by cycles of ion sputtering and annealing. This methodology
has the advantage that the surface quality can be checked during preparation by
standard surface science techniques such
as LEED and Auger electron spectroscopy.
UHV preparation methods have been particularly useful in the study of Au(hkl)
electrodes as the surfaces are relatively inert and the crystal can be transferred in
air from the UHV system to the electrochemical cell. An alternative method for
4.1.2.3
the preparation of Au(hkl) is to use the

ame-annealing technique [17]. After pretreatment to produce a at, well-oriented
surface, the crystal is heated in an hydrogen or butane ame and then allowed
to cool in air before transfer to the cell.
This method, or variations of it, can produce clean surfaces as evidenced by the
subsequent observation of the Au surface
reconstructions in electrolyte.
UHV preparation of crystal surfaces
is time consuming and can be difcult
for some metals. Furthermore, for more
reactive metals than Au, it is transfer
of the crystal after surface preparation
that is vital, as it is important to minimize the interaction of the surface with
the atmosphere. The development of the
ame-annealing method by Clavilier and
coworkers [18] has become of critical importance in surface electrochemistry. The
hydrogen ame-annealing method has
been developed to incorporate a cooling
procedure whereby the crystal is cooled in
a hydrogen or argon atmosphere before a
drop of electrolyte is placed on the surface
for transfer to the electrochemical cell. By
using puried gases, the contact with oxygen is minimized during preparation and
this procedure has been used successfully
for the preparation of Pt(hkl) surfaces.
Unfortunately other metal surfaces,
such as Cu, Ag, and Ni, cannot be prepared
by ame-annealing methods. The most
successful method for these surfaces has
been to use a long preanneal of the
surface in a forming gas, such as hydrogen,
followed by a short electrochemical etch,
rinsing with water and then direct transfer
into the electrochemical cell [1921]. The
exact conditions for the etching procedure
vary for different metals and even for
different surface orientations of the same
metal. For the results included in this
report, readers are referred to the reference
305
306
material for precise descriptions of the

surface preparation that was followed in
each case.
4.1.3
Surface Structures
Modern surface crystallographic studies

have shown that on the atomic scale,
most clean metals tend to minimize
their surface energy by two kinds of
surface atom rearrangements relaxation
and reconstruction [2226]. In this review,
the term surface reconstruction applies to
the case in which there is lateral (i.e. in the
surface plane) movement of surface atoms
such that the surface layer has a symmetry
that is different from that of the underlying
bulk of the crystal. Hence, the surface
layer has a two-dimensional unit cell
that is different from the corresponding
two-dimensional unit cell of a layer in
the bulk. The periodicity of the surface
can be dened by Woods notation;
for example, an unreconstructed surface
would be termed as (1 1), whereas if
the surface unit cell size was doubled
in one of the primary vector directions,
it would be termed as (2 1), and so
on. On the other hand, surface relaxation
applies to the case in which the surface
layer is in a (1 1) state but the layer
is displaced along the surface normal
direction from the position expected for
bulk termination of the crystal lattice. In
this section, both surface reconstruction
and surface relaxation effects are described
with specic examples chosen to illustrate
the phenomena as they are observed in the
electrochemical environment.
4.1.3.1
Surface Reconstruction
4.1.3.1.1 Gassolid
Interface Under
UHV conditions, the clean low-index faces
of the fcc metals, Pt, Ir, and Au, have

all been shown to reconstruct under certain conditions of sample temperature and
surface preparation [24]. Given that the
hexagonal (111) lattice planes form the
most dense surfaces they have the least tendency to reconstruct and only the Au(111)
surface exhibits a reconstruction at room
temperature. The reconstruction is rather
complex and involves a small increase in
the surface density (23 surface atoms instead of 22), which leads to a large unit cell,
the (23 3) structure, which is known

as the chevron or herringbone reconstruction [27.] In contrast, the most open
(110) surfaces have a strong tendency to
reconstruct, exhibiting (1 n) periodicities [24]. The (110) surfaces of Pt, Ir, and
Au usually reconstruct to form a (1 2)
or (1 3) periodicity. After some initial
controversy, it is now well established
that the (1 2) structure is characteristic of clean surfaces. These structures are
called missing-row structures since every other row in the topmost atomic layer
is missing. Hence, in the transition from
the (1 1) to the (1 2) state, there is a
50% decrease in the density of the surface layer and signicant mass transport
must occur. Theoretical calculations have
indicated that the (1 2) reconstruction is
stabilized on the 5d transition metal (110)
surfaces due to the participation of the d
electrons in bonding in the solid and the
decrease of the kinetic energy of delocalized electrons at the surface [28]. For Au
and Pt, contaminants seem to play the
dominant role in the appearance of the
(1 3) structure, that is, it seems that the
(1 3) structures obtained in UHV are
impurity-stabilized [29].
Perhaps the most intriguing reconstruction occurs on the (001) surfaces.
For example, under UHV conditions, the
Au(001) surface exhibits a hexagonal reconstruction, that consists of a single,

buckled, slightly distorted overlayer that
is aligned close to the [110] bulk direction and is often referred to as a 5 20
reconstruction although, strictly speaking,
the overlayer is incommensurate [30, 31].
Theoretical studies of the Au(001) surface suggest that the reconstruction forms
because the energy gained from the formation of the more compact reconstructed
top layer is greater than the energy lost
as a result of the loss of registry with
the substrate. The energy difference between the relaxed (1 1) surface and the
reconstructed surface is calculated to be
0.2 eV atom1 [32]. X-ray measurements
of the specular CTR at room temperature
have shown that the reconstructed Au layer
is 1.25 times denser than a bulk Au monolayer and that it is expanded away from
the expected bulk position by 20% of
the bulk lattice spacing [30]. Grazing incidence X-ray studies have shown that the
overlayer is aligned approximately with the
[110] direction although the detailed inplane structure is dependent on the rate
of cooling of the surface [31]. For slowly
cooled samples, the overlayer is almost exclusively rotated by 0.8 from the [110]
direction with a very small fraction of the
domains unrotated. The in-plane surface
diffraction pattern for this structure corresponds to a hexagon of spots centered
around the origin of reciprocal space as
shown in Fig. 5. In the [110] direction,
the peaks are centered at (1.206, 1.206, 0).
The corresponding real-space structure is
shown schematically in the lower part of
the gure.
4.1.3.1.2 Solidsolution Interface Establishing the relationship between a surface
structure in electrolyte and that observed in
UHV has always been problematic, largely
due to the difculty in preparing clean surfaces and transferring them to the electrochemical environment (see Sect. 4.1.2.3).
Ex situ techniques, whereby the electrode
is emersed from solution and transferred
back into UHV for surface analysis, cannot give denitive results due to the loss of
potential control and/or solution species.
The advances in the application of in situ
methods, for example, scanning tunneling
microscopy (STM) [17, 21] and X-ray scattering, have alleviated this emersion gap
problem, and provided denitive structure
determination of metal surfaces in electrolyte and under potential control.
Studies of the reconstruction of (111)
metal surfaces date back to the early
1970s, when it was reported that UHVprepared Pt(111)-(1 1) surface remained
intact after contact with solution [33, 34].
More recently systematic, in situ X-ray scattering [35] and STM [21, 28] studies have
demonstrated that well-ordered P(111)(1 1) structures, prepared either in UHV
or by the ame-annealing procedure, are
stable in aqueous solutions. As for Pt(111),
the (111) surfaces of Ag and Cu exhibit
a (1 1) termination in the absence of
any strongly adsorbing species, such as
anion adsorption or UPD metals. In contrast, the Au(111) single-crystal surface
exhibits a potential-dependent reconstruction that is unique among the fcc metal
surfaces. As in UHV, the surface forms
the (23 3) structure at negative electrode potentials whereas at positive potentials the unreconstructed (1 1) surface
is observed. The driving force for the
(1 1) (23 3) phase transition was

proposed to be due to an induced surface
charge density of 0.07 0.02 electrons per
atom, that is, it was postulated that the
formation of the surface reconstruction is
driven by the induced surface charge [36.]
Alternatively, it has been proposed that
307
308
d = 0.206
(1, 1, 0)c
(1, 0)h
1.206
(0, 0, 0)
(1, 1, 0)c
(0, 1)h
(a)
(b)
1.6
1.6
1.6
g
b
(c)
(d)
Fig. 5
The in-plane diffraction pattern for the

Au(001) surface. (a) The unreconstructed phase;
(b) the distorted hexagonal phase; (c) the
rotated, distorted hexagonal phase. In each case,
the lled symbols represent the scattering from
the underlying Au(001) face with square

symmetry and the open symbols the scattering
due to the hexagonal surface reconstruction.
(d) A real-space schematic picture of a rotated
domain of the reconstruction.
adsorption/desorption of solution species

is responsible for the changes in the surface structure [37]. Considering that the
adsorption of ions from supporting electrolyte is closely related to the charge
redistribution at the interface, it is very
difcult to conclude unambiguously what
the real driving force for the potentialinduced reconstruction of Au(111) is. At
present, we consider the issue to be unresolved; for further details the readers
are refered to Kolbs review on reconstruction phenomena at metalelectrolyte
interfaces [17].
The structural behavior of the (001) surfaces in aqueous electrolytes has been
more controversial. For example, the rst
ex situ experiments indicated that the
Pt(001) 5 20 structure transforms
into the (1 1) phase upon contact with
aqueous solution or even when in contact with water vapor [38]. In contrast,
Zei and coworkers found that the UHVprepared reconstructed Pt(001) surface
remained stable in contact with electrolyte, although no convincing evidence of
potential-induced reconstruction was ever
detected [39]. X-ray scattering measurements have unambiguously demonstrated
that the Pt(001)-(1 1) surface is stable in
the potential range between the adsorption of hydrogen and hydroxyl species and
the reconstructed surface was never observed [40]. In contrast with Pt(001), the
Au(001) surface in electrolyte exhibits a
potential-induced reconstruction that is
very similar to the temperature-induced
reconstruction in UHV [7]. To illustrate
the general phenomenon of surface reconstruction in electrolyte, representative
results for this system are now presented
to demonstrate the power of the X-ray
diffraction technique in elucidating the
potential-induced structural transformation of the Au(001) surface.
The potential stability of the gold surface reconstruction in the electrochemical
environment has been studied by CV,
ex situ emersion LEED experiments [17],
STM, and X-ray diffraction [7]. In both
acid and alkaline solutions, there is good
agreement between the experimental techniques that the 5 20 reconstruction
(see Fig. 5) is formed at cathodic potentials and that it can be reversibly lifted
and formed upon cycling the applied
potential anodically. Figure 6 shows representative in-plane X-ray diffraction results
in the form of rocking scans through
the reciprocal lattice positions indicated

in Fig. 5. Figure 6(c) shows a scan through
the (1, 1, 0.4) position that corresponds
to a point on the nonspecular (1, 1,
l) CTR. This scan is representative of
the in-plane mosaic spread of the crystal surface, which is 0.5 full-width at
half-maximum (FWHM). Figures 6(a, b)
show scans at the (1.206, 1.206, 0.4) position in 0.1 M KOH (0.35 V) and 0.1 M
HClO4 (0.05 V) solutions, respectively;
the electrode potential chosen so that the
surface is in the reconstructed state. In
0.1 M KOH, the scan shows a single broad
peak aligned with the Au(110) direction. In
0.1 M HClO4 , however, there is no longer
a central peak but the scan displays two
peaks rotated by 0.75 from the [110]
direction. These results indicate good general agreement with UHV studies of the
Au(001) surface [30, 31], although the rotated peaks observed in 0.1 M HClO4 are
unaccompanied by the strong central peak
observed in the UHV studies, whereas in
0.1 M KOH no rotated domains are observed at all. We should note that the
results shown in Fig. 6 were obtained after
numerous cycles of the electrode potential
in both electrolytes during which the reconstruction was lifted and reformed and
so the results shown are representative of
true equilibrium conditions.
Figure 7 shows some XRV measurements, which demonstrate the potential
range of stability of the 5 20 reconstruction in the different electrolytes.
Figure 7(a) shows the intensity dependence at the position of one of the rotated
peaks in 0.1 M HClO4 , Fig. 7(b) shows
the intensity dependence at the unrotated
peak position in 0.1 M KOH, and Fig. 7(c)
shows data for the same solution obtained
at (0, 0, 1.3), a position close to the antiBragg position along the specular CTR
309
0.1 M KOH
0.16
0.12
(a)
0.12
0.1 M HClO4
0.11
Intensity
[a.u.]
310
0.10
0.09
(b)
0.30
(1, 1, 0.4)
0.20
0.10
3
(c)
f
[degrees]
Rocking scans through: (a) (1.206, 1.206, 0.4) in 0.1 M KOH (at
E = 0.35 V); (b) (1.206, 1,206, 0.4) in 0.1 M HClO4 (E = 0.05 V);
and (c) (1, 1, 0.4), a CTR position for a Au(001) electrode. These
positions are illustrated in Fig. 5. The closed symbols in (a) show the
measured prole in both solutions after the lifting of the reconstruction
at positive potentials.
Fig. 6
in which the reconstruction of the surface causes an increase in the scattered

intensity due to the increased density of
the surface layer. The results show the
reversible lifting and formation of the reconstruction in both electrolyte solutions.
Because of the slow kinetics in the formation of the reconstructed surface, at

the sweep rate used in these measurements, 2.5 mV s1 , there is considerable
hysteresis as the potential is cycled. Previous kinetic studies [8] have shown that

0.19
0.1 M HClO4
0.18
0.17
0.16
(a)
0.15
0.14
Intensity
[a.u.]
0.1 M KOH
0.12
0.10
(b)
(0, 0, 1.3)
0.40
0.20
0.0
(c)
0.4
0.8
1.2
E
[V]
XRV measurements at: (a) the rotated peak position in 0.1 M HClO4 (see
Fig. 6b); (b) (1.206, 1.206, 0.4) in 0.1 M KOH; and (c) at (0, 0, 1.3), a position on
the specular CTR, in the same solution as (b).
Fig. 7
the reconstruction forms over a timescale

of 15100 s, which is consistent with the
data shown in Fig. 7. The fact that the data
at the CTR position, Fig. 7(c), shows the
same hysteresis implies that the kinetics

are controlled by the large-scale movement
of the Au surface atoms rather than any
local ordering effects. As with the Au(111)
311
surface, it is difcult to propose an unique

driving force for the surface reconstruction
although it is interesting that the details of
the reconstructed state (i.e. rotated vs. unrotated domains) appear to depend on the
electrolyte.
The rst insight into the structure of
the (110) surfaces in electrolyte was also
obtained from ex situ studies [17]. For example, the UHV-prepared Pt(110)-(1 2)
surface was found to be stable upon contact with several solutions, if the potential
cycling was restricted to certain potential
regions. More recently, the stability and
structure of the Pt(110) surface in several electrolytes has been examined by in
situ X-ray diffraction [41, 42]. The following results were obtained in 0.1 M NaOH
solution, chosen so as to avoid any strongly
adsorbing anions. The potential was cycled over the range 0.10.7, V, which
corresponds to the hydrogen adsorption
region (<0.3 V) and the region in which reversible adsorption of OH species occurs.
Over this entire potential range, diffraction peaks consistent with the presence
of a (1 2) reconstruction were observed.

Figures 8(a, b) show longitudinal (along
[0 k 0]) and rocking scans (approximately
along [h 0 0]) through the (0, 1.5, 0.1) superlattice Bragg reection. The solid lines are
ts of a Lorentzian lineshape to the data,
which allow the correlation length of the
(1 2) unit cell to be derived. From these
measurements, and from similar scans
through other superlattice reections, a
correlation length of 350 A along the
[1 0 0] direction and 250 A along the
[0 1 0] direction is calculated, in which
these correspond to the parallel and perpendicular directions to the rows in the
missing-row model of the (1 2) reconstruction. The larger correlation length
corresponds to the limit of the instrumental resolution and so this only puts
a lower limit on the domain size. Interestingly, there is no evidence of any shifts in
the (0, 1.5) or (0, 2.5) in-plane reections
along the [0 1 0] direction, in contrast with
the previously reported measurements of
the UHV-prepared surface [43]. This shift
was linked to the presence of randomly
3.0
2.5
Intensity
[a.u.]
312
2.0
1.5
1.0
1.48
0.4
1.52
(h, 0, 0.1)
(a)
(b)
0.0
0.4
f
[degrees]
The measured X-ray intensity at (0, 1.5, 0.1): (a) along the [0 1 0] direction; and
(b) along the [1 0 0] direction for a Pt(110) electrode in 0.1 M NaOH (E = 0.1 V). The
solid lines are ts of a Lorentzian lineshape to the data.
Fig. 8
distributed single-height steps on the surface. The Lorentzian lineshape does not
perfectly produce the data in the tails of the
Bragg reection (Fig. 8a) and this could be
due to a residual step distribution. It is
important to emphasize that the (1 2)
surface reconstruction was stable over a
wide potential range, for example, even at
1.2 V there is still a strong (110)-(1 2)
diffraction pattern. Upon sweeping the
potential negatively, however, the (1 2)
reconstruction is nally lifted as the surface atoms move to accommodate the
oxide reduction. It should also be noted
that the Pt(110)-(1 2) structure was so
stable in aqueous electrolytes that adsorption of CO on this surface did not
induce the (1 2) (1 1) transition
that is observed in UHV upon adsorption of CO [44]. The high stability of
the Pt(110)-(1 2) structure implies that
the potential-induced mobility of Pt surface atoms at the solidliquid interface
is signicantly reduced compared to the
temperature-induced mobility in UHV.
In contrast with Pt(110), the reconstruction at the Au(110)liquid interface can
be reversibly lifted and reformed, implying that the mobility of gold surface
atoms is appreciable even in contact with
electrolyte. Depending on the nature of
the supporting electrolyte, however, either
the (1 2) or (1 3) reconstruction was
found to be stable at negative potentials. In
particular, Magnussen and coworkers [45]
demonstrated that in perchloric acid the
(1 2) phase was stable at the potential
of zero charge (denoted hereafter as the
pzc). Positive of the pzc, the reconstruction
is lifted and the Au(110)-(1 1) structure
is clearly observed in STM experiments.
The observed structural transformation
is highly reversible, the (1 2) structure
being fully restored when the applied potential is negative of the pzc. Because of
the fact that the phase transition was observed at the pzc, it was proposed that the
structural transformation of the Au(110)
surface was governed by the specic (contact) adsorption of anions. In contrast, in
an X-ray diffraction experiment, a (1 3)
reconstruction was observed in salt solutions (0.1 M NaCl) [46]. As for Au(111),
Ocko and coworkers postulated that the
formation of the surface reconstruction
was driven by the induced negative surface charge. It was also suggested that
the (1 3) reconstruction was stabilized
over the (1 2) structure due to the negative surface charge, by analogy to UHV
results showing that Ag(110) underwent
a (1 1) (1 2) transition after the
adsorption of K on the surface [47]. As
for the other Au surfaces, it is very difcult to resolve unambiguously the driving
force for the potential-induced structural
transformation, especially given that surface charge and anion adsorption are so
strongly correlated.
Surface Relaxation
The relaxation of metal surfaces under
UHV conditions has been well studied
by utilizing LEED I V analysis and a
signicant database now exists [23]. For
unreconstructed low-index single-crystal
surfaces, it is often found that the
outermost layer of atoms is contracted
toward the second layer, compared to the
bulk layer spacing. Adsorption on the clean
surface in UHV, however, usually reverses
this trend and results in the outermost
layer of metal atoms being expanded away
from the second layer. The experimental
data has led to interesting theoretical
work aimed at establishing a fundamental
understanding of surface relaxation [48].
The same concepts can be applied to the
M(hkl)electrolyte interface, and indeed
relaxation has also been observed at the
4.1.3.2
313
314
Ag(hkl), Au(hkl), and Pt(hkl) surfaces in

electrochemical environments [40, 46, 49].
As opposed to UHV systems, in the
electrochemical environment, it is not
possible to perform LEED experiments
and surface relaxation effects can only be
probed using X-ray diffraction techniques.
As shown in Sect. 4.1.2.1, the analysis
of CTR data is relatively simple to
perform and can provide accurate values
of surface relaxation in both the UHV
environment and when the surface is in
contact with the electrolyte. Additionally,
the potential dependence of the surface
relaxation can be probed by performing
XRV measurements at suitable reciprocal
lattice positions (see Sect. 4.1.2.1). The
results can give good insight into the
nature of the interaction between species
that can be adsorbed from the electrolyte
solution and the metal surface.
Toney and coworkers used X-ray scattering to probe the distribution of water
molecules at the Ag(111)electrolyte interface [49, 50]. In addition to observing a
large increase in the density of the water layer adjacent to the Ag surface, they
also observed an electrorestrictive effect
in which the topmost Ag atomic layer
was contracted into the surface (by 0.03 A)

relative to the bulk Ag(111) spacing by stepping from 0.23 to 0.52 V (vs. an Ag/AgCl
reference electrode) in 0.002 M NaF solution. This effect was obtained by modeling
of nonspecular CTR data as the scattering contained no contribution from the
incommensurate water adlayers. In a recent X-ray scattering study [51], Lucas and
coworkers have used similar techniques
to probe the relaxation of the Au(111)
surface in its unreconstructed state. The
results have been correlated with surface
stress measurements and density functional theory (DFT) calculations that show
corresponding changes in the density of
states of the Au surface atoms at the Fermi

level.
Of the three low-index Pt surfaces that
we have studied, the Pt(110)-(1 2) has
the most dramatic dependence of the surface relaxation on the applied electrode
potential [41, 42]. In the following discussion, the surface relaxation effects are
interpreted as being caused by the specic nature of the adsorption sites on the
Pt(110)-(1 2) surface and the strength of
the adsorbatemetal bonding, the two of
course being interrelated. The results were
obtained in alkaline solution (Fig. 9), in
which the surface relaxation is attributed
to the adsorption of Hupd and OHad [41].
To conrm the dominant role of these adsorption processes, the charge associated
with hydrogen adsorption and hydroxyl adsorption is also shown in Fig. 9. As shown
in the previous section, the (1 2) surface
reconstruction was stable over the entire
potential range. However, changes in the
diffracted intensities were observed, particularly along the CTRs. Such changes,
in this case at (0, 0, 1.55) on the specular CTR, are shown in the lower part of
Fig. 9. The changes in intensity indicate a
potential dependence of the relaxation at
the Pt(110) surface. Measurements at (0,
0, 2.45) showed the opposite behavior to
those at (0, 0, 1.55), that is, the curve was
inverted. These results suggest that additional expansion of the surface is caused
both by hydrogen adsorption and hydroxyl
adsorption. As discussed in Sect. 4.1.2.1.1,
relaxation in the surface normal direction
induced by Hupd at 0.05 V can be probed
by modeling CTR data. The specular CTR,
(0, 0, l) and nonspecular CTR, (0, 1, l),
data are shown in Fig. 10. In order to
model the data, the structural parameters
derived by Vlieg and coworkers [52] for the
Pt(110)-(1 2) surface in UHV were used,
allowing the vertical displacements in the
Change density
[c cm2]
200
100
0
10 A
QH
QOH
(a)
Side view
5.2
(0, 0, 1.55)
5.0
X
Intensity
[a.u.]
4.8
4.6
4.4
Top view
4.2
0.0
(b)
0.2
0.4
0.6
0.8
1.0
E
[V]
Left: A schematic illustration of the Pt(110)-(1 2) surface showing the

directions of the atomic relaxations derived from ts to the X-ray scattering data. In
the top view, the letter X indicates the most likely site for hydrogen adsorption.
Right: (a) The CV of a Pt(110) disk electrode measured in an RRDE conguration
(20 mV s1 ). The symbols correspond to the charge associated with hydrogen and
hydroxyl adsorption. (b) The XRV measured at (0, 0, 1.55) for the positive
(symbols) and negative (dashed line) sweep directions (2 mV s1 ).
Fig. 9
top two Pt layers, the buckling in the third

layer and an increased DebyeWaller factor (or roughness) of the rst and second
layers to vary. The best t to the data (obtained by a least-squares method) is shown
by the solid lines in Fig. 10. The dashed
line represents a calculation of the CTRs
according to the structural model proposed
by Vlieg and coworkers [52]. The biggest
discrepancy between the calculation for
the model derived in previous work and
the measured data occurs at the anti-Bragg
position on the specular CTR. It was only
possible to reproduce the data by including
an expansion of the topmost layer spacing.
The magnitude of the relaxations derived

from the best t are as follows: 0.35 A expansion in the topmost layer (25% of the
layer spacing), an inward displacement of
0.08 A in the second layer (6%), a buckling
of 0.08 A in the third layer, and root-meansquare roughness of 0.41 and 0.16 A in the
rst and second layers, respectively (the
The
bulk thermal roughness is 0.05 A).
relaxation directions are indicated in the
schematic model of the structure in Fig. 9.
The most striking difference between
the structural model for the Pt(110)(1 2) reconstruction in electrolyte and
the model derived from UHV studies is
315
Intensity
[a.u.]
101
102
(a)
103
101
Intensity
[a.u.]
316
102
103
0.5
(b)
1.0
1.5
2.0
2.5
l
[rlu]
The measured CTR data: (a) (0, 0, l); and (b) (1, 0, l) for the
Pt(110) electrode at an electrode potential of 0.1 V. The dashed lines are
calculated for a contracted (1 2) missing-row model that was found in
UHV studies [52] and the solid lines are ts to the data indicating a 20%
expansion of the surface [41].
Fig. 10
in the relaxation of the topmost Pt atom

in the unit cell. The contraction of the
surface observed in vacuum studies has
been explained on the basis of electrostatic
considerations in which electrons are
transferred from the top atom towards the
missing row with a resultant electrostatic

force that pulls the ion cores into the
surface [53]. In electrolyte, the situation
is inherently more complex because of the
presence of water, ions, and the electric
eld at the interface. In addition, the data
described above were taken in a potential

region in which 1 monolayer of hydrogen
is adsorbed onto the surface. Theoretical
studies of hydrogen adsorption onto Pt
surfaces have indicated that the most
likely site for adsorption is the threefold
coordinated site (indicated by the letter X
in the top view of the surface in Fig. 9),
with the d orbitals of the neighboring
Pt atoms making a major contribution
to the bonding. Full occupation of these
sites agrees with the hydrogen coverage
measured electrochemically. Interestingly,
adsorption of hydrogen into these sites
was proposed as the mechanism for
an increased corrugation of the Pt(110)
surface in UHV, in which a 20% expansion
was observed by He diffraction [54].
4.1.4
Adsorbate-modied Surfaces
It is well established that the adsorption

of atoms and/or molecules on clean
metal surfaces has a dramatic effect on
the surface structure. The presence of
the adsorbed layer usually alters the
metal surface relaxation (see previous
section) and can cause clean metal surfaces
either to reconstruct or, more commonly,
to deconstruct, that is, return from a
reconstructed state to the (1 1) phase.
The thermodynamic driving force for
adsorbate-induced restructuring is the
formation of strong adsorbatesubstrate
bonds that are comparable to or stronger
than the bonds between the substrate
atoms. Within the framework of this
article, we present a selected review
of some recent studies of adsorption
phenomena on single-crystal transition
metals. The section is split into two parts:
the rst dealing with the adsorption of
anions, in particular bromide, and the
second describes metal UPD and gives
selected examples of coadsorption between

metal UPD layers and anions.
Anion Adsorption
The adsorption of anions on metal electrodes has been one of the major topics
in surface electrochemistry. Anion adsorption has an important, and generally
adverse, effect on the kinetics of electrochemical reactions and thus needs to be
understood in some detail. Most of the
recent progress has come from the combination of conventional electrochemical
methods with ex situ and in situ surfacesensitive probes. Of particular interest are
chemisorbed (also called contact adsorbed
or specically adsorbed) anions, whose adsorption is controlled by both electronic
and chemical forces. In this report, we aim
to highlight some of the general themes of
this work by focusing on the adsorption of
bromide anions onto Au and Pt electrodes.
For the Au(001) and Au(111) electrodes,
the focus is on the structural transitions
that are observed as a result of atomic size
effects. For Pt(111) and Pt(001), RRDE data
are combined with X-ray scattering measurements to correlate structural changes
with the potential-dependent surface coverage of the anion species.
4.1.4.1
4.1.4.1.1 Au(111)/Br X-ray

diffraction
and STM studies of the adsorption
of bromide (and other halides) onto
Au single-crystal electrodes have been
reported in a series of papers from the
Brookhaven group [5557]. Figure 11(a)
shows a CV for a Au(111) electrode
in deaerated 0.1 M HClO4 + 0.1 M NaBr
solution. As discussed by Kolb [9],
the peaks at 0.12 V (anodic sweep)
and 0.18 V (cathodic sweep) are
related to the lifting/forming of the
Au surface reconstruction, which is
317
3
E0
Current
[A]
2
1
0
1
2
3
0.4 0.2
0.0
0.2
0.4
0.6
0.8
(0, 0)
(0, 1)
(i) (1, 1) (X, X)
(ii)
(1, 0) (3X, 0)
<T, 1>
EAg/AgCl
[V]
(a)
<1, 1>
(b)
4.25
4.20
BrBr spacing OBr

[]
318
4.15
4.10
4.05
(c)
4.00
(a) The CV for a Au electrode in 0.1 M HClO4 + 0.1 M NaBr. (b) The
in-plane X-ray diffraction pattern observed at potentials positive of E0 .
Positions A, B, and C correspond to the rst-, second-, and third-order
diffraction peaks from the bromide adlattice, respectively (lled circles). Also
shown is a schematic illustration of the corresponding bromide structure.
(c) The bromidebromide nearest-neighbor spacing (obtained from the X-ray
measurements) as a function of the applied potential and solution
concentration; 0.1 M (circles), 0.033 M (squares), 0.01 M (triangles), and
0.001 M (diamonds). The lines are polynomial ts to the data. (This gure is
taken from Ref. [55].)
Fig. 11
induced by a small amount of bromide

adsorption/desorption. The major change
in the bromide surface coverage occurs
in the potential range 0.2 to +0.2 V,
that is, from the reconstruction peak
across the broad shoulder feature. At
higher potentials, there is a small
current associated with additional specic
adsorption of bromide. At the position
marked by E0 (0.42 V), there is a
small reversible peak, which is correlated
with an orderdisorder transition in
the bromide adlayer. The ordering in
the bromide adlayer was probed by
X-ray diffraction measurements [55]. At
potentials above 0.2 V, the Au surface is
unreconstructed, however, no scattering
from an ordered bromide structure was
detected when the potential was below
E0 . When the potential was raised above
E0 , the in-plane diffraction pattern shown
schematically in Fig. 11(b) was obtained.
This pattern exhibits six hexagonally
arranged pairs of diffraction peaks at the
rst-, second-, and third-order positions
(marked by A, B, and C, respectively,
in the gure). Rocking scans at these
positions showed the presence of two
rotated peaks (i.e. equidistant from the
principal Au reciprocal lattice vector, as for
the surface reconstruction of Au(001) in
alkaline electrolyte Sect. 4.1.3.1.2). This
diffraction pattern indicates the presence
of a rotated-hexagonal adlayer structure,
shown schematically in Fig. 11(b).
The rotation angle and the radial peak
positions for the bromide structure depend
on the applied electrode potential over the
range of which the structure was stable. In
fact, the rotation angle ranged from 3.2
to 4.7 , whereas the position of the rstorder peak varied from 0.683 to 0.716 [in
units of Au(1,0,0)] as the potential was
changed from 0.42 to 0.76 V. The change
in the radial peak position can be correlated
with the change in the bromidebromide

nearest-neighbor spacing that is shown
in Fig. 11(c) for four different solution
concentrations of NaBr. At the potential at
which the adlayer is formed (0.42 V), the
BrBr spacing is 4.24 A and it decreases
down to a saturation value of 4.03 A as
the potential is increased. This saturation
value is close to the van der Waals diameter
The fact that the
of bromide (3.704.00 A).
electrocompression is continuous and no
discontinuities are observed in Fig. 11(c)
implies that the adlayer does not lock-in to
any high-order commensurate structures
as this would give rise to a xed BrBr
spacing in a certain potential range. Fits
to the 0.1 M NaBr data to a third-order
polynomial function are shown by the
solid line in Fig. 11(c). The dashed lines
were obtained by successive potential
shifts of 65 mV per logarithm of NaBr
concentration. Such a shift is consistent
with the 59 mV shift expected for a
nominal charge transfer of one electron
from the bromide to the Au electrode, that
is, the electrosorption valency, = 1.
4.1.4.1.2 Au(001)/Br Unlike the continuous electrocompression of the bromide
adlayer observed on Au(111), the bromide adlayer on Au(001), a substrate
with square symmetry, has been observed to undergo a commensurateincommensurate transition in which a
commensurate c( 2 2)R45 structure

transforms continuously to an incommen
surate c( 2 p)R45 structure [57]. The
CV for Au(001) in 0.05 M NaBr is shown
in Fig. 12. Three sharp peaks (labeled as
P1, P2, and P3) are observed as the potential is swept between 0.25 and 0.6 V;
P1 corresponds to the lifting of the Au
reconstruction to leave the surface in the
(1 1) state, P2 corresponds to the forma
tion of the c( 2 2)R45 structure (as
319

15
0.10
P1
0.08
10
c(2 22)
0.04
5
0.02
P2
Intensity
[a.u.]
0.06
c(2 p)
Current
[A cm2]
320
P3
0.00
0
P3
P2
5
0.3
0.0
0.3
0.6
ESCE
[V]
The CV for a Au(001) electrode in 0.05 M NaBr solution and the
corresponding X-ray intensity at the position (0, 1, 0.1) in which scattering from a
bromide adlayer was observed. (Taken from Ref. [57].)
Fig. 12
K
(2, 2)
(c)
pa
22a
2a
(a)
(2, 2)
2a
(b)
(d)
Fig. 13 Real-space model (a, b) of the commensurate c( 2 2)R45 and
incommensurate c( 2 p)R45 bromide structures observed on Au(001). The

open circles correspond to bromide atoms. (c, d) show the corresponding
in-plane reciprocal space patterns in which the squares are scattering from the
Au substrate and the circles are Br reections. In (d), the peaks move outward
along K with increasing potential. (Taken from Ref. [57].)
shown in Fig. 12 by the intensity changes

at the (1, 0, 0.1) reciprocal space position which is where scattering from such a
structure would arise), and P3 corresponds
to the commensurate-incommensurate
phase transition. The c( 2 2)R45

structure is shown schematically in
Fig. 13(a), which indicates that the surface
coverage by bromide is = 1/2 and that
the structure is close to an hexagonal arrangement, despite the square symmetry
of the underlying substrate. This implies
that the elastic interactions between the
relatively large Br ad-atoms (which would
favor hexagonal packing) dominate over
the adsorbatesubstrate interaction. The
observed in-plane diffraction pattern is
shown in Fig. 13(c) in which the squares
correspond to substrate reections and the
circles to Br reections. A similar Br structure was also observed for vapor-deposited
Br on Au(001) [58].
At potentials positive of P3 (0.38 V), the
bromide adlayer undergoes a commensurate-incommensurate transition that was
signied by the movement of the loworder diffraction features continuously and
uniaxially outward with increasing potential. As shown in the diffraction pattern
(Fig. 13d), reections were only observed
at (1, 0.5 + /2), (0, 1 + ), and (2, 1 + ),
along with symmetry equivalents ( denotes the incommensurability). The realspace model of the structure is shown in
Fig. 13(b). For the domain shown, the bromide lattice is commensurate along H (the
2 direction) and incommensurate along

K. As the potential increases above 0.38 V,
the value of increases continuously and
this corresponds to movement of the bromide atoms along the directions shown
by the dashed lines in Fig. 13(b). In the
incommensurate phase, the Br nearest
neighbor spacing is xed at 4.078 A,
whereas the next-nearest-neighbor spacing
decreases from 4.56 A ( = 0, commensurate) to 4.14 A ( = 0.13). This latter

spacing is close to the minimum value
observed on the Au(111) surface (4.02 A).

Pt(111)/Br In contrast to anion
adsorption onto gold electrodes, where a
stable double-layer region occurs over a
wide potential range, on platinum surfaces
there is always strong competition between
anions and either H or OH adsorption.
This competition tends to inhibit the
formation of ordered anion structures and
can make the adlayers difcult to observe
by X-ray scattering measurements. The
interaction of bromide ions with Pt(hkl)
surfaces has been studied extensively at the
solidelectrolyte interface, perhaps even
more so than at the vacuum interface.
The modern in situ structural probes of
STM and X-ray diffraction have played a
signicant role in understanding bromide
ion adsorption on Pt(hkl) surfaces. What
we will emphasize in this review is the
importance of the surface geometry in
the Pt-bromide energetics, and thus in the
ordering of the adlayers.
Quantitative measurements of the coverage by Brad on Pt(111) surface were obtained by purely electrochemical methods,
as described in detail in Ref. [59]. Briey,
by utilizing the ring-shielding properties
of the RRDE, Sect. 4.1.2.2, it was possible to determine the potential-dependent
surface coverage by bromide and its electrosorption valence ( ) on Pt(111). The
electrosorption valence
4.1.4.1.3
id
(ir irfs )/N
= 0.99
(11)
is a measure of the degree of discharge of

the ion upon adsorption, that is = 1 for
a univalent ion corresponds to complete
discharge. Figure 14 shows that as the
321
Pt(111) + 8.105 M Br
Disk current
[A]
(900 rpm: 50 mV/s: 0.1 M HClO4)
0.4
q Br
[ML]
10
(a)
0.3
0.2
0.1
0.0
0.2 0.0
(c)
Ring current
[A]
322
Br
0.2
0.4
E
[V]
+1.08 V Br + e
2
ir
1
0.2
(b)
0.0
0.2
0.4
E
[V]
RRDE results for a Pt(111) electrode in 0.1 M

HClO4 + 104 M Br : (a) Disk current; (b) corresponding ring
current; and (c) the potential-dependent bromide coverage
obtained from the results using Eq. (9).
Fig. 14
Pt(111) disk potential is swept across the

rst voltammetric peak, the associated adsorption of bromide is demonstrated by the
concomitant decrease of the ring current
below its unshielded value. The same observation is made for the subsequent, more
positive peak, which clearly relates this process to further bromide adsorption on the
Pt(111) disk. Following these two characteristic voltammetric peaks, the Pt(111)
disk current diminishes to a double-layerlike structure above 0.2 V. At the same
time, the ring current remains below ir
until the positive potential limit is reached,
manifesting the continuous adsorption of

bromide on the Pt(111) disk. The qualitative correspondence between ring and disk
currents may be evaluated quantitatively in
terms of Br according to Eq. (9); to avoid
the mass transport resistances addressed
in Sect. 4.1.2.2, the bromide adsorption
isotherm is extracted from the negativegoing sweep. The resulting %Br versus E
is shown in Fig. 14(c). As evident from
the shape of curve, the adsorption of Br
(Br 1) is a two-step process, the rst
being displacement of the Hupd state in
a narrow interval of potential, the second
a continuous compression until a closepacked adlayer with a maximum surface

coverage of 0.44 Br per Pt surface atom
is reached at 0.6 V. As demonstrated below, combining the purely electrochemical
measurements of coverage with the structures observed by X-ray diffraction enables
a reasonable physical picture of the adsorption process to emerge.
In a preliminary report of the results for bromide adsorption onto Pt(111),
it was shown that an incommensurate
(3 3) hexagonal bromide adlayer is
present on the Pt(111) surface in the potential range 0.050.7 V [60]. From ts
to the scattering prole at the lowestorder diffraction peak, the peak position
as a function of the electrode potential

indicated that the hexagonal adlayer underwent a continuous electrocompression.
By repeating the Br /Pt(111) experiments
a number of times, however, it became
apparent that the behavior of this structure was more complex than originally
proposed [61]. Figure 15 is typical of the
results obtained, showing scans along (h,
0, 0.1) through the rst-order diffraction
peak from the hexagonal bromide overlayer at different electrode potentials. At
0.2 V, there is a very strong peak due to
the (3 3) bromide adlayer located exactly
at the commensurate (3 3) position (do and an additional
main size, D 100 A)
weak peak at slightly higher wave vector
0.05
0.65 V
0.55 V
0.04
0.35 V
Intensity
[a.u.]
0.03
0.25 V
0.02
0.01
X-ray diffraction scans along (h,

0, 0.1) for Br /Pt(111) at different
electrode potentials. The data were
obtained by subtraction of a background
scan performed at 0.2 V. The solid
lines are ts to the data and the dashed
lines mark the positions h = 0.67 and
h = 0.71.
0.2 V
Fig. 15
0.00
0.66
0.68
0.70
(h, 0, 0.1)
0.72
323
324
(h = 0.68). Upon stepping the electrode

potential to 0.25 V, the intensity distribution shows only a very weak double peak
lineshape. At 0.35 V, the scattered intensity is further reduced and it is almost
impossible to distinguish any measurable
feature from the background scattering.
Upon stepping to higher potential, a weak
peak gradually emerged at h = 0.715, becoming stronger as the potential was raised
to a positive potential limit of 0.65 V. In
Fig. 15, rocking scans through the peak, at
any potential, showed the intensity to be
centered along the Pt1, 0, 0 lattice vector
and there was no evidence of any rotation of the adlayer. Stepping the potential
to 0.2 V always caused the peak to disappear, consistent with the voltammetry
results, which indicate complete desorption of bromide at this potential (see the
isotherm in Fig. 14).
The isotherm for bromide adsorption
(Fig. 14c) shows that increasing the potential in the double-layer region causes a
continuous increase in the bromide coverage. The data shown in Fig. 15 and in
other experiments, however, are not consistent with the notion that the additionally
adsorbed bromide leads to a continuous
compression of the incommensurate bromide adlayer, as this would cause the sharp
peak at h = 0.67 to shift gradually to higher
wave vector. An alternative explanation of
our results is that bromide forms a series
of high-order commensurate (HOC) structures on the Pt(111) surface, that is, at all
potentials the structure corresponds to a
close-packed monolayer but that the periodicity depends critically on the ratio of
the bromide and Pt lattice parameters. At
0.2 V, the unit cell corresponds to a (3 3)
structure with a basis of four Br atoms. At
0.6 V, the unit cell is a (7 7) structure
containing 25 Br atoms, that is, with a 7 : 5
ratio of the Br and Pt lattice parameters.
Both of these unit cells are relatively small

(8.33 and 19.39 A for the (3 3) and (7 7)
structures, respectively) and so the adlayer is well ordered. In between these two
phases, we note that HOC structures with
intermediate BrBr atomic spacings must
[62].
have much larger unit cells (>30 A)
Given that the maximum domain size that
this could explain
is observed is 60 A,
the decreased peak intensity (or disorder)
in the intermediate potential region as the
Br ad-atoms do not have enough mobility
to form well-ordered structures with large
unit cells.
4.1.4.1.4 Pt(001)/Br System As for the
Pt(111)-Br system, the surface structure
and potential-dependent surface coverages
by Br on Pt(001) were obtained from the
combination of X-ray scattering and RRDE
measurements [63]. The effect of adsorbed
bromide on the voltammetric features of
Pt(001) recorded in 0.1 M HClO4 is easily
observed by comparing the voltammetry
with and without Br , as shown in Fig. 16.
Close inspection of the voltammetry in
Fig. 16(a) reveals that formation of a monolayer of hydrogen in the solution containing bromide occurs through two distinctive
voltammetric features: a main sharp peak
at 0.1 V and a small peak at 0.175 V.
These two peaks correspond to simultaneous hydrogen adsorption/desorption
with bromide desorption/adsorption on
(001) terrace sites and most probably on
(001) (111) terrace-step sites, respectively [63]. Further inspection of the RRDE
results reveals that, the major adsorption
peak on Pt(001) is not followed with a
second sharp peak, as was found for the
adsorption of Br on Pt(111). The qualitative correspondence between ring and disk
currents may be evaluated quantitatively
in terms of Br as shown in the previous
discussion for Br/Pt(111). The resulting
Disk current
[A]
Pt(001) + 104 M Br
(900 rpm; 50 mV/s; 0.1 M HClO4)
[C cm2]
300
Qd
200
Qr/N
20
(c)
100
(a)
Ring current
[A]
2
Br
+1.08 V
0.2
Br2 + e
0.0
0.2
0.4
E
[V]
i
r
0.2
(b)
0.0
0.2
0.4
E
[V]
RRDE results for a Pt(001) electrode in 0.1M HClO4 + 104 M

Br : (a) Disk current; (b) corresponding ring current; and (c) the
potential-dependent charge due to bromide adsorption obtained from
(a) and (b).
Fig. 16
bromide adsorption isotherm (charge) versus disk potential is shown in Fig. 16(c)
yielding a maximum coverage of 0.42 ML
at the positive potential limit of 0.5 V.
In the X-ray scattering experiments, no
superlattice peaks were found at any potential for Br/Pt(001) in agreement with
the STM results of Bittner and coworkers [64]. Despite the lack of adsorbate
structures with long-range order in the
surface plane, information about the surface normal structure (the PtBr spacing)
and local bonding sites for specically
adsorbed anions can be obtained from
analysis of the specular and nonspecular
CTRs [63]. Figures 17(a, b) show the specular (0, 0, l) CTR and nonspecular (1, 0, l)
CTR measured at an electrode potential of
0.2 V, where no bromide (or based on the
adsorption isotherm for Br , a negligible
325
Intensity
[a.u.]
10
0.1
(a)
(b)
1.4
Ratio I0.2 V/I0.15 V
326
1.2
1.0
0.8
0.6
1
(c)
(0, 0, l)
1
(d)
(1, 0, l)
CTR data for the Br /Pt(001) system: (a) (0, 0, l); and
(b) (1, 0, l) measured at 0.2 V where no Br is adsorbed on the
surface. The solid lines are a t to the data including relaxation
of the Pt surface. (c, d) show the changes in intensity measured
at 0.2 V as referenced to the data at 0.2 V. The ts to the data
are described in the text.
Fig. 17
amount) is expected to be present on the

surface, but the coverage by hydrogen is
essentially unity. The solid lines represent
a simultaneous t to the data in which the
layer spacing between the rst and second
Pt layers and the occupancy and roughness (in the form of a static DebyeWaller
factor) of the topmost Pt layer were varied.
The best t gave an occupancy of 0.81 in the
surface Pt layer, a root-mean-square (r.m.s)
roughness, = 0.09 A and a surface expansion of the Pt lattice by 1.4% of the bulk
Figures 17(c, d)
lattice spacing (0.03 A).
show the changes in the CTR data that
occur upon the initial stage of bromide
adsorption at ca. 0.2 V, displayed as a
ratio of intensities along the CTR at 0.2 V

(Figs. 17(a, b)) and 0.2 V. Systematic
changes in the data are apparent, which
must be accounted for by changes in the
Pt surface expansion and/or the presence
of the adsorbed Brads . In calculating a t
to the ratio data, the total Br coverage was
xed to 0.35 (to be consistent with the
RRDE results) and the partial Brads occupation of the three low-energy adsorption
sites (hollow, bridge, and on-top) was varied. The only other parameters were the
BrPt vertical separation and the Pt surface expansion. The best t is shown by the
solid lines in Figs. 17(c, d). The expansion
of the topmost Pt layer is reduced to 1% and
the occupations (and heights) of bromide

in the hollow, bridge, and on-top sites are
0.08 (2.0 A),
and 0.09 (2.2 A),
0.17 (1.8 A),

respectively, with an r.m.s roughness of
With such a limited data set there
0.19 A.
is obviously some ambiguity involved in
assigning occupancies to the adsorption
sites. However, a t to the specular CTR
data with Brads located at a single height
above the surface yielded an unreasonably
and this
large r.m.s roughness ( = 0.7 A)
suggests a distribution of heights consistent with the different adsorption sites.
The covalent radii of Pt and Brads are 1.14
respectively, which gives caland 1.30 A,
culated vertical separations of 1.60, 2.08,
and 2.44 A for the hollow, bridge, and ontop sites, respectively. These values are in
good agreement with the CTR results and
certainly indicate that the PtBr bond is
covalent in nature.
Summary of Anion Adsorption on
Au(hkl) and Pt(hkl) Surfaces As evident
from the above discussion, the differences
between the different anion structures are
a consequence of two fundamental properties: the strength of the metalanion
interaction, and the symmetry of the anions with respect to the atomic geometry
of the surface. The adsorption of bromide
onto Au(111) leads to the formation of
a rotated, hexagonal incommensurate adlayer, which undergoes compression and
rotation as the potential is increased [55].
Bromide adsorption onto Pt(111) also leads
to the formation of a hexagonal, incommensurate adlayer, however, the adlayer
is aligned with the Pt lattice at all potentials and appears to form a series of
HOC phases rather than exhibit continuous electrocompression [61]. In relation
to the presence of HOC phases on the
Pt(111) surface, in Sect. 4.1.4.2, we show
4.1.4.1.5
that copper UPD onto Pt(111) in the presence of halide anions also causes the
formation of an aligned, incommensurate
hexagonal CuBr (or CuCl) bilayer structure prior to formation of a full copper
monolayer.
The absence of any ordered structures
for the Brad adlayer on Pt(001) contrasts
with the ordered structures observed on
Au(001), Au(111), and Pt(111). The ordered Brads adlayer structures observed
on Pt(111) and Au(111) are incommensurate, which means that there is no
preference for any particular site. However, the energy minimum for adsorption
at the hollow sites on (001) surfaces is
considerably deeper than for the hollow sites on the (111) surfaces. This
explains the appearance of the com
mensurate c( 2 2)R45 phase on the

Au(001) surface, which is observed until increased bromide coverage drives the
commensurateincommensurate transition [57]. Initial adsorption of bromide into
random hollow sites on Pt(001), therefore, seems likely, up to a maximum
coverage determined by the lateral repulsion between bromide anions. Given
that the secondnear-neighbor distance
for hollow sites is approximately 3.9 A,

this maximum coverage is quite low and
further adsorption probably leads to occupation of other sites, as suggested by the
CTR results. As the potential is increased,
the PtBr bond becomes stronger, thus
further reducing the mobility of the adsorbed atoms. This may be the mechanism
that prevents the formation of any Brads
structure with long-range order on the
Pt(001) surface.
Underpotential Deposition
There are many different methods for producing bimetallic catalytic surfaces. Until
recently, UHV-prepared surfaces had an
4.1.4.2
327
328
enormous advantage over the surfaces prepared by electrochemical methods because

an UHV system equipped with modern
surface-sensitive techniques provided microscopic structural information in a relatively direct fashion. Recently, however,
the emergence of surface characterization
techniques such as X-ray scattering and
STM/atomic force microscopy (AFM), operating under electrochemical conditions,
has allowed the electrodeposition method
to become equally important in the synthesis of model bimetallic structures. In
this section, we will focus on bimetallic
Au and Pt single-crystal surfaces produced
by the UPD solution phase method. UPD
corresponds to the electrochemical adsorption, often of one monolayer, that occurs at
electrode potentials positive of the Nernst
potential below which bulk metal adsorption occurs [16]. Numerous experiments
have shown that the UPD layer can dramatically alter the chemical and electronic
properties of the interface. The UPD layer
is also the rst stage of bulk metal deposition and its structure, therefore, can
strongly inuence the structure of the
bulk deposit. Early studies of UPD using polycrystalline substrates have, more
recently, been extended to single-crystal
substrates, which not only allows the role
of surface atomic structure to be explored,
but also permits the study of the interface
structures by diffraction-based techniques,
such as surface X-ray scattering. Quite a
few systems have now been studied using
this technique and this has led to a greater
understanding of the physics determining
the structure of the UPD layer, in particular with regards to the role of the electrode
potential and of various other adsorbing
species that can be present in solution.
Many relevant papers in UPD studies
are referenced in the review articles by
Kolb [65] and Adzic [66]. For our purposes
here and to follow on from Sect. 4.1.4.1,

we focus on the inuence of anions on Cu
UPD onto Au(111), Pt(111), and Pt(001).
Cu UPD on Au(111) in Sulfuric
Acid Solution Cu UPD onto Au(111) in
sulfuric acid has been an archetypal system
for the study of anion coadsorption during
metal UPD. The UPD of Cu occurs in
two stages and there has been good
agreement that in the second stage a full
monolayer of Cu is formed. The rst stage,
however, has been somewhat controversial
and a unique structural model was not
conrmed until a detailed X-ray diffraction
study by Toney and coworkers [67]. This is
partly because the Cu2+ ion is not fully
discharged to neutral Cu0 but carries an
unknown charge that makes the surface
coverage by Cu difcult to determine by
cyclic voltammetry. The X-ray scattering
experiment followed on from an ex situ
LEED and Auger spectroscopy study in
which it was proposed that the Cu atoms
formed a honeycomb structure with ( 3
3)R30 symmetry and a surface coverage

of 2/3 ML per Au surface atom [68].
Toney and coworkers conrmed these
results by performing a detailed surface
X-ray diffraction study of the structure
involving the measurement of the inplane structure factors, the l dependence
of some of the fractional-order reections
and the bulk CTRs passing through the Au
Bragg reections [67]. Figure 18(a) shows
the in-plane diffraction pattern for the
( 3 3)R30 structure together with

some of the fractional-order rod data. The
data in Fig. 18 immediately ruled out the
( 3 3)R30 structure proposed from

STM and AFM experiments (a triangular
Cu adlayer) [69] as such a at adlayer
would give rise to fractional-order proles
that were far more intense and fall off
slowly with L. The data in Fig. 18 is
4.1.4.2.1

(1/3 1/3) & (4/3 4/3) rods
(02)
1000
( 23 43 )
(01)
(11)
(11)
( 23 23 )
( 13 23 )
1 1
( 43 13 )
(3 3)
(00)
(10)
(1/3 1/3) rod, data

(1/3 1/3) rod, flt
(4/3 4/3) rod, data
(4/3 4/3) rod, flt
800
( 43 43 )
F2
[eu]
( 23 53 )
(20)
600
400
200
(23 13 )
( 43 23 )
(a)
(2/3 4/3) & (2/3 5/3) rods
300
500
F2
[eu]
F2
[eu]
200
(4/3 1/3) rod, data

(4/3 1/3) rod, flt
600
400
300
200
100
100
0
0
(c)
(4/3 1/3) rods

700
(2/3 4/3) rod, data

(2/3 4/3) rod, flt
(2/3 5/3) rod, data
(2/3 5/3) rod, flt
400
L
[r/u]
(b)
500
L
[r/u]
0.5
(d)
1.0
1.5
2.0
2.5
3.0
L
[r/u]
Fig. 18
X-ray diffraction data for Cu/Au(111) in

sulfuric acid solution at the rst stage of Cu
deposition: (a) In-plane diffraction pattern;
(b) (1/3, 1/3, l) and (4/3, 4/3, l); (c) (2/3, 4/3,
l) and (2/3, 5/3, l); (d) (4/3, 1.3, l). The solid lines
are ts to the data according to the structural
model shown in Fig. 19. (Taken from Ref. [67].)
only consistent with a model consisting

of atoms at different heights above the
Au(111) surface.
Toney and coworkers considered a range
of structural models containing both
ordered Cu and sulfates and proposed
the model shown in Fig. 19. In this
model, Cu atoms form a honeycomb lattice
(2/3 ML coverage) and sulfate molecules
are adsorbed in the honeycomb centers
above the Cu atoms. Both Cu atoms
and sulfate molecules occupy fcc threefold
hollow sites on the Au(111) lattice. Three
oxygen atoms of each sulfate molecule are
bonded to Cu atoms and the remaining
oxygen atom points away from the surface.

From the detailed analysis of the data, it
was possible to determine several bond
lengths in the structure. The CuO bond
a distance comparable
length was 2.15 A,
to the Cu O bond lengths in organic

compounds, which implied that the sulfate
molecules were chemically bonded to the
Cu ad-atoms hence stabilizing the partial
charge of the Cu. It should be noted that
this open Cu structure is not found in UPD
studies of Cu on Au(111) in perchloric
acid solution thus conrming the crucial
role of the sulfate anions in the UPD
process.
329
330

Interfacial structure of
Cu/Au(111) derived from the data in
Fig. 18. (a) Top view and (b) side view.
The large light gray, medium gray, small
light gray, and lled spheres represent
Au, Cu, S, and O atoms, respectively.
(Taken from Ref. [67].)
Fig. 19
(a)
(b)
Cu UPD on Pt(111) in Sulfuric

Acid Solution The interpretation of processes associated with the formation of the
Cu monolayer on Pt(111), and the nature
of the Pt(111)Cu structure, have been
the subject of considerable controversy.
Overviews with some different perspectives can be found in Refs. [7080]. The
RRDE method was successfully applied to
investigate the kinetics of Cu2+ deposition
and to determine the potential-dependent
surface coverage by Cuupd (%Cu(upd) ). The
lower curve in Fig. 20(a) shows a cathodic
sweep of the cyclic voltammogram (CV)
measured on the disk electrode in an
RRDE for UPD of Cu onto Pt(111) in
sulfuric acid solution. The presence of
sulfate anions in solution causes a splitting of the voltammetric Cu deposition
peak. This result is in agreement with
previous studies that, depending on the
solution concentrations and sweep rates,
4.1.4.2.2
can show single peaks or double peaks in

both sweep directions. The lled circles in
Fig. 20(a) correspond to the Cu coverage
obtained by potentiostatic measurements,
in which the amount of Cu deposited on
the disk electrode in a potential step from
Ei (indicated in the gure) is assessed
from the corresponding change in the constant Cu2+ ux to the ring electrode. The
guide to the eye (solid line) shows that in
the potential region in between the two
peaks, the Cu coverage is constant in the
range 0.60.7 ML (i.e. Cu atoms per Pt
surface atom).
The structural details pertaining to Cu
UPD have been derived from in situ
X-ray scattering measurements [81]. The
specular CTR results in Fig. 20(b) clearly
indicate that UPD occurs in a two-stage
process and the sharpness of the peak
implies that the structure formed in the
middle potential region is present over

(a) The lower curve shows the
cathodic sweep of the cyclic
voltammogram (CV) for a Pt (111) disk
electrode in an RRDE conguration in
0.05 M H2 SO4 + 105 M Cu. The
closed circles represent the Cu coverage
(atoms per surface Pt atom) assessed
from changes in the ring current during
Cu deposition onto the disk electrode as
the potential is stepped (for each point)
from Ei . (b) Changes in the X-ray
scattering at (0, 0, 3.9), lower panel, (0,
1, 2.5), middle panel, and (1/3 4/3, 1.5),
top panel, as the electrode potential is
swept at 2 mV s1 . The solid lines and
dashed lines are for the positive-going
(anodic) and negative-going (cathodic)
sweep directions, respectively.
Fig. 20
Pt(111)
0.8
0.05 M H2SO4
5 105 M Cu2+
w = 900 rpm
Cu
0.6
0.4
0.2
i
[A]
Ei
0.0
0.0
0.1
0.2
0.3
0.4
E
[V]
(a)
0.30
(1/3, 4/3, 1.5)
0.20
100
0.10
Intensity
[a.u.]
0.40
(0, 1, 2.5)
0.35
0.30
0.25
0.20
(0, 0, 3.9)
0.25
0.20
0.0
(b)
a very narrow potential range. Indeed,

at 0.15 V weak peaks, which could be
indexed to a ( 3 3)R30 superlattice,

diffraction patterns were observed. For
example, rocking scans through the (2/3,
2/3, 2), (1/3, 1/3, 1.5), and (1/3, 4/3, 1.5)
positions showed broad peaks, FWHM
of 0.84 , 0.49 , and 1.72 , respectively,
which gives an ordered domain size in the
0.1
0.2
0.3
E
[V]
This correlation length is

range 3060 A.
at the limit of detection by X-ray scattering
and the weak intensities coupled with the
narrow stable potential range precludes a
detailed structural analysis as performed
for the Au/Cu/H2 SO4 system [67]. It
should be noted, however, that there are a
number of similarities, most noticeably
that the superlattice structure factors
331
are not measurable at small out-of-plane

momentum transfer. This implies that
a similar structure may be responsible
for the observed diffraction pattern, an
assumption that can be tested by analysis
of CTR data. In the top part of Fig. 20(b)
is shown the X-ray scattering intensity at
(1/3, 4/3, 1.5) as the potential is swept at the
same rate as the data in the lower gures.
Although the signal to noise is poor, it can
clearly be seen that the scattering from the
( 3 3)R30 structure is only present

in the same narrow region highlighted by
the changes at (0, 0, 3.9).
To gain more insight into the nature
of the Cu UPD structures, the (0, 0,
l), (1, 0, l) and (0, 1, l) CTRs were
measured at 0.0 and 0.15 V in order to
derive structural models. Such results are
relatively insensitive to the orientation of
any adsorbed sulfate molecules and so the
sulfate anion was approximated to a single
molecule (i.e. single ion core) consisting of
one sulfur and two oxygen atoms (Z = 64).

Additional sensitivity to the coverage and
location of the Cu atoms was obtained by
performing the CTR measurements with
incident X-ray energies of 8.78 keV (200 eV
below the Cu K adsorption edge) and
8.94 keV (5 eV below the Cu K adsorption
edge). In simulating the CTR data, the
Cu atomic form factor in the scattering
equations was replaced by the form of
Eq. (7). At each electrode potential, a
simultaneous t to the CTR data (E =
8.78 keV) and the ratio data set (R =
I8.78 keV /I8.94 keV ) was performed. The
structural parameters allowed to vary were
the Cu coverage, PtCu surface normal
spacing, sulfate coverage, and Cu-sulfate
surface normal spacing, along with their
respective enhanced DebyeWaller-type
roughness. Results of ts to the specular
CTR data and ratio data at electrode
potentials of 0.0 and 0.15 V are shown
in Fig. 21 by the solid lines.
I8.78 keV/I8.34 keV
1.3
E = 0.15 V
E = 0.0 V
1.2
1.1
1.0
0.9
1.0
The measured specular CTR

scattering for an incident X-ray energy of
8.78 keV (lower panels) and the ratio
between a similar data set measured at
8.94 keV (upper panels) at electrode
potentials of 0.15 V (left) and 0.0 V
(right). The solid lines are ts to the
data to a single structural model and the
ratio data set is calculated by xed
changes in the Cu atomic form factor.
Fig. 21
Intensity
[a.u.]
332
0.1
(0, 0, l ) (units of c*)
Use of the anomalous scattering technique enhances the accuracy of the Cu

coverage determination. At 0.15 V, the Cu
coverage is 0.60 with an adsorbed sulfate
layer of coverage 0.22 on top of the Cu
layer. By comparison with the structural
model derived for the ( 3 3)R30 Cusulfate structure on Au(111) (Fig. 19), it
can be seen that the CTR results are in
good agreement with that study. The bond
lengths obtained for the Cu-sulfate bilayer
are in good agreement with the Au study
in which it was proposed that the sulfate
is chemically bonded to the hexagonal Cu
layer. At 0.0 V, the results give a Cu coverage of 0.81 ML per Pt surface atom with
0.21 ML of sulfate anions still adsorbed on
top of the Cu layer. Because of the increase
in Cu coverage, the sulfate anions are located at an increased height above the Cu
adlayer presumably as the hollow sites in
the hexagonal Cu adlayer have been lled
by Cu ad-atoms. It is interesting that the
Cu coverage derived by the CTR measurement are in good agreement with those
obtained from the RRDE experiments and
it appears, therefore, that it is not possible to complete a full Cu monolayer on
the Pt(111) surface. The lower limit of
potential (0.0 V) was chosen to avoid the
formation of Cu+ via one-electron reduction of Cu2+ without further reduction
to metallic copper. Fits to the nonspecular CTR data at both potentials (0.0 and
0.15 V), again to both raw data and ratio
data sets, gave occupation of both types of
threefold hollow sites, fcc and hexagonal
close packed (hcp), although the fcc sites
are favored (fcc/hcp = 0.51/0.09 at 0.15 V
and 0.67/0.16 at 0.0 V). In contrast with the
Au(111) system in which the Cu ad-atoms
occupied only the fcc sites, it appears that
on Pt(111) the energy difference between
these sites may be smaller. Occupation of
both types of threefold hollow site may
be responsible for the weak ordering of

the intermediate Cu-sulfate bilayer phase
and for the inability to complete the Cu
monolayer at 0.0 V.
Cu UPD onto Pt(111) in the Presence of Bromide Anions The RRDE results for Cu UPD on Pt(111) in solution
containing Cu2+ and Br ions are illustrated in Fig. 22. As the disk potential is
swept across the Cu UPD region, both Cu
UPD peaks recorded on the disk electrode,
rst at 0.25 V and the second at 0.15 V, are
mirrored by the decrease in the ring current from its ir value. Following these two
characteristic Cu UPD peaks, the Pt(111)disk current diminishes to a capacitive
current below 0.1 V, manifesting the completion of the Cu monolayer on the Pt(111)
disk electrode; consistent with the return
of ir to ir in the same potential region.
The fact that the ring-shielding current, ir ,
mirrors the current on the disk electrode
clearly indicates that both cathodic processes are primarily related with Cu UPD.
The same observation is made for the subsequent positive sweep, at which there are
two distinctive processes seen from the
ring electrode; the rst as a shoulder at
ca. 0.375 V and second as a sharp peak
at 0.425 V, both clearly related to stripping of Cu from the Pt(111) disk electrode.
The total amount of Cu deposited by UPD
was evaluated quantitatively by integration
of the ring current, yielding a maximum
Cu coverage of 0.95 ML (almost identical
with the value inferred from disk measurements, the insert in Fig. 22). This rather
small difference in the charge assessed
from the disk and the ring measurements
(ca. 5 C cm2 ) suggests that the anion coadsorption contributes insignicantly to
the Coulombic charge passing the interface during Cu deposition.
4.1.4.2.3
333
20
0.0
900 rpm
5 mVs1
qdisk
qring
qCu
[ML]
Disk current
[A]
15
0.5
10
1.0
0.0
0.1
0.2
0.3
0.4
E
[V]
0
0.1 M HClO4
5 105 M Cu2+
1 104 M Br
(a)
ir
Ring current
[A]
334
Cu2+ + 2e
Cu
2 A
0.0
(b)
0.275 V
0.2
0.4
E
[V]
Top: CV for Cu UPD on a Pt(111) disk electrode in solution

containing Br . Bottom: ring-electrode currents recorded with the ring
being potentiostated at 0.275 V. Insert: copper surface coverages
assessed from the disk electrode (d ) and the ring electrode (r ) during
the positive-going sweep.
Fig. 22
The ring-shielding properties of the

RRDE can also be used to assess the
mass ux of Br from and to the Pt(111)
disk electrode during Cu UPD [74, 79].
Figure 23 shows that starting from the
positive potential limit, the ring current at
0.45 V equals ir indicating the absence
of either Br adsorption or desorption on
the disk. In fact, at this potential, the
surface is almost fully covered by Brad ,
as indicated by the bromide adsorption

isotherm inferred from solution free of
copper, insert of Fig. 23. In the potential
region between 0.45 and 0.3 V, a small
desorption of Br (2 C cm2 ) just
preceding Cu deposition is demonstrated
by the concomitant increase of the ring
current above its unshielded value. As
the disk potential is swept across the
rst UPD peak, the ring current remains
qBr
[ML]
0.50
900 rpm
20 mVs1
Pt(111)
0.25
Disk current
[A]
Pt(111)/Cu
0.0
0.0
0.2
0.4
E
[V]
0.1 M HClO4
1 104 M Cu2+
1 104 M Br
5 A
Ring current
[A]
I R = 16.2 A
1 A
Br
+1.06 V
0.0
Br2 + e
0.2
0.4
E
[V]
Top: Potentiodynamic curve for Cu UPD on a Pt(111) disk
electrode. Bottom: Ring current corresponding to the change in the
Br ux during copper deposition. Insert: Surface coverage of
bromide at the disk electrode with and without copper present in
solution (assessed from ring transients).
Fig. 23
almost constant, manifesting an absence

of either Br adsorption or desorption;
the second Cu UPD peak is accompanied
by the increase in the ring current
from ir , with completion of the Cu
monolayer causing the ring current to
return back to its unshielded value. The
resulting Br from integration of the ring
currents versus the potential of the Pt(111)
surface in a solution containing Cu2+ , and
the corresponding values for the Pt(111)
surface in the solution free of Cu2+ , are
shown in the insert of Fig. 23. Note that Br
in between the two Cu UPD peaks is equal

to the coverage following the complete
stripping of Cu from the Pt(111) surface.
Note also that after complete stripping of
Cu, the Br coverage is exactly the same as
in the case when the adsorption of Br
on Pt(111) was monitored in a solution
free of Cu. Remarkably, the total change in
adsorption/desorption of Br during Cu
UPD is rather low, ca. 4 C cm2 .
X-ray scattering measurements of the
Cu UPD layer on Pt(111) in solution containing bromide anions are summarized
335
in Fig. 24 [77, 80]. With the potential at

0.225 V, four symmetry-independent inplane peaks were observed at (0, 0.735),
(0.735, 0.735), (0, 1.47), (0.735, 1.47), as
shown by the lled circles in the in-plane
diffraction pattern illustrated in Fig. 24.
The changes in the scattering intensity
of the fundamental peak (Fig. 24c) show

that the structure responsible for this peak
is potential dependent and present only
in the potential region between the two
Cu UPD peaks. The location of the fundamental peak at 0.735 (and higher-order
peaks) shifted to 0.742 upon increasing the
2.0
0.12
Intensity
Intensity
0.14
0.10
0.08
1.5
1.0
0.5
0.06
2
0.0
(O k)
0
f
[]
(a)
0.4
0.0
0.4
f
[]
(b)
(h O)
0.77
6
4
2
0
(c)
Peak position (h)
Integrated intensity
336
0.76
CuBr
CuCl
0.75
0.74
0.73
0.2
0.3
E
[V]
0.4
0.2
(d)
0.3
0.4
E
[V]
The in-plane X-ray diffraction pattern observed at 0.225 V for Cu

UPD onto Pt(111) in solution containing Br . The Pt CTRs are represented
by lled circles and the open circles correspond to scattering from the
4 4 structure. Rocking scans through (a) (0, 0.741 0.1) and (b) (1, 0,
0.1). (c) Integrated intensities obtained from scans such as the one shown
in (a) for the CuBr bilayer (lled symbols) and CuCl bilayer (open circles)
phases. (d) Corresponding position of the rst-order diffraction peak from
the CuBr and CuCl bilayer phases as a function of the electrode potential.
Fig. 24
potential stepwise, as shown in Fig. 24(d).

The comparable data for chloride solution
is also shown. The location and symmetry of the in-plane peaks indicates that
the adsorbed species form a hexagonal superstructure that is incommensurate with
the underlying platinum lattice; the nearest commensurate unit cell is (4 4),
which would have a fundamental peak
at 0.750. Note that the chloride structure has a unit cell that is smaller than
(4 4), and the bromide structure has one
that is larger than (4 4), illustrating in
a rather dramatic way the dominant role
played by the anion and Cu-anion bonding in determining the structure of this
intermediate phase. The inserts (a) and
(b) in Fig. 24 show two rocking scans
( scans) through the (0, 0.735, 0.1) and
(1, 0, 0.1) peaks, respectively. Both peaks
have Lorentzian lineshapes with the width
of the (0, 0.735) peak corresponding to a
somewhat smaller
domain size of 200 A,
than the estimated terrace size of the
Additional inforPt(111) surface, 1000 A.
mation about the composition of the unit
cell of this phase was obtained by utilizing anomalous scattering methods, that
is, by measuring the relative intensities
of the four in-plane peak intensities close
to the Cu K adsorption edge (8979 eV).
In general, the intensities of the relatively
strong (0.735, 0.735) and (0, 1.47) peaks
were weaker close to the Cu edge, indicating that the adsorbed superstructure
does not consist of only Cu atoms (which
would result in the reduction of the intensity of all in-plane peaks near the Cu
edge) but must also include Br (for more
details, see Ref. [80]). The simplest realspace structure consistent with the data
has an hexagonal unit cell with the lattice spacing (0.735)1 times the Pt lattice
and a two atom basis
spacing, or 3.36 A,
containing one copper and one bromide
atom arranged in two hexagonal layers,

one of Cu and one of Br, each with the
coverage of 0.53 ML.
Cu UPD onto Pt(001) in the
Presence of Bromide Anions In pure
perchloric acid, Cu is deposited as p(1 1)
islands on Pt(001) in a one-step adsorption
process as evidenced by a single reversible
change in the scattered X-ray intensity
at (1, 0, 0.1), an anti-Bragg point
on the nonspecular CTR [82]. The effect
of bromide adsorbed onto the surface
(i.e. from solution) is to broaden the
potential range of Cu deposition, although
there is no evidence of a stagewise
deposition process. This contrasts with
results obtained on the Pt(111) surface
in which two distinct steps in Cu UPD
were observed (see previous section). With
the electrode potential held at 0.43, 0.12,
and 0.13 V, in each case allowing enough
time for the system to reach an equilibrium
state, a search was performed for inplane X-ray scattering peaks. In agreement
with the results in Sect. 4.1.4.1.4, no
peaks were found at 0.43 V, where only
Br is adsorbed onto the surface. In
addition, no peaks were found at 0.12 V,
an intermediate potential in which on the
Pt(111) surface under identical conditions
the (4 4) structure was observed. At
0.13 V, however, peaks at reciprocal
lattice positions, which could be indexed
to a c(2 2) structure, were found. The
in-plane scattering in reciprocal space is
represented in Fig. 25 together with some
rocking scans at the various reciprocal
lattice points that are indicated, (a) (1/2,
1/2, 0.1), (b) (0, 1, 0.1), and (c) (1/2, 3/2,
0.1). As can be seen from Fig. 25, all of
the measured c(2 2) reections were
relatively broad compared to the peak at
(1, 0, 0.1) on the Pt CTR and, from
Lorentzian ts to the data (solid lines
4.1.4.2.4
337

Pt(bragg)
Pt(CTR)
2
c(2 2)
Intensity
[a.u.]
338
(a)
0.4
(b)
0.0
0.4
f
[degrees]
(c)
A representation of the in-plane X-ray scattering measured at

0.13 V in solution containing 0.1 M HClO4 + 103 M
Cu2+ + 102 M KBr. The solid circles correspond to the measured
c(2 2) reections and the squares to the location of bulk Pt CTRs
and Pt Bragg reections. The lower gures show scans through the
indicated reciprocal lattice points at: (a) (1/2, 1/2, 0.1); (b) (0, 1, 0.1);
and (c) (1/2, 3/2, 0.1). In each case, the solid lines are ts of a
Lorentzian lineshape to the data.
Fig. 25
in Fig. 25), a domain size in the range

3060 A for the c(2 2) structure was
calculated. The c(2 2) diffraction pattern
was only formed if the electrode potential
was held in the region below 0.12 V
and even then, only if the potential was
approached from more negative potential.
Once formed, the c(2 2) structure was
stable at negative potential until the onset
of bulk Cu deposition.
In the previous section, anomalous
scattering techniques were used to show
that the (4 4) structure observed in the

rst stage of Cu deposition on Pt(111)
contained both Cu and Br in the unit cell.
Given that the c(2 2) diffraction pattern
in this study was obtained at 0.13 V, a
potential in which a signicant amount of
Cu is adsorbed onto the surface, similar
measurements were made at several l
values along the (1/2, 1/2, l) and (1/2, 3/2,
l) rods. Measurements of the integrated
intensities at 8779 eV, 200 eV below the
Cu K adsorption edge, and 8974 eV, 5 eV
below the Cu edge, were performed but

showed absolutely no dependence on the
incident X-ray energy. Given that the
dispersion corrections to the Cu atomic
form factor change signicantly over this
energy range, this result implies that no
Cu is contained in the c(2 2) unit cell.
It seems likely, therefore, that the c(2 2)
structure consists of an ordered Br lattice
that is formed on top of a pseudomorphic
Cu layer.
To obtain detailed structural information CTR measurements were modeled by
a c(2 2) Br adlayer on top of a p(1 1)
pseudomorphic Cu monolayer, where the
respective coverages, surface normal spacings, and roughnesses were allowed to vary
in order to t the data. The CTRs passing
through the bulk Pt Bragg reections include contributions to the scattering both
from the Cu and Br adlayers and the Pt
lattice. To further test the model, the CTRs
were measured with incident X-ray energies of 8779 and 8974 eV and the structural
parameters were simultaneously rened
to t both the raw CTR data set (measured at 10 keV) and the intensity ratio
I8779 eV /I8974 eV . This method increases
the sensitivity of the tting procedure to
the Cu layer as shown in Sect. 4.1.4.2.2.
CTR data taken with the potential held at
0.13 V, where the c(2 2) structure was
present, is shown in the lower panels of
Fig. 26. The top panels in Fig. 26 show
the ratio data sets, I8779 eV /I8974 eV , which
clearly indicates the sensitivity of the CTR
measurements to the Cu adlayer. The solid
lines in Fig. 26 correspond to the results
of a simultaneous t to these data sets in
which the Cu coverage, Cu , PtCu spacing, dPt Cu , coverage of Br in a c(2 2)
adlayer, that is, two Br atoms per c(2 2)
unit cell, Br , CuBr spacing, dCu Br , and
the Cu and Br roughnesses were varied. The results are listed in Table 1. A
Tab. 1 Structural parameters to the t to

the data in Fig. 26. The coverages, Cu and
Br , are with respect to a full copper
monolayer (one Cu atom per surface Pt
atom) and a full c(2 2) Br adlayer (0.5 Br
atoms per surface Pt atom)
Cu
dPt Cu
Cu
1.0 0.06
1.75 0.05 A
0.13 0.05 A
Br [c(2 2)]
dCu Br
Br
0.9 0.1
1.79 0.08 A
0.3 0.2 A
schematic of the structure is shown in

Fig. 26.
Although the presence of Br anions
considerably slows the kinetics of ordering,
at negative potential (0.13 V) a complete
pseudomorphic p(1 1) Cu monolayer
is formed on the Pt(001) surface. When
this monolayer is completed, Br forms
a c(2 2) overlayer on top of the Cu
monolayer. The CuBr spacing implies
that, as was the case for the PtBr bond for
Br adsorption onto Pt(001) (Sect. 4.1.4.1.4),
the metalhalide bond is covalent in
nature. The near-neighbor spacing in the
which is similar to
Br adlayer is 3.92 A,
the near-neighbor spacing of Br adsorbed
onto Pt(111) at positive electrode potential
(Sect. 4.1.4.1.3). The Br coverage on the
pseudomorphic Cu monolayer is nearly
identical to the coverage measured on the
Pt surface at positive potential and implies
that close to a full Br monolayer remains
on the electrode surface during Cu UPD. It
is interesting that the bromide layer forms
a structure with long-range order on the Cu
monolayer and not on the Pt(001) surface
in solution free of Cu.
A Physical Model for Cu UPD
The results in this section have demonstrated that Cu UPD is a multistage
4.1.4.2.5
339

The measured CTRs at an
electrode potential of 0.13 V in 0.1 M
HClO4 + 103 M Cu2+ + 102 M KBr
where the c(2 2) structure is present.
Upper panels: The ratio between CTR
data sets taken with incident X-ray
photon energies of 8779 and 8974 eV,
that is, I8779 eV /I8974 eV . The solid lines
are results of a simultaneous t to all of
the data according to the structural
model that is schematically illustrated.
The open circles are Pt substrate atoms,
the shaded circles are Cu atoms, and
the black circles are Br atoms that form
the c(2 2) structure. The side view
indicates the surface normal spacings
listed in Table 1.
I8779 keV/I8974 keV
Fig. 26
1.6
1.4
1.2
1.0
0.8
0.6
101
Intensity
[a.u.]
340
102
103
1
2
3
(0, 0, l)
2
3
(1, 0, l)
dCu-Br
dPt-Br
process that, irrespective of the orientation

of surface and/or the nature of specifically adsorbing anions, is governed by
a complex interplay of Cuupd substrate,
Cuupd anion, and the substrateanion
energetics. The structure of the adlayer
depends strongly on the anion in the
supporting electrolyte. In perchloric acid,
the UPD layer on Pt(111) grows as progressively larger patches of Cu having
the Pt lattice constant, that is, a pseudomorphic adlayer. From the electrocatalysis
standpoint, the Cuupd ClO4 system is
potentially the most interesting, because
a true bimetallic surface exists at which

both Pt and Cu atoms can do catalysis. In either sulfate or halide-containing
solutions, a multistep deposition occurs
with the formation of ordered anion adlattices. Possible Pt(111)Cuupd anion and
Pt(001)Cuupd anion structural models
as a function of Cuupd coverage are shown
in Fig. 27. This model is very similar to
previously proposed models for Cu UPD
on Pt(111) [75] and Au(111) [83], and is
based on the enhanced adsorption of halide
and other anions on Cuupd modied Pt
sites. Clearly, Cuupd is either sandwiched
Pt(111)
jd
[A cm2]
200
3.
- Anions
(4 4)
100
- Cu
0
4.
O2
2.
1.
(a)
3.
Pt(001)
jd
[A cm2]
100
(2 2)
4.
0.2
0.4
(b)
0.6
0.8
1.0
E vs RHE
[V]
A cartoon schematic of the Cu UPD process on Pt in the presence of anion

species. States 14 represent increasing Cu coverage, that is, consistent with a
negative sweep of the electrode potential.
Fig. 27
between the Pt surface and anions or is

in contact with anions adsorbed on the
adjacent Pt sites [77]. States 1a and 1b show
a close-packed bisulfate/halide layer that is
present on the Pt(111) and Pt(001) surfaces
at potentials positive of Cu UPD. States
2a and 2b refer to the initial deposition
of Cuupd on Pt(111) and Pt(001) surfaces
on which the Cu2+ ion is only partially
discharged. At higher surface coverages
by Cuupd (0.5 < %Cu < 1), anions are either entirely displaced from the surface
by Cuupd (state 3a) or Cuupd and anions
form two-layer structures in which anions are adsorbed on both Pt as well as
Cuupd sites (state 3b in Fig. 27). Note that
state 3a is representative of the ordered
(4 4) Clupd Clad (Brad ) bilayer structure.
In contrast with the Pt(111)Cuupd anion
system, on Pt(001), in the same potential
341
342
range (0.10.4 V), a mixture of Cu and Br

atoms is present on the Pt(001) surface
in a disordered structure. Presumably the
strong afnity of adsorbates for the Pt
fourfold hollow site prevents the formation of a long-range ordered bilayer phase
(for example, a (001) plane of Cu(I)Br) as
adsorbate mobility is reduced on the (001)
substrate. The nal step in Cu UPD is
the lling-in of the Cuupd monolayer
to form a bilayer phase: that is, a pseudomorphic (1 1) Cuupd monolayer with
either a disordered anion adlayer on the top
of Pt(111)Cuupd [80, 81], state 4a, or an
ordered c(2 2) bilayer Cuupd Brad structure on the Pt(001) surface [82], state 4b.
In both Cl and Br solutions, Cu UPD
onto Pt(111) proceeds via the same intermediate phase. In sulfuric acid solution,
however, the equivalent intermediate state
on Pt(111) corresponds to the forma
tion of the honeycomb ( 3 3)R30

Cuupd HSO4 structure [81] that consists
of both 0.22 ML bisulfate anions and
0.66 ML of Cuupd in the unit cell. The same
structure was proposed for the Au(111)sulfate system [67, 68]. The formation of
the intermediate state (state 2) occurs over
a very narrow region of potential. As will
be shown in Sect. 4.1.5, the existence of
bare platinum sites in state 2 (Fig. 27) is
essential for the adsorption of O2 , and this
rst-stage formation plays the key role in
the activity of the PtCuupd interface for
the ORR.
4.1.5
The Oxygen-reduction Reaction (ORR)
The main problem in the development of

an efcient electrochemical energy conversion device (fuel cell) is in the sluggish
ORR kinetics even on the most catalytically
active electrode materials, for example, on
platinum group metals measurable currents are obtained only below 1 V. It is

customary, therefore, in kinetic studies of
the ORR to use the current density at a
xed potential, for example, at 0.9 V, as
a measure of the reaction rate instead of
the exchange current density, for example,
at the reversible potential. Given that the
reversible potential of the ORR is 1.23 V,
virtually all reversible electrode potentials
for solid elements are less positive, and
consequently not many substrates may
even be considered as oxygen electrodes.
Besides the stability of electrode materials, the ORR is accompanied with many
other complications, such as the necessity of cleaving the strong OO bond (ca.
497 kJ mol1 ) to produce the maximum
efciency in the 4e process. Complete
evaluation of electrocatalysis for the ORR
must, therefore, accommodate many contributing factors, including variations of
the free energy of activation, the heats
of adsorption of reactive intermediates,
and the preexponential surface coverage
term, (1 %ad ). In this section, closely
following the microscopic description of
clean and modied metal surfaces (presented in Sects. 4.1.3 and 4.1.4), we discuss
recent efforts towards a fundamental understanding of electrocatalysis of the ORR
on these surfaces.
Reaction Pathway
The ORR is a multielectron reaction that
may include a number of elementary steps
involving different reaction intermediates.
Of the various reaction schemes proposed
for the ORR [84], the simplied version
of the scheme given by Wroblowa and
coworkers [85] appears to be the most
effective one to describe the complicated
reaction pathway by which O2 is reduced
at metal surfaces:
4.1.5.1

k1
O2
O2, ad
k2
H2O2, ad
k4
k3
H2O
k5
H2O2
On the basis of this reaction scheme, O2

can be electrochemically reduced either
directly to water with the rate constant
k1 without intermediate formation of
H2 O2,ad (so-called direct 4e reduction)
or to H2 O2,ad with the rate constant k2
(series 2e reduction). The adsorbed
peroxide can be electrochemically reduced
to water with the rate constant k3 (series
4e pathway), catalytically (chemically)
decomposed on the electrode surface
(k4 ), or desorbed into the bulk of the
solution (k5 ). Although a number of
important problems pertaining to the
interpretation of the reaction pathway
for the ORR on metal surfaces have yet
to be resolved, it appears that a series
pathway via an (H2 O2 )ad intermediate
may be operative on pure Pt, Au, and
Cu and their bimetallic surfaces. This
can be considered as a special case
of the general mechanism in which k1
is essentially zero, that is, there is no
splitting of the OO bond before a
peroxide species is formed. Peroxide, on
the other hand, may (k5 = 0) or may
not (k5 = 0) be further reduced to water.
In either case, the rate-determining step
appears to be the addition of the rst
electron to O2,ad . The rate expression is
then [86, 87],

i = nF KcO2 (1 %ad )x exp

exp
1G
RT
F E
RT
(12)
where n is the number of electrons, K is

the rate constant, cO2 is the concentration
of O2 in the solution, %ad is the total
surface coverage by anions (%ad) and OHad
(%OHad ), x is either 1 or 2 depending on
the site requirements of the adsorbates,
i is the observed current, E is the
applied potential, 1G is the standard
free energy (chemical) of activation
(1G = 1G0 + 1G0% %ad ), 1G0 is the
standard free energy of activation at
zero coverage, 1G0% is the difference
in the heats of adsorption of reactance
and products in the rate-controlling step
(1G0% = 1G00 + r%ad ), and are the
symmetry factors (assumed to be 1/2),
and r is a parameter characterizing the
rate of change of the apparent standard
free energy of adsorption of reaction
intermediates with the surface coverage by
adsorbing species. In deriving Eq. (12), it
is assumed that the reactive intermediates,
(O2 )ad and (HO2 )ad , are adsorbed only
to low coverage, that is, they are not a
signicant part of %ad . In the following
sections, this reaction pathway and rate
expression are used to analyze the effects of
various factors on the kinetics of the ORR
on Pt(hkl), Au(hkl), and Cu(hkl) surfaces.
ORR on Pt(hkl) Surfaces
During the past two decades, a great deal
of work in surface electrochemistry has
been aimed at elucidating the role of the
local symmetry of surface atoms in electrocatalysis, particularly for the kinetics of the
ORR on platinum single-crystal surfaces.
It is now well established that the kinetics
of the ORR on Pt(hkl) surfaces are sensitive
to the surface structure [88, 89] and arise
because of structure-sensitive adsorption
of spectator species, such as Hupd [89, 90],
OHad [91], HSO4(ad) [90], Clad [92], and
Brad [87]. Within the limited scope of this
report, it will not be possible to review all
4.1.5.2
343
344
of these results. Rather, we will show, using representative examples, the kind of
information that can improve our understanding of the role of the local symmetry
of platinum surface atoms in O2 reduction electrocatalysis. There are two general
observations concerning the structure sensitivity of the ORR: (i) the same activation
energy in both acid (at the reversible potential ca. 42 kJ mol1 [93]) and alkaline
solution (at 0.8 Vca. 40 kJ mol1 [94]) has
been found for all three Pt(hkl) surfaces;
(ii) the structural sensitivity is most pronounced in electrolytes in which there is
strong adsorption of anions. As a consequence, the structure sensitivity of the
ORR on Pt(hkl) surfaces is mainly determined by the preexponential (1 %ad )
coverage term.
ORR at the Pt(hkl)halide Interface For the purpose of demonstrating
the importance of the (1 %ad ) term in
the kinetics of the ORR on Pt(hkl), two
representative sets of polarization curves
are shown in Figs. 28 and 29. Figure 28
shows that in the presence of Cl , the
variation in activity increases in the order (001) < (110) < (111). Figure 28(a)
shows that the ORR on the Pt(111) surface in 0.1 M HClO4 + 103 M Cl is
strongly inhibited, with an activity in a
solution containing Cl being several orders of magnitude lower than in 0.1 M
HClO4 [88, 92]. Figure 28(a) also shows
that between 0.3 < E < 0.9 V, the ring
currents were a small fraction of the disk
currents, implying that on the surface
highly covered with Clad , oxygen reduction proceeds almost entirely through the
4e reduction pathway. The appearance
of peroxide oxidation currents on the ring
electrode begins at potentials negative of
0.25 V, and parallels with the adsorption
of hydrogen on Pt(111). In contrast with
4.1.5.2.1
Pt(111), the ORR on Pt(110) (Fig. 28b) and

Pt(001) (Fig. 28c) in Cl -containing solution is accompanied quantitatively with
peroxide formation, implying that within
the same potential limits H2 O2 is formed
as an intermediate on the Pt(110) and
Pt(001) electrodes covered with Clad . On
Pt(001) the percentage of H2 O2 produced between 0.25 < E < 0.6 V during
the ORR is almost constant and is close
to 25%. Note that, at potentials more
negative than E = 0.3 V, the surface coverage with Clad is reduced substantially,
but the O2 -reduction currents sharply decrease to diffusion-limiting currents for
ca. 2e per O2 molecule at the negative
potential limit. In the same potential region, the O2 -reduction currents on the disk
electrode are accompanied quantitatively
by the H2 O2 -oxidation currents on the
ring electrode, reaching maximum peroxide production of ca. 90% at 0.075 V.
In Cl -free solution, however, a 3e reduction is observed at the same potential
limit [90]. This may suggest that Clad is
always present on Pt(001) covered by Hupd
(as also observed for Brad on Pt(001) see
Sect. 4.1.4.1.4) thus, the availability of Pt
atoms required for the breaking of the
OO bond are signicantly reduced and
the ORR proceeds entirely through the 2e
peroxide pathway on the Pt(001)-Hupd -Clad
surface. The amount of peroxide formed
on the Pt(110) surface is substantially
smaller, Fig. 28(b), ca. 2.5%, and no increase in peroxide formation was observed
in the Hupd region.
As demonstrated recently [92], the key to
resolving the structural sensitivity of the
ORR and the HOR on Pt(hkl) surfaces in
Cl -containing solution is to be found in
understanding the interplay between the
electronic properties (the work function)
of the substrate and the match between
the Pt surface atomic geometry and the
IR
[A]
50
iD
[mA cm2]
0
2
4
6
Pt(111)
IR H2O2
iD
;
[mA cm2] [A] [%]
(a)
2
0
2
4
Pt(110)
IR H 2 O2
;
[A] [%]
(b)
0.05 M H2SO4 + 103 M Cl
0.05 M H2SO4
50
iD
[mA cm2]
0
2
4
Pt(001)
6
0.2
0.4
0.6
0.8
E
(c)
[VRHE]
Polarization curves for Pt single-crystal electrodes in

0.05 M H2 SO4 with and without 103 M Cl : (a) Pt(111); (b) Pt(110);
and (c) Pt(001).
Fig. 28
adsorbing anions. The high activity of the

Pt(111)-Clad surface for the ORR can be
attributed to the highest work function
(the pzc) of the Pt(111) surface. An obvious
consequence of this positive value of the

pzc is that the onset of the desorption of
Clad occurs at more positive potentials on
this surface. The fact that Clad inhibits
345
346
the initial adsorption of O2 molecules,

but does not affect the pathway of the
reaction implies that the interaction of
Clad with the close-packed (111) surface
is not very strong. On the other hand,
the lower pzc of Pt(001) would require a
larger shift in the thermodynamic driving
force to displace Clad from the surface,
accounting for a strong deactivation of
the ORR on the Pt(001)-Clad surface. The
substantial production of peroxide on the
Pt(001)-Clad surface is related to strong
adsorption of Clad on the fourfold hollow
(001) sites, (similar to the mechanism
that frustrates the ordering of Br on the
Pt(001) surface see Sect. 4.1.4.1.4) with a
corresponding reduction in the available
pairs of platinum sites required for the
breaking of the OO bond. This model
may seems to conict with the case of
Pt(110), in which the kinetics of the ORR
is higher than on the Pt(001) surface even
though the pzc of the (110) surface has a
lower value and interaction of Clad with low
coordinated (110) sites is as strong as on
Pt(001). The key to resolving this seeming
contradiction is, however, in the structure
of the Pt(110) surface itself (as described in
Sect. 4.1.3). Clad is most probably adsorbed
in the troughs or missing rows of the
reconstructed (110) surface and, hence, O2
adsorption and the breaking of the OO
bond may take place on the top rows of
platinum sites (the rst atomic layer in the
structural model shown in Fig. 9).
Clad is not observed to form any ordered
structures on Pt(hkl) surfaces, in contrast
with Brad on Pt(111) see Sect. 4.1.4.1.3).
Furthermore, the potential dependence
surface coverage by Clad cannot be discerned from RRDE measurements. As a
consequence, it is difcult to nd the
relationship between the activity of the
ORR and the potential dependence of
Clad surface coverage/structures. To get
insight into this relationship, the ORR has

recently been studied on the Pt(111) surface in solution containing Br anions.
As demonstrated in Sect. 4.1.4.1.3, the
potential dependence of bromide adsorption/surface structure was obtained from a
combination of RRDE and X-ray scattering
measurements.
Figure 29(b) summarizes a family of polarization curves for the ORR on Pt(111) in
0.1 M HClO4 + 104 M Br along with a
representative polarization curve recorded
in solution free of Br anions. In all experiments, the ring was potentiostated at
0.95 V, where oxidation of peroxide arriving at the ring is under diffusion control,
Sect. 4.1.2.2. In pure perchloric acid, at
low overpotentials, the ring currents were
a very small fraction of the disk currents,
implying that on the surface covered with
reversible adsorbed OHad (the so-called
buttery region) oxygen reduction proceeds almost entirely through the direct
4e pathway. The appearance of peroxide
oxidation currents on the ring electrode
begins at potentials negative of 0.0 V, and
parallels the adsorption of hydrogen on
Pt(111). At the negative potential limit,
the limiting current corresponds to exactly
two-electron reduction of O2 . Figure 29(b)
shows that the ORR on the Pt(111) surface is strongly inhibited by Br , with
an activity in the solution containing Br
being several orders of magnitude lower
than in pure perchloric acid. Figure 29(b)
shows that O2 reduction on the Pt(111)-Br
surface starts at ca. 0.3 V (%Br 0.4 ML
and %OH 0.0 ML in Fig. 29a) and is under combined kinetic-diffusion control between 0.1 < E < 0.3 V. Figure 29(b) also
shows that peroxide was detected on the
ring electrode in the same potential region, indicating that H2 O2 is formed as
an intermediate on the electrode covered with Br 0.40.35 ML. At potentials
900 RPM
10 A
0.4
qBr
[Ml]
0.3
0.2
Pt(111)
0.1 M HClO4
0.1
0.1 M HClO4 + 8 103 M Br

H ads/des on Pt(111)/Br
(a)
0.0
IR
[A]
150
100
50
0
Pt(111)/Br
2500
1600
900
400
RPM
ID
[mA]
400
1.0
2500
2.0
0.2
(b)
0.0
0.2
0.4
0.6
E
[V]
(a) CV for Pt(111) in 0.1 M HClO4 (dashed line) showing

the effect of Br in solution (solid line) and the effect on the
adsorption/desorption of hydrogen (dotted line). The Br coverage is
shown by the dot-dashed line. (b) Polarization curves for the ORR
on Pt(111) in 0.1 M HClO4 + 104 M Br .
Fig. 29
more negative than E = 0.1 V, the surface coverage with Br decreases substantially, for example, from Br 0.35 ML
to Br 0.25 ML, and the O2 -reduction
currents increase to a diffusion-limiting

current for 3e per O2 molecule. In the
same potential region, the O2 -reduction
currents are accompanied quantitatively by
347
348
the H2 O2 -oxidation currents on the ring

electrode, implying that in the presence
of Br anions O2 reduction does not proceed entirely through the 4e reduction,
as found for the Pt(111)-Clad system [92].
It was proposed that strongly adsorbed
Br can simultaneously suppress the initial
adsorption of the O2 molecule and the formation of pairs of platinum sites needed
for the breaking of the OO bond [87].
On the basis of this proposition, the physical model that appears to rationalize the
ORR on the Pt(111)-Brad electrode is one
in which the active sites for the adsorption of molecular O2 are the small number
of bare platinum sites that are created
when desorption of Brad takes place from
the Pt(111) surface. Finally, by comparing the ORR in Fig. 29 with the potential
stability of the Brad structure in Fig. 15,
one can nd no correlation between the
ordered surface structure and the kinetics
of the ORR.
The analysis for the ORR on Pt(111)
has shown that the Tafel plot of mass
transport corrected currents for the ORR at
the Pt(111) surface in solution containing
Br is ca. 118 mV dec1 . The fact that
the Pt(111)-Brad surface has a single slope
of 118 mV dec1 strongly suggests that
the apparent standard free energy of
adsorption of the reaction intermediates is
not affected by the neighboring Brad , that
is, the adsorption of reaction intermediates
that are formed during the ORR on
the OHad -free Pt(111) surface obeys the
Langmuirian conditions. In contrast, the
Tafel plot in 0.1 M HClO4 solution
free of Br deviates signicantly from
120 mV dec1 , similarly to what is found
in the literature for the O2 reduction
on a polycrystalline Pt electrode [95]. A
number of theories have been proposed
to explain the transition in the Tafel
slope on polycrystalline Pt electrodes,
which is usually from ca. 60 mV dec1

at low current densities to 120 mV dec1
at high current densities [86]. The change
in the slope has been attributed to the
change from Temkin to Langmuirian
conditions for the adsorption of reaction
intermediates [95], or as being due to a
change in the surface coverage by PtOHad ,
which effects the adsorption of molecular
O2 [96]. In both theories, the rst chargetransfer step has been proposed as the
rate-determining step. It has, however,
been very difcult to obtain denitive
evidence that will support either theory,
since both the nature of intermediates
in the ORR have never been probed
experimentally and the surface coverage
by OHad species was poorly dened. By
eliminating the surface heterogeneity with
single-crystal electrodes, it was proposed
that there are two modes of action of
the OHad state on the kinetics of the
ORR on the Pt(111) surface [97]: (i) OHad
can block the adsorption of O2 on active
platinum sites, that is, they compete for
the same sites; and (ii) OHad can alter the
adsorption energy of intermediates (the
r%ad follows the Temkin conditions) that
are formed during the ORR on the bare Pt
sites neighbored by the OHad .
ORR at the Pt(hkl)Cuupd Interfaces Studies of the kinetics of the ORR
on platinum single-crystal electrodes modied by UPD metal ad-atoms are of interest
primarily from a fundamental science perspective. Since it is possible to begin the
experiment with a clean Pt(hkl) surface in
which the ORR is well known, the change
in both rate and reaction pathway with the
addition of another metal to the surface,
usually in a highly structured manner, can
provide some insight into the ORR on metals that are otherwise difcult to study, for
example, Cu, Pb, and Bi. Only the results
4.1.5.2.2
on Cuupd -modied surfaces are reviewed

here because trace levels of copper may be
present in real systems, mostly as a corrosion product, thus the effects of Cu UPD
on Pt will serve as a model system to understand why a small amount of Cu UPD
has a devastatingly inhibiting effect on the
ORR. The rst result for the ORR on the
Pt(111)-Cuupd electrode was published by
Abe and coworkers [76]. By using STM and
hanging meniscus rotating disk electrode
(HMRDE) methods, the authors demonstrated a close correlation between the
inhibition of the ORR with the microstructure of Cuupd adlayer. In sulfuric acid
solution, threefold hollow platinum sites
are still accessible for O2 adsorption inside
the honeycomb ( 3 3)R30 structure

that forms at ca. 0.5 V, but the reduced
geometry favors 2e reduction to peroxide
versus 4e reduction to water. The correlation between the surface structure of
Cuupd and the kinetics of the ORR can
be examined by using a combination of
RRDE and X-ray scattering measurements
(the X-ray diffraction measurements are
described in Sect. 4.1.4.2). These results reveal that there is no correlation between the
ORR kinetics and the ordered Cuupd surface structures irrespective of the nature
of the specically adsorbing anions [98].
The effect of Cuupd on the ORR at the
Pt(111) interface in 0.05 M H2 SO4 and
0.05 M H2 SO4 + 103 M Cl are shown
in Figs. 30(a, b). For comparison, the polarization curves for the ORR in Cu2+ -free
solutions under the same experimental
conditions are shown as dotted curves.
Clearly, in a solution containing Cu2+
ions, the ORR is strongly inhibited in both
Cl -free and Cl -containing solutions.
Very similar behavior was also observed
on a polycrystalline electrode [99, 100] and
Pt(111) [76, 98] in different acid solutions.
Given that Cu UPD is a rather slow
process, the ORR currents at the Cuupd modied surfaces can also be measured
under steady state conditions, see the
closed circles in Fig. 30. In these experiments, the ORR is more inhibited than
in the potentiodynamic experiments. As
shown in Fig. 30, the deactivation is more
pronounced in the presence of Cl ions,
presumably due to stronger interaction of
specically adsorbed chloride (Clad ) than
bisulfate (H2 SO4(ad) ) anions with the PtCuupd surface. Analysis of the RRDE data
(Fig. 30) also revealed that the mechanism
for the ORR on Cuupd -modied electrodes
is the same as on unmodied Pt(111),
that is, proceeds mostly as a 4e reaction pathway with negligible solution
phase peroxide formation, (ca. 3%). It
is surprising, however, that a relatively
small amount of Cuupd has such a devastating effect on the rate of ORR. This
anomalously large inhibition by a very
small amount of Cuupd is attributed to
enhanced anion adsorption on platinum
atoms adjacent to Cuupd atoms [98], see
Sect. 4.1.4.2 and Fig. 27. The predominant
role of anions in determining the rate of the
ORR on Cuupd -modied platinum singlecrystal surfaces is supported by the fact
that the activity of the ORR decreases in
the same sequence as the Cu-anion bond
strength increases: PtCuupd HSO4,ad
PtCuupd Br . The model that rationalizes these results is one in which the active
sites for the adsorption of molecular O2
are the small number of platinum islands created in state 2 in Fig. 27. It is
important to note that in the Cu UPD
potential region, Cuupd is always covered
with anions, even at the most negative
potentials at which Pt(001) and Pt(111)
are covered by nominally 1 ML of Cuupd .
It is obvious, therefore, that is impossible to determine the true catalytic activity
of an anion-free pseudomorphic (1 1)
349
Cu
IR
[A]
6
0
0
10
20
0.2
Ei
0.2
0.4
0.6
[VRHE]
(b)
[VRHE]
1.0
0.8
0.05 M H2SO4 + Cl
10
0
0
0.05 M H2SO4 + 8.105 M

Cu2+ + 1.103 M Cl
0.8
100 A cm2
Pt(111)
20
0.0
0.5
0.6
0.05 M H2SO4 + 8.105 M Cu2+

0.05 M H2SO4
0.4
50 A cm2
Pt(111)
1.0
1.0
Fig. 30
The effect of Cuupd on the ORR at the Pt(111) interface in (a) 0.05 M H2 SO4 and (b) 0.05 M H2 SO4 + 103 M Cl , represented at the bottom of
(a) and (b). In both cases, the solution contains 104 M Cu2+ . The corresponding Cu coverages (open circles) obtained from RRDE experiments (disk
currents are shown by solid lines) are represented at the top.
(a)
iD
[mA cm2]
Cu
IR
[A]
iD
[mA cm2]
350
(b)
Iring
[A]
2.5
0.8
2.0
1.5
1.0
0.5
0
0.0
50
0.6
Edisk vs SCE
[V]
0.4
Idisk
[A]
Cu(111)
@ 1600 rpm
0.8
10
0.4
0.2
0.5 M H2SO4
0.1 M HClO4
Edisk vs SCE
[V]
0.0
0.0
0.8
0.6
0.4
2 0.8
0.4
0.0
0.2
0.5 M H2SO4
0.1 M HClO4
Edisk vs SCE
[V]
Edisk vs SCE
[V]
Cu(001)
@ 1600 rpm
Idisk
[A]
4
0.0
0
0
50
100
(d)
(c)
Fig. 31
Polarization curves (positive sweeps) of the ORR on (a, b) Cu(111) and (c, d) Cu(001) electrodes in 0.5 M H2 SO4 (dashed lines) and 0.1 M
HClO4 (solid lines). Inserts: The corresponding CVs obtained on the disk electrodes in 0.1 M HClO4 in the absence of O2 .
(a)
Idisk
[mA]
Iring
[A]
Idisk
[mA]
100

351
352
Cuupd monolayer. One way to overcome

this problem is to study the ORR on Cu
single-crystal surfaces.
ORR on Cu(hkl) Surfaces
In contrast with the extensive studies of the
ORR on Pt(hkl) surfaces, there has been
no substantial fundamental study of the
effects of anion adsorption on the kinetics
of the ORR on Cu(hkl) surfaces. Very
recently, by utilizing the RRDE technique,
Brisard and coworkers [101] have shown
that the ORR on Cu(111) and Cu(001)
surfaces in sulfuric acid solution is a
structure-sensitive process, see Fig. 31.
As for Pt(hkl), an interpretation of the
variation in the activity of this process with
the different low-index crystal surfaces
of Cu can be presented on the basis
of the premise of the structure-sensitive
adsorption of sulfuric acid anions on
Cu(hkl) surfaces, for example, as for Pt(hkl)
with the (1 %ad ) term.
A comparison of the polarization curves
(positive sweeps) of the ORR on Cu(111)
and Cu(001) at the same rotation rate
(1600 rpm), Fig. 31, shows relatively small,
for example, less than a factor of 2, differences in activity, but very different reaction
pathways below 0.4 V. Since very little
is known about the nature of adsorbed
anions on Cu(hkl) in sulfuric acid solution, or how the adsorption sites might
differ between the different surfaces, it is
difcult to rationalize the structural sensitivity of the ORR on Cu(hkl). For the
Cu(111)-SO4,ad interface, both Wilms and
coworkers [102] and Li and Nichols [103]
found by STM that a complex incommensurate ordered structure of sulfate anions
formed at the Cu(111) surface throughout the 0.1 to 0.4 V potential region in
H2 SO4 . This ordered structure does not
appear to lie along any high symmetry
directions of the Cu(111) substrate but it
4.1.5.3
forms a particularly stable adlayer between

0.3 < E < 0.1 V. In addition, by utilizing a potentiostatic STM image, it was
shown that disordering of this structure
coincides with the onset of formation of
the cathodic peak, for example, with the
desorption of (bi)sulfate anions at 0.35 V
in Fig. 31(a). Given that the onset of the
ORR on the Cu(111) surface in Fig. 31(a)
is observed at the same potential, it was
suggested that the kinetics and reaction
pathway of the ORR is controlled by the
number of Cu bare sites on which the
adsorption of O2 molecules may take place.
Although structural information regarding
the surface coverage of (bi)sulfate anions
on Cu(001) is lacking, it is reasonable to
suggest that the ORR pathway on Cu(001)
is the same as on Cu(111).
The overall inhibiting effect of anion
adsorption on the kinetics of the ORR
was determined by comparing the ORR on
Cu(111) and Cu(001) in 0.5 M H2 SO4 and
0.1 M HClO4 [101]. Figure 31(a) shows
that in 0.1 M HClO4 the initial reduction of
O2 molecules is associated with production
of peroxide on the ring electrode (Fig. 31b).
A comparison of the voltammetric curve of
Cu(111) in 0.1 M HClO4 solution free of
O2 (inset of Fig. 31b) with the polarization
curves for the ORR in Fig. 31(a), suggests that the amount of peroxide formed
on the disk electrode is determined by
the potential-induced change in the surface coverage of OHad on the Cu(111)
surface [101]. Thus, the decrease in the
surface coverage by OHad on Cu(111) results in an increase in the rate of the ORR
on the Cu(111) surface, as observed for
the ORR on Pt(hkl) in the potential region of the structure-sensitive adsorption
of OHad [97]. Further inspection of Fig. 31
shows that the combined kinetic-diffusion
control of the ORR on Cu(111) and Cu(001)
is followed with the smooth transition to a
well-dened diffusion-limiting current for

the 4e reduction (0.65 < E < 0.3 V),
consistent with an absence of peroxide on
the ring in this potential region. The fact
that the ORR on Cu(111) and Cu(001) is
orders of magnitude lower in sulfuric acid
solution than in pure perchloric acid, indicates that the compact layer of strongly
adsorbed bisulfate anions suppresses the
initial adsorption of the O2 molecule on
these surfaces, for example, the kinetics of
the ORR on Cu(hkl) is determined by the
(1 %ad ) term in Eq. (12).
ORR on Au(hkl) Surfaces
The ORR on gold single-crystal surfaces,
was also found to be structure sensitive
in both alkaline and acid solutions. In
fact, the rst observation that the rate of
the ORR is structure sensitive was made
from measurements of the ORR on gold
single-crystal surfaces in alkaline solution [104, 105]. In acid solutions [105, 106],
the order of activity increases in the
sequence: Au(111) < Au(110) < Au(001).
In contrast with Pt(hkl) and Cu(hkl) surfaces, the ORR on Au(hkl) surfaces in
acid solution is a very slow process. Because of the very weak interaction of O2
molecules with gold, two-electron reduction with 100% production of peroxide
was commonly observed on all Au(hkl)
surfaces. The ORR on Au(hkl) in alkaline
solution represents one particularly interesting example of a structure-sensitive
reaction. In independent studies by two
research groups [104, 105], the Au(001)
surface in alkaline solution was found
to be the most active of the low-index
surfaces (by nearly four orders of magnitude) and was reported as even more
active than polycrystalline platinum under
the same experimental conditions. In spite
of efforts by many research groups to determine the origin of this catalytic activity,
4.1.5.4
the reason for this remarkable effect remains unclear. It was reported [104] that
the unique catalytic activity of the Au(001)
surface only appears in the potential region in which OHad is adsorbed on the
surface, see Fig. 32. Figure 32 also shows
that at potentials negative of the potential
for Au(001)-OHad formation, the surface
provides only 2e reduction. On the basis of this evidence, it was proposed that
the change in the catalytic activity of the
Au(001) surface from 4e to 2e reduction was due to the existence of a AuOH
layer. On the other hand, it was also proposed that the higher activity of Au(001)
for the ORR is related to the dissociative
adsorption of HO2 anions on the fourfold symmetry sites that occur only on the
(001)-(1 1) surface [105]. Consequently,
the change in the oxygen-reduction pathway from 4e to 2e reduction is due
to potential-induced reconstruction of the
(1 1) surface to a structure similar to the
hex reconstruction found in UHV. The
results in Sect. 4.1.3.2.1 (Fig. 7) indicate
that the (1 1) hex reconstruction
is not the dominant mechanism for this
change in reaction pathway since the transition in reduction kinetics occurs in a
much narrower potential range than the
(1 1) hex transition. Furthermore,
the time constant for the two transitions
differs by approximately two orders of magnitude, with the structural transition being
the slower.
Summary of the ORR
Just as in heterogeneous catalysis, the
ultimate challenge in electrocatalysis is
to relate the microscopic details of adsorbed states of reaction intermediates
to the macroscopic measurement of kinetic rates. There are many strategies
that may be employed in this endeavor.
Here we develop the relation between
4.1.5.5
353
Current
100 A cm2
900
1600
2500 rpm
2 mA cm2
1.2
1.0
0.00
0.8
0.25
0.50
0.6
0.75
Reconstruction fraction
1.4
Intensity (normalized)
354
0.4
1.2
0.8
0.4
0.0
0.4
1.00
0.8
E
[V]
Top: Polarization curves for the ORR on Au(001) in 0.1 M
KOH at different rotation rates. The dashed line shows the CV.
Bottom: Corresponding changes in the X-ray scattering signal at (0,
0, 2.3) where the scattered intensity depends on the presence of the
surface reconstruction, that is, deconstruction leads to an increase
in the intensity. The fraction of the surface that is reconstructed is
shown on the right-hand axis.
Fig. 32
the kinetics of the ORR and two factors: a chemical-dependent part, 1G =

1G0 + 1G0% %ad , and the preexponential (1 %ad ) term. The rst term is mainly
determined by the nature of interaction
of O2 molecules and reaction intermediates with the electrode surface, that is,
by the MO2 energetics and the ability
of the substrate (M) to break the OO
bond. The preexponential (1 %ad ) term
is, however, mainly determined by the interplay between the electronic properties
(the work function, for example, the pzc)

of the substrate and the match between
the surface atomic geometry of the single crystals and the geometry of adsorbing
spectator species (Hupd , OHad , and anions), that is, by the ability of the surface
to adsorb molecular O2 . In the following
discussion, these two terms are used to
analyze the difference in the kinetics of
the ORR on Pt(hkl), Au(hkl), and Cu(hkl)
and the structure sensitivity of the ORR on
M(hkl) surfaces. One should keep in mind,
however, that a compensation effect exists

between the preexponential factor and the
heat of activation.
Given that the energetics of O2 molecules and intermediates on Pt(hkl), Au(hkl),
and Cu(hkl) in solution is not known, the
1G = 1G0 + 1G0% %ad term cannot be
determined on these electrodes. However,
one can use the values of thermodynamic
functions that are obtained from UHV
measurements and test them in the electrochemical system for consistency. Two
general features in the energetics of the
O2 interaction clearly emerge from the
UHV studies: (i) although the heat of
adsorption of oxygen increases in the
order Pt(hkl) < Cu(hkl)<<<Au(hkl), the
heats of O2 adsorption are relatively insensitive to the surface structure of the
substrate [107, 108]; and (ii) the energetics of interaction are strongly coverage
dependent [23]. The coverage dependence
on single-crystal surfaces arises primarily from adsorbateadsorbate repulsive
interactions rather than surface heterogeneity [109]. As we demonstrate below,
on the basis of UHV-like arguments
for the M-O2 energetics and the potentialdependent surface coverage of spectator
species (from Sects. 4.1.3 and 4.1.4), one
can develop a very reasonable relation between the kinetics of the ORR and the
parameters 1G% , and (1 %ad ).
We rst discuss the structure sensitivity
on the same metal substrates. The fact that
the kinetics of the ORR on Pt(hkl) (Figs. 28
and 29), Cu(hkl) (Fig. 31), and Au(hkl)
(Fig. 32) are orders of magnitude lower
in acid solutions containing strongly adsorbing anions indicates that the structure
sensitivity is due primarily to the structuresensitive adsorption of spectator species,
for example, by the (1 %ad ) term. The
(1 %ad ) term is determined by both the
pzc of the substrate as well as the match
between surface atomic geometry of the

single crystals and the geometry of adsorbing spectator species. The importance of
the pzc in controlling the (1 %ad ) term
is obvious from the Pt(hkl)Clad system
in which, due to the high anodic pzc of
Pt(111) and the ability of this surface to
desorb Clad at the most positive potentials,
this surface is the most active for the ORR.
In sulfuric acid solution, however, an exceptionally large deactivation is observed
at the Pt(111) surface. For this system, the
(1 %ad ) term is affected more with the
symmetry factors than with the pzc of the
surface; for example, due the strong adsorption of the (bi)sulfate anion, which is
induced by the geometrical match between
the hexagonal (111) surface and the C3v geometry of the oxygen atoms of the sulfate
anions, the ORR is strongly deactivated on
the Pt(111) surface. The exceptionally low
activity of Cu(111) and Au(111) in sulfuric
acid solution is also related to the strong interaction of bisulfate anions with the (111)
surface atoms.
The most direct probe of the pzc
effects in the kinetics of the ORR can
be obtained by comparing the catalytic
activities of Pt(hkl) versus Cu(hkl). On
both surfaces, in solution free of strongly
adsorbing anions, the reaction mechanism
is the same (the 4e reduction), the
formation of peroxide appears to precede
the breaking of the OO bond [91]. Owing
to the relatively high energy of adsorption
of oxygen on Cu (ca. 200 kJ mol1 on
Cu vs. 350 kJ mol1 on Pt [23]), the
exceptionally large difference in catalytic
activity between Cu and Pt cannot be
explained by a difference in the isosteric
heat of adsorption of O2 on these two
surfaces. We propose, therefore, that the
main contribution to the deactivation of
the ORR on Cu(hkl) is due to the (1 %ad )
term, which is controlled by the low
355
356
pzc of Cu, leading to high coverages of

oxide/anions at low overpotentials and the
inability of Cu to adsorb O2 molecules.
It is interesting that although the pzc of
Au is much higher than on Cu, Au(hkl) is
catalytically inactive in acid solution and
the ORR proceeds as a 2e process [105].
This can be attributed to the very low heat
of adsorption of O2 on Au, which is well
established in UHV [23]. Therefore, in acid
solutions, the main difference in the ORR
pathway on Pt(hkl) and Cu(hkl) (ca. 4e
reduction) versus Au(hkl) (2e reduction)
lies rather in the chemical 1G term
than the (1 %ad ) term. While the 1G
term is determining the reaction pathway,
the (1 %ad ) term is determining the
availability of Au sites for O2 adsorption.
Because of the high pzc of Au, desorption
of anions/oxides occurs at more positive
potentials than on Cu and, in turn, the
onset potential for the ORR is more positive
on Au(hkl) than on Cu(hkl) surfaces. What
is inuencing the electrocatalytic activity of
Au(001) in alkaline solution is, however,
still a mystery. Further work is needed
in order to understand the ability of the
Au(001) surface to break the OO bond at
such low overpotentials.
attain surface sensitivity. We have highlighted the applicability of this technique to

studies of the electrochemical interface, in
particular, with the ability to probe changes
in the surface atomic structure as the
electrode potential is changed and different reaction conditions prevail. The X-ray
technique is particularly powerful when
combined with RRDE measurements because RRDE can supply element-specic
adsorption isotherms, which are key to
the interpretation of the X-ray data. In
turn, structural results obtained from Xray diffraction measurements can give
vital insight into the kinetics of electrochemical reactions, such as the oxidationreduction reaction, which are crucially dependent on the exact nature of
the available adsorption sites on an electrode surface.
Several future developments of the
experimental techniques and materials are
anticipated. For example, the study of
bimetallic crystals relevant to applications
in electrocatalysis, nanoparticle systems,
and further applications of spectroscopic
techniques in combination with X-ray
scattering studies. The work highlighted in
this chapter lays the foundation for these
developments.
4.6
Conclusion
Acknowledgment
In this chapter, we have attempted to show

that in situ structural analysis techniques
and traditional electrochemical measurements can be combined to obtain detailed
insight into electrochemical reactions on
single-crystal electrode surfaces. Thus, it
is possible to establish a link between
macroscopic electrochemical phenomena
and atomic-scale surface structures. X-ray
diffraction is a relatively new technique
that takes advantage of the high brightness of synchrotron radiation sources to
This work was supported by the Director, Ofce of Energy Research, Materials
Sciences Division (MSD) of the U.S.
Department of Energy (DOE) under contract no. DE-AC03-76SF00098. Research
was carried out in part at SSRL, which
is funded by the Division of Chemical
Sciences (DCS), U.S. DOE and at the
XMaS UK-CRG beamline at the ESRF,
Grenoble. CAL is grateful for the support of an EPSRC Advanced Research
Fellowship.
References
1. R. Feidenhansl, Surf. Sci. Rep. 1989, 10,
105.
2. P. H. Fuoss, S. Brennan, Annu. Rev. Mater.
Sci. 1990, 20, 365.
3. I. K. Robinson, D. J. Tweet, Rep. Prog. Phys.
1992, 55, 599.
4. I. K. Robinson, Phys. Rev. B 1986, 33, 3830.
5. B. E. Warren, X-ray Diffraction, Dover Publications, New York, 1990.
6. M. G. Samant, M. F. Toney, G. L. Borges
et al., J. Phys. Chem. 1988, 92, 220.
7. B. M. Ocko, J. Wang, A. Davenport et al.,
Phys. Rev. Lett. 1990, 65, 1466.
8. I. M. Tidswell, N. M. Markovic, C. A. Lucas
et al., Phys. Rev. B 1993, 47, 16 542.
9. C. A. Lucas, J. Phys. D: Appl. Phys. 1999, 32,
A198.
10. M. F. Toney, D. G. Wiesler, Acta Crystallogr., Sect. A 1993, 49, 624.
11. C. A. Lucas, N. M. Markovic, P. N. Ross,
Surf. Sci. 2000, 448, 65.
12. C. A. Lucas, N. M. Markovic, B. N. Grgur
et al., Surf. Sci. 2000, 448, 77.
13. E. D. Specht, F. J. Walker, Phys. Rev. B 1993,
47, 13 743.
14. H. Stragier, J. O. Cross, J. J. Rehr et al.,
Phys. Rev. Lett. 1992, 69, 3064.
15. W. J. Albery, M. L. Hitchman, Ring-Disc
Electrodes, Clarendon Press, Oxford, 1971.
1980.
17. D. Kolb, Prog. Surf. Sci. 1996, 51, 109.
18. J. Clavilier, J. Electroanal. Chem. 1980, 107,
211.
19. M. Samant, M. Toney, G. Borges et al., Surf.
Sci. 1988, 193, L29.
20. Y. S. Chu, I. K. Robinson, A. A. Gewirth,
Phys. Rev. B 1997, 55, 7945.
21. O. M. Magnussen, J. Scherer, B. M. Ocko
et al., J. Phys. Chem. 2000, 104, 1222.
22. G. A. Somorjai, M. A. Van Hove, Prog. Surf.
Sci. 1989, 30, 201.
23. G. A. Somorjai, Introduction to Surface
Chemistry and Catalysis, John Wiley & Sons,
New York, 1993.
24. P. A. Thiel, P. J. Estrup in CRC Handbook
of Surface Imaging and Visualization (Ed.:
A. T. Hubbard), CRC Press, Boca Raton,
Fla., 1995.
25. R. I. Masel, Principles of Adsorption and

Reaction on Solid Surfaces, John Wiley &
Sons, New York, 1996.
26. M. A. Van Hove, K. Hermann, P. R. Watson
et al., Surf. Rev. Lett. 1999, 6, 805.
27. A. R. Sandy, S. G. J. Mochrie, D. M. Zehner
et al., Phys. Rev. B 1991, 43, 4667.
28. K.-M. Ho, K. P. Bohnen, Phys. Rev. Lett.
1987, 59, 1833.
29. I. K. Robinson, P. J. Eng, C. Romainczyk
et al., Phys. Rev. B 1993, 47, 10 700.
30. B. M. Ocko, D. Gibbs, K. G. Huang et al.,
Phys. Rev. B 1991, 44, 6429.
31. D. Gibbs, B. M. Ocko, D. M. Zehner et al.,
Phys. Rev. B 1990, 42, 7330.
32. N. Takeuchi, C. T. Chan, K. M. Ho, Phys.
Rev. Lett. 1989, 63, 1273.
33. R. M. Ishikawa, A. T. Hubbard, J. Electroanal. Chem. 1976, 69, 317.
34. E. Yeager, W. E. OGrady, M. Woo, J. Electrochem. Soc. 1978, 125, 348.
35. I. M. Tidswell, N. M. Markovic, P. N. Ross,
36. J. Wang, B. M. Ocko, A. J. Davenport et al.,
Science 1992, 255, 1416.
37. K. P. Bohnen, D. M. Kolb, Surf. Sci. 1998,
407, L629.
38. F. T. Wagner, P. N. Ross, J. Electroanal.
Chem. 1983, 150, 141.
39. M. S. Zei, N. Batina, D. M. Kolb, Surf. Sci.
1994, 306, L519.
40. I. M. Tidswell, N. M. Markovic, P. N. Ross,
Phys. Rev. Lett. 1993, 71, 1601.
Phys. Rev. Lett. 1996, 77, 4922.
42. N. M. Markovic, B. N. Grgur, C. A. Lucas
et al., Surf. Sci. 1997, 384, L805.
43. I. K. Robinson, Y. Ku, L. C. Feldman, Phys.
Rev. B 1984, 29, 4762.
44. T. Gritsch, D. Coulman, R. J. Behm, G. Ertl,
Phys. Rev. Lett. 1989, 63, 1086.
45. O. M. Magnussen, J. Wiechers, R. J. Behm,
Surf. Sci. 1993, 289, 139.
46. B. M. Ocko, G. Helgeson, B. Schardt et al.,
Phys. Rev. Lett. 1992, 69, 3350.
47. C. L. Fu, K. M. Ho, Phys. Rev. Lett. 1989, 63,
1617.
48. P. J. Feibelman, Surf. Sci. 1996, 360, 297.
49. M. F. Toney, J. N. Howard, J. Richer et al.,
Nature 1994, 368, 444.
Surf. Sci. 1995, 335, 326.
357
358

51. R. J. Nichols, T. Nouar, C. A. Lucas,
W. Haiss, W. A. Hofer, Surf. Sci. 2002, 513,
263271.
52. E. Vlieg, I. K. Robinson and K. Kern, Surf.
Sci. 1990, 233, 248.
53. M. W. Finnis, V. Heine, J. Phys. F 1974, 4,
L37.
54. E. Kirsten et al., Surf. Sci. 1990, 231, L183.
55. O. M. Magnussen, B. M. Ocko, J. X. Wang
et al., J. Phys. Chem. 1996, 100, 5500.
56. B. M. Ocko, O. M. Magnussen, J. X. Wang
et al., Physica B 1996, 221, 238.
57. B. M. Ocko, O. M. Magnussen, J. X. Wang
et al., Phys. Rev. B 1996, 53, R7654.
58. E. Bertel, E. P. Netzer, Surf. Sci. 1980, 97,
409.
59. H. A. Gasteiger, N. M. Markovic, P. N.
Ross, Langmuir 1996, 12, 1414.
Surf. Sci. 1996, 340, L949.
Phys. Rev. B 1997, 55, 7964.
62. P. Zeppenfeld, U. Becher, K. Kern et al.,
Phys. Rev. B 1992, 45, 5179.
63. N. M. Markovic, C. A. Lucas, H. A. Gasteiger et al., Surf. Sci. 1996, 365, 229.
64. A. M. Bittner, B. Wintterline, B. Beran et al.,
Surf. Sci. 2001, 335, 291.
65. D. M. Kolb in Advances in Electrochemistry and Electrochemical Engineering (Eds.:
H. Gerischer, C. W. Tobias), John Wiley &
66. R. R. Adzic in Advances in Electrochemistry and Electrochemical Engineering (Eds.:
H. Gerischer, C. W. Tobias), John Wiley &
Phys. Rev. Lett. 1995, 75, 4472.
68. Z. Shi, J. Lipkowski, J. Electroanal. Chem.
1994, 364, 289.
69. O. M. Magnussen, J. Hotlos, R. J. Nichols
et al., Phys. Rev. Lett. 1990, 64, 2929.
70. J. H. White, H. D. Abruna, J. Phys. Chem.
1990, 94, 894.
71. R. Michaelis, M. S. Zei, R. S. Zhai et al., J.
72. R. Durand, R. Faure, D. Aberdam et al.,
73. N. Markovic, P. N. Ross, Langmuir 1993, 9,
580.
74. N. M. Markovic, H. A. Gasteiger, P. N.
Ross, Langmuir 1995, 11, 4098.
75. R. Gomez, H. S. Yee, G. M. Bommarito
et al., Surf. Sci. 1995, 335, 101.
76. T. Abe, G. M. Swain, K. Sashikata et al., J.

77. N. M. Markovic, C. Lucas, H. A. Gasteiger
et al., Surf. Sci. 1997, 372, 239.
78. A. C. Finnefrock, L. J. Buller, K. L. Ringland
et al., J. Am. Chem. Soc. 1997, 119,
11 703.
79. N. M. Markovic, H. Gasteiger, C. Lucas
et al., Surf. Sci. 1995, 335, 91.
80. I. M. Tidswell, C. A. Lucas, N. M. Markovic
et al., Phys. Rev. B 1995, 51, 10 205.
Phys. Rev. B 1997, 56, 3651.
Phys. Rev. B 1998, 57, 13 184.
83. S. Wu, Z. Shi, J. Lipkowski et al., J. Phys.
Chem. 1997, 101, 10 310.
84. K. Kinoshita, Electrochemical Oxygen Technology, John Wiley & Sons, New York,
1992.
85. H. Wroblowa, Y. C. Pan, J. Razumney, J.
86. M. R. Tarasevich, A. Sadkowski, E. Yeager
in Comprehensive Treatise in Electrochemistry (Eds.: J. O. M. Bockris, B. E. Conway,
E. Yeager et al.), Plenum Press, New York,
1983.
87. N. M. Markovic, H. A. Gasteiger, B. N.
Grgur et al., J. Electroanal. Chem. 1999, 467,
157.
88. F. El Kadiri, R. Faure, R. Durand, J. Electroanal. Chem. 1991, 301, 177.
89. N. M. Markovic, R. R. Adzic, B. D. Cahan et al., J. Electroanal. Chem. 1994, 377,
249.
Ross, J. Phys. Chem. 1995, 99, 3411.
Ross, J. Phys. Chem. 1996, 100, 6715.
92. V. Stamenkovic, N. M. Markovic, P. N.
Ross Jr, J. Electroanal. Chem. 2000, 500,
44.
93. B. N. Grgur, N. M. Markovic, P. N. Ross Jr,
Can. J. Chem. 1997, 75, 1465.
94. T. J. Schmidt, B. N. Grgur, N. M. Markovic et al., J. Electroanal. Chem. 2001, 500,
36.
95. A. Damjanovic in Comprehensive Treatise
in Electrochemistry (Eds.: J. O. M. Bockris,
B. E. Conway), Plenum Press, New York,
1969.
96. M. R. Tarasevich, Electrochimiya 1973, 9,
578.

97. N. M. Markovic, P. N. Ross Jr in Interfacial Electrochemistry-Theory, Experiments and
Applicatoions (Ed.: A. Wieckowski), Marcal
Dekker, New York, 1999.
98. N. M. Markovic, B. N. Grgur, C. A. Lucas,
P. N. Ross, Electrochim. Acta. 1998, 44,
10091017.
99. G. Kokkinidis, D. Jannakoudakis, J. Electroanal. Chem. 1984, 162, 163.
100. S. A. S. Machado, A. A. Tanaka, E. R. Gonzales, Electrochim. Acta 1991, 36, 1325.
101. G. M. Brisard, N. Bertrand, P. N. Ross et al.,
102. M. Wilms, P. Broekmann, C. Stuhlmann
et al., Surf. Sci. 1998, 416, 121.
103. W. H. Li, R. J. Nichols, J. Electroanal. Chem.
1998, 456, 153.
104. R. R. Adzic, N. M. Markovic, V. B. Vesovic,

J. Electroanal. Chem. 1984, 165, 105, 121.
105. J. D. E. McIntyre, Peck in The Physics
and Chemistry of Electrocatalysis (Eds.:
J. D. E. McIntyre, M. J. Weaver, E. Yeager),
The Electrochemical Society, Pennington,
N.J., 1984.
106. R. R. Adzic in Electrocatalysis (Eds.: J. Lipkowski, P. N. Ross), Wiley-VCH, New York,
1998.
107. G. Ertl, N. Neumann, K. M. Streit, Surf. Sci.
1977, 64, 393.
108. P. A. Thiel, R. J. Behm, P. R. Norton et al.,
J. Chem. Phys. 1983, 78, 7448.
109. W. T. Lee, L. Ford, P. Blowers et al., Surf.
Sci. 1998, 416, 141.
359
360
4.2
Theoretical Aspects Associated with

Charge-transfer Kinetics across Interfaces
between Two Immiscible Electrolyte
Solutions
David J. Fermn, Henrik Jensen, and Hubert
H. Girault
Ecole Polytechnique Federale de Lausanne,
Lausanne, Switzerland
4.2.1
Introduction and Scope
Charge-transfer reactions at interfaces between two immiscible electrolyte solutions (ITIES) have been the subject of
a tremendous amount of research from
an experimental and theoretical point of
view in the last thirty years. The relevance
of this kind of molecular interfaces has
been addressed in various review articles
and books, expanding from fundamental
charge transfer at interfaces to drug delivery in biological systems, ion-selective
electrodes, sensors, articial photosynthesis, and so forth [15]. The scope of this
chapter is to review the theoretical models
developed in recent years for rationalizing
the kinetic behavior of electron and ion
transfer across ITIES. This chapter does
not intend to provide an exhaustive account of experimental developments and
specic models but an overview of the
more general theoretical descriptions in a
critical fashion. In doing so, we have made
an attempt to present the basic description
of each model employing a homogenized
terminology.
The discussions will be mostly centered
on ideal systems in which the reactants
are considered as particles with given
redox properties and solvation energies
present at the interfacial region between
two dielectric media. In the case of
electron transfer (ET), we shall concentrate
on models within the framework of the

transition state theory. On the other hand,
ion transfer will be examined not only from
the viewpoint of charge transfer in polar
media but also in terms of ion transport in
heterogeneous systems.
4.2.2
Heterogeneous Electron Transfer across

ITIES
ET reactions across a liquid|liquid junction are undergoing a renaissance as a

result of the increasing interest in molecular electrochemistry. The pioneering work
by Samec [6, 7], Kharkats [8], Girault and
Schiffrin [9], and later by Marcus [1012]
established fundamental aspects that experimental evaluation is yet to ascertain.
For instance, the dependence of the bimolecular ET rate constant on the Galvani
potential difference and the distance separating the redox species are among the
most controversial issues in this area.
Throughout the 1980s and early 1990s,
few experimental works were reported
on the dynamics of ET as probed by
cyclic voltammetry and electrochemical
impedance spectroscopy employing fourelectrode potentiostats [1321]. However,
the analysis of these data was rather limited
because of the primitive knowledge of the
structure and potential distribution across
ITIES. More recently, the introduction of
spectroelectrochemical techniques, Scanning Electrochemical Microscopy (SECM),
and dynamic photoelectrochemistry has
provided new insights into the correlation
between structure and reactivity at these
interfaces. As a result, not only has our understanding of the fundamental properties
of ITIES signicantly increased but we are
also able to envisage novel potential applications as well as rationalize important
aspects concerning life-sustaining redox
4.2 Charge-transfer Kinetics across Interfaces between Two Immiscible Electrolyte Solutions
biological processes [1]. In this section, we

shall highlight fundamental aspects in relation to the energetic and dynamics of
heterogeneous ET bringing about some
relevant examples.
ferricenium/ferrocene (Fc+ /Fc) has been

employed in numerous studies in organic
solvents as a reference potential [22]. In
the case of the water|DCE junction, the
following reaction can be envisaged:
Thermodynamic Aspects in
Connection to Heterogeneous ET
The thermodynamic driving force for an
ET process across liquid|liquid junctions is
determined by the difference in the redox
potential of both species as well as the
Galvani potential difference. Considering
the general reaction
+
+ Fc+
(DCE) H(w) + Fc(DCE)
(4)
where the subscript w and DCE correspond to the water and 1,2-dichloroethane
electrolyte phases, respectively. In order to
estimate the formal Gibbs energy for Reaction (4), the following thermodynamic
cycle can be introduced:
kobs
+
Fc+
(DCE) Fc(w)
4.2.2.1
Ow
1
+ R2
kobs
Rw
1
+ O2
(1)
the equality in the electrochemical potential determines that the ratio of the
activities of the redox species is linked to
the Galvani potential difference by
w
cO1 cR2
RT
w
w
=
ln
(2)
o
o et
w c
nF
cR1
O2
where the formal potential, w

o et , is
determined by

,
,w

w
EO1/R1
o et = EO2/R2
SHE
SHE
(3)
By denition, both redox potentials in
Eq. (3) have to be referred to a common
reference potential, in this case that of the
standard hydrogen electrode (SHE). Obviously, some thermodynamic assumptions
have to be introduced in order to evalu ,
ate [EO2/R2
]SHE , which denotes the formal
redox potential of a redox couple in the
organic phase versus the SHE in water.
A simple approach to this problem
,
involves the estimation of [EO2/R2
]SHE for
a given redox couple that can be further
used as a reference potential in the organic
phase. For instance, the redox couple

1
2 H2(w)
1
2 H2(w)
(5)
+
+ Fc+
(w) H(w) + Fc(w) (6)
Fc(w) Fc(DCE)
(7)
From Eqs. (4 to 7), it follows that

,w

EFc,+ /Fc
= w
o Fc+ + EFc+ /Fc
SHE
SHE
w
GFc
(8)
where the three terms correspond to the

formal transfer potential of ferricinium,
the formal redox potential of Fc+ /Fc in water versus the SHE, and the Gibbs energy
associated with the partition of ferrocene
between the two phases, respectively. Independent evaluation of the three con
stants provided EFc,+ /Fc = 0.638 0.020 V
versus SHE [5]. Taking this value as a reference, the redox potential of several couples
in DCE often found in the literature can
be summarized as exemplied in Table 1.
These data allow predicting whether the
formal ET potential as dened by Eq. (3)
is effectively located within the ideally
polarizable window. This condition, also
referred to as redox potential matching,
is essential for studying ET processes at
ITIES under potentiostatic control.
361

Formal redox potential versus SHE in
water for species often employed in
electrochemical studies at the
water|1,2-dichloroethane interface
Tab. 1
[E ,DCE ]SHE /V
Redox couple
Decamethylferrocene+/0
Dimethylferrocene+/0
Diferrocenylethane+/0
Butylferrocene+/0
Ferrocene+/0
Trianysilamine+/0
Tri(bromophenyl)amine+/0
Tetraphenylborate0/
Tetrakis(4-chlorophenyl)borate0/
Tetracyanoquinodimethane0/
Chloranil0/
Benzoquinone0/
0.07
0.55
0.55
0.56
0.64
0.76
1.10
0.93
1.15
0.29
0.17
0.43
The condition of redox potential matching is illustrated in Fig. 1. Considering the

concentration ratio of ferri/ferrocyanide
indicated in the cell representation and
,
EFe(CN)
3 /Fe(CN) 4 as 0.36 0.01 V, the
6
value of w
o et in the presence of Fc
should be located close to 0.16 V. The
cyclic voltammogram features a response
with a half-wave transfer potential that co
incides with w
o et . The signal observed
at more negative potentials corresponds
to the transfer of tetrapropylammonium
employed as internal reference for the estimation of the Galvani potential difference

(w
o TPA+ = 0.093 V [23]). It should be
noted that the half-wave potential associated with the ET across the ITIES is
not necessarily correlated to the formal

Cyclic voltammogram at the

water|1,2-dichloroethane interface in the presence of
ferricyanide/ferrocyanide couple and ferrocene (Fc) at
10 mV s1 . This experiment was performed with a
four-electrode potentiostat, while the electrochemical
cell can be represented as
Fig. 1
Li2 SO4
1.5 M
Fe(CN)6
0.1 M
Ag Ag2 SO4 Fe(CN)6 4 0.01 M

TPA+
104 M
(aq)
40
TPA+
20
BTPPACl
0.01 M
0.001 M
AgCl Ag
0.01 M
20
ET
40
60
80
0.2
BTPPATPBCl
Fc 4 103 M LiCl
106 i
[A]
362
0.0
0.2
w
of
[V]
(DCE)
(aq)
where BTPPA and TPBCl stand for

bis(triphenylphosphoranylidene)-ammonium and
tetrakis-(4-chlorophenyl) borate, respectively. The
surface area of the liquid|liquid junction was 1.53 cm2 .
+
0.4 The cation tetrapropylammonium (TPA ) was
introduced in the aqueous as a reference for the
Galvani potential difference. (Reprinted from Ref. [5]
with permission from Marcel Dekker, INC.)
ET potential. This process involves the

diffusion of four species, consequently,
if the concentration of the reactants in
both phases is similar, then rather large
peak-to-peak separations could be observed [24]. In the particular case of Fig. 1,
the redox couple in the aqueous phase is
largely in excess with respect to the organic phase, therefore its concentration
can be taken as constant and the ET
rate is only controlled by the diffusion of
tetracyanoquinodimethane (TCNQ). Under these conditions, commonly referred
to as the constant composition approximation, the aqueous phase behaves as a
solid electrode with the Fermi level dened by the electrochemical potential of
the ferri/ferrocyanide couple.
Another well-studied redox system at
liquid|liquid interfaces involves the electron acceptor TCNQ [14, 15, 25, 26]. Cyclic
voltammograms obtained in the presence

of an excess of ferri/ferrocyanide are exemplied in Fig. 2(a). As expected for a
quasi-reversible ET process, the peak current increases linearly with the square root
of the scan rate. The spectral changes
obtained in situ by Total Internal Reection from the DCE phase are also
contrasted in Fig. 2(b) as a function of the
Galvani potential difference. The spectra
evolution shows the bands associated with
the radical TCNQ generated from the
heterogeneous ET reaction. Indeed, the
differential spectroscopic signal at 670 nm
exhibits the same potential dependence of
the current [26]. This close correlation between electrochemical and spectroscopic
behavior indicates that the dynamic of
ET can be evaluated by both approaches.
Furthermore, consistent data for this system has also been gathered employing
180
106 i
[A]
Cyclic voltammograms of the

heterogeneous reduction of
tetracyanoquinodimethane (TCNQ) by
the ferricyanide/ferrocyanide couple
across the water|DCE interface at
various scan rates (a). The composition
of the electrochemical cell is as in Fig. 1
but with a ratio 0.01/0.4 M
Fe(CN)6 3 /Fe(CN)6 4 and 6 104 M
TCNQ in DCE. Absorption spectra
obtained in total internal reection from
the DCE phase as the potential is swept
from 0.190 to 0.260 V, with a difference
40 mV between each spectrum (b). The
spectra show the main features of the
radical TCNQ as the potential is
swept toward the reduction potential.
(Fig. 2a was reprinted from Ref. [25]
with permission from Royal Society of
Chemistry. Fig. 2b was reprinted from
Ref. [26] with permission from
Elsevier Science.)
Fig. 2
180
0.24
v: 49, 36, 25, 16, 9 mV/s

0.14
0.04
0.06
0.16
0.26
w
o f
[V]
(a)
Absorption
2.0
1.5
1.0
0.5
0.0
600
(b)
700
800
Wavelength
[nm]
900
1000
363
364
techniques based on SECM at nonpolarizable interfaces [27, 28]. Consequently, we

shall adopt this system as an example for
some of the discussions in connection to
fundamental aspects on ET.
Electron-transfer Kinetics
As in any interfacial redox process, the rate
of heterogenous ET across ITIES is determined by the interfacial concentration of
the reactants and the phenomenological
bimolecular ET rate constant (kobs ). Expressions concerning the mass transport
of species involved in the redox processes
have been addressed for a variety of geometries, for example, planar diffusion [24],
spherical diffusion at microinterfaces [29],
as well as for approach curves familiar
from techniques based on SECM [3035].
In this section, we shall concentrate instead on the physical aspects underlying
the rate constant of heterogeneous ET. In
particular, the dependence of kobs on the
thermodynamic driving force as well as
the nature of the activation energy will be
highlighted.
4.2.2.2
The Phenomenological Butler

Volmer Expression Considering symmetric energy surfaces for the initial and
nal states of the reaction described in
Eq. (1), the activation energy will be affected by a fraction obs of the Galvani
potential difference
4.2.2.2.1
w
Gact = Gact,eq obs (w
o o eq )
(9)
where Gact,eq and w
correspond
to
o eq
the Gibbs activation energy and the Galvani potential difference at the equilibrium
state. The effect on the ET rate constants
can be phenomenologically described as

obs nFw

o
kobs = kobs exp
(10)
RT
and

(1 obs )nFw

o
kobs = kobs exp
RT
(11)
where the preexponential parameters can
be evaluated at the standard potential

obs nFw
o
k = kobs exp
RT

(1 obs )nFw
o
= kobs exp
RT
(12)
From Eqs. (10 to 12), the faradaic current
as a function of the applied potential can
be expressed in terms of the concentration
of each of the reactants at the respective
interfacial reaction planes (position along
the z-axis),

w
I = nFAkobs cO1
(zO1 )cR2 (zR2 )

w
obs nF (w
o o )
exp
RT

w
(zR1 )cO2 (zO2 )
nFAkobs cR1

w
(1 obs )nF (w
o o )
exp
RT
(13)
Equation (13) can be rewritten in terms of
w
the overpotential = w
o o eq , as
w (z )c (z )
cO1
O1 R2 R2
w c
cO1
R2

obs nF
exp
RT
I = I0
w (z )c (z )
cR1
R1 O2 O2
w c
cR1
O2

(1 obs )nF
exp
(14)
RT
where
w
w
I0 = nFAkobs [cO1
cR2 ]1obs [cO1
cR2 ]obs
(15)
If we assume that the concentrations at
the reaction planes are equal to that in
the bulk, Eq. (14) is simplied to the
corresponding ButlerVolmer expression
for ET at ITIES,

obs nF
I = I0 exp
RT

(1 obs )nF
exp
(16)
RT
This simplistic derivation opens up some
important questions that we shall address
in the next sections. Firstly, the observed
ET rate constant needs to be considered in
terms of microscopic parameters such as
the distance separating the redox species,
the adiabatic properties of the systems,
and the thermodynamic driving force. If
the reactants feature ionic species, the concentration proles at the interfaces arising
from polarization effects can introduce
substantial deviations from Eq. (16). Another parameter to be dealt with from rst
basis is obs . Does this parameter only reect the fraction of the Galvani potential
difference acting on the Gact as indicated
in Eq. (9)?
The Rate Constant of Electron
Transfer We shall commence by assuming that the liquid|liquid interface can be
described in terms of a sharp boundary between water and a lower dielectric
medium. The single ET process of Eq. (1)
involves the approaching of the reactants
to the interface as schematically represented in Fig. 3. The observed rate of ET is
given by
4.2.2.2.2
w
cR2
jobs = kobs cO1
where kobs is the phenomenological bimolecular ET rate constant (units of

cm4 mol1 s1 ). This rate constant effectively corresponds to an integral overall
distance l and the characteristic angles
dening the position of the reactants at
the interface (Fig. 3). Following a Marcus
zR2
(17)
aR2
zR2
aR2
l
l
qmax
z=0
z=0
q
a O1
zO1
(a)
a O1
(b)
(a) Schematic representation of the coordinates for the calculations of the hypervolume
as described by Marcus in Ref. [10]. (b) The maximum value of the angle for a given zR2 and l
arrangement.
Fig. 3
365
366
type of analysis [7, 912],

Gact
kobs = exp
kB T
(18)
where is the transmission coefcient,

is the frequency of molecular motion,
and is the volume per unit of area
in which the reactants are separated by
a distance (l + l). l is dened as the
region where the contribution to the total
ET is maximized. As discussed later, the
activation free energy Gact is dependent
on the reorganization energy term as
well as the work terms associated with
approaching reactants and separating the
products. If we consider the arrangement
depicted in Fig. 3, it follows that [10]

=
l=aR2 +aO1
laO1
zo =aR2
max (zo ,l) 2

=0
l2
l=aR2 +aO1
k(l)dl
k(lmax )
= 2(aO1 + aR2 ) 3
(19)
where corresponds to the angle along the

plane of the interface. Assuming that the
ions cannot penetrate the phase boundary,
max can be dened as the maximum angle
obtained for a given value of z1 and l
as depicted in Fig. 3(b). In Eq. (19), the
parameter k(l) is given by

Gact (l)
(20)
k(l) = (l) exp
kB T
and lmax corresponds to the characteristic distance at which the weighting
factor k(l) is maximized. Drawing analogies with the behavior observed for ET
in bulk liquids, Marcus suggested that
k(l) can go through a maximum at
(21)
Equation (21) is associated with the limiting case for a sharp phase boundary
between both liquids. Marcus also estimated the hypervolume for the case in
which both reactants are allowed to penetrate the interfacial region in such a way
that their centers can effectively reach the
liquid boundary. For such conguration,
the limits of the integrations in Eq. (19)
are different and approaches to
(aO1 + aR2 )3 1
=0
k(l)
sin d d dzo dl
k(lmax )

(l aR2 aO1 )2 l
=
a distance close to aO1 + aR2 , and exponentially decrease as l increases for

larger distances. Consequently, taking
k(l)/k(lmax ) = exp[(l aO1 aR2 )], it
follows that the integration in Eq. (19) results in
(22)
If we consider that the sum of the ionic

radii is of the order of 10 A and = 1 A 1 ,
it can be estimated that the hypervolume,
and consequently the rate constant, is
approximately two orders of magnitude
larger for the case in which the ions are
allowed to penetrate the interfacial region.
Marcus proposed that the analysis involving Eqs. (18 to 22) was also valid in
the adiabatic limit. Smith and coworkers criticized this point on the basis of
the fact that the transmission coefcient
exhibits a complex dependent on the distance separating the redox species, and
therefore should be considered in the integration step [36]. Let us consider the
LandauZener transmission coefcient
=
2
(1 + )
(23)
where
= 1 exp(2)
(24)
and
|H |2 3/2
=
2h(kB T )1/2
(25)
where the parameter H corresponds to

electronic coupling matrix element

(l aO1 aR2 )
H = H0 exp
(26)
2
and is the local reorganization energy.
From classical transition state theory, the
LandauZener approximation accounts
for the probability of ET upon multiple
passages through the transition state
region of the energy surface. Equation (23)
is equal to 1 for adiabatic ET processes and
decreases toward the nonadiabatic limit.
On the basis of this spatial dependence
of the transmission coefcient as well as
the Gibbs activation energy (as discussed
in the next section), it is clear that a
more complete model will include the
integration not only of the hypervolume
as implicitly done for Eqs. (21 and 22) but
also of the local ET rate constant and even
the concentration prole of the reactants
at the interface [36].
The meticulous analysis by Smith and
coworkers allows illustrating the effect of
each of the spatially dependent parameters
on the integrand of Eq. (19) [36]. Including
all terms in the integrand, it follows

kobs =
laO1 max (zo ,l) 2
l=aR2 +aO1 zo =aR2 =0
exp
=0

Gact (l) 2
l gR1 (z)gO2 (z)
kB T
sin d d dz dl
(27)
where g is the corresponding concentration prole along the z-axis. Neglecting

distance dependence on the Gibbs energy of activation and the concentration
proles, the solution of Eq. (27) in the
nonadiabatic limit (i.e. = 4) can be

approximated to
kobs 2(aO1 + aR2 ) 3

Gact
exp
kB T
(28)
which is identical in form to the previous

Marcus derivation (Eq. 18) taking the
hypervolume as in Eq. (21). A numerical
comparison between the limiting case
given by Eq. (28) and the full integration of
Eq. (27) is displayed in Fig. 4 as a function
of the parameter . It is observed that
away from the nonadiabatic limit, Eq. (28)
provides bimolecular rate constants one or
two orders of magnitude smaller than for
the complete model.
A further renement introduced by
Smith and coworkers involves the resolution of the general integrand (27) taking
into account the distance dependence of
the reorganization energy. As discussed in
the next section, the solvent reorganization
term is not only dependent on the distance
separating the redox couples but also on
their relative positions with respect to the zaxis. Fig. 5 shows the value of the integrand
calculated from the nonadiabatic limit of
the Marcus expression, as well as from
the general model excluding concentration proles. In general, these simulations
show that the integrand is more sharply
dependent on the distance l in the generalized model. In order to illustrate the
contribution of the various physical elements involved in the ET dynamics, we
shall look closer at the reorganization energy as well as the interfacial distribution
of reactants (which is connected to the
characteristic work terms for approaching
reactants to the interface).
367

1017
3
ks
[cm4s1]
1018
2
1019
1020
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
0.8
1.0
1.2
1.4
(a)
1015
3
1016
ks
[cm4s1]
368
1017
1
1018
1019
0.0
(b)
0.2
0.4
0.6
Calculated electron-transfer rate constants as a function

of the adiabatic parameter for aO1 = aR2 = 3 A and
= 1 eV. The parameter was evaluated as 0.25 and 1 A 1 for
(a) and (b), respectively. Curve 1 was estimated from Eq. (28),
employing the approximation = LZ (l = aO1 + aR2 ). Curve 2
is obtained from the full numerical integration of Eq. (27).
Curve 3 was calculated as in curve 2, but assuming a
concentration prole of the two reactants that increases toward
the interface. (Reprinted from Ref. [36] with permission from
Elsevier Science.)
Fig. 4
The Solvent Reorganization Energy The Gibbs energy of activation already included in Eq. (18) can be described
in terms of the total reorganization energy
4.2.2.2.3
(), the concentration independent term of

the Gibbs energy of ET (Get ), as well as
work terms for approaching reactants (wp )
and separating products (ws ) to and from
Full integrand (C(z) assumed bulk)

all curves divided by 3 peak height
1.0
z=3
0.8
z = 3.25
0.6
Integrand curves
for z = 3 to 4.75
in 0.25 increments
Ho = 0.01 eV
z = 3.5
0.4
0.2
0.0
0
(a)
3
R a1 a2
Marcus integrand (C(z) assumed bulk)

all curves divided by 3 peak height
1.2
z=3
1.0
z = 3.25
0.8
z = 3.5
Integrand curves
for z = 3 to 4.75
in 0.25 increments
Ho = 0.01 eV
0.6
0.4
0.2
0.0
0
(b)
3
R a1 a2
Fig. 5 Evaluation of the full integrand of Eq. (27) as a function of

the distance l employing Eq. (29) for Gact , Eq. (32) for s , and
Eq. (26) for H (a). Other parameters were evaluated as
op
op
ws = os = 30, w = o = 2, H0 = 0.01 eV, = 1 A 1 , and

Get = 0. The integrand corresponding to Eq. (19) under identical
condition is shown in (b). The dependence of the integrand on the
distance is sharper in (a) than in (b) owing to the inclusion of the
distance dependent of Gact and s . (Reprinted from Ref. [36] with
permission from Elsevier Science.)
369
370
the interface
Get + ws wp
1+
4
Gact =
(29)
where Get is given by

w
Get = nF (w
o o et )
(30)
The term in Eq. (30) corresponds to the

so-called driving force for the ET process.
Considering that reactants and products
involve ionic species, the work terms (wp )
and (ws ) feature a chemical part associated
with their solvation energies and an electrostatic term that is determined by the
local potential distribution. This electrostatic term is rather important as it may
account for part of the potential dependence of the phenomenological ET rate
constant. We shall give a closer look to this
point in the next section.
The contribution of the solvent reorganization energy to the total as considered by
Kharkats [8] and later by Marcus [11]. The
expressions obtained appeared not entirely
consistent, but successive revision established that the key aspect lies in separating
the static and optical terms of the integral
of the electric displacement vectors over
the volume of the two liquids system [37]

1
1

(Df Di )2 dv
s =
20
st
20
Vw Vo
Vw Vo
dielectric media and for the general geometrical arrangement depicted in Fig. 3(a),
provides an expression for s of the form

(ne)2 1
1
1
s =
s
40 oop
o 2aR2

(ne)2 1
1
1
+
op s
40 w
w 2aO1

(ne)2
1
+
40 4zR2

op
op
s
o w
os w
op op
op
s)
o (o + w ) os (os + w

(ne)2
1
+
40 4zO1

op
op
s s
w o
w
o
op op
op s s
(w + os )
w (w + o ) w

(ne)2 2
1
1
op
op s
s
40 l
o + w
o + w
(32)
where op and s are the optical and static
dielectric constants, n is effective number
of electrons transferred, and e is the
op
op
elementary charge. Taking w = o =
s = s = s , Eq. (32) reduces to
op and w
o
the well-known expression obtained for
homogeneous ET,
hm
s
static
(Df Di )2 dv
op

n2
1
1
1
=
+
40 2aO1
2aR2
l

1
1
s
(33)
op
If we take 1,2-dichloroethane as the organic

solvent, Eq. (32) can be simplied to
optical
(31)
where Di and Df are the displacement

vectors for the initial and nal states,
respectively. Integration of Eq. (31), considering a sharp boundary between two
w|DCE
n2
40
0.189 0.275
+
aR2
aO1

0.029 0.014 0.494
+
zR2
zO1
l
(34)
It can be noticed that the dominant factors

in Eq. (34) are related to the ionic size and
the distance separating the redox species.
Assuming 0.5 nm for the ionic radii of both
reactants in a face-to-face conguration
w|DCE
and a total distance l of 1 nm, s
can
be estimated to be of the order of 0.7 eV.
This magnitude is considerably larger than
the common internal reorganization terms
for species in which no substantial changes
are seen in the structures during the redox
process. Therefore, it is expected that the
solvent reorganization energy will provide
the most important contribution to the
total .
Marcus also calculated the chemical part
of the work term for the formation of the
precursor complex [11]

(zO1 )2
(zR2 )2
1
wp =
s
40
4do os
4dw w

s
2 zO1 zR2
os
+
(35)
w
s + s
s + s
w
l w
o
o
It should be noticed that this expression is
obtained for a sharp boundary between
the dielectric; therefore it may not be
directly applicable in the mixed-solvent region model. The equivalent expression for
the successor complex (ws ) can be easily
obtained by replacing the charges of the
reactants by the corresponding products.
For the sake of comparison, if we introduce the parameters employed earlier for
w|DCE
estimating s
, it can be obtained that
they are of the order of 10 and 80 meV,
respectively. For this calculation we have
also assumed that zO1 = zO2 = +1 and
zR1 = zR2 = 0. It can be clearly seen that
wp and ws are expected to be rather small
w|DCE
in comparison to s
, therefore the
activation energy (see Eq. 29) is mostly
controlled by the latter. It should also be
mentioned that Benjamin and Kharkats
have developed expressions for s for all
possible locations of the reactants at the

interface, including the case in which
one of them is allowed to cross from
one phase to the other [38]. This analysis provides a quantitative comparison
between the homogeneous limit (Eq. 33)
and the corresponding heterogeneous case
(Eq. 32).
Potential Dependence of the Heterogeneous Electron-transfer Rate Constant
Some of the earlier works devoted to rationalize the dependence of the observed ET
rate constant on the applied potential were
based on basic models for charge transfer
in polar media. Two of the main issues
arising from these considerations are the
structure of the liquid|liquid boundary and
the distribution of the electrostatic potential. In the pioneering work by Samec in
1979, the potential dependence of kobs was
considered assuming a model involving
two diffuse layers separated by an ion-free
compact layer [7]. The region delimited by
the inner layer establishes the distance for
maximum probability of ET (R ). In this
case, the element R is estimated by the
space volume of the ion of radius aO1
located in one of the reaction plane and
certain molecular volume adjacent to the
phase boundary,
4.2.2.2.4
R = 2aO1 AVm
(36)
The density probability of the precursor

complex, (R ), can be obtained by a
distribution of the type
(R ) = aO1 aR2

zO1 zwO1 + zR2 zoR2
exp
kT
(37)
where zO1 and zR2 correspond to the
position of the edges of the inner layer.
From Eqs. (36 and 37), and considering
371
372
that the ET probability at the precursor

complex is determined by Gact (Eq. 29),
the rate constant of ET can be evaluated as

zO1 zwO1 zR2 zoR2
RT

zO1
nzR2
exp
(38)
RT
pzc
kobs = kobs exp
where
pzc
kobs
= 2aO1 Vm B exp
4RT

w
nFo et
exp
RT
(39)
1 nF zO1

+
(zR2 w
o et ) (40)
2
4
and B involved the overlap integral of

the wave function associated with the
initial and nal states. Although the
preexponential term is not fully developed
as discussed in Sect. 4.2.2.2.2, the whole
quantity can be experimentally accessed by
measurements of the phenomenological
rate constant at the potential of zero
charge (pzc). On the other hand, a very
important implication from Eq. (38) is that
the dependence of the kobs on the Galvani
potential difference may only arise from
changes in the interfacial concentration of
the reactants and not by the Gibbs energy
of activation.
Girault and Schiffrin proposed a model
in which the interface features a mixedsolvent region [9]. To approach this region, the reactants and products undergo
changes in their solvation energies and
ionic atmosphere. Consequently, the work
terms associated with the formation of
the precursor complex and dissociation of
the successor complex also feature some
potential independent parameters,
wp = wp + zO1 FzwO1 + zR2 FzoR2

(41)
ws = ws + zO2 FzoO2 + zR1 FzwR1

(42)
The terms wp and ws are associated with
the changes in solvation energy and ionic
atmosphere experienced by the reactants
and products in the mixed-solvent region.
The second and third terms in Eqs (41
and 42) reect the work involved to
displace the ions in and out of the
electried interfaces. Effectively, zwO1
and zoR2 are the potential drop developed
between the reaction plane and the bulk
of the corresponding electrolyte phase.
In the original analysis by Samec [7],
these electrostatic components provide
the most important contributions to the
work terms. This is simply because of
the assumption that chemical changes in
the ionic environment can be neglected
outside the inner compact layer.
The approach followed by Girault and
Schiffrin [9] was based on the encounter
preequilibrium model, previously employed for ET in condensed phase [39] and
metal|electrolyte interfaces [40]. It follows
from this model,
kobs = Kpe ket
(43)
where Kpe is the equilibrium constant for

forming the precursor complex, which is
given by

wp
e
Kp = R exp
(44)
RT
Similar to the previous consideration, R
corresponds to the reaction zone where
the molecules feature the appropriate
conguration for the ET step. Considering
Eqs. (20, 29, 30, and 41 to 44), an
expression for kobs identical in form
to Eq. (38) can be obtained but with a

preexponential factor of the form
kobs = R eff exp
pzc

exp
4RT

exp
wp
RT
(wp ws + nFw
o et )

far more complex than what is described

by current modeling.
Selected Experimental Results
The rst quantitative evaluations of heterogeneous electron transfer rate constants
were performed employing cyclic voltammetry and electrochemical impedance
spectroscopy at polarizable interfaces [13,
14, 17, 21]. Careful choice of redox couples
and electrolyte composition was essential
in order to avoid interference from coupled ion-transfer processes. From these
early studies, rate constants typically of the
order of 1022 cm4 s1 were measured for
the ET between hexacyanoferrate and various redox couples in the organic phase. A
common aspect to all these measurements
is the redox potential matching, which allows to onset the ET reaction within the
polarizable window commonly obtained
for liquid|liquid junctions. Obviously, this
boundary condition does not permit studying redox processes featuring large values

of Get . As discussed later, dynamic studies far from equilibrium provide very
useful information on fundamental parameters such as the preexponential factor
in Eq. (28) and the reorganization energy
(Eq. 32).
The particular case of hexacyanoferrate
and TCNQ (see Fig. 2) have been studied not only by electrochemical impedance
spectroscopy [14] but also by Potential Modulated Reectance Spectroscopy
(PMR) [26], SECM [42], and Microelectrochemical Measurements at Expanding
Droplets (MEMED) [27, 28]. Considering
that the composition of the electrolyte
varied for all these approaches, rate constants between 1023 and 1022 cm4 s1
were consistently obtained for this system. Cheng and Schiffrin demonstrated
that these gures are consistent with the
basic expressions derived by Marcus for a
4.2.2.3
RT
(45)
here eff is the effective electron hopping
frequency. It should also be noticed that if
the potential independent components of
the work terms are negligible with respect
to the reorganization energy and the thermodynamic driving force, Eq. (45) reduces
to an expression similar to Eq. (39). Apparently, this might be physically plausible if
w|DCE
, wp , and
we compare the values of s
ws as derived from Eqs. (34 and 35).

As we mentioned previously, the potential dependence of kobs can arise from
changes in the Gibbs energy of ET or
from concentration polarization effects.
Schmickler approached this problem by
simulating the potential dependence of
the distribution of ionic species using a
lattice-gas model [41]. These simulations
showed that as far as the distance separating the redox species is smaller than the
Debye length of each space charge region,
the potential dependence would be mostly
controlled by the concentration polarization phenomena. This conclusion appears
more consistent with the mixed-solvent
region model, which allows a close approach between the two reactants. Some
experimental reports have shown deviations from the ButlerVolmer behavior,
which could reect concentration polarization phenomena. However, as reviewed in
the next section, a clear dependence of the
ET rate constant and the Galvani potential
difference have been observed, suggesting
that the potential and ionic distribution are
373

Potential dependence of the
electron-transfer rate constant (kobs )
pzc
normalized to the value at the pzc (kobs )
2.5
for the TCNQ reduction by
2.0
hexacyanoferrate at the water|DCE
interface. Full line was estimated from
1.5
Experimental
Eq. (38), assuming that the potential
drop across the inner layer is negligible
1.0
0 V). Dependence of the
(zzO1
R2
potential
drop across the ET reaction
0.5
plane with the Galvani potential
Model
difference, as obtained from
0.0
0.20 0.15 0.10 0.05 0.00 recalculation of the two trends in (a),
assuming a transfer coefcient of
7
0.5 (b). These results suggest that
6
approximately 30% of the Galvani
potential difference is developed
5
between the reacting species under the
experimental conditions dened in
4
Fig. 2. (Reprinted from Ref. [26] with
3
permission from Elsevier Science.)
Fig. 6
pzc
ln (kobs/kobs)
3.0
(a)
O1
102 zR2
f
[V]
374
2
1
0
0.20 0.15 0.10 0.05
(b)
0.00
wof
[V]
sharp liquid|liquid boundary Eqs. (28 and

32) [14].
The potential dependence of the observed ET rate constant from hexacyanoferrate (II) to TCNQ is shown in Fig. 6 [26].
The rate constants were estimated from
chronoabsorptometric and PMR measurements under the conditions described in
Fig. 2. Deviations from the ButlerVolmer
behavior are evident at potentials away
from the pzc, which were also pointed
out by Cheng and Schiffrin in a previous
paper [14]. The theoretical curve shown in
Fig. 6 illustrates a simulation of the potential dependence of the observed rate
constant normalized to the value at the
pzc employing Eq. (38). The potential drop
between the plane at zO1 and zR2 was initially neglected and only the contribution
associated with concentration polarization
was considered. The interfacial potential

was estimated using a PoissonBoltzman
distribution of the ionic species. The
substantial underestimation of the experimental points can be taken as evidence that
a nite fraction of the applied potential is
developed between the reaction planes zO1
and zR2 . Assuming a transfer coefcient
= 0.5, Fig. 6(b) shows that the local potential could be up to 30% of the total
Galvani potential difference.
The dependence of the observed ET rate
constant on the thermodynamic driving
force indeed remains rather controversial.
Unlike electrochemical and spectroelectrochemical approaches at polarizable interfaces, techniques based on SECM and
dynamic photoelectrochemical measurements have allowed probing rate constants
far away from equilibrium conditions. In
SECM, redox-active species are generated

in the proximity of the interface by an approaching microelectrode. The feedback
current at the microelectrode tip contain
information on the rate of ET at the
liquid|liquid boundary. On the other hand,
the photoinduced ET involving dyes and
redox species at either side of the interface manifest itself as photocurrent responses [5, 4347]. These photoresponses
are proportional to the ratio between the
rate constants of excited state decay and
heterogeneous ET. The common point
between these two approaches is that
highly reactive species are generated at
the interface, either by light or by an approaching microelectrode; therefore large
driving forces for the ET can be attained.
Tsionsky and coworkers studied the
dynamics of ET from various redox couples located in the aqueous phase to the
oxidized zinc tetraphenylporphyrin in benzene by SECM. The driving force was
varied by changing the concentration ratio
of a common ion (ClO4 ) and by using
different redox couples in the aqueous
phase. It was observed that for driving
forces smaller than 0.5 eV, the potential
dependence of the observed rate constant
exhibits a ButlerVolmer behavior. Ignoring concentration polarization effects, this
dependence is consistent with Eq. (38). For
larger driving forces, the observed rate
constant becomes potential independent
suggesting a diffusion control regime. It
should be mentioned that the redox couples in the aqueous phase are in large
excess with respect to the zinc porphyrin
in the organic phase in order to avoid coupling of diffusion proles in both phases
(the constant composition approximation).
Interestingly, a different behavior is observed in the presence of a C-10 surfactant
as depicted in Fig. 7(a). The adsorbed
lipid layer effectively increases the distance
separating the redox couple, therefore increasing the solvent reorganization energy
(see Eq. 32). In this case, a decrease in
kobs with increasing potentials at high values of driving force has been observed.
This phenomenon was taken as rst evidence of the Inverted Marcus Region at a
polarized interface. However, the limited
number of experimental points does not
allow quantifying parameters such as the
reorganization energy term.
Barker and coworkers employed essentially the same approach but reducing the
concentration of the redox couple in the
aqueous phase. Although the analysis of
the approaching curves is further complicated by the coupling of diffusion regimes
in either side of the interface, the most
important aspect is that the overall rate
of ET decreases. Consequently, rate constants of fast reactions, which previously
appeared diffusion-controlled, were accurately measured. This analysis provided a
dependence of the rate constant on the
driving force as illustrated in Fig. 7(b). The
authors rationalized the limited number of
experimental points employing Eqs. (18,
29, and 30), providing values for the reorganization energy and the preexponential
factor of 0.55 eV and 8.3 1020 cm4 s1 ,
respectively. In principle, the value for s is
consistent with Eq. (32), assuming a faceto-face approach and for l aO1 + aR2 .
The previous works based on SECM have
not addressed the specic contributions of
the Galvani potential difference and the
formal ET potential to the thermodynamic
driving force. Our recent work based on
photoinduced ET at externally polarizable interfaces quantitatively tackles this
problem employing the thermodynamic
relations discussed in Sect. 4.2.2.1 [48].
The advantage of this approach is that
the driving force can be changed potentiostatically as well as by changing the redox
375

Fig. 7 Driving force dependence of
kobs obtained for the reduction of
1
ZnPor+ in benzene by various redox
couples in the aqueous phase as probed
2.0
by SECM in the presence of a full
monolayer of C10-lipid (a). As the
2
driving force increases, kobs increases in
2.5
the presence of Fe(CN)6 4 (curve 1),
but decreases for Co (II) sepalchrate
(curve 2) and V2+ (curve 3). A similar
3.0
analysis is presented in (b) but in the
500 absence of the C-10 lipids and for
1500
1000
o
o
substantially smaller concentration of
E + w j
the aqueous redox couple. The curve
[mV]
in (c) was obtained from photocurrent
2.5
measurements at the polarizable
2.0
Fe(CN)63-/4water|DCE interface in the presence of
water-soluble porphyrin dimer
1.5
(ZnTPPS : ZnTMPyP) and ferrocene
1.0
(Fc), dimethylferrocene (DMFc),
FeEDTA-/20.5
butylferrocene (ButylFc),
Mo(CN)63-/4diferrocenylethane (DfcEt), and
0.0
decamethylferrocene (DCMFc). (Figs. a,
0.5
Ru(CN)63-/4b and c were reprinted from Refs. [32,
34], respectively, with permission from
1.0
0 100 200 300 400 500 600 700 800 900 the American Chemical Society.)
E1/2
[mV]
1.5
logk f
[cm s1]
log (k12)
[cm s1 M1]
(a)
(b)
10.0
ket
[M1 s1 cm]
376
1.0
: Fc
: DMFc
: ButylFc
: DFCEt
: DCMFc
0.1
0.01
0.001
0.0
(c)
0.5
1.0
1.5
2.0
Geto'
[eV]
couple acting as quencher in the organic

phase. From the photocurrent analysis, the
phenomenological rate constant of ET as a
function of the driving force is illustrated
in Fig. 7(c). It can be observed that the
ET from the redox couple in the organic
phase to the excited state of the porphyrin
dimer exhibits a rather high reorganization
energy, which shift the inverted region beyond 1 eV. The number of experimental
points obtained from a family of ferrocene
derivatives allows a condent analysis in
terms of Eq. (29). This rather high reorganization energy can also be explained in
terms of Eq. (34), assuming aFc 0.2 nm,
apor 0.7 nm, and l 1 nm.
Finally, it is rather interesting to notice that Figs. 7(b and c) provide preexponential factors of the order of 5
1020 cm4 s1 . Comparing with the simulations in Fig. 4, it can be concluded that
the experimental data is consistent with
a highly nonadiabatic limit ( 0), in
which the expressions derived by Marcus
and Smith and coworkers do converge.
This low electronic coupling between the
redox species may arise from the difference in the solvation properties, which
induces interfacial separation distances of
the order of 1 nm.
activated process and as a pure transport

process. Usually, the ion transport across a
liquid|liquid interface has been described
using the NernstPlanck equation [60, 61]
or the Langevin equation [5759, 65]. The
basic ideas of these independent theories
will be reviewed starting from rst principles and a more generalized approach will
be outlined. In a recent theoretical work the
transport description has been combined
with a microscopic theory on the mechanism of ion transfer [68]. Other models for
ion transfer will be dealt with briey in a
separate paragraph [55, 6466, 79].
4.2.3
4.2.3.1
Ion-transfer Reactions
Ion-transfer reaction at interfaces between

immiscible electrolyte solutions (ITIES)
has been an area of active research for
more than two decades [4952]. Several
research groups have developed theories
dealing with the fundamental features
of ion-transfer processes at (ITIES) [7,
5368]. The theoretical work has often
been initiated by advances in experimental
techniques used to obtain kinetic information on the ion-transfer kinetics [53,
6974]. More recently, computer simulations have given important information
on the microscopic surface structure [51,
57, 7578]. The microscopic nature of
the ion-transfer event has been discussed
in the theoretical works on ion transfer, and different factors that are potentially important for the ion-transfer event
at ITIES have been suggested: elasticity of the inner layer, surface structure
in general, the electric potential energy
(dependent on the Galvani potential),
hydrodynamic friction, and changes in
solvation.
There are two main approaches to describe ion-transfer reactions, that is, as an
Kinetics of Ion-transfer Reactions

Based on the Assumption of an Activated
Ion-transfer Event
The simplest kinetic approach is to consider an ion-transfer reaction as an activated process as schematically shown in
Fig. 8, but we shall consider the transfer
of the ion across the complete interface
at which the Galvani potential difference between the two phases varies. This
derivation may be termed the global approach.
For this simple kinetic model, the
Gibbs energy in the adjacent phases is
dened by the standard electrochemical
potential, which can be dened as the
concentration independent part of the
electrochemical potential.
The current is dened by

I = zFA k cw k c
(46)
where k and k are the forward and

backward global rate constants with units
of cm s1 . The bulk concentrations are
given by cw and c . The charge of the
ion is z, A is the surface area, and F is
the Faraday constant. At equilibrium, the
Galvani potential difference is equal to the
standard transfer potential dened by
377

Interface
Gibbs
energy
378
m0, w + zFfw
RT ln(co/cw)
m0, o + zFfo
Reaction coordinate
Standard Gibbs energy prole for global ion-transfer
reactions.
Fig. 8
w
o i =
o,w
Gtr,i
zi F
(47)
Go,w
tr,i
is the standard Gibbs

where
energy of transfer from water to the
organic phase. At the standard transfer
potential the activation energy barrier is
regarded as symmetrical. In the standard
case at equilibrium we have cw = c = c
w
and w
o eq = o . When we apply a
Galvani potential difference different from
the standard transfer value, only a fraction,
, of the driving force is active on lowering
the energy barrier:

w
w
I = zFAk c ezF (o o )

w
w
ezF (1)(o o )
(48)
In the general case, this equation becomes

w
w
I = zFAk cw ezF (o o )

w
w
c ezF (1)(o o ) (49)

If this simple kinetic approach can be
used as a primary model, it is difcult
to ascribe a physical meaning to the rate
constants. Indeed, these rate constants
correspond to the transfer of an ion across
the entire interfacial region composed of

two back-to-back diffuse layers. The only
merit of this global approach is to yield
an expression that in essence can be
considered as a ButlerVolmer equation
for an ion-transfer reaction. Many groups
have observed this ButlerVolmer behavior, as originally suggested by Gavach [53].
Figure 9 illustrates the results obtained by
means of chronocoulometry for the transfer of acetylcholine.
Interfacial Reaction Independent of
the Local Electric Field
Another way to treat an ion-transfer
reaction is to consider the reaction as
an elementary jump across a thin
interfacial layer, which we will call the
reaction plane, but it may also be called
inner layer or mixed-solvent layer
depending on the interfacial model used.
The interface (about 1 nm thick) comprises
back-to-back diffuse layers.
With the notation of Fig. 10, the current
now reads

I = zFA k ca k cb
(50)
4.2.3.2
where k and k represent the rate

constants of the ion-transfer reaction
0.5
log k
[cm s1]
a = 0.48
1.0
a = 0.53
1.5
150
100
50
50
(wof wof)
[mV]
Tafel plots for ion transfer of acetylcholine. From water to DCE () and from DCE to water ().
Reaction plane
Gibbs
energy
Fig. 9
m0,w + zFfa
a
RT ln(c b/c a)
m0,o + zFfb
b
Reaction coordinate
Standard Gibbs energy prole for an elementary
ion-transfer reaction.
Fig. 10
Ion
in water
Scheme 1
Preequilibrium
Ion at
interface
in position a
Transfer
Ion at
interface
in position b
Preequilibrium
Elementary steps for kinetically controlled ion transfer.
Ion in
organic phase
379
380
across the reaction plane and where

the concentrations ca and cb are the
concentrations on both sides of the
reaction plane. In a rst approximation,
we can assume that the rate constants are
potential-independent and that the only
consequence of the interfacial polarization
is to vary the interfacial ion concentration
as described by Boltzmann statistics:
ca = cw ezFw /RT
a
cb = c ezFo /RT
b
(51)
(52)
In the application of these two equations it

is assumed that there is a preequilibrium
between the reaction plane and the
bulk regions.
It is important to realize that the application of the above scheme and Eqs. (51 and
52) are based on the assumption that the
ion transfer is slow compared to diffusion
of ions to a and b from bulk solutions. By
substitution in Eq. (50), we obtain

a
I = zFA k cw ezFw /RT

b

k c ezFo /RT
(53)
Furthermore, if the positions a and b are
close, we can assume that the electrical
potential difference between these two
points is negligible. In this case, Eq. (53)
reduces to

w
I = zFA k cw ezFo /RT

w

k c e(1)zFo /RT (54)
The proportion of the Galvani potential
difference occurring in the aqueous phase
is termed . At equilibrium, at the standard
transfer potential, we have
w
zFwo
RT = k e(1)zFo /RT
ke
=k
(55)
If we further assume that does not

vary with w
o (which is a very rough
approximation), the following expression
for the current ensues:

w
w
I = zFAk cw ezF (o o )/RT

w
w
c e(1)zF (o o )/RT (56)

This result is similar to what was obtained
in the simple approach (Eq. 49). The
fundamental difference is that now it is
possible to relate k 0 to the microscopic
features of the ion transfer.
Interfacial Reaction Dependent on
the Electric Field
Still considering an ion-transfer reaction
as an interfacial process, Eq. (50) remains.
However, we will now assume that the rate
constants depend on the applied Galvani
potential difference according to
4.2.3.3
k = k0 ezFb /RT
a
k = k0 e(1)zFb /RT
(57)
(58)
In these equations it is assumed that

transition state theory can be applied
in the description of the ion transfer.
By substitution in Eq. (50), the following
equation is obtained:

a
I = zFA k0 ca ezFb /RT

a
k0 cb e(1)zFb /RT (59)

If we still assume that there are equilibriums between the interfacial position and
the bulk, we have

b
a
I = zFA k0 cw ezFa /RT ezFw /RT

b
b

k0 c e(1)zFa /RT ezFo /RT
(60)
We can dene the potential # such that

#a = ba
(61)
b# = (1 )ba
(62)
Equation (60) then reads

#
I = zFA k0 cw ezFw /RT

k0 c ezF# /RT
(63)
The coefcient can be dened as
#w = w
(64)
#
(65)
(1 )w
Equation (63) can now be written as

w
I = zFA k0 cw ezFo /RT

w

k0 c e(1)zFo /RT (66)
At the formal standard transfer potential,
we have
w
zFwo /RT
= k0 e(1)zFo /RT
k0 e
=k
(67)
By dening k in this way, we recover

an expression of the current similar to
Eq. (49). This approach may be compared
with the of work Koryta [54] and Samec [7].
In the latter, an activation energy and a
solvent reorganization energy are explicitly associated with the rate constant of
ion transfer. (The microscopic transport
of ions across the ITIES may also be
described using the Langevin equation,
see Refs. [58, 59], which relates the microscopic rate of particle motion, v, of a
particle of mass m with coordinate x and
time t, to the regular potential of the eld
U (x, t), a phenomenological friction coefcient, , and a function A(x, t) describing
the random force caused by uctuations in

the condensed medium:

dv
1 dU (x, t)
= v
+ A(x, t)
dt
m
dx
The derivative of U with respect to position is the regular force acting on the
particle. However, this so-called stochastic
approach results in similar NernstPlanck
type equations albeit with microscopic
diffusion coefcient differing from the traditional macroscopic diffusion coefcient.
The microscopic diffusion coefcient, D,
is related to , through D = (kB T )/(m).)
Kinetics of Ion-transfer Reactions
Based on a Transport Description
Another way to treat ion-transfer reactions
is to consider them as a special case of ionic
conductivity across an interface. From a
phenomenological viewpoint, anionic ux
can be written [80]:
4.2.3.4
Ji = ci u i grad i
(68)
By taking into account that the electrochemical potential can be considered as

the sum of a standard term, which is solvent dependent, a concentration/activity
term, and an electrical term, we have
Ji (x) = ci u i (x)[gradi (x)

+ RT grad ln ci (x)
+ zi F grad(x)]
(69)
which is identied as the NernstPlanck

equation:
Ji (x) = RT u i (x)gradci (x) zi F ci (x)
u i (x)grad[(x) i (x)]
with i (x) = i /zi F .
(70)
381
382
To integrate this equation, we can

rewrite it in the following way:
Ji (x) = RT u i (x)

grad ci ezi F [(x)i (x)]/RT
ezi F [(x)i (x)]/RT
(71)
If we consider the ion-transfer reaction as

the crossing of the inner layer, we have to
integrate this equation between a in the
water phase and b in the oil phase.

Ji
b
ezi F [(x)i (x)]/RT

dx = RT
u i (x)
a

d ci ezi F [(x)i (x)]/RT (72)
If (x) i (x) varies linearly through the

interface, by dening
zi F
[(b) i (b)]
RT
zi F
ya =
[(a) i (a)]
RT
yb =
(73)
(74)
we can write
zi F [(x) i (x)]
x
= yb + (ya yb )
RT
L
(75)
If it is assumed that the ionic mobility is
constant through the interface, we have
Ji
u i
eyb +(ya yb )x/L dx = RT

d ci ezi F [(x)i (x)]/RT (76)
Solving this integral leads to

L
LJi
Ji
e(ya yb )x/L dx =
eyb
u i RT
u
i RT
0
ya

y
b
e e
= RT [cia eya cib eyb ]

(ya yb )
(77)
We then obtain an equation for the

ion ux:

u i RT cia eya cib eyb
Ji =
(ya yb )
L
eya eyb
(78)
The parameter y is dened as
ya yb
zi F
=
(a ab i )
2
2RT b
(79)
The ion ux can now be expressed as

2u i RT y cia e2y cib
Ji =
L
e2y 1
y=
u i RT y
(ca ey cib ey ) (80)
L sinh y i
The approach outlined above is essentially

equivalent to that suggested by Kakiuchi [60], which again relies on the work
of Goldman for ion transfer across membranes. Expressions for the ion-transfer
rate constants may be obtained as below
if we recall that the ionic ux across the
interface can be described according to
Eq. (50).

yey
u i RT
(81)
k =
L
sinh y

y
u i RT
ye
k =
(82)
L
sinh y
Since it may be noted that
yey
yey
= lim
=1
y0 sinh y
y0 sinh y
lim
(83)
It immediately follows that a useful

denition for ki is

u i RT
(84)
ki =
L
The equations describing k and k can

easily be incorporated into Eq. (50) to
describe the entire current potential characteristic. Note that in this treatment the
ion transfer is treated as a transport process and u i is a phenomenological mobility

of the ion in the inner layer. This model
has successfully accounted for experimental data [71] and the experimental values
of and k can be given phenomenological meanings. Within the framework
of the model suggested by Kakiuchi, the
ion-transfer rate constant can be related to
the microscopic structural features of the
liquid|liquid reaction plane describing the
ionic mobility and interface width.
By analogy to the derivation of Eqs. (56
and 66), we can obtain an expression for Ji
on the basis of the bulk concentration of
the ion in water and oil. The concentration
of the ion at a and b may be expressed
according to Eqs. (51 and 52) if we again
assume the formation of a preequilibrium
according to Sch. 1, or in other words
that the ion transfer/transport across the
liquid|liquid interface is slow compared to
the diffusion/migration of the ion in the
diffuse layers. Under such conditions the
following expression for Ji ensues:
Ji =
u i RT y w zi Faw /RT
c e
L sinh y i
a
ezi F (b b i )/2RT
ci ezi Fo /RT

b
ezi F (b b i )/2RT

a

(85)
We can dene a potential # such that

a# = #b = 12 ab
(86)
leading to the following expression:

Ji =
u i RT y
L sinh y

a
#
ciw ezi F (w +1/2b i )/RT

a
#
ci ezi F (o 1/2b i )/RT

(87)
In general, 1/2ab i can be approximated

by 1/2w
o i and we can therefore write

Ji =
u i RT y
L sinh y

#
w
ciw ezi F (w +1/2o i )/RT

#
w
ci ezi F (o 1/2o i )/RT

(88)
Finally, by dening as the fraction of the

total potential drop between the two bulk
phases occurring from bulk water to the
middle point of the potential drop between
a and b, we arrive at the following result:
Ji =
u i RT y
L sinh y

w
w
ciw ezi F (o 1/2o i )/RT
w ]/RT
ci ezi F [(1)o 1/2o i

w
(89)
It should be noted that in order to apply
this rened model in general it is necessary
to know the potential drop across the inner
layer. We can recover Eq. (56) if we make
the assumption that is independent
of w
o and dene k according to the
following equation valid at equilibrium at
the formal standard transfer potential:
u i RT y
w
w
ezi F (o i 1/2o i )/RT =
L sinh y
u i RT y
L sinh y
w 1/2w )/RT
o i
ezi F ((1)o i
=k
(90)
Note, however, that this approach leads to

a different dependence of k on w
o i .
An interesting point concerns the fact that
when is close to 0.5, the dependence of

k on w
o i vanishes.
383
384
General Treatment of Ion-transfer

Reactions at ITIES
In recent theoretical work on ion transfer
at ITIES, the ion transfer through the diffuse layer has been analyzed [62, 63, 67].
In the following we will assume that the
ion transport from bulk solution to the
interface can be described according to
the NernstPlanck equation. No specic
assumptions regarding the ion distribution or the potential prole will be made.
The ion ux across the diffuse layers can
then be obtained by solving the following equations:
a
Ji
ezi F (x)/RT dx = RT
4.2.3.5
u w
i

Ji
b
d(ezi F (x)/RT )
zi F (x)/RT
u i
(91)
dx = RT
d(ezi F (x)/RT )
(92)
Note that in these equations the integration

limits and represent the bulk
phases of the oil and water phase,
respectively, which is identied as the
position from the interface in which the
Galvani potentials equal the bulk phase
values. Upon integration, the following
expressions are obtained:
a z F a /RT

w
ciw ezi F /RT
ci e i

Ji = RT u w
a
i
ezi F (x)/RT dx

RT u w
a
i
=
cia ezi Fw /RT ciw
(93)
w
A

b
ci ezi F /RT cib ezi F /RT

Ji = RT u i
ezi F (x)/RT dx
=
RT u i
ci cib ezi Fo /RT

(94)
where Aw and A are given by

Aw =
A =

=
ezi Fw /RT dx
(95)
ezi Fo /RT dx
b
b
ezi Fo /RT dx
(96)
The concentrations of the ion at a and b

are then obtained as

Ji Aw
a
w
cia = ezi Fw
+
c
(97)
i
RT u w
i

Ji A
b
+
c
(98)
cib = ezi Fo
i
RT u i
In the case of small Aw and A , the traditional Boltzmann distribution is recovered.
This situation corresponds to the case in
which the potential drops between a and
w and between b and o are small.
These expressions can be inserted into
Eq. (50) to obtain the following result:
w zi Faw /RT
k ci e
b
k ci ezi Fo /RT
(99)
Ji =
Aw zi Faw /RT
1+ k
e
RT u w
i
A
b
k
ezi Fo /RT
RT u i
Note that a potential dependence may
easily be incorporated into an expression
for k and k as in Eqs. (57 and

58). A ButlerVolmer type expression
is obtained (Eq. 53) when Aw and A
can be approximated to zero in Eqs. (97
and 98). Alternatively, Eqs. (97 and 98)
can be incorporated into Eq. (80) if the
ion transfer is described as a pure
transport process:
(ciw ey ezi Fw /RT
u i y
b
ci ey ezi Fo /RT )
L sinh y
w
Ji =
A
a
1+
ey ezi Fw /RT
RT u w
i

A y z Fb /RT u i y
i
o
e
e
RT u i
L sinh y
(100)
As in Eq. (80), we have here specied a
linear potential prole across the interface.
If we dene k and k as in Eq. (81

and 82), we recover Eq. (99). The potential
dependence on the rate constant is,
however, somewhat different from what
is obtained from expressions based on
transition state theory (Eqs. 57 and 58).
In a practical situation it can be assumed
that a and b are approximately equal. Then
the potential drop in the aqueous phase is
w
simply obtained as w
a o .
Regardless of whether the transfer or
transport approach is followed, the ux
equation can, in principle, be solved for any
given potential prole in the diffuse layers.
Usually, the potential prole is quite well
described using the PoissonBoltzmann
equation, which in the case of a z : z
supporting electrolyte can be solved analytically. More advanced theories may also
be applied [81, 82]. In a recent theoretical
study based on the idea of determining
the permeabilities of the diffuse double
layers and the interface independently,
different potential proles were analyzed.
The PoissonBoltzmann potential prole,
a stepwise linear potential prole, and a
potential step prole were compared. Interestingly, even a very simple stepwise
potential prole is hardly distinguishable
from the linear potential prole and the relatively complicated PoissonBoltzmann
a
potential distribution when realistic parameters of the ion and diffuse layers
are used [62]. From these considerations
it may thus be concluded that traditional
steady state current voltage curves generally do not give much information about
the inner layer processes.
In general, experimental studies on iontransfer reactions lead to values of apparent rate constants. In this connection it is
interesting to note that the nominator in
Eq. (99) is essentially similar to the expression for Ji in Eq. (66) obtained from the
simple ButlerVolmer approach. It thus
immediately follows that the correction
of an experimentally determined apparent
rate constant based on Eq. (66) (equivalent
to the Frumkin correction used for ET at
solid electrodes) is not in direct agreement
with the more general treatment leading to
Eq. (99). This point was rst recognized by
Senda [63], who termed the denominator
of Eq. (99) the Levich correction.
The Marcus Model for Ion Transfer
at Liquid|Liquid Interfaces
Recently, Marcus has attempted to treat
the ion transfer as a combined activationcontrolled and transport-controlled process [68]. In the Marcus theory for ion
transfer between two immiscible solutions, the ion transport is described using
a NernstPlanck type equation [57]. This
equation leads to an expression of k similar to that obtained by Kakiuchi [60],
4.2.3.6
k =
uRT
D
=
L
L
(101)
The experimentally determined k leads to

a phenomenological diffusion coefcient
that is in the order of 108 cm2 s1 ,
assuming an interface thickness of 10 A.

A central point in the Marcus theory
is the experimentally and theoretically
veried presence of protrusions at the
385
386
liquid|liquid interface. It is assumed that

the ion transfer occurs via attachment of
the ion to a protrusion in one phase.
The probability of forming a protrusion
of height h is given by P (h). The
probability of forming a protrusion is
related to the Gibbs energy of formation of
the protrusion.
The overall process of transfer of an ion
I from a solvent A to a solvent B can be
shown schematically as
AnA I + nB B nA A + IBnB
(102)
Here nA and nB denotes a measure of the

solvation of I in A and B.
The attachment and detachment of an
ion to a protrusion in A is described by
A . Using these
the rate constants k A and k
rate constants as well as the rate constant
of detachment of the ion in solvent B, it is
possible to describe the ion ux across the
inner layer.
A
ci (za )]P (ha )ha (103)
J = [k A ci (za ) k
B
J = k
ci (zb )P (hb )hb
(104)
The overall rate constant of ion transfer is

obtained as
1
krate
1
A
kassn
1
Ak
Keq
diff
1
B kB
Keq
diss
(105)
In this equation the rate constants and
equilibrium constants may be estimated
using the transition state theory. In this
treatment an activation energy can easily
be incorporated in the expressions for k A
B using the transition state theory.
and kdiss
A
kassn
= k A P (ha )ha
A
Keq
=
2
1
2g r P (ha )ha
B
=
Keq
2
1
2g r P (hb )hb
e[(zb )(za )]/kB T
(106)
(107)
(108)
The electrochemical potential of the ion

at zb and za is given by (zb ) and (za ),
respectively. Within the framework of a
transition state model, the (x, y) oscillation
of the ion transverse to the protrusion
is considered. The thermally averaged
mean square displacement or the (x, y)
oscillator is given by r 2 . The surface
density of protrusions is termed and
g is a structural factor. It is instructive
to consider one of the limiting situations
Ak
in which Keq
diff is the rate-controlling
step. In this case the expression for krate
reduces to
A
kdiff
krate = Keq
= 12 gr 2 P (ha )ha kdiff (109)

The expression for kdiff is essentially the
same as that derived by Kakiuchi for an
ion-transfer reaction given in Eq. (82). It
thus immediately follows that the apparent
good agreement between experimentally
determined ion-transfer rate constants
and Kakiuchis model is also consistent
with the Marcus model. Furthermore,
the small phenomenological diffusion
coefcient needed to t the data to Eqs. (80
to 84) may simply be explained by a small
A.
Keq
The Marcus theory is difcult to apply
directly in a practical situation, owing to
lack of knowledge of the probability of
the formation of protrusions. One way
to overcome this problem could be to
employ capillary wave theory for the interface between two immiscible electrolyte
solutions. Recently, theoretical efforts have
been made to describe capillary waves at
soft electried interfaces [83]. It may be
possible to use such theories to quantify
the value of P (h)h. One of the major
complications is related to the fact that the
surface tension is dependent on the Galvani potential difference between the two
phases, which means that P (ha )ha is

potential-dependent. Theoretical [77] and
experimental [84, 85] studies have been
undertaken to clarify the dependence of
the surface structure on an electric eld,
but more studies are clearly needed before quantitative predictions can be made.
In particular, a clear-cut experimental evidence of a link between protrusions at
liquid|liquid interfaces and ion transfer is
still lacking.
Other Models
Another model that treats the ion transfer
across ITIES as an activated transport process by applying classical transition state
theory has been suggested by Girault and
Schiffrin [55]. Also in this case the equation describing the ion ux across the
interface is similar to that obtained by
the ButlerVolmer approach. Naturally,
the interpretation of the values and k
is somewhat different in that case. Experimentally [85] and theoretically [79], the effect of solvent viscosity on the ion-transfer
process has been conrmed indicating the
importance of this parameter.
Aoki has modeled ion transfer as a combined transport and activated process [65].
In this approach the activation energy for
ion transfer is a combination of an electric potential energy, an activation energy
due to solvation changes, and an energy
required to form a pore into the interface.
The microscopic ion transport is described
using the Langevin equation. In certain
cases, such as when viscous effects are
pronounced, the theory is in good agreement with the ButlerVolmer equation.
A limiting factor in this theory is related to the fact that it cannot successfully
account for the experimentally observed
concave variation of the logarithm to the
ion-transfer rate constant with the Galvani
potential difference.
4.2.3.7
In a model put forward by Indenbom,

the ITIES is considered as an elastic
lm in which forces of interfacial tension
counteract the forces of an external
electric eld [64]. This causes a potentialdependent compressed grooved structure
of the interface at which ion transfer takes
place when the compression reaches a
critical state. The validity of this picture
will rely on detailed molecular dynamics
studies of the microscopic features of ion
transfer across ITIES. In this connection it
is relevant to mention the work of Cunnane
and coworkers in which the kinetics of
ion transfer across a monolayer adsorbed
at a liquid|liquid interface is discussed in
terms of the energy required to form a pore
of a size similar to that of the transferring
ion [56].
Schmickler has treated the ion-transfer
process within the framework of a latticegas model [66]. In this model the interface
is treated as an extended region in which
the two solvents mix. The reaction rate
is generally obtained using the transition
state theory in combination with Kramers
theory depending on whether frictional
forces are important.
The central idea in Schmicklers model
is to treat the liquid|liquid system as
an innite three-dimensional lattice. The
fraction of sites occupied by the two kinds
of solvent molecules at a position z are
termed f1 (z) and f2 (z).
1
for z 0
z/
2e
f1 (z) =
1
for z 0
2ez/
(110)
f2 (z) = 1 f1 (z)
(111)
In these equations, represents the width

of the interface.
The nal result is shown to be in
good agreement with a ButlerVolmer
387
388
behavior. Schmicklers treatment may be

useful in a theoretical modeling of the
transfer process, in particular to determine
the potential of mean force acting on the
ion. Qualitative experimental predictions
may also be made. This model can,
for instance, successfully account for the
dependence of the solvation energy on
the ion-transfer rate constant. It may be
noted that no assumptions regarding the
microscopic structure of the liquid|liquid
interface are made in this description
and, for instance, surface tension is
not considered explicitly. An advantage
of this relatively simple model is that
additional complications such as ion
pairing of supporting electrolyte ions can
be described [86].
4.2.4
Conclusion and Outlook
The analysis presented in this review

appears to show that while classical transition state theory provides a good account
for experimental observation concerning
ET kinetics, the picture is less clear for
ion-transfer processes. Our understanding remains hindered by the primitive
knowledge of the ITIES structure at a
microscopic level, despite current efforts
based on computer modeling [41, 51, 66,
87]. The potential distribution across the
interface is also an essential piece of information that is yet to be claried in the
liquid|liquid context. Furthermore, these
analyses have excluded any specic interaction of ions at the interface, which can
also exert a profound effect on the reactivity of the system. In order to address
all these aspects, it is crucial to develop
novel spectroscopic techniques coupled to
electrochemical systems.
Dynamic studies of heterogeneous ET
across ITIES consistently suggest that
these processes are strongly nonadiabatic.

The maximum values of rate constants
measured either by SECM or photoelectrochemical techniques are of the order
of 1019 cm4 s1 , which coincide with
the nonadiabatic limit obtained from the
models outlined in Sect. 4.2.2. The low
electronic coupling between reactants is a
result of the typical separation distances
at the interface. These distances are also
responsible for reorganization energies
close to 1 eV as estimated from experimental data.
Undoubtedly, a key development that
will clarify basic aspects in connection to
ET dynamic is the effective control on the
distance separating the redox species. It
can be envisaged that redox couples linked
by functionalized spacers, which allow
highly ordered self-assembling structures,
could provide very valuable information on
the rate of interfacial ET and the effect of
the interfacial potential. In this case, electronic coupling between reactants could
be quite substantial and sophisticated
modeling will be required to fully rationalize these ndings. Ultrafast photoinduced
processes can be effectively approached by
spectroscopic Pump-Probe methods down
to the femtosecond timescale.
The ion-transfer theories outlined in this
chapter are all based on idealized models.
In practice, complications in the measurements related to adsorption of the transferring ion or chemical reactions associated
with the ion transfer arise. Detailed theoretical descriptions of the microscopic
nature of such complications are still lacking. One of the most interesting feature
of simple ion transfers concerns the fact
that they always appear fast, and indeed
as the experimental techniques have been
improved, the measured values have increased even for similar ion transfers. One
way to clarify the microscopic features of
ion-transfer reactions could be to perform

photoinduced ion-transfer reactions in a
controlled way. Such an experiment could,
for instance, be carried out using a surfactant that upon illumination generates an
ion that can transfer across the interface.
References
1. F. Reymond, D. J. Fermn, H. J. Lee et al.,
2. A. G. Volkov, D. W. Deamer, Liquid-Liquid
Interfaces: Theory and Methods, CRC Press,
Boca Raton, Fla., 1996.
3. A. G. Volkov, D. W. Deamer, D. I. Tanelian
et al., Liquid Interfaces in Chemistry and
Biology, John Wiley & Sons, New York, 1998.
4. R. Lahtinen, D. J. Fermn, K. Kontturi et al.,
5. D. J. Fermn, R. Lahtinen in Dynamic Aspects
of Heterogeneous Electron Transfer Reactions
at Liquid|Liquid Interfaces (Ed.: A. Volkov),
Marcel Dekker, Boca Raton, Fla., 2001,
pp. 179228.
6. Z. Samec, V. Marecek, J. Weber, J. Electroanal. Chem. 1978, 96, 245247.
7. Z. Samec, J. Electroanal. Chem. 1979, 99,
197205.
8. Y. I. Kharkats, Sov. Electrochem. 1976, 12,
1257.
9. H. H. J. Girault, D. J. Schiffrin, J. Electroanal.
Chem. 1988, 244, 1526.
10. R. A. Marcus, J. Phys. Chem. 1990, 94,
41524155, 7742.
10501055.
20102013.
13. Y. Cheng, D. J. Schiffrin, J. Electroanal.
Chem. 1991, 314, 153163.
14. Y. Cheng, D. J. Schiffrin, J. Chem. Soc.,
Faraday Trans. 1993, 89, 199205.
15. Y. F. Cheng, D. J. Schiffrin, J. Chem. Soc.,
Faraday Trans. 1994, 90, 25172523.
16. V. J. Cunnane, G. Geblewicz, D. J. Schiffrin,
17. G. Geblewicz, D. J. Schiffrin, J. Electroanal.
Chem. 1988, 244, 2737.
18. S. Kihara, M. Suzuki, K. Maeda et al., J.
19. E. Makrlik, Z. Phys. Chem. 1987, 268,
212214.
20. M. Lher, R. Rousseau, E. Lhostis et al., C. R.

Acad. Sci. Paris, t. 322, Serie IIb 1996, 322,
5562.
21. Q. Z. Chen, K. Iwamoto, M. Seno, Electrochim. Acta 1991, 36, 291296.
22. S. Daniele, M. A. Baldo, C. Bragato, Electrochem. Commun. 1999, 1, 3741.
23. H. H. Girault, see the web site dcwww.ep.ch
for a comprehensive list of free energy of ion
transfer at various liquid|liquid interfaces.
24. A. A. Stewart, J. A. Campbell, H. H. Girault
et al., Ber. Bunsen-Ges. Phys. Chem. 1990, 94,
8387.
25. Z. Ding, P.-F. Brevet, H. H. Girault, Chem.
Commun. 1997, 20592060.
26. Z. F. Ding, D. J. Fermin, P. F. Brevet et al., J.
27. J. Zhang, C. J. Slevin, L. Murtomaki et al.,
Langmuir 2001, 17, 821827.
28. J. Zhang, P. R. Unwin, J. Phys. Chem. B 2000,
104, 23412347.
29. B. Quinn, R. Lahtinen, L. Murtomaki et al.,
30. A. J. Bard, D. E. Cliffel, C. Demaille et al.,
Ann. Chim. 1997, 87, 1531.
31. A. L. Barker, J. V. Macpherson, C. J. Slevin
et al., J. Phys. Chem. B 1998, 102, 15861598.
32. A. L. Barker, P. R. Unwin, S. Amemiya et al.,
J. Phys. Chem. B 1999, 103, 72607269.
33. M. Tsionsky, A. J. Bard, M. V. Mirkin, J.
Phys. Chem. 1996, 100, 17 88117 888.
34. M. Tsionsky, A. J. Bard, M. V. Mirkin, J. Am.
Chem. Soc. 1997, 119, 10 78510 792.
35. C. Wei, A. J. Bard, M. V. Mirkin, J. Phys.
Chem. 1995, 99, 16 03316 042.
36. B. B. Smith, J. W. Halley, A. J. Nozik, Chem.
Phys. 1996, 205, 245267.
37. H. H. Girault, J. Electroanal. Chem. 1995,
388, 93100.
38. I. Benjamin, Y. I. Kharkats, Electrochim. Acta
1998, 44, 133138.
39. N. Sutin, Acc. Chem. Res. 1982, 15, 275.
40. J. T. Hupp, M. J. Weaver, J. Electroanal.
Chem. 1983, 152, 114.
428, 123127.
42. B. Liu, M. V. Mirkin, J. Am. Chem. Soc. 1999,
121, 83528355.
43. D. J. Fermn, Z. Ding, H. Duong et al., J.
Phys. Chem. B 1998, 102, 10 33410 341.
44. D. J. Fermn, H. Doung, Z. Ding et al., Electrochem. Comm. 1999, 1, 2932.
45. D. J. Fermn, H. Duong, Z. Ding et al., Phys.
Chem. Chem. Phys. 1999, 1, 14611467.
389
390

46. D. J. Fermn, H. Duong, Z. Ding et al., J.
Am. Chem. Soc. 1999, 121, 10 20310 210.
47. H. Jensen, J. J. Kakkassery, H. Nagatani
et al., J. Am. Chem. Soc. 2000, 122,
10 94310 948.
48. N. Eugster, D. J. Fermn, H. H. Girault, J.
Phys. Chem. B 2002, 106, 34283433.
49. P. Vanysek, Electrochim. Acta 1995, 40,
28412847.
50. H. H. Girault, Mod. Aspects Electrochem.
1993, 25, 162.
51. I. Benjamin, Annu. Rev. Phys. Chem. 1997,
48, 407.
52. Z. Samec, T. Kakiuchi, Adv. Electrochem. Sci.
Eng. 1995, 4, 297361.
53. C. Gavach, B. DEpenoux, F. Henry, J. Electroanal. Chem. 1975, 64, 107115.
54. J. Koryta, P. Vanysek, M. Brezina, J. Electroanal. Chem. 1977, 75, 211228.
55. H. H. J. Girault, D. J. Schiffrin, J. Electroanal.
Chem. 1985, 195, 213227.
56. V. J. Cunnane, D. J. Schiffrin, M. Fleischmann et al., J. Electroanal. Chem. 1988, 243,
455464.
57. I. Benjamin, J. Chem. Phys. 1992, 96,
577585.
58. Y. Y. Gurevich, Y. I. Kharkats, J. Electroanal.
Chem. 1986, 200, 316.
59. Z. Samec, Y. I. Kharkats, Y. Y. Gurevich, J.
60. T. Kakiuchi, J. Electroanal. Chem. 1992, 322,
5561.
61. T. Kakiuchi, J. Noguchi, M. Senda, J. Electroanal. Chem. 1992, 327, 6371.
62. K. Kontturi, J. A. Manzanares, L. Murtomaeki, Electrochim. Acta 1995, 40, 29792984.
63. M. Senda, Electrochim. Acta 1995, 40,
29932997.
64. A. V. Indenbom, Electrochim. Acta 1995, 40,
29852991.
65. K. Aoki, Electrochim. Acta 1996, 41,
23212327.
426, 59.
67. L. Murtomaki, A. K. Kontturi, D. J. Schiffrin,

68. R. A. Marcus, J. Chem. Phys. 2000, 113,
16181629.
69. Z. Samec, V. Marecek, J. Weber et al., J.
70. P. D. Beattie, A. Delay, H. H. Girault, Electrochim. Acta 1995, 40, 29612969.
71. T. Kakiuchi, Y. Takasu, J. Phys. Chem. B
1997, 101, 5963.
72. Z. F. Ding, F. Reymond, P. Baumgartner
73. Z. F. Ding, R. G. Wellington, P. F. Brevet
74. D. J. Fermn, Z. Ding, P. F. Brevet et al., J.
75. I. Benjamin, Chem. Rev. 1996, 96, 14491475.
76. I. Benjamin, Science 1993, 261(5128),
15581560.
77. K. J. Schweighofer, I. Benjamin, J. Electroanal. Chem. 1995, 391, 110.
78. P. A. Fernandes, M. N. D. S. Cordeiro, J. A.
N. F. Gomes, J. Phys. Chem. B 1999, 103,
62906299.
79. R. Ferrigno, H. H. Girault, J. Electroanal.
Chem. 2001, 496, 131136.
80. Z. Samec, V. Marecek, D. Homolka, J. Electroanal. Chem. 1985, 187, 3151.
81. Q. Cui, G. Zhu, E. Wang, J. Electroanal.
Chem. 1995, 383, 712.
82. L. I. Daikhin, A. A. Kornyshev, M. Urbakh, J.
83. Z. H. Zhang, I. Tsuyumoto, S. Takahashi
et al., J. Phys. Chem. A 1997, 101, 41634166.
84. Z. H. H. Zhang, I. Tsuyumoto, T. Kitamori
10 28410 287.
85. Y. Shao, H. H. Girault, J. Electroanal. Chem.
1990, 282, 5972.
86. C. M. Pereira, W. Schmickler, F. Silva et al.,
87. I. Benjamin, ACS Symp. Ser. 1994, 568,
409422.
4.3 The Inuence of Ad-atom Adsorption on Reaction Rates and Mechanisms
4.3
The Inuence of Ad-atom Adsorption on

Reaction Rates and Mechanisms
Georgios Kokkinidis
Aristotle University of Thessaloniki, Thessaloniki, Greece
4.3.1
Introduction
It is well known that many chemical reactions do not by themselves occur at a

signicant rate, although thermodynamics predict that they are very favorable. The
rate of such reactions, however, may be
increased by several orders of magnitude
by a catalyst. The catalyst (homogeneous
or heterogeneous) is a substance, which alters the rate of a chemical reaction without
itself being consumed or generated in the
process. Similarly, there are many electrochemical reactions that also do not proceed
with a sufcient rate because of poor kinetics. For these reactions, it is again
necessary to nd an active electrode/catalyst
or, better, an electrocatalyst, otherwise high
overpotentials should be applied in order to promote the reaction. Thus, the
term electrocatalysis was introduced for the
heterogeneous charge-transfer reactions
that take place at the electrode/electrolyte
solution interface. The objective of electrocatalysis is to discover electrocatalysts that
allow electrode reactions to occur with a
sufcient rate at potentials very close to the
equilibrium potential and, in some cases,
to improve the selectivity with respect to
the reaction products.
The catalytic role of an electrode is
revealed by writing the rate of an electrochemical reaction as follows [1]:

j
kT
Go# /RT
=
= c
e
nF
h
eF
/RT (for a cathodic process)
(1)
This equation makes it clear that electrocatalysis consists of two terms: an electrical
one (the term depending on the potential difference across the interface) and
a chemical one (the Arrhenius term),
which depends on the nature of the electrode material.
Although by applying high overpotentials one can increase the rate of an
electrochemical reaction on a given electrode/catalyst even by more than 10 orders
of magnitude, the activity of the electrode,
the Arrhenius term, still remains the most
important fact in electrocatalysis. This is
obvious, since the energy efciency of any
cell is determined mainly by the overpotentials necessary at the anode and cathode.
The lower the overpotentials, the better
is the energy efciency. Furthermore, the
application of high overpotentials usually
introduces problems due to the interference of undesired reactions (e.g. the
hydrogen or the oxygen evolution reaction,
the electrodissolution of the electrode,
etc.).
The catalytic activity of an electrode
depends on its electronic and perhaps
crystallographic properties. When the reactants or intermediates are adsorbed on
the electrode surface, the reaction rate will
depend additionally on the heat of adsorption, on the strength of bond-forming or
bond-breaking, and on the properties of
the adsorption isotherm. All these factors
may be changed by the modication of the
electrode surface from adsorbed species
existing in the solution, not participating
in the reaction. These species can sometimes improve the catalytic activity of the
bare material. For example, heavy metal
atoms (e.g. Pb, Tl, and Bi) deposited in the
underpotential region have been shown
to increase the catalytic activity of noble metal electrodes (e.g. Pt, Au, Ag, Pd,
etc.) for several technologically interesting
391
392
electrochemical reactions. Such reactions

are the reduction of oxygen, the oxidation of organic fuels, the reduction of
aromatic nitro compounds, and so forth.
Although noble metals are considered to
be the best electrocatalysts in contemporary electrocatalysis, their modication by
foreign metal ad-atoms deposited in the
underpotential region may improve their
catalytic activity for certain electrochemical reactions.
4.3.2
Metal Ad-atoms Deposited in the

Underpotential Region
Foreign metal ad-atoms can be deposited

on an electrode electrochemically in the
so-called underpotential region, that is,
the potential region positive to the reversible Nernst potential for the bulk
metal deposition.
E = Eads ENernst 0
(2)
The underpotential deposition (abbreviated to UPD) of a metal M on a foreign

metal substrate M is a process that takes
place according to the reaction
M z+ + ze
Mads
(3)
This Faradaic adsorption of M z+ ions dissolved in the solution of the supporting

electrolyte leads to the formation of suband monolayer coverages of discharged M
ad-atoms and constitutes the rst stage
in electrocrystallization. The UPD process
was studied rather extensively because of
its importance in investigating a wide variety of electrochemical phenomena, such
as adsorption on electried interfaces,
charge-transfer reactions, surface diffusion, nucleation and growth, double-layer
changes, and, of course, because of its
fundamental and practical signicance in

electrocatalysis.
The UPD process was studied for a large
number of systems using both polycrystalline and monocrystalline electrodes as
substrates and the progress of understanding all related parameters was covered in
depth in several review articles [25]. Here
we will give only an outline of some thermodynamic and structural aspects, which
will be useful for the better understanding
of the role of the UPD ad-atoms as reaction
modiers in electrocatalysis.
The equilibrium potential for the UPD
process (Reaction 3) may be formulated
as [3]

fM z+ cM z+
RT
ln
Eads () = E o +
zF
fads
(4)
where fM z+ and fads are the activity
coefcients of the metal ion in solution
and the bulk deposit (fads = 1 for 1),
cM z+ is the metal ion concentration
in solution, and the UPD ad-atom
coverage. Since takes values between
0 and 1, Eq. (4) yields an equilibrium
potential that is always more positive
than the corresponding Nernst potential
for the bulk metal deposition. Moreover,
Eq. (4) implies that the submonolayer
equilibrium potential shifts toward more
positive potentials by 60/z mV dec1 for
constant with increasing the metal ion
concentration in solution.
There are two excess parameters, the
charge = ( /E) and the surface concentration = ( /)E from which
the electrosorption valency may be dened
and experimentally determined:

1
(5)
=
F E
The symbols in all the above expressions
have their usual meaning.
The electrosorption valency expresses

the charge ow during the electrosorption
process at a constant potential. It takes
values ranging between zero and the
ionic charge z. Values very close to z
show that the adsorbed species are almost
completely discharged, while values of
lower than z indicate adsorption with
partial charge transfer. The adsorbed
species are recognized as ad-atoms when
takes values very close or equal to z.
Saturate coverages with ad-atoms are obtained at potentials close to the potential for
the bulk metal deposition. The maximum
surface concentration is about max
2 109 mol cm2 . This surface concentration corresponds to a maximum amount
of charge max z 200 C cm2 [3, 4].
The extent of the UPD region depends
on the binding energies between the
adsorbed atoms and the atoms of the
substrate metal. There exists a linear
relationship between
Ep and
, where
is the difference in the work functions

between the substrate and the adsorbed
metal. This linear correlation is expressed
by a semiempirical Eq. (3):
Ep = 0.5

(6)
Ep is dened as the difference between

the bulk and the most positive adsorption peak of the corresponding cyclic
voltammogram for the UPD. The linear correlation holds for systems with
polycrystalline substrate electrodes, while
for systems with single-crystal electrodes
some deviations were observed [3]. This
behavior suggests that the binding energies of the UPD are not only determined
by the difference in the electronegativities
expressed as
but they are also inuenced by the lateral interactions that are
governed by the crystallographic orientation of the substrate.
From theoretical considerations of bond

formation in electrosorbates it was concluded that the underpotentially deposited
metal ad-atoms can approximately be
considered to be covalently bonded and
nearly completely discharged if the difference in the Paulings electronegativities
of the substrate and adsorbed metal is
|
| 0.5 [6]. That means these systems
have electrosorption valency /z 1. Systems in which the electrosorption valency
is equal to the ionic charge are those resulting from the UPD of heavy metal ions
on noble and transition metal electrodes.
There are a number of parameters that
can actually determine the structure of the
UPD monolayers [4]. The most outstanding ones are certainly the structure of the
substrate, the ratio between the diameter
of the adsorbed species and the distance
of two adjacent adsorption sites, and of
course the degree of the UPD ad-atom
coverage. At low surface coverages a UPD
metal monolayer consists of isolated M
ad-atoms or partially discharged and desolvated ions, which are randomly distributed
on the substrate owing to the lateral repulsion. With the increase in the degree
of coverage some ordering occurs and at
high surface coverages the UPD layer may
form a sequence of two-dimensional periodic superlattice structures with electronic
properties closely approaching those of the
bulk metal.
On single-crystal electrodes the UPD
deposits often form well-dened, ordered monolayers. Hypothetical superlattice structures have been proposed, which
rely on the charges determined for the various peaks observed in the cyclic voltammograms. The anions of the supporting
electrolytes inuence the shape and the
position of the UPD peaks in the cyclic
voltammograms, although the superlattice structures are not always affected by
393
394
interactions between the ad-atoms and

the anions adsorbed on the UPD monolayer [7].
In the beginning, conventional electrochemical techniques had been used to
study the UPD, both on polycrystalline
and single-crystal electrodes. The work
done on single-crystal electrodes included
both low index planes and stepped surfaces. The technique developed by Clavilier
et al. [8, 9] (the so-called ame annealing
technique) has led to a revolution in electrochemical studies, since single crystals
of platinum and gold can be relatively easily formed and precisely oriented using an
optical technique. Single-crystal surfaces
were previously prepared in an ultrahigh
vacuum (UHV) chamber by sputtering
and annealing, and their structure was
dened by low energy electron diffraction
(LEED) [1012]. The results of the various
techniques were compared mainly in connection with the hydrogen adsorption on
platinum [13, 14].
The electrochemical methods provide
only indirect information about the structure of the UPD ad-atom layers in an
atomic scale. In order to obtain more direct information about the structure of the
UPD ad-atom layers, many investigators
have adopted the use of in situ techniques,
in which the electrode surface is examined with surface science methods. The
methods mainly used are in situ X-ray
diffraction [7, 1517], in situ scanning tunneling microscopy (STM) [1823], and in
situ atomic force microscopy (AFM) [24].
These methods have provided detailed information on the atomic structure, the
thermodynamic stability, and the dependence of the structure on the potential
for several UPD systems. However, little
attempt has been made (and this mainly
concerns the reduction of oxygen) to correlate directly the structure of the ad-atom
layers to their inuence on the catalytic

activity of the substrate. The reason is that
the superlattice structures of the ad-atom
layers are inuenced by the strong adsorption of intermediates involved in the
mechanism of many electrocatalytic reactions. For example, the rate of formic acid
oxidation on Pt modied by Pb ad-atoms is
different if lead is adsorbed before or after
formic acid [25].
4.3.3
Electrocatalysis by Ad-atoms
Modied electrodes with ad-atoms (usually

deposited in the underpotential region)
exhibit enhanced electrocatalytic activity
for several categories of electrochemical
reactions. The most extensively studied
reactions are those related to the development of low-temperature fuel cell
technology, namely, the reduction of oxygen and the oxidation of organic fuels. The
ad-atoms may inuence the rate and the
mechanism of electrochemical reactions
through [2629]
1.
2.
3.
4.
the third-body effect,

the prevention of poison formation,
the bifunctional mechanism, and
the modication of the electronic properties of the electrode surface.
In the third-body mode of action, the

ad-atoms are considered as inert species,
actually not participating in the reaction.
Their presence acts only to diminish the
places where the poison formation reaction can take place. The poisoning reaction
requires greater number of neighboring
sites (say Pt surface atoms) not occupied by ad-atoms than the main reaction
that proceeds in parallel through reactive
(not strongly adsorbed) intermediates. The
enhanced reactivity in the presence of adatoms results from decreased poisoning

and, thereby, an increase in the rate of the
reaction branch, which requires smaller
number of neighboring sites.
Blocking of poison formation through
the third-body mechanism is physical. Adatom coverages at around 0.5, although
permitting the adsorption/dehydrogenation step (which is necessary for initiating
the oxidation of an organic molecule),
prevent the strong adsorption of intermediates, which poison the electrocatalyst
surface. In some cases, it is believed
that adsorbed hydrogen is involved in the
poison formation reaction. The ad-atoms
block the adsorption of hydrogen, thus inhibiting the poison reaction. The required
coverage by the ad-atoms should be greater
than 0.5 in the hydrogen potential region
to avoid poison formation.
The bifunctional mechanism refers not
only to catalysts modied by ad-atoms
but also to electrocatalysis on bimetallic
surfaces. The oxidation of methanol on
PtRu alloy surfaces is a typical example of bifunctional mechanism. The role
of the modier is to supply oxygen for
the oxidation of either reactive or poisoning intermediates. The potential in which
increased reaction rates are observed is
governed by the potential in which the
adjacent modier or ad-atom adsorbs oxygen, which should be less positive than
that of platinum. The bifunctional mechanism reduces the oxidation overpotential,
which is particularly important for fuel
cells. The bifunctional mechanism refers
also to catalysts on which chemisorption
of reducible groups and reductive cleavage
of bonds might increase the electrocatalyst
selectivity. A bridge adsorption for the reactive intermediate is assumed to explain
the selective bond cleavage, in which a
substrate atom and an adjacent ad-atom

are involved.
Finally, the ad-atoms may cause local electronic changes on the substrate,
thus affecting the reactantsubstrate interactions. The role of ad-atoms on the
modication of the catalytic properties of
metal surfaces has been extensively studied [3032]. The behavior of these systems
has been rationalized in terms of changes
in adsorption bond strength of the reactant species caused by changes in catalyst
(metal) work function, which results from
the electronic interactions of the ad-atoms
with the active metal surface [3032].
4.3.4
Reactions Catalyzed by Ad-atoms

Reduction of Oxygen
The reduction of oxygen is one of the
most extensively studied electrochemical
reactions. There are two reasons for the
increased interest for this reaction, rstly,
its fundamental signicance with respect
to electrode kinetics and, secondly, its role
in the area of electrochemical energy
conversion and corrosion. The kinetics and
mechanism of oxygen electroreduction
greatly depends on the electrode material.
Oxygen reduction is also a very sensitive
reaction to the surface structure, as shown
by studies on single-crystal electrodes
of different metals. There are several
extended reviews and book chapters,
which cover all these aspects of oxygen
electroreduction [3337].
In aqueous solutions, oxygen reduction
is considered to proceed by two overall
reaction pathways [38].
4.3.4.1
The direct four-electron pathway:

O2 + 4H+ + 4e 2H2 O
(acid solutions)
(7)
395
396
O2 + 2H2 O + 4e 4OH
(alkaline solutions) (8)
O
M
O O
O O
The peroxide pathway:

O2 + 2H+ + 2e H2 O2
(2H+ + 2e )
2H2 O
(acid solutions)
(9)
OH
O2 + H2 O + 2e HO2
(H2 O + 2e )
3OH
(alkaline solutions)
(10)
Depending on the electrode material, the
reduction of oxygen occurs by one or the
other reaction pathway. The two reaction
pathways may also take place in parallel.
The most powerful technique for the
quantitative determination of the extent
of these reactions was proved to be the
rotating ring-disc electrode voltammetry,
which allows the detection of hydrogen
peroxide on the ring electrode.
On some electrodes, hydrogen peroxide cannot undergo further reduction and
is obtained as the nal reaction product. Electrodes on which the two-electron
reduction to hydrogen peroxide predominates are Au, Hg, and various carbon
materials. Electrodes on which the fourelectron reduction to water is predominant
include Ag, Pt, and the Pt family metals.
Platinum is considered to be the best electrocatalyst for oxygen reduction, since the
reduction on this electrode occurs with the
lowest overpotential. The reaction pathway that is more favorable depends on the
way molecular oxygen is adsorbed on the
metal surface. Three different models for
adsorption have been proposed, the Grifths model, the Paulings model, and the
bridge model [38, 39]:
The end-on adsorption (Paulings model)

favors the two-electron reduction of O2 to
H2 O2 . The bridge model, with the two
bonds with two metal sites, favors the
direct four-electron reduction of O2 to H2 O
through rupture of the OO bond.
Metal ad-atoms deposited in the underpotential region strongly inuence the rate
and mechanism of oxygen reduction. Positive and/or negative catalytic effects were
observed depending on the substrate electrode. The work done was reviewed by
Adzic [26] and Kokkinidis [27]. It was reported that Pb, Tl, and Bi ad-atoms catalyze
the reduction of oxygen on Au and carbon electrodes. The half-wave potential
is shifted toward less negative potentials and the two-electron reduction to
hydrogen peroxide is transformed to the
four-electron reduction to water. The catalytic action is mainly observed in alkaline
electrolytes and at potentials of low UPD
ad-atom coverages. Similar behavior was
also observed on ruthenium [40]. Ruthenium partially oxidized, which is inactive
for O2 reduction in alkaline solutions, can
be made very active by Pb and Tl ad-atoms.
Traces of Pb and Tl are sufcient to activate the four-electron reduction on this
electrode.
In contrast to positive catalysis on gold
and carbon electrodes, Pb, Tl, and Bi adatoms cause a negative catalytic effect on
Pt and Ag electrodes. The four-electron
reduction on the bare metals changes to
two-electron reduction on the electrodes
covered by the modiers. This partial
inhibition occurs at negative potentials, in
the mass-transport control region, where
almost complete UPD ad-atom coverages
are obtained. Finally, the reduction of
reects the concentration-dependence of

the UPD process according to Eq. (4).
Figure 2 shows ring-disc measurements
for O2 reduction on Au and Au/Tlads
in an alkaline solution, obtained from
Ref. [44] (Fig. 2a), and on Pt and Pt/Pbads
in an acid solution, obtained from Ref. [45]
(Fig. 2b). The ring currents indicate the
production of hydrogen peroxide as a
function of the disc potential. The positive
catalytic effect observed on Au and Au(hkl)
electrodes was attributed by Juttner [46] to
oxygen on Pt [41] and Au [42] is completely

inhibited by Cu ad-atoms deposited at
underpotentials.
Figure 1 shows, as an example, the
rotating-disc voltammograms of oxygen
reduction on Ag(100) in the absence
and in the presence of Pb2+ ions in
solution [43]. The cyclic voltammogram for
the UPD of Pb on the Ag(100) singlecrystal electrode is given for comparison.
It should be noted that the observed
potential shift of the inhibition effect
8
j
[A cm2]
4
0
4
(b)
8
0.0
j
[mA cm2]
0.4
(c)
E vs NHE
[V]
0.1
0.2 5
5.10
5.103
cPb2+/M
0.8
4
2 3
1
1.2
(a)
0.2
0.0
0.2
E vs NHE
[V]
(a) Currentpotential curves of O2 reduction on Ag(100) in

aerated solution of 0.5 M HClO4 + xM Pb(ClO4 )2 . ( = 1 mV s1 ;
f = 25 Hz). x/M: (1) 0; (2) 5 105 ; (3) 5 104 ; (4) 5 103 ; and
(5) 5 102 . (b) Cyclic voltammogram for the UPD of Pb on
Ag(100) in 0.5 M HClO4 + 5 103 M Pb(ClO4 )2 . ( = 10 mV s1 ).
(c) Potential variation of the descending part of curves 2 to 5 versus
cPb 2+ . (Reproduced from Ref. [43] with permission.)
Fig. 1
397
0.2
iR
[A]
iR
[mA]
0.8
0.1
0.0
Pt
Pt/Pb
0.0
0.8
0.0
Au
Au/TI
400
iD
[mA]
iD
[mA]
398
0.4
1.6
0.8
2.4
1400 rpm
1.2
0.6
(a)
0.4
0.2
0.0
ED vs Hg/HgO
[V]
0.0
(b)
0.4
0.8
ED vs RHE
[V]
Fig. 2
(a) Currentpotential curves of O2

reduction on a AuAu rrde in 0.1 M NaOH
without (full line) and with (dashed line)
6.6 106 M Tl(I) ( = 10 mV s1 ). ER = 0.1 V.
(b) Current-potential curves of O2 reduction on a
PtPt rrde in 0.2 M HClO4 without (full lines)

and with (dashed lines) 2 105 M Pb2+
( = 40 mV s1 ). ER = 1.2 V. (Reproduced from
Refs. [44, 45] with permission.)
the change of the adsorption conguration

from end-on on bare Au to bridge
adsorption on Au partially covered by
the modiers AuOOM (M = Pb, Tl,
Bi). As mentioned earlier, this mode of
adsorption favors the OO bond cleavage.
At high ad-atom coverages the bridge
adsorption does not exist any more and
oxygen is reduced to H2 O2 as on bulk Pb,
Tl, and Bi electrodes.
The inuence of the structure of the
ad-atom layers on the reduction of oxygen was examined by using in situ optical
and scanning probe microscopies. With
these techniques reliable information on
the structure of the ad-atom layers can be
obtained during the course of oxygen or
hydrogen peroxide reduction. In most systems studied, the UPD adlattice structures
are not affected by the presence of molecular oxygen in solution. On the basis of
AFM observations, Chen and Gewirth [24]

reported that the maximum catalytic activity for H2 O2 electroreduction on Au(111)
modied by Bi UPD monolayers is observed in a narrow potential region, where
the (2 2)Bi structure is obtained. At
potentials negative to this region, the rectangular Bi structure is formed and the
reduction of H2 O2 is almost completely
suppressed. The (2 2)Bi structure enables the heterobimetallic bridge adsorption (AuOOBi), which, according to
the views of Juttner [46], may explain the
enhanced reactivity.
More recent studies combined the in
situ techniques with the use of singlecrystal electrodes in a rotating geometry,
the so-called rotating hanging meniscus electrode (RHME) [4749] or in the
most desired geometry of a rotating ringdisc electrode. This combination provides
unique possibilities for studying structural effects in O2 electrocatalysis. Adzic

et al. [50] studied the effect of Tl adlayers
on the reduction of oxygen on Au(111) in
acid solutions by using X-ray diffraction
and the RHME. They found that the reduction mechanism changes from a fourelectron reduction at a low Tl coverage to a
two-electron reduction at a higher Tl coverage. The (2 2)Tl structure is conducive
to the four-electron reduction, while the
close-packed rotated hexagonal Tl phase,
which exists at more negative potentials, is
responsible for the two-electron reduction.
The reduction of oxygen does not change
the Tl coverage, although it decreases the
in-plane diffraction intensity. This indicates that O2 molecules interact with the
Tl ad-atoms and provide direct evidence
that oxygen reduction is, indeed, an innersphere and not an outer-sphere reaction.
Finally, it is worth mentioning the
pronounced inhibition effect of oxygen reduction on Au(100) and Pt(111) electrodes
caused by bromide adsorption (Fig. 3). It
is well known that the specic adsorption of halide anions inhibit the reduction
of oxygen, but information of the inhibition mechanism in an atomic level was
attempted for the rst time by Adzic
and
Wang
[51]. As shown in Fig. 3, the
c( 2 2 2)R45 adlayer structure on
Au(100) and the (3 3)Br structure on
Pt(111) cause a complete inhibition of
the two-electron reduction on Au(100) and
the four-electron reduction on Pt(111) surfaces, respectively. The reduction can take
place only at potentials more negative than
the low potential limit of the existence
of these structures. The stability of the
structure of the Br adlayers during oxygen
reduction was checked by measuring the
X-ray intensity as a function of the potential in the absence and in the presence of
oxygen in solution.
Reduction of Nitro and Nitroso

Compounds
The reduction of nitro compounds has
played an important role in the development of organic electrochemistry. The
nitro group is one of the most easily reducible groups. In aqueous solutions the
general reduction mechanism, rst proposed by Haber, is
4.3.4.2
(2e + 2H+ )
RNO2 RNO
H2 O
(2e + 2H+ )
RNHOH
(2e + 2H+ )
RNH2
H2 O
(11)
According to this mechanism, nitro compounds are reduced to the corresponding

hydroxylamines or amines by an irreversible four- or six-electron process that
proceeds through nitroso intermediates.
Nitroso intermediates are not stable on the
electrode surface because their reduction
potential is more positive than that of the
initial nitro compounds. Azoxy, azo, and
hydrazo derivatives can also be obtained
by electrolysis, particularly in alkaline solutions, as a result of chemical follow-up
reactions rather than electrochemical reactions. The electrochemistry of all these
compounds has been discussed in reviews [5254]. Among the different nitro
compounds, nitrobenzene has played a
central role and its reduction is one of the
most thoroughly studied electrochemical
reaction. The reduction involves adsorption of reactants and intermediates and
follows direct electron-exchange mechanism on most cathodes except of platinum,
in which the reduction proceeds by an electrocatalytic hydrogenation mechanism.
The inuence of ad-atoms on the reduction of nitro and nitroso compounds on
399

(i)
Au(100), Br
(i)
Pt(111), Br
50
0
i
[A]
j
[A cm2]
0
(3 3)Br
4
50
0.8
0.0
0.4
0.0
(ii)
c(2 22)
0.0
0.4
i
[mA]
Intensity
(ii)
j
[mA cm2]
400
(iii)
Pt(111), Br
0.8
Au(100), Br
0.2
Pt(111)
Au(100)
1.2
1.4
0.0
(a)
0.4
0.8
E vs NHE
[V]
0.0
(b)
0.4
0.8
E vs NHE
[V]
(a) (i) Voltammetry curve of the Au(100) electrode in 0.1 M HClO4 + 20 mM Br

( = 10 mV s1 ). (ii) X-ray intensity at (1/2, 1/2, 0.1) position in the absence (dashed line) and
in the presence (full line) of oxygen. (iii) O2 reduction on Au(100) without and with 20 mM Br
( = 20 mV s1 ). (b) (i) Voltammetry curve of the Pt(111) electrode in 0.1 M HClO4 + 20 mM
Br ( = 20 mV s1 ). (ii) O2 reduction on a hanging meniscus rotating Pt(111) disc electrode
in 0.1 M HClO4 without and with 20 mM Br ( = 20 mV s1 ; f = 625 rpm). (Reproduced
Fig. 3
Ag, Au, and Pt electrodes has been studied

during the last two decades [5561]. It was
found that ad-atom layers of Pb, Tl, and Bi
deposited at underpotentials have a strong
inuence on the performance of the above
substrates for the nitro and nitroso group
electroreduction, both in terms of the improvement of the catalytic activity and the
catalytic selectivity of these electrodes.
Reduction of nitrobenzene on Ag/Pb(upd)
and Au/Pb(upd) The electrocatalytic inuence of Pb(upd) adlayers on the
reduction of nitrobenzene on Ag [55] and
Au [56] electrodes was studied in acidic
solutions. On the bare metals, nitrobenzene is reduced to aniline. The UPD of

Pb on Ag and Au causes a partial inhibition of nitrobenzene reduction. The step of
the reduction of the intermediate hydroxylamine to aniline is completely inhibited
when the silver and gold substrates are
screened off by a lead monolayer. Figure 4
shows the polarization curves for the nitrobenzene and nitrosobenzene reduction
on Ag(111) and Ag(111)/Pb(upd) electrodes. The limiting current of nitrobenzene falls from the six-electron level
to the four-electron level at potentials
more negative than the third UPD peak
j
[mA cm2]
0.0
0.1
2 34 5
0.2
(a)
j
[mA cm2]
0.0
0.1
1
(b)
0.2
j
[mA cm2]
0.1
0.0
0.1
0.2
(c)
0.0
0.2
0.4
E vs NHE
[V]
Steady state Currentpotential curves for (a) PhNO2

(5 104 M) and (b) Ph-NO (5 104 M) reduction on Ag(111)
in 0.2 M LiClO4 + 7.5 103 M HClO4 in methanol without
Pb(ClO4 )2 (1) and with Pb(ClO4 )2 ; (2) 5 104 ; (3) 103 ;
(4) 2.5 103 ; and (5) 5 103 M. (c) Cyclic voltammogram for
the UPD of Pb on Ag(111) in the same electrolyte with
2.5 103 M Pb(ClO4 )2 ( = 10 mV s1 ). (Reproduced from
Ref. [55] with permission.)
Fig. 4
at which complete ad-atom coverage is

obtained. In this potential region, lead
forms on Ag(111) a rotated hexagonal
close-packed structure as revealed by
X-ray diffraction [15] and STM [20] studies. Also, the second reduction wave
of nitrosobenze, which corresponds to
the reduction of phenylhydroxylamine to
401
402
aniline, is completely suppressed in the

same potential region.
Reduction of aromatic nitro and nitroso compounds on Pt/M(upd) In contrast to silver
and gold electrodes, interesting catalytic
effects were found for the reduction of
aromatic nitro compounds on Pt surfaces
modied by Pb, Tl, and Bi UPD adlayers
in aqueous solutions [5759]. On clean Pt
the reduction of nitro compounds occurs
in the potential region where hydrogen
is evolved and no separate well-formed reduction waves are obtained. On Pt/M(upd)
the reduction begins at more positive potentials and diffusion-controlled limiting
currents are obtained. Generally, the catalysis of the reduction of nitro compounds
was attributed [57] to the change of the
nitro group reduction mechanism from
catalytic hydrogenation on clean Pt to direct electron-transfer mechanism on Pt
surfaces covered by the UPD adlayers, as
on metals with high hydrogen overvoltage
(Scheme 1).
In the case of phenyldinitromethane [58]
the UPD-modied surfaces exhibit also
remarkable electrocatalytic selectivity as
regards the nal reduction products under
electrolysis conditions. From the two
competitive reaction routes
NO2
PhCH
+ 2e + 2H+
NO2
PhCH2NO2 + HNO2
(12)
(Pt): Ph-NO 2 + 4Pt
(Pt/Mads): Ph-NO 2
Scheme 1
**
Ph-N
*
O
*
O
2e + 2H+
Ph-NO
H2O
2Hads
H2O
2e + 2H+
NO2
PhCH
+ 2e + 2H+
NO2
NOH
PhC
NO2
+ H2O
(13)
Reaction (12), producing phenylnitromethane, predominates on Pt/Tl(upd),
while on Pt/Pb(upd) the major reaction is the reduction to -nitrobenzaldoxime.
Also, in the case of the reduction of
benzofuroxan [59] the UPD adlayers of Pb,
Tl, and Bi strongly modify the catalytic
activity and selectivity of the platinum
electrode. Figure 5 shows the rotatingdisc voltammograms for benzofuroxan
reduction on Pt and Pt/M(upd), M = Pb,
Tl, Bi.
On bare platinum the reduction proceeds via catalytic hydrogenation to
o-nitroaniline (Scheme 2), which is reduced further to o-phenylenediamine. On
the other hand, on UPD-modied platinum surfaces, the reduction occurs at
much more positive potentials and follows
the electronation mechanism, that is, the
direct exchange of electrons between the
dinitroso tautomeric form and the modied electrode surface. The rst step is now
the reduction to o-benzoquinone dioxime
that appears as a stable intermediate over
a wide potential range (0.60 to 0.45 V)
before it is reduced further to the nal
products.
*** *
Ph-N-O
mHads
r.d.s
Ph-NHOH
2e + 2H+
Ph-NH2
H2O
Products
Reduction of nitrobenzene on Pt and Pt/Mads (M = Pb, Tl).

0.0
(a)
0
1.5
(b)
3.0
4
j
[mA cm2]
j
[mA cm2]
(c)
0
4.5
(d)
0.2
6.0
0.0
0.4
0.8
1.2
E vs NHE
[V]
7.5
0.0
0.4
0.8
E vs NHE
[V]
Current-potential curves for benzofuroxan (103 M) reduction on a
Pt rotating-disc electrode in 0.5 M HClO4 without (1) and with (2) 103 M
TlClO4 ; (3) 103 M Pb(ClO4 )2 ; (4) 103 M Bi(ClO4 )3 ( = 10 mV s1 ;
f = 25 Hz). Inset: Cyclic voltammograms for the UPD of Pb, Tl, and Bi on
Pt in 0.5 M HClO4 . (a) Without ions; (b) with 5 104 M Pb2+ ; (c) with
5 104 M Tl+ ; and (d) with 5 104 M Bi3+ . Scan rate = 50 mV s1 .
(Reproduced from Ref. [59] with permission.)
Fig. 5
Pt
NO2
O*
O + 2Pt
N
2Hads
N*
NH2
O*
**
N
NO2
O*
+ 4Pt
mHads
Products
NH2
NH2
Pt/M(upd)
O
N
2e + 2H+
O
N
NOH
NO
NO
NOH
Scheme 2 Reduction of benzofuroxan on Pt and Pt/M(upd)

(M = Pb, Tl, Bi).
403

O2N
N
N
NO2
NO2
O2N
NO2
N
H
The behavior of these compounds differs

from that of nitrobenzene. The strong
electron-attracting heterocyclic nuclei, on
the one hand, renders the nitro group more
easily reducible and, on the other hand,
helps to stabilize the hydrated form of the
nitroso intermediates.
Reduction of heterocyclic nitro compounds

on Au/M(upd) Overlayers of Pb and
Tl ad-atoms, deposited on Au in the
underpotential region, exert pronounced
catalytic effects on the reduction of
some N -heterocyclic nitro compounds [56,
60, 61].
160
0
(i)
3
4
10
20
ER = 1.6 V
w1/2
[s1/2]
0.2
jD
[mA cm2]
(ii)
0.1
0.0
1
2
10
ER = 0.5 V
20
30
2
3
0.1
0.2
0.0
0.4
E vs SHE
[V]
(a) (i) Current-potential curves for the

reduction of 3-nitro-1,2,4-triazole (103 M) on a
Au disc electrode in 0.2 M HClO4 without
(dashed lines) and with (full lines) 5 104 M
Pb(ClO4 )2 ( = 20 mV s1 ). f/Hz: (1) 12.5;
(2) 25; (3) 50; and (4) 83.3. The inset shows the
jL versus 1/2 ( = 2 f) plots at (1) 0.1 V and
(2) +0.21 V. (ii) Cyclic voltammogram for the
Fig. 6
2
1
(a)
80
1
2
n=4
0
Au
Au Pb
n=2
iR
[A]
1
2
iR
[A]
jL
[mA cm2]
j
[mA cm2]
j
[mA cm2]
404
0.8
(b)
0.0
0.4
ED vs SHE
[V]
UPD of Pb on Au in 0.2 M
HClO4 cPb 2+ = 5 104 M; = 50 mV s1 .
(b) AuAu rotating ring-disc (N = 0.18)
measurements for 3-nitro-1,2,4-triazole
(103 M) in 0.2 M HClO4 + 5 104 M
Pb(ClO4 )2 ( = 20 mV s1 ). f/Hz: (1) 12.5;
(2) 25; and (3) 50. (Reproduced from Ref. [56]
with permission.)
reaction, the limiting current exhibits kinetic character.

On Au partially covered by Pb ad-atoms
(E = 0.4 to 0.2 V), an electrocatalytic
mechanism appears to be operative. The
reduction proceeds through chemisorption of the nitro group and reductive
cleavage of one of the two NO bonds,
which leads directly to the hydroxylamino
compound and gives diffusion-controlled
four-electron limiting currents. As in the
case of oxygen reduction, a bridge adsorption of the nitro group with a bridge
complex involving a substrate (Au) and an
adsorbed atom (M = Pb, Tl) was assumed
to explain the strong chemisorption on Au
The effect of covering a gold disc with

Pb ad-atom layers on the reduction of
3-nitro-1,2,4-triazole is shown in Fig. 6(a).
Nitrotriazole undergoes a four-electron reduction both on Au and Au/Pb(upd) [56].
However, the reduction mechanism depends on the UPD ad-atom coverage.
On bare Au and Au surfaces with almost complete ad-atom layer (E < 0.15 V),
the four-electron reduction follows the
ECE mechanism (Scheme 3). The reaction
proceeds through a N ,N -dihydroxyamine
derivative (the hydrated form of the nitroso intermediate) that is reduced further
after an irreversible loss of water. Because of the slow rate of the dehydration
Au and Au/M(upd) ( high coverage)
OH
OH
N
NO2
N
2e + 2H
OH
kfast
kslow
N
N
NHOH
NO
N
2e + 2H+
H
Au/M(upd) (low coverage)
O
Ar
N
2e
r.d.s
Ar
HO
2H+
OH
2e + 2H+
fast
ArNHOH + H2O
Ar
N
(M = Pb, Tl)
Au M
Scheme 3 Reduction mechanism of 3-nitro-1,2,4-triazole on Au and Au/M(upd)
(M = Pb, Tl).
H2O
405
406
partially covered by ad-atoms. Analogous

behavior was also observed with the other
heterocyclic nitro compounds [60, 61].
The ring-disc measurements presented
in Fig. 6(b) conrm the above mechanisms. The nal product of the reduction of nitrotriazole both at potentials of
diffusion-controlled and kinetically controlled limiting currents is hydroxylaminotriazole, which is collected by the ring
electrode at ER = 1.6 V. In the potential
region, where the current is kinetically
controlled, the intermediate nitrosotriazole was detected at ER = 0.5 V. In fact,
the nitroso compound detected on the ring
electrode is not that formed on the disc,
because it is rapidly reduced further, but
that produced from the dihydroxyamine
intermediate upon its dehydration across
the gap from disc to ring. Dihydroxyamine
itself is not oxidized back to nitrotriazole
on the ring, as one would expect, because
it is rapidly stabilized by intramolecular
hydrogen bonds.
The reduction of the above N -heterocyclic nitro compounds on Au modied by
ad-atoms may by used as model reactions
for studying the surface geometric and
electronic structure in electrocatalysis.
By using Au, Ag, or Pt single-crystal
electrodes, it could be possible to correlate
the catalytic activity with the superlattice
structures of the ad-atoms.
Oxidation of Small Organic
Molecules
The electrooxidation of simple onecarbon molecules (C1 molecules), such
as methanol, formaldehyde, and formic
acid, became the subject of considerable
and continuous interest. This is quite
reasonable since these compounds, particularly methanol, are potential fuels for
fuel cells. The complete oxidation of the
4.3.4.3
rst two fuels to CO2 requires the donation of oxygen, which comes either
from water or hydroxyl ions depending
on the pH of the solution. In aqueous
acid solutions the overall reaction may be
written as
CHx Oy + (2 y)H2 O
CO2 + [2(2 y) + x](H+ + e ) (14)
Since all hydrogen atoms must be abstracted from the fuel, the electrocatalyst
must support the dehydrogenation. Electrodes that can be used as efcient anodes
in C1 electrocatalysis are those with great
afnity toward dehydrogenation reactions,
namely, Pt and some other metals of the
Pt-Group. Unfortunately, the current densities obtained with these electrodes are
very low owing to poisoning effects. The
main effect of poisoning is to block sites
on the electrode surface.
Many efforts have been made to identify
the nature of the poisoning species. The
most successful techniques are the in situ
IR reectance spectroscopies. Intensive
work has been done both on polycrystalline
and single-crystal electrodes. The work
done has been reviewed by Bewick and
Pons [62], Beden and Lamy [63], and more
recently by Sun [64]. It is now recognized
that CO (either linearly or bridge-bonded
to the surface) is the poison and not formyl
species (COH), as believed previously [65,
66]. Carbon monoxide is formed by dehydration in the case of HCOOH and by
dehydrogenation in the case of HCHO and
CH3 OH. The CO formation reaction constitutes the one branch of the so-called
dual mechanism, originally suggested
by Capon and Parsons for the electrochemical oxidation of C1 molecules [65, 66]. The
other branch is the reaction through reactive intermediates.

[P]
C1 molecules
CO2
[RI]
[P] is the poison and [RI] the reactive

intermediates. All species involved in the
mechanisms may be considered as reactive
intermediates, if they are not strongly
bonded to the surface and are oxidized
to CO2 .
The importance of the oxidation of C1
molecules in electrochemical energy conversion has prompted intense research
efforts to nd new more active electrocatalysts or to improve the activity of
existing catalysts, for instance, platinum.
Efforts to enhance the catalytic properties of metal electrodes toward C1 electrooxidation include both bulk alloys and
pure metal surfaces modied by ad-atoms
(modiers), usually deposited in the underpotential region.
Generally, the ad-atoms cause positive
catalytic effects with signicant enhancement of the electrocatalytic activity of platinum in several cases. Several reviews have
been published on this subject [2628,
67]. Very recently Ross [68] and Jarvi and
Stuve [29] have discussed the more recent advances in our understanding of
the fundamentals of C1 electrocatalysis
by ad-atoms. The different types of enhancement (third-body effect, bifunctional
mechanism, poison destabilization, and
electronic modication) are well documented. The new information obtained
from the in situ spectroscopic studies
about the nature of poisons and the dependence of their coverages on potential,
as well as the use of single-crystal electrodes with dened surface structure and
specic reactivity, enables a deeper insight in the electrocatalysis by ad-atoms.
As a general rule, one can say that, except for methanol, the more susceptible
the electrode to poisoning, the more important is the enhancement caused by

ad-atoms.
Formic acid electrocatalysis by ad-atoms
The electrooxidation of formic acid on
noble metals, such as Pt, Pd, Rh, Ir, is
a classical example of a self-poisoning
reaction. The currents obtained are very
low and depend strongly on the potential
sweep direction (anodic or cathodic) and
the crystallographic orientation of the
electrode surface. The Pt(111) face with
the hexagonal lattice structure was found
to be the surface that gives the higher
currents and is less sensitive to poisoning
effects [69, 70]. Foreign metal ad-atoms
cause striking catalytic effects on the
oxidation of formic acid on polycrystalline
and single-crystal electrodes. Ad-atoms
may be deposited either by UPD or
irreversibly by immersing the electrode
into the metal solution at open circuit.
The early work was reviewed by Adzic [26],
while the most recent work by Jarvi and
Stuve [29].
Modiers that can enhance the electrocatalytic activity of Pt, Pd, Rh, and
Ir electrodes are Pb, Tl, Bi, Cd (see
Ref. [26] and references therein). Among
these modiers, lead appears to be the
stronger promoter in formic acid electrocatalysis [26, 7174]. Figure 7 shows the
effect of the UPD of Pb on formic acid
electrooxidation on Pt. As seen, very high
currents are obtained on Pt/Pb(upd) in
the range between 0.3 and 0.7 V. Another
important observation is the coincidence
of the currents during the positive and
the reverse negative potential scans, which
is considered as an evidence of reduced
poisoning effects.
The more recent studies on formic acid
electrooxidation refer to Pt single-crystal
electrodes modied by ad-atoms with
407

80
Pt
Pt/Pb
60
j
[mA cm2]
408
40
20
0
0.0
0.4
0.8
1.2
1.6
E vs NHE
[V]
Voltammetric curves for formic acid (0.265 M)
oxidation on a Pt electrode in 1 M HClO4 without (dashed
line) and with (full line) 103 M Pb2+ ions. Scan rate
= 50 mV s1 . (Reproduced from Ref. [71] with permission.)
Fig. 7
controlled coverage, which are adsorbed

irreversibly on the Pt surface. Ad-atoms
deposited in this way include Pd [75, 76],
Pb [77], Sb [78], As [79], and Bi [78, 80].
Palladium is the stronger promoter for
formic acid oxidation. Indeed, adsorbed
Pd drastically enhances the electrocatalytic
activity of the Pt(100) electrode toward
formic acid oxidation (Fig. 8). High current densities are obtained at low potentials
(0.22 V vs. RHE), where bare Pt is not
active. The catalytic activity reaches a maximum at Pd = 0.6 [75]. Multilayers of
Pd, epitaxially grown on Pt single-crystal
electrodes, also exhibit enhanced catalytic
activity toward formic acid oxidation [76]
compared to that of massive Pd singlecrystal electrodes. Palladium overlayers

show higher resistivity against CO poisoning than bare Pd, which was attributed
to electronic effects.
Formaldehyde electrocatalysis by ad-atoms
Poisoning effects during the electrooxidation of formaldehyde were mainly
observed in acid solutions. Ad-atoms deposited at underpotentials enhance the
electrocatalytic activity of the Pt electrode
toward HCHO oxidation [8183]. According to Motoo and Shibata [82], ad-atoms
that do not adsorb oxygen (e.g. Pb, Tl,
Bi) have only a small inuence resulting
(a)
j
[mA cm2]
j
[mA cm2]
20
10
0.1
0.0
0.5
E
[VRHE]
0.1
0.1
0.5
0.9
j
[mA cm2]
j
[mA cm2]
20
10
0.2
0.0
0.2
0.1
(b)
0.5
0.5
E
[VRHE]
0.9
E vs RHE
[V]
Fig. 8 (a) Cyclic voltammograms of Pt(100) (dashed line) and

Pt(100)/Pdads ( = 0.32) (full line) in 0.25 M HCOOH + 0.5 M
H2 SO4 . (b) Cyclic voltammograms of Pd (dashed line) and
Pt(100)/Pdads ( = 1.06) (full line) in the same electrolyte. The
insets show blank voltammograms in 0.5 M H2 SO4 . (Reproduced
from geometric factors (third-body mechanism), while oxygen adsorbing ad-atoms

(e.g. Ru, Ge, Sn, Sb) give rise to a
greater enhancement. The ad-atoms facilitate the donation of oxygen through
the bifunctional mechanism, thus lowering the oxidation overpotential.
Methanol electrocatalysis by ad-atoms The
oxidation of methanol takes place at potentials more positive than that needed
for formic acid. Even on platinum, which
is the most active electrocatalyst, the oxidation occurs at high overpotentials and
signicant currents are obtained only in
the potential region of Pt-oxide formation.

Many efforts have been made to promote
the activity of Pt by adding a second metal,
either as ad-atoms or by alloying it with Pt.
These studies are extensive and overviews
of the literature may be found in Refs. [26,
28, 29, 68].
Here, we will restrict ourselves to pointing out that ad-atoms, such as Pb and
Bi, which are strong promoters for formic
acid electrocatalysis, only slightly increase
the activity of Pt for methanol oxidation [8386]. The most successful promoter for methanol electrocatalysis is Ru
and perhaps Sn, albeit not as ad-atoms.
409
410
Pt-Ru alloys are now recognized as the

best electrocatalysts for methanol oxidation exhibiting lower overpotentials and
long-term stability, both important factors
for direct methanol fuel cells. In a series of
excellent papers, Gasteiger et al. [8790]
have provided valuable information regarding the reactivity of PtRu alloys. The
effect of Ru on the kinetics of CH3 OH
electrooxidation was rationalized in terms
of the bifunctional action and a series reaction mechanism. The Pt sites act as centers for the adsorption/dehydrogenation
of methanol-producing poison and carbonaqueous intermediates, and Ru sites
act as centers for adsorbing oxygencontaining species at less positive potential
than Pt. The oxygen species on Ru sites oxidize both CO and the reactive carbonaqueous intermediates by surface diffusion reactions. The surface composition of Ru alters the balance between the reaction rates
of CO formation and oxidation with the 7
to 10% Ru surface being the optimum surface composition for enhanced reactivity.
Oxidation of Bigger Organic
Molecules
Other organic molecules, whose oxidation
can be catalyzed by ad-atoms, include primary and secondary alcohols [85, 9193],
diols and polyols [94100], and different monosaccharides [101105]. Aldehydes are oxidized at more positive potentials
compared to the oxidation potentials of
primary alcohols and their oxidation is
catalyzed only by oxygen-adsorbing adatoms [106, 107].
Generally, submonolayers of ad-atoms
of heavy metals (such as Pb, Tl, Bi,
Cd, etc.) deposited in the underpotential
region increase the catalytic activity of Pt
toward the electrooxidation of alcohols and
sugars both in aqueous, acid, and alkaline
solutions. The catalytic effect is more
4.3.4.4
pronounced in alkaline solutions, in which

the currents on the modied surfaces are
much higher compared to those obtained
on platinum without ad-atoms. It should
be noted that in alkaline media the
catalytic effect is observed only at very
low concentrations (around 106 M) of
heavy metal ions in the solution. At higher
concentrations, negative catalytic effects
are observed rather than positive catalysis.
In general, the main products formed
during the oxidation of primary and
secondary alcohols are the corresponding
aldehydes and ketones, while the main
product of the oxidation of glucose is
gluconic acid.
In acid solutions the catalytic effect
caused by ad-atoms is characterized by
the shift of the onset of oxidation toward
less positive potentials and an increase
of the peak current density by a factor
of 2 to 5, depending on the kind of adatom and the organic substance being
oxidized. In alkaline solutions the catalytic
effect is mainly recognized by an immense
increase of the oxidation current, especially
with Pb and Bi ad-atoms. In alkaline
solutions the poisoning effect is not
so signicant, since strongly adsorbed
intermediates are not involved in the
oxidation mechanism.
On the contrary, in acid solutions
strongly adsorbed species are supposed
to be formed during the oxidation of alcohols and sugars. The catalytic effect
was generally interpreted in terms of a
decrease of the electrode poisoning. The
nature of the poisoning species produced
during the dissociative chemisorption of
aliphatic alcohols [108111] and ethylene
glycol [112] on Pt has been studied by
in situ reectance spectroscopies. It was
demonstrated that CO is the main poisoning species and that the coverage of the
surface by adsorbed CO decreases with
the increase of the carbon chain. The formation of CO results from the dissociative
chemisorption and the C1 C2 bond breaking. In the case of D-glucose oxidation on
Pt, the poisoning species is believed to be
a gluconolactone type intermediate [113,
114], which is also considered as the reactive intermediate producing gluconic acid
upon hydrolysis.
O
R
O Pt
OH
C1 + 2Pt
(2H+ + 2e)
C1 Pt
R
O
R
C1 O + 2Pt
(15a)
OH
O OH
HO
OH
+ 2Pt
OH
OH
O O
HO
OH
OH
Pt
Pt
OH
O
HO
OH
O + 2Pt
Hydrolysis
OH
(15b)
However, more recent studies by in situ IR
spectroscopy [115, 116] have demonstrated
that the catalytic decomposition of glucose
on Pt produces CO, which should be the

major adsorbate responsible for poisoning
of the electrode surface.
The structure of the organic molecules
should also be considered in interpreting the catalytic effects caused by the
modiers. This can be deduced from the
comparative studies (performed in acid solutions) of the UPD effect on the oxidation
of two different series of organic compounds with the same functional group,
such as aliphatic primary alcohols [85] and
monosaccharides [103]. The lack of uniform catalytic behavior (based on current
densities) leads to the conclusion that besides the third-body effect of the ad-atoms
that is physical in nature the specic interacting forces of the reaction intermediates
with the modied electrode surface must
also play a signicant role in the electrocatalytic process.
Finally, ad-atom-modied electrodes
may also exhibit selectivity regarding the
products obtained after electrolysis. Selective oxidation is particularly important in
organic electrosynthesis. Two examples of
selective oxidation of organic compounds
will be discussed: the oxidation of glycolaldehyde in acid solutions on Pt and
Pt/Sbads [117] and the oxidation of gluconic acid in alkaline solutions on Pt and
Pt/Pbads [118].
On smooth Pt the main product of glycolaldehyde oxidation is glyoxal, while on
Pt/Sbads it is glycolic acid [117] (Scheme 4).
The Sb ad-atoms increase the reaction
rate of both the oxidation of the CHO
group to COOH group and the oxidation of
the CH2 OH group to CHO group, but the
enhancement of the rst reaction is much
Pt
Selective oxidation
of glycolaldehyde on Pt and
Pt/Sbads .
Scheme 4
CH2OHCHO
Pt/Sbads
CHOCHO (glyoxal)
CH2OHCOOH (glycolic acid)
411
412
more signicant than that of the second

one. Thus, the CHO group is selectively
oxidized on the Pt electrode modied by
the Sb ad-atoms.
In the case of D-gluconic acid [118], analysis of the oxidation products by ionic chromatography has shown that on smooth
Pt the main products are oxalic, tartaric,
5-keto-D-gluconic, and D-glucuronic acids.
On Pt modied by Pb ad-atoms, D-gluconic
acid is selectively oxidized to 2-keto-Dgluconic acid, which is obtained with
relatively high yield (33%). The reaction
rate and the orientation of D-gluconic acid
depend on the nature of the electrode. On
bare Pt the reaction rate is very low and
the oxidation occurs preferably at C6 yielding D-glucuronic acid. On Pt/Pbads the
rate of reaction is signicantly increased
and the oxidation occurs at C2 producing 2-keto-D-gluconic acid. On Pt/Pbads
no D-glucuronic acid was obtained. This
means that oxidation at C6 does not occur at all, indicating that the presence of
ad-atoms changes the way that the reacting molecules link and interact with the
electrode surface.
Reversible Redox Reactions
It is known that some redox systems show
poor reversibility on platinum. This is due
to the fact that a monolayer of chemisorbed
species from the redox components is
formed that hinders the exchange of
electrons between the rest of the diffusing
molecules and the electrode.
Underpotentially deposited ad-atom layers of Pb, Tl, and Bi improve markedly
the reversibility of the redox systems: hydroquinone/p-benzoquinone [119,
120], pyrocatechol/o-benzoquinone [120],
adrenaline/adrenalinequinone [121], hexahydroxybenzene/tetrahydroxy-1,4-benzoquinone [122], p-benzoquinone dioxime/
p-dinitrosobenzene [123], as well as the
4.3.4.5
oxidation of some dienolic compounds,

such as ascorbic acid [124, 125] and rhodizonic acid [122]. Figure 9 presents an
example of how the UPD of Tl affects the
cyclic voltammetric behavior of the system pyrocatechol/o-benzoquinone on the
Pt electrode.
The improvement of the reversibility
of the redox systems was attributed by
Kokkinidis et al. [120, 122] to the change
of the reaction mechanism. The reaction
mechanism changes from an inner-sphere
mechanism involving adsorbed intermediates on bare Pt to an outer-sphere one
without complications from adsorption
of the reacting molecules on Pt/Mads
(M = Pb, Tl, Bi).
Inorganic Electrochemical
Reactions
Inorganic electrochemical reactions inuenced by ad-atoms include the oxidation
of hydrazine and the reduction of some
oxo-anions.
The inuence of the UPD of heavy metals on the oxidation of hydrazine on Pt was
studied in aqueous acid and alkaline solutions [126]. Adlayers of Pb, Tl, Bi, Cu,
Cd, Ag cause a complete inhibition of
the four-electron oxidation of hydrazine
to nitrogen. On the Pt-modied surfaces
the oxidation commences at sufciently
positive potentials where appreciable dissolution of ad-atoms occurs.
On the contrary, UPD monolayers of
Pb, Tl, and Bi markedly catalyze the reduction of chromates (Cr2 O7 2 ) [127] and
persulfates (S2 O8 2 ) [127, 128] on Pt. The
E1/2 of the reduction waves shifts toward
more positive potentials and diffusioncontrolled limiting currents are attained
within the UPD region. The catalytic
activity was interpreted in terms of preventing the covering of the Pt surface
by species resulting from the dissociative
4.3.4.6
jp
[mA cm2]
j
[mA cm2]
3
2
1
0
20
10
v
[mV s1]
1
Ep
[V]
0.93
0.81
0.79
0.89
1.4
Pt
0.91
Pt/TI
0.77
1.0
0.6
0.2
log (v )
[V s1]
0.2
0.6
1.0
E vs NHE
[V]
Cyclic voltammograms for pyrocatechol (2.5 103 M) on Pt in
0.5 M HClO4 without (dashed lines) and with (full lines) 103 M TlClO4 .
Scan rate /mV s1 : (1) 50; (2) 100; (3) 200; (4) 300; and (5) 400. The
insets show plots of jp versus 1/2 and Ep versus log without (1) and
with (2) TlClO4 . (Reproduced from Ref. [120] with permission.)
Fig. 9
chemisorption of Cr2 O7 2 and S2 O8 2

anions, thus leading to faster electron
exchange for the charge-transfer step involved in the reaction mechanisms. In the
case of dichromate reduction in acid solutions, a remarkable discontinuity in the
voltammetric behavior was observed when
the potential was scanned in the positive
direction. This behavior is related to the anodic desorption of the UPD monolayers,
which in the presence of Cr2 O7 2 ions occurs very abruptly owing to the competitive
adsorption of dichromates.
Finally, the reduction of NO3 anions
on Ag(hkl) electrodes is strongly inhibited
by Pb and Tl ad-atom layers [129]. The
inhibition effect was correlated to the superlattice structures of the UPD ad-atoms
on the different Ag single-crystal surfaces.
4.3.5
Conclusion
The inuence of ad-atoms on reaction rates

and mechanisms has attracted the interest
of many researchers over the last 25 years.
The most extensively studied reactions are
the reduction of oxygen and the oxidation
of organic fuels because of the importance
of these reactions in electrochemical energy conversion. The challenge was to
improve the catalytic activity of Pt and
413
414
to develop new, more efcient electrocatalysts for these reactions. On the other
hand, studies on ad-atom-modied electrodes have allowed us to understand how
surface geometric and electronic structure
inuences catalytic activity and reactivity.
Signicant progress has been made toward this direction by using single-crystal
electrodes with well-dened structures of
substrate and ad-atom layers in combination with in situ spectroscopies and
scanning probes. These techniques have
provided information on reaction intermediates (reactive or poisons) and surface
properties under operating conditions that
can help us to directly correlate the surface structure and composition with the
catalytic activity. We already know enough
about oxygen reduction, but there is still
much to learn about the oxidation of organic fuels.
Except the enhancement of the electrocatalytic activity, ad-atoms may also
improve the selectivity of the surface of
the electrocatalysts regarding the nal reaction products. Little work has been done
in this direction. Probably in the future,
electrodes modied by ad-atoms will be
more attractive as potential catalysts in selective electrocatalysis. The improvement
of the selectivity would require thorough
mechanistic studies in order to get a
better understanding of the various factors that may inuence the selectivity of
these electrodes.
References
1. J. O.M. Bockris, A. K. N. Reddy, Modern
Electrochemistry, Plenum Press, New York,
1977, Vol. 2.
2. W. J. Lorenz, H. D. Herrman, N. Wuthrich
11671177.
3. D. M. Kolb in Advances in Electrochemistry and Electrochemical Engineering (Eds.:
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
H. Gerischer, C. W. Tobias), Wiley, New

York, 1978, pp. 125271, Vol. 11.
K. Juttner, W. J. Lorenz, Z. Phys. Chem. N.F.
1980, 122, 163185.
S. Szabo, Int. Rev. Phys. Chem. 1991, 10,
207248.
J. W. Schultze, F. D. Koppitz, Electrochim.
Acta 1976, 21, 327343.
M. F. Toney, J. G. Gordon, M. G. Samant
et al., Langmuir 1991, 7, 796802
J. Clavilier, R. Faure, G. Guinet et al., J.
J. Clavilier, D. Armand, S. C. Sun et al., J.
A. T. Hubbard, R. M. Ishikawa, J. Katekaru,
P. N. Ross, J. Electrochem. Soc. 1979, 126,
6777.
P. N. Ross, F. T. Wagner in Advances in Electrochemistry and Electrochemical Engineering
(Eds.: H. Gerischer, C. W. Tobias), Wiley,
New York, 1984, pp. 69112, Vol. 13.
R. R. Adzic in Modern Aspects of Electrochemistry (Eds.: R. E. White, J. O. M. Bockris,
1990, pp. 163236, Vol. 21.
R. Parsons, G. Ritzoulis, J. Electroanal.
Chem. 1991, 318, 124.
M. G. Samant, M. F. Toney, G. L. Borges
et al., J. Phys. Chem. 1988, 92, 220225.
M. F. Toney, J. G. Gordon, M. G. Samant
et al., Phys. Rev. 1992, B45, 93629374.
J. Wang, B. M. Ocko, A. J. Davenport et al.,
Phys. Rev. 1992, B46, 10 32110 338.
M. P. Green, K. J. Hanson, D. A. Scherson
et al., J. Phys. Chem. 1989, 93, 21842187.
M. Binggeli, D. Carnal, R. Nyffenegger
et al., J. Vac. Sci. Technol. 1991, B9,
19851992.
U. Muller, D. Carnal, H. Siegenthaler et al.,
Phys. Rev. 1992, B46, 12 89912 901.
X. Gao, A. Hamelin, M. J. Weaver, Phys.
Rev. 1992, B46, 70967102.
K. Sashikata, N. Furuya, K. Itaya, J. Electroanal. Chem. 1991, 316, 361368.
D. M. Kolb, Electrochim. Acta 2000, 45,
23872402.
C.-H. Chen, A. A. Gerwirth, J. Am. Chem.
Soc. 1992, 114, 5439, 5440.
X. Xia, T. Iwasita, J. Electrochem. Soc. 1993,
140, 25592565.
R. R. Adzic in Advances in Electrochemistry and Electrochemical Engineering (Eds.:
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
H. Gerischer, C. W. Tobias), Wiley, New

York, 1984, pp. 159260, Vol. 13.
G. Kokkinidis, J. Electroanal. Chem. 1986,
201, 217236.
R. Parsons, T. VanderNoot, J. Electroanal.
Chem. 1988, 257, 945.
T. D. Jarvi, E. M. Stuve in Electrocatalysis
(Eds.: J. Lipkowski, P. N. Ross), Wiley, New
York, 1998, pp. 75153.
N. D. Lang, S. Holloway, J. K. Norskov,
Surf. Sci. 1985, 150, 2428.
M. P. Kiskinova, Stud. Surf. Sci. Catal. 1992,
70, 1137.
C. G. Vayenas, I. V. Yentekakis in Handbook of Heterogeneous Catalysis (Eds.: G. Ertl,
H. Knozinger, J. Weitkamp), Wiley-VCH,
Weinheim,
New
York,
1997,
pp. 13101325, Vol. 3.
A. Damjanovic in Modern Aspects of Electrochemistry (Eds.: J. O.M. Bockris, B. E.
Conway), Plenum Press, New York, 1969,
pp. 369483, Vol. 5.
M. R. Tarasevich, A. Sadkowski, E. Yeager
in Comprehensive Treatise of Electrochemistry (Eds.: B. Conway, J. O.M. Bockris,
E. Yeager et al.), Plenum Press, New York,
1983, pp. 301332, Vol. 7.
K. Kinoshita, Electrochemical Oxygen Technology, Wiley, New York, 1990.
A. J. Appleby, J. Electroanal. Chem. 1993,
357, 117179.
R. Adzic in Electrocatalysis (Eds.: J. Lipkowski, P. N. Ross), Wiley, New York, 1998,
pp. 197242.
E. Yeager, Electrochim. Acta 1984, 29,
15271537.
P. Fischer, J. Heitbaum, J. Electroanal.
Chem. 1980, 112, 231238.
N. Anastasijevic, Z. M. Dimitrijevic, R. R.
Adzic, Electrochim. Acta 1992, 37, 457464.
G. Kokkinidis, D. Jannakoudakis, J. Electroanal. Chem. 1984, 162, 163173.
T. Abe, G. M. Swain, K. Sachikata et al., J.
A. Swetanova, K. Juttner, J. Electroanal.
Chem. 1981, 119, 149164.
R. Amadelli, N. Markovic, R. Adzic et al., J.
S. A. S. Machado, A. A. Tanaka, E. R. Gonzalez, Electrochim. Acta 1994, 39, 25912597.
K. Juttner, Electrochim. Acta 1984, 29,
15971604.
F. El Kadiri, R. Faure, R. Durand, J. Electroanal. Chem. 1991, 301, 177188.
48. N. M. Markovic, R. R. Adzic, B. Cahan

249259.
Ross, J. Phys. Chem. 1995, 99, 34113415.
50. R. R. Adzic, J. Wang, B. M. Ocko, Electrochim. Acta 1995, 40, 8389.
51. R. R. Adzic, J. X. Wang, Electrochim. Acta
2000, 45, 42034210.
52. W. Kemula, T. M. Krygowski in Encyclopedia of Electrochemistry of the Elements (Eds.:
A. J. Bard, H. Lund), Marcel Dekker, New
York, 1979, pp. 77208, Vol. 13.
53. H. Lund in Organic Electrochemistry (Eds.:
M. M. Baizer, H. Lund), 3rd ed., Marcel
Dekker, New York, 1991, pp. 401432.
54. A. Cyr, P. Huot, J.-F. Marcoux et al., Electrochim. Acta 1989, 34, 439445.
55. G. Kokkinidis, K. Juttner, Electrochim. Acta
1981, 26, 971977.
56. G. Kokkinidis, A. Papoutsis, G. Papanastasiou, J. Electroanal. Chem. 1993, 359,
253271.
57. G. Kokkinidis, P. Jannakoudakis, Electrochim. Acta 1984, 29, 821828.
58. G. Kokkinidis, E. Coutouli-Argyropoulou,
59. G. Kokkinidis, N. Argyropoulos, Electrochim. Acta 1985, 30, 16111620.
60. A. Papoutsis, G. Kokkinidis, J. Electroanal.
Chem. 1994, 371, 231239.
61. G. Kokkinidis, A. Kelaidopoulou, J. Electroanal. Chem. 1996, 414, 197208.
62. A. Bewick, B. S. Pons in Advances in Infrared and Raman Spectroscopy (Eds.:
R. J. H. Clark, R. E. Hester), Wiley Heyden,
London, 1985, pp. 163, Vol. XII.
63. B. Beden, C. Lamy in Spectroelectrochemistry-Theory and Practice (Ed.: R. G. Gale),
Plenum Press, New York, 1988, pp.
189261, Chap. 5.
64. S.-G. Sun in Electrocatalysis (Eds.: J. Lipkowski, P. N. Ross), Wiley, New York, 1998,
pp. 243290.
65. A. Capon, R. Parsons, J. Electroanal. Chem.
1973, 44, 17.
1973, 45, 205231.
67. B. Beden, J.-M. Leger, C. Lamy in Modern
Aspects of Electrochemistry (Eds.: J. O.M.
Bockris, B. E. Conway, R. E. White), Plenum
Press, New York, 1992, pp. 97264, Vol. 22.
415
416

68. P. N. Ross in Electrocatalysis (Eds.: J. Lipkowski, P. N. Ross), Wiley, New York, 1998,
pp. 4374.
69. J. Clavilier, R. Parsons, R. Durand et al., J.
70. R. R. Adzic, A. V. Tripkovich, W. OGrady,
Nature 1982, 296, 137, 138.
71. R. R. Adzic, D. N. Simic, A. R. Despic et al.,
72. R. R. Adzic, D. N. Simic, A. R. Despic et al.,
73. D. Pletcher, V. Solis, J. Electroanal. Chem.
1982, 131, 309323.
74. M. Shibata, S. Motoo, J. Electroanal. Chem.
1985, 188, 111120.
75. M. J. Llorca, J. M. Feliu, A. Aldaz et al., J.
76. M. Baldauf, D. M. Kolb, J. Phys. Chem. 1996,
100, 11 37511 381.
77. S. A. Campbell, R. Parsons, J. Chem. Soc.,
Faraday Trans. 1992, 88, 833841.
78. E. Herrero, J. M. Feliu, A. Aldaz, J. Electroanal. Chem. 1994, 368, 101108.
79. A. Fernandez-Vega, J. M. Feliu, A. Aldaz
229240.
80. E. Herrero, A. Fernandez-Vega, J. M. Feliu
81. D. M. Spasojevic, R. R. Adzic, A. R. Despic,
82. S. Motoo, M. Shibata, J. Electroanal. Chem.
1982, 139, 119130.
83. M. Beltowska-Brzezinska,
J. Heitbaum,
W. Vielstich, Electrochim. Acta 1985, 30,
14651471.
84. B. Beden, F. Kadirgan, C. Lamy et al., J.
1987, 229, 385394.
87. H. A. Gasteiger, N. Markovic, P. N. Ross
et al., J. Phys. Chem. 1993, 97, 12 02012 029.
88. H. A. Gasteiger, N. Markovic, P. N.
Ross et al., J. Electrochem. Soc. 1994, 141,
17951803.
90. N. M. Markovic, H. A. Gasteiger, P. N. Ross
91. G. Kokkinidis, A. Papoutsis, J. Electroanal.
Chem. 1989, 271, 233247.
92. D. Sazou, N. Xonoglou, G. Kokkinidis, Collect. Czech. Chem. Commun. 1986, 51,
24442454.
93. C. T. Hable, M. S. Wrighton, Langmuir
1993, 9, 32843290.
95. F. Kadirgan, B. Beden, C. Lamy, J. Electroanal. Chem. 1982, 136, 119138.
96. F. Kadirgan, B. Beden, C. Lamy, J. Electroanal. Chem. 1983, 143, 135152.
97. M. Shibata, N. Furuya, M. Watanabe, J.
98. G. Pierre, A. Ziode, M. L. Kordi, Electrochim
Acta 1987, 32, 601606.
99. P. Ocon, B. Beden, H. Huser et al., Electrochim. Acta 1987, 32, 387394.
100. P. Ocon, B. Beden, C. Lamy, Electrochim.
Acta 1987, 32, 10951101.
101. M. Sakamoto, K. Takamura, Bioelectrochem.
Bioenerg. 1982, 9, 571582.
102. N. Xonoglou, G. Kokkinidis, Bioelectrochem.
Bioenerg. 1984, 12, 485498.
103. G. Kokkinidis, N. Xonoglou, Bioelectrochem.
Bioenerg. 1985, 14, 375387.
104. N. Xonoglou, I. Moumtzis, G. Kokkinidis,
105. G. Kokkinidis, J.-M. Leger, C. Lamy, J. Electroanal. Chem. 1988, 242, 221242.
1985, 187, 151159.
1986, 201, 2332.
108. B. Beden, M. C. Morin, F. Hahn, C. Lamy,
109. H. Hitmi, E. M. Belgsir, J.-M. Leger et al.,
110. S.-G. Sun, D.-F. Yang, Z.-W. Tian, J. Electroanal. Chem. 1990, 289, 177187.
111. E. Pastor, S. Wasmus, T. Iwasita et al., J.
112. F. Hahn, B. Beden, F. Kadirgan et al., J.
113. S. Ernst, J. Heitbaum, C. H. Hamann, Ber.
114. Yu. B. Vassiliev, O. A. Khazova, N. N. Nikolaeva, J. Electroanal. Chem. 1985, 196,
105125, 127144.
115. T. T. Bae, E. Yeager, J. Electroanal. Chem.
1991, 309, 131145.
116. F. Largeaud, K. B. Kokoh, B. Beden et al., J.
117. M. Shibata, N. Furuya, M. Watanabe, J.

118. K. B. Kokoh, P. Parrot, E. M. Belgsir et al.,
Electrochim Acta 1993, 38, 13591365.
119. R. Wetzel, L. Muller, P. Gottlich, Z. Phys.
Chem. 1977, 258, 528532.
120. G. Kokkinidis, J. Electroanal. Chem. 1984,
172, 265279.
121. M. Sakamoto, K. Takamura, Bioelectrochem.
Bioenerg. 1983, 10, 251260.
122. G. Kokkinidis, D. Sazou, I. Moumtzis, J.
123. G. Kokkinidis, G. Papanastasiou, J. Electroanal. Chem. 1988, 257, 239255.
124. K. Takamura, M. Sakamoto, J. Electroanal.

Chem. 1980, 113, 273283.
125. M. J. Walls, J. M. Feliu, A. Aldaz et al., J.
126. G. Kokkinidis. P. Jannakoudakis, J. Electroanal. Chem. 1981, 130, 153162.
127. G. Kokkinidis, D. Sazou, J. Electroanal.
Chem. 1987, 237, 137144.
128. G. Kokkinidis, D. G. Zatkas, D. Sazou, J.
129. C. Mayer, K. Juttner, W. J. Lorenz, J. Appl.
Electrochem. 1979, 9, 161169.
417
418
4.4
Electrochemistry in Micelles and

Microemulsions
James F. Rusling
Departments of Chemistry and Pharmacology
University of Connecticut, Storrs, CT, USA
4.4.1
Introduction
Analytical and synthetic electrochemistry

is usually done in homogeneous aqueous solutions or organic solvents. However, there are advantages in some cases
in using microheterogeneous uids such
as micellar solutions and microemulsions [13]. These include solubilization
of organics and biomolecules, manipulation of electrode surface properties,
stabilization or destabilization of reactive
intermediates [4, 5], control of reaction
pathways, and enhancement of rates of
mediated reactions [6, 7]. Advantages also
accrue from the low cost [8] and low toxicity
of micellar solutions and microemulsions
compared to pure aprotic solvents often
used for organic electrochemistry. Electrochemical methods can also be used to
characterize micelles and microemulsions
via the measurement of diffusion coefcients of electroactive probes [4, 5, 9].
Micelles and microemulsions depend
on surfactants for their internal structure.
Surfactants (surface-active agents or detergents) are amphiphilic molecules with
charged or polar head groups attached to
one or more hydrocarbon tails (Fig. 1). Micelles are aggregates that form in water
when concentrations of surfactants exceed
a critical micelle concentration (CMC). Microemulsions are microheterogeneous uids made from oil, water, and surfactants,
and sometimes cosurfactants. They appear
homogeneous to the eye but contain dynamic heterogeneous nanostructures that
can be tuned for specic applications by

adjusting the relative amounts of the components [13].
The aim of this chapter is to describe electrochemistry in micelles and
microemulsions in fundamental and practical terms. A major focus is on the use of
these media to purposely inuence the
desired outcome of electrochemical reactions. The chapter also describes how
electrochemical methods can be used for
structural characterization of these uids.
In the next section (Sect. 4.4.2), we discuss
structures, properties, and dynamics of micelles and microemulsions. Subsequent
sections present relevant aspects of direct electrochemistry and electrochemical
catalysis in micelles and microemulsions.
4.4.2
Surfactant-stabilized Microheterogeneous
Fluids
Structures of Micelles and
Microemulsions
4.4.2.1
Surfactants and Micellar Structure

The formation of micellar aggregates from
ionic surfactants in water causes sharp
discontinuities in conductivity and surface tension. Water-soluble surfactants
form spherical or globular micelles at
concentrations near the CMC (Fig. 2).
Micelle formation is thermodynamically
favored in water for water-soluble surfactants such as cetyltrimethylammonium
bromide (CTAB), sodium dodecyl sulfate (SDS), and Brij 35, and is driven
by hydrophobic interactions between the
tails [10]. The charged or polar head groups
face the water phase and the hydrocarbon
tails reside in the interior of the micelle.
For ionic surfactants in water, the head
group region is only partly neutralized
by counterions, setting up an interfacial
4.4.2.1.1
4.4 Electrochemistry in Micelles and Microemulsions

CH3
+
N
Br
CH3
DDAB-didodecyldimethylammonium bromide
O
O
P
O
DHP-dihexadecylphosphate
CH3
CH3
Br
CH3
CTAB-cetyltrimethylammonium bromide
Na+ SO4
SDS-sodium dodecylsulfate
CH2
OOC
CH
OOC
CH2
O3POCH2CH2N+(CH3)3
PC-phosphatidylcholine (lecithin)
N
+
Br
Cetylpyridinium bromide
HO(CH2CH2O)23
Brij-35 polyoxyethylene(23)dodecyl ether
Fig. 1 Amphiphilic molecules (surfactants) used to make
micellar solutions and microemulsions.
potential on the order of 100 mV [3]. This

interfacial region consists of an electrical double layer (i.e. head groups and
counterions) called the Stern layer and
provides binding sites for solutes of the
opposite charge. If salt is added to a micellar solution, the surface potential is
partly neutralized and the micelle grows
larger. Micelles having a single narrow

size distribution are called monodisperse.
Hydrophobic cores of globular micelles
typically have diameters of 1 to 3 nm.
Micellar core volume (Vc ) can be estimated from the average number of surfactant molecules per micelle, that is, the
aggregation number (Nag , ca. 30200),
419
420

Water
Water
Water
Water
Globular micelle
Oil
Oil
Water
Oil
Rod-shaped micelle
Oil
Inverse micelle
Oversimplied representations of micellar structures.

Surfactants are represented with dark circles for head groups and
curved lines for hydrocarbon tails.
Fig. 2
via [10]
Vc = m (27.4 + 26.9n )A
(1)
for a single chain surfactant m = Nag for

a double chain surfactant m = 2Nag and
n is the number of carbon atoms in the
tail less one. Solubilization of nonpolar
molecules occurs in hydrophobic regions
of the micelles but near the Stern layer [3].
At intermediate concentrations considerably above the CMC, rod-shaped micelles
often form. At very high concentrations,
lamellar aggregates and other condensed
phases form. In organic solvents, some
surfactants form spherical inverse micelles
with the head groups facing small pools
of trapped water, with hydrocarbon tails
facing the bulk solvent phase (Fig. 2).
Structures of micelles depend on the
structures and concentrations of the surfactant molecules. The surfactant packing
parameter, v/a0 lc , can be used to predict

the shape of the aggregate formed [11]. To
a rst approximation, a0 is the area of
the head group, and the other terms are
found from
lc = 1.5 + 1.26nA
(2)
v = 27.4 + 26.9n A
(3)
where n is the number of carbons in

the tail of length lc , and n = n 1. For
a double chain molecule, n is multiplied by 2. Values of v/a0 lc < 1/3 suggest
spherical micelles, 1/3 > v/a0 lc > 1/2 predicts rod-shaped micelles, 1/2 > v/a0 lc >
1 suggests lamellar structures including bilayers, and v/ao lc > 1 implies that reverse
micelles will form in organic solvents.
The packing parameter is inuenced to
some extent by added salt, which decreases
average head group area, and by other experimental variables [12].

Water
Water
Oil
Oil
Water
Water
Oil
Water
o/w microemulsion
Oil
Oil
Oil
Bicontinuous microemulsion
Water
Oil
Oil
w/o microemulsion
Fig. 3
Oversimplied representations of microemulsion structures.
4.4.2.1.2 Microemulsion Structure Surfactant molecules residing at oilwater

interfaces lower the interfacial free energy
to nearly zero and drive the formation of
optically clear, thermodynamically stable
microemulsions [15]. Charged or polar
head groups of the surfactants (Fig. 1) face
the water phase and hydrocarbon tails extend into the oil phase. Microemulsions
can contain droplets of oil in a continuous water phase (oil-in-water or o/w
microemulsions) or droplets of water in
a continuous oil phase (water-in-oil or w/o
microemulsions). These droplets have diameters in the range 10 to 100 nm, larger
than most micelles but much smaller
than droplets in emulsions. As with micelles, the surface charge on the droplets
leads to interfacial potentials on the order

of 100 mV.
A third type of microemulsion does
not contain droplets but features continuous oil and water phases intertwined in
dynamic extended networks (Fig. 3). Conductivity is imparted to microemulsions
by ionic surfactants or salt in a continuous
water phase. Thus, o/w and bicontinuous
microemulsions can be made in conductive forms suitable for electrochemical
experiments.
The surfactant packing parameter v/a0 lc
can be used to predict what types of
nanostructures in microemulsions can
form using specic surfactants. Amphiphilic molecules with 0.5 < v/a0 lc < 1
tend to form three-component o/w and
421
422
bicontinuous microemulsions. Surfactants

with v/a0 lc < 0.5 usually require cosurfactants such as long-chain alcohols to form
microemulsions and those with v/a0 lc > 1
tend to form w/o microemulsions. These
rules provide guidelines, but the type of
microemulsion obtained also depends on
ratios of oil, water, cosurfactant, and surfactant [2].
Dynamics
A major inuence on the mass transport of electroactive solutes ensues when
4.4.2.2
they are bound to micelles, microemulsion

droplets, or interfacial surfactant layers
in bicontinuous microemulsions. Binding
can occur by coulombic or hydrophobic
interactions. An understanding of the inuence of solute binding on electrochemical experiments relies on the knowledge
that micelles and microemulsion nanostructures exist in mobile equilibria with
solutes and surfactant monomers (Fig. 4).
Ejection of a monomer or solute from globular aggregates occurs on the microsecond timescale, and recapture is diffusion
Solute
+
Solute
Fig. 4
Solute
Dynamic equilibria involving micelles, solutes and surfactant monomers.
controlled, that is, on the nanosecond

timescale. Recapture rates do not depend
strongly on solute structure, but exit rate
constants range from 103 to 107 s1 , and
decrease with the size and solubility of the
solute [3, 4].
Multiply charged ions have a strong
coulombic attraction to oppositely charged
micelles; for example, Ru(bpy)3 2+ [bpy =
2, 2 -bipyridine] binds strongly to SDS
micelles. Such solutes reside on the micelle surface, although the ligands can
contribute hydrophobic interactions to
the binding. Nonpolar solutes like aromatic hydrocarbons reside in hydrophobic regions close to the Stern layer of
the micelles. Spectroscopic studies show
that their microenvironments are more
polar in ionic micelles than in hydrocarbon oils but less polar than in
water [13].
Surfactant Aggregates on Solid
Surfaces
Adsorption of surfactants on charged
metal oxides is understood better than on
solid electrodes and instructive parallels
can be drawn [14]. For ionic surfactants
of opposite charge to the metal oxide
surface, the rst stage of adsorption
from very dilute aqueous solutions is
thought to result from neutralization of
the surface charge [1517]. Well below
the CMC, surfactants adsorb individually
with little intermolecular interaction. As
the concentration increases, adsorption
increases sharply and surface aggregates
begin to form. It was suggested that at
this point the surface is characterized
by patches of bilayers [15, 18]. In the
inner adsorbate layer, head groups face
the solid surface. The outer layer of the
surfactant has its charged head groups
facing the water phase. Once the high
charge density patches are covered with
4.4.2.3
bilayers, other regions of the surface begin

lling in.
Atomic force microscopy (AFM) [1921]
provides a more detailed molecular picture of adsorbates at concentrations above
the CMC. Adsorbed cationic surfactant C14 TAB (tetradecyltrimethylammonium bromide) was arranged in surface
micelles on silica and in long cylinders on mica. Double-chained DDAB was
organized in bilayers on mica. These
structures result from the balance of adsorbatesurface interactions and aggregate
curvature controlled by intermolecular interactions.
Ionic surfactants above the CMC form
full coverage aggregates on metal and
carbon electrodes [4, 22, 23]. Multilayers
may form at extreme potentials of the
opposite sign of the surfactant head group.
Developing a denitive molecular picture
of the aggregate structure based only
on electrochemical results is difcult.
Spectroscopy of the electrodeadsorbate
solution interface has helped elucidate
structural features [4, 14].
A few general features of the surfactant adsorbate structure can be extracted
from the information available. At potentials negative of the point of zero
charge (PZC) on hydrophilic metal and
carbon electrodes, cationic surfactants adsorb head down. Positive of the PZC,
anionic surfactants adsorb head down.
Nonionic surfactants may adsorb head
down on hydrophilic electrodes on either
side of the PZC. Adsorption of cationic
surfactants may be head down even at
potentials positive of the PZC. This may involve adsorbed anions such as chloride on
the electrode. Surface aggregate structures
above the CMC may include bilayers, surface micelles, or cylinders depending on
the nature of the surfactant, the electrode
surface, and the applied potential.
423
424
4.4.3
Electrochemistry in Micellar and

Microemulsion Systems
Mass Transport Diffusion
Mass transport rates of electroactive solutes can be affected greatly by micelles
and microemulsions. In cyclic voltammetry, for the reaction of a reversibly reduced
or oxidized solute (O or R),
4.4.3.1
O + ne = R
(4)
the peak current (Ip ) in a quiet solution

depends on the apparent diffusion coefcient D as given by the RandlesSevcik
equation [24]:
Ip = 0.4463nFAC
nF
RT
1/2
1/2 D 1/2
(5)
where n is the number of electrons
transferred per molecule, F is Faradays
constant, A is the electrode area, C* is the
solute concentration, R is the gas constant,
T is the temperature in Kelvin, and is
the scan rate. Equation (5) predicts a linear
plot of Ip versus for a reversible reduction
or oxidation. The slope of this plot can be
used to estimate D in a micellar solution
or microemulsion.
Linear sweep voltammetry at ultramicroelectrode disks of radius r < 10 m under
mass transport control, usually achieved
at scan rates <50 mV s1 , provides a limiting current IL that depends directly on
D [25]:
IL = 4nF D C r
(6)
A number of other electrochemical methods can also be employed to obtain D [4, 9].
D is inuenced by binding to micelles,
microemulsion droplets, or surfactant
monolayers at o/w interfaces, as well
as obstruction effects. D for watersoluble, unbound electroactive ions can

be interpreted in terms of obstruction by
oil droplets in o/w microemulsions [5, 9]
as in Eq. (7):
D = D0 (1 c )p+1
(7)
where c is the volume fraction of the oil

in the microemulsion, D0 is the diffusion
coefcient of the free ion in water, and p is
a tted coefcient. D values of unbound
water-soluble ions in microemulsions are
typically 40 to 80% of those in water.
Diffusion of electroactive species residing
in a single microphase of bicontinuous
microemulsions is somewhat obstructed
by the o/w network.
Solutes bound to micelles, droplets, or
interfacial regions can have D values
tenfold smaller than their D values in the
corresponding homogeneous solvent. For
aggregates with a single size distribution,
where the equilibrium between the bound
and the free solute is fast with respect
to the timescale of the electrochemical
experiment, D is given by a two-state
model [4, 5]:
D = fa D 0 + fb D 1
(8)
where fa is the mole fraction of the

unbound solute, fb is the fraction of
the bound solute, D0 is the diffusion
coefcient of the free solute, and D1
is that of the micelle or droplet in the
uid. Equation (8) represents the usual
case in micelles and microemulsions
when D is estimated by cyclic, rotating
disk, or ultramicroelectrode voltammetry
at relatively low scan rates [4].
D1 can be used to estimate the
hydrodynamic radius (rh ) of a micelle or microemulsion droplet via the
reaction:
StokesEinstein equation:
D1 =
kT
6rh
M + mX = MXm
(9)
where k is Boltzmanns constant, is

the viscosity, and T is the temperature
in Kelvin. D1 can be obtained by using
equilibrium expansions of Eq. (7) describing D versus total concentration of an
electroactive probe, CX [4, 5]. At typical
surfactant and probe concentrations, more
than one electroactive probe (X) can bind
to a micelle or droplet (M) in the overall
Km
(10)
[MXm ]
=
[M][X]m
(11)
Assuming [X] CX , that is, that the probe

is almost entirely bound, Eq. (8) can be
expressed as
D =
D0
1 + CM K m CXm1
CM K m CXm1 D1
1 + CM K m CXm1
(12)
10
107D
[cm 2 s1]
0.5
1.5
2.5
3.5
Fc
[mM]
Inuence of the concentration of ferrocene (Fc) on D measured by cyclic voltammetry
at a glassy carbon electrode in 0.15 M CTAB/0.1 M tetraethylammonium bromide. Points are
experimental: line is best t to Eq. (12) with n = 3 by nonlinear regression, giving
D1 = 3.3 107 cm2 s1 for the rod-shaped micelles. (Reproduced with permission from
Ref. [4], Copyright 1994 by Marcel Dekker.)
Fig. 5
425

12
106D
[cm2 s1]
426
0.00
0.50
1.00
1.50
B12
[mM]
Inuence of concentration of cobalt(II) corrin complex vitamin B12r
on D measured by linear sweep voltammetry at a 10-m radius carbon
ultramicroelectrode in w/o microemulsion made with 0.2 M aerosol AT,
4 M water, and isooctane. Points are experimental: line is best t to
Eq. (12) with n = 3 by nonlinear regression giving
D1 = 0.63 106 cm2 s1 for the water droplets. Using this value with
Eq. (9) gave a hydrodynamic radius of 7.5 nm. (Adapted with permission
from Ref. [27], Copyright by American Chemical Society.)
Fig. 6
Equation (12) holds for micelles and w/o

or o/w microemulsions with single size
distributions. It describes the dependence
of D on CM at constant CX and on CX
at constant CM . K is an apparent binding
constant for a single probe molecule.
Plots of D versus CX at constant CM
are useful for determining D1 because
the aggregate size is likely to remain
constant for the full set of CX . When
CM is increased, aggregate size and shape
may change [3, 4]. If K is too large,
no dependence of D versus CX will be
observed.
Figure 5 illustrates agreement of D
versus CX data with Eq. (12) for rod-shaped
micelles using micelle-soluble ferrocene

(Fc) as the electroactive probe. Fitting
Eq. (12) to the data is typically done with
D0 , D1 , and K as parameters. A series
of nonlinear regressions are done, each
with a different xed integer value m =
2, 3, . . . , q, until the minimum standard
deviation of the regression in the series is
found [26].
Figure 6 illustrates the use of Eq. (12)
for a w/o microemulsion. Ultramicroelectrodes were used to obtain D for the water
droplets using a water-soluble electroactive
probe. Results of the regression analysis
allow an estimate of the hydrodynamic
radius of the spherical droplet.
Inspection of Figs 5 and 6 shows that at

high probe concentration, limiting values
of D tend toward D1 . Thus, a reasonable
estimate of D1 for monodisperse systems
can be obtained by measuring D at
relatively large probe concentrations, for
example, 2 to 5 mM. For polydisperse
systems, additional equilibria must be
considered. A three-state model was used
to develop a version of Eq. (12) for two
micellar size distributions, and applied to
systems with coexisting globular and rodshaped micelles [4].
Formal Potentials
Binding electroactive solutes to micelles or
microemulsion droplets inuence formal

potentials (E o ) estimated by voltammetry. The pseudophase equilibrium approach

can predict E o in the microheterogeneous
media. Equilibria of micelles or droplets
with both forms of the redox couple
must be considered via a square scheme
(Scheme 1) for the reversible redox couple A/B.
The pseudophase equilibrium model considers solutes to be dissolved in continuous and discrete (micelles or droplets)
phases in a way similar to the partition
in conventional two-phase systems of oil
and water [3, 4]. The equilibrium distribution constants are
4.4.3.2
K0 =
[AW ]
[AM ]
and
Kr =
[BW ]
[BM ]
Subscript W refers to the water or

continuous phase, and subscript M refers
to the micellar or droplet phase. In general,
the reversible midpoint or formal potential

r ) is
in the microheterogeneous uid (EM
r
EM
= EW +
RT
K0 (1 + Kr )
ln
nF
Kr (1 + K0 )
D
RT
ln 0
2nF
Dr
where D s are measured diffusion coefcients in the microheterogeneous solu

tions, EW
is the formal potential of A/B
in the continuous phase, and the other
constants have their usual electrochemical
meanings. There are several limiting cases
allowing simplications:
Case 1. A and B exist mainly in the
bound form, and Kr , K0 1.
r
EM
= EW +
RT
K0
ln
nF
Kr
(14)
Examples include multiply charged metal

complexes ML3+ /ML2+ in SDS micelles [28].
Case 2. B is entirely bound but A is
not bound.
D
RT
RT
1 + Kr
ln
ln 0
nF
Kr
2nF
Dr
(15)
An example is ferrocene bound to
positively charged alkyltrimethylammonium micelles, where oxidation product ferricinium ion dissociates when
formed [29].
r
EM
= EW +
Electrode Kinetic Effects

Attempts were made several decades ago
to describe the inuence of adsorbed
surfactants on the apparent heterogeneous
4.4.3.3
AM + ne
K0
Scheme 1
(13)
M + AW + ne
BM
Kr
M + BM
427
428
electron transfer rate constant k using

Eq. (16) [4, 9]:
k = k0 (1 ) + k1
(16)
where k0 is the rate constant on a bare

electrode, k1 is the rate constant on a
fully coated electrode, and is fractional
surface coverage of surfactant. While many
systems deviate from Eq. (16) at large ,

a decreased k in micellar solutions and
microemulsions is often found.
For a reactant molecule or ion in
a micellar solution or microemulsion,
predictions of electron transfer kinetics
at electrodes need to consider [14]: (1) the
distance between the electrode and the
reactant, (2) the environment surrounding
the reactant at the time of electron
transfer, (3) structure and dynamics of
surfactant aggregates on the electrode, and
(4) dynamics of interactions of the reactant
with surfactant structures on the electrode
and with micelles. A molecular picture
of these events during electron transfer
is by no means clear, and quantitative
predictions are not possible at this time.
A qualitative view of the above factors is
given in the following paragraphs.
When small electroactive ions or molecules are bound to larger aggregates like
micelles or microemulsion droplets, the
reactant (probe) is transported to the electrode along with the larger, slower diffusing aggregate. Equation (12) describes the
inuence of concentration of surfactant
or reactant on electrochemically measured
diffusion coefcients. At [X] >2 mM, the
measured apparent diffusion coefcient
D approaches the diffusion coefcient of
the micelle D1 . This implies electrolysis
of one reactant X per micelle. This electrolyzed X could reside within MXn , or be
released by dissociation, as illustrated in
Eqs. (17 and 18) for an oxidation:
#
MXn
$
MXn1 + X
(17)
#
X $
X + e (at electrode) (18)
When MXn reaches the electrode, either

an electron is donated directly from one
of the bound Xs or dissociation (Eq. 18)
yields a free X that donates the electron.
Both coulombic and hydrophobic interactions of reactants with adsorbed
surfactant on electrodes are important
in determining electron transfer kinetics. Reactants in micellar solutions and
microemulsions can be preconcentrated
into adsorbed surfactant lms on electrodes [30], yielding mixed layers of reactants and nonelectroactive surfactants.
Coulombic effects in micellar solutions
may result in small kinetic enhancements
when ionic reactants interact with oppositely charged surfactant adsorbed on
electrodes. Partial inhibition of electron
transfer can occur by coulombic repulsion if the charge sign on the reactant
and adsorbed surfactant are the same. Hydrophobic molecules and ions may show
a small amount of preconcentration on
the electrode.
Studies of oxidations in micellar CTAB
and SDS media [31] suggested average
microenvironments for ferrocene during
electron transfer in micellar solutions that
are more polar than in apolar solvents
but less polar than in water. Similar
microenvironmental polarity is found for
hydrophobic solutes in micelles [13].
Studies of electroactive surfactants 1
to 4 (Fig. 7) in dilute or micellar solutions have provided insight into adsorbate
structure, interactions, and dynamics at
electrodes [14]. Ordered lms were formed
on glassy carbon electrodes from dilute
solutions of 3. Single-molecule residence
times on the electrode in micromolar
solutions were 4.5 s for Fc-C8, 14 s for FcC12, and 66 s for Fc-C16. Bilayers of 3 were
adsorbed on electrodes from microemulsions made with 3, oil, water, and alcohol.

Comparisons of electrochemical k values of surfactants 1 and 2 in microemulsions and micellar solutions helped establish qualitative pictures of dynamics
at relevant electrode-uid interfaces. Rate
constants for oxidations of ferrocene (Fc)
2-Fc (1) and 5-Fc (2) were similar in homogeneous DMF and DMSO on Pt and glassy
carbon electrodes [32, 33]. However, in
aqueous CTAB micelles, electron transfer
rates were in the order Fc > 2-Fc > 5-Fc,
with tenfold differences in successive values. This was attributed to 2-Fc and
5-Fc achieving head down orientations on
the electrode prior to electron transfer.
Adsorbed CTA+ on the electrode seems to
help order 1 and 2 on the electrode prior
to electron transfer.
In a bicontinuous CTAC microemul
sion, ferrocene had a k twice as large
as 1 and 2 [33], but values for 2-Fc and
5-Fc were similar. These relatively small
differences in electron transfer rates in
the microemulsion cannot be explained by
head downtail up orientation of 1 and
2 at the time of electron transfer as proposed for micellar solutions. The results
suggest an increased disorder and mobility in the electrodeuid interface in the
CTAC microemulsion compared to micellar CTAB solutions.
Alcohols are often used as cosurfactants in microemulsions, and insight has
been obtained from the electrochemistry
of ferrocene alcohols (4). Oxidations of
FcOHC10 , FcOHC14 , and FcOHC18 [34]
were nearly reversible and controlled by
diffusion in microemulsions of DDAB,
CTAC, or SDS. In micellar solutions,
electrode reactions were more complex
and reected strong adsorption of the
ferrocene alcohols onto the electrode.
CH3
CH3
N +
OOCFc
CH3
(1)
2-Fc
CH3
CH3
N+
OOCFc
CH3
(2)
5-Fc
CH3
CH3
CH2
(CH2)n1CH3
Fc-Cn, n = 8,12,16
(3)
Fc
CH3
HO
N
Fc
(CH2)n1CH3
(4)
FcOH-Cn, n = 10,14,18
Electroactive
ferrocene-containing surfactants.
Fig. 7
In microemulsions, the alcohols are distributed between the oilwater interfaces

and the oil phase of the microemulsions,
with increases in chain length favoring
binding at the oilwater interface. Results
of these and other studies [3537] suggest
that the presence of sufcient cosurfactant can improve interfacial uidity and
facilitate electron transfer at electrodes.
In general, then, surfactant aggregates
on electrode surfaces can control electron
transfer kinetics. Aggregates such as
bilayers, cylinders, or surface micelles
adsorb onto electrodes in solutions with
surfactant concentrations above the CMC.
429
430
On hydrophilic electrodes, head down

adsorption of surfactants seems to be
preferred.
The electroactive reactant may replace
adsorbed surfactant at a site on the electrode and approach the electrode closely.
In micellar CTAB, this may occur with
reactants 1, 2, and 3. However, adsorbed
reactant is unlikely to replace all of the
adsorbed nonelectroactive surfactant in
micellar solutions well above the CMC, and
mixed adsorbate layers form on the electrode. Alternatively, if surfactants remain
strongly adsorbed, hydrophobic reactants
could enter the surface lm and approach
the electrode to within roughly one head
group diameter prior to electron transfer.
Both possibilities can be inferred for specic experimental systems. Comparisons
of ferrocene, 1 and 2 suggested that underivatized ferrocene may approach the
electrode within the distance of one head

group. Small reactants typically have k
values in micellar solutions consistent
with this view [4]. Thus, depending on
the ability of the solute to compete with
nonelectroactive surfactant for sites on the
electrode, electron transfer can take place
to adsorbed reactant or to reactant that approaches the electrode to within a distance
of roughly one surfactant head group.
There are several molecular scenarios
for the delivery of micelle-bound reactants into adsorbed surfactant lms on
electrodes. One possibility is dissociation
(Eq. 17) followed by entry of the reactant
into the aggregate lm on the electrode,
orientation near the surface, and electron transfer. Making the analogy between
these latter processes and the adsorption
rates of 3, entry into the lms and orientation is expected to occur on a millisecond
timescale.
Alternatively, a micelle in solution could
join with aggregates on the electrode
surface, bringing the reactant close enough

to the electrode for electron transfer,
analogous to the joining of micelles and
submicellar aggregates on the millisecond
timescale. This process may be indistinguishable from Eqs. (17 and 18) by
voltammetry [14].
A picture of the microemulsionelectrode interface emerges through comparison with the micellar case. Although
the surfactant is adsorbed onto electrodes
from both types of uids, the adsorbate
in microemulsions is likely to be more
disordered and more uid. Results for ferrocene alcohols 4 suggest that there is no
long-chain alcohol cosurfactant strongly
adsorbed onto electrodes. However, some
alcohol may be distributed into a dynamic
surfactant layer on the electrode. Oils in
some microemulsions could also penetrate the surfactant layer on the electrode,
perhaps inducing disorder and increased
uidity. However, much more information is needed before a comprehensive
understanding of electron transfer processes at electrodes in microemulsion is
reached.
4.4.4
Pathway Control of Organic Reactions

Direct Electrolyses
Control of organic reaction pathways can
be achieved by using micellar media and
microemulsions. In the following sections,
selected cases are presented rather than a
comprehensive review.
Products of the reduction of nitrobenzene
in water depend on pH [38]. In neutral and
weakly acid media, phenylhydroxylamine
is always the product. In stronger acid
(pH< 4), phenylhydroxylamine is reduced
to aniline. In alkaline media, dimeric
products azoxybenzene (5), azobenzene
(6), or hydrazobenzene (7) are formed.
4.4.4.1

O
N+
H
N
N
H
Under conditions where phenylhydroxylamine is the product, a single

four-electron voltammetric peak is found
for nitrobenzene. However, at pH > 10
containing surface-active camphor or
gelatin, a single one-electron wave for reduction of nitrobenzene to its anion radical
is found, with a subsequent three-electron
wave at more negative potentials [4]. Similar behavior occurs in SDS micelles
and was attributed to hydrophobic-based
stabilization of the nitrobenzene anion
radical by the micelles [39]. Surfactant solutions allow electrochemical generation
of relatively stable radical anions of nitrobenzene and its derivatives, and have
been used for electron spin resonance
(ESR) studies.
Electrolysis of nitrobenzene at 1.5 V
versus Ag/AgNO3 in DDAB/hexane/water
microemulsions, even when acidied,
gave dimers 5 or 6 as products [40]. No
hydroxylamine or aniline was detected,
suggesting that the reaction takes place
at sites of low proton availability.
Direct reductions of organohalide pollutants have been done in solutions containing ionic and nonionic surfactants [41],
but often with low current efciencies.
One interesting approach involved the
use of an acid-labile nonionic surfactant,
1% oil, and water for the dechlorination
of hexachlorobenzene. This allowed facile
recovery of lesser-chlorinated products by

breaking the microemulsion with acid [42].
Micelles can also facilitate one-electron
dimerizations over two-electron reductions. An example is electrohydrodimerization of activated olens, including the
commercially important conversion of
acrylonitrile to adiponitrile in the Monsanto process, which employs an emulsion
of acrylonitrile, water, and short-chain
tetraalkylammonium salts [43]. High selectivity for the dimer adiponitrile can also
be achieved by using millimolar concentrations of nonionic surfactants such as
Triton X-100 [44]. Surfactant adsorbed on
the electrode is thought to create reaction
sites of low proton availability, shutting
down the competing two-electron reduction that requires protons. Other organic
dimerizations also benet from using micellar solutions [4].
A variety of electrochemical oxidations
and reductions have been examined in micellar solutions. For further information,
the reader is directed to comprehensive
reviews [4, 22, 30, 41, 45].
Only a few other direct electrochemical reductions have been studied in
microemulsions. Reductions of naphthalene and biphenyl resulted in selective
reduction of a single benzene ring in
the polyaromatic hydrocarbon [40], as in
8 to 12. Products 8 and 9, in addition to
biphenyl, were also found from catalytic
reduction of polychlorinated biphenyls in
microemulsions [46].
Electrochemical Polymerization in
Microemulsions
Microemulsions may have advantages over
conventional solvents for synthesizing
conducting polymers. W/o microemulsion
droplets can serve as microreactors to control polymer growth kinetics and particle
size [47].
4.4.4.2
431
432
10
11
12
W/o microemulsions of water, aniline,

light petroleum, and nonionic surfactant
Empilan NP-5 have been utilized for electrochemical polymerization of aniline to
polyaniline [48]. Improved homogeneity
and conductivity was achieved compared
with polyaniline grown in water. The heterogeneous nature of the microemulsion
directed the mode of polymer growth
and improved conductivity and structural features.
Poly(p-phenylene) was made by electrochemical polymerization in o/w microemulsions [49] of benzene, sulfuric
acid, and anionic, cationic, or neutral
surfactant. Benzene radical cation was stabilized by the anionic surfactant, resulting
in polymer with less cross-linking, smaller
particle size, and a relatively narrow size
distribution. With cationic surfactants,
the radical cation destabilized the water
droplets and led to a broader size distribution of polymer particles.
Stepwise polymer lm formation in
a microemulsion has been explored [50].
The rst step was the partial polymerization of a microemulsion containing
acrylamide, styrene, pentanol, water, and
SDS. Addition of potassium persulfate
and azo-bis-isobutyronitrile gave a viscous mixture that was used to coat the
electrode surface. Evaporation of pentanol left a highly porous surface onto
which pyrrole was electrochemically polymerized. This porous composite had a

large surface area for electrochemical reactions and allowed greater permeation by
reactants.
Nonconductive w/o microemulsions
were used for electrochemically induced
polymerization at a specially designed
solid polymer working electrode. Using a microemulsion of water, toluene,
and sodium bis(2-ethylhexyl) sulfosuccinate (AOT), acrylamide was polymerized
via persulfate reduction [51]. An applied
potential initiated polymerization, which
continued for several hours after the power
was off. This resulted in a latex suspension with comparable molecular weight
and particle size to those obtained by thermal or UV initiation. The degree of stirring
controlled the particle size (7130 nm).
The inuence of structure and concentration of cosurfactant amides in the above
w/o microemulsions was investigated by
ultramicroelectrode voltammetry [36]. Oxidation of amides was detected only above
a threshold concentration. The threshold
decreased as the chain length of amide increased because longer chain amides were
more extensively adsorbed at the oilwater
interface. Increasing the amount of acrylamide makes the interfacial region more
uid and increases the permeability to electroactive species so that electron transfer
can occur more easily [37, 52].
Mediated Reactions in Micellar
Solutions
Micellar rate enhancement of thermal and
light-initiated biomolecular reactions often occurs via preconcentration of the two
reactants in the micelles [3]. In electrochemical catalysis (Scheme 2), the analogous situation can occur for participants
in biomolecular reactions coadsorbed into
surfactant lms on electrodes. Interfacial
4.4.4.3

Scheme 2 Simple pathway for
electrochemical catalysis.
preconcentration in micellar solutions can

be exploited in electrochemical catalysis in
which a catalyst (mediator, P/Q) is added
to the system to shuttle electrons between
the electrode and a reactant that is otherwise difcult to oxidize or reduce [30].
Rate-determining steps (r.d.s.) in electrochemical catalysis are often biomolecular,
for example, the reaction between Q and
A in Scheme 2.
The observed rate of a chemical reaction
in micellar solutions is the sum of rates
in the continuous and micellar phases [3].
Consider a bimolecular chemical reaction
between Q and A in which the reactants
are totally bound to micellar aggregates.
The rate constant kobs , obtained on the
basis of amounts of Q and A in the total
system volume (Vt ), is given by
kobs =
kM [Q]M [A]M
[Q][A]
(19)
where kM is the rate constant in the

micellar phase, denoted by subscript M.
The concentration of the reactant in the
micellar phase is the total concentration
divided by the volume fraction of the
micelles, M , so that
P + e
Q+A
P+B
kM
2
M
(20)
kobs =
Equation (20) shows that kobs is enhanced

by compartmentalization of all the reactants into micellar volume Vt M . In electrochemical catalysis in micellar solutions,
large rate enhancements are observed
when the reaction occurs in surfactant
aggregates on the electrode surface [4,
30]. In this case, the compartmentalization volume where the biomolecular r.d.s.
occurs is that of the surfactant lm on
the electrode.
Catalytic reductions of organohalides
are examples of electrochemical catalysis that can give large rate enhancements
in micellar solutions [4, 30]. Anthracene
derivatives and metal complexes catalyze
these two-electron carbonhalogen cleavage reactions. The pathway is shown in
Scheme 3. Catalyst P is dissolved in the
micellar solution or immobilized on the
electrode. At applied potentials near E ,
P is reduced to Q, which reacts with aryl
halide ArX, regenerating P. This latter reaction is often the r.d.s. Subsequent steps
yield hydrocarbon ArH. Since catalyst P
is regenerated in the pathway, the voltammetric peak current of P is larger in the
P + e
Q
k1
ArX + Q
E (at electrode)
ArX + P
k2
ArX
Electrochemical
catalytic reduction of aryl
halides.
Scheme 3
Ar + H+ + e
Ar + P
ArH
433
434
presence of ArX. This catalytic current

can be used to estimate rate constants.
Mediator 9-phenylanthracene (9-PA)
was used to reduce 4-bromobiphenyl in
CTAB solutions on Hg electrodes [4, 30].
Voltammetry showed a large preconcentration of 9-PA in a thick lm of CTAB on the
Hg surface. Reduction of 9-PA at 2.2 V
versus SCE gave the anion radical that was
stabilized by the positively charged CTAB
lm. The observed k1 for reaction of 9-PA
anion radical with 4-bromobiphenyl was
107 M1 s1 in 0.1 M CTAB compared
to 300 M1 s1 in DMF, demonstrating
a large rate enhancement in the micellar solution. Electrolysis using this system
showed a clean, high-yield conversion of
4-bromobiphenyl to biphenyl.
Other catalysts with E values more
positive than 2 V showed smaller rate
enhancements because a thick CTAB lm
apparently does not form on Hg electrodes
at these potentials. Negatively charged
clays on electrode surfaces have been
used to form lms of cationic micelles
on electrodes for organohalide reductions
positive of 2 V. Details of these and other
mediated reactions in micellar solutions
are found in reviews [4, 30].
Mediated Reactions in
Microemulsions
4.4.4.4
Reductive Dehalogenations Microemulsions are usually more useful than

micelles for electrochemical synthetic applications because larger amounts of polar
and nonpolar reactants can be solubilized. Electrochemical catalysis has been
used in microemulsions for the electrolytic conversion of organohalide pollutants to hydrocarbons [53] using mediators
such as metal phthalocyanines and cobalt
complexes. Microemulsions were used
for the complete electrochemical catalytic
4.4.4.4.1
dechlorination of polychlorinated biphenyl

(PCB) mixtures [46, 5456] and the pesticide DDT [56, 57] and were also employed
for mediated electrolytic dechlorination of
PCBs adsorbed onto soils and clays [55].
The reactions were most successful with
a lead cathode in a simple, undivided,
two-electrode electrochemical reactor operated under constant-current using zinc
phthalocyanine as a mediator.
The catalytic conversion of trans-1,2dibromocyclohexane (DBCH) to cyclohexene with macrocyclic cobalt complexes [6,
58] was used as a probe to investigate the
kinetics of mediated electrochemical reactions in microemulsions. The P/Q catalyst
couples were macrocyclic cobalt complexes
CoII L/CoI L such as cobalt corrins, salen,
porphyrins, and phthalocyanines.
Br
Br
Co(II)L, 2e
+ 2 Br
(21)
In bicontinuous DDAB microemulsions,
CoII L mediators reside in the water
phase and DBCH in the oil phase, and
the reaction (Eq. 21) probably occurs at
the o/w interface. Cyclic voltammetry
(Fig. 8) shows the reversible CoII /CoI
reductionoxidation peaks of the mediator
and an increase in reduction current when
DBCH is added. Direct reduction of DBCH
occurs at a much more negative potential.
The catalytic current was used to estimate the apparent rate constant k1 for
the DBCH reaction with CoI L [58]. A

linear plot of log k1 versus E of the
mediator (Fig. 9) in DDAB microemulsions and organic solvents was obtained.
A similar linear plot was found for the
catalytic reduction of benzyl bromide in
microemulsions [59]. These plots suggest
that the bimolecular reactions in these
microemulsions are controlled mainly by

50
b
I
[A]
25
a
0
0.60
0.80
1.00
1.20
1.40
1.60
1.80
E vs SCE
[V]
Cyclic voltammograms at 0.1 V s1 on glassy carbon electrodes in a DDAB
microemulsion: a) 0.4 mM Co(salen) alone; b) 0.4 mM Co(salen) +1.5 mM DBCH; and c)
1.5 mM DBCH alone without catalyst. (Adapted with permission from Ref. [58], Copyright
by American Chemical Society.)
Fig. 8
the intrinsic activation free energy of the

reaction via E , and not by partition between oil and water phases. This situation
requires sufciently large o/w interfacial
area of the uid.
As with micelles, incorporation of reactants into an adsorbed surfactant layer on
an electrode [6, 7] can lead to high reactant
concentrations in a restricted reaction volume and enhanced rates of bimolecular reactions. Conversely, when reactants in a bimolecular r.d.s. reside separately in oil and
water phases in microemulsions with insufcient interfacial area or slow partition
dynamics, the reaction rate may be slower
than in a homogeneous solution [27].
The reduction of DBCH was also effected by using carbon electrodes coated
with catalytic lms made by covalently linking poly-L-lysine (PLL) onto oxidized carbon electrodes, then attaching a reversible
cobalt corrin catalyst [60]. Covalent linkage
was necessary for stability of the catalytic
lms in microemulsions. Conversion of
DBCH to cyclohexene in microemulsions

was achieved with turnover numbers
much higher than with dissolved catalysts. The catalytic properties of these
PLLCo lms could be controlled by microemulsion composition [61], and anionic
micelles formed within the cationic lms
in SDS microemulsions.
Turnover rates for the reduction of
DBCH to cyclohexene mediated by the
PLLCo lm in bicontinuous SDS microemulsions were controlled by the difference between the reduction potential

of the reactant and E of the catalyst in
the lm, similar to dissolved cobalt complex catalysts. High conductivity and low
viscosity of the bulk microemulsion also
facilitated fast catalyst turnover.
4.4.4.4.2 Bond Forming Reactions Mediated electrochemical synthesis in microemulsions can be used to construct
bonds. Carboncarbon linkages between
an alkyl halide RX and an activated olen
435
6
Co(salen)
B12
CoOEP
log k1
[M1 s1]
436
4
CoTPP
2
CoPcTS
1.2
1.0
0.80
0.60
0.40
E ' vs SCE
[V]
Inuence of catalyst formal potential (E ) on log k1 for reaction of CoI L with DBCH
in a bicontinuous microemulsion (%) of DDAB/water/dodecane (21/39/40) and in
dimethylformamide () for dissolved catalysts vitamin B12, Co(salen), cobalt
phthalocyaninetetrasulfonate (CoPCTS), cobalt tetraphenylporphyrin (CoTPP), and cobalt
octaethylporphyrin (CoOEP). Points from reactions in the microemulsion (%) represent
apparent k1 values for 0.4, 0.5, 1.0, and 2.0 mM catalyst in order of decreasing log k1 .
(Adapted with permission from Ref. [58], Copyright by American Chemical Society.)
Fig. 9
CoIIL + e
RX + CoIL
CoIL (at electrode)

k1
RCoIIIL + X
RCoIIIL + (H+, e or hn) + CH2 = CHZ

Scheme 4
R CH2CH2Z + CoIIL
Mediated electrochemical bond formation.
are made by using CoII L complexes as

mediators as in Scheme 4.
Here, X represents a halogen atom
and Z is an electron-withdrawing group.
Electrochemically reduced mediator CoI L
reacts with alkyl halide RX to yield
intermediate RCoIII L, which is cleaved
at a more negative potential or by visible
light. The resulting alkyl anion or radical

is trapped by an activated olen to form a
carboncarbon bond.
Several examples of mediated bond
formation are given below. Conjugated
additions of alkyl iodides to 2-cyclohexen1-one gave 3-alkyl cyclohexanones 13
(Scheme 5) [62] in 70 to 80% yields in

Scheme 5
Co(I)L + RI
O
RCo(III)L + I
O
hn
RCo(III)L +
+ Co(II)L
R
13
microemulsions using 0.85 V + light.

Poorer yields were obtained at 1.45 V in
the dark because of competing reductions.
Cyclization of 4-bromobutyl-2-cyclohexen-1-one (14) to 1-decalone (15) in microemulsions was also mediated by vitamin B12 (Scheme 6) [62, 63]. Cis- and
trans-1-decalone were obtained in 90%
yields with either light or electrochemical
cleavage of the alkylcobalt bond in microemulsions and in DMF. Microemulsions gave a stereoselective ratio of 93 : 7
trans:cis, but poor stereoselectivity was obtained in homogeneous DMF. Preferential
formation of thermodynamically favored
trans-1-decalone involved equilibration of
isomers via ketoenol tautomerism catalyzed by hydroxide ions formed by the
coelectrolysis of water.
Catalysis using vitamin B12 at 1.5 V

versus SCE at carbon electrodes in
CTAB and SDS bicontinuous microemulsions converted 2-(3-bromopropyl)-2-cyclohexen-1-one (16) to the 5-endo-trig cyclization product 4-hydrindanone (17) in
62 to 70% yields (Scheme 7) [64]. Similar
conditions gave only 7 to 19% of 17
in DMF, MeOH, or MeOH/water. The
microemulsions seem to provide reaction
sites of low proton availability to inhibit
protonation of a key carbanion intermediate whose protonation would prevent
cyclization.
Finally, phase transfer catalysis
was achieved in DDAB microemulsions. Dichlorocarbene was generated at a carbon cathode from
dichlorodibromomethane used as the
O
O
Br
Co(III)L
+ Co(I)L
+ Br
14
O
O
Co(III)L
hn
+ Co(II)L
15
Scheme 6
O
Br
Co(II)L
1.5 V
Scheme 7
16
17
437
438
oil constituent of a microemulsion [65].

Tetrabutylammonium bromide helped to
catalyze this reaction in the oil phase. The
carbene reacted with cyclohexene to generate 7,7-dichloro-bicycloheptane. The best
current efciency was 80%.
4.4.5
Summary
This chapter describes the electrochemistry of small reactants dissolved in micellar solutions and microemulsions. A
major inuence of these microheterogeneous uids on reversible reactants is
slowing down mass transport. These phenomena enable electrochemical probes to
be used to characterize aggregate mass
transport and size in the uids. Tuning the
compositions of micelles and microemulsions can control pathways and kinetics of
direct organic reactions, polymerizations,
and mediated electrochemical reactions.
Acknowledgments
The authors work on micelles and microemulsions described in this chapter was
initially supported by US PHS Grant No.
ES03154 from the National Institute of Environmental Health Sciences, NIH. More
recent synthetic aspects were supported by
Grants nos. CTS-9306961, CTS-9632391,
and CTS-9982854 from NSF. The contents
of this chapter are solely the responsibility of the author and do not necessarily
represent ofcial views of NIH or NSF.
References
1. P. L. Luisi, L. Magid, CRC Crit. Rev. Biochem.
1987, 20, 409474.
2. M. Bourrel, R. S. Schechter, Microemulsions
and Related Systems, Marcel Dekker, New
York, 1988.
3. J. H. Fendler, Membrane Mimetic Chemistry,

John Wiley & Sons, New York, 1982.
4. J. F. Rusling in Electroanalytical Chemistry
(Eds.: A. J. Bard), Marcel Dekker, New York,
1994, pp. 188, Vol. 19.
5. J. F. Rusling in Modern Aspects of Electrochemistry (Eds.: B. E. Conway, J. OM. Bockris),
Plenum Press, New York, 1994, pp. 49104,
No. 26.
6. J. F. Rusling, D.-L. Zhou, J. Electroanal.
Chem. 1997, 439, 8996.
7. J. F. Rusling in Reactions and Synthesis in
Surfactant Systems (Eds.: J. Texter), Marcel
Dekker, New York, 2001; in press.
8. S. Friberg, Adv. Colloid Interface Sci. 1990, 32,
167182.
9. R. A. Mackay, Colloids Surf. 1994, 82, 123.
10. C. Tanford, The Hydrophobic Effect, 2nd ed.,
11. J. Israelachvili, Intermolecular and Surface
Forces, 2nd ed., Academic Press, San Diego,
Calif., 1992.
12. D. F. Evans, D. J. Mitchell, B. W. Ninham, J.
Phys. Chem. 1986, 90, 28172825.
13. K. Kalyanasundarum, Photochemistry in Microheterogeneous Systems, Academic Press,
New York, 1987.
14. J. F. Rusling, Colloids Surf. 1997, 123124,
8188.
15. J. H. Harwell, J. C. Hoskins, R. S. Schecter
et al., Langmuir 1985, 1, 251.
16. M. A. Yeskie, J. H. Harwell, J. Phys. Chem.
1988, 92, 2346.
17. P. Chandar, P. Somasundaran, N. Turro, J.
Coll. Interface Sci. 1987, 117, 31.
18. J. F. Scamehorn, R. S. Schecter, W. H. Wade,
J. Colloid Interface Sci. 1982, 85, 463.
19. S. Manne, H. E. Gaub, Science 1995, 270,
1480, 1481.
20. S. Manne, J. P. Cleveland, H. E. Gaub et al.,
Langmuir 1994, 10, 4409.
21. E. J. Wanless, W. A. Ducker, J. Phys. Chem.
1996, 100, 3207.
22. T. C. Franklin, S. Mathew in Surfactants
in Solution (Eds.: K. L. Mittall), Plenum
Publishing, New York, 1989, pp. 267286,
Vol. 10.
23. N. Shinozuka, S. Hayano in Solution Chemistry of Surfactants (Eds.: K. L. Mitall), Plenum
Publishing, New York, 1979, pp. 599623,
Vol. 2.
Methods, John Wiley & Sons, New York, 1980.

25. M. Fleishmann, S. Pons, D. R. Rolison, P. P.
Schmidt, Ultramicroelectrodes, Datatech,
North Carolina, 1987.
26. J. F. Rusling, T. F. Kumosinski, Nonlinear
Computer Modeling of Chemical and Biochemical Data, Academic Press, San Diego,
1996.
27. A. Owlia, Z. Wang, J. F. Rusling, J. Am.
Chem. Soc. 1989, 111, 50915098.
28. Y. Ohsawa, Y. Shimazaki, S. Aoyagui, J. Electroanal. Chem. 1980, 114, 235.
29. Y. Ohsawa, S. Aoyagui, J. Electroanal. Chem.
1982, 136, 353.
30. J. F. Rusling, Acc. Chem. Res. 1991, 24,
7581.
31. A. P. Abbott, C. L. Miaw, J. F. Rusling, J.
32. A. P. Abbott, G. Gounili, J. M. Bobbitt et al.,
J. Phys. Chem. 96, 11 09111 095.
33. G. Gounili, J. M. Bobbitt, J. F. Rusling, Langmuir 1995, 11, 28002805.
34. X. Zu, J. F. Rusling, Langmuir 1997, 13,
36933699.
35. E. Garcia, S. Song, L. E. Oppenheimer et al.,
Langmuir 1993, 9, 2782.
36. E. Garcia, J. Texter, J. Colloid Interface Sci.
1994, 162, 262.
37. B. Antalek, A. J. Williams, E. Garcia et al.,
Langmuir 1994, 10, 4459.
38. A. J. Fry, Synthetic Organic Electrochemistry,
2nd ed., Wiley, New York, 1989.
39. G. L. McIntire, D. M. Chiappardi, R. L. Casselberry et al., J. Phys. Chem. 1982, 86,
2632.
40. H. Carrero, J. Gao, J. F. Rusling et al., Electrochem. Acta 1999, 45, 503512.
41. N. J. Bunce, S. G. Merica, J. Lipkowski,
Chemosphere 1997, 35, 27192726.
42. S. G. Merica, C. E. Banceu, W. Jedral et al.,
Environ. Sci. Technol. 1998, 32, 15091514.
43. D. Pletcher, F. C. Walsh, Industrial Electrochemistry, 2nd ed., Blackie Academic, London, 1993.
44. M. R. Moncelli, F. Pergola, G. Aloisi et al., J.
45. R. A. Mackay, J. Texter, (Eds.), Electrochemistry in Colloids and Dispersions, VCH Publishers, New York, 1992.
46. S. Zhang, J. F. Rusling, Environ. Sci. Technol.

1993, 27, 13751380.
47. J. F. Rusling, C. J. Campbell in Encyclopedia of Surface and Colloid Science (Ed.:
A. Hubbard), Marcel Dekker, New York,
2002, 41434161.
48. H. S. O. Chan, L. M. Gan, C. H. Chow et al.,
J. Mater. Chem. 1993, 3, 11091115.
49. A. Manna, K. Bandyopadhyay, K. Vijayamohanan et al., Langmuir 1998, 14, 8490.
50. D. A. Kaplin, S. Qutubuddin, Synth. Met.
1994, 63, 187194.
51. E. Garcia, L. E. Oppenheimer, J. Texter in
Electrochemistry of Colloids and Dispersions
(Eds.: R. A. Mackay, J. Texter), VCH Publishers, New York, 1992.
52. E. Garcia, S. Song, L. E. Oppenheimer et al.,
Colloids Surf., A 1995, 94, 131136.
53. J. P. Zelina, J. F. Rusling in Environmental
Analysis and Remediation (Eds.: R. A. Meyers), John Wiley & Sons, New York, 1998,
pp. 15671583.
54. E. Couture, J. F. Rusling, S. Zhang, Trans.
Inst. Chem. Eng.(U.K.) 1992, 70:B, 153157.
55. S. Zhang, J. F. Rusling, Environ. Sci. Technol.
1995, 29, 11951199.
56. J. F. Rusling, S. Schweizer, S. Zhang et al.,
Colloids Surf. 1994, 88, 4149.
57. S. Schweizer, Q. Huang, J. F. Rusling,
Chemosphere 1994, 28, 961970.
58. D.-L. Zhou, J. Gao, J. F. Rusling, J. Am.
Chem. Soc. 1995, 117, 11271134.
59. D.-L. Zhou, H. Carrero, J. F. Rusling, Langmuir 1996, 12, 30673074.
60. D.-L. Zhou, C. K. Njue, J. F. Rusling, J. Am.
Chem. Soc. 1999, 121, 29092914.
61. C. K. Njue, J. F. Rusling, J. Am. Chem. Soc.
2000, 122, 64596463.
62. J. Gao, J. F. Rusling, D.-L. Zhou, J. Org.
Chem. 1996, 61, 59725977.
63. J. Gao, C. K. Njue, J. K. N. Mbindyo et al., J.
Electroanal. Chem. 1999, 464/1, 3138.
64. J. Gao, J. F. Rusling, J. Org. Chem. 1998, 63,
218, 219.
65. C. J. Campbell, J. F. Rusling, Langmuir 1999,
15, 7416, 7417.
439
443
5.1
Electrocatalysis
Enrique Herrero, Juan M. Feliu, Antonio Aldaz
Universidad de Alicante, Alicante, Spain
5.1.1
Introduction: Concept of Electrocatalysis

and Electrocatalyst
The concept of a catalyst is quite clear in

chemistry. It was rst dened by Ostwald
in the nineteenth century as a substance
that only modies the velocity of a chemical reaction without suffering any chemical
change in the process. This denition is
based on a comparison between the rates
of the reaction in the presence and in the
absence of the catalyst. Two different types
of catalysis can be dened: homogeneous
catalysis, in which the catalyst and all the
species involved in the reaction are in the
same phase and heterogeneous catalysis,
when the catalyst constitutes a different
phase than that containing the reaction
species and the reaction takes place on the
surface of the catalyst. An inaccurate extrapolation of the denition of the catalyst
to electrocatalyst would indicate that all the
electrode materials are electrocatalyst since
the electrode reactions are heterogeneous
reactions in which an inert material (the
electrode) is always present and normally
does not suffer any chemical change in the
process. However, the right denition of

an electrocatalyst requires the comparison
of rates in the presence and the absence of
the electrode material. The problem arises
just in the possibility of performing this
comparison since, in most cases, there is
no homogeneous electrochemical reaction
with which to compare.
Although the denition of catalyst cannot be directly extrapolated for that corresponding to electrocatalyst, the concepts
underlying the denition can be effectively
used. In a catalyzed reaction, there is a specic interaction between the catalyst and
some species involved in the chemical reaction. The specic interaction is then the
key point in the catalysis denition and,
thus, the electrocatalyst can be dened as
an electrode material that interacts specifically with some species involved in the
reaction and remains unaltered after the
reaction. This denition allows a good distinction between electrocatalyzed and nonelectrocatalyzed reactions: the existence of
specic surface interactions is characteristic of an electrocatalyzed reaction.
It is implicit in the denition of the
electrocatalyst that this specic alters the
energetics of the reaction path, that is, it
changes the activation energy, and thus
the velocity of reaction. The lack of experimental or calculated rates for the reaction
in absence of an electrocatalyst is then
444
5 Kinetics and Mechanism of Selected Electrochemical Processes
not necessary to dene an electrocatalytic

reaction. Moreover, if several electrode materials are used and the reaction rates for
a given reaction depends on the electrode
material (and/or surface structure), it can
be said that these electrodes are acting as
a true electrocatalyst with different efciency. These different reaction rates can
only be justied if there is a specic interaction between the electrode and the species.
On the other hand, if the reaction is insensitive to the electrode material, these
electrodes are not interacting with the reaction species and are not electrocatalysts.
When comparing the catalytic activity of
several electrodes, the comparisons have
to be made either at constant overpotential
and measuring the current density (value
that is proportional to the reaction rate)
or at constant current density and measuring the overpotential. Of course, the
nal goal is always higher rates at lower
overpotentials.
As a rst step, it is important to
dene which reactions are susceptible to
be catalyzed. In principle, the reaction
rates of any process can be increased.
In a heterogeneous process, such as
electrode reactions, the diffusion of the
active species to the electrode may be
the rate-determining step of the whole
process. In that case, any improvement
in the rate of the electron-transfer step
would not produce any change in the
overall rate of the process, since the masstransfer process is still the limiting step.
The electrode reaction will behave then as
a reversible diffusion-controlled reaction.
Whenever the reaction in the actual
experimental conditions is not diffusion
controlled, it may be interesting to nd a
better electrocatalyst for it. The criteria for
dening a reaction as diffusion controlled
depends on the technique employed for the
study. According to the technique, several
parameters can be measured. For instance,

when using a stationary technique in
which the diffusion layer is constant, the
limiting diffusion current for any species
can be written as
jlim,i = ki ci
(1)
where ci is the bulk concentration of i

and ki is the mass-transfer coefcient,
which is a function of the diffusion
coefcient and can be calculated for
the different electrochemical techniques.
For a general reaction Ox + ne = R,
the relationship between the stationary
current and the potential is
E=E +
+
RT
kR
ln
nF
kOx
(j jlim,Ox )
RT
ln
nF
(jlim,R j )
(2)
Whenever a process studied with a stationary technique follows Eq. (2), the process
is mass controlled and improvement in the
reaction rates will not increase the overall
reaction rate. If not, it may be interesting
to nd a better electrocatalyst.
Improvement of the reaction rates is
not the only attribute sought in an
electrocatalyst. Often, a higher selectivity
versus any competitive reaction that may
take place is also required. For that reason,
catalysts given higher reaction rates but
lower material yields are normally not
used. Thus, a good electrocatalyst is the
one that gives high current densities at
low overpotential and with high selectivity
for the desired product.
When studying the electrocatalytic behavior of a given material, several properties may inuence its overall performance:
the electronic properties, the surface structure, and the different types of sites present
on it, the particle size, surface composition, and so on. The variety of these
5.1 Electrocatalysis
properties makes it difcult to understand

what the role of the electrocatalyst in a
reaction is. This way any interpretation
may be proposed. In this short chapter,
we will focus on the most fundamentals
aspects of the electrocatalysis, that is, how
the different properties of a given material relate to its electrocatalytic behavior.
In order to understand the fundamentals of electrocatalysis, it is important to
limit and control the number of possible
variables for a given experiment. Singlecrystal metal electrodes are normally used
in these studies, since they possess a very
well-dened structure and allow to establish a correlation between surface structure
and its catalytic activity. For that reason,
we will emphasize the results obtained on
well-dened metal electrode surfaces. Despite the use of single-crystal electrodes in
the last decades, the number of studies is
rather limited and only a few systems have
been investigated systematically. However,
it is worth noting that the results shown
here will have to be later applied to the
real surfaces employed in more applied
processes. Several general articles about
electrocatalysis can be found in the literature [18].
5.1.2
Electronic Effects in Electrocatalysis

5.1.2.1 Specic Interaction and Transfer
Rates
The specic interaction between the electrode material and some of reacting species
has to be dened formally by the overlap of
the electronic states of the reactant and the
electrode material at the transition state.
A weak overlap electron-transfer reaction
is dened as that in which the energy of
the transition state is not affected by the
proximity of the reacting species to the
electrode. Conversely, in strong overlap
reactions the energy of the transition state

is affected greatly by the proximity of the
electrode and the reacting species. A strong
overlap takes place when specic interaction between the surface and some of the
reactant species exists.
These situations are depicted in Fig. 1
for a multistep reaction Ox + ne = R.
In this gure, A denotes the transition
state and P and S, respectively, denote the
precursor and successor states that precede
and follow the transition state. The full line
represents the energetics of the process for
a large distance between reactive species
and electrode. If the reacting species come
closer to the electrode surface and these
species have attractive interactions with the
electrode surface, the dashed and dotted
lines are obtained, where the dotted line
represents a stronger overlap than the
dashed line. In this simplest model, the
transition state corresponds to the same
reaction coordinate.
These curves can be related to the innerand outer-sphere reactions pathways. The
full line represents the outer-sphere reaction, whereas the inner-sphere reactions
are depicted by the dashed and dotted lines.
As can be seen, outer-sphere reactions
are always weak overlapping reactions,
that is, reactions in which the electrode
is not acting as an electrocatalyst. Innersphere reactions can be either strong or
weak overlap reactions, depending on the
properties of the reactant species and electrode material. From a chemical point of
view, a chemical bond between the reactant species and the electrode is formed
in the strong overlap cases, that is, the
reactant species are chemisorbed on the
electrode surface.
From a practical point of view, it is important to distinguish between electrocatalyzed and nonelectrocatalyzed reactions.
In addition to the criteria given above, in
445
A
1
Energy
446
2
Ox
3
P
G
Reaction coordinate
Schematic free energy prole scheme for an electron
electroreduction. O, oxidized species; R, reduced species; A, transition
state; and P and S, the precursor and successor states, respectively, that
) for a large distance
precede and follow the transition state. (
between the reactive species and the electrode. (-----) Energetics of the
process in which the reacting species come closer to the electrode surface
and these species have attractive interactions with the electrode surface.
( . . . . ) Same as a dashed line but with a stronger overlap.
Fig. 1
situ infrared and Raman spectroscopies

allow the detection of adsorbed species
on the electrochemical interface. These
species present vibrational frequencies
that depend on the electrode material and
that are different than those found for
the species in solution, a clear indication
that the species are interacting with the
electrode surface. In other cases, the detection of intermediate adsorbed species
is not possible and indirect methods have
to be used to determine if the electrode
is acting as an electrocatalyst. Normally, if
the reaction rates depend on the electrode
material and its surface structure, it can

be said that the electrode is acting as
an electrocatalyst. However, different reaction rates on different electrode materials
may be the consequence of changes in
the specic reactant solvation at the interface or electrostatic double-layer effects,
and not a consequence of a strong overlap
transfer [9].
In the simplest cases, it is possible
to compare rates for homogeneous and
heterogeneous electron-transfer reactions.
According to the Marcus theory for
homogeneous outer-sphere reactions [10],
the reaction rate is given by

k1 = A1 exp
4RT
(3)
where is the reorganization factor and

A1 , the preexponential factor [10]. Analogously, the reaction rate for heterogeneous
reactions is

k2 = A2 exp
(4)
2RT
Dividing Eqs. (3 and 4) leads to
k1
=
A1
k2
A2
1/2
(5)
When the reaction rates for the homogeneous and heterogeneous reaction rates
fulll Eq. (3), it is clear that the reaction
is taking place through an outer-sphere
mechanism, and the electrode material is
not acting as a true electrocatalyst.
Adsorption and Electrocatalysis
As stated previously, adsorption of some
species involved in the electrochemical reaction is the key parameter to consider
an electrode material as an electrocatalyst.
Therefore, it is important to relate the activity of such a material with its adsorption
properties. Let us consider rst the simplest reaction A + e B, in which A is
the only species adsorbed on the catalyst,
and try to relate the reaction rate to the
adsorption properties of the surface S. A
simple reaction mechanism is
5.1.2.2
S + A S A
1
2
S A + e S + B
where S stands for an
site. Assuming that the
(6)
adsorption
adsorption
step (Reaction (1)) takes place through

Langmuir kinetics and k1 , k1 > k2 and
using the steady state approximation, the
reaction rate can be written as
=
k2 K1 CA
1 + K1 CA
(7)
where K1 is the equilibrium constant for

Reaction (1), and k2 the rate constant for
step 2. Now, it is desirable to express both
the constants as a function of the Gibbs
energy of adsorption (G1 ). K1 can be
readily written as

G1
K1 = exp
(8)
RT
k2 can be expressed as a function of the
activation energy (EA2 ) and the electrode
potential, E:

EA2
k2 = k2 exp
RT

2 F (E E )
exp
(9)
RT
where is the symmetry factor for
the electron-transfer step and E is the
standard potential for the overall reaction.
For a given series of homologous reactions
(in this case, the reaction S A + e B
using different surfaces), the activation
energy can be written as a function of
the Gibbs energy change of the reaction
using the Brnsted relationship [11]:
EA = EA G
(10)
where is the transfer coefcient (generally close to 0.5) and EA is the intrinsic
activation energy for the surface in which
the Gibbs energy of the reaction of the
adsorbed species is zero. Substituting
Eq. (10) into Eq. (9), the following equation
can be obtained:
447
448

EA2
k2 = k2 exp
RT

2 F (E E )
exp
RT

E + 2 G2
= k2 exp A2
RT

2 F (E E )
exp
(11)
RT
The Gibbs energy of the second step (G2 )
of the overall process is related to the
overall Gibbs energy (GT ) and the Gibbs
energy of adsorption (G1 ) according to
G2 = GT G1
(12)
where G1 depends on the electrode

material. Substituting Eq. (12) into Eq. (11)

E + 2 GT 2 G1
k2 = k2 exp A2
RT

2 F (E E )
exp
(13)
RT
and grouping all the constant terms for the
series of homologous reactions (GT and
) into k2a , the following expression can

EA2
be obtained:

2 G1
k2 = k2a exp
RT

2 F (E E )
(14)
exp
RT
Substituting Eqs. (8 and 14) into Eq. (7)
yields

(1 2 )G1
k2a exp
RT

=
G1
CA
1 + exp
RT

F (E E )
exp
CA (15)
RT
As can be seen, the reaction rate

is a function of the Gibbs energy of
adsorption, the electrode potential, and
the concentration of A. In fact, the reaction
rate can be divided into two separate terms:
one containing the dependence of G1
and CA and the other one that depends
on the electrode potential. If the catalytic
activity of several surfaces is compared at
constant potential, the reaction rate will
depend only on G1 and CA . In Fig. 2,
the term

(1 2 )G1
a
k2 exp
CA
RT
=

k2 =
(16)
G1
CA
1 + exp
RT
has been plotted versus the Gibbs energy
=
of adsorption. It follows that k2 , and obviously the reaction rate, increases with
the Gibbs energy of adsorption, reaches
a maximum, and then decreases, giving
a curve known as the volcano curve. This
kind of curve is very common in catalyzed
processes, not only in the simple process
that has been considered here, since most
reaction kinetics can be described by mathematical expressions similar to Eq. (14),
and therefore have the same qualitative
behavior. This behavior demonstrates that
adsorption increases reaction rates provided that its energy is not too high. In such
case, the adsorbed intermediate is too stable to react on to the products. This idea is
called, in heterogeneous catalysis, the Principle of Sabatier [12]. The comparisons in
electrochemistry are performed using the
exchange current density (j0 ) instead of the
reaction rate. It can be easily demonstrated
that j0 follows the same dependence on
G1 that the reaction rate, and therefore, volcano curves are obtained. In fact,
such curves have been obtained for several
electrochemical processes, such as hydrogen evolution [13], ethylene oxidation [14],
log (k#2)
G 1
Fig. 2
Generalized representation of k2 versus G1 .
formic acid oxidation [15, 16], and oxygen

reduction [17]. In all these cases, the reactions proceed through mechanisms much
more complicated than those studied here
and for instance, the nature of the relevant
adsorbed species for hydrogen evolution is
unclear [18]. An additional complication of
the electrochemical processes is the possible dependence of G1 with the applied
potential. Even in the simplest case, this
dependence arises either from the specic
interaction of the reacting species with the
electrode material, or from the specic interaction of any competing species (water,
anions) present in solution. In fact, the
concept of the adsorption isotherm of a
molecule in solution is more complicated
than that in the gas phase and has to take
into account the presence of other species.
Additional comments on this problem will
be made in the next paragraph.
In other processes that involve two

adsorbed species that react (for instance,
LangmuirHinselwood mechanisms), the
dependence of the reaction rate with the
adsorption energy is more complicated
since two adsorption energies are involved.
For this case, the ideal electrocatalyst is
the one that has intermediate adsorption
energy for both adsorbed species. If the
chemistry of the two adsorbed species is
different, which is normally the case, it
is difcult to nd a surface that has the
right combination of adsorption energies
for both species.
CO oxidation (among other oxidation
reactions of interest in electrocatalysis)
belongs to this type of reactions. It has
been normally proposed that adsorbed CO
reacts through a LangmuirHinselwood
mechanism with adsorbed OH from
water dissociation. Therefore, this reaction
449

500
500
400
400
300
300
200
200
100
100
0
100
0
100
200
200
300
300
400
400
500
(a)
j
[A cm2]
j
[A cm2]
450
0.0
0.5
1.0
E vs. RHE
[V]
500
1.5
(b)
0.0
0.5
1.0
1.5
E vs. RHE
[V]
Voltammetric proles of CO oxidation on CO-saturated 0.1 M HClO4

solutions. (a) Au(110) electrode. The dashed line corresponds to the voltammetric
prole obtained in the supporting electrolyte alone. Scan rate 100 mV cm1 .
(Reproduced with permission from Ref. [24]). (b) Pt(111) electrode. Scan rate
20 mV cm1 . (Reproduced with permission from Ref. [25].)
Fig. 3
requires a metal that adsorbs OH and

CO with moderate energies. We can then
compare, in this case, the performance
of platinum and gold for the oxidation.
To assess the interactions between the
CO molecule and the electrode surface,
in situ Fourier Transform Infrared (FTIR)
spectroscopy has been used. For platinum
electrodes, strong absorption bands from
adsorbed CO have been detected [1921],
even after CO has been eliminated from
solution, whereas, in the case of gold, very
weak absorption bands are only observed
when CO is present in solution [22, 23].
Therefore, it can be concluded that CO is
strongly adsorbed on the platinum surface
in comparison to gold surfaces. Although
the adsorption energy for OH on both
surfaces is different, being lower for the
more noble metal (gold), the main factor
in the differences of both electrocatalytic

activities is the CO-adsorption energies.
In spite of the fact that the catalytic
performance of these metals towards CO
oxidation is far from optimum (G values
for the reaction indicates that solution
CO should be oxidized at 0.0 V), the
oxidation of CO from solution on gold
takes place at less positive potentials than
on platinum (Fig. 3) [24, 25]. The low
activity of the platinum surface is related
to the strong adsorption of CO, which
hinders its oxidation. In this case, the
strong adsorption not only stabilizes the
adsorbed CO but also prevents OH from
adsorption since it completely covers the
platinum surface.
Other reaction mechanisms present
additional complications. That is the
case of the parallel-path mechanisms,
especially when an intermediate is strongly

adsorbed on the surface poisoning the
active sites and thus preventing further
reaction of the molecule. These parallelpath mechanisms are often found for the
oxidation of organic molecules in which
the poisoning intermediate is normally
the adsorbed CO. The simplest parallelpath mechanism is that of the oxidation
of formic acid. Schematically, it can
written that
Active intermediate
E1
HCOOH
CO + H2O
E2
CO2
E2 > E1
CO2
(17)
The simplicity of this reaction scheme
makes formic acid oxidation a model
reaction in electrocatalysis. For these
reactions, it is important to assess the
catalytic activity of the surface for both
paths (specially the determination of
the reaction path through the active
intermediate) in order to investigate the
catalytic effects of a given surface. The
problem is the reciprocal interference
of the paths that makes difcult the
evaluation of the reaction rates of both
paths separately. The reaction rate for the
active intermediate path can be obtained
only in the absence of surface poison. In
order to achieve that condition, several
strategies have been used. The simplest
strategy is to measure the current density.
After cleaning the surface of the poison
(that normally can be achieved at potentials
above 0.75 V), the electrode potential is
then stepped to a lower value at which
the oxidation of the molecule takes place.
As the electrode surface at time zero (just
after the step) is free from the poison, the
current extrapolated at t = 0 is the reaction
rate through the active intermediate path.
This method has been used for formic acid
and methanol [26, 27]. This strategy has
also been used in combination with a linear

potential sweep in the technique known as
pulsed voltammetry [28, 29] (Fig. 4).
Another important characteristic of
these reactions is the different behavior
of the molecules when adsorbed on electrochemical and ultrahigh vacuum (UHV)
environments. In UHV, formic acid dissociates on a platinum surface to yield
only H2 and CO2 , and CO is not a reaction intermediate [30, 31]. The differences
can be attributed to the presence of water
molecules on the electrochemical environments. The presence of water, a polar
molecule, guides the adsorption process
of the reacting molecule. Methanol adsorption on Pt(111) surfaces provides a
clear example of such changes induced by
the presence of water (or a solvent). In
UHV, methanol adsorbs on the Pt(111)
surface through the oxygen atom after
losing the hydrogen bonded to it. In electrochemical environments, the rst step
in the methanol oxidation is the breaking
of one out of the three CH bonds, and
the molecule adsorbs on the surface with
the OH group pointing outwards [26]. The
polar nature of the OH group and its interactions with the water molecule (also
polar) prevent the adsorption through the
oxygen group.
5.1.3
Surface Structure Effects

Role of the Surface Symmetry in
Electrocatalysis
It is clear that adsorption processes are
structure sensitive, that is, the adsorption
of species depends on the structure
of the surface site. This change in
structure may involve several changes:
rst, the surface will have a different
energy, as revealed by the dependence
of the work function with the surface
5.1.3.1
451
20
i
[mA cm2]
452
0.2
0.5
E RHE
[V]
(a) Pulsed votammogram; and (b) stationary voltammogram for a
),
Pt(100) electrode in 0.5 M H2 SO4 + 0.1 M HCOOH solution (
pulsed voltammogram (-----). Pulse voltammetry conditions: sampling
window time: 0.30 s; poison oxidation pulse length: 0.360 s; potential of
pulse: 0.95 V; time between the end of the poison oxidation pulse and the
opening of the sampling window: 0.015 s. Sweep rate: 10 mV s1 .
(Reproduced with permission from Ref. [28].)
Fig. 4
structure. As generally accepted, there is

a linear relationship between the work
function and the potential of zero charge
(PZC) of the electrode [32, 33], which
in turn plays an important role in the
adsorption process of the different species
and in the properties of the double
layer. In this way, the changes in the
surface energy plays a double role in
the electrocatalysis, one linked to the
changes in the adsorption energy of the
molecules and the second role related to
the changes in the double-layer structure.
The interfacial structure has been proved
to affect the oxidation/reduction of the
charges species at potentials close to the
PZC, altering the typical ButlerVolmer

kinetics [34].
Second, the changes in surface geometry
will lead to changes in the geometry of the
adsorption sites and the distance between
them. The effect of the changes in the
site geometry can be studied easily using
low-index single-crystal electrodes, since
they have a dened surface structure and
few different types of sites. For instance,
(111) surfaces of face centered cubic (fcc)
metals have on top, bridged-bonded and
two different types of three hollow sites
(tetrahedral and octahedral), whereas (100)
surfaces have on top, bridged-bonded and
four hollow sites. If the most favorable site
for the adsorption of a given species is

the hollow sites, there will be signicant
differences in the energy of adsorption
in the (111) and (100) surfaces, owing
to the different geometry of the sites,
and thus an effect in the electrocatalysis
of the species will be observed. In other
cases, the species may bond to the surface
through more than one surface site. In
this case, the right surface requires not
only the right adsorption sites, but also
these sites have to be at a distance
compatible with the geometry of the
adsorbing molecule.
In most cases, it is almost impossible
to decouple the effects of the surface
energy changes and the changes in the site
geometry. Moreover, in electrochemistry,
adsorption is always a competitive process
since there are other species, like the
solvent or the supporting electrolyte, which
can be adsorbed on the electrode surface
apart from the electroactive species. In
some cases, these competing adsorption
processes govern the adsorption of the
electrocatalytically active species. Owing
to the intrinsic difculty in assigning
the changes of the adsorption to any
specic property of the surface, we will
term the effect related to the surface
properties as surface structure effects. It is
clear that the interpretation becomes more
difcult when polycrystalline materials are
to be considered.
As aforementioned, surface structure effects on adsorption are well known in
electrochemistry. A typical example is
CO adsorption on low-index single-crystal
platinum surfaces. For CO, three different adsorption geometries exist: on-top,
bridge-bonded, and multibonded. The relative population of each adsorption species
and the surface structure of adsorbate layer
are dependent on the surface structure
and metal (see for instance Refs. [1921]
(Fig. 5)). Moreover, multibonded CO, in

which CO is coordinated to three platinum
atoms, is only observed on the Pt(111)
surface. Additionally, the formation of
multibonded CO not only requires a threefold symmetry surface but also long-range
order [35]. Experiments performed using
stepped surfaces show that the multibonded CO does not appear on surfaces
having short terraces with (111) symmetry. This effect suggests that long-range
order domains will have a different reactivity. Additionally, the IR frequencies
for on-top CO are also dependent on the
surface symmetry [1921]. Changes in the
IR frequency are associated to changes in
the bonding energy of the CO molecule
to the surface. If the adsorption energy
of the surface is different for the same
adsorption site in surfaces with different
geometry, the energy of the surface is playing a role in the adsorption process of
CO. This example clearly illustrates the
interrelation between the different properties (symmetry, energy, etc.) of a given
surface regarding its adsorption properties
and the difculty to assign the changes in
its adsorption properties to a given structural factor.
The changes in the adsorption properties
are not only restricted to changes in the
binding geometry for the adsorption or
surface structure. When the species has
several groups that are liable to bond to the
surface, changes in the surface structure
may lead to changes in the binding
group. That is the case for urea adsorbed
on platinum single-crystal electrodes. On
Pt(100), urea is always bonded to the
surface through the two nitrogen atoms of
the molecule [36]. On the other hand, the
adsorption is coverage sensitive on Pt(111).
For low urea coverages, the adsorption
takes place trough only one nitrogen atom,
whereas for high urea coverages, the
453
454

Pt (111)
Pt (100)
qCO = 0.16
Pt (110)
qCO = 0.16
qCO = 0.2
qCO = 0.85
qCO = 1.0
4 103 a.u.
qCO = 0.6
Rh (111)
Rh (100)
qCO 0.2
qCO = 0.75
qCO = 0.24
qCO = 0.75
1800
2000
Rh (110)
qCO = 0.25
2000
1800
qCO = 1.0
2000
1800
n
[cm1]
FTIR spectra for CO irreversibly adsorbed on platinum and rhodium low-index
surfaces in 0.1 M HClO4 at 0.25 V versus SCE at saturation and low CO coverages. The
reference spectrum was acquired at 0.450.5 V after CO oxidation. (Reproduced with
permission from Ref. [19].)
Fig. 5
O
C
HN
NH
Pt (100)
O
C
HN
NH 2
Pt (111)
Low coverage
NH 2
HN
C
O
Pt (111)
High coverage
adsorption is through the oxygen of the

carbonyl group [37] (Fig. 6).
The identication of the different adsorbed species and its relation with the
catalytic activity of the surface is much
more troublesome. The problem generally
arises from the short life of the adsorbed
intermediates, which makes difcult its
detection with the usual techniques. In
spite of this fact, the effects of the surface
structure on electrocatalysis are beginning to be well documented. The typical
example is again formic acid oxidation
Fig. 6 Proposed adsorption modes of
urea on Pt(111) at low and high urea
coverages and on Pt(100). (Reproduced
with permission from Ref. [37].)
on platinum single-crystal electrodes. The

effect of the surface structure affects both
the direct and the poisoning paths for
the formic acid oxidation [38]. Among the
three basal planes, the lowest poisoning rate corresponds to Pt(111), which
also shows the lower catalytic activity [39,
40]. When comparing the intrinsic activity of Pt(111) and Pt(100) in sulfuric acid
medium, it can be seen that the current
densities for the Pt(100) surface at 0.5 V
versus RHE are 10 times higher than
those obtained on Pt(111) [28, 29]. Similar behavior is obtained for methanol on
platinum electrodes (Table 1) [27]. These
results clearly indicate that the reactivity
of the molecule depends on the electrode
surface structure.
Methanol oxidation also illustrates the
effect of the competing adsorption on
the electrocatalysis. The presence of a
strong adsorbing anion on the supporting electrolyte hinders the oxidation of
the molecule (Fig. 7). Thus, the current
densities for methanol oxidation in phosphoric acid solution are lower than those
obtained in sulfuric acid, being the highest values those obtained in perchloric acid
solution [26, 27]. This is because the oxidation of methanol molecules takes place
at potentials in which the anions are adsorbed on the electrode surface. Therefore,
Maximum voltammetric peak current
densities (jCV ), and instantaneous current
densities at 0.2 V (jE=0.2 ) and 0.4 V (jE=0.4 ) (vs.
Ag/AgCl) for methanol oxidation in 0.1 M
H2 SO4 + 0.1 M HClO4 for Pt(111), Pt(110), and
Pt(100) electrodes. Data taken from Ref. [27]
Tab. 1
jCV (mA cm2 )

jE=0.2 (mA cm2 )
jE=0.4 (mA cm2 )
Pt(111)
Pt(110)
Pt(100)
0.74
0.57
4.8
24.5
6.5
17.0
4.2
5.2
6.1
the rst step in the adsorption process of

the species of interest is the replacement
of the adsorbed anions. The stronger the
anion adsorption, the more difcult the
displacement of the anion by the incoming
molecule is, resulting in a lower current. The voltammetric characteristics for
formic acid or methanol oxidation remain
essentially the same, but lower currents
are obtained when anions are strongly adsorbed on the electrode surface.
Until this point, the use of low-index
crystal surface electrodes has allowed
to get insight into the electrocatalytic
phenomena. However, real surfaces are far
from being perfect and comprise several
types of sites, including step and kink sites,
and the size of the ordered domains is
rather limited. For that reason, it is also
important to rationalize the effect of such
sites in the reaction, using stepped and
kinked surfaces. Because of its complexity,
there are few examples of these effects
in the literature, but the research on the
subject is expected to become more and
more important.
The presence of defects in a given surface structure may also play an important
role in the electrocatalytic process. Normally, long-range ordered domains are
separated from each other by step or kink
sites. The substrate atoms that belong to
kink or step sites have special adsorption
properties since they have a lower coordination number, that is, they have a lower
number of nearest neighbor substrate
atoms than the surface atoms in a wellordered domain. These properties may result in an increased catalytic activity. In order to model the real surfaces, stepped and
kinked surfaces are used since they have a
well-dened distribution of these sites and
they can be prepared reproducibly.
A clear example of the role of defect
sites in electrocatalysis can be found in
455

0.4
0.2
0.0
0.2
0.4
0.6
0.8
0.6
0.8
i
[mA cm2]
2.0
1.5
1.0
0.5
(a)
0.0
i
[mA cm2]
1.00
0.75
0.50
0.25
(b)
0.00
0.50
0.40
i
[mA cm2]
456
0.30
a3
0.20
0.10
a1
0.00
0.4
(c)
a2
0.2
0.0
0.2
0.4
E (Ag/AgCl)
[V]
Voltammetric proles for the Pt(111) electrode in: (a) 0.1 M

HClO4 + 0.2 M CH3 CH2 OH; (b) 0.1 M H2 SO4 + 0.2 M
CH3 CH2 OH; and (c) 0.1 M H3 PO4 + 0.2 M CH3 CH2 OH. Scan rate
50 mV s1 . (Reproduced with permission from Ref. [27].)
Fig. 7
the oxidation of adsorbed CO on Rh(111)

electrodes [41]. For well-ordered surfaces,
the stripping of the adsorbed CO and
the recovery of the clean surface is
quite difcult. However, this process is
considerably easier in a stepped surface.
This fact clearly indicates the necessity of

defect sites in the surface structure of the
Rh(111) surface in order to oxidize the
adsorbed layer.
Glucose oxidation is a clear example
of the effects of the domain size of the
electrode in the electrocatalysis [4245].

This molecule is relatively big and has
several groups that can be attached to the
electrode surface. Consequently, it prefers
relatively at areas to adsorb and oxidize.
For this reason, the maximum catalytic
activity is found for the Pt(111) [4244]
and Pt(100) electrodes [44]. If stepped
surfaces are used, the oxidation current is
proportional to the terrace size [44], a clear
indication of the inuence of the domain
sizes in the reaction (Fig. 8). Moreover,
an additional oxidation process at low
potentials appears that corresponds to the
oxidation on the step sites at which glucose
adsorption would be more reactive.
Chiral Surfaces
As aforementioned, defect sites (kink
sites) may have an important role
in electrocatalysis. Kink sites have an
j
[mA cm2]
5.1.3.2
additional property they are chiral. This

means that it is possible to prepare two
nonoverlapping specular surfaces having
the same type of kink site. Analogously to
the chiral molecules, the kink sites have
been termed as R or S depending on its
symmetry [46, 47]. If an enantiomer interacts with a chiral site, the reactivity can be
dependent on the symmetry of the chiral
site, leading to the formation of different reaction products and/or to different
reaction rates. Therefore, chiral surfaces
exhibit stereoselective activity.
In order to observe the stereoselective
behavior of the surface, the kink site and a
chiral center of the enantiomeric molecule
have to be involved in the adsorption
process that yields the reaction products.
Moreover, owing to the symmetry rules,
crossed activity has to be observed, that
is, the behavior of an R surface with an S
(100)
(27, 1, 1)
(19, 1, 1)
(11, 1, 1)
(711)
1.0
0.5
(511)
(311)
0
0.5
1.0
E RHE
[V]
Voltammetric proles for the oxidation of 0.01 M glucose in 0.1 M
HClO4 on different Pt(s):[n(100) (111)] stepped surfaces. Scan rate
50 mV s1 . (Reproduced with permission from Ref. [44].)
Fig. 8
457
500
400
300
200
100
0
100
Current density
[A cm2]
Current density
[A cm2]
.2
.6
.4
.6
.2
.8 .9
E (Pd/H ref)
[V]
.6
.8 .9
500
400
300
200
100
0
100
0
(d)
.4
E (Pd/H ref)
[V]
(b)
Current density
[A cm2]
.2
500
400
300
200
100
0
100
.8 .9
500
400
300
200
100
0
100
0
(c)
.4
E (Pd/H ref)
[V]
(a)
Current density
[A cm2]
458
.2
.4
.6
.8 .9
E (Pd/H ref)
[V]
Voltammograms for glucose oxidation on the two chiral Pt(643) electrodes in 0.05 M
H2 SO4 supporting electrolyte. (a) Pt(643)S electrode in 5 mM D-glucose; (b) Pt(643)S electrode
in 5 mM L-glucose; (c) Pt(643)R electrode in 5 mM D-glucose; and (d) Pt(643)R electrode in
5 mM L-glucose. Scan rate: 50 mV s1 . (Reproduced with permission from Ref. [47].)
Fig. 9
molecule should be the same as that of an S

surface with an R molecule. Such behavior
has been observed with L- and D-glucose
and the R and S Pt(643) surfaces [47, 48].
As can be seen in Fig. 9, the voltammetric
prole for the oxidation of the molecule
depends on the surface symmetry and the
glucose isomer. The main difference is
the apparition of a new peak at 0.31 V in
the S D and R L combinations, which is
absent in the other two cases. The current
density at 0.31 V increases with the kink
density, thus proving its relationship with
the site symmetry.
5.1.4
Surface Composition Effects
In many cases, the most active metals

available have only a limited catalytic
activity for the reaction considered. In

those cases, an alternative is to try to
modify the surface properties of the metal
in order to increase its activity. In most
cases, this can be achieved by depositing
a foreign ad-atom on the surface. This
particular case can be compared to alloys
but is should be remembered that bulk
properties are different, that is, the bulk
phase is a pure metal and only the surface
has a different composition in the case
of a modied electrode. It is assumed
that the foreign ad-atom will maintain
the surface structure of the substrate,
leading to the growth of a well-dened
surface layer. The new species on the
surface can affect the desired reaction in
three different ways. (1) It may change the
electronic properties of the surface in such
a way that the adsorption of the reactant
j
[A cm2]
j
[mA cm2]
100
0.5
E (RHE)
[V]
10
0.5
E (RHE)
[V]
Quasi-stationary voltammetric proles of Pt(100) (-----) and
) electrodes in 0.25 M
PdPt(100) (Pd = 0.32) (
HCOOH + 0.5 M H2 SO4 solution. Inset: blank voltammograms of the
same electrodes in 0.5 M H2 SO4 solution. Scan rate 50 mV s1 .
(Reproduced with permission from Ref. [49].)
Fig. 10
is modied and the catalytic properties

of the surface are enhanced. (2) It also
may affect the availability of adsorption
sites required for a given reaction, that
is, blocking the possibility of a poisoning
reaction. (3) It may provide new reaction
sites for some of the reactive species, acting
as a bifunctional catalyst. We will explain
these effects with some examples.
5.1.4.1 Changes in the Electronic
Properties of the Metal
When a foreign ad-atom is deposited on a
surface, the electronic properties of the ensemble ad-atomsurface atom are different from that of the original surface atoms.
One of the clearest examples is that of
formic acid oxidation on Pt(100) + Pd [49].
On the unmodied Pt(100) electrode,
the voltammetric prole for formic acid
oxidation is shown in Fig. 10 (dashed line).

As can be seen, no current is recorded in
the positive-going scan, since the surface
is completely covered by the poison (CO).
In the negative-going scan, the maximum
current is round 17 mA cm2 at 0.45 V
(close to the intrinsic activity of Pt(100)
electrodes). When palladium is deposited
on the electrode surface, the oxidation peak
shifts to 0.21 V, both in the positive- and
negative-going scans (Fig. 10, full line).
This peak reaches its maximum when the
palladium coverage is ca. 0.5, with an intensity of ca. 55 mA cm2 . The current
density at 0.21 V for the Pd-modied surface is much higher than the intrinsic
activity at this potential obtained for the
unmodied Pt(100) surface and for a palladium electrode (negligible currents are
recorded in both scans for the unmodied
459
460
electrode at this potential, see Fig. 4), and

the oxidation potential has diminished
ca. 0.3 V.
It is worth noting that the catalytic
effect of the ad-atom-modied electrodes
can be different from that found for
an alloy of these metals, since even in
the case of solid solutions in the whole
composition range, the surface structure
is not the same. In the alloy, there is a
microscopic mixture of the two species
participating in the alloy, whereas in
the ad-atom-modied electrodes, the adatoms are deposited on top of the surface.
Moreover, the surface concentration of the
different metallic atoms in the alloy can be
completely different from that obtained
in the bulk, making the interpretation
of the results obtained with the alloy
electrodes more difcult. In this respect,
much more research is needed in order to
fully understand the different behavior of
the alloys.
Ensemble Effects
In the previous section, we have discussed
the cases in which the ad-atom plays
a positive inuence on the catalytic
properties of the surface. In other cases,
the ad-atom may not exert any effect. The
only consequence of the presence of the
ad-atom is that some of the active catalytic
sites of the surface become blocked. It
is said that the ad-atom is acting like a
third body. In some cases, a third-body
effect may result in a surface that has
some interesting properties. That is the
case when the reaction mechanism has
parallel paths, or there is a competing
reaction. If the site requirements for the
different paths are different, the catalytic
activity of the ad-atom-modied surface
may be altered. These effects are known
as ensemble effects, since they require a
given structure of the adlayer to appear.
5.1.4.2
In the case of formic acid oxidation,

this effect is well documented. It has been
proposed that the dissociation of formic
acid to yield water and CO (the poisoning
intermediate) requires two adjacent sites,
whereas the direct oxidation through the
active intermediate demands only one
site [50]. If the ad-atom deposits randomly
on the electrode surface, there will be
a coverage at which the distribution of
the ad-atoms does not leave two adjacent
sites. In that case, the dissociation reaction
to yield the CO molecule is completely
inhibited. Of course, the reaction through
the active intermediate exhibits lower
currents, since the number of available
adsorption sites on the surface are lower,
but the electrode presents the advantage
of a very low poisoning rate and stable
currents with time. A good example of this
behavior is found for selenium-modied
Pt(111) surfaces [50]. As the selenium
coverage is increased, the currents at 0.5 V
and the total poison accumulation decrease
linearly, as expected for ad-atoms acting
only as a third body. At a selenium coverage
of 0.28 (coverage at which 84% of the
surface platinum sites have been blocked
by the ad-atom) the poison formation is
completely inhibited whereas the surface
still oxidizes formic acid [50].
Bifunctional Catalysts
For some reactions, two or more species
have to be adsorbed on the surface for
the reaction to proceed. That is the case,
for instance, of CO oxidation on metal
electrodes. As aforementioned, CO and
an oxygen-containing species have to be
adsorbed on the electrode surface. For
such cases, electrocatalysis is much more
difcult, since the adsorption energy for
the different species has to be the adequate
for the reaction to proceed. In some cases,
the adsorption energy is good for one of
5.1.4.3
the reactant species but not for the other,

resulting in low current densities for the
process. If it is possible to nd an adatom that has more adequate adsorption
energy for the second adsorbing species,
the surface will exhibit better catalytic
properties. In this case, the reaction
will take place preferentially between one
species adsorbed on the original surface
and the second species adsorbed on the
ad-atom. That kind of catalyst is called a
bifunctional catalyst.
The typical example of a bifunctional
catalyst is the ruthenium-modied platinum surfaces for the oxidation of
methanol [5158]. Of all the metals studied, platinum is the one that displays the
best catalytic activity for methanol oxidation. However, its catalytic activity is
still too low for practical purposes. The
problem arises from the necessity of the
adsorption of a second species able to
transfer the oxygen group required for its
complete oxidation to CO2 . Ruthenium is
known to have a better afnity for OH than
platinum but its interaction with methanol
is very poor [59]. Thus, a platinum surface, which has good interaction energy
with methanol, modied with ruthenium,
whose interaction with OH is better than
that of platinum, exhibits better catalytic
properties than platinum.
The observed catalytic enhancement is
potential dependent, since the adsorption
energy for OH is also potential dependent.
The energy of OH adsorption increases
with the potential. Ruthenium adsorbs OH
at less positive potentials than platinum,
and therefore the onset for methanol oxidation takes place at lower potentials [56].
If the potential is increased, the adsorption
energy for OH will increase. At a specic
potential, the energy of OH adsorption will
be too high for the reaction, and the modied surface will display lower catalytic
activity than the platinum surface [56].

This kind of bifunctional mechanism is
also operative for CO oxidation on the
same electrode surface.
It is important to establish if the
mechanism of the observed catalytic effect
played by an ad-atom is electronic or if
the ad-atom is acting as a bifunctional
catalyst. The problem arises from the fact
that both effects can be considered as
short-range effects, that is, the catalytic
enhancement is only observed in the
surface substrate atoms closest to the adatom. For the electronic effects, it is known
that the effect induced by an ad-atom
extends to the rst and second row or
neighboring atoms to the ad-atoms [60].
For the bifunctional catalyst, the pair
responsible for the catalytic effect is the
combination of an active site in the adatom and an active site in the surface
close to it. This way, the determination
of the electrocatalytic mechanism requires
a detailed knowledge of the distribution
of the ad-atom on the surface and
the condition under which the catalytic
enhancement is found. These questions
are currently under investigation.
That is the case, for instance, of CO oxidation in surfaces modied with arsenic
and bismuth (Fig. 11) [61, 62]. Both adatoms catalyze the oxidation of adsorbed
CO layers. From the measured IR frequencies for CO, it is clear that arsenic and
bismuth alter the surface energy of the
electrode. However, the frequency shifts
are in opposite directions, the effect of
arsenic is a blue shift of the frequencies
whereas the presence of bismuth causes a
red shift [62]. If the effective mechanism
of these ad-atoms were an electronic effect,
the effects of bismuth and arsenic would be
the opposite. Owing to the fact that the electrocatalytic effect is only observed at potentials at which the ad-atoms adsorb OH, the
461
j
[A cm2]
30
20
10
25
75
50
t
[s]
(a)
100
j
[A cm2]
462
50
0
(b)
15
10
20
25
t
[s]
CO-oxidation transients for Pt(111) at 0.70 V for

(a) As = 0.04 and CO = 0.68; (b) Bi = 0.10 and
), As = 0.00 and CO = 0.56 (- - - - - -).
CO = 0.56 (
(Reprinted with permission from Ref. [62].)
Fig. 11
main electrocatalytic mechanism of these

ad-atoms is a bifunctional mechanism.
Modeling of the Catalytic Effects
The last point in electrocatalytic studies is
trying to model the catalytic effect of the
ad-atoms to achieve a full knowledge of its
5.1.4.4
effect and to be able to predict its catalytic

properties regarding other molecules or
other surface ad-atoms systems. Owing
to the complexity of the simulations, it
is important to have a well-dened system in which the experimental data are
well understood. The most comprehensive
simulations use clusters to model the
surface and tried to determine the most

favorable adsorption sites and the activation energy for the process. However, as
a result of the long computational times
required to calculate the different parameters of the process, the size of the cluster
are limited to a small number of metal
atoms and to one or two different species
adsorbed on the cluster surface in most
of the cases. Therefore, the results obtained with these simulations represent a
semiquantitative approach to the process.
Moreover, the fact that the solvent is absent from most of the simulations and
the lateral interactions between adsorbed
species are not taken into account, the
applications to electrochemical problems
are more limited. However, some studies have tried to calculate the activation
energy for the CO-oxidation process in a
platinum surface with and without ruthenium, obtaining a potential dependence of
the activation energy [63, 64].
In other cases, a simpler simulation
based on statistical models is preformed.
This kind of simulation was performed to
determine the effect of the different adatoms in the oxidation of formic acid on
platinum electrodes [65]. In this model, the

ad-atoms can play an effective catalysis on
the process or act as a simple third body.
In the rst case, the activity of the surface
is directly proportional to the number of
pair ad-atomsurface sites, whereas in the
latter case, the activity is proportional to
the number of unoccupied surface sites.
Since oxidation of formic acid takes place
though a parallel-path mechanism, the
effects of different levels of poisoning are
also considered. For the cases in which the
surface is completely covered by poison,
both types of ad-atoms (the catalytically
effective ad-atoms and the third-body adatoms) produce similar qualitative effects,
that is, both types increases the current for
the oxidation of formic acid [65]. Of course,
the catalytic enhancement is higher in the
case of the ad-atoms that modies the
electronic properties of the surface, since
the global effect will be the combination
of the electronic enhancement and the
third-body effect (any ad-atom always acts
as a third body). This is the case, for
instance, for the Pt(100) surfaces modied
with ad-atoms [6567]. For the surfaces
with low poisoning, that is, the Pt(111)
50
j
[mA cm2]
40
Comparison between
experimental () and theoretical
currents versus coverage curves for the
electrooxidation of formic acid at
BiPt(111) electrode in 0.25 M
HCOOH + 0.5 M H2 SO4 . (Reproduced
with permission from Ref. [65].)
Fig. 12
30
20
10
0
0.0
0.2
0.4
0.6
qBi
0.8
1.0
463
464
electrode, the distinction between the 14. A. T. Kuhn, H. Wroblowa, Trans. Faraday.
Soc. 1967, 63, 1458.
effects of both types of ad-atoms is clear.
For the electronically effective ad-atoms,
1973, 44, 239.
the catalytic activity of the surface increases 16. R. R. Adzic, D. N. Simic, A. R. Despicm
with the ad-atom coverage until it reaches a
maximum when 75% of the initial surface 17. A. J. Appleby, Catal. Rev. 1970, 4, 221.
sites have been covered, that is, bismuth- 18a. S. Schuldiner, J. Electrochem. Soc. 1963, 110,
332.
modied Pt(111) surfaces (Fig. 12) [65, 68]. 18b. S. Schuldiner, J. Electrochem. Soc. 1968, 115,
For the ad-atoms acting only as a third
362.
body, the current of the oxidation of formic 19. S.-C. Chang, M. J. Waver, Surf. Sci. 1990,
238, 142.
acid diminishes with the ad-atom coverage,
20. S.-C. Chang, J. D. Roth, Y. Ho et al., J. Elecas observed with the selenium-modied
tron Spectrosc. Relat. Phenom. 1990, 54/55,
Pt(111) surfaces [65].
1185.
References
1. A. J. Appleby in Comprehensive Treatise
of Electrochemistry (Eds.: B. Conway,
J. OM. Bockris, E. Yeager et al.), Plenum
Press, New York, 1983, pp. 173239, Vol. 7.
2. L. I. Kristalik in Advances in Electrochemistry and Electrochemical Engineering (Eds.:
P. Delahay, C. Tobias), John Wiley & Sons,
New York, 1970, pp. 283339, Vol. 7.
3. R. Parsons, Proc. Conf. 75th Anniv. Real
Soc. Esp. de Fis. Y Quim, Madrid, 1981,
pp. 352358.
4. J. OM. Bockris, A. Reddy, Modern Electrochemistry, Plenum Press, New York, 1970.
5. J. OM. Bockris, S. U. M. Khan, Surface Electrochemistry, Plenum Press, New York, 1973.
6. R. Adzic in Modern Aspects of Electrochemistry (Eds.: R. E. White, J. OM. Bockris,
1990, pp. 163236, Vol. 21.
7. J. Lipkowski, P. N. Ross, (Eds.), Electrocatalysis, Wiley-VCH, New York, 1998.
8. A. Wieckowski, (Ed.), Interfacial Electrochemistry, Marcel Dekker, New York, 1999.
9. M. J. Weaver in Chemical Kinetics (Ed.:
R. G. Compton), Elsevier, Amsterdam, 1987,
pp. 160, Vol. 27.
10. R. A. Marcus, J. Electroanal. Chem. 2000, 483,
2.
11. J. N. Brnsted, K. J. Pedersen, Z. Phys. Chem.
1923, 108, 185.
12. R. I. Masel, Principles of Adsorption and Reaction on Surfaces, John Wiley & Sons, New
York, 1996.
13. S. Trasatti, J. Electroanal. Chem. 1972 39, 163.
21. S.-C. Chang, J. D. Roth, M. J. Waver, Surf.

Sci. 1991, 244, 113.
22. S.-C. Chang, A. Hamelin, M. J. Weaver, J.
Phys. Chem. 1991, 95, 5560.
23. S.-C. Chang, A. Hamelin, M. J. Weaver, Surf.
Sci. 1990, 239, L543.
24. G. J. Edens, A. Hamelin. M. J. Weaver, J.
Phys. Chem. 1996, 100, 2322.
25. A. Wieckowski, M. Rubel, C. Gutierrez, J.
26. K. Franaszczuk, E. Herrero, P. Zelenay et al.,
J. Phys. Chem. 1992, 96, 8509.
27. E. Herrero, K. Franaszczuk, A. Wieckowski,
J. Phys. Chem. 1993, 97, 9730.
28. J. Clavilier, J. Electroanal. Chem. 1987, 236,
87.
30. M. R. Columbia, P. A. Thiel, Surf. Sci. 1990,
235, 53.
31. M. R. Columbia, A. M. Crabtree, P. A. Thiel,
J. Am. Chem. Soc. 1992, 114, 1231.
32. S. Trasatti in Advances in Electrochemistry
and Electrochemical Engineering (Eds.:
H. Gerischer, C. W. Tobias), Wiley Interscience, New York, 1977, pp. 213321,
Vol. 10.
33. W. Schmickler, Interfacial Electrochemistry,
Oxford University Press, New York, 1996.
34. A. N. Frumkin, Z. Electrochem. 1955, 59, 809.
35. A. Rodes, R. Gomez, J. M. Feliu et al., Langmuir 2000, 16, 811.
36. V. Climent, A. Rodes, J. M. Orts et al., Langmuir 1997, 13, 2380.
37. V. Climent, A. Rodes, J. M. Orts et al., J.
38. C. Lamy, J. M. Leger, C. Clavilier et al., J.
39. S. Motoo, N. Furuya, Ber. Bunsen-Ges. Phys.
Chem. 1987, 91, 457.
40. S. Motoo, N. Furuya, J. Electroanal. Chem.
1985, 184, 303.
41. R. Gomez, J. M. Orts, J. M. Feliu et al., J.
42. K. Popovic, A. Tripkovic, N. Markovic et al.,
43. K. Popovic, A. Tripkovic, N. Markovic et al.,
45. A. Rodes, M. J. Llorca, J. M. Feliu et al., Anal.
Quim. 1996, 92, 118.
46. C. F. McFaden, P. S. Cremer, A. J. Gellman,
Langmuir 1996, 12, 2483.
47. G. A. Attard, A. Ahmadi, J. M. Feliu et al.,
Langmuir 1999, 15, 1420.
48. G. A. Attard, A. Ahmadi, J. M. Feliu et al., J.
Phys. Chem. B 1999, 103, 1381.
50. M. J. Llorca, E. Herrero, J. M. Feliu et al., J.
51. M. Watanabe, S. Motoo, J. Electroanal. Chem.
1975, 60, 267.
52. K. Franaszczuk, J. Sobkowski, J. Electroanal.
Chem. 1992, 327, 235.
53. T. Iwasita, F. C. Nart, W. Vielstich, Ber.

et al., J. Phys. Chem. 1993, 97, 12 020.
57. W. Chrzanowski, A. Wieckowski, Langmuir
1997, 13, 5974.
58. T. Frelink, V. Visscher, J. A. R. van Veen,
Surf. Sci. 1995, 335, 353.
59. B. J. Kennedy, A. W. Smith, J. Electroanal.
Chem. 1990, 293, 103.
60. P. J. Fiebelman, D. R. Ramman, Phys. Rev.
Lett. 1984, 52, 61.
61. E. Herrero, J. M. Feliu, A. Aldaz, J. Catal.
1995, 152, 264.
62. E. Herreo, A. Rodes, J. M. Perez et al., J.
63. A. B. Anderson, E. Grantscharova, J. Phys.
Chem. 1995, 99, 9143.
64. A. B. Anderson, E. Grantscharova, S. Seong,
J. Electrochem. Soc. 1996, 143, 2075.
65. E. Leiva, T. Iwasita, E. Herrero et al., Langmuir 1997, 13, 6287.
67. J. Clavilier, A. Fernandez-Vega, J. M. Feliu
68. J. Clavilier, A. Fernandez-Vega, J. M. Feliu
465
466
5.2
CO, Formic Acid, and Methanol Oxidation

in Acid Electrolytes Mechanisms and
Electrocatalysis
Wolf Vielstich
Instituto de Quimica de Sao Carlos-SP,
USP, Brasil
CH3 OH + 8OH CO3 2
5.2.1
+ 6H2 O + 6e
Scope of the Chapter
The anodic oxidation of methanol was considered as one of the most interesting
subjects in electrochemistry during the
last 15 years. Besides very characteristic
reaction pathways and electrocatalytic effects, the methanol molecule with its four
hydrogen atoms is the basis of a highenergy density liquid fuel. The reaction of
methanol with oxygen follows the chemical route
CH3 OH + 1.5O2 CO2 + 2H2 O
(1)
with a reaction enthalpy of 726.6 kJ mol1
and with a Gibbs Free Energy of
702.5 kJ mol1 under standard conditions.
The high-energy density of a combination
of liquid methanol and liquid oxygen was
already used since 1940 by Wernher von
Braun, for the propulsion of the rst rocket
in space. But in the 1960s, the electric energy onboard the space ship was not taken
from a methanol/oxygen fuel cell, but from
a hydrogen/oxygen system. At this time,
the electric data of a methanol/oxygen fuel
cell had not been sufcient for this application. Especially in acid solution, the rate
of methanol oxidation
CH3 OH + H2 O CO2 + 6H+ + 6e
(2)
in combination with oxygen reduction at
the cathode
1.5O2 + 6H+ + 6e 3H2 O
was very low, even with platinum as a

catalyst [1].
By the above reasons, the possibility of
using methanol in an alkaline electrolyte
was studied, showing a much better rate
of oxidation at more suitable potentials [2].
But the following equation:
(3)
(4)
shows that part of the electrolyte is

consumed and carbonate ions are formed.
After consumption, the cell has to be lled
again. Nevertheless, some applications of
an alkaline system have been studied [3].
It is interesting to note that at platinum,
formate is the main intermediate of the
reaction [4].
The electrochemistry for CO and formic
acid also are different in alkaline solution.
As opposed to acid solutions, two chemical
pathways introduce a catalytic conversion
of CO and OH ions to hydrogen as
active material. From the adsorbed state,
carbonate ions and/or formate can be
formed [5, 6].
COad + 2OH CO3 2 + 2Had (5)
COad + OH HCOO
(6)
Formate reacts with an additional OH

ion also to carbonate ions and hydrogen
HCOO + OH CO3 2 + 2Had
(7)
In both cases, the charge-transfer step is
the oxidation of hydrogen.
The interest in the oxidation of CO,
formic acid, and methanol in acid media
did come relatively late. It was only
in 1986 that the European Community
started a project for an acid methanol
fuel cell, a direct methanol fuel cell,
DMFC [7, 8]. This development is still
5.2 CO, Formic Acid, and Methanol Oxidation in Acid Electrolytes Mechanisms and Electrocatalysis
going on. In the following text, we will

therefore concentrate our survey mainly
on methanol. But CO and formic acid are
directly connected with this approach, due
to the fact that formic acid is formed in one
pathway of methanol oxidation, and CO
is playing an important role as adsorbed
product or main intermediate, depending
on operating conditions such as electrode
potential, catalyst, and temperature.
In acid media, platinum is an essential
part of a catalyst, because at platinum the
adsorbed methanol molecule is especially
able to start different reaction pathways
already at room temperature, while, for
example, at ruthenium a temperature of
about 60 C is necessary. For this reason,
the reactions of CO and formic acid have
to be studied on ruthenium and PtRu
catalysts also.
In the following paragraph, we will start
with some thermodynamic considerations
and then we will give a survey on CO
oxidation and the reaction pathways of
formic acid and methanol oxidation.
5.2.2
Basic Facts
G = H T S
Thermodynamics, Open-circuit
Potentials
From thermodynamics, one obtains the
electrode potential on the basis of a
reversible process at the surface of the
catalyst. Our top example is the hydrogen
electrode with its high exchange current at
platinum in acid solution:
(8)
For our model species, carbon monoxide,

formic acid, and methanol, such an equilibrium does not exist. Nevertheless, the
open-circuit potential observed for formic
acid and methanol is not too far from the
thermodynamic values. Thermodynamic
(9)
The reaction of methanol and oxygen,

for example, can be an electrochemical
one (Eqs. 13), split into two separated
charge-transfer processes at the anode
and cathode. Then we have as maximum
possible electromotive force
E=
G
nF
(10)
This is the potential difference between

the oxygen and the methanol electrode.
With the thermodynamic potential of
oxygen, +1.23 V versus hydrogen, we now
can place the thermodynamic data for
methanol in relation to the hydrogen
reference electrode. In addition, we can
calculate the ideal efciency of the
methanol/oxygen reaction by relating G
to the change in enthalpy:
id =
5.2.2.1
Had + H2 O H3 O+ + e
data are obtained via the Gibbs free energy

change G, accompanying a chemical reaction at the absolute temperature T , and
is related to the corresponding changes in
enthalpy and entropy
G
[%]
H
(11)
Doing the same for CO and HCOOH, we

can list our data in Table 1 below.
While the theoretical potential of CO
and methanol are positive to the hydrogen
reference, for formic acid the data is negative. But experiments show (Sect. 5.2.3)
that open-circuit potentials are much more
positive than the thermodynamic values,
especially in the case of formic acid. This
is a consequence of the missing reaction
equilibrium. In addition, the increase in
rate positive to the open-circuit potential
is much less than in the case of hydrogen
oxidation. The rate of charge transfer at
a given potential depends strongly on the
catalyst and its surface structure.
467
468

Tab. 1 Thermodynamic data for the oxidation reactions of hydrogen, CO, formic acid, and methanol
under standard conditions at 25 C, standard potential 0 , and ideal efciency
Fuel
Reaction
H0
[kJ mol1 ]
G0
[kJ mol1 ]
0 [V]
Hydrogen
H2 + 12 O2 H2 Ol
286.0
237.3
0.000
83.0
CO
CO + 12 O2 CO2
283.1
257.2
0.163
90.9
Formic Acid
HCOOH + 12 O2 CO2 + H2 Ol
270.3
285.5
0.251
105.6
726.6
702.5
0.015
96.7
Methanol
CH3 OH +
3
2 O2
CO2 + 2H2 Ol
Reaction Pathways
In the introduction (Sect. 5.2.1), we
learned already that in alkaline solution
CO shows two possible reaction pathways.
In acid solution this is not the case. But we
have now to discuss the different pathways
for formic acid and methanol.
As already proposed by Capon and
Parsons [9], the anodic reaction of formic
acid at platinum proceeds indeed via
a dual-path mechanism. One pathway
is producing CO2 in a direct way via
dehydrogenation
5.2.2.2
HCOOH CO2 + 2H+ + 2e (12)

and in a parallel pathway a surfaceblocking residue is formed via dehydration,
HCOOH COad + H2 O
(13)
followed by the oxidation of the adsorbate

as the rate-determining step:
H2 Oad OHad + H+ + e
(14)
and
COad + OHad CO2 + H+ + e
(15)
The direct pathway possibly involves a
reactive intermediate like COOH [9]. This
step is still unknown, but the formation
of hydrogen atoms has been shown using
a palladium membrane as sensor [10, 11].
On the other hand, CO is a very well-known

intermediate (or product, depending on
potential and catalyst used). The existence
of the direct path above is proven by
an experiment with upd-lead [12]. It was
shown that at a coverage of 84% of
the platinum surface (1) an additional
formation of adsorbate is not possible and
(2) on 16% free surface sites the oxidation
of formic acid continues. A second proof
of the dual-pathway model has been made
via a Differential electrochemical mass
spectrometry (DEMS) experiment and it
will be discussed in Sect. 5.2.3.1.
Methanol oxidation takes place via at
least two different pathways. The main
route is producing CO again as a poisoning adsorbate. In a second alternative, formic acid is formed, which reacts
in solution with methanol to methylformate (Sect. 5.2.3). The rst route leads
to hydrogen-containing species as formulated by Bagotzki and coworkers [13]:
CH3 OH CH2 OHad + H+ + e
(16)
or
CH2 OHad H2 COad + H+ + e
or
CH2 OHad CHOHad + H+ + e
(17)
CHOHad COHad + H+ + e
or
CHOHad COad + 2H+ + 2e
(18)
or
H2 COad HCOad + H+ + e (19)
The presence of adsorbed CO was
shown by in situ Fourier Transform
Infrared Spectroscopy (FTIRS) [14] and
the presence of adsorbed COH by
FTIRS [15], the relative amounts depending on methanol concentration. COH is
favored by low methanol concentrations
as proved via Thermal Desorption Spectroscopy (TDS) [16a]. For the complete
oxidation to CO2 , again OHad is required
as in the above case of formic acid.
The existence of formic acid formation
on a second path has been proved by the
DEMS-technique (Sect. 5.2.3). A possible
step of formation could be
CHOHad + OHad HCOOH
+ H+ + e or
CHOHad + H2 O HCOOH
+ 2H+ + 2e
(20)
The rate of HCOOH formation is still

unknown. The DEMS technique shows
only the part of HCOOH that has been
transformed to mass 60 (HCOOCH3 ) via
the reaction with methanol in solution
(Sect. 5.2.3). As shown by several authors,
we have last not least a pathway via
formation of formaldehyde. The rate of
formation is not easy to be measured [16b].
Bifunctional Mechanism of
Methanol Oxidation
The kinetic of methanol oxidation [17] is
determined by the fact that the product
5.2.2.3
of Reaction (18), adsorbed CO, requires

a catalyst other than platinum, while on
the other side Pt atoms are necessary for
the dehydrogenation steps. The blocking
residue CO needs a surface with the
easy formation of an OH-bond out of the
surrounding water molecules, and this at
low potentials. A model catalyst for this
procedure is ruthenium (Sect. 5.2.3.3). At
room temperature, this metal is ineffective
as opposed to the rst steps in methanol
oxidation. Therefore, for this bifunctional
function of the reaction to the nal
products of CO2 and water, one needs
a surface arrangement of different metal
sites, for example, clusters containing
Pt and Ru atoms. The principle of the
bifunctional action is illustrated in Fig. 1.
The metal sites for adsorption and dehydrogenation should be platinum or even
better, a PtRu combination (see below); the
sites for OHad formation should preferably be ruthenium. OHad formation on
platinum needs a much higher anodic
potential. A special question arises: what
is the most suitable distribution of the
two different surface sites, requiring three
neighboring Pt sites close to at least one
Ru site, supposed to be? In the following paragraphs, we will learn that for a
smooth model surface we have to use an
alloy, the optimum ratio of Pt : Ru sites depending on operating conditions such as
temperature and methanol concentration.
5.2.3
Oxidation of CO, Formic Acid, and

Methanol at Pt Metals
Results via CV and MSCV
Online observation of mass signals from
volatile reaction partners offers, in addition to a currentpotential plot, useful information on the studied reaction
(DEMS technique, in Ref. [1820]). For
5.2.3.1
469
Methanol
O
C
H
Pt-elektrode
CO
P
STO
H+
e e
CO2
H+
H2 O
O C
H+
Fig. 1 Illustration of the different reaction steps during methanol oxidation on a

model catalyst surface, from left to right: three sites methanol adsorption,
dehydrogenation, adsorption of blocking CO intermediate, formation of OHad from
a water molecule, reaction with CO and desorption as CO2 .
Current density
[A cm2]
mass signals also, cyclic voltammograms

(CVs) can be resolved up to more than
100 mV s1 [21]. Besides the amount of
30
III
IV
II
15
information given in a rst survey, a continuous control of electrode reactions as

function of time is possible.
In the following text, we will characterize the oxidation of our small organic
molecules by CVs and simultaneously by
mass spectrometric cyclic voltammograms
(MSCV) as well as by the respective potential scans.
Carbon Monoxide Oxidation
The current/potential behavior of CO on a
porous Pt surface is demonstrated after CO
adsorption in a ow cell experiment [22]
via a potential scan in a solution free of
CO, starting at 50 mV versus RHE [23].
Figure 2 shows a rst low current region
(peak I) between 0.3 and 0.55 V, in the
upper part, well below oxide formation
on platinum. The following three peaks,
5.2.3.1.1
0
CO2
m /e = 44
III
II
IV
5.1012A
Mass signal
470
0.0
0.2
0.4
0.6
Potential vs. RHE

[V]
0.8
Fig. 2 Simultaneous recording of

current and mass signal e/m = 44 for
the oxidation of adsorbed CO during a
potential scan at Pt in 0.05 M HClO4 ,
Ead = 50 mV, tad = 10 min, 10 mV s1 .
Dached trace: supporting electrolyte
only [23].
5.2.3.1.2 Methanol Oxidation The mechanism of methanol oxidation with its

parallel pathway characteristic is much
more complicated than CO oxidation. We
therefore begin our study with simultaneous registration of CVs and MSCVs.
Figure 4 shows results for porous Pt electrodes in a 1 M CH3 OH sulfuric acid
solution. Comparing part (a) and part (b) of
the gure, it is clear that the products CO2
and also HCOOH are following the current over the total experimental region up
Fig. 3 Potentiodynamic proles for the
oxidation of CO adsorbate on
single-crystal Pt surfaces, obtained by
the ow cell technique as in Fig. 2, but
tad = 3 min [23].
to 1.4 V. Note the two orders of magnitude

difference in the mass signal intensity for
CO2 and methyl formate. Obviously, for
the oxidation of methanol also, platinum
shows a high overpotential of ca. 0.6 V
(compare with the thermodynamic data in
Sect. 5.2.2). We will discuss the formation
of current peaks in the CVs later, under
Sect. 5.2.3.2.3
Interesting new information can be obtained by taking a rst anodic scan, after
contacting the electrode with methanol
at 0.05 V, where adsorbed hydrogen still
inhibits methanol adsorption. Figure 5
shows the resolution of reaction steps
as a function of the potential [26, 27].
The increase in CO2 formation not before 400 mV demonstrates that the current
40
Pt(111)
20
Current density
[A cm2]
II to IV, are close together between

0.6 and 0.7 V. The relative contribution
of these peaks to the total charge is
strongly dependent on the structure of the
polycrystalline surface. The lower curve
with the mass signal of CO2 shows that in
all cases, a close relation exists between
the current and the mass signal. The
peaks therefore are related to COad and
the oxidation product is CO2 only. We
learn from the beginning of CO oxidation
at 0.3 V that at room temperature, the
overpotential for CO oxidation amounts
to more than 100 mV (Table 1).
In a second experiment, we turn from
a porous Pt surface to the well-dened
surface of a Pt(111) single crystal [23, 24].
The potential scans of Fig. 3 show that
the overpotential is the same as for the
polycrystalline surface, and peak I appears
in both cases again, but the multiplicity
of the large current peak is missing. The
main oxidation peak on Pt(100) is much
sharper and more positive than for Pt(111).
From this it follows that for a given
oxidation time, the CO coverage is always
smaller on Pt(111) than on Pt(100).
Pt(100)
150
100
50
0.2
0.6
1.0
Potential vs. RHE

[V]
471
j
[A cm2]
(a)
Fig. 4 Methanol oxidation at porous Pt

in 1 M CH3 OH and 0.5 M H2 SO4 .
(a) CV with 20 mV s1 ;
(b) simultaneously recorded mass
signal m/e = 44 (CO2 ); and (c) MSCV
for m/e = 60 (HCOOCH3 ) [25].
0.8
0.4
0.0
i ion 1012
[A cm2]
m /e = 44
CO2
In closing our rst view on methanol

oxidation in acid media, we should give
some results about the inuence of surface
structure on the electrocatalytic activity.
For this purpose, we have plotted, in
Fig. 6, the rst potential scans (in anodic
and cathodic direction) for methanol
oxidation at the three main single surfaces
of platinum [2629]. Pt(111) and Pt(110)
show the highest activity at low potentials.
A strong early dehydrogenation obviously
occurs on Pt(100). For recent details, see
T. Iwasita in Ref. [26, 27]. The discussion
of the current maximum on Pt(111) will
be done with the help of information from
FTIRS, mentioned below in Sect. 5.2.3.2.
(b)
m /e = 60
HCOOCH3
i ion 1014
[A cm2]
472
0
0.0
(c)
0.5
1.0
1.5
E vs. RHE
[V]
during the anodic peak between 200 and

350 mV is due to the oxidation of hydrogen atoms out of the dehydrogenation
step [10, 11]. The desorption of hydrogen
atoms that are already present on the surface, is involved also. The oxidation of
methanol up to CO2 during a rst potential
scan starts only above 450 mV.
5.2.3.1.3 Formic Acid Oxidation Cyclic

voltammograms taken at platinum in a
DEMS experiment [30, 31] show the CO2
signal following the current as in the
case of Methanol oxidation (Fig. 7). In the
potential range up to 1.5 V, we have again
three anodic current peaks, but the bulk
oxidation begins for formic acid at a much
lower potential. Studying the activity of
the single-crystal surfaces [29, 32, 33], one
nds strong activity for Pt(111) (Fig. 13),
and a very low one on Pt(110) and
Pt(100). Adzic and coworkers [33] have
demonstrated that the addition of the
three CVs combines to the diagram for
polycrystalline platinum.
The existence of a dual-pathway mechanism can easily be demonstrated by using
labeled H13 COOH in the following DEMS
experiment (Fig. 8) [30, 31]. After cycling
in the basic electrolyte, changing to 0.04 M
0.3
Current
[mA]
0.2
0.1
0.0
(a)
3.0
m /e = 44
MI 1011
[A]
2.5
2.0
1.5
1.0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
Potential vs. RHE

[V]
(b)
Fig. 5 First potential scan at a porous platinum layer after contact

with solution at +0.05 V, CV and MSCV as in Fig. 4, but 0.2 M
CH3 OH/0.1 M HClO4 , 10 mV s1 : (a) current; (b) mass signal for
CO2 , traced line: base signal without methanol [26, 27].
H13 COOH/0.5 M H2 SO4 solution under

potential control at 0.2 V for 3 min. It
follows exchange of the solution rst
against 0.5 M H2 SO4 and nally against
H12 COOH/0.5 M H2 SO4 . Then the potential scan is started in the anodic direction.
The mass signals for the oxidized adsorbate (m/e = 45) start only above 0.5 V,
before this happens HCOOH from the
bulk is oxidized to 12 CO2 (m/e = 44). The
direct pathway is active at the covered
surface. About the adsorbates, we now will
learn more from IR experiments in the
next paragraph.
5.2.3.2
Study of Adsorbates via FTIRS
5.2.3.2.1 Carbon Monoxide Carbon monoxide adsorbed at the interface electrode/electrolyte shows a strong surfacesensitive behavior [34]. We will show
this for Pt(111) as a model substrate.
Depending on the potential, CO can
be adsorbed in three different forms:
linearly bonded (atop), bridge bonded,
and threefold bonded. With the shift
in the strength of the CO bond, the
frequency (or alternatively, the wave number) are varying in a clear manner.
473

80
Pt(100)
Current
[A]
60
40
8
6
4
2
0
2
20
0.1
0.2
0.3
0.4
0
50
Pt(110)
40
Current
[A]
30
2
20
1
0
10
1
2
0.1
0.2
0.3
0.4
30
Current
[A]
474
Pt(111)
20
10
0
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Potential
[V]
Fig. 6 First potential scan after methanol adsorption at 0.05 V
as in Fig. 5, but single-crystal surfaces Pt(100), Pt(110), and
Pt(111), 50 mV s1 [26, 27].
In Fig. 9, this fact is shown for the

case of a CO-saturated 0.1 M HClO4
solution and two different potentials,
(a) 0.01 V and (b) 0.34 V versus RHE.
The reference spectrum used during the
in situ FTIR experiment was 0.75 V

RHE [35].
The positive going bands are due to the
loss of CO molecules, the negative going
band at 2345 cm1 results from the CO2
2.0
I / 103
[A]
1.5
1.0
0.5
0
0.5
0.5
1.0
1.5
E rhe
[V]
Ii /
[A]
1011
m /e = 44
3
0.5
1.0
1.5
E rhe
[V]
Fig. 7 Formic acid oxidation at porous Pt, CV, and MSCV as in
Fig. 4, but 12.5 mV s1 and 0.01 M HCOOH solution [30, 31].
formed during the potential step to 0.75 V

and trapped in the thin layer between
electrode and IR-window (Chapter 3.1 of
this volume). The thin-layer thickness was
minimized, such that at least 90% of the

CO2 produced was from electrooxidation
of the adsorbate and not from the solutionphase CO.
475

Fig. 8 CV and MSCV of formic acid
oxidation as in Fig. 7, but using
preadsorbed H13 COOH in order to form
a labeled adsorbate up to saturation
coverage, followed by oxidation of
H12 COOH at the covered surface. The
second MSCV below shows, in addition,
the rst cathodic and the second anodic
scan [30, 31].
10.0
I /104
[A]
7.5
5.0
2.5
E ad
0.0
0.5
1.0
m /e = 44
Ii
[A]
/1011
1.5
E rhe
[V]
2.5
E ad
m /e = 45
0.5
Ii
1.0
1.5
E rhe
[V]
m /e = 44
[A]
/1011
476
0.5
1.0
1.5
E rhe
[V]
Spectrum (A) shows a sharp CO band

at 2066 cm1 attributed to atop CO
and a much weaker band at 1773 cm1
for threefold adsorbed CO. At 0.34 V,
we have a drastic change from the
presence of threefold-bonded to bridgebonded CO at 1850 cm1 . By taking data
over a wider range of potentials, one

gets the change in band intensity as a
function of the potential for the species
studied [35].
Comparing the IR data from Fig. 10
with adlayer structures obtained from
scanning tunneling microscopy (STM),
.01
2073
1850
.005
(b)
Absorbance
[a.u.]
2066
1773
(a)
.005
.01
2400
2200
2000
1800
Wavenumbers
[cm1]
Fig. 9 Potential-difference infrared spectra in a CO-saturated 0.1 M
HClO4 solution at (a) 0.01 V and (b) 0.34 V versus RHE, reference was
made by stepping to 0.75 V RHE [35].
one understands that adlayer structure

and bonding properties reect each other.
Obviously, bridge- and threefold-bonded
CO requires the presence of a (2 2) or
( 19 19) structure.
Not only does the integrated band
intensity change with the potential, but
the frequency (or wave number) also shifts.
Figure 11 shows the potential dependence
of these two band parameters for atop
CO on polycrystalline ruthenium. In this
case, we follow the stripping of a saturated
adlayer formed at 0.3 V using a pure 0.1 M
HClO4 as electrolyte. Relative to platinum,

the wave numbers are shifted to lower
values. Oxidation to CO2 starts at Ru more
than 100 mV earlier than on Pt, a fact that
will be discussed in Sect. 5.2.3.3.
5.2.3.2.2 Formic Acid In situ IR spectra
taken during the rst anodic polarization
at the Pt(111) model electrode in HCOOH
solution give important new information.
After contacting the solution at 0.05 V, using a ow cell procedure [22], a series of
spectra was collected at xed potentials
477

0.25
CO2
Atop CO
Bridge CO
3-fold CO
0.20

[a.u.]
478
0.15
Diffuse
(2 2)
0.10
( 19 19)
0.05
0.00
0.3
0.2
0.1
0.0
0.1
0.2
0.3
Potential vs SCE
[V]
Fig. 10 Plot of integrated infrared band intensities for linearly, bridge- and
threefold-bonded CO adsorbed on Pt(111), together with spectra for CO2 ,
obtained along with stepping to the 0.75 V versus RHE reference; all data were
taken with CO-saturated 0.1 M HClO4 [35].
(Fig. 12). Atop CO (near 2050 cm1 ) and

bridge-bonded (at 1870 cm1 ) are formed.
Obviously both types of adsorbate are
formed with increasing intensities up to
a maximum of 0.3 to 0.45 V. It has been
established that on this surface no interconversion between bridged and atop
CO occurs with increasing positive potential. [37]. As shown in further studies [38],
the direct pathway of HCOOH oxidation
via dehydrogenation is the main pathway
on Pt(111).
The Anodic Maximum of Formic
Acid Oxidation below 0.5 V Small organic
molecules as formic acid, formaldehyde,
and methanol do show anodic maxima when the applied polarization is
reversed also. At potentials above 0.8 V
5.2.3.2.3
at polycrystalline Pt, the current decay

during the positive going scan was explained in terms of the formation of
platinum oxide [39] and the increase in
current during the scan in negative direction as due to HCOOH oxidation on
the clean surface after oxide reduction.
This interpretation, however, cannot explain decays at potentials near 0.5 V as,
for example, observed for formic acid
(Fig. 13). For methanol in sulfuric acid,
Markovic and Ross [40] state that sulfate
adsorption, which is a reversible process,
hinders the adsorption of organic species.
Herrero and coworkers [41] stated that
sulfate/water interactions [42, 43] are responsible for this effect. But the reversible
behavior is observed in H3 PO4 also [41],
and even in the presence of a much less
adsorbable electrolyte as HClO4 (Fig. 13).
Fig. 11 Potential dependence of band

parameters for spectra obtained during
CO stripping of a saturated adlayer,
formed at 0.3 V RHE on polycrystalline
Ru in 0.1 M HClO4 : (a) integrated band
intensities, normalized with the
maximum value obtained after complete
oxidation of CO at 0.8 V, the CO2 band
at 2341 cm1 was calculated with the
reference spectrum taken at 0.1 V where
no CO2 is formed; (b) CO stretch
wave numbers [36].

[a.u.]
CO
0.8
0.4
CO2
0.0
Wavenumber
[cm1]
2025
2010
1995
0.0
0.2
0.4
0.6
0.8
E vs RHE
[V]
6
Pt(111)
CO2
Band intensity
[a.u.]
4
CO ( 3)
2
= CO ( 3)
0
0.0
0.2
0.4
0.6
0.8
1.0
Potential
[V]
Fig. 12 Integrated IR band intensities of COL (atop), COB (bridge-bonded), and
CO2 during the rst anodic scan at Pt(111) in 0.1 M HCOOH/0.1 M HClO4 as
function of potential [38].
479
1.6
1.4
Current
[mA]
480
1.2
1.0
0.8
0.6
0.30
0.35
0.40
0.45
0.50
Potential vs RHE
[V]
Fig. 13 CV for a Pt(111) electrode in 0.1 M HCOOH/0.1 M
HClO4 [44, 45]. The potential is reversed at 0.5 V, sweep rate
50 mV s1 .
A rst satisfactory explanation of the symmetric current/potential response can be

given via a recent infrared spectroscopy
study [44, 45].
Potential-dependent bands for the OH
stretching and HOH bending mode of water have been analyzed and interpreted
in terms of the interactions of water
molecules with the electric eld of the double layer and with the metal surface [46].
In short, the shift of the bending mode in
the potential region of 0.4 to 0.9 V indicates O-adsorbed water molecules being
oriented from a tilted to perpendicular
position at the surface (Fig. 14). With increasing potential, water dipoles acquire
a conguration of minimum interaction
energy with the electric eld in the double

layer. In this position, participation of the
3a orbital of the H2 O molecule results in
an increased HOH-bond angle and, consequently, in a red shift of the bending
mode. Obviously, the interaction of water with the platinum surface seems to
be stronger than assumed so far. The interaction leads to a partial dissociation of
water, a phenomenon observed for several
adsorbed weak acids. Since the dissociation of water produces OH, the interaction
described can play an important role in the
electrocatalytic properties of the system.
It is interesting to observe that using
the Pt(111) surface, a symmetric current
response is present, even when the
1640
Bandcenter frequency
[cm1]
1630
1620
1610
1600
dHOH
1590
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Potential
[V]
Fig. 14 Potential dependence of the wave number for the in-plane
deformation (scissor mode) of water adsorbed at Pt(111) [44, 45].
potential scan is reversed already near

0.5 V. A similar behavior was reported by
Capon and Parsons [9] for polycrystalline
platinum. Obviously, the process for
the current decay depends reversibly on
the potential. This can be explained
assuming that a stronger H2 O interaction
is unfavorable for HCOOH adsorption.
In addition, the onset of current near
0.2 V, well below the beginning of CO
oxidation, proves that the direct reaction
pathway of HCOOH oxidation is active.
This reaction model is supported by the
fact that the CV of CO oxidation shows
also a maximum in the negative scan, if
the energy of adsorption is low as in the
case of silver [45], but not for the strongly
adsorbing metal platinum [44, 45].
5.2.3.2.4 Methanol The adsorption model presented above for HCOOH emphasizes that organic species have to
compete with water for an adsorption site,
that is, we assume an energy barrier for
the desorption of water comparable with

that of the studied organic molecules.
TDS data of gas/solid interfaces indicate a chemisorbed state of water at
170 to 180 K [47, 48]. Assuming a rstorder kinetic and a preexponential factor
of 1013 s1 , the bond strength of water
should be in the order of 42 kJ mol1 [49].
The peak temperatures for desorption at
Pt(111) are 190 K for CH3 OH, 170 K for
HCOOH, and 200 K for C2 H5 OH [47, 48].
Accordingly, the adsorption energy for the
undissociated molecules should be in the
same order as that of water. Calculated
from TDS data, the bond strength of CO
lies in the order of 108 kJ mol1 [50]. Assuming the same relationship in ultrahigh
vacuum (UHV) as at the electrochemical
interface, a competition with water for Pt
sites can be expected from small organic
molecules, but not from CO. And indeed, the absence of a reactivation peak
had been observed on voltammograms
at polycrystalline [51] and single-crystal
platinum [52] in CO-saturated solution.
481
0.25 V
0.30 V
Fig. 15 In situ FTIR spectra at

Pt(111) in 0.5 M CH3 OH/0.1 M
HClO4 at different potentials,
reference spectra at 0.1 V versus
RHE (256 scans, 8 cm1
resolution) [53].
0.35 V
H2O
0.40 V
0.50 V
1710
1826
1230
0.55 V
2059
482
0.60 V
HCOOCH3
COL
COB
R /R 0 = 0.01
CO2
2400
2000
1600
1200
Wavenumber
[cm1]
Figure 15 shows in situ FTIR spectra

during the oxidation of methanol on
Pt(111) in 0.5 M CH3 OH/0.1 M HClO4
at potentials between 0.25 and 0.60 V [53].
Besides CO2 production, formation of atop
CO and bridged CO, one has bands due to
HCOOCH3 formation near the interface
at 1710 and 1230 cm1 . The development
of the (integrated) band intensities with
electrode potential is given in Fig. 16. As
in the case of HCOOH oxidation (Fig. 12),
both types of CO adsorbates are formed
with a maximum near 0.5 V together with
an early CO2 production, obviously due to
a direct pathway of methanol oxidation.

Adding ruthenium clusters to the Pt(111)
surface (for details see the next paragraph),
the weak band of bridge-bonded CO is not
more observed and CO2 production starts
ca. 100 mV earlier.
Infrared spectroscopy is able to prove
the development of two more multibonded
adsorbates, the earlier-discussed COHad ,
and a CHx OHad species. These bands
have been discovered years later than the
different CO adsorbates because of their
relatively slow development with time [15].
The time-resolved formation is presented
3.2
0.16
Pt(111)
0.12
2.4
COL
0.08
1.6
COB
CO2
Band intensity
[a.u.]
0.04
0.8
0.00
3.2
Pt(111)/Ru(45%)
2.4
0.08
CO2
1.6
0.04
0.8
COL
0.00
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.0
0.8
Potential vs RHE
[V]
Integrated IR band intensities of COL (atop), COB
(bridge-bonded), and CO2 during methanol oxidation at Pt(111) and at
Ru(45%)/Pt(111) in 0.5 M CH3 OH/0.1 M HClO4 as function of
potential [53].
Fig. 16
in Fig. 17 for smooth polycrystalline Pt

from 1 M CH3 OH/0.1 M HClO4 solution.
Atop CO signals are given for comparison.
It is important to note that an adsorption
band for a species like HCO, which was

discussed also as an alternative [54], has
to be expected according to calculations
near 1900 cm1 . On the other hand, the
483
484

In situ FTIR spectra (1024
scans, 8 cm1 resolution) at smooth
polycrystalline platinum, time-resolved
formation of atop CO (ca. 2050 cm1 ),
threefold-bonded COH (1256 cm1 ),
and CHx OH (1200 cm1 ), adsorbed
from 1 M CH3 OH and 0.1 M HClO4 ,
adsorption potential 350 mV, reference
potential 50 mV versus RHE [15].
Fig. 17
t
[min]
5
10
20
30
40
R /R
2100
1900
1300
0.1%
1100
Wavenumber
[cm1]
experimental data for COH ts well with

the theoretical value [5558]. In addition,
the preliminary note, favoring HCO [54]
was found to be an experimental error. The
nature of the adsorbate was later clearly
shown to be COH, via a DEMS experiment
(Ref. [1820]) by using labeled isotopes [7].
Additional experiments with the same
electrolyte, but using single-crystal surfaces, show that for Pt(111) and Pt(100)
the 1260 cm1 COH band starts developing near 150 mV versus RHE [26, 27].
Metal Alloys and Binary Metal
Catalysts
It is now well established that Pt metal
combinations such as PtRe, PtSn, PtIr, and
5.2.3.3
especially PtRu show a drastic enhancement in catalytic activity for methanol

oxidation. The rst results have been published already more than 30 years ago,
see for example Ref. [59] on PtRu, [60] on
PdRu, [61, 62] on PtMo and PtSn, but especially [63] on PtRu, on Raney metals of
Pd, Ru, Pt, Rh, Ir, Os and their mixtures.
Different results have been obtained by the
different authors, but the positive effects of
ruthenium in combination with platinum
were never in doubt.
The mechanism of the ruthenium effect was rst described by Watanabe and
Motoo [17], postulating a bifunctional mechanism in which platinum serves as catalyst
for a dissociative methanol adsorption and
ruthenium (in the form of islands, clusters, or atoms) for the formation of a
Ru-OH species, being used nally for a
successive oxidation of methanol adsorbates (Fig. 1). It has to be noted that at
room temperature ruthenium does not
adsorb methanol [64], the rst step of adsorption takes place in the presence of Pt
sites only.
In the following, we discuss the PtRu
system as a model catalyst. Binary PtRu
electrocatalysts are presently studied in
many different forms, PtRu alloys [6568],
Ru electrodeposits on Pt [69, 70], PtRu
codeposits [7173], Ru adsorbed on Pt [74],

and Ru evaporated on Pt [75]. All these
materials present an enhanced activity for
methanol oxidation.
The determination of the surface composition of PtRu samples has been the matter
of several investigations. The simple use
of CVs has the handicap that features,
clearly associated with the Ru coverage
as in the case of hydrogen for platinum,
do not exist (Fig. 18). Watanabe and Motoo [17] suggested the formation of RuOH
in the potential range between 0.34 and
0.9 V of the CV. They dened the ratio
100
50
50
0
0
50
50
Current density
[A cm2]
(a) 0% Ru
(d) 63% Ru
100
50
50
50
50
(b) 11% Ru
(e) 13% Ru (H2red.)
50
50
50
50
(c) 39% Ru
0.0
0.2
0.4
0.6
0.8
(f) PtRu (90 : 10) alloy

0.0
0.2
0.4
0.6
0.8
1.0
Potential vs RHE
[V]
Fig. 18 CVs at 50 mV s1 for Pt(111), different Ru-modied Pt(111) surfaces and a
PtRu alloy in 0.1 M HClO4 , 25 C; (a) pt(111), (bd) using spontaneous Ru
adsorption, (e) Ru adsorption and simultaneous reduction by hydrogen [75].
485
between the charge for RuOH formation

and the charge for H atom desorption
as a measure for the degree of coverage
with Ru atoms. Frelink and coworkers [76]
assumed a two-electron process for RuO
formation. Since methanol oxidation is a
surface-sensitive process, IR spectroscopy
using single-crystal surfaces covered with
submonolayers of ruthenium is an obvious approach for a better understanding [2629].
For a rst evaluation of the PtRu surface
composition, we compare in Fig. 18 the
CVs of a clean Pt(111) surface with the CVs
of Ru-modied Pt(111) and a PtRu alloy.
For Pt(111), the characteristic features for
hydrogen adsorption/desorption and the

so-called anomalous states are observed.
As the surface becomes covered with
increasing amounts of Ru, the sharp peak
at 0.78 V decreases and the charge for the
formation of Ru oxides in the double-layer
region increases. The sample prepared
via Ru reduction with hydrogen (Fig. 18e)
exhibits a sharp peak at ca. 0.1 V. The
forms of the CVs of Fig. 18(bd) conrm
the lack of clear features indicating welldened surface processes that could be
used for coverage determinations. For Rumodied Pt(100), the existence of more
than one surface process has been clearly
observed [77].
30
0.50
15
0.25
0
15
0.00
Pt
Pt
50
25
0
25
PtRu(50 : 50)
Current density
[mA cm2]
Current density
[A cm2]
486
1.2
0.8
0.4
PtRu(50 : 50)
0.0
30
0.6
0
0.3
30
Ru
60
0.0
0.2
0.4
0.6
0.8
Potential vs RHE
[V]
Ru
0.0
1.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Potential vs RHE
[V]
Fig. 19 CVs for oxidative stripping of CO (left side) and oxidation from
CO-saturated solution (right side), 0.1 M HClO4 ; for stripping saturation of
surface with CO at 0.3 V; dashed: second scan coinciding the CV of a clean
surface in the supporting electrolyte, 50 mV s1 [36].
5.2.3.3.1 CO Oxidation For the characterization of electrode activity, usually

two different CVs are taken: a stripping
voltammogram (left side of Fig. 19) or a
voltammogram in CO-saturated solution
(right side of Fig. 19). As can be seen in
the gure, the optimum PtRu composition is very different in both cases. For
a stationary rate of CO oxidation from solution, pure ruthenium is by far the best
catalyst. When taking a stripping voltammogram, clearly a binary metal surface is
the better option. At room temperature,
one nds an optimum composition near
Pt : Ru = 1 : 1 [78] up to Pt : Ru = 1 : 2 [79],
depending on potential. This fact demonstrates that at a CO-free surface the overall
reaction, adsorption and oxidation, is best
catalyzed by a one-to-one addition of
platinum atoms to ruthenium. For the
oxidation of methanol, at least at room
temperature, this is not the case, probably
because of the fact that for adsorption of
the methanol molecule, three neighboring

Pt sites are needed.
5.2.3.3.2 Oxidation of Methanol The oxidation of methanol via a binary catalyst
will be discussed for PtRu combinations
as a model system. As known from early
investigations [59, 63] and already demonstrated here with the IR results of Fig. 16,
PtRu surfaces are strongly enhancing the
oxidation of methanol. In order to approach the behavior in the stationary case,
we analyze the activity mainly in the form
of i(t)-curves.
In Fig. 20, electrocatalytic activities at
room temperature are compared as function
of Ru percentage at the surface, for Rumodied Pt(111) and UHV prepared alloy
electrodes [75]. Disc-shaped PtRu alloys
(JohnsonMatthey) of 5 mm diameter and
2 mm thickness were polished to a mirror
nish and treated in UHV by Ar sputtering
and heating according to Ref. [80]. After
120+/12 A cm2
100
Plot of current densities for

methanol oxidation from amperometric
curves at 0.5 V after 300s and for
different PtRu surfaces [75]. ()
Ru/Pt(111) formed by spontaneous
adsorption, () Ru-modied Pt(111) via
Ru adsorption and simultaneous
reduction by hydrogen, () Ru coverage
by UHV vapor deposition, data obtained
after 20 min, () UHV-prepared PtRu
alloys (see text), () data of
UHV-prepared PtRu alloys from Ref. [67]
for comparison.
Fig. 20
Current density
[A cm2]
40
31+/2.4 A cm2
10
20
40
Ru%
60
80
100
487
488
application of a potential step from 0.05

to 0.5 V, the current was measured after
20 min [75]. For spontaneous adsorbed
Ru, a pronounced growth in activity
is observed with increasing coverage,
followed by a broad maximum near
30 A cm2 between 15 and 50% Ru.
Related to clean Pt(111) (see in gure),
the activity is increased by a factor of ca.
50. Ru deposits obtained via reduction of
Ru with hydrogen or via evaporation of Ru
present a three times lower activity than
spontaneous adsorbed Ru. Highest data
are obtained by the alloy samples. Values
measured by Gasteiger and coworkers [67]
are also plotted in the gure. A maximum
of activity of ca. 120 A cm2 is observed
between 10 and 40% Ru. Again three
times more activity is shown than using
spontaneous adsorbed ruthenium.
According to a model suggested for the
bifunctional mechanism, a surface structure having one Ru atom neighboring
three Pt sites represents the optimum geometry for methanol oxidation [81]. On
account of this, it was predicted that
activity/composition plots should have a
maximum near 10%. The experimental
results do not support such a prediction.
This is not surprising since an ordered surface structure as ideally proposed [81] does
not exist in real systems, even for ordered
structures like Pt(111). In this respect, it is
noteworthy that Pt deposited onto Ru(001)
under UHV conditions segregates into
clusters also [82]. In addition, the model
calculation giving a sharp peak near 10%
Ru [81] only holds if the dissociative adsorption of methanol is rate determining.
But the benet of a homogeneous distribution of Pt and Ru sites in contrast to
an island structure must be emphasized.
Thus, in Fig. 20, the alloys are present by a
factor of three higher currents than spontaneous adsorbed Ru/Pt(111) electrodes.
But so far, a direct proof for a homogeneous distribution of Pt and Ru in the

alloy does not exist.
The data of Fig. 20 show that for alloys,
under the experimental conditions of room
temperature and 0.5 M CH3 OH concentration, the rate of reaction is more or less
independent of surface composition in the
range between 10 and 40% Ru. Within this
range, neither the number of Pt sites (necessary for methanol dissociation) nor of
Ru sites (necessary for water dissociation)
can be regarded as rate-limiting factors.
Obviously, the kinetic limitation is caused
by the reaction between adsorbed CO and
RuOH. Therefore, a homogeneous distribution of Pt and Ru atoms must have a
strong inuence on the rate. Indeed, the
optimum distribution is that of alloys. For
samples forming Ru islands, the method
of preparation must be chosen in such
a way that the diameters of the islands
are as small as possible. In order to have
a rst approximation of cluster size, in
situ IR spectroscopy can be of help, as
shown below.
Cramm and coworkers studied the COIR signal on Ru-modied Pt(111) surfaces [83]. With increasing Ru coverages
from 25 to 60%, an increasing signal for
CO adsorbed at Ru sites was observed.
Near the 2070 cm1 band for atop CO on
Pt sites, a new band near 2010 cm1 clearly
develops. The inuence of the preparation procedure on the CO-Ru adsorption
band is shown in Fig. 21. From the two
electrodes with adsorbed Ru-clusters, only
for ruthenium adsorbed via hydrogen a
band at 1965 cm1 is clearly observed.
The difference in signal should be attributed to the formation of different size
Ru islands, depending on the preparation
method. Larger islands seem to be formed
by the H2 -reduction procedure. This justies the relatively low activity of the latter
Fig. 21 Comparison of CO bands in
FTIR spectra for methanol on different
samples of Fig. 20: Pt(111), Pt(111) with
spontaneously adsorbed Ru(39%), with
Ru reduced by hydrogen and PtRu alloy,
15% of Ru each [75].
Pt(111)
1834
Pt(111)/Ru(ads)
2057
Pt(111)/Ru(H2)-red.
2060
1965
PtRu alloy
R /R0 = 2 103
2048
2250
2000
1750
Wavenumber
[cm1]
sample. Obviously, surface structure and

Pt/Ru distribution determine the electrocatalytic activity.
The behavior of the PtRu system
for methanol oxidation changes strongly
with increasing temperature. At 40 to
60 C, ruthenium can adsorb and oxidize
methanol also. Therefore, less Pt sites
are necessary for the steps of adsorption
and dehydration. Results on PtRu surfaces
show indeed a shift of the optimum PtRu
ratio to higher Ru contents near 50% [84].
For porous PtRu clusters on carbon black
at 60 C, Watanabe and Motoo [17] have
shown an optimum PtRu ratio near 50%.
Beside the successful system PtRu,
many other platinum metal combinations
have been checked as catalyst for methanol

oxidation, for example, PtPd, PtAu, PtIr,
and PtOs [85], or even multimetal systems like PtRuOs and PtRuOsIr [86]. The
current/potential plots do not markedly
differ and are therefore difcult to analyze. In addition, a current normalization
via mgPt cm2 or mgMetal cm2 or
via BET was performed. A more realistic determination of the active surface
has been suggested using oxidative stripping of CO [73]. Investigations on smooth
PtSn alloys [81, 87] have been negative in
spite of the fact that a number of publications show increased activities for Sn,
electrodeposited on platinum (for a survey,
see [88]).
489
Binary catalysts other than PtRu show

increased activity for the oxidation of CO
and HCOOH also. For CO, the alloy
Pt3 Sn alloy was studied in detail and with
interesting results [81, 87].
Ad-atom Effects on Formic Acid
Oxidation Submonolayers of some metal
atoms, irreversible adsorbed or underpotentially deposited (upd) [89] on a
metal substrate, can present electrocatalytic properties different from those of the
pure metals. They may be similar to those
of alloys or codeposited metals [90]. Typical examples are submonolayers of As, Sn,
Bi, Tl, and Pb on platinum [9193]. Since
a upd layer can exist within a given potential range only, this modied catalyst
may not fulll the conditions for technical
application.
A noticeable increase in activity on
formic acid oxidation was already observed 30 years ago [94], at polycrystalline
platinum in the presence of 105 M Pb
5.2.3.3.3
ions in solution. These ions form surface

ad-atoms at potentials of interest between
ca. 100 and 500 mV RHE. Figure 22 shows
CVs obtained at smooth Pt, with and
without the addition of Pb2+ ions to the
electrolyte. The current is increased by two
orders of magnitude. As already discussed
in Sect. 5.2.2.2, the Pb ad-atoms catalyze
the direct pathway of formic acid oxidation. To our knowledge, the PtPb system
is the only upd catalyst with successful stationary data. At 10 mA cm2 and at room
temperature, a cathodic shift of potential
by 200 mV was observed for hours [95].
Ad-atom electrocatalysis in the case of
formic acid oxidation can be divided into
two terms: (1) inhibition of surface poisoning by CO formation (indirect pathway),
and (2) true enhancement of the rate of
oxidation (direct pathway). For most adatoms, these effects are mixed. An example
for the mixed case is the Sb ad-atom adsorption on Pt(100) [96]. The reaction takes
place on single Pt sites and randomly
12
j
[mA cm2]
490
0
0.0
0.5
1.0
1.5
2.0
E (RHE)
[V]
Fig. 22 Cyclic Voltammograms for formic acid oxidation at smooth
polycrystalline platinum, without addition (dotted lines) and with lead ions
in the electrolyte (full lines); 1 M HCOOH/1 M H2 SO4 , 105 M Pb2+ ,
50 mV s1 [94].
30
1.0
I/Imax
Ip
[mA cm2]
0.8
20
0.6
0.4
10
0.2
0.0
0.0
0.2 0.4 0.6 0.8
Sb
(a)
0.2
0.4
0.6
0.8
1.0
0.8
1.0
Adatom
(b)
1.0
40
30
I/Imax
I
[mA cm2]
0.8
20
0.4
10
0.2
0.0
0
0.5
(c)
0.6
1.0
Bi
0.0
(d)
0.2
0.4
0.6
Adatom
Fig. 23 Comparison of experimental and

simulated current/adatom plots for the
oxidation of formic acid. (a) Sb-modied Pt(100)
surface [96]; (b) simulated curve for a pure
third-body effect (current proportional to the
number of isolated Pt sites) [97]; (c) Bi-modied

Pt(111) surface [101]; (d) simulated curve for a
true catalytic effect (current proportional to
number of ad-atom-free Pt sites [97].
distributed ad-atoms may block neighboring places, making some isolated sites
available for the direct oxidation. In addition, the rate of reaction per free Pt site
is enhanced. In Fig. 23, we have in (a, b)
a comparison between the experimental
current/Sb plot and a simulation under the conditions of randomly distributed
ad-atoms and a reaction rate proportional
to the number of isolated Pt sites [97]. All

other places are occupied by ad-atoms or by
poison. This effect had been called thirdbody effect [98]. On well-dened surfaces,
some ad-atoms form geometric arrangements. Thus, Se atoms can form a (2 2)
layer on Pt(111), exhibiting isolate Pt sites,
surrounded by Se atoms occupying threefold places [99]. In such an arrangement,
491
492
the third-body effect is particularly effective

(ensemble effect). A true enhancement effect
with a current density of 3.2 mA cm2 (at
the poison-free surface) and 40 mA cm2
at a Bi = 0.8 was observed for Bi on
Pt(111) [96]. This represents an increase in
rate per free Pt site by a factor of ca. 70. In
addition, Bi coverages of only 10% totally
inhibit the formation of CO [100]. The rate
of reaction depends, for this system, on
the number of ad-atom/free site pairs. In
this case, in Fig. 23(c, d), the experimental plot [101] is again compared with the
simulated curve [97].
5.2.4
Surface Structure and Methanol

Electrocatalysis
From the discussions in the above chapters, we learned already that surface
structures are of special importance for
methanol oxidation as a result of the required bifunctional catalyst properties. A
rst step to use STM images to gain a
better insight into this matter was done
by Cramm and coworkers [83]. Electrodeposits of Ru islands on Pt(111) in 0.1 M
HClO4 were characterized by STM images. Islands of monoatomic height and
between 2 to 5 nm of diameter are shown
for two different Ru coverages. More recent
results will be presented in the following paragraph.
Ex Situ STM Images Taken during
Methanol Oxidation via UHV Transfer
In order to follow the morphology of a
Ru/Pt(111) surface during methanol oxidation, the following modication of an
Omicron standard preparation and analysis chamber was made [102]. The manipulator was extended by a platinum shield
to avoid foreign metal deposition. The external load-lock chamber was modied to
5.2.4.1
allow insertion of a mini electrochemical

cell. With these additions, a UHV transfer
from the electrochemical cell to the STM
position and back was possible.
An example of the results obtained is
given in Fig. 24 (a, c), which show the morphology shortly after preparation of the
electrode. After UHV transfer to the electrochemical cell, methanol was oxidized
at this surface at 500 mV for 50 min. It
followed cleaning with water and UHV
transfer back to the STM. From Fig. 24
(b, d), it can be concluded that almost no
change in morphology took place.
In a second experiment, the activity of
different PtRu model surfaces is characterized via current/time curves after
a potential step from 300 to 500 mV
(Fig. 25) [103]. Because of the pronounced
differences in activity, the plots are given in
a half logarithmic scale. Two layer islands
(A) and smooth Ru surface alloys (B) show
strong decay and lower currents than Ru
monolayers on smooth Pt(111) (C). Ion
bombardment before (D) or after (E) Ru
island formation increases the activity. But
highest current densities were obtained for
the PtRu(85 : 15) electrode (F) with the by
far nest distribution of Ru and Pt atoms
in the surface.
All currenttime plots in Fig. 25 show
a marked decay with time. This phenomenon will be discussed in the
next paragraph.
Current Decay at Smooth Surfaces
When comparing the catalytic properties
of different smooth surfaces, it is essential
to have a suitable measure or denition
for the catalytic activity itself. For this
purpose, currenttime curves at potentials
between, 0.35 V and 0.5 V versus RHE,
had been used [67, 104]. As a common
feature for all model catalysts, the resulting
i(t)-curves show a pronounced decay at
5.2.4.2
50 nm
50 nm
(b)
(a)
20 nm
20 nm
(c)
(d)
Morphology of a Ru/Pt(111) electrode taken via STM before and after 50 min
methanol oxidation at 500 mV in 0.5 M CH3 OH/0.1 M HClO4 . Potential step from 50 to
500 mV RHE [103].
Fig. 24
the beginning, which becomes less steep

after ca. 10 min (Figs. 25 and 26). All
smooth probes present a current decay
that was also observed for spontaneously
adsorbed Ru at Pt(111) [74, 75]. In most
publications, the current density after
10 to 30 min has been regarded as a
good measure for the catalytic activity
of the respective surface. However, in
all cases, it is clearly visible that at this
time a steady state was yet not reached,
this way of dening the catalytic activity
appearing thus unsatisfactory. Same decay
curves as in Fig. 26 have been reported
for all three single-crystal Pt surfaces
and potentials below 550 mV [74, 105].
Already at 600 mV, Stuve and coworkers
report about a stationary current at Pt
(111), 10s after potential step from 55 to
600 mV [106].
Three main points should be discussed

in connection with the above observation.
First, there is a pronounced decay of the
current occurring at the beginning of the
experiment. The current decreases by a factor of about two during the rst 10 min of
application of the potential step. A current
limitation by mass transport is unlikely for
CH3 OH concentrations of 0.5 M; current
densities in the 100 mA cm2 range could
be achieved from that point of view.
A second important aspect of the
currenttime response is the fact that the
current does not look like it is going to
reach some steady state value even after
30 minutes. In fact, the decay is observed
over hours [104].
The third and probably most interesting feature is that in many cases the
currenttime curves measured on freshly
493

Alloy
1000
Current density
[A cm2]
F
100
E
D
C
10
Structure of PtRu model

electrodes and catalytic activity for
methanol oxidation after a potential step
from 300 to 500 mV RHE; current/time
curves under conditions as in
Fig. 24 [103]; (A) Two layer Ru island
formation on Pt(111) by Ru evaporation
at 400 K; (B) Smooth Ru/Pt(111) surface
alloy; (C) Ru evaporated on smooth
Pt(111); (D) Ru evaporated on a rough
Pt(111) surface; (E) Ru evaporated on
Pt(111) and roughened via ion
bombardment; (F) Sputtered PtRu alloy
(85 : 15), STM image not possible.
Fig. 25
B
A
0
12
18
Time
[min]
40
30
20
Pt(111)
10
0
300
Current density
[A cm2]
494
Pt : Ru
50 : 50
200
100
0
400
Pt : Ru
75 : 25
200
0
400
Pt : Ru
85 : 15
200
0
10
Time
[min]
15
20
Fig. 26 Currenttime curves after

potential step from 300 to 500 mV RHE
for different smooth catalysts as
indicated; 0.5 M CH3 OH/0.1 M HClO4 ;
room temperature [104].
0.020
Current
[A]
0.015
0.010
0.005
0.000
200
400
600
800
1000
1200
1400
Time
[s]
Fig. 27 Current/time curve for a porous PtRu electrodeposited (81 : 19) on
a gold substrate. The steps at ca. 4 min and 14 min are caused by the
production of CO2 bubbles [104].
prepared rough electrode (also on technical

fuel cell electrodes) exhibit no decay
(Fig. 27). The electrode for this experiment was prepared on a gold substrate
(c Electrodeposit from 20 mM H2 PtCl6 +
2 mM RuCl3 + 0.1 M HCl; E = 100 mV
versus RHE : t = 15 min.) and according
to energy dispersive X-ray spectroscopy
(EDXS) the catalyst composition was
Pt : Ru = 81 : 19 [104]. The constant current response observed here is comparable
to that reported by Chu and Gilman [107]
for methanol oxidation at porous PtRu
alloys as well as by Aramata and Masuda [108] for porous PtRuAu alloys and
50 C. In agreement with this behavior,
Motoo and coworkers [109] measured a
constant cell voltage during polarization
of a methanol-zinc cell at 125 mA cm2
and 40 C. However, this behavior of
rough electrodes cannot be generalized
and, depending on experimental variables
during their preparation, technical PtRu

electrodes may present a slow loss of
activity. Thus, the main difference between
both types of materials is that smooth electrodes always present such a current decay.
The difference in behavior of both types of
materials is probably related to the microstructure of the surface.
Rough surfaces, as those of technical
electrodes, are characterized by the presence of a large amount of defects (steps,
kinks, etc.) and it is well known that such
defects exhibit enhanced catalytic activity for the oxidation of organic residues.
Recently, Souza and coworkers [110] observed that electrodeposited PtRu electrodes do present a higher capacity of
breaking the CC bond and producing
CO2 than do smooth materials of the same
composition. In this context, it could be
possible that hydrogenated residues easily undergo oxidation to CO2 on defect
495

0.75
Potential vs RHE
[V]
(a) Applied potential program

0.50
0.25
0.00
250
(b) Current response
Current density
[A cm2]
496
200
150
100
10
20
30
40
50
60
70
80
90 100 110
Time
[min]
Fig. 28 Current/time curves for a smooth PtRu(85 : 15) alloy as in
Fig. 26, but with load interruptions by turning back the potential to its
initial value [104].
sites at rough electrodes while smooth

surfaces present a lower activity towards
this process.
For a better understanding of the rst
strong decay on smooth electrodes, the
potential was repeatedly changed between
75 and 500 mV [104]. After resetting the
potential to 500 mV (Fig. 28), the current
is higher than at the end of the preceding
wave, but the initial value is not recovered.
At the beginning of each cycle, we observe
an initial fast decay. The deactivation of
the smooth PtRu surface has at least
two components a partially reversible
one and a second one causing a slow
loss in activity that cannot be reversed at
low potentials.
The initial decay, being a reversible

process, could be related to the formation
of Ru oxides such as RuO2 or RuO3 , their
relative ratio depending on potential [111].
These oxides, having Ru in a high valence
state, must have a covalent character and
it is doubtful as to whether or to what
extent they can act as oxygen donors.
However, the presence of oxides at low
potentials is an indication of the propensity
of Ru to adsorb and dissociate water. Thus,
adsorbed OH could be present at the Ru
surface, being the oxygen donor for the
oxidation of organic compounds. X-Ray
Photoelectron Spectroscopy (XPS) data for
electrodes immersed at 300 and 50 mV
versus RHE support the assumption
that oxidation of the Ru component is

a quasi-reversible process, undergoing
composition changes in the timescale of
minutes [112].
The slow deactivation of smooth surfaces
could be due to a further accumulation
of carbon monoxide blocking active sites.
But IR data show that a Pt(111) surface becomes saturated with CO after 2 min, applying a potential step from 50 to 500 mV.
On Ru-modied Pt(111), the formation of
adsorbed CO is even faster [104]. On the
other hand, at sufcient anodic potentials,
for example above 1 V RHE, adsorbed poisoning particles obviously can be oxidized.
This fact suggests that Cn Hm compounds,
possibly formed during methanol adsorption [56, 110], could be responsible for the
slow deactivation.
The slow decay follows the relation
1
1
= + bt
j (t)
j
(21)
with j = j (t = 0) and b as a characteristic

constant. The current must be proportional to the number of active surface sites,
N (t):
j (t) = k N (t)
(22)
Eq. (21) suggests a second-order kinetic on
N (t), which slowly causes the blockage of
the surface
2
dN (t) = kN (t)
(23)
Plotting the reciprocal current versus the

time, one observes that for t > ca. 10 min
all curves for smooth alloys follow a
function in the described manner [104].
Interestingly, literature results measured
at 60 C follow the same behavior [67].
Using the gradient of the reciprocal
current, b, as a parameter characterizing
the deactivation, one observes that the
effective activity is, in all cases, reciprocal
to the value of b [102, 104].
Rate-determining Step of Methanol

Oxidation
Coming back to the well-analyzed intermediates of methanol oxidation, one
has to consider the following scheme for
the mechanism:
5.2.4.3
Path one
(1) CH3 OH
CO + 4H+ + 4e
(24)
+
(2) Me-H2 O Me-OH + H + e

(25)
(3) CO + Me-OH CO2 + H+ + e
(26)
Path two
(4) CH3 OH + H2 O
HCOOH + 4H+ + 4e
(27)
(5) HCOOH + CH3 OH

HCOOCH3 + H2 O
(28)
(6) HCOOCH3 + 2H2 O

2CO2 + 8H+ + 8e
(29)
Path three
(7) CH3 OH
HCHO + 2H+ + 2e
(30)
(8) HCHO + H2 O
CO2 + 4H+ + 4e
(31)
Path four
(9) CH3 OH
COH + 3H+ + 3e
(32)
(10) COH + Me-OH

CO2 + 2H+ + 2e
(33)
Path four is observed at low concentrations

of methanol only (<0.1 M), and path two
will cover not much more than 10% of
the total process. Therefore, for the discussion of the rate-determining step, we
497
have to consider the different issues of

path one. If Step (24) is rate determining, adsorption and/or dissociation of the
methanol molecule should be slow, that is,
the current should depend on methanol
concentration. Otherwise the reactive desorption of the surface-blocking CO is the
slowest part of the reaction. In this case, the
current must be independent of methanol
concentration; Step, (25 or 26) being the
slowest process.
Literature data show that depending
on the experimental conditions, both
phenomena can be observed. Chu and
Gilman [107] did nd for a PtRu(48 : 52)
alloy at 25 and 60 C, and Schmidt and
coworkers [113] for a porous PtRu(52 : 48)
surface at 25 C, an increase in current by
a factor of about 1.3, changing methanol
concentration from 0.5 M to 2.0 M. For a
porous PtRuAu alloy (Pt : Ru = 50 : 50) at
50 C, Aramata and Masuda [108] report
a much stronger dependence on concentration between 0.05 and 0.5 M CH3 OH.
Recently, ValdecirPaganin and coworkers [114] did study the dependence on
methanol concentration for ve different porous catalysts and with ve different methanol concentrations between
0.1 and 3.0 M. Interestingly, for an ETEK PtRu(50 : 50) catalyst, again a strong
inuence of methanol concentration was
found. For PtRu(84 : 16) and PtRu(75 : 25),
a higher activity was observed, being
almost constant between 0.5 and 3 M
(Fig. 29). The last data are in good agreement with the plots in Fig. 20. Obviously,
in the high activity PtRu region and for
25 C, the rate-determining reaction is the
oxidation of adsorbed CO (Steps 2 and 3
in our scheme for the mechanism) while
outside this region (e.g. for the catalysts
PtRu(50 : 50)) the available Pt sites are not
enough for a sufcient rate of adsorption
of methanol. For Pt(90 : 10) and of course
for pure platinum, the number of Pt sites
is high, but the bifunctional mechanism is
hindered or even not possible at all.
For a methanol concentration smaller
than ca. 0.3 M, we obviously have in
all cases the effect of a to low rate of
methanol adsorption. In the region of
0.012
Pt75Ru25/C
0.010
Pt84Ru16/C
inorm.
[A cm2]
498
0.008
Pt50Ru50/C-(E-TEK)
0.006
Pt92Ru10/C
0.004
Pt/C-(E-TEK)
0.002
0.5 V vs. RHE
0.000
0.0
0.5
1.0
1.5
2.0
2.5
CH3OH conc.
[mol L1]
Fig. 29 Current density, taken 30 min after applying the load, as a
function of methanol concentration for different porous catalysts at
25 C [114].
3.0
N-shaped currentpotential
curves i() ( potential of the working
electrode) and load lines (U )/RE ; U
applied voltage, RE ohmic resistance:
(a) single stationary case; (b) three
stationary states (bistability); (c) plot of
current i against the external voltage U.
Fig. 30
(a)
(U j)/RE
Potential j
(U j)/RE
General Conditions for

Electrochemical Instabilities
Common for the three molecules CO,
HCOOH, and CH3 OH is the occurrence
of maxima in the CV even at concentrations above 1 M in solution, that is,
without mass-transfer limitations (Figs. 6
and 7). According to the early study
of Bonhoeffer [115] and Franck [116], a
negative slope in the steady state current/potential plot (decrease of current
with increase of overvoltage, di/d < 0,
that is, a negative impedance Z of the
interface, Fig. 30) is one physical condition for the phenomenon of oscillations
in the case of an applied constant current.
Using a potentiostatic circuit, the additional condition/z/ < RE has to be met, in
order to observe bistability or oscillations
(see below).
For a better understanding, we study
the equivalent circuit of Fig. 31. The
output of a potentiostat delivers the
voltage U , applied between the working
5.2.5.1
i (j)
Instabilities Oscillations during the

Oxidation of Formic Acid, Methanol, and
Carbon Monoxide
i (j)
5.2.5
electrode and the reference electrode.

The working electrode is represented by
W and faradaic
double-layer capacity CD
W
impedance ZF (). Finally the ohmic
resistance RE is the electrolyte resistance
between the working electrode and the
Luggin capillary of the reference electrode.
The resistance of the counterelectrode
and the resistance between the Luggin
capillary and the counterelectrode does not
have to be considered in a potentiostatic
circuit. The applied voltage U is composed
of the voltage drop over RE , iRE , and
of the potential difference between
the working and reference electrode, the
electrode potential, U = iRE + .
(b)
Potential j
i (j)
highest catalytic activity and between 0.5

and 3.0 M of methanol concentration, no
change in current is observed, that is,
step one is not rate-determining. The rate
seems to be proportional to expressions
such as Pt-CO/Ru-CO and Ru-OH/Pt-OH.
Changing PtRu composition between ca.
15 and 40% Ru, both expressions may
compensate each other.
Voltage U
(c)
499
500

Equivalent circuit for the
connection of a potentiostat to a
working and reference electrode of an
electrochemical cell, with double-layer
i
capacities CDW , CDR , and faradaic
impedances ZFW , ZFR of a working and
i = 0 reference electrode, ohmic resistance R
E
(in most cases electrolyte resistance
between the working electrode and end of
R
C D the Luggin capillary of the reference
electrode, and possible external
resistance in connection with the
working electrode), U applied voltage,
i = iC + iF .
Fig. 31
iC
W
CD
RE
ZF
iF
ZF
The current through the interface electrode/electrolyte has two pathways, iF and
iC . For the respective differential equation follows:
i=
(U )
= iC + iF
RE
= ACD
d
+ iF ()
dt
(34)
with A as the electrode surface. For a

stationary state, with iC = 0, we may have
the electrode potential SS , and Eq. (34)
reduces to
(U SS )
i = iF (SS ) =
RE
(35)
Two of the possible load lines (U )/RE

against are plotted for different values
of U and RE in Fig. 30(a, b). The load
line in Fig. 30(a) shows one intersection
with the N -shaped i() curve, that is,
one steady state. In Fig. 30(b), the two
outer intersections are necessarily stable,
being on branches with positive slopes. A
perturbation of the middle state leads to
one of the branches with positive slopes,
and the system is therefore called bistable.
Finally, Fig. 30(c) shows the plot of the
current against the applied voltage U ,
indicating the inuence of the voltage
drop iRE .
The stability of the potentiostatic circuit

is checked by applying a small potential
perturbation :
iF () = iF (SS + )

diF
(36)
= iF (SS ) +
d =SS
where (diF /d) is by denition equal
to Z 1 . Introducing Eqs. (36 and 34),
it follows:
i = iC + iF
= ACD
i =
d
diF
+
and
dt
d
(U )
=
, or
RE
RE
= ACD
+ ZF 1 , or
RE
dt
d()
= (ACD )1 (ZF 1 + RE 1 ), or
dt
= (t = 0) exp{(ACD )1
(ZF 1 + RE 1 )t}
(37)
Equation (37) shows the time dependence

of the perturbation. It follows that SS
becomes stable as long as ZF 1 has positive
values, that is, as long as
ZF > 0
(38)
For ZF > 0, one has always a function

exp(-at).
On the other hand, the system becomes
unstable as soon as the exponent is positive,
that is, for
ZF < 0 and |ZF | < RE
(39)
In the case that the current is held

constant (galvanostatic circuit, which can be
approximated with a potentiostatic circuit
by introducing a large ohmic resistance
RE ) and the voltage U is variable, the
condition for instability reduces to Eq. (38),
because the second part of the conditions
in Eq. (39) is met by denition.
For the evaluation of possible perturbations, that is, of the presence of negative
faradaic impedances ZF , we now follow
the approach of Koper [117] and introduce
the ButlerVolmer expression for our currentpotential curve
iF () = nFA()cs ()k()
(40)
with n number of electrons, A available

electrode area, cs concentration at the
interface, and k() the electrochemical
rate constant.
From Eq. (40), it follows for the
faradaic impedance

diF
dk()
= nF Acs
ZF 1 =
d
d

dcs
dA
+ Ak()
+ cs k()
(41)
d
d
We may have three causes for an N shaped currentpotential curve, which
possesses a negative impedance in a
limited potential region, that is, ZF 1
becoming negative values:
1. The rate constant k() decreases with
if either a species that catalyses the
reaction at low potentials is desorbing
at higher overpotentials or a substance

that hinders the reaction adsorbs more
strongly at higher potentials, as for
example, in the case of water molecules.
2. A coulombic repulsion between electroactive species at the electrode, that
is, dcs /d < 0.
3. A decrease in the active surface A, for
example, by adsorption of a species that
completely inhibits the reaction, or by
formation of a passivating layer.
Especially, the last point has to be
considered in our investigation of small
molecules producing strongly adsorbed CO.
Comparing Fig. 30 with Fig. 4, we can
assume that in the potential region between 1.0 and 1.3 V, methanol oxidation
could be an example for a negative differential resistance (NDR) system. In addition,
from Fig. 30, it can be concluded that
N -shaped systems with a negative differential resistance can oscillate at constant
potential (current oscillations), but not at
constant current. As we see below under
Sect. 5.2.5.3 and Fig. 35, indeed, current
oscillations are to be observed. The surface
properties are modied by an electrochemical reaction alone, by the formation of
platinum oxide. This Pt-O formation is followed by a chemical reaction between the
oxide and methanol in solution. CO2 is produced, and the current at the free surface
sites increases to form new surface oxide, until the surface coverage approaches
its maximum and the chemical reaction
becomes stronger again. And the cycle
starts anew. A characteristic of the (simple)
NDR oscillators is that only one potentialdependent process is involved, that is, the
electrochemical oxide formation.
An electrochemical oscillator works under galvanostatic conditions (potential oscillations) if the mechanism involves two
501
502
different potential-dependent processes.

In general, potential oscillations appear
in the region of the positive part of the
steady state curve. A simple example of
the phenomenon are the potential oscillations to be observed during the oxidation
of hydrogen at platinum in acid solution,
using mixtures of a large amount of
hydrogen with some percent of carbon
monoxide [118]. After the application of a
constant current, the potential of hydrogen oxidation increases from values below
100 mV RHE up to ca. 800 mV, because of
the poisoning of Pt sites, more and more
with adsorbed CO. The oxidation of CO,
starting slowly above 600 mV, becomes really effective near the maximum of the
oscillations. At the free platinum sites,
hydrogen is adsorbed, and the potential
drops down steeply versus the reversible
hydrogen potential again, a new cycle
starts. The potential-depending rates of
adsorption and desorption of the blocking
adsorbate CO are next to hydrogen oxidation the second potential-dependent
process. But, where is the point of negative resistance the negative slope of one
of these two processes? It is the wellknown branch of the steady state curve
of hydrogen oxidation above ca. 800 mV
RHE. As already discussed in Sect. 5.2.3.2,
in the case of formic acid (Fig. 13), a
competition for free surface sites between hydrogen and adsorbed water takes
place. Koper and Sluyters have shown that
for such instabilities, an inductive loop
and/or a negative faradaic impedance in a
Niquist plot are characteristic (Fig. 36) and
called hidden negative differential resistance
(HNDR) [119, 120]. The mathematical description of an HNDR oscillator has been
given, for instance, by Krischer [121125]
after studying, as a model system, hydrogen oxidation at a rotating Pt disc in 0.5 M
H2 SO4 /102 M Cl with the addition of
small amounts of metal ions [121123].

Here, we have the HNDR in form of decreased hydrogen oxidation via potentialdependent Cl ion adsorption.
Potential Oscillations during
Formic Acid Oxidation
The explanation for the instability at the
electrochemical interface during formic
acid oxidation is very similar to that of the
above-mentioned H2 /CO model. Coming
back to the dual-path process of formic acid
oxidation (Eqs. 1214), the direct oxidation
to CO2 is equivalent to hydrogen oxidation,
and the pathway via CO formation replaces
the CO adsorption from the gas phase. The
origin of an NDR is the competition for adsorption at free sites between HCOOH and
water the bond strength water-metal becoming stronger with increasing potential
(compare Fig. 13). The pathway via dehydration (Step 13) takes care that the NDR
is hidden in a certain potential region.
Potential oscillations during formic acid
oxidation had been observed already in
the 1920s [126]. As Fig. 32 shows [127], the
upper potential limit is between 800 and
900 mV, as in the case of H2 /CO oscillations. Introducing a model simulation,
only the four (or ve) rate constants of
Eqs. (1214) have to be employed. This has
been done quite successfully by Okamoto
and coworkers [128], reproducing the ngerprints of their experimental results.
The surface coverages of CO and H2 O
being of importance, Okamoto applied the
following equations [128]:
5.2.5.2
dCO
= k2 (1 CO H2 O )
dt
k4 (CO H2 O )
(42)
dH2 O
= k3 CO (1 CO H2 O )
dt
k3 H2 O k4 (CO H2 O )
(43)
0.8
0.4
0.0
10
12
14
70
72
74
76
78
80
Potential
[V]
0.8
0.4
0.0
0.8
0.4
148
150
152
154
156
Time
[min]
Potential
[V]
1.0
0.5
0.0
0
10
15
20
Time
[sec]
Fig. 32 Potential oscillations during formic acid oxidation at a smooth polycrystalline
platinum electrode: (a) 5 M HCOOH/0.5 M H2 SO4 , applied current density of
1.5 mA cm2 ; (b) 1 M HCOOH and 0.5 M H2 SO4 , 0.25 mA cm2 [127].
1
dE
=
[I 2Fh{k1 (1 CO H2 O )
dt
Cd
+ k4 (CO H2 O )}]
(44)
with k1 to k4 rate constants, including the

concentrations of formic acid and water,
k3 the rate constant of the backward
reaction, I the applied current, Cd the

double-layer capacity, E the electrode
potential, F the Faraday constant, and h
the number of sites on a unit surface area
of platinum.
Different types of potential oscillations
had been observed during formic acid
oxidation at a rotating electrode, that is, with
503
504
additional control of mass transfer at the

interface [129].
Potential-controlled current oscillations
do occur at single-crystal platinum surfaces
also [130132]. And nally, a special case
of mass-transfer inuence was reported by
Raspel and Eiswirth [133]. Starting with a
low concentration of protons in a formic
acid solution, the pH in the double layer
of a hanging meniscus at platinum singlecrystal surfaces can move to lower values.
In this case, again current oscillations
at constant potentials are found. It is
important to note that the oscillations can
be stopped by stirring the solution. This observation supports the above assumption
that mass-transfer processes like diffusion
can be responsible for the occurrence of
oscillations also.
Interestingly, new impedance results
(Fig. 33) show that a Niquist plot can
also be an indication for dynamic instability [131, 132]. For earlier observations, using CH2 O in 0.1 M NaOH on
a rhodium rotating disk electrode (RDE),

see Ref. [134].
Current Oscillations during
Methanol Oxidation
As already mentioned in the introduction,
platinum oxides can be reduced via a
chemical reaction with a fuel like methanol
in solution. Figure 34 shows the effect,
observed also by introducing HCOOH
and CH2 O for comparison [127, 135]. At
constant methanol concentration, the rate
of this reaction depends on the amount
of oxide coverage. On the other hand, at
the applied potential, the current for oxide
formation is decreasing with coverage.
After reaching a maximum value, the rate
of chemical oxide reduction decreases, and
then the current of oxide formation again
increases in this part of the cycle. Figure 35
shows current oscillations observed for
gas diffusion electrodes in contact with
a methanol solution at 1.2 and 1.5 V
5.2.5.3
Im Z 8
[]
6
4
2
5 Hz
8
Re Z
[]
10 kHz
4
4
6
0.1 Hz
4
6
8
Fig. 33 Niquist plot for formic acid oxidation on platinum,

obtained at +740 mV versus SHE, that is, in the region of the
positive slope of i() (compare Fig. 7), showing an inductive
loop and negative faradic impedance; 0.1 M
HCOONa/0.033 M H2 SO4 , indicating a dynamic
instability [131, 132], as already mentioned under
Sect. 5.2.5.1.
Reduction of Pt oxide, formed
in 2.5 M H2 SO4 , by dipping the
platinized Pt electrode in 0.1 M
fuel/2.5 M H2 SO4 solution [127, 135].
Fig. 34
1.2
1.0
Potential relative to jHo2

[V]
RHE [136]. As the gure suggests, the cycle

frequency follows the applied potential as
parameter.
Two additional comments are of importance. (A) The oscillations of Fig. 35 cannot
be observed at smooth platinum surfaces.
Obviously, the diffusion in the electrolytelled part of the pores of the gas diffusion
electrode, used in the experiment, is essential for the effect. (B) Indeed, if we disturb
this mass-transfer process in the pores of
the above-studied porous methanol electrode, the oscillations immediately vanish.
Disturbance is possible by applying oxygen
to the gas chamber of the porous electrode.
This results in a strong chemical reaction
with the methanol in the pores, forming
CO2 [136]. On the other side, applying nitrogen to the gas chamber has no effect on
the oscillations.
For a somewhat complicated case,
Schell and coworkers [137] have studied
voltammetric responses, using a rotating
platinum disc in alkaline solution. Oscillation had been observed for potentials above 700 mV RHE. Note, that
in alkaline solution, the formation of
PtOH starts already on leaving the
hydrogen region [138]. High-order periodic responses were obtained in experiments in which the system was allowed to
relax under xed conditions, following the
transfer of the electrode to the methanol
solution.
And again, potential oscillations of different types between 600 and 900 mV
RHE are observed also, using smooth platinum in 0.5 M CH3 OH/0.5 M
H2 SO4 [139].
1.4
0.8
0.6
0.4
CH3OH
0.2
HCOOH
CH2O
0.0
[min]
Current Oscillations during CO

Oxidation at a Rotating Disc
According to a recent experiment, using a
rotating platinum electrode [140], current
oscillations are observed also during potentiostatic oxidation of carbon monoxide.
Figure 36 shows the current/time behavior after application of 0.95 V on a rotating
polycrystalline platinum electrode. Oscillations can only be observed in a small
potential window between 0.9 and 1.0 V
and in the presence of the rotations of the
electrode. The current density seems to be
small enough to ensure that the potential drop at the interface is only negligibly
disturbed by the ohmic drop in the solution. Therefore, the sources of instability
should be of a chemical nature. The oscillations are probably connected to the
initial stages of platinum oxide formation
and the interference on this process by
CO adsorption and/or oxide reduction.
Because of the low solubility of CO in
the electrolyte, the rate of mass transfer to
5.2.5.4
505
0.20
Current density
[A cm2]
c
0.16
0.12
0.08
50
100
150
200
250
300
Time
[s]
Fig. 35 Current Oscillations during methanol oxidation at a Pt catalyzed gas
diffusion electrode (0.4 mgPt cm2 ), supplied with nitrogen to the gas side;
2.0 M CH3 OH/0.5 M H2 SO4 ; controlled potentials versus RHE (a) 1.0; (b) 1.2;
and (c) 1.5 Volt [136].
0.15
0.4
0.12
I
[mA cm2]
506
0.3
0.00
0.2
3000
3500
4000
0.1
0.0
0
1000
2000
3000
4000
t
[s]
Fig. 36 Current/time transient of a rotating platinum electrode (1.000 rpm) in
CO-saturated 0.1 M HClO4 at 0.95 V versus RHE. Initial instants after potential
application, the insert shows a time window [140].
5000
the surface (e.g. via convection) is of importance. At a rotating electrode, one has
above 1.0 V versus RHE limiting current,
being constant also during the negative
scan and passing the sharp maximum of
the anodic scan (Fig. 4 in Ref. [140]).
5.2.6
Nonelectrochemical Pathway of Methanol

Oxidation
During investigations of DMFCs, it was

observed that as a result of the diffusion of methanol molecules through the
electrolyte to the three-phase boundary of
the cathode, the potential of the oxygen
electrode may shift to more negative
values. In addition, a higher consumption
of methanol was found [57, 88]. It was assumed that this effect results from simultaneous oxygen reduction and methanol
oxidation at the platinum surface used,
the oxygen reduction being only slightly

poisoned by the adsorbed methanol. This
explanation was supported by investigations of Chu and Gilman using rotating
Pt disc electrodes [141]. But, recently it
has been shown that a chemical pathway, the purely heterogeneous reaction of
methanol and oxygen to form carbon dioxide, has to be considered also [136]. This
can be studied best in the presence of large
amounts of both reactants. Therefore, a
higher methanol concentration was used
in the electrolyte near the oxygen cathode.
In the experiments of Fig. 37, we present
rst oxygen reduction (curve (a), oxygen
supplied on the gas side of the Pt catalyzed
diffusion electrode only) without methanol
in the electrolyte. Of course, a standard
oxygen-reduction behavior results. For
comparison, in curve (b), a 2 M methanol
solution is used and nitrogen is supplied
to the gas chamber of the gas diffusion
Current density
[A cm2]
0.1
CH3OH
b, c
0.0
O2 + CH3OH
d
0.1
O2
0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Potential vs RHE
[V]
Fig. 37 Stationary current/potential curves of a Pt catalyzed gas diffusion
electrode as in Fig. 33. Points taken after 3 to 5 min, room temperature;
(a) 0.5 M H2 SO4 , oxygen supplied to the gas chamber only; (b) 2.0 M
CH3 OH/0.5 M H2 SO4 , nitrogen supplied to gas chamber and solution;
(c) methanol solution and nitrogen supplied to the gas chamber as in (b),
but oxygen supplied to the solution; (d) methanol in solution and oxygen
supplied to the gas chamber [136].
507
508
electrode. The current/potential plot of a

methanol anode is obtained. With curve
(c), the effect of adding oxygen to the
methanol solution is studied. Because of
the low oxygen solubility in the electrolyte,
practically no change is observed because
of the minimum oxygen reduction current.
With curve (d) nally, we have the
case of a strong heterogeneous chemical
reaction of methanol with oxygen at the
platinum surface, supplying methanol to
the solution and oxygen to the gas chamber
of the diffusion electrode. Below 0.7 V
curve (d) is comparable with curve (a), but
the currents are up to 30 mA cm2 smaller
than those without methanol in solution.
The anodic part shows methanol oxidation,
but again with a substantial loss in the rate.
If the strong effect of methanol, presented
in the difference between curve (d) and (a),
is due to a chemical reaction, the product
of this reaction, CO2 , must be found also
below 0.5 V, where no anodic oxidation of
methanol takes place at the platinum.
The carbon dioxide produced was collected by precipitation at three tubes in
series with 0.5 M Ba(OH)2 + 0.1 M BaCl2
solution and the determination of concentration was made by potentiometric titration with standard acid (0.5 M HCl) [142].
The formation of CO2 was followed at different potentials between 0.1 and 0.5 V for
the conditions of curves (a), (b), and (d) at
intervals of 60 min. Only for the conditions
of curve (d), methanol in the electrolyte and
O2 supplied to the gas diffusion electrode,
CO2 was found, ca. 1.8 103 mol h1 at
a difference in current between (a) and
(d) of ca. 35 mA cm2 . Obviously, the difference in current is due to methanol
consumption via a chemical pathway at
the three-phase boundary of the Pt gas
diffusion electrode.
Recently, another method was suggested
to measure the current efciency for CO2
production during methanol oxidation,

using online mass spectrometry in combination with a dual thin-layer ow through
cell [143].
As shown decades ago, formaldehyde
also is obtained via a nonelectrochemical
reaction at Pt surfaces ( [25, 143] and
references therein). In acid solution,
formaldehyde reacts with water to form
the gem-diol H2 C(OH)2 [134]. Using a
sensitive uorescence assay, Korzeniewski
and Childers [144] followed the formation
of gem-diol during the oxidation of
methanol out of a 15 mM CH3 OH/0.1M
HClO4 solution at smooth platinum for
different methanol oxidation potentials.
Above 500 mV RHE, a strong drop of the
conversion of methanol to the gem-diol
is observed.
References
1. A. Kutschker, W. Vielstich, Electrochim. Acta
1963, 8, 985989.
2. W. Vielstich, Chem.-Ing.-Tech. 1963, 35, 362.
3. E. Guth, E. J. Haase, H. G. Plust et al., Proc.
21st Ann. Power Sources Conf., Atlantic City,
1967, p. 29ff.
4. O. Bloch, M. Prigent, J. C. Balaceanu, The
Electrochemical Society Meeting, Indianapolis, 1961, Ext. Abstr. No 116.
5. G. Gruneberg, Dissertation Uni Braunschweig, 1958.
6. W. Vielstich, Fuel Cells, John Wiley & Sons,
London, UK, 1970, pp. 103107.
7. T. Iwasita, W. Vielstich, E. Santos, J. Electroanal. Chem. 1987, 229, 367376.
8. A. Hamnett, Catal. Today 1997, 38,
445457.
1973, 45, 205.
10. V. S. Bagotzki, Yu. B. Vasiliev, Electrochim.
Acta 1966, 11, 1439.
11. W. Guther, W. Vielstich, Electrochim. Acta
1982, 27, 811816.
12. X. H. Xia, T. Iwasita, J. Electrochem. Soc.
1993, 140, 25592565.
13. V. S. Bagotzki, Yu. B. Vasiliev, O. K.
Khasova, J. Electroanal. Chem. 1977, 81, 229.
14. A. Bewick, K. Kunimatsu, S. Pons, J. W.
Russel, J. Electroanal. Chem. 1984, 160, 47.
15. T. Iwasita in Advances in Electrochemical Science and Engineering (Eds.: H. Gerischer,
C. W. Tobias), Wiley-VCH, Weinheim, Germany, 1995, pp. 123216, Vol. 4.
16a. S. Wilhelm, T. Iwasita, W. Vielstich, J. Electroanal. Chem. 1987, 238, 383391.
16b. K. I. Ota, Y. Nakagawa, M. Takahashi, J.
17. M. Watanabe, S. Motoo, J. Electroanal.
Chem. 1975, 60, 267273.
18. B. Bittins, E. Cattaneo, P. Konigshoven
et al. in Electroanalytical Chemistry (Ed.: A. J.
Bard), Marcel Dekker, New York, 1991,
pp. 182219, Vol. 17.
19. H. Baltruschat in Interfacial Electrochemistry
(Ed.: A. Wieckowski), Marcel Dekker, New
York, 1999, pp. 577597.
20. P. A. Christensen, A. Hamnett, Techniques
and Mechanisms in Electrochemistry, Blackie
Academic, Chapman & Hall, London, 1994,
pp. 217224, 278283.
21. O. Wolter, C. Giordano, J. Heitbaum et al.,
Proc. Symp. Electrocatalysis, The Electrochemical Society, Pennington, N.J., 1982,
pp. 235253.
22. P. Stonehart, Electrochim. Acta 1973, 18, 63.
23. E. Santos, E. Leiva, W. Vielstich et al., J.
24. L. H. Leung, A. Wieckowski, M. J. Weaver,
J. Phys. Chem. 1988, 92, 69856990.
25. T. Iwasita, W. Vielstich, J. Electroanal.
Chem. 1986, 201, 403408.
26. X. H. Xia, T. Iwasita, F. Ge et al., Electrochim. Acta 1996, 41, 711718.
27. T. Iwasita in Handbook of Fuel Cells (Eds.:
W. Vielstich, A. Lamm, H. Gasteiger), John
Wiley & Sons, Chichester, UK; in press.
28. J. Clavilier, C. Lamy, J. M. Leger, J. Electroanal. Chem. 1981, 125, 249.
29. S.-Ch. Chang, L. H. Leung, M. Weaver, J.
Phys. Chem. 1990, 94, 60136021.
30. J. Willsau, Dissertation Uni Bonn, 1985,
pp. 8084.
31. J. Willsau, J. Heitbaum, Electrochim. Acta
1986, 31, 943948.
32. J. Clavilier, R. Parsons, R. Durand et al., J.
33. R. R. Adzic, A. V. Tripkovic, W. E. OGrady,
Nature 1982, 296, 137.
34. R. R. Adzic in Handbook of Fuel Cells (Eds.:
W. Vielstich, A. Lamm, H. Gasteiger), John
Wiley & Sons, Chichester, UK; in press.
35. I. Villegas, M. J. Weaver, J. Chem. Phys.

1994, 101(2), 16481660.
36. W. F. Lin, T. Iwasita, W. Vielstich, J. Phys.
Chem. B 1999, 103, 32503257.
37. F. Kitamura, M. Takahashi, M. Ito, Surf. Sci.
1989, 223, 493.
38. T. Iwasita, X. H. Xia, E. Herrero et al., Langmuir 1996, 12, 42604265.
39. J. Giner, Ber. Bunsen-Ges. Phys. Chem. 1959,
63, 386.
40. N. Markovic, P. Ross, J. Electroanal. Chem.
1992, 330, 499.
J. Phys. Chem. 1994, 98, 5074.
42. O. M. Magnussen, J. Hagebock, J. Hotlos
et al., Faraday Discuss. 1992, 94, 329.
43. G. J. Edens, X. Gao, M. J. Weaver, J. Electroanal. Chem. 1994, 375, 357.
44. T. Iwasita, X. H. Xia, H. D. Liess et al., J.
Phys. Chem. B 1997, 101, 75427547.
45. A. Rincon, M. C. Perez, G. Orozco et al.,
Electrochem. Commun. 2001.
46. T. Iwasita, X. H. Xia, J. Electroanal. Chem.
1996, 411, 95.
47. B. A. Sexton, K. Rendulic, A. E. Hughes,
Surf. Sci. 1982, 121, 181.
48. G. B. Fisher, J. L. Gland, Surf. Sci. 1980, 94,
446455.
49. P. A. Thiel, T. E. Madey, Surf. Sci. Rep. 1987,
7, 211385.
50. R. W. McCabe, L. D. Schmidt, Surf. Sci.
1977, 66, 101124.
51. J. A. Caram, C. Gutierrez, J. Electroanal.
Chem. 1991, 305, 259.
52. A. Wieckowski, M. Rubel, C. Guiterres, J.
53. T. Iwasita, X. H. Xia, H.-D. Liess, W. Vielstich, J. Phys. Chem. B 1997, 101,
75427547.
54. J. Willsau, J. Heitbaum, J. Electroanal.
Chem. 1985, 185, 181183.
55. J. M. Bowmann, J. S. Bittmann, L. B. Harding, J. Chem. Phys. 1986, 85, 911.
56. S. Wilhelm, T. Iwasita, W. Vielstich, J. Electroanal. Chem. 1987, 238, 383391.
57. A. Hamnett in Interfacial Electrochemistry
York, 1999, pp. 843883.
58. W. Vielstich, P. A. Christensen, S. A. Weeks
327333.
59. J. O.M. Bockris, H. Wroblowa, J. Electroanal. Chem. 1964, 7, 428.
509
510

60. O. A. Petrii, B. I. Podlovchenko, A. N.
Frumkin et al., J. Electroanal. Chem. 1965,
10, 253.
61. J. A. Shropshire, J. Electrochem. Soc. 1965,
112, 465.
62. J. A. Shropshire, J. Electrochem. Soc. 1967,
114, 773.
63. H. Binder, A. Kohling, G. Sandstede in
From Electrocatalysis to Fuel Cells (Ed.:
G. Sandstede), University of Washington
Press, Seattle, 1972, pp. 4379.
64. K. Franasczcuk, J. Sobkowski, J. Electroanal.
Chem. 1992, 327, 235.
65. T. Iwasita, F. C. Nart, W. Vielstich, Ber.
Bunsen-Ges. Phys. Chem. 1990, 94,
10301034.
et al., J. Phys. Chem. 1993, 97, 12201229.
17951803.
68. N. Markovic, H. A. Gasteiger, P. N. Ross
69. K. A. Friedrich, K. P. Geyzers, U. Linke
123128.
1998, 14, 19671970.
71. M. Krausa, W. Vielstich, J. Electroanal.
Chem. 1994, 379, 307314.
72. M. P. Hogarth, J. Munk, A. K. Shukla et al.,
J. Appl. Electrochem. 1994, 24, 85.
73. J. P. Iudice de Souza, T. Iwasita, F. C. Nart
et al., J. Appl. Electrochem. 2000, 30, 4348.
74. W. Chrzanowski, H. Kim, A. Wieckowski,
Catal. Lett. 1998, 50, 6975.
75. T. Iwasita, H. Hoster, A. John-Anacker
et al., Langmuir 2000, 16, 522529.
76. T. Frelink, W. Visscher, J. A. R. van Veen,
Surf. Sci. 1995, 335, 353360.
1997, 13, 59745978.
78. H. Gasteiger, N. Markovic, P. N. Ross et al.,
J. Phys. Chem. 1994, 98, 617625.
79. W. F. Lin, M. S. Zei, M. Eiswirth et al., J.
Phys. Chem. 1999, 103, 69686977.
80. H. A. Gasteiger, P. N. Ross, E. J. Cairns,
Surf. Sci. 1993, 293, 6780.
81. P. N. Ross in Electrocatalysis (Eds.: J. Lipkowski, P. N. Ross), Wiley-VCH, New York,
1998, pp. 4374.
82. F. Bautier de Mongeot, M. Scherer,
B. Gleich et al., Surf. Sci. 1998, 411,
249362.
83. S. Cramm, K. A. Friedrich, K. P. Geyzers

et al., Fresenius J. Anal. Chem. 1997, 358,
189192.
85. A. Hamnett in Interfacial Electrochemistry
York, 1999, 843883.
86. E. Reddington, A. Sapienza, B. Gurau et al.,
Science 1998, 280, 17351739.
87. K. Wang, H. Gasteiger, N. Markovic et al.,
88. S. Wasmus, A. Kuver, J. Electroanal. Chem.
1999, 461, 1431.
89. D. M. Kolb in Adv. Electrochem. a. Electrochem. Engn. (Eds.: H. Gerischer, C. W.
Tobias), John Wiley & Sons, New York,
1978, p. 125, Vol. 11.
90. R. A. Hess, Ch. C. Liang, US Patent
3 340 097, 1967 (applied 1964).
91. M. Watanabe, S. Motoo, J. Electroanal.
Chem. 1977, 78, 243.
92. R. R. Adzic, Proc. Symp. Electrocatal. 1982,
82, 309.
94. E. Schwarzer, W. Vielstich, Chem.-Ing.-Tech.
1973, 45, 201.
95. A. Castro-Luna, T. Iwasita, W. Vielstich, J.
97. E. Leiva, T. Iwasita, E. Herrero et al., Langmuir 1997, 13, 6287.
98. M. Angerstein-Kozlowska, B. McDougall,
B. E. Conway, J. Electrochem. Soc. 1973, 122,
756.
99. M. Llorca, E. Herrero, J. M. Feliu et al., J.
100. E. Herrero, A. Fernandez-Vega, J. M. Feliu
102. H. Hoster, T. Iwasita, H. Baumgartner
et al., PCCP 2001, 3, 337346.
103. H. Hoster, doctor thesis, Uni-Bw Munchen
2002.
104. H. Hoster, T. Iwasita, H. Baumgartner
et al., J. Electrochem. Soc. 2001, 148, A496.
105. K. Franaszczuk, E. Herrero, P. Zelenay
et al., J. Phys. Chem. 1992, 96, 85098516.
106. S. Srimarulu, T. D. Jarvi, E. M. Stuve, Electrochim. Acta 1998, 44, 11271134.
107. D. Chu, S. Gilman, J. Electrochem. Soc. 1996,
143, 1685.
108. A. Aramata, M. Masuda, J. Electrochem. Soc.
1991, 138, 19491957.
109. M. Watanabe, M. Uchida, S. Motoo, ECS
Meeting, Boston, Mass., Ext. Abstr. 339,
1986.
110. I. A. Rodrigues, J. P. I. Souza, F. C. Nart,
Langmuir 1997, 13, 6829.
111. R. Liu, H. Iddir, Q. Fan et al., J. Phys. Chem.
B 2000, 104, 35183531.
112. H. Kim, I. Rabelo de Moraes, G. TremiliosiFilho et al., Surf. Sci. 2001, 474, L203L212
113. T. J. Schmidt, H. Gasteiger, R. J. Behm,
Electrochem. Commun. 1999, 1, 14.
114. V. A. Paganin, W. Lizcano-Valbuena, E. R.
Gonzales, Electrochim. Acta 2000, 47,
37153722.
115. K. F. Bonhoeffer, G. Vollheim, Z. Naturforsch. 1953, 8b, 406.
116. U. F. Franck, R. FitzHugh, Z. Elektrochem.
1961, 65, 156.
117. M. T. M. Koper, Electrochim. Acta 1992, 37,
17711778.
118. T. Yamazaki, T. Kodera, Electrochim. Acta
1991, 36, 639646.
119. M. T. M. Koper, J. H. Sluyters, J. Electroanal. Chem. 1991, 303, 73.
120. M. T. M. Koper, J. H. Sluyters, J. Electroanal. Chem. 1994, 371, 149.
121. K. Krischer in Modern Aspects of Electrochemistry (Eds.: J. OM. Bockris, B. E. Conway,
R. White), Kluwer Academic Publishers,
New York, 1999, pp. 1142, Vol. 32.
122. K. Krischer, N. Mazouz, P. Grauel, Angew.
Chem. 2001, 113, 842863.
123. K. Krischer, N. Mazouz, P. Grauel, Angew.
Chem., Int. Ed. Engl. 2001, 40, 850861.
124. K. Krischer, M. Lubke, W. Wolf et al., Electrochim. Acta 1995, 40, 69.
125. M. Wolf, K. Krischer, M. Lubke et al., J.
126. E. Muller, S. Tanaka, Z. Elektrochem. 1928,

34, 256.
127. W. Vielstich, Fuel Cells, John Wiley & Sons,
London, UK, 1970, pp. 8891.
128. H. Okamoto, N. Tanaka, M. Naito, Chem.
Phys. Lett. 1996, 248, 289295.
129. F. N. Albahadily, M. Schell, J. Electroanal.
Chem. 1991, 308, 151173.
130. N. Markovic, P. Ross, J. Phys. Chem. 1993,
97, 97719778.
131. P. Strasser, M. Eiswirth, G. Ertl, J. Chem.
Phys. 1997, 107, 9911003.
132. P. Strasser, Interface 2000, 9, 4652.
133. F. Raspel, M. Eiswirth, J. Phys. Chem. 1994,
98, 76137618.
134. J. F. Walker, Formaldehyde, 3rd ed., Reinhold Publishing, New York, 1964.
135. J. E. Oxley, G. K. Johnson, B. T. Buzalski,
136. W. Vielstich, V. A. Paganin, F. Lima et al.,
J. Electrochem. Soc. 2001, 148, A502.
137. Y. Xu, A. Amini, M. Schell, J. Phys. Chem.
1994, 98, 12 75912 767.
138. E. Santos, M. C. Giordano, J. Electroanal.
Chem. 1984, 172, 201.
139. M. Krausa, W. Vielstich, J. Electroanal.
Chem. 1995, 399, 712.
140. D. C. Azevedo, A. L. N. Pinheiro, E. R.
Gonzales, Electrochem. Solid-State Lett. 2002,
5, A5154.
141. D. Chu, S. Gilman, J. Electrochem. Soc. 1994,
141, 1770.
142. G. H. Jeffery, J. Bassett, J. Mendham et al.,
Vogels Textbook of Quantitative Chemical
Analysis, 5th ed., John Wiley & Sons, New
York, 1989, p. 297.
143. H. Wang, Th. Lofer, H. Baltruschat, J.
144. C. Korzeniewski, Ch. L. Childers, J. Phys.
Chem. B 1998, 102, 489492.
511
512
5.3
Electrochemical Nucleation and Growth

Benjamin R. Scharifker and Jorge Mostany
Universidad Simon Bolvar, Caracas,
Venezuela
5.3.1
Phase Formation on Electrodes
The onset of stable structures, known as

nuclei, is characteristic of the phase formation phenomena [1]. These structures
are aggregates of atoms or molecules, and
they become centers for the propagation of
the two-dimensional or three-dimensional
new phase. The phase formation process
is supported by material originated from
the bulk of the mother phase and invariably, during electrochemically driven
phase formation processes, occurs also
with chemical reactions taking place at the
interface between the developing phase
and the electrolyte solution. A sequence of
events lead to the formation and growth of
the new phase; these include adsorption of
precursors on the electrode substrate, frequently accompanied also by charge transfer, formation of two- or three-dimensional
nuclei, and the growth of the new phase.
The supersaturation (or excess Gibbs energy) of the system decreases with the
formation and growth of the new phase;
hence the overall process is categorized as
a rst-order phase transformation.
An appropriate description of the energetics and kinetics of the nucleation
process requires consideration of the equilibrium properties of the bulk phases
involved as well as contributions that arise
from the formation and growth of the interface between them. Clusters that constitute
the seeds of the phase transformation are
usually composed of a small number of
atoms or molecules. For such small entities, the surface or interfacial properties
will play dominant roles in preserving the

excess energy and, in some cases, this
might result in a long lasting metastable
equilibrium, with indenite preservation
of the supersaturated state. Thus, in terms
of classical thermodynamics, the total free
energy of a spherical drop in equilibrium
with its vapor is obtained from the properties of the condensed bulk phase as
well as the interface. The Kelvin equation [2] ln(p/p ) = 2 L /rkT states that
the excess vapor pressure of a drop varies
with the surface tension or free energy of
the surface, , and the molar volume of
the condensed phase L , and also with the
inverse of the radius of curvature of the
surface, 1/r; p is the equilibrium vapor
pressure of the at surface between phases
with the same chemical composition. According to this expression, small clusters
will not be stable unless the pressure of the
vapor surpasses their equilibrium vapor
pressure, and this will invariably occur at a
value higher than the equilibrium pressure
of the bulk phase. Under such conditions,
the increasing volume of the new phase
reduces the free energy, whereas the free
energy change due to the expansion of the
surface is always positive. Therefore, the
combined contributions due to formation
of a bulk phase and a surface produce a
maximum of
G, occurring under given
conditions at a certain size of the aggregate
of the new phase, thus dening the critical
radius r . The critical radius must be exceeded for a nucleus to grow irreversibly,
and this may happen through uctuations
in temperature, pressure or composition,
or, in electrochemical systems, by changes
of the electrical potential.
For a simple electrochemical process, Mz+
+ ze = M(ads) , the equi(sol)
librium potential Erev is given by
the Nernst equation Erev = E +
RT /zF ln(aMads /aMn+,sol ), where E is the
eq
5.3 Electrochemical Nucleation and Growth
standard potential of the redox equilibrium

between bulk metal and a solution of Mn+ ,
aMads is the equilibrium surface activity of
eq
ad-atoms, and aMn+,sol is the activity of Mn+

in solution. This quantity may be modied
at will varying the electrode potential to
a new value E = Erev + , where is the
overpotential at which the surface activity
of M satises the relation aMads /aMads =
eq
exp(zF /RT ). It can be shown that both
the energy of formation of the critical nucleus,
G = 16 3 M 2 ()/3 2 (zF )2
(the function () of the contact angle
between the nucleus and the substrate
will be dened later), as well as its radius,
r = 2/zF , depend on the overpotential [3]. This illustrates the possibility
of controlling, rapidly, precisely, and reversibly, the conditions that determine the
occurrence of electrochemical nucleation
processes.
A critical cluster of atoms is at a maximum of free energy. Considering then
that the appearance of critical nuclei is
the limiting step of the nucleation process
and that the probability of their formation
would be proportional to exp(
G /kT ),
the nucleation rate will be proportional to
exp(16 3 M 2 ()/3 2 z2 2 kT ), where
M is the molar mass of the deposit and
is its mass density. The preexponential
factor derives from kinetic arguments, considering that the populations of clusters of
different sizes are in a steady state, determined by a sequence of reactions such as
n1
m1
n+1
m+1
(1)
Thus, consecutive addition (and detachment) of atoms leads eventually to the
formation of supercritical nuclei that grow
irreversibly and are not further considered
in the steady state equations. Becker and
Doring [4] provided an analytical solution

considering a system of constant composition in which supercritical clusters are
reintroduced into the system as the equivalent amount of discrete units. A steady
state expression of the nucleation rate follows, Js = Zm cm , where m is the net
probability of addition of an atom per unit
time from a critical cluster of size m , and
cm , the equilibrium concentration of critical clusters, is related to the monomer concentration through the Bolztmann equation, cm = c1 exp(
Gm /kT ) and the
nondimensional Zeldovich factor Z, which
accounts for the fact that the steady state
concentration at m is only 1/2 of the concentration at equilibrium, and that critical
clusters may still decay [5].
Classical nucleation theory uses macroscopic properties characteristic of bulk
phases, like free energies and surface
tensions, for the description of small clusters These macroscopic concepts may lack
physical signicance for typical nucleus
sizes of often a few atoms as found
from experimental studies of heterogeneous nucleation. This has prompted the
development of microscopic models of the
kinetics of nucleation in terms of atomic
interactions, attachment and detachment
frequencies to clusters composed of a
few atoms and with different structural
congurations, as part of a general nucleation theory based on the steady state
nucleation model [6]. The size of the critical nucleus follows straightforwardly in
the atomistic description from the logarithmic relation between the steady state
nucleation rate and the overpotential. It
has been shown that at small supersaturations, the atomistic description corresponds to that of the classical theory of
nucleation [7].
Given that in electrochemical systems the supersaturation of a developing
513
514
phase is determined by the overpotential,

experimental studies of the kinetics of
phase formation processes are generally conducted under potentiostatic conditions. At constant potential then, it
is generally considered that the steady
state nucleation rate Js maintains a constant value A, which depends on the
applied overpotential. The nucleation frequency A may be determined from
the analysis of the current that represents the overall rate of electrodeposition.
Analyses of current transients, together
with direct observation of the electrode
surface, constitute the most common
methods for the study of nucleation
processes.
In many cases of electrochemical phase
formation reactions, the charge-transfer
step is fast and the rate of growth of
mature supercritical nuclei is well described by mass transport of electrodepositing species to hemispherical growth
centers [8]. The current attending the
growth of an isolated nucleus is proportional to the square root of its age,
I (t) = zF (2Dc)3/2 (Mt/)1/2 , where c is
the molar concentration and D is the
diffusion coefcient of electrodepositing
species in solution. On the other hand,
the rate of appearance of nuclei on a
surface with a number density N0 of active sites for nucleation may be expressed
as dN/dt = A(N0 N ), where A is the
nucleation rate per site, from which the
number density of nuclei on the surface as a function of time will be given
by N = N0 [1 exp(At)]. Depending on
the frequency at which nuclei appear and
the corresponding rate at which active
sites are depleted, the nucleation process may be classied as instantaneous
or progressive. When A is very large
(A 1/t), depletion of nucleation sites
occurs at very early stages, the above expression soon reduces to N = N0 , and
nucleation is instantaneous. Conversely,
for very small A, the number density of
nuclei increases initially linearly with time,
N = N0 At, and nucleation is said to be
progressive.
In multiple nucleation conditions, the
current density may be expressed as a
convolution of the growth current of individual nuclei and their birth rate on the
surface. In the very initial stages, nuclei
are small in both size and number, and
they grow independently of each other.
The current transient is then given
by
3/2 (Mt/)1/2 N %( At)
i(t) = zF (2Dc)
0
[9], where %( At) is the Dawson

integral,
At 2
At
e d.
%( At) = 1 (e / At) 0
For, At 20, %( At) 1 and the equation above describes a current density
proportional to N0 t 1/2 , corresponding to
instantaneous nucleation; for
At 0.2
nucleation is progressive, %( At)
(2/3)At and the current density is proportional to (2/3)N0 At 3/2 [10, 11].
An overall description of the nucleation
and growth process valid for all times
is in general complicated by interactions
among growing centers. When growth
is diffusion controlled, then overlap of
the spherical diffusional elds supporting
the growth of individual nuclei must be
considered, eventually collapsing into a
single planar eld to the plane of the electrode. The many-body problem that ensues
has been approached following several
strategies. General models not restricted
to the instantaneous case, and allowing
determination of A and N0 from current transients, consider two-dimensional
projections of the hemispherical diffusional elds around individual nuclei onto
the plane of the electrode [12] and estimate their overlap using the Avrami
theorem [13]. The current density is then

given by [11]:
i(t) =
zF Dc
(Dt)1/2
(At)1/2
eAt
2
1
e d
(At)1/2 0

1 eAt
1
At

cM 1/2
3/2
1 exp (2) D

1 eAt
(2)
N0 t 1
At
The expression above describes appropriately experimental current transients

obtained in studies of electrodeposition
processes for numerous systems. The electrodeposition of Hg may be considered
as a model system for the study of the
fundamentals of electrochemical phase
formation, and is briey discussed here
(see Sect. 5.3.4.4). Analysis of current transients allows determining the values of
the nucleation rate and the number density of active sites. The experimentally
observed dependence of the nucleation
rates on overpotential and the charge
transferred during the electrodeposition
reaction are in agreement with fundamental theoretical postulates. The number
density of nuclei observed microscopically
on the surface, on the other hand, are in
good agreement with those predicted from
the nucleation rates obtained. Overall, experimental studies show the dominant
role of deposit|substrate interactions during the electrochemical phase formation
phenomena.
Detailed knowledge of the mechanisms
and kinetics of phase formation are of practical interest for applications in surface
nish, electrometallurgy, electrocatalysis,
and more recently in microelectronics

and the manufacture of nanostructures.
The experimental conditions of the electrochemical processes (including current
density, potential, electrolyte composition,
or the use of complexing agents) have
substantial effects on the morphology
and properties of the electrodeposits [14].
In particular, interest in developing efcient and environmentally friendly energy
sources has motivated the study of the
catalytic properties of micro- and nanometallic clusters of noble and transition
metals. These clusters may be deposited
on metal surfaces or dispersed in threedimensional structures, such as porous
matrices, polymeric membranes, or thin
conductive polymer lms, to catalyze reactions such as oxygen reduction or the oxidation of hydrogen or organic molecules.
Because of the strong effects of surface
structure on reactivity, available theories
of phase formation have become especially
useful for controlling the size, morphology, density, and spatial distribution of
metallic electrodeposits on electrodes.
5.3.2
Thermodynamics of Electrochemical Phase

Formation
The thermodynamic treatment of phase

transitions is usually based upon the
equilibrium properties of the bulk phases
involved. A description of the formation
of a new phase requires consideration
of at least two additional factors, the
contribution of changes in the surface
states to the total energy, and the inuence
of these changes on the rate of formation
of the new phase.
Transformation of one phase into another will occur only when both phases
are not at mutual equilibrium; the greater
the deviation from equilibrium, the greater
515
516
the driving force for the transformation. If

both phases are well developed, that is, if
they are sufciently large for the contributions of surface effects to be ignored, then
the slightest deviation from equilibrium
will cause a displacement of the boundary between phases, to counterbalance the
deviation. However, if the phases are not
well developed, that is, when one of them
does not even exist, then the initial surface contributions to the total energy can
be sufciently important so as to impede
the formation of a new phase. Such systems are in metastable equilibrium, and
may remain in such condition for considerable periods of time. Common examples
of metastable equilibria are supersaturated
vapors (whether overcooled or overcompressed) and supercooled liquids, some of
which, as in the case of glasses, remain
indenitely as stable amorphous solids.
The new phase generates from the
metastable system necessarily in the form
of nuclei. These are small clusters of
atoms or molecules that in the prevailing conditions have developed into a size
sufciently large to grow spontaneously,
ensuring their own viability and, eventually, the stability of the new phase. The
intensive properties of nuclei differ from
the bulk phase only because of their small
size. This point of view is not necessarily correct since the properties, structure,
and even the composition of small clusters
may not be identical to those of the corresponding bulk phase, but the notion of
a nucleus determined essentially by its
size is useful to relate the macroscopic and
microscopic descriptions of phase formation, as described below.
Leaving aside for the moment the
problem of the origin of the nucleus
and considering only the conditions for
its further growth, the effect of its small
size is manifested by an overwhelmingly
large ratio of surface to volume, which

for macroscopic bodies tends to zero. Lord
Kelvin [2] demonstrated that the larger the
vapor pressure in equilibrium with a small
drop of liquid at a given temperature, the
smaller the radius r of the drop. If the drop
is not sufciently small, then its surface
energy may be expressed as the product of
the surface area 4r 2 , and the macroscopic
surface tension, , corresponding to r
. The thermodynamic potential of the
total system, composed of vapor V and
liquid L, is expressed by the equation,
G = V dnV + L dnL + dS
(3)
where nV and nL represent the number of

molecules on V and L, V and L are the
chemical potentials referred to a molecule
in the bulk phase at the given temperature
and pressure, and S is the surface area
of the drop. Thermodynamic equilibrium
determined by the condition G = 0 and
for a closed system,
V L + 4
d(r 2 )
=0
dnL
(4)
Designating the volume occupied by a

molecule in the liquid phase as L , then
nL = 4r 3 /3L , and,
V L +
2 L
=0
r
(5)
In the limiting case of r , this

expression reduces to the ordinary equilibrium condition between macroscopic
phases, V = L . Differentiating Eq. (5)
and considering that dV = V dp and
dL = L dp, where V is the volume per
molecule in the gaseous phase, then

1
(6)
(V L )dp = 2 L d
r
Neglecting L in comparison with V
and considering the vapor as an ideal
gas, for which V = kT /p, where k is

Boltzmanns constant and T the temperature, then

p
2 L
ln
=
(7)
p
rkT
where p/p is the ratio between the
pressures exerted over the surface of
radius of curvature r and a at surface,
of innite radius of curvature. This is
the starting point of the classical theory
of nucleation. According to Eq. (7), small
clusters of the new phase are characterized
by an excess vapor pressure. If the
supersaturation pressure is larger than p,
then the cluster will continue growing;
if it were less than p, it would tend
to evaporate. Since every cluster had
to be initially of molecular size (and
therefore p p ), it still remains to
be explained how a nucleus of a new
phase ever appears. The reason is that
all systems at nite temperature undergo
uctuations, which eventually produce
a cluster with an equilibrium vapor
pressure less than or equal to that of
the supersaturated state. Hence a new
phase will be formed, and according to
classical theory, will continue growing. But
before entering into a detailed discussion
of the matter, we will describe briey some
fundamental concepts necessary to discuss
equilibrium and phase transformations in
electrochemical systems.
5.3.3
Electrochemical Nucleation Kinetics
Electrochemical nucleation can be considered as a series of partial reactions in

which at least one step involves charge
transfer by electrons or ions through the
solution|electrode interface. The rate of the
process is determined by the potential difference existing across the double layer in
reactions such as
Mz+
(solution) + ze(electrode)
M(new phase)
(8)
for metal ion reduction to the corresponding metallic phase, or
M(solid) + aXz
(solution)
MXa(new phase)
+ aze
(electrode)
(9)
for anodic deposit formation.
According to ErdeyGruz and Volmer [15], the supersaturation c/c , where
c is the concentration of ions in solution
in equilibrium with a surface of radius of
curvature r, and c is the corresponding
concentration of ions in equilibrium with
a planar surface, is directly determined by
the overpotential, = E Erev , where E
is the electrode potential and Erev is the
equilibrium potential, in the absence of net
current ow for a single electrode reaction.
For the metal electrode|solution interface,

RT
c
=
(10)
ln
zF
c
where F is Faradays constant. This
relation expresses that through external
control of the electrical potential, the
conditions governing the electrochemical
nucleation process can be precisely and
reversibly controlled.
The Classical Nucleation Model
From Eq. (3), the Gibbs energy of formation of a nucleus may be formally described
as [3]
5.3.3.1
G = Gvolume + Gsurface
(11)
517
518
where
Gvolume

4
r 3
GV
=
3
(12)
where
GV is the Gibbs energy of
formation of the bulk phase per unit
volume, and
Gsurface = 4r 2
(13)
Gvolume is always a negative quantity,

the thermodynamic way of expressing the
spontaneity of the process, arising from
the greater stability of the new phase
with respect to the initial state. On the
other hand, the interfacial contribution,
Gsurface , is necessarily positive; otherwise
the new phase would immediately become
dispersed in the initial phase. For small
values of r, the second term predominates,
and the total change in Gibbs energy
involved in the formation of a nucleus is
also positive, that is, (
G/r)small r > 0.
However, for larger values of r, the
rst term of Eq. (11) predominates and
(
G/r)large r < 0. It is then evident
that a particular value of the radius
r = r exists, in which the Gibbs energy
change is maximum, (
G/r)r = 0.
Thus equating the rst derivative of
Eq. (11) to zero,

G
= 4(r )2
GV
r r
+ 8(r ) = 0
(14)
from which r , known as the critical

radius, is obtained as
r =
GV
(15)
Clusters with radius of curvature less

than r will minimize their energy by
disappearing in the initial phase, while
larger ones will grow spontaneously. A
cluster with radius r is designated as

a critical nucleus an entity in precise
equilibrium with the initial phase as
implied by (
G/r)r = 0, with equal
probabilities of growing or decaying.
If r r + r, the cluster will grow
irreversibly; alternatively, if r r r, it
will disappear. Equation (15) is a direct
consequence of Eq. (7), and is known as
the GibbsKelvin equation, relating the
Gibbs energy of an interface with it radius
of curvature. The reversible work required
to form a critically sized cluster from the
supersaturated phase can be obtained from
Eq. (14), integrating between r = 0 and
r = r [3],

16
3
Ghomo =
3
(
GV )2

=
(16)
4(r )2
3
which has been designated here as
Ghomo (in homogeneous phase) to emphasize that Eq. (16) refers to the work
of formation of a nucleus from the
bulk phase.
A comparison of Eqs. (15 and 7) reveals
that both the radius of the critical nucleus
and the work needed to form it are
functions of the ratio c/c . When c = c ,
that is, when the solution is saturated
in relation to an interface of innite
radius of curvature, the work required to
form a nucleus tends to innite, and the
phase transition does not occur. As the
supersaturation ratio c/c increases, the
radius of the critical nucleus, and hence its
reversible work of formation, diminishes.
Supersaturation is one of the principal
factors determining the work required to
form a critical nucleus, but there are
other factors worth considering too. In
the majority of cases and especially under
electrochemical conditions, formation of
a new phase occurs on a surface. The
interaction between the substrate and the

new phase has a strong inuence on
the reversible work of formation of the
critical nucleus, thus nucleation occurs
onto those sites on the surface with the
minimal work of formation of critical
clusters, the so-called active sites, to be
treated further below.
Consider the deposition of a liquid
phase in the solid|solution interface. The
liquid surface makes contact with the solid
at a characteristic angle , the contact
angle [16], dened as the angle between
the tangent planes to the solid and liquid
phases on the contact line, as shown in
Fig. 1.
At any point along the equilibrium
line between the contacting phases, there
are three surface tension forces i,j : 1,2
between the liquid drop and the solution,
at the contact angle with respect
to the surface; 1,3 between the solid
and the solution, directed along the
plane of the surface, and 2,3 due to
the solid|liquid interface, also along the
plane but opposing 1,3 . To maintain
equilibrium, the solid reacts with an
additional force perpendicular to the
surface and directed towards the bulk of
the solid. The equilibrium condition is
expressed by the Young equation
2,3 1,2 cos 1,3 = 0
(17)
A second relation between the three

forces is obtained using Dupres equation,
dening the reversible work W2,3 needed
to separate a unit area of the solid|liquid
interface. Conservation of energy requires
that [17]
2,3 + W2,3 = 1,3 + 1,2
subtracting Eq. (17) from Eq (18),
W2,3 = 1,2 (1 + cos )
Fig. 1
(19)
This quantity is maximum when = 0 ,

that is, when the liquid completely wets
the solid and W2,3 is zero when =
180 and the liquid evades the solid,
a characteristic behavior that constitutes
an effective method to determine the
interaction energy between surfaces.
The work of formation of a liquid drop
onto a surface may be written as
G =
GV V + 1,2 S1,2
+ 2,3 S2,3 + 1,3 S1,3
(20)
where V is the volume of the drop and

S1,2 , S2,3 and S1,3 are the surface areas
of the different interfaces involved. To
calculate the work of formation of the
critical nucleus, it is sufcient to know the
relation () between the volume of the
spherical shell in contact with the solid, v,
and the volume of a sphere with an equal
s1, 2
s1,3
(18)
s2, 3
Model of a liquid drop on a solid surface.
519
520
radius of curvature, Ve , in equilibrium

with the supersaturated solution. The
heterogeneous work of formation of the
critical nucleus on the surface would be
given by the corresponding work to form a
sphere with the same radius of curvature,
Eq. (16), multiplied by ()
Ghetero =
Ghomo ()

16(1,3 )3
=
()
3(
GV )2
(21)
The volume of the spherical shell is v =

(2 3 cos + cos3 )r 3 /3 whereas that
of the sphere is Ve = 4r 3 /3, thus,
(2 3 cos + cos3 )
v
=
Ve
4
(22)
0 () 1, nucleation at an interface
requires always less energy than within
the bulk phase.
To describe electrochemical phase formation processes, it is necessary to establish the role of the overpotential on
the formation of nuclei. Under electrochemical conditions, the Gibbs energy of
interfacial phase formation
GV , appears
from the charge transfer of ionic species
across the double layer, and can be dened
in terms of the product of the net charge
of formation of a cluster and the applied
overpotential:
() =
GV =
zF
M
(23)
where is the density of the deposit and M

is the molar mass of the electrodepositing
species. Then, the free energy of critical
nucleus formation may be expressed as
Ghetero =
16 3 M 2 ()
3(zF )2
(24)
where, to simplify the notation, 1,3 has

been written as . From this expression,
r =
2 M
zF ||
(25)
Figure 2 shows the free energy of hemispherical clusters ( = 90 , () = 1/2) of

mercury as a function of the radius, for
different values of the overpotential.
Nucleation on Electrode Surfaces
The probability of formation of a
critical nucleus is proportional to
exp(
G /kT ) [18], hence the rate
of nucleation will be proportional
to exp(4 (r )2 ()/3kT ) and this
exponential term is central to the theory
of nucleation; a full description of the
nucleation rate however requires also
determining the preexponential factors.
According to Farkas [19], each cluster is
characterized by its number of molecules,
n, and through kinetic arguments it
is possible to evaluate the evolution of
n. Since n is an integer, it can only
vary in 1 due to random addition or
loss of molecules. n may in principle
change in 2 or even more units due
to attachment of dimers, trimers, and
so forth, but these events will be
infrequent and may be neglected. This
argument leads to an expression for
the nucleation rate that contains an
indeterminate integration constant, that
Becker and Doring managed to eliminate
reintroducing clusters attaining a given
size, bigger than critical, as an equivalent
number of monomers [4]. In this way, a
steady state is maintained even though
the system is closed, and under these
conditions the number of clusters of a
given size remain constant. Thus, a steady
state distribution of cluster sizes may be
dened, and the nucleation rate is then
5.3.3.2

30 mV
3.0
45 mV
1018 G*
[J]
2.0
60 mV
1.0
90 mV
0.0
1.0
0
10
15
20
25
108
[r cm1]
Fig. 2 Free energy of formation of hemispherical mercury clusters from
Hg2 2+ aqueous solution at different overpotentials, as a function of
nucleus size. = 300 dyn cm1 , obtained from electrocapillary curves of
= 0.8 V.
Hg in KNO3 solution at EHg(II)/Hg
expressed as the rate of aggregation of

molecules to the critical nucleus, times the
steady state concentration of clusters of
critical size.
The treatment of Zeldovich [5] does
not differ in essence from the already
stated argument and allows for a more
precise evaluation of critical parameters
with a minimal of assumptions. It follows considerations from the Kramers
theory of transition states, according to
which the passage through the activated
state is subject to random external inuences, similar to those affecting the
Brownian movement of particles in intense elds [20]. Thus, the system contains
a constant number of molecules suffering
statistical uctuations at temperature T .
If Cn (t) is the concentration of clusters
of n molecules at time t, then, allowing only transitions between contiguous
neighbors, that is, from n to n 1 or
n + 1,
n1

n1
n
n+1

n
(26)
n+1
The change in concentration of clusters of n molecules may be written as dCn(t)/dt = n1 Cn1 (t) (n +
n )Cn (t) + n+1 Cn+1 (t), which has the
form of Kolmogorov differential equation
for Markov processes in discrete number
space and continuous time [21]. n and
n are respectively the net probabilities
of incorporation or loss of molecules by
a cluster per unit time, and these may
be dened formally as the aggregation or
detachment frequencies times the surface
area of the cluster of n molecules. Given
the small size of the clusters, n and n
are not simple functions of n and in general they are unknown. However, if n
and n are not functions of time, then
an equilibrium distribution Cn of cluster sizes exists, such that dCn /dt = 0 for
Cn (t) = Cn , and the following differential
521
522
equation may be stated if we consider that

in equilibrium, the number of clusters
that progress from n to n + 1 equals those
decaying from n + 1 to n per unit time:

Cn (t)
Cn (t)

=
n Cn
(27)
t
n
n
Cn
This expression holds if the time during
which the macroscopic phase transition
occurs is large compared to the characteristic times of the accretion and decay
processes, n1 and n1 .
Equation (27) is the FokkerPlanck
equation and describes diffusion in the
presence of an external force, n plays a
role similar to that of the diffusion coefcient in the Fick equation; for the special
case of Cn = C = constant, substituting
n for D, we obtain Ficks law,
2 Cn (t)
Cn (t)
=D
t
n2
(28)
However, the eld of force or gradient in the Gibbs energy introduced by

Eqs. (1113) must be, in general, retained
to describe the random diffusion of clusters through the distribution of their sizes.
The equilibrium concentration of n-mers
is given by
Cn = C1 exp
Gn
kT

(29)
where
Gn is the standard Gibbs energy
of formation of a n-mer from n monomers
and C1 is the concentration of monomers
in the initial phase. Referring to Eq. (16)
describing the Gibbs energy of a critical
cluster (cf. Fig. 2), it is clear that Cn is
minimal at n = n (n is the number
of molecules constituting the critical
nucleus) or r = r and that
G =
Gn therefore constitutes the activation

Gibbs energy for nucleation. Substituting
Eq. (29) in Eq. (27),

Cn (t)
Cn (t)
=
n
t
n
n
+
n Cn (t) (
Gn )
kT
n

(30)
To solve this equation, it is convenient

to dene a critical region of values
of n in which the Gibbs energies of
clusters differ from the maximum
G
less than kT, the energy of thermal
uctuations (cf. Fig. 3), extending a cluster
size range of (2kT / )1/2 , where =
( 2
Gn /n2 )n .
Since the Gibbs energy of activation is maximal at the critical nucleus,
(
Gn /n)n = 0 and to a rst approximation the variation of
G within the
critical region may be neglected; the second term in the right-hand side of Eq. (30)
will be small compared to the rst and
a steady state solution is readily found at
xed supersaturation and when clusters
attaining a given supercritical size are instantaneously extracted from the system
and replaced by an equivalent amount
of monomers. Under these conditions,
Cn (t)/t = 0 and the steady state nucleation rate Js may be expressed as

Cn
(31)
Js = n
n n
which is clearly a diffusion equation
that may be solved approximately with
the limiting layer model. Thus replacing
(Cn /n)n by Cn /
n, where
n is the
width of the critical region, (2kT / )1/2
and taking account that diffusion occurs at
both sides of the critical region, then [22]
Js =
2n Cn
(2kT / )1/2
(32)
Thus the steady state nucleation rate is

expressed as the product of the equilibrium
0.655
kT
1018 G q
[J]
0.650
0.645
0.640
0.635
0.630
0.625
30
35
40
45
50
55
60
n
[atoms]
Fig. 3 Critical region around n within which the energies of hemispherical
mercury clusters at = 90 mV differ from
G less than kT.
concentration of critical clusters, Cn , the

rate of attachment of monomers to critical
clusters, n , and a nonequilibrium
factor, known as the Zeldovich factor Z,
which takes into account that the steady
state concentration of n is only 1/2 of the
equilibrium value and that supercritical
clusters may still decay. The Zeldovich
factor is thus dened as

2
1

Gn
(33)
Z=
2kT
n2
n
The treatment as outlined above provides
a good approximation of the steady state
nucleation rate under conditions in which
G may be calculated. The principal

difculty in nding the nucleation rate is in
evaluating the equilibrium concentration
of clusters in terms of their formation free
energies. The classical way of realizing this
is to consider all clusters as chemically
equivalent,
Gn is obtained from the
Gibbs energy change for their formation
from monomers, as a sum of volume
and surface terms (cf. Eq. 11).
For small clusters, this procedure is

prone to criticism on several grounds.
The thermodynamic properties of small
clusters, for instance, surface tension or
composition, are supposed to be identical
to those of the equilibrium bulk phase. In
particular, it is considered that clusters are
of homogeneous chemical composition,
but this is hardly so in clusters that are
only a few nm in diameter, given that for
them the width of the interfacial boundary would be comparable to the radius
of curvature. It is then inappropriate to
extrapolate the normal notion of surface
tension to such small clusters [23]; since
the surface tension arises from the steep
discontinuity between phases, it should be
lower at high curvatures [24]. The surface
width of the liquidvapor transition region
has been calculated, and for simple liquids
is about seven or eight molecular diameters [25]. This means that the thermodynamic properties of a system composed
of, for example, argon vapor and argon
clusters smaller than 300 molecules will
differ from those of a system composed of
523
524
bulk liquid and vapor phases at the same

temperature. This is the basis of the diffuse drop model [26], according to which
the nucleus presents uniform properties at
low supersaturations, with interfacial energies independent of radius of curvature.
At increasing supersaturation, the work
of formation decays progressively, the
interface becomes diffuse, and the composition of the nucleus becomes nonuniform
throughout. At even higher supersaturation, the composition at the center of the
nucleus approaches that of the external
phase; the radius rst decreases but then
passes through a minimum and then diverges, as the work of formation tends to
zero at very high supersaturation. Hence,
unless the interfacial energy can be determined, the Gibbs energy of formation
of small clusters, needed for the estimation of the nucleation rate, will remain
unknown. In an experimental study of the
electrochemical nucleation of metals from
molten salts, Hills and coworkers [27] observed that the radius of critical nuclei
passed through a minimum as the supersaturation increased. The diffuse liquid
drop model predicts this behavior, but
its adoption shifts the problem from determining the surface tension of small
clusters to the no simpler task of calculating the Gibbs energy of formation of
a nonhomogeneous phase. An alternative
description is to introduce in the classical
expression of the energy of formation of
the surface, Eq. (11), the surface tensions
between the liquid and the supersaturated
phase at the different overpotentials [28],
instead of the equilibrium surface tension
between the liquid and the supersaturated
phase. This is a remarkable result; extrapolation of macroscopic properties of
materials to small clusters conducts to reasonable agreement between experimental
data and the kinetic equations derived

from the classical theory of nucleation.
Microscopic Approach to Phase
Formation: The Atomistic Model
As discussed in the previous section, the
liquid drop model confronts serious difculties in describing the nucleation rate
when the critical size is of just a few atoms,
as is commonly found in experimental
studies of electrochemical nucleation. For
such small clusters, not only is it inappropriate to use the macroscopic surface
energy to calculate the nucleation rates,
the concept of interfacial energy itself is
open to question. For instance, all atoms
are on the surface in clusters smaller than
six atoms. On this basis, Walton [29] developed an expression for the concentration
of clusters using statistical mechanics,
avoiding the concept of surface energy
and deriving the nucleation rate from
purely kinetic considerations. Waltons treatment follows closely the classical work of
Volmer [30] but using partition functions
instead of classical thermodynamics to determine the equilibrium concentration of
clusters. The expression obtained for Cn
is [29]
n

C1
En
Cn
=
exp
(34)
N0
N0
kT
5.3.3.3
where N0 is the number density of

adsorption sites on the surface, which is
assumed to be much higher than C1 , the
number density of adsorbed atoms, and
En = n n1 is the energy of formation
of the cluster, taking as zero the potential
energy of the adsorbed atom. En is found
minimizing the formation energies of all
possible congurations of n atoms in
all possible orientations, which may be
accomplished, for example, by trial and
error. The nucleation rate is calculated
assuming that all clusters larger than
critical are immediately extracted from

the system; the critical nucleus is dened
as that cluster for which the probability
of growing is less than or equal to
1/2 but, upon incorporating one atom,
attains a probability of growing larger
than 1/2. The probability of supercritical
clusters decaying and turning subcritical
is considered as very slim and is therefore
neglected. The nucleation rate is then
expressed as
n
Ra2
n
a

[(n + 1)Qad + En QD ]
exp
kT
(35)
where R is the incidence rate of atoms
to the surface, n is the capture cross
section of critical nuclei, a is the distance
between adsorption sites, is an attempt
frequency, Qad is the bond energy with the
surface, and QD is the activation energy for
surface diffusion. Since n can only adopt
integer values, the same size of critical
nucleus should operate over a range of
temperatures, and discontinuities in the
temperature-dependence of the nucleation
rate are to be expected upon changes
of the critical size along temperature
intervals. Regarding the clustering process
as an innite chain of intermediates
with relative concentrations determined
by their respective rates of growth and
decay [31] avoids altogether having to
consider the notion of critical nucleus but,
unfortunately, does not lead to analytical
expressions for the nucleation rate without
a number of additional assumptions,
although it is possible to obtain explicit
expressions for certain limiting cases [32].
Stoyanov [6] reexamined Waltons atomistic theory of heterogeneous nucleation
and derived the nucleation rate avoiding
J =R
both the notion of critical nucleus as well

as that of equilibrium concentration of
clusters, starting from the expression for
the nucleation rate obtained by Becker and
Doring [4], that is,
+1 C1
Js =
1+
m

i=2
2 3 i
+2 +3 +i
(36)
where +i and i are the frequencies

of attachment and detachment of atoms,
respectively, from a cluster of i atoms,
given by
+i = i

QD
Ci
exp
N0
kT
and

(Ei Ei1 + QD )
i = i exp
kT
(37)
where i is the number of ways of forming
a cluster of size i + 1 incorporating one
ad-atom to a cluster of size i and i is the
number of ways an atom can detach from
a cluster of size i. Given that for each birth,
i i + 1, there is a corresponding death,
i + 1 i, i and i are related by
i = i+1
(38)
The sum in the denominator in Eq. (36) is

taken to an arbitrary size m that may be
dened as, for example, the minimal size
experimentally observable. Thus Js turns
into the rate of conversion of the largest
invisible clusters into detectable clusters
and, since the rate in the steady state is
equal for clusters of all sizes, the size of
the critical nucleus does not need to be
dened. Substituting Eqs. (37) and (38) in
Eq. (36), the steady state nucleation rate is
525
526
obtained as

i
m

Js = +1 C1
(j N0
1+
i=2 j =2
exp(Ej /kT )
exp(Ej 1 /kT )/j +1 C1
(39)
Since E1 = 0 and 1 = 2 , the product in

the denominator may be expanded, and
+1 C1

i1
m

Ei
2 N0
1+
exp
C1
kT
i=1 i
(40)
Equation (40) reduces into Eq. (35) if any
of the terms in the sum in the denominator
is much larger than others, in which case
the notion of critical nucleus is recovered.
The rate of electrochemical phase formation according to the atomistic model
was derived by Milchev and coworkers [7]
considering that the frequencies of attachment of atoms to form stable clusters of n
atoms and detachment to form unstable
clusters of n 1 atoms, is given by
Js =
+n 1 = k+n 1 c

Un 1 (1 e )ze0 (E0 + )
exp
kT
(41)
and
n =
k
Un + e ze0 (E0 + )
exp
kT
(42)
where Un 1 and Un are the energies
of transfer of electrodepositing ions from
solution to clusters and vice versa, c is
the concentration of electrodepositing ions
in solution, e is the electronic transfer
coefcient, e0 is the charge of the electron,
and E0 is the equilibrium potential of the
electrodeposition reaction. According to

Eq. (41), Un Un 1 = n represents
the separation work of an atom from the
cluster of n atoms. Furthermore, if the
number density N0 of sites available for
adsorption of atoms on the surface is in
large excess with respect to the number
density N1 of adsorbed atoms, then this
latter is given by

ze
N0 N1
0
exp
(43)
N1 =
N0
kT
so that

kn 1
n
n
=
exp
+n 1
k+n 1 c
kT

ze0 (E0 + )
exp
(44)
kT
and the nucleation rate is expressed as

Un (1 e )ze0 E0
Js = ZcN0 exp
kT

%(n )
exp
kT

(n + 1 e )ze0
exp
(45)
kT
where Z contains the ratio of products of
frequencies of attachment and detachment
from clusters up to the critical and %(n) =
n1/2 ;i represents the difference in
energy of n atoms when they are part of an
innitely large crystal on the one hand, and
when they form an independent cluster on
the electrode surface on the other, that is,
%(n) represents the surface energy of the
cluster of n atoms.
5.3.4
Experimental Studies of Electrochemical

Phase Formation
Metal electrodeposition occurs at the interface between an electronically conducting substrate and an ionically conducting
electrolyte. The sequence of events leading

to the formation of a new phase may include the adsorption of metal ad-atoms,
as well as two-dimensional or threedimensional nucleation, followed by the
growth of the bulk deposit. The particular mechanism for the growth of
the new phase on the conductive substrate is determined by a balance of the
interactions between ad-atoms and the
substrate, <(Meads S), and the crystallographic compatibility of the substrate
with the bulk metal deposit [1]. Under
quasi-equilibrium conditions, in which
low supersaturations and the absence of
kinetic limitations might be assumed,
phase formation occurs through two fundamental mechanisms: if the energy of
metal|substrate interactions dominates,
<(Meads S) < <(Meads Me), then the
concentration of ad-atoms Meads is small
and the formation of three-dimensional
clusters is favored. On the other hand,
if <(Meads S) > <(Meads Me), then the
degree of commensurability between the
crystalline habits of the substrate and the
adsorbate determines if the new phase
grows as epitaxial bidimensional layers or
as three-dimensional nuclei.
Three-dimensional Nucleation
Experimental studies of electrochemical
phase formation may be carried out by direct observation of the number of nuclei
on the electrode surface at different stages
of the growth process [33, 34], or by relating the electrical current with the number
density of nuclei on the electrode [8, 35],
in which case it is necessary to establish
also the rate of growth of the deposit. In
many cases of electrochemical phase formation reactions, particularly during the
electrodeposition of metals from molten
salts [36], or aqueous solutions, [37], the
charge-transfer step is fast and the rate
5.3.4.1
of growth of mature, supercritical nuclei,

are well described by the mass transfer
of the electrodepositing species to growth
centers. Nuclei are small and they may
be regarded as ultramicroelectrodes [38];
the diffusion-controlled current ux to an
isolated hemisphere of radius r0 on the
surface is [8, 39]
I = 2zF Dcr0
(46)
Where zF is the molar charge transferred

during the electrodeposition process and
D is the diffusion coefcient of the
electrodepositing species in the bulk of
the solution. The same current also causes
the growth of the deposit, thus, following
Faradays law:
I =
zF dV
M dt
(47)
zF
dr
2r02
M
dt
(48)
and accordingly,
2zF Dcr0 =
from which the radius of an isolated

nucleus of age t is given by

2DcMt
(49)
r0 =
and the growth current of an isolated

nucleus as a function of time is
I (t) =
zF (2Dc)3/2 M 1/2 t 1/2

1/2
(50)
In general, experimental studies of the

kinetics of formation of metallic phases
on electrodes are performed potentiostatically, stepping the potential from a positive
to a negative value with respect to the reversible potential of the electrodeposition
reaction. Under potentiostatic conditions,
it is generally assumed that both the steady
state nucleation rate Js and the growth rate
527
528
of nuclei will remain a constant, with rates

depending on the applied overpotential.
The value A of the steady state nucleation
rate can be obtained from analysis of the
current, which represents the overall rate
of the electrodeposition process. For a real
surface at constant overpotential, the rate
of formation of nuclei over a given number of active sites N0 may be expressed
as [12]
dN
= (N0 N )
dt
(51)
Integration with initial condition N = 0 at

t = 0 gives the number density of nuclei
as a function of time:
N = N0 [1 exp(At)]
(52)
Typical values of N0 are 104 cm2 <

N0 < 1010 cm2 , frequently dependent of
the potential but always smaller than
the number density of atoms on the
surface, of ca. 1015 cm2 [40, 41]. For
small values of A ( 1/t), Eq. (52) reduces
to N = N0 At and the nucleation process
is said to be progressive, while for large
values of A, N = N0 and the saturation
number density of nuclei is achieved
immediately after applying the potential
step, and nucleation is instantaneous.
Under conditions of multiple nucleation,
the current density can be expressed
as the convolution of the growth rates
of individual nuclei at time t after
having born at time u (Eq. 50) and
their appearance on the surface, dN/du =
AN0 exp(Au), Eqs. (51 and 52) [9]:
i(t) =
zF (2Dc)3/2 M 1/2
1/2
t
dN
(t u)1/2
du
dt
0
(53)
The solution of Eq. (53) may be expressed as

i(t) =
zF (2Dc)3/2 M 1/2 N0 t 1/2

% (54)
1/2
where
eAt
%=1
(At)1/2
(At)1/2
e d
(55)
%(At)1/2 is Dawsons integral, a tabulated function [42]. For, At 20, % 1

and Eq. (54) describes the current density
corresponding to instantaneous nucleation, proportional to N0 t 1/2 ; for At 0.2,
% (2/3)At and the current density is
proportional to (2/3)N0 At 3/2 , corresponding to the limiting case of progressive
nucleation.
Equation (54) and the limiting cases of
instantaneous and progressive nucleation are valid only during the initial stages
of the electrocrystallization process, when
the number and size of the nuclei are sufciently small so as for growth to occur independently from each other. The general description of the current transient requires
consideration of interactions among nuclei [912, 4348]. Initially, as discussed
above, nuclei grow independently of each
other and the process is controlled by
hemispherical diffusion to isolated nuclei the short time behavior. After
some time, the diffusion elds around nuclei overlap and eventually the growth of
the deposit is controlled by linear diffusion
towards the planar electrode surface the
long time behavior. To overcome the
difculties encountered in describing this
many-body problem, some approaches
have taken into account the overlap of
three-dimensional diffusional elds using
Avramis theorem [13], either considering
slices [45] or two-dimensional projections
on the electrode surface in the form of
circular diffusion zones [9, 11, 12, 43,

44]. Other treatments have avoided the
Avrami approach altogether, considering
mean eld approximations [46] or using
statistical mechanics to calculate nucleus
growth [48], but restricting consideration
to only the limiting case of instantaneous
nucleation. Other developments [49, 50]
have considered the changes in concentration occurring around nuclei during their
diffusion-control growth and the effect of
these changes on the nucleation rates and
on the spatial distribution of nuclei, but
do not discuss how to determine the nucleation rate A and the number density of
active sites N0 from experimental data.
Models allowing to determine A and
N0 obtain the current transient from
the material ux to free, noninteracting, growth centers, considering circular
diffusion zones around them, with timedependent radii rd . As shown in Fig. 4,
these are two-dimensional projections of
three-dimensional elds that dene, for
a hemispherical nuclei of radius r0 , an
equivalent area toward which the same
amount of matter that diffuses spherically
to a three-dimensional nucleus diffuses by
planar diffusion.
The mass balance results in
3/2
= (2)
cM
1/2
(t u) (57)
However, nuclei do not grow independently of each other and their interactions
results in overlap of the diffusion elds
around them. Overlap may be accounted
for using Avramis theorem, [13]
d = 1 exp(ex )
(58)
where d is the fraction of electrode surface

covered by planar diffusion zones and ex
is the extended coverage, that is, the
fraction of surface that would be covered
in absence of overlap,
ex =
3/2 (2Dc)3/2 M 1/2

AN0
1/2
t
(t u) exp(Au) du
(59)

ex = 2
(56)
(a)

rd2
that after integration yields
zF (2Dc)3/2 M 1/2 (t u)1/2

=
1/2
zF Dc(rd2 )
D(t u)1/2
where the term at the left is the current to

a single hemispherical nucleus, Eq. (52),
and the right-hand side term is the Cottrell
current for a plane electrode of equivalent
area rd2 , generated at the moment of birth
of the nucleus. Accordingly, the radial
ux (ux density area) equated to the
planar ux to the equivalent diffusion
zone, denes its area as [12]
2MDc
1/2
N0 (Dt)1/2 t 1/2 >
= At>
(b)
(60)
(c)
(a) Hemispherical diffusion elds; (b) projection to the

electrode surface as planar diffusion zones; and (c) their
overlap on the plane of the surface.
Fig. 4
529
530

cM 1/2
3/2
1 exp (2) D

1 eAt
(66)
N0 t 1
At
where = (2)3/2 D(cM/)1/2 N0 /A and

>=1
(1 eAt )
At
(61)
The coverage of the electrode surface with

diffusion zones, considering overlap, is
then given by the following expression:
d = 1 exp[(At 1 + eAt )] (62)
The instantaneous limit corresponds to
the 0 limit, high nucleation rate A on
small number density of active sites N0 ,
while the progressive situation results
when the value of A is low and/or N0
is elevated, that is, when . The
current density is obtained as the planar
diffusive ux to an electrode of fractional
area d :
zF Dc
i(t) =
(63)
d
The length of the diffusion layer is obtained from the balance between Eqs. (54
and 63) [11]
zF Dc
zF (2Dc)3/2 M 1/2 N0 t 1/2
%=
ex
1/2
(64)
as
>
= (Dt)1/2
(65)
%
According to this expression [11], the
rate of expansion of the diffusion layer
depends on the nucleation rate A, but
not on the density of active sites. For
instantaneous nucleation, = (Dt)1/2 ,
whereas for progressive nucleation, =
(3/4)(Dt)1/2 . From Eqs. (54, 60 and 64),
the current density is described as [11, 12]
i(t) =
zF Dc
(Dt)1/2
(At)1/2
eAt
2
e d
1/2
(At)
0

1 eAt
1
At
Two-dimensional
Electrocrystallization
Two-dimensional phase formation occurs
preferentially when a strong substratemetal interaction exists, a process that
typically involves the formation of growth
centers a few atoms thick, that expand and
coalesce to form a monolayer that serves
as a precursor deposit to subsequent twodimensional metal layers.
As discussed before, for a determinate
supersaturation level
, ion transfer
from solution to electrode across the electrochemical double layer occurs accompanied by the corresponding free energy
change,
GV (n) = nze||. A fraction
(n) of this energy is consumed in the
formation of the new nucleus|solution
and nucleus|surface interfaces, giving
for the total change of Gibbs free energy associated with the formation of a
cluster comprising n atoms,
GV (n) =
nze|| + (n). This expression has a
minimum
G for the equilibrium form
of the crystal with the lower surface energy i among the i crystalline faces, as
determined by the GibbsCurie expression [51], = ;i i Ai = min (for constant
volume V ).
In the case of epitaxial grow over
an ideal substrate, the surface energies
of the two-dimensional clusters reduce
to border energies i (J cm1 ), as the
emergence of the new phase is balanced
by disappearance of substrate area. Under
these conditions, the 2D crystal satises
the GibbsCurie condition for the excess
free energy, in analogy to the threedimensional case; = ;i i Li = min (at
5.3.4.2
constant area), where Li is the border

length of the 2D crystal. By appropriate
considerations for simple geometrical
forms [52], the number of atoms of the
critical 2D nucleus may be obtained from
n =
b2
(ze||)2
(67)
where b is a geometrical factor, the area

occupied by an atom in the cluster surface,
and the mean value of the border energies
for the i crystalline faces, from which
the free energy of the critical 2D nuclei
is
G = ze||n The VolmerWeber
expression applied to the two-dimensional
case gives

b2
(68)
J = A2D exp
ze||kT
where A2D may be considered constant, although it encloses the potential-dependent
contributions coming from the Zeldovich
factor Z and the probability of attachment
to critical nuclei n n .
Analysis of ln(J ) versus 1/|| relationships from experimental data allows to
determine the nucleation parameters n
and
G for the critical nucleus, through
the relation [52]
d ln J
zF
1 d
G
=
= n
d||
KT d||
RT
(69)
The description of the current transient

arising from nucleation and growth of
two-dimensional monolayers, considers as
limiting step the attachment of atoms or
molecules to the edge of the growing twodimensional centers. The required charge
to form a disc-shaped nuclei of radius r
and height h is Q(r) = zF r(t)2 h/M,
from which the instantaneous current
i(r, t) = zF 2r(t)h/M)dr(t)/dt may be
obtained differentiating with respect to
time. If k (in moles cm2 s1 ) is the
rate of attachment of atoms to the growth

center, the radial expansion rate would
be dr(t)/dt = Mk/ and the temporal
dependence of the radius r(t) = Mkt/.
These expressions lead to the following
equation for the current associated with the
growth of an isolated disc-shaped nucleus:
i2D (t) =
2nF k 2 hM
t
(70)
As previously considered with respect

to the three-dimensional nucleation situation, the expression for the current
transient corresponding to multiple twodimensional nucleation requires consideration of the coalescence of growth centers,
which diminishes the edge length available for attachment of new atoms or
molecules, and produces a decay of the current. By means of the Avrami theorem (see
Sect. 5.3.4.1), the following expressions are
obtained for the limiting cases of instantaneous and progressive two-dimensional
nucleation:
I2D,inst =
I2D,prog =
2zF MhN0 k 2 t

N0 M 2 k 2 t 2
exp
(71)
2
zF MhN0 k 2 t 2

AN0 M 2 k 2 t 3
exp
3 2
(72)
Spatial Distribution of Nuclei

The fundamental reason to study nucleation kinetics under potentiostatic
conditions is to maintain constant supersaturation and hence a constant nucleation
rate, except for nonstationary effects occurring at the onset of the potential step.
However, the irreversible growth of the
5.3.4.3
531
nuclei diminishes the concentration of

electrodepositing species and inhibits the
nucleation rate in their vicinity. Thus the
number of nuclei that ultimately attain
stable growth over the surface depends on
the local variations of the nucleation rate,
as a result of changes in concentration
occurring during the growth process. At
sufciently high overpotentials and in the
presence of excess supporting electrolyte,
the local nucleation rate at a given moment t and at a distance r from the center
of a nucleus of age = (t u) and radius
r0 = (2DcM/)1/2 1/2 is given by [53]
Thus, the probability of forming a

nucleus somewhere in the electrode surface at a given moment t is affected by
the presence of a nucleus i located at
a distance ri ; this may be expressed as
pi (ri , t) = A(ri , t)/A0 . If the effects of
several nuclei in the vicinity of the site
were independent, then the joint nucleation probability would be expressed as
p(ri , t) = Dpi (ri , t), but the radial symmetry of the diffusional hemispherical
elds generated by the growth of nuclei
is robust [54], thus the concentration of
electrodepositing species is determined by
the nucleus that exerts the more intense
diffusional eld on the particular site, acting then as the most inuential neighbor,
and superseding the effects induced by
other smaller or more distant neighbors.
Then the local nucleation rate is not determined by the combined effects of all
the nuclei surrounding the particular site,
but only by the most inuential neighbor [50]. This most inuential neighbor
A(r, t)

n +1
(2DcM/)1/2 1/2
(73)
= A0 1
r
where A0 is the noninhibited nucleation
rate at a sufciently large distance from a
growing nucleus. The local variation of the
nucleation rate due to the growth of nuclei
is shown in Fig. 5.
rd/r0
rd /r0
1.0
n* = 0
0.8
A(r, t)/A0
532
2
0.6
5
0.4
10
0.2
0.0
20
10
10
20
r/r0(t)
Local nucleation rate around a growing nucleus of radius r0 ,
for different critical nucleus sizes. Also shown is the exclusion zone of
radius rd , dening regions inside (A = 0) and outside (A = A0 ) the
exclusion zone.
Fig. 5
and its effects on the nucleation process

may be identied dening a circular zone
around each nucleus, inside which formation of further nuclei is excluded. The
overall rate of nucleation will diminish as
a result of the decreased supersaturation
caused by the growth of nuclei, and will
eventually vanish as the exclusion zones
expand and overlap, covering all the area
available. In Fig. 5, the size of the exclusion
zone has been identied with that of the
diffusion zones considered on the current
transient analysis. Under this assumption,
the saturation number density of nuclei on
the electrode surface is given by

NS = AN0
exp (2)3/2 D
cM
1/2
N0
(Au 1 eAu )
A
eAu du
(74)
Evaluation of NS from Eq. (74) requires

numerical methods, but for the particular case of progressive nucleation, that is, when nucleation is slow
and occurs on a large number density of active sites, then [55] NS =
1/2

. Values of NS
AN0 /[2(8cM/)1/2 D]
obtained from this expression agree favorably with experimental number densities
of Pb, Ag, and Hg nuclei obtained by direct microscopic observation of electrode
surfaces [5658].
Local variations of the nucleation rate affect not only the saturation number density
of nuclei, but also their spatial distribution [59, 60], and this can be studied analyzing images of the electrode surface. The
spatial distribution of nuclei is represented
in Fig. 6 by the probability density of the
distances between nearest neighbors, in
nondimensional, normalized form. For
(2pNs)1/2p(r)
1.0
0.8
0.6
0.4
0.2
0.0
(2pNs)1/2r
Normalized probability density of nearest neighbor distances
for 2.2 106 drops/cm2 of mercury, electrodeposited onto vitreous
carbon at 220 mV from 0.01 mol dm3 Hg2+ aqueous solution
(
) [58]. Also shown are the probability densities corresponding to
uniformly distributed drops (thin line) and excluded nucleation for
distances less than rd = [(8 cM/)1/2 D(t u)]1/2 from each drop
(heavy line), obtained from digital simulations [59].
Fig. 6
533
534
uniformly distributed nuclei, the nearest

neighbor distribution can be obtained as
follows [56]. The probability of nding the
nearest neighbor at a distance r from a
given nucleus, P (r), can be expressed as
the product of the probability of not nding one at a distance less than r, and the
probability of nding one between r + dr:

r
dP (r) = p(r)dr = 1
p(u) du
0
2NS r dr
(75)
This may be now expressed as dP (r)/[1

P (r)] = 2NS rdr, and after integration
becomes P (r) = 1 exp(NS r 2 ). Thus,
p(r)dr = 2NS r exp(NS r 2 )
(76)
The probability density in Eq. (76) has maximum value for rm = 1/(2NS )1/2 , and we
may express it in nondimensional form in
terms of this characteristic distance, as
(2NS )1/2 p(r) = (2NS )1/2 r
exp( 12 [(2NS )1/2 r 2 ])
(77)
This normalized distribution does not depend on the number density of nuclei on
the electrode surface, and only denotes
the ordering pattern of nuclei on the electrode surface. If nuclei were, for example,
ordered on a square array, then the distance between nearest neighbors would
1/2
be 1/NS and the probability of nding the nearest neighbor would change
from 0 to 1 at a nondimensional distance
1/2
NS /(2NS )1/2 = (2)1/2 = 2.507. In
a hexagonal array, all nearest neighbors are
separated by a nondimensional distance
(4/ 3)1/2 = 2.694. For a random uniform distribution, the probability density
is maximum at the characteristic distance,
that is, (2NS )1/2 /(2NS )1/2 = 1. In
nondimensional coordinates, the maximum of the distribution shifts to higher
distances as the ordering of the nuclei on

the plane increases. As is shown in Fig. 6,
inhibition of nucleation in the vicinity of
growing nuclei induces a certain degree of
ordering, which may be advantageous, for
example, in the preparation of catalysts to
reduce sintering and to assist accessibility
of the catalytic sites by reacting species.
In general [50, 5658], electrodeposited
nuclei at low overpotentials are uniformly
distributed, reecting a random distribution of active sites on the electrode surface,
while at higher overpotentials the spatial
distribution of nuclei is affected by inhibition of the nucleation rate in their vicinity.
Experimental Examples
Mercury electrodeposition is a model system for experimental studies of electrochemical phase formation. On the one
hand, the product obtained is a liquid drop,
corresponding very well with the liquid
drop model of classical nucleation theory.
Besides, electron transfer is fast [61] and
therefore the growth of nuclei is controlled
by mass transport to the electrode surface [44]. On the other hand, the properties
of the mercury|aqueous solution interface
have been the object of study for over a
century and hence are fairly well understood. The high overpotential for proton
reduction onto both mercury and vitreous
carbon favor the study of the process over
a wide range of overpotentials. In spite
of the complications introduced by the
equilibrium between the Hg2+ , Hg2 2+ ,
and Hg species, this system offers an
excellent opportunity to verify the fundamental postulates of the electrochemical
nucleation theory. In fact, the dependence
of the nucleation rate on the oxidation
state of the electrodepositing species is
fully consistent with theory: critical nuclei appear with similar sizes and onto
similar number densities of active sites
5.3.4.4
either from Hg2 2+ or Hg2+ solutions, as

expected from the identical interactions between deposit and substrate in both cases.
This demonstrates the dominant role of
the interaction between the surface and
the deposit in the phase formation process, as discussed before. Fig. 7 shows a
family of current transients obtained for
mercury electrodeposition from Hg(I) solution onto vitreous carbon at different
potentials.
A similar family of theoretical current
transients, obtained from evaluation of
Eq. (66) with the known values of z, F , c,
M, and corresponding to the reduction
of Hg(I) to Hg(0), is shown in Fig. 8.
As shown in both the experimental and
theoretical transients, the currents rst
increase as a result of the formation of mercury drops and their growth controlled by
radial diffusion, decaying at longer times
due to the onset of semi-innite planar diffusion. The diffusion coefcients may be
obtained either from i versus t 1/2 plots
for sufciently long times, at which linear

dependence is observed in accordance
with Cottrell behavior, or by hydrodynamic
methods using rotating electrodes. In the
particular case of mercury, nucleation may
be realized also from Hg(II) solutions, and
Fig. 9 shows the nucleation rates as a function of the overpotential from both Hg(I)
and Hg(II) solutions, obtained from the
analysis of current transients.
According to Eq. (45), the atomistic
theory predicts a linear dependence of
the nucleation rate with overpotential for
constant size of the critical nucleus, as
shown in Fig. 7. Also, in accordance with
Eq. (45), the value of the slope doubles
for mercury deposition from Hg(II) as
compared with deposition from Hg(I).
The slopes of logarithmic plots of the
nucleation rate as a function of the
overpotential, such as that shown in
Fig. 9, conrm the small sizes of critical
nuclei in electrochemical phase formation
processes. Frequently, as suggested by the
2.5
250
240
230
i
[mA cm2]
2.0
220
210
1.5
200
190
1.0
0.5
0.0
10
15
20
25
30
t
[s]
Fig. 7 Current transients recorded during Hg electrodeposition onto
vitreous carbon from 0.01 mol dm3 Hg2 2+ solutions in KNO3
1 mol dm3 at the indicated overpotentials (in mV) [58].
535
2.5
g
f
i
[mA cm2]
2.0
e
d
c
1.5
b
a
1.0
0.5
0.0
10
15
20
25
30
t
[s]
Current transients according to Eq. (66), for mercury
electrodeposition from 10 mM Hg2 2+ solutions, with D = 105 cm2 s1 ,
at nucleation rates A and number densities of active sites N0 ,
respectively, of: (a) 0.02 s1 and 0.4 106 cm2 ; (b) 0.04 s1 and
0.6 106 cm2 ; (c) 0.06 s1 and 0.8 106 cm2 ; (d) 0.08 s1 and
1.0 106 cm2 ; (e) 0.10 s1 and 1.2 106 cm2 ; (f) 0.12 s1 and
1.4 106 cm2 ; and (g) 0.14 s1 and 1.6 106 cm2 . The broken line
indicates the Cottrell diffusive limiting current to the electrode surface.
Fig. 8
102
101
100
A
[s1]
536
101
102
103
0.15
0.20
0.25
0.30
h
[V]
Nucleation rates of Hg onto vitreous carbon from 0.01 mol dm3
Hg2 2+ () and 0.01 mol dm3 Hg2+ (
) solutions, as a function of the
overpotential [58].
Fig. 9
slopes in Fig. 9, single atoms adsorbed on

the surface grow irreversibly and play the
role of critical nuclei.
Usually, only a small fraction of the surface is active for nucleation, the number
density of active sites is potential dependent, and much less than the atomic
density of the surface [40]. Analysis of
transients by means of Eq. (66) allows
also obtaining the number densities of
sites on the electrode surface. As shown
in Fig. 10, nucleation of mercury either
from Hg2 2+ or Hg2+ solutions takes
place at a given overpotential onto a
similar number of sites, as would be
expected from the identical interactions
between the electrodeposited atoms and
the substrate in both cases. This observation reafrms the dominant role of the
surface|deposit interactions in phase formation phenomena.
We have presented here some of the
fundamental aspects of electrochemical
phase formation processes, and some

illustration has been given by way of
a specic example, the electrodeposition of mercury. Many other theoretical and experimental studies show that
electrochemical methods may be advantageously used for the electrodeposition
of a wide variety of materials. Electrochemical phase formation offers possibilities for controlling the chemical nature,
size, morphology, and spatial distribution of dispersed and nanostructured
deposits of a variety of materials, including metals [62, 63], semiconductors [64],
or conducting polymers [65]. Controlled
monodisperse deposition of materials at
nanoscopic scales, which has been shown
to be achievable with electrochemical
methods, is necessary for optimal performance of a variety of devices and
processes, such as, for example, quantum dots or clean electrochemical energy
converters.
N0
[cm2]
107
106
105
0.15
0.20
0.25
0.30
h
[V]
Number densities of nucleation sites on the surface of vitreous
carbon as a function of the overpotential, obtained from analysis of current
transients obtained during mercury deposition from 0.01 mol dm3 Hg2 2+
() and 0.01 mol dm3 Hg2+ (
) aqueous solutions [58].
Fig. 10
537
538
References
1. E. Budevski, G. Staikov, W. J. Lorenz, Electrochemical Phase Formation and Growth,
2. W. Thomson, Proc. R. Soc. Edinb. 1870, 7, 63.
3. J. W. Gibbs, Collected Works (Thermodynamics), Yale University Press, New Haven, 1948,
Vol. I.
4. R. Becker, W. Doring, Ann. Phys. 1935, 24,
719.
5. J. B. Zeldovich, Acta Phys. Chem. USSR 1943,
18, 1.
6. S. Stoyanov, Thin Solid Films 1973, 18, 91.
7. A. Milchev, S. Stoyanov, R. Kaischev, Thin
Solid Films 1974, 22, 255.
8. G. J. Hills, D. J. Schiffrin, J. Thompson, Electrochim. Acta 1974, 19, 657.
9. M. Sluyters-Rehbach, J. H. O. J. Wijenberg,
E. Bosco et al., J. Electroanal. Chem. 1987,
236, 1.
10. B. R. Scharifker, J. Mostany, M. PalomarPardave et al., J. Electrochem. Soc. 1999, 146,
1005.
11. L. Heerman, A. Tarallo, J. Electroanal. Chem.
1999, 470, 70.
12. B. R. Scharifker, J. Mostany, J. Electroanal.
Chem. 1984, 177, 13.
13. M. Avrami, J. Chem. Phys. 1939, 7, 1103.
14. W. J. Basirun, D. Pletcher, A. SarabyReintjes, J. Appl. Electrochem. 1996, 26, 873.
1931, 157, 165.
16. T. Young, Philos. Trans. R. Soc. (Londres)
1805, 1, 65.
17. R. Aveyard, D. A. Haydon, An Introduction to
the Principles of Surface Chemistry, Cambridge
University Press, Cambridge, 1973, p. 74.
18. M. Volmer, A. Weber, Z. Phys. Chem. 1926,
119, 277.
19. L. Farkas, Z. Phys. Chem. 1927, 125, 236.
21. E. Parzen, Stochastic Processes, Holden-Day,
San Francisco, 1962.
22. J. Frenkel, Kinetic Theory of Liquids, Clarendon Press, Oxford, 1946.
23. E. A. Guggenheim, Trans. Faraday Soc. 1940,
36, 408.
24. R. C. Tolman, J. Chem. Phys. 1949, 17, 333.
25. K. S. C. Freeman, I. R. McDonald, Mol. Phys.
1973, 26, 529.
26. J. W. Cahn, J. E. Hilliard, J. Chem. Phys.
1959, 31, 688.
27. G. J. Hills, D. J. Schiffrin, J. Thompson, Electrochim. Acta 1974, 19, 671.

28. R. McGraw, H. Reiss, J. Stat. Phys. 1979, 20,
385.
29. D. Walton, J. Chem. Phys. 1962, 37, 2182.
30. M. Volmer, Z. Elektrochem. 1929, 35, 555.
31. G. Zinsmeister, Vacuum 1966, 16, 529.
32. R. M. Logan, Thin Solid Films 1969, 3, 59.
33. R. Kaischew, B. Mutaftchiew, Electrochim.
Acta 1965, 10, 643.
34. S. Toschev, A. Milchev, K. Popova et al., C.R.
Acad. Bulg. Sci. 1969, 22, 1413.
35. M. Fleischmann, H. R. Thirsk, Electrochim.
Acta 1960, 2, 22.
36. F. Lantelme, J. Chevalet, J. Electroanal. Chem.
1981, 121, 311.
37. B. R. Scharifker, G. J. Hills, J. Electroanal.
Chem. 1981, 130, 81.
38. B. R. Scharifker in Modern Aspects of Electrochemistry (Eds.: J. O. M. Bockris, B. E. Conway, R. E. White), Plenum Press, New York,
1992, p. 467, Vol. 22.
39. S. Fletcher, J. Chem. Soc., Faraday Trans. 1
1983, 79, 467.
40. J. Mostany, J. Mozota, B. R. Scharifker, J.
41. E. Michailova, A. Milchev, J. Appl. Electrochem. 1990, 21, 170.
Mathematical Functions, Dover Publications,
New York, 1965, p. 298.
43. G. Gunawardena, G. Hills, I. Montenegro
44. B. Scharifker, G. Hills, Electrochim. Acta
1983, 28, 879.
45. E. Bosco, S. K. Rangarajan, J. Electroanal.
Chem. 1982, 134, 213.
46. P. A. Bobbert, M. M. Wind, J. Vlieger, Physica A 1987, 146, 69.
47. M. V. Mirkin, A. P. Nilov, J. Electroanal.
Chem. 1990, 283, 35.
48. M. Tokuyama, Physica A 1990, 169, 147.
49. A. Milchev, W. S. Kruijt, M. SluytersRehbach et al., J. Electroanal. Chem. 1993,
362, 21.
50. E. Garca-Pastoriza, J. Mostany, B. R. Scharifker, J. Electroanal. Chem. 1998, 441, 13.
51. P. Curie, Bull. Soc. Mineral. 1885, 8, 145.
52. R. Kaischev, Bull. Acad. Bulg. Sci. Phys. 1951,
2, 191.
53. W. S. Kruijt, M. Sluyters-Rehbach, J. H. Sluyters et al., J. Electroanal. Chem. 1994, 371,
13.

54. L. C. R. Alfred, K. B. Oldham, J. Electroanal.
Chem. 1995, 396, 257.
55. B. R. Scharifker, Acta Cient. Venez. 1984, 35,
211.
56. A. Serruya, J. Mostany, B. R. Scharifker, J.
Chem. Soc. Faraday Trans. 1993, 89, 255.
57. A. Serruya, B. R. Scharifker, I. Gonzalez
et al., J. Appl. Electrochem. 1996, 26, 451.
58. A. Serruya, J. Mostany, B. R. Scharifker, J.
59. B. R. Scharifker, J. Mostany, A. Serruya, Electrochim. Acta 1992, 37, 2503.
60. J. Mostany, A. Serruya, B. R. Scharifker, J.

61. P. Bindra, A. P. Brown, M. Fleischmann
62. J. L. Fransaer, R. M. Penner, J. Phys. Chem.
B 1999, 103, 7643.
63. G. Oskam, P. C. Searson, J. Electrochem. Soc.
2000, 147, 2199.
64. R. M. Penner, Acc. Chem. Res. 2000, 33, 78.
65. M. F. Suarez, R. G. Compton, J. Electroanal.
Chem. 1999, 462, 211.
539

Volume 2 Interfacial Kinetics and Mass Transport

Uploaded by

Copyright:

Available Formats

Volume 2 Interfacial Kinetics and Mass Transport

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Volume 2 Interfacial Kinetics and Mass Transport

Uploaded by

Copyright:

Available Formats

3

The CurrentPotential Relationship

section of this volume covers kinetics and

This rst chapter to Volume 2 Interfacial

The rst reaction is an electron transfer

which is proportional to the rate of the

Since we are dealing with heterogeneous

Figure 1 depicts a reproduction of the

1.1 The CurrentPotential Relationship

Thermodynamics of Electrochemical Cells

electrons transferred from the electrode,

The electrode potential can be expressed by

where M is the inner or Galvani potential

where O is the soluble oxidized and

where E is the standard potential of

where aO and aR denote the activities of O

The ButlerVolmer Equation

In this section, we shall derive an expression for the currentpotential relationship

the rate of which is considered to be

which, for reaction order one for reactant

where kf and kb are the corresponding

According to the absolute reaction rate

1.1 The CurrentPotential Relationship

Fig. 2 Schematic representation of the effect of electrode potential

For convenience, we divide the free

Gf (E) = Gf (Ee ) + F E

Thus from Eq. (16), the net current density

We note that at equilibrium i = 0

Eq. (27), we obtain the ButlerVolmer

1.1 The CurrentPotential Relationship

Plots of log |i/io | versus calculated with Eq. (29).

and since E = G /F , the Tafel dependence is a linear free-energy relationship

a linear i relationship for the ow of

which resembles the ohmic current

which is a measure of io and is also

that it cannot account for the curvature

Theory of Electron Transfer

the vibrational, i contributions:

Assuming the solvent as a dielectric

where qjr and qj are the equilibrium

1.1 The CurrentPotential Relationship

where wo and wR are the electrostatic

distance like in intramolecular electron

species to the electrode, x:

where ko is the preexponential rate coefcient and the tunneling coefcient,

rates of electron transport across organic

Fig. 4 (a) Schematic representation of

1.1 The CurrentPotential Relationship

Electrode reactions also characterized by exponential currentpotential

metal electrode unlike the inversion found

relationship are interfacial ion processes

that stabilizes the cation in solution. Note,

where a metal ion at the surface metal

therefore the total current that ows at the

1.1 The CurrentPotential Relationship

are also examples of interfacial ion-transfer

Multiple Electron Transfer

In general, reactions at electrodes involving more than one electron-transfer events

reorganization in the metal-ligand vibration scale, 1013 s or slower, the probability

1. The rst electron-transfer step is rds,

Gf (E) = Gf (Ee ) + F E

and since E = G /F , the Tafel dependence is a linear free-energy relationship

For small overpotentials, || RT /