Electrochimica Acta 131 (2014) 160–167

Contents lists available at ScienceDirect

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

Electrochemical characterization of Polymer Electrolyte Membrane

Water Electrolysis Cells
C. Rozain, P. Millet ∗
Institut de Chimie Moléculaire et des Matériaux d’Orsay, UMR CNRS n◦ 8182, Université Paris Sud 11, 15 rue Georges Clémenceau, 91405 Orsay cedex, France

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

Article history: The purpose of this paper is to report on the electrochemical characterization of Polymer Electrolyte
Received 8 August 2013 Membrane (PEM) water electrolysis cells. Results were obtained using membrane-electrode assemblies
Received in revised form 18 January 2014 containing unsupported IrO2 catalyst at anode for the oxygen evolution reaction (OER) and carbon-
Accepted 18 January 2014
supported platinum nano-particles at the cathode for the hydrogen evolution reaction (HER). Roughness
Available online 31 January 2014
factors of anodes and cathodes have been determined using an internal reference electrode. Individ-
ual cell voltage contributions have also been measured as a function of operating current density. Cell
Keywords:
impedance spectra have been measured at different cell voltages along the polarization curve. It is shown
water electrolysis
polymer electrolyte
that charge transfer processes are major cell impedance contributors at voltages up to 1.8-1.9 V. At higher
hydrogen, impedance. cell voltages, cell impedances are mainly resistive. It is shown that the impedance associated with the
HER is negligible and that the two time-constants observed on experimental impedance spectra can both
be attributed to the OER. Possible mechanism options are discussed. Finally, some results related to the
EIS characterization of PEM water electrolysis stacks are also reported.
© 2014 Elsevier Ltd. All rights reserved.

1. Introduction 25 ◦ C - 1000 ◦ C temperature range which is accessible to con-

ventional chemical technology. The Gibbs free energy change of
Increasing shares of renewable energy sources (intermit- the water splitting reaction becomes negative at approximately
tent electricity) in the energy mix are calling for appropriate 2,500 K. Therefore, the direct thermal dissociation of water is dif-
energy storage solutions. In spite of excellent charging/discharging ficult to achieve because few materials can sustain such high
efficiencies, electrochemical devices such as batteries offer inap- operating temperatures and because the separation of gaseous
propriately low storage capacities and the need for chemical energy hydrogen and oxygen in such conditions remains a very chal-
carriers is commonly acknowledged. Among a handful of potential lenging task. Thermo-chemical cycles can be used to reduce the
candidates, molecular hydrogen (H2 ) is usually considered as one operating temperature [4]. Besides thermal processes, electric-
of the most promising. There are few natural reserves of gaseous ity can also be used to split water at much lower temperatures.
hydrogen which has to be extracted from natural feedstocks. In the Among three main water electrolysis technologies (alkaline, PEM
industry sector, hydrogen is produced from natural hydrocarbons and Solid Oxide) [5], the Polymer Electrolyte Membrane (PEM) pro-
(natural gas, oil and coal), using steam reforming or gasification cess, which is specifically considered in this paper, is perceived
processes. The annual production is in the 50- billion tons range as a key process for transforming zero-carbon electricity sources
[1]. As a result, large amounts of carbon dioxide, a gas which is into the supply of zero-carbon hydrogen and oxygen for mis-
considered as responsible for increasing greenhouse effects and cellaneous end-uses [6]. In state-of-art technology, a PEM water
associated climate changes, are emitted in the atmosphere (0.3–0.4 electrolyzer can now operate at 1 A.cm−2 with a high-heating-
m3 CO2 /m3 H2 ) [2]. Water is a CO2 -free source of molecular hydro- value (HHV) efficiency close to 75-80% and the multiple A.cm−2
gen and a lot of R&D is carried out on the subject in view of the range is becoming accessible. In addition, PEM water electrolysis
so-called “hydrogen economy” [3]. The total amount of energy offers the possibility of delivering pressurized hydrogen to facili-
required to split one mole of liquid water into gaseous hydro- tate hydrogen storage in pressurized containers [7]. Main scientific
gen and oxygen is (in standard conditions), H◦ = +285 kJ.mol−1 . contributions in the field have been recently reviewed [8]. Increas-
The enthalpy change remains approximately constant over the ing R&D investments are being made by public funding agencies
and private companies to develop electrolysis units in the 100-250
Nm3 H2 /hour delivery range and in the 104 -105 hours operation
∗ Corresponding author. Tel.: +33 1 6915 4812. range that would allow PEM water electrolysis to compete with
E-mail address: [email protected] (P. Millet). the more mature alkaline process. Efficiency and durability issues

0013-4686/$ – see front matter © 2014 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.electacta.2014.01.099
 C. Rozain, P. Millet / Electrochimica Acta 131 (2014) 160–167 161

are therefore critical and a better understanding of the perform- current collectors. In the anodic compartment, the current collector
ances of conventional membrane-electrode assemblies (MEAs) is (label 3 in Fig. 1-a) is usually a millimeter thick porous titanium disc
required. obtained by thermal sintering of titanium powder. The microstruc-
In this context, the purpose of this paper is to report on the elec- ture of the disc (in terms of porosity and tortuosity) is a key factor
trochemical characterization of PEM water electrolysis cells using that needs to be optimized because this is the place where two
MEAs that contain Nafion as solid polymer electrolyte, iridium opposite flows (liquid water towards the anode and oxygen gas
oxide as anodic catalyst for the oxygen evolution reaction (OER) and away from the anode) coexist [10]. Inappropriate porosities may
carbon-supported platinum as cathodic catalyst for the hydrogen yield mass transport limitations. Optimal porosity values may dif-
evolution reaction (HER). fer as a function of the nominal operating current density since,
according to Faraday’s law, the gas production is directly related to
the current flowing across the cell. In the cathodic compartment,
2. Experimental a similar titanium disc can be used. Alternatively (since titanium
discs are rather expensive), a gas-diffusion-layer (GDL) similar to
2.1. Chemicals and materials those found in PEM fuel cells can also be used (label 3 in Fig. 1-a).
The microstructure of the cathodic current collector is less critical
MEAs were prepared using PTFE-reinforced Nafion (du Pont) than the anodic one because at the cathode, the hydrogen gaseous
as solid polymer electrolyte (SPE). Unsupported iridium dioxide production and the flow of liquid water (due to the electro-osmotic
(Surepure Chem.) was used at the anode for the OER (loading ≈ drag across the cell, i.e. hydration water molecules transported by
1.5 mg.cm−2 ). Carbon-supported platinum nano-particles (Sigma protons) are both co-linear, away from the MEA. Two titanium end-
Aldrich Co.) was used at the cathode for the HER (loading ≈ 0.5 mg plates delimit the PEM cell (labels 5 and 5 in Fig. 1-a). Biphasic
Pt.cm−2 ). Catalytic inks containing catalysts powders were sprayed liquid–gas mixtures can circulate in channels or machined grooves
automatically over the surface of the SPE using a catalytic ink in the thickness of the end-plates, just like in PEM fuel cells. Alter-
printer (ExactaCoat, Sono Tek Co.). natively, less-expensive titanium grids can also be used as cell
spacers (labels 4 and 4 in Fig. 1-a). Individual PEM cells (usually
2.2. Electrochemical characterization in the 4–7 mm thick range) can be stacked together in order to
increase gas production. A total of fifty to sixty cells is nowadays
A computer-controlled ModuLab Potentiostat/Galvanostat common good practice in industrial systems whereas in alkaline
(Solartron Analytical Co.) equipped with an external 12V-60A water electrolysis technology, several hundred cells can be clamped
booster was used to perform EIS measurements on laboratory together.
cells in the 100 kHz-1 mHz frequency range. Short (10 cm long)
electric cables were used to minimize parasite impedance effects.
A VPM3 Multipotentiostat (Biologic Co.) was used to perform 3.2. Electrical equivalent circuit of a PEM water electrolysis cell
batch EIS measurements on the PEM stack. EIS spectra reported
here were not corrected for any possible parasite inductive effects. Fig. 1-b shows the equivalent electrical circuit of the PEM water
Experimental impedance spectra were fitted using the EC-Lab electrolysis cell. Each charge transfer interface is modelled using
software (Biologic Co.). the parallel connection of a polarization resistance and a con-
stant phase element (Q) [11]. Constant phase elements are used to
take into account deviations of double layers from ideal capacitive
3. Detailed Description Of The Pem Water Electrolysis Cell behaviours due to surface roughness, polycrystallinity and anion
adsorption (sulfonate ions in PEM cells). Diffusion impedances are
3.1. Cell components added in series to account for possible mass transport limitations
of hydrogen and oxygen gases away from the interfaces across
Fig. 1-a shows the cross-section of the PEM water electroly- porous current collectors. Interface impedances are also series con-
sis cell. The active component at the center of the cell (label 1 nected to ohmic resistances to account for electronic (metals) and
in Fig. 1 a) is the MEA. MEAs are made of a thin (usually ≈ 50- ionic (SPE) conductivity of cell components. It will be shown in
200 ␮m thick) proton-conducting polymeric film (also called solid the following that the ohmic resistance of the PEM cell is a critical
polymer electrolyte or SPE) with two porous catalytic layers plat- parameter that can quite significantly impact the overall perform-
ted on each side. In conventional PEM systems, the SPE is usually ances of the PEM cell and that needs to be minimized. Details of the
made of perfluoro-sulfonated polymer, as in fuel cell technology. different circuit components are:
Nafion® from DuPont [9] is a standard SPE material. The central Rc  and Ra  (.cm2 ): electronic resistance of electron-
grey area in Fig. 1a is the PTFE fibre which is used for mechanical conducting metallic cell components at cathode and anode
reinforcement. The acidity and ionic conductivity of Nafion® prod- respectively.
ucts is equivalent to that of 1 M sulfuric acid aqueous solutions. Rel (.cm2 ): ionic resistance of the SPE.
This is why platinum-group metals (PGMs) are used as electrocat- Rc ct (.cm2 ): cathodic polarization resistance associated with
alysts (non-PGM metals are prone to corrosion/dissolution in such the HER.
strong acidic media). The SPE is surface coated with two catalytic Ra ct (.cm2 ): anodic polarization resistance associated with the
layers, a thin (0.1-1 ␮m thick) layer of carbon-supported platinum OER.
nano-particles at the cathode for the HER (label 2 in Fig. 1a) and a Qc dl (F.cm−2 .sn-1 ): cathodic constant phase element of the cath-
thicker (1-5 ␮m thick) layer of un-supported IrO2 particles at the ode/electrolyte interface; n is related to the angle of rotation of a
anode for the OER (label 2 in Fig. 1a). Catalysts particles must be purely capacitive line on complex plane plots [11].
strongly bonded to the SPE surface to sustain the large mechanical Qa dl (F.cm−2 .sn-1 ): anodic constant phase element of the
constraints that occur during operating due to the massive produc- anode/electrolyte interface.
tion of gases. In addition, catalytic layers must be sufficiently porous Zc D (.cm2 ): cathodic diffusion impedance due to H2 transport
to let gases go away from interfaces, and sufficiently conductive away from the cathode.
in order to let current flow homogeneously and to reduce para- Za D (.cm2 ): anodic diffusion impedance due to O2 transport
site ohmic resistances. The MEA is clamped between two porous away from the anode and/or to H2 O transport to the anode.
162 C. Rozain, P. Millet / Electrochimica Acta 131 (2014) 160–167

Fig. 1. (a) cross-section of a PEM water electrolysis cell (thicknesses not to scale); 1- PTFE-reinforced SPE; 2- cathodic catalytic layer; 2 : anodic catalytic layer; 3- cathodic
current collector; 3 - anodic current collector; 4- cathodic spacer; 4 : anodic spacer; 5- cathodic bipolar plate; 5 - anodic bipolar plate; (b) associated equivalent electrical
circuit (see text for details).

3.3. Reactions mechanisms in place of bulk Pt particles: the mean diffusion path is limited to
the nanometer scale due to the geometry of the catalytic powder,
3.3.1. HER and carbon superficial diffusion is unlikely to happen due to strong
Today, owing to the extended work of pioneering electro- carbon-H interactions and low H diffusivity. On particularly acti-
chemists such as Frumkin, Conway and Parsons, only two reactions vated Pt surfaces, the rate determining step has been reported to
paths involving two steps each are regarded as likely to occur in be the diffusion of molecular hydrogen away from the interface
acidic media (in Table 1, Had denotes a surface-adsorbed hydrogen [15]. Because of the high reversibility of the H+ /H2 redox couple on
ad-atom). Had species are formed by proton electronation (Volmer Pt in acidic media, transport of gaseous hydrogen across the porous
step 1). Molecular hydrogen (gaseous) evolves either by direct current collectors of PEM water electrochemical cells could possi-
chemical recombination of surface hydrogen ad-atoms (Tafel step bly have a negative effect on the kinetics of the HER, in particular
2) or by electrochemical recombination (Heyrovsky step 2 ). When at high current densities. However, there is no indication of such
Pd or Pd-based materials are used as electrocatalysts [12], the HER effects reported in the literature.
is in competition with the hydrogen insertion reaction (HIR) that
brings two additional reaction steps: 3.3.2. OER
surface sub-surface Many different mechanisms have been reported in the literature
MHad < > M + Hab (step3)
to account for the kinetics of the OER in acidic media on metallic
oxides [16]. The most often quoted are described in Table 2. On
Hab sub-surface < > Hbulk (step4) Pt electrodes, mechanism 1 usually prevails. Experimental transfer
coefficients for the OER and the ORR (oxygen reduction reaction)
where Hab denotes an absorbed hydrogen atoms in sub-surface
at high current density are both close to 0.5 (corresponding to a
regions. Fully hydride Pd particles are also good HER catalysts.
Tafel slope of 118 mV.dec−1 at 25 ◦ C and 130 mV.dec−1 at 80 ◦ C)
According to Frumkin et al. [13], steps (1) and (3) form one sin-
[17]. With 4 electrons exchanged in the overall reaction, this cor-
gle step whereas according to Bockris et al. [14], they are separated
responds to a stoechiometric number ␯ = 4, a value consistent with
steps.
mechanism 1, the first step being rds. However, Pt is not a good
On platinum, in acidic media, there is a general agreement
catalyst for the OER (and is thus never used for application in PEM
that the HER mechanism is that of a fast proton discharge fol-
water electrolysis cells), the exchange current density (approxi-
lowed by rate-determining chemical desorption (mechanism 1).
mately 10−9 A.cm−2 per cm2 of real interface area [17]) being far
The exchange current density of the HER on Pt in acidic media
is 10−3 A.cm−2 . The same situation is likely to occur at the cath-
ode of PEM cell when carbon-supported Pt nano-particles are used Table 2
two main OER mechanisms on Pt and IrO2 (S = reacting site).

Table 1 Mechanism 1: oxide path (Pt) Mechanism 2: Krasil’shchikov

two main HER mechanisms on Pt in acidic media. path (IrO2 )

Mechanism 1: Mechanism 2:
4 Pt + 4 H2 O → 4 PtOH + 4 H+ + 4e- S + H2 O → S-OH + H+ + 1 e−
H+ + 1 e− + M < > MHad surface H+ + 1 e− + M < > MHad surface (step 1) (step 1)
(step 1) (step 1) 4 PtOH → 2 PtO + 2 PtH2 O (step 2) S-OH → S-O− + H+ (step 2)
2 MHad surface < > H2 (g) + 2 M H+ + 1 e− + MHad surface < > 2 PtO → O2 + 2 Pt (step 3) S-O− →S-O + e− (step 3)
(step 2) H2 (g) + M (step 2 ) S-O → S + ½ O2 (step 4)
 C. Rozain, P. Millet / Electrochimica Acta 131 (2014) 160–167 163

Fig. 3. typical CV measured at the IrO2 anode of a PEM water electrolysis cell
(1.8 mg.cm−2 ). v = 50 mV.s−1 .

Fig. 2. (a) typical CV measured at the C-Pt cathode of PEM water electrolysis cell;
carbon-supported platinum nano-particles deposited with the
(b) CV on smooth Pt foil, 1 M H2 SO4 . v = 50 mV.s−1 .
automated ink printer are approximately 350-400, depending on
coating parameters. In conventional MEAs, the roughness factor
too small. Iridium metal or iridium oxides are known as the best of PEM cathodes usually lies in the 100-600 range, depending on
OER promoters in acidic media. The exchange current density mea- the morphology of catalysts particles, thickness of catalytic layer
sured on smooth iridium electrodes is approximately one thousand and plating process [24]. Half-cell measurements can also be used
times higher (≈ 10−6 A.cm−2 per cm2 of real interface area [18]) to probe anodes. A typical CV measured on a PEM water elec-
than on Pt. According to the literature, this is the Krasil’shchikov trolysis anode containing unsupported IrO2 is plotted in Fig. 3.
path of Table 2 [19] which is most frequently found to occur on Integration of the cyclic voltamogramm in the 100-1200 mV poten-
iridium oxide electrodes [20,21]. tial range yields a coulombic charge that can be used to evaluate
the IrO2 -SPE roughness factor. The reference charge for IrO2 is
4. Results And Discussion 310 ␮C.cm−2 [25]. Roughness factors measured in the course of this
study on MEAs containing unsupported IrO2 particles are ≈ 700-
4.1. Half-cell and full-cell electrochemical characterization 800. Although IrO2 particles used in the anodic catalytic layer are
larger than the Pt nano-particles used at the cathode, the difference
In cells using a liquid electrolyte, individual electrode elec- of roughness is due to the difference of thickness of catalytic lay-
trochemical characteristics can be measured separately using a ers. It should be noted here that fr values tend to decrease markedly
reference electrode. In a PEM water electrolysis cell, the anode-to- when the potential scan rate increases, indicating that some react-
cathode distance is very small (usually in the 0.05-0.2 mm range, ing sites are less accessible than others for the charge transfer
depending on the thickness of the SPE) and the electrolyte is a reactions.
solid. It is therefore not possible to introduce a conventional (e.g.
saturated calomel electrode) reference electrode or even a cap- 4.1.2. Polarization curves and cell voltage components
illary in order to separately probe anode or cathode responses. An interesting characteristic of PEM water electrolysis is that
However, several techniques described in the literature are using PEM cells can be operated in the multi-Amp current density range
a so-called “internal reference electrode” to by-pass this problem without mass transport limitations (Fig. 4-a), as long as the struc-
[17,22]. ture and porosity of current collectors is appropriate. Industrial
PEM water electrolyzers usually operate in the 0.8-2.5 A.cm2 range
4.1.1. Roughness factors but lab-scale experiments have already shown that much higher
Fig. 2-a shows the cyclic voltamogram measured at the cath- current densities (up to 10 A.cm−2 ) can be reached. The differen-
ode of a PEM water electrolysis cell using an internal reference tial cell resistance (in .cm2 ) is obtained by taking the first order
electrode (RHE = Reversible Hydrogen Electrode). For comparison,
Fig. 2-b shows the CV measured at a platinum foil in 1 M sulphuric
acid. In a PEM cell, the shape of the CV is usually somewhat dis-
torted due to several reasons: (i) oxygen traces (due to oxygen
cross-permeation from the anode) are reduced at the cathode in the
potential range of the CV and as a result, this tend to bend the CV
towards negative currents; (ii) the response of carbon-supported
Pt-nanoparticles contains an increased capacitive contribution due
to the highly developed carbon surface that smooth out plat-
inum features; (iii) it is difficult to neatly clean the SPE; traces
of alcohols used during the plating process tend to adsorb at the
surface of the cathode and are difficult to remove. Nevertheless,
the main features of the PEM Pt-cathode are sufficiently well-
marked to evaluate (at least approximately) the roughness factor
fr of the interface which is the dimensionless ratio of the total Pt-
electrolyte interface in cm2 divided by the geometrical surface of
the MEA in cm2 . The reference charge measured on a smooth Pt
electrode is 210 ␮C.cm−2 [23]. Cathodic roughness factors mea- Fig. 4. (a) polarisation curve measured on a PEM water electrolysis cell at 80 ◦ C; (b)
sured in the course of this study on MEAs containing commercial cell resistance of the PEM cell as a function of operating current density.
164 C. Rozain, P. Millet / Electrochimica Acta 131 (2014) 160–167

Fig. 5. individual voltage terms of the polarisation curve of Fig. 4-a in the restricted
0-1 A.cm−2 range.

Fig. 6. EIS impedance spectra measured at different cell voltages and 80 ◦ C on a 23

derivative of the polarisation curve. A plot of the cell resistance cm2 PEM water electrolysis cell.
as a function of operating current density is provided in Fig. 4-b
The maximum cell resistance obtained is 250 m.cm2 . High fre-
quency cell impedance measurements made with and without MEA (Fig. 7). This is again an indication of non-ideal capacitive behaviors
show that the electronic resistance of metallic cell components [26]. It would be tempting to attribute one semicircle to each elec-
is approximately 120-150 m.cm2 . The resistance of the MEA is trode process (the HER at the cathode and the OER at the anode).
approximately 100-130 m.cm2 (see plot b, Fig. 4), in agreement However, half-cell EIS measurements reveal (Fig. 8) that the con-
with values reported in the literature for this kind of SPEs. Half- tribution of the HER to the overall cell impedance is small. At low
cell measurements are also useful to separately plot the voltage frequencies (close to stationary conditions), it can be seen that the
drop of each cell voltage contributor as a function of operating polarization resistance of the cathode is much lower (≈ 8 times)
current density (Fig. 5). Figs. 4 and 5 show that at low current den- than the polarization resistance of the anode. This is not a surprise
sity, in the so-called activation domain, ohmic drops are negligible since the HER in acidic aqueous media is known to be a fast process
and the current–voltage relationship is controlled by the charge- [15] (see section 3.3). On the contrary, the OER which is a two-
transfer kinetics of both HER and OER, the OER overvoltage being electron process is rather slow and strongly irreversible at ambient
significantly larger than the HER one. At higher current densities, temperature [27]. From the characteristic frequencies at the sum-
the current voltage relationship is mostly linear in shape. In this mit of the arcs of Fig. 8, time constants of the HER and OER are
domain, impedances of charge transfer processes at both anode and obtained (0.003 s and 0.4 s respectively). Taking the diameters of
cathode become constant and the ohmic resistance of the cell (the
sum of electronic and ionic resistances) becomes the predominant
term.

4.1.3. Cell impedances and reaction mechanisms

As can be seen from Fig. 6, shapes and magnitudes of cell
impedance spectra change significantly with the applied cell volt-
age. At low cell voltage (U < E◦ = 1.18 V at 80 ◦ C), charge transfer
resistances are infinite and impedance spectra are mostly capaci-
tive in shape. As discussed above, this is not the behaviour of a pure
capacitance (a line parallel to the imaginary axis would be expected
in that case) because of the 3D structure of the electrode/electrolyte
interfaces, the description of which requires the use of constant
phase elements (CPEs) [11,26].
As the cell voltage increases, polarization resistances tend to
decrease. Increasing faradaic (stationary) currents flow across the
cell and impedance spectra gradually bend down toward the real
axis at low frequencies. Depressed charge transfer semicircles
are obtained. Diameters of these semicircles (which provide a
direct measure of the overall, ie. anode + cathode, polarization resis-
tances), tend to decrease when the cell voltage further increases.
When the cell voltage is sufficiently high (> 1.8 V), polariza-
tion resistances become extremely small and the cell impedance
becomes mostly ohmic (cell resistances are the sum of electronic
and ionic internal resistances). In the intermediate cell potential
range (1.4–1.6 V), the impedance arc can be described by two Fig. 7. zoom of EIS impedance spectra of Fig. 6 for Ecell (mV) = 1- 1400; 2- 1425; 3-
depressed semicircles having their centers below the Re(Z) axis 1450; 4- 1475; 5- 1500; 6- 1525 mV.
 C. Rozain, P. Millet / Electrochimica Acta 131 (2014) 160–167 165

4.2. Role of cell voltage and operating temperature on cell

impedances

Considering the characteristics of experimental impedance

spectra (one arc attributed to the OER that can be described by
two depressed semicircles, no mass transport limitations), the
equivalent electrical circuit of Fig. 1 has been simplified by consid-
ering only the series connection of two (RQ) circuits: R (RHF QHF )
(RLF QLF ). More explicitly:
1 1
Zcell ≈ R˝ + n
+ n
(1)
(j ω) QHF + R1 (j ω) QLF + R1
HF LF
√
In Eq. 1, j = −1, ␻ is the pulsation in rad.s−1 , R (in .cm2 ) is
the ohmic resistance of the cell obtained from the high frequency
Fig. 8. impedance spectra of (•) the cathode (HER), (o) the anode (OER) and (+) total
cell measured at 80 ◦ C and 1.4 V. (−) best fit using Eq. (1). intercept of the impedance spectra with the real axis in Nyquist
plots (R = Rc  + Rel + Ra  ), RHF and RLF (in .cm2 ) are the polariza-
tion resistances of the high (HF) and low (LF) frequency depressed
semicircles respectively, and QHF and QLF (in F.cm−2 .sn-1 ) are their
the arcs as the value of polarization resistances and neglecting CPE constant phase elements. The impedance of a CPE is [11,26]:
effects, then the capacitance of each interface can be estimated.
1
After correction for the roughness factors (¸ 4.1.1), the following ZQ = (2)
n
capacitance values are obtained: CHER ≈ 60 ␮F.cm−2 and COER ≈ Q (j ω)
50 ␮F.cm−2 . They are in agreement with literature data for smooth As discussed above, Q is a frequency independent constant in
metal/electrolyte interfaces. It can be concluded from these differ- F.cm−2 .sn-1 and n is a factor ranging between 0 and 1. Q tends to a
ent observations that, in the equivalent electrical circuit of Fig. 1, pure capacitance C as n tends to 1.
the charge transfer resistance of the cathodic process (Rc ct ) can Experimental impedance spectra measured at different cell vol-
be neglected and that the features of experimental impedance tages (Figs. 6 and 7) and different operating temperatures (not
diagrams are mainly due to the anode. The two depressed semi- represented here), have been fitted with Eqs. 1 and 2. From the
circles observed on the experimental arcs can be attributed to two best fits, polarization resistances of both high (RHF ) and low (RLF )
mono-electronic steps. These observations are consistent with the frequencies have been determined as a function of the cell voltage
Krasil’shchikov mechanism reported in the literature for the OER over the 0-90 ◦ C temperature range. Results are plotted in Fig. 9.
on iridium oxide (Table 2) but additional research work is still Polarization resistances tend to decrease with increasing tempera-
required to clarify the point. Mass transport limitations across cur- ture and cell potential (anode potential). At cell voltages larger than
rent collectors were not observed over the multi-Amp.cm2 range approximately 1.8-1.9 V, cell losses due to charge transfer processes
under investigation. This is an indication that the porous current become negligible and the impedance of the PEM cell is mostly
collectors have the appropriate porosity. Therefore, it can be con- ohmic. CPE’s exponents were found to be less than unity, close to
cluded that PEM water electrolysis cells can be characterized by 0.9, in agreement with literature values for rough interfaces [26].
measuring their entire cell impedance and that (at least in a first Interface capacitances have been determined from CPE values as
approximation), impedance spectra are due to the OER which is the described in Ref. [28]. As discussed above, the capacitance of the
rate-determining step (rds). low frequency semicircle was found to be consistent with that of the

Fig. 9. plot of polarization resistances of the cell as a function of cell voltage for different operating temperatures (a) RHF in .cm2 ; (b) RLF in .cm2 (no IR correction).
166 C. Rozain, P. Millet / Electrochimica Acta 131 (2014) 160–167

stacks. They can also be used to quantify and follow degradation

processes during long-term operation.

5. Conclusion

Results reported in this paper are related to the electrochem-

ical characterization of PEM water electrolysis cells. They were
obtained using membrane-electrode assemblies containing unsup-
ported IrO2 catalyst at the anode for the oxygen evolution reaction
and carbon-supported platinum nano-particles at the cathode for
the hydrogen evolution reaction. Roughness factors of anodes and
cathodes have been obtained from cyclic voltammetry experiments
using an internal reference electrode. Cell impedance spectra have
been measured at different cell voltages along the polarization
curve (1.4 to 1.9 V) and at different operating temperatures (0 to
90 ◦ C). It is shown that charge transfer reactions are major cell
impedance contributors at cell voltages up to approximately 1.8-
Fig. 10. Picture of a PEM water electrolysis stack (source CETH2 Co.). 1.9 V). At higher cell voltages, the cell impedance is mainly resistive
due to the ohmic resistance of metallic cell components and ionic
resistance of the polymer electrolyte. No mass transport limitation
OER double layer but no clear physical meaning could be attributed was observed over the multi Amp.cm−2 range of operating cur-
to the capacitance of the other time constant at high frequency. rent density under investigation. Two different time constants are
The assumption that the OER in PEM water electrolysis cell pro- observed on experimental impedance cell spectra. It is shown that
ceeds according to the Krasil’shchikov mechanism, as reported in they can both be attributed to the two-electrons oxygen evolution
the literature for iridium oxide electrodes, is still speculative and reaction because the kinetics of the hydrogen evolution reaction
a deeper analysis of the OER mechanism is still required to clarify is much higher. The two polarization resistances have been mea-
the point. sured over the conventional ranges of operation of PEM water
electrolyzers (1.4-1.9 V and 0-90 ◦ C). Results are consistent with
4.3. EIS characterization of PEM stacks the Krasil’shchikov reaction mechanism but a deeper analysis is still
required for confirmation. Finally, some results related to the EIS
EIS is also an interesting tool for the characterisation and characterization of large PEM water electrolysis stacks using 600
optimization of PEM water electrolysis stacks. Fig. 10 shows the cm2 membrane electrode assemblies are also reported. It is shown
photograph of a PEM water electrolysis stack that can deliver up that EIS measurements can be used to assess the homogeneity of
to 10 Nm3 H2 /hour, equipped with 600 cm2 MEAs. The high fre- the electrical environment and to analyze degradation processes.
quency cell resistances (ohmic internal cell resistance) and the In terms of perspectives, there is still a need to elucidate the OER
anodic charge densities of each individual cell (obtained by integra- mechanism and to determine the rate determining step.
tion of CVs measured on the anode side) measured at open potential
voltage and in the early stages of the activation process are plot- AcknowledgementS
ted in Fig. 11. There is a significant discrepancy of datapoints from
one cell to the other and cell resistances are high because SPEs are Financial supports from the French Agence Nationale de la
not yet fully hydrated. At the end of the activation process that Recherche, within the frame of the Progélec program (Aitoiles
may take several days, cell resistances tend to values close to 250 project ANR-2011-PRGE-001-03) and the Emergence program
m.cm2 (as for laboratory experiments) and are more homoge- (HydroPEM project ANR 11 EMMA 048 01), and from the French
neous throughout the stack (not shown here). This is an illustration “Conseil General de l’Essonne” within the frame of the ASTRE pro-
that such routine EIS measurements can be used to evaluate the cedure are gratefully acknowledged.
homogeneity of electrical environment in PEM water electrolysis
References

[1] O. Bicakova, P. Straka, Production of hydrogen from renewable resources and

its effectiveness, Int. J. Hydrogen Energy 37 (2012) 11563.
[2] N. Muradov, Hydrogen via methane decompositio: an application for decar-
bonization of fossil fuels, Int. J. Hydrogen Energy 26 (2001) 1165.
[3] J. Andrews, B. Shabani, Re-envisioning the role of hydrogen in a sustainable
energy economy, Int. J. Hydrogen Energy. 37–2 (2012) 1184.
[4] L. Xiao, S.–Y. Wu, Y.–R. Li, Investigation of thermo-chemical cycles for hydrogen
production, Renewable Energy 41 (2012) 1.
[5] P. Millet in: Electrochemical Technologies for Energy Storage and Conversion,
chapter 9, Water electrolysis for hydrogen generation, R-S. Liu, X., Sun, H., Liu,
L. Zhang and J. Zhang Editors, J. Wiley & Sons (2011).
[6] Hydrogen energy and fuel cells. Directorate General for Research, report EUR
20719 EN, European Commission, Brussels, 2003.
[7] P. Millet, R. Ngameni, S.A. Grigoriev, N. Mbemba, F. Brisset, A. Ranjbari, C. Etié-
vant, PEM water electrolysis: from electrocatalysis to stack development, Int.
J. Hydrogen Energy 35 (2010) 5043.
[8] M. Carmo, D.L. Fritz, J. Mergel, D. Stolten, A comprehensive review on PEM
water electrolysis, Int. J. Hydrogen Energy 38 (2013) 4901.
[9] C. Heitner-Wirguin, Recent advances in perfluorinated ionomer membranes:
structure, properties and applications, J Membrane Sci. 120 (1996) 1.
[10] S.A. Grigoriev, P. Millet, S.A. Volobuev, V.N. Fateev, Optimisation of porous cur-
Fig. 11. plot of anodic charge density and high frequency cell impedance as a func- rent collectors for PEM water electrolysers, Int. J. Hydrogen Energy 34 (2009)
tion of cell number in the stack of Fig. 10. 4968.
 C. Rozain, P. Millet / Electrochimica Acta 131 (2014) 160–167 167

[11] G.J. Brug, A.L.G. Van Den Eeden, M. Sluyters-Rehbach, J.H. Sluyters, The anal- [20] J.F.C. Boodts, V.A. Alves, L.A. Da Silva, S. Trasatti, Kinetics and mechanism of oxy-
ysis of electrode impedances complicated by the presence of a constant gen evolution on IrO2–based electrodes containing Ti and Ce acidic solutions,
phase element, J. Electroanal. Chem. And Interfacial Electrochem. 176 (1984) Electrochim. Acta 39 (1994) 1585.
275. [21] S. Trasatti, In: The oxygen evolution reaction, H. Wendt Ed., Electrochemical
[12] S.A. Grigoriev, M.S. Mamat, K.A. Dzhus, G.S. Walker, P. Millet, Platinum and pal- hydrogen technologies, Elsevier, Amsterdam, 1990.
ladium nano-particles supported by graphitic nano-fibers as catalyst for PEM [22] P. Millet, Water electrolysis using EME technology: electric potential distribu-
water electrolysis, Int. J. Hydrogen Energy 36 (2011) 4143. tion inside a Nafion membrane during electrolysis, Electrochim. Acta 39 (1994)
[13] A.N. Frumkin, in: P. Delahay (Ed.), Advances in Electrochemistry and Electro- 2501.
chemical Engineering, 3, Interscience, New-York, 1963. [23] S. Trasatti, O.A. Petri, Real surface area measurement in electrochemistry, Jour-
[14] J.O’M. Bockris, J. Mc Breen, L. Nanis, The hydrogen evolution kinetics and hydro- nal of Electroanalytical Chemistry 327 35 (1992) 3–376.
gen entry into ␣-iron, J. Electrochem. Soc. 112 (1965) 1025. [24] P. Millet, T. Alleau, R. Durand, Characterization of membrane electrodes assem-
[15] J.O’M. Bockris, A.K.N. Reddy, in: Modern Electrochemistry, Plenum Press, New- blies, for solid polymer electrolyte water electrolysis, J Appl. Electrochem. 23
York, 1970. (1993) 322.
[16] J.O’M. Bockris, Kinetics theory of adsorption intermediates in electrochemical [25] R.F. Savinell, R.L.Z. Iii, J.A. Adams, Electrochemically Active Surface Area Voltam-
catalysis, J. Chem. Phys. 24 (1956) 817. metric Charge Correlations for Ruthenium and Iridium Dioxide Electrodes,
[17] P. Millet, R. Durand, M. Pinéri, Preparation of new solid polymer elec- Journal of The Electrochemical Society 137 (1990) 1–6.
trolyte composites for water electrolysis, Int. J. Hydrogen Energy 15 (1990) [26] E. Barsoukov, J.R. Macdonald, in: Impedance Spectroscopy, Theory, Experiment,
245. and Applications, Wiley Ed., 2005.
[18] A. Damjanovic, A. Dey, J.O’M. Bockris, Electrode kinetics of oxygen evolution [27] A. Damjanovic, Electron transfer through thin anodic films in oxygen evolution
and dissolution on Rh, Ir and Pt-Rh alloy electrodes, J. Electrochem. Soc. 113 at Pt electrodes in alkaline solutions, Electrochim. Acta 37 (1992) 2533.
(1966) 739. [28] C. Hsu, F. Mansfeld, Concerning the Conversion of the Constant Phase Element
[19] A.I. Krasil’shchikov, Zh. Fiz. Khim 37 (1963) 273. Parameter into a Capacitance, Corrosion 57 (747) (2001).