Electrochimica Acta 211 (2016) 420–429

Contents lists available at ScienceDirect

Electrochimica Acta
journal homepage: www.elsevier.com/locate/electacta

Simulations and Interpretation of Three-Electrode Cyclic

Voltammograms of Pseudocapacitive Electrodes
Henri-Louis Girard a , Bruce Dunn b , Laurent Pilon a,∗
a
University of California Los Angeles, Henry Samueli School of Engineering and Applied Science, Mechanical and Aerospace Engineering Department,
420 Westwood Plaza, Los Angeles, CA 90095, USA
b
University of California Los Angeles, Henry Samueli School of Engineering and Applied Science, Material Science and Engineering Department,
410 Westwood Plaza, Los Angeles, CA 90095, USA

a r t i c l e i n f o a b s t r a c t

Article history: This study aims to provide physical interpretation of cyclic voltammograms obtained from pseudoca-
Received 7 April 2016 pacitive electrodes using three-electrode systems. It presents numerical simulations based on a recent
Received in revised form 11 June 2016 continuum model able to reproduce experimental CV curves obtained with different Nb2 O5 -based elec-
Accepted 13 June 2016
trodes and LiClO4 in propylene carbonate as electrolyte. First, the respective contributions of faradaic
Available online 23 June 2016
and electric double layer charge storage mechanisms were clearly identified along with the associated
faradaic and capacitive regimes. This was further illustrated by comparing CV curves for pseudocapacitive
(Nb2 O5 ) and EDLC (carbon) electrodes. Transition from the faradaic to the capacitive regime was caused
by Li+ starvation at the electrode/electrolyte interface and the formation of ClO− 4
electric double layer.
Moreover, the effects of electrode crystallinity on CV curves were reproduced and interpreted in terms
of enhanced transport properties for lithium intercalation and electrode electrical conductivity. Finally,
the electrode thickness featured an optimum corresponding to a compromise between accommodating
large amounts of intercalated lithium and minimizing the potential drop across the electrode to drive
faradaic reactions.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction For that reason, there has been substantial interest in the synthe-
sis and characterization of materials that exhibit pseudocapacitive
The most common electrochemical capacitors (ECs), gener- charge storage [2–5].
ally known as supercapacitors or electric double layer capacitors Pseudocapacitive materials have often relatively poor solid-
(EDLCs), store charge physically by the adsorption of ions in a state diffusion and electrical conductivity which limits the practical
thin double layer at the interface between the electrode, typically thickness of the electrode and its energy and power densities. In
carbon, and the liquid electrolyte [1]. It is well established that order to mitigate the slow intercalation and/or the potential drop
these carbon-based supercapacitors possess a number of attrac- across the electrode several approaches can be envisioned. First,
tive properties including fast charging, the delivery of significantly the thickness can be adjusted to reduce the electrode electrical
more power than batteries, and long cycle life. Pseudocapacitance resistance and the intercalation distance. The materials can also be
represents a second capacitive energy storage mechanism. Pseu- synthesized with different crystalline structures to achieve larger
docapacitors store charge both physically in the electric double electrical conductivity and faster intercalation [6]. Our recent work
layer (EDL) and chemically by changing the oxidation state of the with Nb2 O5 (orthorhombic form) established that upon lithium
electrode material [2–5]. insertion, this material was able to exhibit high levels of energy
The energy density associated with redox reactions can be 10 storage, nearly 120 mAh g−1 at a charging rate of 1 minute (nomi-
times greater than that of double layer processes. Thus, pseu- nally 60C) [5]. We characterized the electrochemical properties of
docapacitor materials offer the prospect of achieving the energy the Nb2 O5 and determined that the kinetics of charge storage was
densities of batteries combined with thepower density of EDLCs. unusually fast thanks a crystalline network offering easy transport
pathways. Another key feature with this material is that there is
no phase transformation upon lithiation [7]. Indeed, phase transi-
tions lead to slower and less reversible ion diffusion in the electrode
∗ Corresponding author. Tel.: +1 (310) 206 5598; fax: +1 (310) 206 2302. material indicated by the presence of additional peaks in the cyclic
E-mail address: [email protected] (L. Pilon). voltammetry curves [7].

http://dx.doi.org/10.1016/j.electacta.2016.06.066
0013-4686/© 2016 Elsevier Ltd. All rights reserved.
 H.-L. Girard et al. / Electrochimica Acta 211 (2016) 420–429 421

Electrochemical studies of pseudocapacitor materials are gen-

Nomenclature erally of the three-electrode type. In this configuration, the
pseudocapacitor material serves as the working electrode, with a
ai Effective ion diameter of species i (nm) porous carbon electrode as the counter electrode and a reference
b b-value, power constant in jt = a vb electrode that controls the potential at the working electrode. For
ci Ion concentration in the electrolyte of species i the research on Nb2 O5 , where we were investigating the insertion
(mol/L) and de-insertion of lithium ions, Li metal served as both the counter
ci,max Maximum ion concentration, ci,max = 1/NA a3 (mol/L) and the reference electrode [5]. The benefit of three-electrode stud-
Cs,int Integral areal capacitance (F/m2 ) ies is that one is able to characterize a single material, in contrast
c1,P Concentration of intercalated Li in the pseudocapac- to experiments with a traditional two-electrode device where the
itive electrode (mol/L) properties of a given electrode material are inferred. Our previous
c1,P,0 Initial concentration of intercalated Li simulation studies that were used to characterize electric double
c1,P,max Maximum concentration of intercalated Li, layer capacitors with asymmetric electrolytes or hybrid pseudoca-
c1,C,max = m /M (mol/L) pacitors were based on two-electrode configurations [8–12]. While
Di Diffusion coefficient of ions in electrolyte (m2 /s) these studies are extremely useful for providing insight on device
D1,P Diffusion coefficient of intercalated Li in the pseu- characteristics, they are not designed to contribute detailed under-
docapacitive electrode (m2 /s) standing of the electrode material, which is frequently the focus of
e Elementary charge, e = 1.602 × 10−19 C experimental studies on pseudocapacitors. Indeed, the current in
F Faraday constant, F = eNA = 9.648 × 104 C mol−1 two-electrode hybrid pseudocapacitors is controlled by the carbon
H Stern layer thickness (nm) electrode whereas the current in three-electrode measurements
jC Magnitude of predicted capacitive current density is only limited by redox reactions and interfacial and transport
(A/m2 ) phenomena through or near the pseudocapacitive electrode. Con-
jF Magnitude of predicted faradaic current density sequently, the equilibrium potential and the state of charge may
(A/m2 ) vary significantly and must be accounted for.
jT Magnitude of predicted total current density (A/m2 ) The objective of the present study is to connect numerical
k0 Reaction rate constant, Equation simulations directly to experimental studies by simulating three-
(14)(m2.5 mol−0.5 s−1 ) electrode configurations with a Nb2 O5 working electrode. Our
L Electrolyte thickness (nm) approach involves the use of numerical simulations based on a con-
LP Thickness of the pseudocapacitive electrode (nm) tinuum model previously described [10,11]. The results provide
m Stoichiometric number of intercalated Li, Lim Mp Oq physical interpretation of the experimentally measured cyclic
M Molecular weight of the fully intercalated active voltammograms obtained from three-electrode measurements on
electrode material (g/mol) Nb2 O5 electrodes and characterize the interplay between faradaic
nc Cycle number and EDL charge storage mechanisms. Furthermore, this study illus-
NA Avogadro constant, NA = 6.022 × 1023 mol−1 trates how the physical model and the associated simulation tools
Ni Ion flux of species i (mol m−2 s−1 ) can be used to optimize the design of individual pseudocapacitive
p, q Stoichiometric numbers of the metal and oxygen in electrodes.
the metal oxide Mp Oq
Ru Universal gas constant, Ru = 8.314 J mol−1 K−1
T Local temperature (K) 2. Analysis
t Time (s)
v Scan rate of the cyclic voltammetry (V/s) 2.1. Schematic and Assumptions
x Location in one-dimensional space (m)
zi Ion valency of species i Fig. 1 shows a schematic of the simulated domain and of the
associated coordinate system. The domain consisted of a pseudo-
Greek symbols capacitive electrode of thickness LP made of metal oxide Mp Oq and a
˛ Transfer coefficient, Equations (13) and (14) binary and symmetric electrolyte of thickness L containing lithium
0 Vacuum permittivity, 0 = 8.854 × 10−12 F m−1 salt in a solvent at high concentration. The electrolyte was further
r Relative permittivity of electrolyte divided in two regions (i) a Stern layer of thickness H defining the
 eq Equilibrium potential difference (V) distance of closest approach of the ions to the electrode and (ii)
 H Potential drop across the Stern layer (V) a diffuse layer where the ions were free to move under the com-
D Debye length (m) bined influence of electrodiffusion and steric repulsion [13]. The
 Overpotential,  =  H −  eq (V) origin x = 0 was located in the bulk of the electrolyte far from the
 Density of the fully intercalated pseudocapacitive electrode, i.e., L » D where D is the Debye length characterizing
electrode material (kg/m3 ) the thickness of the double layer. For a binary symmetric elec-
P Electrical conductivity of the electrode (S/m)  1/2
trolyte it is defined as D = 0 r Ru T/2F 2 z12 c1,∞ . The model
CV Cycle period (s)
used in this study was based on the generalized modified Poisson-
Electric potential (V)
Nernst-Planck (GMPNP) model for binary symmetric electrolyte
min , max Minimum and maximum of the potential win-
with faradaic reactions. This model and the underlying assumptions
dow (V)
were described in detail in Ref. [10] and need not be repeated. In
s Imposed cell potential (V)
brief, the main assumptions include (1) the Stern layer thickness
H was taken as the radius of solvated ClO− 4
ions (2) the dielec-
Superscripts and subscripts
tric constant of the electrolyte was taken as constant and equal
∞ Refers to bulk electrolyte
to 64.4 corresponding to PC [14], (3) the reversible redox reac-
i Refers to ion species i (i = 1 for Li± and i = 2 for ClO− )
4 tion occurred at the electrode/electrolyte interface transferring Li+
P Refers to the pseudocapacitive electrode
from the electrolyte into the Mp Oq matrix, (4) Li+ intercalation,
accompanied by reduction of the transition metal cations in the
422 H.-L. Girard et al. / Electrochimica Acta 211 (2016) 420–429

Fig. 1. Schematic of the simulated one-dimensional pseudocapacitive electrode consisting of a redox-active pseudocapacitive electrode in a three-electrode setup with LiClO4
in propylene carbonate as electrolyte. The dashed line encloses the computational domain simulated.

electrode, was treated as one-dimensional diffusion processes, (5) coefficient D1,P , the initial Li concentration c1,P,0 , the reaction rate
The physicochemical properties of the electrodes and the elec- constant k0 , and the equilibrium potential for the redox reaction
trolyte were assumed to be constant, and (6) the electrode material  eq , (iii) the cell dimensions including the electrolyte thickness
did not undergo any phase transition during the lithiation pro- L and the electrode thickness LP , and (iv) the operating conditions,
cess, as observed experimentally [5,7]. The governing equations i.e., the temperature T, the potential window defined by max and
are reproduced in Appendix A for convenience. They consisted of min , and the scan rate v.
the conservation equations governing the spatiotemporal evolu- Table 1 summarizes the values of these input parameters for the
tion of (i) the potential (x, t) in the electrode and electrolyte, (ii) baseline case corresponding to Nb2 O5 electrode with an electrolyte
the ion concentrations c1 (x, t) and c2 (x, t) for Li+ and ClO− 4
ions in the containing Li+ ions in propylene carbonate (PC). The properties
electrolyte, respectively, and (iii) the intercalated Li concentration were taken from the literature whenever possible. The counter-ion
c1,P (x, t) in the electrode.
The three-electrode measurement method consists of (1) the
Table 1
working electrode under investigation, (2) a reference elec-
Electrode and electrolyte properties and dimensions used in the simulations
trode setting the reference potential, and (3) a counter-electrode reported in this study
designed to measure the current without limiting it. This method
Parameters Symbol Value Unit Ref.
ensures that no current flows through the reference electrode and
its potential remains constant. It also enables precise control of the Relative r 64.4 [14]
potential between the electrolyte and the pseudocapacitive work- permittivity
Ion diameter a1 = a 2 0.67 nm [15]
ing electrode. In our simulations, a triangular potential waveform Ion valency z1 = − z2 1
s (t) was imposed at the current collector (x = − L − LP ) [10]. On Ion diffusion D1 = D2 2.6 × 10−10 2
m /s [15]
the other hand, the potential in the bulk of the electrolyte was set coefficient
as (x = 0, t) = 0 V. This set the reference potential for the system Stern layer H 0.5 nm [9]
thickness
in a way similar to the reference electrode in the experimental
Initial bulk ion c1,∞ = c2,∞ 1 mol/L [16–18]
setup. Additionally, the ion concentrations c1 (x, t) and c2 (x, t) were concentration
imposed to be constant at x = 0 and equal to their bulk concentra- Electrode P 10 −4 −3
to 10 S/m [19]
tions, i.e., c1 (0, t) = c2 (0, t) = c1,∞ . This prevented the formation of conductivity
an EDL and ensured that the reference electrode did not pass any Intercalated Li D1,P 10−14 to 10−11 m2 /s [20]
diffusion
current. No condition was imposed on the current flowing at the coefficient
electrode/current collector interface to account for the overcapac- Reaction rate k0 10−8 m2.5 mol−0.5 s−1 [20–22]
itive counter-electrode used experimentally. Transfer coefficient ˛ 1/2
Maximum c1,P,max 32.9 mol/L [23]
intercalated Li
2.2. Constitutive Relationships concentration
Initial intercalated c1,P,0 1 to 5 mol/L [23]
A total of 15 input parameters were needed to solve the gov- Li concentration
Electrolyte L 1 m
erning equations, initial and boundary conditions including (i) the thickness
binary and symmetric electrolyte properties, comprising the per- Electrode thickness LP 20 to 150 nm [27,24]
mittivity r , the Li+ ion diameter a1 , valency z1 , diffusion coefficient Temperature T 298 K
D1 , and bulk concentration c1,∞ , (ii) the pseudocapacitive electrode Potential window max − min 0.9 V [16–18]
Scan rate v 0.5 to 10 V/s [16–18]
properties namely the electrical conductivity  P , the Li diffusion
 H.-L. Girard et al. / Electrochimica Acta 211 (2016) 420–429 423

in the binary symmetric electrolyte had opposite valency z2 = − z1 predictions corresponding to the baseline case (Table 1). Fig. 2
but diameter a2 , diffusion coefficient D2 , and bulk concentration indicates that the overall shapes of the experimentally measured
c2,∞ identical to those of Li+ . Note that the permittivity r of the and numerically predicted CV curves were similar. The CV curves
electrolyte was considered constant and independent of the elec- featured a large current density in the lower portion of the
tric field. Furthermore, the equilibrium potential difference  eq , potential window and a significantly smaller current at higher
also known as the open circuit potential, varies with the oxidation potentials. The evolution of the CV curves with increasing scan rate
state of the electrode [25]. This variation is commonly assumed to also followed a similar trend with the charging and discharging
be linear [25]. The fitting parameters were measured for a MnO2 peak increasing in magnitude and shifting to lower and higher
electrode in KOH electrolyte with potassium intercalation as a func- potentials, respectively. Finally, the shape of the CV curves during
tion of the state of charge c1,P /c1,P,max where c1,P,max corresponds to the potential reversal between charging and discharging became
the maximum intercalated Li concentration [25]. In the absence of more round with increasing scan rate. This can be attributed
experimental data for Nb2 O5 , parameters for MnO2 were used to to the increasingly resistive behavior. Indeed, the potential at
account for the evolution of the equilibrium potential difference the electrode/electrolyte interface (x = − L) became significantly
with the state-of-charge [25] lower and varied more slowly than the potential s (t) imposed
  at the current collector (x = − L − LP ) as the scan rate increased.
c1,P (t)
 eq (t) = 10.5 4− − 39.9. (1) This delayed the EDL formation and reduced the electrode surface
c1,P,max overpotential driving the redox reaction.
The maximum concentration c1,P,max corresponding to fully- To further elucidate these observations, Fig. 3a shows the
lithiated Li2 Nb2 O5 was taken as 32.9 mol/L. In order to preserve numerically predicted (i) capacitive current density jC (t) due to
the continuity of all variables from the imposed initial condi- EDL formation and dissolution, (ii) the faradaic current density jF (t)
tions, the initial equilibrium potential was required to be zero, i.e., associated with reversible redox reactions, and (iii) the total cur-
 eq (t = 0) = 0 V. This condition was satisfied by setting the initial rent density jT (t) = jC (t) + jF (t) as functions of the imposed potential
Li concentration in the electrode as c1,P (0) = c1,P,0 = 6.578 mol/L. In s (t) for the baseline case at scan rate v = 5 V/s. Fig. 3b shows

the baseline case, the pseudocapacitive electrode had thickness the corresponding concentrations c1 (− L + H, t) of the cation Li+
LP = 50 nm, electrical conductivity  P = 10−4 S/m, and Li diffu- and c2 (− L + H, t) of the anion ClO− 4
at the Stern/diffuse layer
sion coefficient D1,P = 10−12 m2 /s, corresponding to realistic values interface while Fig. 3c shows the intercalated lithium concentration
[19,20]. Finally, the numerically predicted CV curves correspond- c1,P at the electrode/electrolyte interface. The CV curve displays two
ing to oscillatory steady-state were compared with experimental regimes: (i) the faradaic regime in the lower portion of the poten-
measurements for electrodes made of Nb2 O5 with LiClO4 in PC as tial window when the total current density was dominated by the
electrolyte [5,6,26]. faradaic current, i.e., jT (t) ≈ jF (t), and (ii) the capacitive regime in
the upper end of the potential window when the capacitive current
3. Results and Discussion dominated, i.e., jT (t) ≈ jC (t), and the overall current magnitude was
smaller than in the faradaic regime. Fig. 3b indicates that the anions
3.1. Voltammogram interpretation ClO− 4
formed the EDL during the capacitive regime. Their concen-
tration c2 (− L + H, t) at the Stern/diffuse layer interface near the
Fig. 2a shows CV curves obtained from three-electrode exper- pseudocapacitive electrode reached the maximum packing con-
imental CV measurements at different scan rates. The electrode centration of 5.52 mol/L quickly after the onset of the capacitive
consisted of T-Nb2 O5 (orthorhombic) nanocrystal mixed 1:1 with regime, marked by a red circle. Symmetrically, the cations Li+ were
carbon black on a stainless steel current collector with 1 mol/L repelled and their concentration c1 (− L + H, t) at the electrode sur-
LiClO4 in PC as electrolyte [5]. Similarly, Fig. 2b plots the numerical face was negligible during the capacitive regime. The steep drop in

Fig. 2. (a) Experimental measurements of the total current IT (t) scaled by the absolute value of the maximum current as a function of imposed potential s (t) for a pseudoca-
pacitive electrode consisting of T-Nb2 O5 (orthorhombic) nanocrystal mixed 1:1 with carbon black on a stainless steel current collector with 1 mol/L LiClO4 in PC as electrolyte
reported in Figure 1a of Ref. [5]. (b) Numerically predicted total current density jT (t) scaled by the absolute value of the maximum current density as a function of imposed
potential s (t) for different scan rates v ranging from 0.5 to 10 V/s.
424 H.-L. Girard et al. / Electrochimica Acta 211 (2016) 420–429

Fig. 3. (a) Numerically predicted capacitive jC , faradaic jF , and total jT current densities as functions of imposed potential s (t) for the baseline pseudocapacitive electrode at
scan rate v = 5 V/s. (b) Concentrations of the cation Li+ c1 (− L + H, t) and of the anion ClO− c (− L + H, t) at the Stern/diffuse layer interface near the pseudocapacitive electrode
4 2
(x = − L + H) as functions of imposed potential s (t). (c) Predicted b-value as a function of imposed potential s (t) for constitutive relationships corresponding to the baseline
case.

the faradaic current observed in the CV curves during discharging 3.2. Comparison with equivalent EDL electrode
was due to the starvation of Li+ ions at the electrode/electrolyte
interface necessary for the faradaic reaction. The onset of Li+ ion Fig. 4a compares experimental CV curves at 2 mV/s in terms of
starvation is indicated by a blue square in Fig. 3a to 3c. Note also specific current iT (in A/g) for an electrode made of carbon nano-
that the simulated scan rates were larger than those used in exper- tubes and for a pseudocapacitive electrode consisting of T-Nb2 O5
imental measurements. This was due to the fact that diffusion (orthorombic) with 1 mol/L LiClO4 in PC as electrolyte and Li metal
limitation and Li+ ion starvation appeared at a much higher scan for the counter and reference electrodes [26]. Similarly, Fig. 4b
rate for the simulated planar electrode than for the actual porous shows the numerically predicted CV curves at 1 V/s for the base-
electrodes. line pseudocapacitive electrode previously defined as well as for
Fig. 3d shows the b-value, obtained from fitting the evolution of an equivalent electrode in which charge storage was exclusively
current density with scan rate according to the power law jT ( s ) = through EDL formation. The density of carbon and Nb2 O5 were
a( s )vb( s ) as a function of imposed potential s (t), as commonly taken as 2250 and 4550 kg m−3 , respectively. Here also, numeri-
performed experimentally [16,27–29]. Fig. 3d indicates that a dip cal and experimental results show similar behavior for both types
in b-value occurred for potentials s (t) around 0.3 V. Such a dip of electrodes. The EDLC or carbon electrode displayed a rectan-
was also observed in three-electrode experimental measurements gular CV curve characteristic of capacitive behavior. On the other
and attributed to the oxidation of the metal oxide [29]. However, hand, the pseudocapacitive electrode featured a large current in
the numerical simulations established that the dip in the b-value the lower portion of the potential window. Interestingly, in both
coincided with the transition between the faradaic and the capac- experimentally measured and numerically simulated CV curves,
itive regimes (red circle) [10,11]. This transition corresponded to the current density of the pseudocapacitive electrode, in the upper
the steep drop in the total current density observed in both exper- end of the potential window, corresponding to the capacitive
imental data and numerical simulations shown in Fig. 2. regime, matched that of the carbon electrode. This confirms that,
 H.-L. Girard et al. / Electrochimica Acta 211 (2016) 420–429 425

Fig. 4. (a) Experimental CV curves in terms of the total current IT (t) scaled by the absolute value of the maximum current as a function of imposed potential s (t) at 2 mV/s.
Here, the pseudocapacitive electrode consisting of 5 nm Nb2 O5 nanocrystal mixed with PVDF (10%) and carbon black (10%) with 1 mol/L LiClO4 in PC as electrolyte and Li
metal for the counter and reference electrodes as well as an equivalent electrode consisting of carbon nanotubes (EDLC). The experimental data was retrieved from Figure
3a of Ref. [26]. (b) Numerically predicted CV curves in terms of the total current IT (t) scaled by the absolute value of the maximum current as a function of imposed potential
s (t) at 1 V/s for the pseudocapacitive electrode considered in the baseline case as well as for an equivalent EDL electrode.

in this potential region, EDL formation was the main mechanism Nb2 O5 . All other properties were the same as in the baseline case
for charge storage. Such comparison provides an experimental way (Table 1). Here also, numerical and experimental CV curves display
to distinguish the capacitive from the faradaic regime and to esti- very similar shapes and trends for both amorphous and crystalline
mate the contribution of EDL formation to charge storage for a electrode materials. They indicate that the capacitance was signif-
given electrolyte. As previously explained, the simulated scan rate icantly larger for the crystalline phase of Nb2 O5 thanks to better
was significantly larger than that imposed experimentally. This transport properties leading to faster intercalation and smaller
resulted in significantly larger current densities. In other words, electrical resistance. However, the current density in the capaci-
simulating planar electrodes replicates experimental measure- tive regime ( s > 0.3 V) appeared to be independent of the phases.
ments only qualitatively but can provide significant insights lacking This can be attributed to the fact that charge storage was achieved
from experimental measurements. Quantitative predictions with only by EDL formation at the electrode/electrolyte interface and to
realistic scan rates and current densities should be achievable by the absence of faradaic reactions and Li+ intercalation.
simulating actual three-dimensional electrode architectures, albeit
at significant computational cost.
3.4. Effect of Electrode Thickness

3.3. Crystalline vs. amorphous pseudocapacitive material Fig. 6a shows the numerically predicted CV curves in terms of
total current density jT as a function of imposed potential s (t) for
Niobium oxide can assume different phases depending on the different values of pseudocapacitive electrode thickness LP ranging
synthesis method. The most common structures are amorphous, from 5 to 150 nm at scan rate v = 1 V/s. All parameters other than LP
pseudo-hexagonal (TT-Nb2 O5 ), and orthorhombic (T-Nb2 O5 ). corresponded to the baseline case (Table 1). Fig. 6a indicates that,
These phases exhibit different electrical conductivity  P and Li as the electrode thickness increased, the magnitude of the peak
diffusion coefficient D1,P . In particular, crystalline phases display current initially increased. However, the shape of the CV curve
larger electrical conductivity than the amorphous phase due to became more rounded and the peak current decreased beyond a
the increased domain size enabling the charge carriers to follow certain thickness. This was due to the increasingly resistive behav-
crystalline pathways [6]. They also feature larger solid-state ior caused by the potential drop across the thicker electrode, as
diffusion coefficient due to the creation of diffusion pathways previously discussed.
in the crystalline structure with low energy barrier for easier Fig. 6b shows the electrode integral capacitance Cs,int at low scan
intercalation [30]. rate (i.e., under quasi-equilibrium) as a function of the thickness of
Fig. 5a shows experimental CV curves at 10 mV/s for pseu- the pseudocapacitive electrode LP . For the parameters of the base-
docapacitive electrodes consisting of (i) T-Nb2 O5 (orthorombic) line case considered in this study, the optimal electrode thickness
nanoparticles or (ii) amorphous Nb2 O5 nanoparticles on a stainless was around LP = 65 nm. These results illustrate the trade-off faced
steel current collector with 1 mol/L LiClO4 in PC as electrolyte and Li when designing a pseudocapacitive electrode between offering a
metal for the counter and reference electrodes [6]. Similarly, Fig. 5b large electrode volume for redox intercalation while maintaining
shows numerically predicted CV curves for v = 5 V/s for a pseudo- acceptable potential drop across the electrode to drive the faradaic
capacitive electrode with (i) high conductivity ( P = 10−3 S/m) and reactions. Among all parameters, increasing the electrode thick-
fast Li diffusion (D1,P = 10−12 m2 /s) simulating crystalline Nb2 O5 ness and/or the diffusion coefficient of Li+ in the pseudocapacitive
and (ii) an electrode with lower conductivity ( P = 10−4 S/m) and electrode had the most beneficial effects on the electrode perfor-
slower diffusion (D1,P = 10−14 m2 /s) corresponding to amorphous mance. Larger electrical conductivity was also beneficial when the
426 H.-L. Girard et al. / Electrochimica Acta 211 (2016) 420–429

Fig. 5. (a) Experimental CV curves in terms of the total current IT (t) scaled by the absolute value of the maximum current as a function of imposed potential s (t) for a
pseudocapacitive electrode consisting of T-Nb2 O5 (orthorombic) nanoparticles on a stainless steel current collector with 1 mol/L LiClO4 in PC as electrolyte and Li metal for
the counter and reference electrodes as well as a pseudocapacitive electrode consisting of amorphous Nb2 O5 in the same conditions. The experimental data was retrieved
from Figure 4a of Ref. [6]. (b) Numerically predicted CV curves in terms of the total current IT (t) scaled by the absolute value of the maximum current as a function of imposed
potential s (t) for a pseudocapacitive electrode with high conductivity ( P = 10−3 S/m) and fast Li diffusion (D1,P = 10−12 m2 /s) as well as for a pseudocapacitive electrode with
low conductivity ( P = 10−4 S/m) and slow diffusion (D1,P = 10−14 m2 /s).

Fig. 6. (a) Numerically predicted total current density jT (A/m2 ) as a function of imposed potential s (t) at v = 1 V/s for different pseudocapacitive electrode thicknesses
ranging from 5 to 150 nm. (b) Integral capacitance Cs,int as a function of electrode thickness. All other parameters are identical to the baseline case.

electrode was relatively thick. This corroborated the findings of our dominated by EDL formation as discussed extensively in our
recent scaling analysis for two-electrode devices captured by the previous studies [10,11]. The transition between the regimes was
dimensionless number f [11]. shown to be caused by Li+ starvation in the electrolyte due to the
electrostatic repulsion associated with the ClO− 4
EDL formation.
4. Conclusion This transition coincided with the dip in the b-value commonly
observed experimentally. This was further illustrated by comparing
This paper presented unequivocal physical interpretation of CV curves in terms of specific current (A/g) for pseudocapacitive
experimentally measured CV curves reported for Nb2 O5 pseudo- (Nb2 O5 ) and EDLC (carbon) electrodes. In addition, the larger
capacitive electrodes. First, the numerical simulations of cyclic current obtained with crystalline vs. amorphous Nb2 O5 electrodes
voltammetry reproduced the behaviors observed experimentally was attributed to the combination of faster Li intercalation and
[5,6,26]. Two regimes were identified in the CV curves (i) a faradaic smaller electrical resistance. Finally, the design trade-off between
regime dominated by redox reactions and (ii) a capacitive regime a thick electrode for achieving large energy storage or capacitance
 H.-L. Girard et al. / Electrochimica Acta 211 (2016) 420–429 427

and a thin electrode for minimizing potential drop across the constant and the universal gas constant, respectively. The maxi-
electrode and resistive losses was illustrated leading to an optimal mum concentration cj,max = 1/(NA a3j ) corresponds to simple cubic
electrode thickness of about 65 nm. Further research should focus ion packing at the electrode surface. The first and second terms of
on quantitative comparison with experimental data by simulating Equation (5) represent the ion flux due to diffusion and electromi-
realistic three-dimensional porous electrode architectures. gration, respectively, while the last term represents a correction
accounting for finite ion size [9]. Finally, the presence of a Stern
Acknowledgements layer near the electrode was accounted for via boundary conditions
[32,9] and no governing equations for (x, t) and ci (x, t) needed to
This material is based upon work supported as part of the be formulated or solved within the Stern layer.
Molecularly Engineered Energy Materials, an Energy Frontier
Research Center funded by the U.S. Department of Energy, Office A.2. Boundary and initial conditions
of Science, Office of Basic Energy Sciences under Award Number
DE-SC0001342. Partial support for the research was also provided The one-dimensional governing Equations (2) to (5) are second-
by the Office of Naval Research. order partial differential equations in space and first-order in time.
Each equation requires two boundary conditions and one initial
Appendix A. Governing Equations and Boundary Conditions condition in each region it is solved.
First, the initial electric potential was assumed to be uniform
A.1. Governing equations across the device and given by

(x, 0) = 0 V. (6)
The local electric potential (x, t) in the pseudocapacitive
(− LP − L ≤ x ≤ − L) electrode is governed by the one-dimensional Initially, the Li+ and ClO−
4
ion concentrations in the electrolyte
Poisson equation expressed as [9,22] (− L + H ≤ x ≤ 0) were taken as uniform and equal to their bulk con-
  centrations satisfying electroneutrality according to
∂ ∂
P =0 (2) c1 (x, 0) = c1,∞ and c2 (x, 0) = −z1 c1,∞ /z2 . (7)
∂x ∂x
Similarly, the initial Li concentration in the pseudocapacitive
where  P is the electrical conductivity of the electrode.
electrode (− L − LP ≤ x ≤ − L) was uniform and equal to c1,P,0 , i.e.,
The local molar concentration of intercalated Li atoms in the
pseudocapacitive electrode, denoted by c1,P (x, t), is governed by c1,P (x, 0) = c1,P,0 . (8)
the mass diffusion equation given by [22,31]
  The potential at the current collector/pseudocapacitive elec-
∂c1,P ∂ ∂c1,P trode interface was imposed as (− L − LP , t) = s (t). During
= D1,P (3)
∂t ∂x ∂x cyclic voltammetry measurements, s (t) varied linearly with time
according to [9]
where D1,P is the diffusion coefficient of intercalated lithium atoms
− v[t − (nc − 1) CV ] for (nc − 1) CV ≤ t < (nc − 1/2)
in the electrode. =
max CV
(9)
s (t)
Moreover, the potential and ion concentrations in the diffuse min + v[t − (nc − 1/2) CV ] for (nc − 1/2) CV ≤ t < nc CV

layer of the electrolyte solution (−L + H ≤ x ≤ 0) are governed by

where nc is the cycle number and CV is the cycle period while max
the generalized modified Poisson-Nernst-Planck (GMPNP) model
and min are the maximum and minimum values of the cell poten-
derived by Wang et al. [9]. For a binary and symmetric electrolyte
tial s (t), respectively. In addition, the potential in the bulk of the
the GMPNP in the diffuse layer is expressed as [9]
electrolyte (x = 0) was set as (x = 0, t) = 0 V. The ion concentrations
c1 (x, t) and c2 (x, t) were imposed to be constant at x = 0 and equal
to their bulk concentrations, i.e., c1 (0, t) = c2 (0, t) = c1,∞ .
The electric potential varied linearly across the Stern layer so
that the electric field at the pseudocapacitive electrode satisfied
[32,9]

∂ 1 
where ci (x, t) is the local molar concentration of ion species i in − (−L + H, t) = (−L, t) − (−L + H, t) (10a)
the electrolyte solution at time t. Here, i = 1 refers to Li+ and i = 2 ∂x H
refers to ClO−
4
. Moreover, F = 96 485 C mol−1 is the Faraday constant, This boundary condition accounted for the presence of the Stern
0 = 8.854 × 10−12 F m−1 is the free space permittivity, and r is the layers at the electrode surface without explicitly simulating them
relative permittivity of the electrolyte solution. The local mass flux in the computational domain [32].
of ion species i, denoted by Ni (x, t) in mol m−2 s−1 , is defined for a The current density at the pseudocapacitive elec-
binary and symmetric electrolyte as [9] trode/electrolyte interface, located at x = − L was equal to the
sum of the capacitive jC (x, t) and faradaic jF (t) current densities (in
∂ci Di Fzi ci ∂ A/m2 ) at the Stern/diffuse interfaces, located at x = − L + H, so that
Ni (x, t) = −Di − [9,21]
∂x Ru T ∂x
⎛ ⎞ ∂
2 −P (−L, t) = jC (−L + H, t) + jF (−L + H, t) (11)
Di ci ∂ ⎝ ∂x
− 2 cj /cj,max ⎠ (5)
1− c /c ∂x where, jC (x, t) is the displacement current density due to the electric
j=1 j j,max j=1
double layer formation at the pseudocapacitive electrode surface
where Di and ai are the diffusion coefficient and the effec- and defined as [33]
tive ion diameter of ion species i in the electrolyte solution, 2
respectively. The temperature is denoted by T (in K) while ∂
jC (−L + H, t) = −0 r (−L + H, t). (12)
NA = 6.022 × 1023 mol−1 and Ru = 8.314 J K−1 mol−1 are the Avogadro ∂x∂t
428 H.-L. Girard et al. / Electrochimica Acta 211 (2016) 420–429

In addition, the faradaic current density jF (t) is typically [4] G. Wang, L. Zhang, J. Zhang, A Review of Electrode Materials for Electrochemical
described by the generalized Frumkin-Butler-Volmer model eval- Supercapacitors, Chem. Soc. Rev. 41 (2012) 797–828.
[5] V. Augustyn, J. Come, M.A. Lowe, J.W. Kim, P.-L. Taberna, S.H. Tolbert,
uated at the pseudocapacitive electrode/electrolyte interface and H.D. Abruña, P. Simon, B. Dunn, High-Rate Electrochemical Energy Stor-
expressed as [13] age through Li+ Intercalation Pseudocapacitance, Nat. Mater. 12 (2013)
518–522.
[6] J.W. Kim, V. Augustyn, B. Dunn, The Effect of Crystallinity on the Rapid Pseudo-
(1 − ˛)z1 F −˛z1 F
jF (t) = jF,0 (t){exp[ ] − exp[ ]} (13) capacitive Response of Nb2 O5 , Adv. Energy Mater. 2 (2012) 141–148.
Ru T Ru T [7] J. Come, V. Augustyn, J.W. Kim, P. Rozier, P.-L. Taberna, P. Gogotsi, J.W. Long, B.
Dunn, P. Simon, Electrochemical Kinetics of Nanostructured Nb2 O5 Electrodes,
J. Electrochem. Soc. 161 (2014) A718–A725.
where  =  H −  eq is the overpotential and  H = (− L, [8] L. Pilon, H. Wang, A. d’Entremont, Recent Advances in Continuum Modeling
t) − (− L + H, t) is the electrical potential drop across the Stern layer of Interfacial and Transport Phenomena in Electric Double Layer Capacitors, J.
near the pseudocapacitive electrode [13]. For electrodes made of Electrochem. Soc. 162 (2015) A5158–A5178.
[9] H. Wang, A. Thiele, L. Pilon, Simulations of Cyclic Voltammetry For Electric
transition metal oxides, the equilibrium potential difference  eq Double Layers In Asymmetric Electrolytes: A Generalized Modified Poisson-
is typically fitted experimentally as a function of the state-of-charge Nernst-Planck Model, J. Phys. Chem. C 117 (2013) 18286–18297.
c1,P /c1,P,max where c1,P,max is the maximum concentration of inter- [10] H.-L. Girard, H. Wang, A. d’Entremont, L. Pilon, Physical Interpretation of
Cyclic Voltammetry for Hybrid Pseudocapacitors, J. Phys. Chem. C 119 (2015)
calated lithium atoms in the pseudocapacitive electrode [34–36]. 11349–11361.
It can be obtained by fitting experimental data for the open-circuit [11] H.-L. Girard, H. Wang, A.L. d’Entremont, L. Pilon, Enhancing Faradaic Charge
potential [21,22], as illustrated by Eq. (1). The so-called exchange Storage Contribution in Hybrid Pseudocapacitors, Electrochim. Acta 182 (2015)
639–651.
current density jF,0 (t) can be written as [21,22]
[12] A. d’Entremont, L. Pilon, Electrochemical Transport Phenomena in Hybrid
Pseudocapacitors Under Galvanostatic Cycling, J. Electrochem. Soc. (2015)
(accepted).
[13] A.J. Bard, L.R. Faulkner, Electrochemical Methods: Fundamentals and Applica-
jF,0 (t) = Fz1 k0 [c1 (−L + H, t)]1−˛ [c1,P,max −c1,P (−L, t)]˛ [c1,P (−L, t)]˛ tions, John Wiley & Sons, New York, NY, 2001.
[14] G.J. Janz, R.P.T. Tomkins, Nonaqueous Electrolytes Handbook, Vol. I, Academic
(14) Press, New York, NY, 1972.
[15] K. Nishikaw, Y. Fukunaka, T. Sakka, Y.H. Ogata, J.R. Selman, Measurement of
LiClO4 Diffusion Coefficient in Propylene Carbonate by Moire Pattern, J. Elec-
where k0 is the reaction rate constant expressed in m1+3 ˛ mol−˛ s−1 . trochem. Soc. 153 (2006) 830–834.
Here, the transfer coefficient ˛ was assumed to be 0.5, correspond- [16] M.C. López, G.F. Ortiz, P. Lavela, R. Alcántara, J.L. Tirado, Improved Energy Stor-
ing to identical energy barriers for forward and backward redox age Solution Based on Hybrid Oxide Materials, ACS Sustainable Chem. Eng. 1
(2013) 46–56.
reactions [13]. In this case, the faradaic current jF (t) has the same
[17] Y.-P. Lin, N.-L. Wu, Characterization of MnFe2 O4 /LiMn2 O4 Aqueous Asymmetric
sign as the overpotential (t). Supercapacitor, J. Power Sources 196 (2011) 851–854.
Moreover, the current collector was impermeable to the lithium [18] E. Lim, H. Kim, C. Jo, J. Chun, K. Ku, S. Kim, H.I. Lee, I.-S. Nam, S. Yoon,
K. Kang, et al., Advanced Hybrid Supercapacitor Based on a Mesoporous
atoms intercalated in the pseudocapacitive electrode so that the
Niobium Pentoxide/Carbon as High-Performance Anode, ACS nano 8 (2014)
mass flux of Li vanished at the pseudocapacitive electrode/current 8968–8978.
collector interface (at x = -L-LP ), i.e., [19] D. Bélanger, T. Brousse, J.W. Long, Manganese Oxides: Battery Materials Make
the Leap to Electrochemical Capacitors, Electrochem. Soc. Interface 17 (2008)
49–52.
∂c1,P [20] A.M. Colclasure, R.J. Kee, Thermodynamically Consistent Modeling of Elemen-
−D1,P (−L − LP , t) = 0 mol m−2 s−1 . (15)
∂x tary Electrochemistry in Lithium-ion Batteries, Electrochim. Acta 55 (2010)
8960–8973.
[21] S.R. Subramanian, V. Boovaragavan, V. Ramadesigan, M. Arabandi, Math-
The mass flux of Li exiting through the pseudocapacitive elec- ematical Model Reformulation for Lithium-Ion Battery Simulations:
trode/electrolyte interface (x = L) was related to the faradaic current Galvanostatic Boundary Conditions, J. Electrochem. Soc. 156 (2009)
260–271.
density jF (t) based on stoichiometry as
[22] G.M. Goldin, A.M. Colclasure, A.H. Wiedemann, R.J. Kee, Three-Dimensional
Particle-Resolved Models of Li-ion Batteries to Assist the Evaluation of Empir-
∂c1,P jF (t) ical Parameters in One-Dimensional Models, Electrochim. Acta 64 (2012)
−D1,P (−L, t) = . (16) 118–129.
∂x z1 F [23] M. Toupin, T. Brousse, D. Bélanger, Charge Storage Mechanism of MnO2 Elec-
trode Used in Aqueous Electrochemical Capacitor, Chem. Mater. 16 (2004)
The mass flux of Li+ ions (i = 1) across the Stern/diffuse layer 3184–3190.
[24] J. Zhu, S. Tang, H. Xie, Y. Dai, X. Meng, Hierarchically Porous MnO2 Microspheres
interface near the pseudocapacitive electrode was related to jF (t) Doped with Homogeneously Distributed Fe3 O4 Nanoparticles for Supercapa-
in the same way such that citors, ACS Appl. Mater. Interfaces 6 (2014) 17637–17646.
[25] P. Guillemet, T. Brousse, O. Crosnier, Y. Dandeville, L. Athouel, Y. Scudeller, Mod-
eling Pseudo Capacitance of Manganese Dioxide, Electrochim. Acta 67 (2012)
jF (t) 41–49.
N1 (−L + H, t) = . (17)
z1 F [26] X. Wang, G. Li, Z. Chen, V. Augustyn, X. Ma, G. Wang, B. Dunn, Y. Lu, High-
Performance Supercapacitors Based on Nanocomposites of Nb2 O5 Nanocrystals
and Carbon Nanotubes, Adv. Energy Mater. 1 (2011) 1089–1093.
Finally, the pseudocapacitive electrode was impermeable to [27] J. Wang, J. Polleux, J. Lim, B. Dunn, Pseudocapacitive Contributions to Electro-
ClO−
4
ions (i = 2) so that chemical Energy Storage in TiO2 (Anatase) Nanoparticles, J. Phys. Chem. C 111
(2007) 14925–14931.
[28] Y.-M. Lin, P.R. Abel, D.W. Flaherty, J. Wu, K.J. Stevenson, A. Heller, C.B.
N2 (−L + H, t) = 0 mol m−2 s−1 . (18) Mullins, Morphology Dependence of the Lithium Storage Capability and Rate
Performance of Amorphous TiO2 Electrodes, J. Phys. Chem. C 115 (2011)
2585–2591.
[29] L. Shao, J.-W. Jeon, J.L. Lutkenhaus, Porous Polyaniline Nanofiber/Vanadium
References Pentoxide Layer-by-Layer Electrodes for Energy Storage, J. Mater. Chem. A 1
(2013) 7648–7656.
[1] P. Simon, Y. Gogotsi, Materials for Electrochemical Capacitors, Nat. Mater. 7 [30] C.-P. Liu, F. Zhou, V. Ozolins, First Principles Study for Lithium Intercalation and
(2008) 845–854. Diffusion Behavior in Orthorhombic Nb2 O5 Electrochemical Supercapacitor,
[2] Y. Zhang, H. Feng, X. Wu, L. Wang, A. Zhang, T. Xia, H. Dong, X. Li, L. Zhang, APS Meeting Abstracts (2012), p 26003.
Progress of Electrochemical Capacitor Electrode Materials: a Review, Int. J. [31] C. Wang, A.M. Sastry, Mesoscale Modeling of a Li-Ion Polymer Cell, J. Elec-
Hydrogen Energy 34 (2009) 4889–4899. trochem. Soc. 154 (2007) 1035–1047.
[3] Z.-S. Wu, G. Zhou, L.-C. Yin, W. Ren, F. Li, H.-M. Cheng, Graphene/Metal Oxide [32] H. Wang, L. Pilon, Mesoscale Modeling of Electric Double Layer Capacitors
Composite Electrode Materials for Energy Storage, Nano Energy 1 (2012) with Three-Dimensional Ordered Structures, J. Power Sources 221 (2013)
107–131. 252–260.
 H.-L. Girard et al. / Electrochimica Acta 211 (2016) 420–429 429

[33] H. Cohen, J.W. Cooley, The Numerical Solution of the Time-Dependent Nernst- [35] G. Sikha, R.E. White, B.N. Popov, A Mathematical Model for a Lithium-ion Bat-
Planck Equations, Biophys. J. 5 (1965) 145–162. tery/Electrochemical Capacitor Hybrid System, J. Electrochem. Soc. 152 (2005)
[34] T. Jow, J. Zheng, Electrochemical Capacitors Using Hydrous Ruthenium Oxide A1682–A1693.
and Hydrogen Inserted Ruthenium Oxide, J. Electrochem. Soc. 145 (1998) [36] H. Kim, B.N. Popov, A Mathematical Model of Oxide, Carbon Composite Elec-
49–52. trode for Supercapacitors, J. Electrochem. Soc. 150 (2003) 1153–1160.