Journal of The Electrochemical

Society

OPEN ACCESS You may also like

- Polymer Electrolyte Water Electrolysis:
Proton Transport in Catalyst Layers of a Polymer Correlating Porous Transport Layer
Structural Properties and Performance:
Electrolyte Water Electrolyzer: Effect of the Anode Part I. Tomographic Analysis of
Morphology and Topology
Tobias Schuler, Ruben De Bruycker,
Catalyst Loading Thomas J. Schmidt et al.

- Critical Review—Identifying Critical Gaps

To cite this article: Ugljesa Babic et al 2019 J. Electrochem. Soc. 166 F214 for Polymer Electrolyte Water Electrolysis
Development
Ugljesa Babic, Michel Suermann, Felix N.
Büchi et al.

- Transient and Steady State Two-Phase

View the article online for updates and enhancements. Flow in Anodic Porous Transport Layer of
Proton Exchange Membrane Water
Electrolyzer
Mateusz Zlobinski, Tobias Schuler, Felix
N. Büchi et al.

This content was downloaded from IP address 193.190.2.250 on 02/05/2023 at 15:24

F214 Journal of The Electrochemical Society, 166 (4) F214-F220 (2019)

Proton Transport in Catalyst Layers of a Polymer Electrolyte

Water Electrolyzer: Effect of the Anode Catalyst Loading
Ugljesa Babic, 1 Elisabeth Nilsson,1 Alexandra Pătru, 1
Thomas J. Schmidt, 1,2,∗
and Lorenz Gubler 1,∗,z
1 Electrochemistry Laboratory, Paul Scherrer Institut, 5232 Villigen PSI, Switzerland
2 Laboratory of Physical Chemistry, ETH Zürich, 8093 Zürich, Switzerland

The understanding of loss terms in polymer electrolyte water electrolysis (PEWE) cells is essential to maximize efficiency and
minimize cost and enable the technology for energy applications, such as “power-to-X”. We adapt the use of the transmission-line
model for porous electrodes in conjunction with electrochemical impedance spectroscopy measurements of the cell in the H2 /N2
mode known for fuel cells to PEWE cells to quantify proton transport resistance and double layer capacitance of the anode catalyst
layer. Anode catalyst loading was varied between 0.05 and 3.2 mgIrO2 cm−2 . A non-linear relationship was found between the anode
IrO2 loading and the proton transport resistance. Catalyst layers with very low IrO2 loading (0.05–0.16 mgIrO2 cm−2 ) had ∼4 times
higher mass-specific activity and ∼2 times higher mass-specific capacitance, revealing an inhomogeneous utilization of the catalytic
material. The kinetic and transport limitations associated with the anode catalyst layer have been correlated with its morphological
features.
© The Author(s) 2019. Published by ECS. This is an open access article distributed under the terms of the Creative Commons
Attribution 4.0 License (CC BY, http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse of the work in any
medium, provided the original work is properly cited. [DOI: 10.1149/2.0341904jes]

Manuscript submitted September 4, 2018; revised manuscript received January 24, 2019. Published February 22, 2019.

The dynamic capabilities and wide current density operation range supported,15–17 high surface area catalysts18–20 or a nanostructured thin
of polymer electrolyte water electrolysis (PEWE) make it attractive film catalyst21 is necessary to enable higher penetration of PEWE. On
for converting renewable electricity from the grid to H2 in the context the other hand, reduction of the IrO2 loading is only justifiable while
of “power-to-X”.1–6 PEWE is especially attractive in power grids with maintaining high voltage efficiency of the PEWE stack.
a high share of fluctuating renewables, where the H2 can be used for Understanding the factors that determine the catalyst utilization is
seasonal energy storage. H2 as an energy vector is versatile since it can necessary to direct the development of the next-generation catalyst
be re-electrified in a fuel cell for grid balancing, used in the mobility layers for the electrochemical water splitting technology. A set of di-
sector in fuel cell electric vehicles, or utilized in chemical industries agnostic tools is required to benchmark catalyst layer performance and
to promote deep decarbonization of the energy sector.1,5 Furthermore, make correlations between the morphology and the overpotential in
produced H2 can be used in a downstream methanation process to the polarization characteristics of a PEWE cell. In this study we present
form synthetic natural gas (SNG) using suitable, preferably renewable a detailed methodology for obtaining the proton transport resistance
CO2 and to feed it into the existing gas grid.6 Moreover, mixtures of H+
(RCLa ) and the double layer capacitance of the CLa from electrochemi-
H2 and CO2 can be used to produce syngas and subsequently liquid cal impedance spectra collected in the H2 /N2 regime. The approach has
hydrocarbon energy carriers via a gas-to-liquid process. been widely used in polymer electrolyte fuel cell (PEFC) technology22
The PEWE technology is currently plagued by issues associated and is adapted here for a PEWE cell. The impedance is analyzed us-
with cost and component durability.7,8 The oxygen evolution reaction ing the transmission-line model (TLM)22–26 to establish a correlation
(OER) in acidic medium on the anode is currently catalyzed by pre- between the physical characteristics of the CLa and the transport prop-
cious materials from the platinum group metals (PGM). However, the erties. The method is then tested on catalyst layers with loading varied
cost of the catalyst materials has been estimated to contribute only between 0.05 and 3.2 mgIrO2 cm−2 . The analysis is accompanied by
8% to the cost of the commercial stack.9 The final cost of the pro- cell voltage breakdown analysis based on the Tafel model previously
duced H2 is not greatly affected by the cost contribution of the stack reported in PEWE and PEFC literature.27–31
components.10 It is dominated by the cost of electricity (currently
∼70% of the hydrogen end cost), and is therefore heavily dependent
on the voltage efficiency of the stack.10 The type of the OER catalyst Experimental
and its loading on the other hand have a large impact the stack voltage
efficiency through the activation overpotential (ηact ).11,12 Cell and test-station.—A test bench developed in-house, equipped
Iridium oxide (IrO2 ) is the state-of-the-art PGM to catalyze the with a SP-150 potentiostat and a VMP3B 80A booster from Biologic
OER in commercial PEWE stacks.7,8 The most active material toward was used to perform experiments. MiliQ water was circulated at a flow
the OER is ruthenium oxide, but the dissolution rates remain too high rate of 400 mL min−1 from a heated gas-water separator through the
for commercial implementation in the aggressive PEWE anode envi- anode PEWE cell compartment. A PEWE cell with an active area of
ronment, in spite of extensive efforts to stabilize it.13 Higher loadings 25 cm2 with parallel flow fields was used to accommodate commer-
are preferred to maintain high stack voltage efficiency and reduce the cial T10 sintered-Ti porous transport layers (PTLs) from GKN. The
end cost of H2 . Maintaining a high IrO2 loading in the anode catalyst electro-osmotically dragged water was discarded periodically from
layer (CLa ) at high market penetration of PEWE in the energy sector the cathodic gas-water separator. Cell housing, inlet and outlet tem-
might not be viable due to the low abundance and limited supply of peratures were monitored during the experiments, and the average
iridium, especially considering it is a by-product of the Cu, Ni and Pt temperature was used to specify the operating temperature.
mining.14 The state-of-the art PEWE stacks (∼4 W cm−2 and 2 mgIrO2
cm−2 ) would require around 500 kgIrO2 /GW. This would allow for CCM preparation.—In-house prepared CCMs were based on the
an annual rate of capacity increase of 2 GW/a, considering the an- Nafion 117 (DuPont) perfluorosulfonic acid membrane. Nafion 117
nual production rate of iridium of 4 tons per annum and assuming a sheets of 10 × 10 cm were immersed in a ∼30 vol% HNO3 (VWR
use of 25% of this for PEWE.8 Reduction of the IrO2 loading using Chemicals) solution at 80°C for 1 h to protonate the sulfonic acid
groups and remove impurities. Boiling steps were repeated 4 times
in MiliQ water to remove the acid from the membrane. Membranes
∗ Electrochemical Society Member. were dried for 12 h before weighing, and afterwards mounted into a
z
E-mail: lorenz.gubler@psi.ch PEEK-based frame for spraying.
 Journal of The Electrochemical Society, 166 (4) F214-F220 (2019) F215

Table I. Properties of the IrO2 /TiO2 based catalyst layers for PEWE. The catalyst layer thickness (dCLa ) is obtained from the CCM cross-sections
observed with SEM. The anode catalyst layer proton transport resistance (RH+ CLa ) and the double layer capacitance (CDL ) values are extracted from
+ H+
the cell impedance measured in the H2 /N2 regime at 60°C. Proton resistivity of the anode catalyst layer (ρHCLa ) was calculated from RCLa and dCLa
values. The tortuosity (τ) of the electrode was calculated using Equation 5. The Tafel slope (b) and the anode catalyst activity (ia ) are extrapolated
from the iR-free polarization curves at 60°C. The mass specific quantities cm m
DL and ia are obtained by normalizing the calculated values to the IrO2
loading.
+
Loading (mgIrO2 cm−2 ) dCLa (μm) H+ ∗
RCLa (mΩ cm2 ) CDL ∗ (F) m ∗ (F g−1 )
cDL ia ∗∗ (A) iam ∗∗ (A g−1 ) b (mV/dec) ρH ∗
CLa (Ω cm) τ∗
0.05 1.0 ± 0.2 38.9 0.1 119 0.07 55 70 ± 2 1062 ± 12 23.7 ± 0.6
0.16 1.1 ± 0.1 6.2 0.5 115 0.14 46 57 ± 1 164 ± 4 4.6 ± 0.7
0.8 2.6 ± 0.8 10.3 1.1 54 0.27 16 54 ± 1 120 ± 8 3.4 ± 0.6
1.2 3.5 ± 0.6 4.4 1.4 45 0.33 12 54 ± 2 38 ± 7 1.1 ± 0.7
1.8 4.3 ± 0.5 25.6 1.8 40 0.65 15 55 ± 1 179 ± 6 5.1 ± 0.5
2.6 5.8 ± 0.3 27.6 3.9 60 0.85 15 56 ± 1 144 ± 6 4.1 ± 0.6
3.2 8.3 ± 0.4 35.7 4.4 55 1.32 20 55 ± 1 129 ± 5 3.7 ± 0.8

Uncertainties for the other calculated values are provided in Figures 4, 5, 8.

∗ Calculated from the EIS in the H2 /N2 regime.
∗∗ Calculated from the iR-free polarization curve at 1.47 V.

The anode catalyst ink was prepared by mixing 0.85 mL iso- rupted. The CLa of the 0.05 mgIrO2 cm−2 anode is very thin, and the
propanol, 2.7 mL miliQ water and 2.54 mL Nafion Solution (Aldrich, agglomerates are often disconnected. The surface catalytic agglom-
5 w% Nafion). The ink was sonicated for 15 minutes. 1 g of IrTiO2 erates of the thick CLa are ionically well interconnected through the
catalyst (Umicore, 66 mol% IrO2 and 34 mol% TiO2 ) was added to the ionomer covering the agglomerates underneath.
solution and sonicated for additional 30 min. The ink was sprayed onto
the membrane using an airbrush pencil (Conrad Electronics Airbrush Transmission line model for porous electrodes.—The AC
Pistol HP-200). The CCMs were dried again for 12 h and subsequently impedance response of the CLa is analyzed using the one-dimensional
weighed. The ink-spraying preparation method yielded anode load- transmission-line model (TLM), typically used for the characterization
ings from 0.05 to 3.2 mgIrO2 cm−2 . A gas diffusion electrode (Johnson of porous electrodes in electrochemical applications, including PEFC
Matthey ELE-0244-0542 with HiSpec 9100 Pt/C 0.4 mgPt /cm2 ) was CLs.22–26 According to the TLM, the CL consists of differential ele-
used as the cathode. The CCMs were assembled in the wet state after ments describing the proton transport resistance, double layer capaci-
immersion in miliQ water for 12 h before testing. tance (CDL ) and charge transfer resistance (RCT ) (Figure 2). The TLM
is often simplified by using an inert gas fed working electrode (WE),
Experimental conditions.—Cells were conditioned at 60°C by and a H2 -fed reference/counter electrode (RE/CE). When the potential
cycling the current density between 1 and 2 A cm−2 until stable per- is applied, the permeating H2 from the RE/CE cell compartment is the
formance and uniform temperatures in the system were achieved. Po- reactant for the permeation-rate limited hydrogen oxidation reaction
larization curves were recorded galvanostatically, while measuring (HOR) on the WE. The measured DC current corresponds to the hy-
the cell impedance at 10 kHz at every current step. The measurements drogen crossover current, which varies with the membrane thickness,
were repeated at 50, 60 and 70°C. PEWE cells were operated in the
H2 /N2 regime to measure the ionic resistance of the anode catalyst
layer. A N2 -stream of 500 Nml min−1 was injected into the work-
ing electrode (WE) water loop, while the reference/counter electrode
(RE/CE) was supplied with humidified H2 at 400 Nml min−1 . We en-
sured no faradaic current was measured before recording the electro-
chemical impedance spectrum (EIS). The impedance in H2 /N2 mode
was recorded from 10 kHz to 300 mHz potentiostatically at 1.0, 1.2
and 1.4 V. No faradaic current was observed at these potentials during
the collection of the impedance spectra. Cyclic voltammograms (CVs)
were measured in the potential range between 0 and 1.4 V vs. RHE at
a sweep rate of 50 mV/s.

Results and Discussion

Morphological analysis.—Post-test CCM samples were cryogeni-
cally fractured in liquid nitrogen and analyzed using a Zeiss Supra
VP55 scanning electron microscope, with an acceleration voltage of 5
kV. CCM cross-sections were previously sputter-coated with a ∼7 nm
Cr-layer for better electrical conductivity. CCM cross-sections were
investigated using the scanning electron microscope (SEM) to deter-
mine the thickness of the CLa with different loadings (Table I), which
was found to increase with 2.15 ± 0.13 μm mgIrO2 −1 cm2 . An airbrush
sprayed CLa consists of catalytic agglomerates 100–400 nm in size,
connected by an ionomer film (Figure 1). The structure contains pores
(voids) with a wide size distribution (0.1–1.5 μm in diameter). The
cross-section of the 0.05 mgIrO2 cm−2 anode shows only a few catalytic
agglomerates. The anode surface images reveal how the catalytic ag- Figure 1. Surface of the anode catalyst layers with a loading of a) 0.05 mgIrO2
glomerates are interconnected laterally. Lower loaded anodes are less cm−2 and c) 1.2 mgIrO2 cm−2 and their respective cross-sections b) and d),
homogeneous in-plane, and the ionic network is more frequently dis- observed under the scanning electron microscope.
F216 Journal of The Electrochemical Society, 166 (4) F214-F220 (2019)

The CDL can be obtained from the limiting impedance at low f.22 In-
stead of assuming constant Im(Z) and averaging it at low f,24 we have
estimated the CDL according to Equation 3 using Im(Z) at the minimum
phase angle (ϕmin ), when the transition to the capacitive behavior sets
in. The ϕmin is obtain at lower frequencies for thicker CLa , as the AC
current needs to penetrate longer pores. When the pores are penetrated,
the ϕ goes through a minimum value and increases asymptotically to
Figure 2. Equivalent circuit of the transmission line model for porous elec- 90o . The uncertainty of averaging is in this way mitigated, since the
trodes, consisting of differential elements describing the proton transport resis- Im(Z) in Equation 2 often does not exhibit a plateau at low f, especially
H+
tance in the catalyst layer RCLa , double layer capacitance CDL and the kinetic with low catalyst loadings.34 Another way to calculate CDL is by divid-
resistance related to the electrochemical reaction RCT . In the H2 /N2 regime, ing the average CV current in the potential range of 1.0–1.4 V by the
the impact of the kinetic resistances becomes negligible (RCT → ∞) due to sweep rate. CDL obtained from the CV appears to be higher compared
the near-absence of the faradaic reaction on the electrode. to the CDL calculated from the EIS in the H2 /N2 regime. (Figure 4a).
Higher CDL measured in the CV can be related to the pseudocapaci-
tance resulting from the fast electrosorption processes while cycling
temperature, pressure and relative humidity conditions in the cell. The the electrode potential. For both methods CDL increases with the anode
crossover current does not contribute to the total impedance, and the loading due to the increase of the absolute catalyst mass and therefore
low charge-transfer resistance of the HOR and the HER reactions al- the surface area.
lows for the simplification of the CLa equivalent circuit (Figure 2). The mass-specific capacitance (cDL m
) normalized to the amount of
The electrode-electrolyte interface in this case behaves as a simple IrO2 in the active area shows little deviation between the anodes
capacitor, which simplifies the analysis of the impedance spectra, as with different IrO2 loading, except for the samples with 0.05 and
the charge-transfer arc following the high-frequency 45° region toward 0.16 mgIrO2 cm−2 where it is approximately 2 times higher. Deviation
low frequencies vanishes (Figure 3).22,23 Deviation from the capacitive from the constant cDLm
could be related to the inhomogeneous utiliza-
behavior at low frequency could result from structural inhomogeneity tion of the catalyst in the case of thicker CLa with higher loading. A
of the electrode.32 Impedance spectra are recorded at 1.0, 1.2 and 1.4 V m
trend similar to the one of CDL and cDL vs. the IrO2 loading is observed
representing a potential region beyond the OER regime.33 (cf. below) when estimating the mass-specific activity from the iR-free
The absence of a faradaic reaction on the WE greatly simplifies polarization curves. It is assumed that the TiO2 in the catalyst layer
the impedance analysis and the estimation of the TLM equivalent is an insulator and that the electronic conductivity of the catalyst is
H+
circuit elements describing RCLa and CDL . The equation describing the ensured through the IrO2 percolation network.35 The capacitances are
impedance response of the CL22 can therefore be simplified according normalized only to the mass of IrO2 , since the contribution of TiO2
to would in this case be negligible. At high f , the 45o region appears in

H+
RCLa   the Nyquist plot (Figure 3). Here, the impedance can be described by
Z( f ) = HFR + coth H+
2π f CDL jRCLa [1]
2π f CDL j 
H+
where j is the imaginary unit (the symbol ‘i’ is used for current density). RCLa
Z( f ) ≈ HFR + [3]
In the special case when the frequency ( f ) tends to zero, the WE 2π f CDL j
impedance becomes
H+
RCLa j followed by the capacitive region at low frequencies, represented by
lim Z ( f ) = lim [Re(Z ) + jIm(Z )] ≈ − [2] the 90o line in the Nyquist plot and the HFR-corrected impedance
f →0 f →0 3 2π f CDL

Figure 3. a) HFR-corrected impedance response of the PEWE cell with 0.8 mgIrO2 cm−2 in the H2 /N2 regime measured at 50, 60 and 70°C, showing the
temperature dependence of RCLa H+
. Impedance of the cell with 3.2 mgIrO2 cm−2 is included to demonstrate the increase of RCLa H+
with the thickness of the anode
catalyst layer. b) Bode phase plot of the cell with 0.8 mgIrO2 cm−2 in the H2 /N2 regime measured at 50, 60 and 70°C, and of the cell with 3.2 mgIrO2 cm−2 at
H+
60°C. The RCLa /3 projection is obtained by first finding fmin at which ϕmin appears in b), and extracting the impedance at fmin from the Nyquist plot in a).
 Journal of The Electrochemical Society, 166 (4) F214-F220 (2019) F217

16 50

14 a) 45
From CV

CLa - Calculation (mOhm cm )

2
12 0.05
from EIS 40
3.2
10
35
CDL (F)

8 2.6
30
6 1.8
25
4
20
2

0 15
0.16
600
10

RH+
b) 0.8
500 1.2
5

0
IrO2)

400
0 5 10 15 20 25 30 35 40 45 50
(F g-1

300 RH+
CLa - Fitting (mOhm cm ) 2
cDL
m

200 H+
Figure 5. Comparison of RCLa values calculated using the fϕmin approach (cf.
Figure 3) and from the graphical extrapolation of the vertical line at low f. The
H+
minimum RCLa is observed for the sample with an IrO2 loading in the range
100
from 0.16 to 1.2 mgIrO2 cm−2 .

0
0.01 0.1 1 10
Figure 3 for samples of 3.2 vs. 0.8 mgIrO2 cm−2 . In this study we have
-2 H+
Anode catalyst loading (mgIrO2 cm ) used both methods and compared the RCLa values obtained for PEWE
anodes with catalyst loading ranging from 0.05 to 3.2 mgIrO2 cm−2 .
H+
Figure 4. a) Double layer capacitance CDL and b) mass-specific capacitance The values of RCLa obtained from the graphical extrapolation from the
m of anode catalyst layers with different catalyst loading estimated from the
cDL Nyquist plot and from the CDL at fϕmin are shown in Figure 5. Graphical
cyclic voltammogram (red) and the impedance in the H2 /N2 regime (black). H+
extrapolation resulted in higher values of RCLa in the case of very low
H+
and very high loadings, while both methods yield similar RCLa for the
−2
given by cells with 1.8 and 2.6 mgIrO2 cm .
H+
The measured RCLa decreases with higher cell temperatures due
H+ to the increased ionic conductivity of the ionomer in the CLa
RCLa 1
Z( f ) ≈ + [4] (13.0/10.3/9.1 mΩ cm2 at 50/60/70o C for 0.8 mgIrO2 cm−2 , respec-
f →0 3 jCDL f
tively, based on data from Figure 3). The contribution of the ionic
H+ H+
RCLa /3 can be extracted from the Nyquist plots of H2 /N2 operated cells resistance of the CLa appears as a loss term of magnitude RCLa /3 in
by projecting the low f impedance vertically to the real axis (Figure the polarization curve, as part of mass transport losses (ηmtx )24,33,38
3).22,32,33 An alternative approach is to determine CDL from Equation Improved proton transport in the CLa would partially account for the
3 and calculate RCLa H+ 24-26
. Graphical extrapolation of RCLaH+
from the temperature dependent ηmtx previously reported by Suermann et al.28
Nyquist plot relies on the presence of a vertical, capacitive impedance The change of ηmtx at 1 A cm−2 between 50 and 70o C was ∼7 mV
at low f, which is not always observed in the measured data.22,33,36 for the CL with 0.8 mgIrO2 cm−2 in this study. The change in the area
Deviations from the low f capacitive behavior can be attributed to resistance of a ∼3 μm thick Nafion ionomer sheet between 50 and
inhomogeneity of the catalyst layer, stemming from the preparation 70°C is in the range of ∼1 mΩ cm2 (measured value in the H2 /N2
process or caused by the coarse PTL surface.32,37 The morphologic regime ∼4 mΩ cm2 ), and can only account for a fraction of ηmtx .
H+
properties of the CL together with the CL/PTL interface properties Stronger sensitivity of RCLa to temperature variations during H2 /N2
might result in electrically insulated regions in the CL with high re- measurements compared to what is reported for Nafion39 could stem
sistance, accessible only at low f.37 Furthermore, the wetting of the from thin film limitations and the water distribution in the CL as a
CL pores could be non-uniform in PEWE,11,33 resulting in distributed factor of the PTL/CL interface, and will be discussed further in the
capacitance through the CL thickness. This can lead to a large variance overpotential breakdown section below.
H+ H+
in the extrapolation of the RCLa (Figure 5). On the other hand, approx- Considering CLa as a sheet electrode, RCLa would decrease with the
25 H+
imation of CDL from Equation 3 requires that –Im(Z) plateaus at low thickness, hence lower IrO2 loading. Two trends of RCLa have been
f.24 If this is not the case, the uncertainty in the calculation of CDL , and observed with varying loading in the H2 /N2 regime. RCLa H+
decreases
with decreasing loading in the range from 3.2 to 1.2 mgIrO2 cm−2 on
H+
subsequently of RCLa , increases. To circumvent the possible issues and
errors in the estimation of CDL , we propose a method of finding the the account of lower CLa thickness, since the proton transport path
CDL values at the frequency at which a minimum in the phase angle ϕ is longer for thicker catalyst layers. The lowest RCLaH+
values were ob-
is observed. This point marks the impedance transition from the 45o served for the IrO2 loading range from 1.2 to 0.16 mgIrO2 cm−2 . Inter-
region to the capacitive behavior. HFR-corrected impedance spectra H+
estingly, RCLa was highest for the sample with 0.05 mgIrO2 cm−2 . Pro-
in H2 /N2 mode are used to calculate the ϕ values (Figure 3), and the ton transport from the catalytically active sites relies both on through-
value of fϕmin is obtained at ϕmin . The method results in a defined knee- plane as well as in-plane transport when the structure is composed of
H+
point in the Nyquist plot based on the pore length, from which RCLa catalytic agglomerates connected by the ionomer. At very low IrO2
and CDL can be calculated. The fϕmin is lower for thicker catalyst layers loading applied using the airbrush method the CLa structure becomes
since the current has to penetrate longer distances, which is evident in less homogeneous, as observed in the SEM images. This results in a
F218 Journal of The Electrochemical Society, 166 (4) F214-F220 (2019)

lower degree of connectivity between the catalyst agglomerates by the 1.75

ionomer network and subsequently higher in-plane resistance contri-
H+
bution to the RCLa . Alternative catalyst structures, such as the nano- 1.70
structured thin film catalyst (NSTF) developed by 3M,21 and different 1.65

EIR-free (V)
application methods40 are necessary to achieve low-loaded anodes
with uniform structure. The deviation of the measured impedance 1.60
from the characteristic 45o slope at high f is more pronounced for 1.55
the CLa with low IrO2 loading (<1.2 mgIrO2 cm−2 ). This behavior of
the impedance has previously been ascribed to electrodes with open, 1.50
wedge-shaped pores,41,42 which is the case for low-loaded anodes with 1.45
inhomogeneous distribution and packing of the catalyst agglomerates. 0.55
H+
Other possible causes of non-linear behavior of RCLa with varying IrO2
loading are discussed in more detail in the section on overpotential 0.50
analysis, in the context of the mass transport losses.
The average proton resistivity of the anode catalyst layers with 0.45

ηact (V)
>0.05 mgIrO2 cm−2 (ρH+ CLa ) was found to be 130 ± 33 Ω cm, using
H+ 0.40
the RCLa and the CLa thickness (dCLa ) values from the SEM cross-
sectional images (Table I). The calculated value is in the range of 0.35
values typically obtained for a fuel cell catalyst layer under fully hu-
midified conditions at low ionomer to carbon ratio.26 The reason for 0.30
−2
a large variation in ρH+
CLa in the case of 0.05 mgIrO2 cm is the strong 0.25 0.05 mgIrO2/cm2
H+
deviation of the calculated RCLa from the decreasing trend with the 0.05 0.16 mgIrO2/cm2
IrO2 loading. ρH+CLa is normally found to be a property of the catalyst 0.8 mgIrO2/cm2
layer that is independent of its thickness.26 Our CLa structure is lat- 0.04
1.2 mgIrO2/cm2
erally very inhomogeneous at lower loadings, and the corresponding
H+
in-plane RCLa contribution results in an increase of the apparent ρH+ ηmtx(V) 1.8 mgIrO2/cm
CLa . 0.03
Taking the ionomer resistivity (ρH+ 2.6 mgIrO2/cm2
i ) from the literature and using the
ionomer volume fraction in the electrode (ɛi ), the tortuosity (τ) can be 0.02 3.2 mgIrO2/cm2
calculated:
H+ 0.01
ρCLa εi
τ(ε) = [5]
ρi 0.00 0.0 0.5 1.0 1.5 2.0 2.5 3.0
We obtained ɛi = 0.314, following the equations given in detail by -2
Liu et al.,26 and assumed ρH+ from 8 to 11 Ω cm26,43 to find that
Current density (A cm )
i
τ goes through a minimum close to 1 for the cell with a loading
Figure 6. EIR-free , ηact and ηmtx for the cells with different IrO2 loadings.
of 1.2 mgIrO2 cm−2 (Table I). The findings from the H2 /N2 measure-
Data were obtained by iR-correcting the polarization curves using the HFR at
ments suggest that the ionomer does not form a well-connected net- 10 kHz measured at 60°C and atmospheric pressure.
work between the catalytic agglomerates especially in the case of
0.05 mgIrO2 cm−2 , which is reflected in the RCLa
H+
and the ηmtx values in
the overpotential analysis. PTL/CLa interface suggest that O2 evolution is localized at the CLa in
direct contact with the PTL, while the catalyst in the PTL void (size:
Overpotential breakdown.—The overpotential breakdown of the 50–500 μm) remained inactive.45–48 The limiting factor toward opti-
PEWE polarization curve was conducted according to the procedure mal catalyst utilization was the insufficient electronic conductivity of
described in the Appendix. Polarization curves of PEWE cells with the catalyst layer.11,44–48 Assuming negligible differences in the in- and
differently loaded CLa have been iR-corrected using the HFR to elim- through-plane electronic conductivity of the CLa , the catalyst material
inate the influence of the variations in ohmic overpotential (η ) that
might stem from small variations in the cell assembly, membrane ionic
resistance and contact resistances with different PTL samples. The 1.8 60
HFR was found to be 257 ± 16 mOhm cm2 for all samples at 60°C, 1.6
without any correlation to the IrO2 loading. Increasing the CLa load- 50
ing results in lower EiR−free (Figure 6), which is lowest for the highest 1.4
loading (3.2 mgIrO2 cm−2 ) and highest for the lowest loading (0.05 1.2
mgIrO2 cm−2 ), as may be expected. The EiR−free values of the cells 40
a (A gIrO2)

with 0.16–3.2 mgIrO2 cm−2 have a constant Tafel slope b of 55 ± 1 1.0

-1
i (A)

mV/dec, and are offset by a factor depending on the CLa loading. The 30
0.8
Tafel slope is higher only in the case of the 0.05 mgIrO2 cm−2 sample
im

with 70 mV/dec. A similar effect of the CLa loading on the Tafel slope 0.6 20
was observed by Bernt et al.,11 albeit already at 0.2 mgIrO2 cm−2 . The 0.4
catalyst activity ia , expressed as the current density at a fixed EiR−free 10
of 1.47 V, increased linearly with the CLa loading, as expected. The 0.2
calculated mass activity iam is in the range of 10.2 ± 1.3 A g−1 for the 0.0 0
cells with CLa loading between 0.8 and 3.2 mgIrO2 cm−2 . However, 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
the iam for the cell with 0.16 mgIrO2 cm−2 is found to be approximately
three times higher (Figure 7). Decreasing the loading even further to Anode catalyst loading (mgIrO2 cm-2)
0.05 mgIrO2 cm−2 results in iam of 54.8 ± 4.2 A g−1 . The deviation
of iam in the case of CLa with a very low IrO2 loading (<0.2 mgIrO2 Figure 7. Activity in the form of the current measured at a fixed iR-free po-
cm−2 ) indicates that the catalyst utilization of higher loaded CLs is tential of 1.47 V (black), and the activity normalized to the IrO2 loading (mass
low.44 Operando optical imaging studies of the O2 nucleation at the activity, red).
 Journal of The Electrochemical Society, 166 (4) F214-F220 (2019) F219

CLa PEM H+
which would affect RCLa , and ultimately ηmtx . As the OER in PEWE
cells occurs below the PTL particles,45–48 the water has to diffuse both
laterally and perpendicularly to the membrane surface (Figure 8) from
H2O the PTL pore to the active sites below the PTL particle.33 Since the
Ideal structure water diffusion through Nafion thin films (< 60 nm) is orders of mag-
O2
nitude slower compared to the bulk,49,50 thicker CLs would result in
longer diffusion distances for the reactant, and thus yield an inhomo-
geneous water distribution in the CL. Development of an anodic PTL
with a microporous layer is necessary to alleviate problems which di-
minish the electronic and ionic connectivity of anode CLs with low
IrO2 loadings.11
H2O
Realistic, high loaded Conclusions
O2
Understanding the relationship between cell performance and the
structural features of the anode catalyst layer in PEWE is necessary
for tailoring next-generation catalyst layers for optimal transport prop-
erties and activity. We provide a methodology to determine the pro-
H+
ton transport resistance in the anode catalyst layer, RCLa , based on
H2O impedance measurements in the H2 /N2 regime by identifying the knee
point in the Nyquist plot corresponding to the minimum phase angle
Realistic, low loaded
observed in the Bode phase plot. The presented diagnostic method is
O2 a universal tool for characterizing catalyst layers for water electrolyz-
ers. Application of the method to anodes with varying IrO2 loading
from 0.05 to 3.2 mgIrO2 cm−2 has revealed that the loading affects RCLa H+
Figure 8. Schematic representation of the PEWE anode catalyst layer. The
ionomer phase conducts protons from the catalytically active surface sites. The
in a non-linear fashion. Overpotential analysis showed that there is a
H+
metallic phase provides the electronic conductivity of the layer. correlation between the trends observed in RCLa and the mass trans-
port overpotential ηmtx versus the IrO2 loading. Low loaded anodes
(<1.2 mgIrO2 cm−2 ) exhibit a high degree of in-plane inhomogeneity,
closer to the PEM would not be fully utilized. Although more catalyst which could explain the non-linear relation between the IrO2 loading
H+
contributes to the OER in the case of anodes with higher IrO2 loading, and RCLa . Other potential causes related to the water management in
the fraction of active material would be lower. This effect is illustrated the catalyst layer are discussed in the context of ηmtx values measured
in Figure 8. with different IrO2 loadings.
The variation of the IrO2 loading affected ηmtx in a similar fash- We have demonstrated that more IrO2 catalyst yields lower activa-
H+
ion as the RCLa obtained in H2 /N2 configuration, i.e. in the absence tion losses, as expected, yet a lower effective utilization of the material.
H+ Moreover, a systematic variation of the IrO2 loading has revealed a
of faradaic current (Figure 9). Since the RCLa contribution appears as
relationship between the activation overpotential ηact and ηmtx that can
a part of the ηmtx ,33,38 a good correlation in the trend of these two
to a certain extent compensate the apparent loss of voltage efficiency
independently determined quantities can be taken as a validation of
H+ in going toward lower IrO2 loading. Optimizing the morphology of
the transmission line model method to determine RCLa in PEWE cells.
the catalyst layer by adapting the preparation method could open a
Structural inhomogeneity of low IrO2 -loaded CLa related to the prepa-
H+ pathway to low loaded catalyst layers with high volumetric utilization
ration method could be the cause of the non-linear behavior of RCLa ,
of the active material.
and consequently ηmtx . Low IrO2 loadings were found to have a sim-
ilar effect on the electronic conductivity of the catalyst layer.11 It is
H+
important to note that RCLa is very sensitive to variations in the water Acknowledgments
content of the ionomer in the CLa .22,24–26 Bernt et al.11 have suggested
Funding by the Swiss Federal Office of Energy (grant-no. 501198-
that the water distribution might be different for thick and thin CLa ,
01) is gratefully acknowledged. TJS thanks Innosuisse and the Swiss
Competence Center for Energy Research Heat & Electricity Storage.
We acknowledge Katrine Elsoe (DTU) for performing preliminary
50 50 series of catalyst layer coating trials in the framework of a research
45 stay at PSI.
CLa/3 (mOhm cm )

40 40
2

35 Appendix
ηmtx (mV)

30 30 Overpotential analysis.—The electrolyzer cell voltage (Ecell ) is composed of the

reversible cell voltage (Erev ), ohmic (η ), kinetic (ηact ), and mass-transport (ηmtx ) over-
25 potentials (Equation A1).27–29,33
20 20 Ecell = Erev +ηact +η +ηmtx = Erev +ηact +i · (RPE M +Rel )+i ·RCLa
H+
+ηmtx−rest [A1]
15
RH+

Erev is given by the Nernst equation for the operating pressure, temperature and the activity
10 1 A cm-2 10 of the species involved in the reaction (Equation A2). We assume unity for the activity of
2 A cm-2 water and ideal gas behavior for the gases. For the temperatures used in this study, 50°C,
5
3 A cm-2 60 C, 70°C, Erev is 1.195 V, 1.186 V, 1.177 V, respectively.
0 0  √ 
0.1 1 10 R·T a(H2 ) · a(O2 )
Erev (p, T ) = Erev
0
(T ) + ln [A2]
2·F a(H2 O)
Anode catalyst loading (mgIrO2 cm-2)
ηohm is the result of losses related to the membrane proton transport processes (RPEM ) and
the contact resistances (Rel ) and follows Ohm’s law, ηΩ = i · (RPEM +Rel ).7,51 These losses
Figure 9. (red) and ηmtx (black) at 1 (), 2(●) and 3 () A
H+
RCLa /3 cm−2 for are obtained when correcting the polarization curve with the high-frequency resistance
anode catalyst layers with different IrO2 loading, 60°C. (HFR), in this case measured at 10 kHz. ηact is related to sustaining the sluggish anodic
F220 Journal of The Electrochemical Society, 166 (4) F214-F220 (2019)

oxygen evolution reaction (OER).52 The cathode reaction is assumed not to be kinetically 19. E. Oakton, D. Lebedev, M. Povia, D. F. Abbott, E. Fabbri, A. Fedorov, M. Nachtegaal,
limiting.38 ηact is obtained by subtracting Erev from the Tafel line, which is calculated from C. Copéret, and T. J. Schmidt, ACS Catalysis, 7, 2346 (2017).
the iR-free cell voltage (EiR−free ) at low current densities (0.01–0.1 A cm−2 ). ηmtx appears 20. D. Lebedev, M. Povia, K. Waltar, P. M. Abdala, I. E. Castelli, E. Fabbri, M. V. Blanco,
at higher current densities and is obtained from the difference of EiR−free and the fitted A. Fedorov, C. Copéret, N. Marzari, and T. J. Schmidt, Chemistry of Materials, 29,
5182 (2017).
Tafel line. Part of ηmtx is related to the proton transport resistance in the CLa ,11,33,38 and
21. K. A. Lewinski, D. F. v. d. Vliet, and S. M. Luopa, ECS Transactions, 69, 893 (2015).
can be obtained from the Nyquist plot of the EIS during H2 /N2 operation.33 22. S. S. Kocha, in Handbook of Fuel Cells, W. Vielstich, A. Lamm, H. Gasteiger, and
H. Yokokawa Editors, p. 538, John Wiley & Sons, (2010).
23. B. E. Conway, in Electrochemical Supercapacitors, p 382, Springer, (1999).
ORCID 24. Y. Liu, M. W. Murphy, D. R. Baker, W. Gu, C. Ji, J. Jorne, and H. A. Gasteiger,
Journal of The Electrochemical Society, 156, B970 (2009).
Ugljesa Babic https://orcid.org/0000-0002-6774-1001 25. W. Gu, D. R. Baker, Y. Liu, and H. A. Gasteiger, in Handbook of Fuel Cells,
Alexandra Pătru https://orcid.org/0000-0002-6555-590X W. Vielstich, A. Lamm, H. A. Gasteiger, and H. Yokokawa Editors, John Wiley &
Thomas J. Schmidt https://orcid.org/0000-0002-1636-367X Sons, Ltd, (2010).
26. Y. Liu, C. Ji, W. Gu, D. R. Baker, J. Jorne, and H. A. Gasteiger, Journal of The
Lorenz Gubler https://orcid.org/0000-0002-8338-6994 Electrochemical Society, 157, B1154 (2010).
27. M. Suermann, T. J. Schmidt, and F. N. Büchi, Electrochimica Acta, 211, 989 (2016).
28. M. Suermann, K. Takanohashi, A. Lamibrac, T. J. Schmidt, and F. N. Büchi, Journal
References of The Electrochemical Society, 164, F973 (2017).
29. H. Ito, T. Maeda, A. Nakano, A. Kato, and T. Yoshida, Electrochimica Acta, 100, 242
1. R. Winkler-Goldstein and A. Rastetter, Green, 3, 69 (2013). (2013).
2. F. N. Büchi, M. Hofer, C. Peter, U. D. Cabalzar, J. Bernard, U. Hannesen, 30. T. J. Schmidt and J. Baurmeister, Journal of Power Sources, 176, 428 (2008).
T. J. Schmidt, A. Closset, and P. Dietrich, RSC Advances, 4, 56139 (2014). 31. H. A. Gasteiger, S. S. Kocha, B. Sompalli, and F. T. Wagner, Applied Catalysis B:
3. B. Bensmann, R. Hanke-Rauschenbach, G. Müller-Syring, M. Henel, and Environmental, 56, 9 (2005).
K. Sundmacher, Applied Energy, 167, 107 (2016). 32. M. C. Lefebvre, R. B. Martin, and P. G. Pickup, Electrochemical and Solid-State
4. M. Sterner and I. Stadler, in Energiespeicher - Bedarf, Technologien, Integration, Letters, 2, 259 (1999).
Springer Vieweg, (2014). 33. U. Babic, T. J. Schmidt, and L. Gubler, Journal of The Electrochemical Society, 165,
5. A. Buttler and H. Spliethoff, Renewable and Sustainable Energy Reviews, 82, 2440 J3016 (2018).
(2018). 34. R. Makharia, M. F. Mathias, and D. R. Baker, Journal of The Electrochemical Society,
6. T. J. Schildhauer and S. M. A. Biollaz, CHIMIA International Journal for Chemistry, 152, A970 (2005).
69, 603 (2015). 35. E. Oakton, D. Lebedev, A. Fedorov, F. Krumeich, J. Tillier, O. Sereda, T. J. Schmidt,
7. M. Carmo, D. L. Fritz, J. Mergel, and D. Stolten, International Journal of Hydrogen and C. Coperet, New Journal of Chemistry, 40, 1834 (2016).
Energy, 38, 4901 (2013). 36. D. Malevich, B. R. Jayasankar, E. Halliop, J. G. Pharoah, B. A. Peppley, and K. Karan,
8. U. Babic, M. Suermann, F. N. Büchi, L. Gubler, and T. J. Schmidt, Journal of The Journal of The Electrochemical Society, 159, F888 (2012).
Electrochemical Society, 164, F387 (2017). 37. M. C. Lefebvre, R. B. Martin, and P. G. Pickup, Electrochemical and Solid-State
9. L. Bertuccioli, A. Chan, D. Hart, F. Lehner, B. Madden, and E. Standen, Study on Letters, 2, 259 (1999).
Development of Water Electrolysis in the EU by E4tech Sàrl with Element Energy Ltd 38. M. Bernt and H. A. Gasteiger, Journal of The Electrochemical Society, 163, F3179
for the Fuel Cells and Hydrogen Joint Undertaking (2014). (2016).
10. DOE Technical Targets for Hydrogen Production from Electrolysis, Office of Energy 39. B. R. Matos, C. A. Goulart, E. I. Santiago, R. Muccillo, and F. C. Fonseca, Applied
Efficiency & Renewable Energy, accessed on 11.07.2018, retrieved from http://energy. Physics Letters, 104, 091904, (2014).
gov/eere/fuelcells/doe-technical-targets-hydrogen-production-electrolysis , (2015). 40. K. E. Ayers, J. N. Renner, N. Danilovic, J. X. Wang, Y. Zhang, R. Maric, and H. Yu,
11. M. Bernt, A. Siebel, and H. Gasteiger, Journal of the Electrochemical Society, 165, Catalysis Today, 262, 121 (2016).
F305 (2018). 41. D. Cericola and M. E. Spahr, Electrochimica Acta, 191, 558 (2016).
12. P. Millet, N. Mbemba, S. A. Grigoriev, V. N. Fateev, A. Aukauloo, and C. Etiévant, 42. H. Keiser, K. D. Beccu, and M. A. Gutjahr, Electrochimica Acta, 21, 539 (1976).
International Journal of Hydrogen Energy, 36, 4134 (2011). 43. S. M. Slade, T. R. Ralph, C. P. d. León, S. A. Campbell, and F. C. Walsh, Fuel Cells,
13. O. Kasian, S. Geiger, P. Stock, G. Polymeros, B. Breitbach, A. Savan, A. Ludwig, 10, 567 (2010).
S. Cherevko, and K. J. J. Mayrhofer, Journal of The Electrochemical Society, 163, 44. C. Rozain, E. Mayousse, N. Guillet, and P. Millet, Applied Catalysis B: Environmental,
F3099 (2016). 182, 153 (2016).
14. Umicore, Precious Materials Handbook | A specialised Book about the 45. J. Mo, Z. Kang, S. T. Retterer, D. A. Cullen, T. J. Toops, J. B. Green, M. M. Mench,
contribution of metal-based materials to crucial applications in line with and F.-Y. Zhang, Science Advances, 2, e1600690 (2016).
specific megatrends, Umicore AG & Co. KG, Hanau-Wolfgang, Germany 46. Z. Kang, J. Mo, G. Yang, S. T. Retterer, D. A. Cullen, T. J. Toops, J. B. Green Jr,
(2012). M. M. Mench, and F.-Y. Zhang, Energy Environ. Sci., 10, 166 (2017).
15. H. Xu, B. Rasimick, M. Riera-Smith, B. Pivovar, S. Alia, K. C. Neyerlin, K. Lewinski, 47. Z. Kang, G. Yang, J. Mo, S. Yu, D. A. Cullen, S. T. Retterer, T. J. Toops, M. P. Brady,
and S. Luopa, High-Performance, Long-Lifetime Catalysts for Proton Exchange G. Bender, B. S. Pivovar, J. B. Green, and F.-Y. Zhang, International Journal of
Membrane Electrolysis, DOE Hydrogen and Fuel Cells Program, Annual Progress Hydrogen Energy, 43, 14618 (2018).
Report, (2015). 48. Z. Kang, G. Yang, J. Mo, Y. Li, S. Yu, D. A. Cullen, S. T. Retterer, T. J. Toops,
16. H.-S. Oh, H. N. Nong, T. Reier, M. Gliech, and P. Strasser, Chem. Sci., 6, 3321 G. Bender, B. S. Pivovar, J. B. Green, and F.-Y. Zhang, Nano Energy, 47, 434 (2018).
(2015). 49. E. M. Davis, C. M. Stafford, and K. A. Page, ACS Macro Letters, 3, 1029 (2014).
17. L. C. Seitz, C. F. Dickens, K. Nishio, Y. Hikita, J. Montoya, A. Doyle, C. Kirk, 50. S. A. Eastman, S. Kim, K. A. Page, B. W. Rowe, S. Kang, C. L. Soles, and
A. Vojvodic, H. Y. Hwang, J. K. Norskov, and T. F. Jaramillo, Science, 353, 1011 K. G. Yager, Macromolecules, 45, 7920 (2012).
(2016). 51. C. Rakousky, U. Reimer, K. Wippermann, M. Carmo, W. Lueke, and D. Stolten,
18. D. F. Abbott, D. Lebedev, K. Waltar, M. Povia, M. Nachtegaal, E. Fabbri, Journal of Power Sources, 326, 120 (2016).
C. Copéret, and T. J. Schmidt, Chemistry of Materials, 28, 6591 52. E. Fabbri, A. Habereder, K. Waltar, R. Kötz, and T. J. Schmidt, Catal. Sci. Technol.,
(2016). 4, 3800 (2014).