Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Advances in Benchmarking and Round Robin Testing For PEM Water

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Applied Energy 352 (2023) 121898

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Advances in benchmarking and round robin testing for PEM water


electrolysis: Reference protocol and hardware
Thomas Lickert a, Stefanie Fischer b, James L. Young c, Selina Klose a, Irene Franzetti a,
Daniel Hahn a, Zhenye Kang c, Meital Shviro b, c, Fabian Scheepers b, Marcelo Carmo b, d,
Tom Smolinka a, Guido Bender c, Sebastian Metz a, *
a
Fraunhofer Institute for Solar Energy Systems ISE, 79110 Freiburg, Germany
b
Forschungszentrum Jülich, GmbH, IEK-14: Electrochemical Process Engineering, 52425 Jülich, Germany
c
National Renewable Energy Laboratory (NREL), Golden, CO 80401, USA
d
Nel Hydrogen, 10 Technology Drive, Wallingford, CT 06492, USA

H I G H L I G H T S

• A detailed measurement protocol for PEM water electrolysis is developed.


• A test cell for harmonized measurements in non-harmonized test benches is presented.
• For a set of CCM and PTLs, in-house and across-sites reproducibility is quantified.
• Reference polarization curve data and impedance spectroscopic data are presented.
• Differences in conditioning behavior and thermal management are analyzed.

A R T I C L E I N F O A B S T R A C T

Keywords: While the number of publications in the PEM water electrolysis community increases each year, no common
PEM water electrolysis ground concerning reference hardware (test cells and test bench) and testing protocols has been yet established.
Harmonized test protocol This would, however, be necessary for the comparability of experimental results. First attempts for such refer­
Benchmarking
ence hardware and procedures have been made in the framework of the Task 30 Electrolysis within the Tech­
Reproducibility
Conditioning of catalyst coated membrane
nology Collaboration Programme on Advanced Fuel Cells (AFC TCP) of the International Energy Agency (IEA).
AFC TCP Task 30 Since then, improvements of both the test hardware (test cell and components) as well as the measurement
protocol were identified, and a revised methodology and key results based on a comprehensive measurement
series have been obtained. A detailed protocol for testing commercial reference components with a reference
laboratory test cell developed in-house by Fraunhofer ISE is presented. For evaluation of the protocol and the
hardware, it was tested at three different institutions at the same time. Impedance spectroscopic and polarization
data was acquired and analyzed. The obtained differences in performance were calculated to give the community
an expectation window to compare own data to. Finally, the importance of a thorough temperature control and
the conditioning phase are demonstrated.

1. Introduction cells, stacks, assembled with all kinds of components and materials [5].
A comprehensive review on early works in PEMWE is given in the works
Polymer electrolyte membrane water electrolysis (PEMWE) will of Carmo et al. [6]. Since then, numerous further works with different
experience a rapid market ramp-up in the future as green hydrogen will hardware and test conditions have been published [7–16]. More than
play a key role in transforming the energy system towards a carbon free ever, harmonization and benchmarking activities are crucial to make
system [1–4].More and more companies, universities and scientific group to group comparisons possible and increase the trust that litera­
groups are producing a large amount of data on different kinds of test ture results are reproducible and accurate.

* Corresponding author.
E-mail address: Sebastian.metz@ise.fraunhofer.de (S. Metz).

https://doi.org/10.1016/j.apenergy.2023.121898
Received 31 May 2023; Received in revised form 1 August 2023; Accepted 2 September 2023
Available online 16 September 2023
0306-2619/© 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
T. Lickert et al. Applied Energy 352 (2023) 121898

Since the last publication by the Task 30 Electrolysis benchmarking institutions that were measured with their own hardware as long as one
group [16] of the AFC TCP, significant improvements in cell perfor­ validated measurement with the reference hardware exists. The second
mance [5,7,17–21], catalyst systems and loadings [19,20,22–31], will generally improve the reproducibility of creating experimental re­
porous transport layers (PTLs) [32–37] and durability have been re­ sults in general. Both together would enable the meaningful comparison
ported [9,11,38]. across the entire PEMWE research community and help to bridge the
Iridium and platinum are still the most wildly used catalysts and “valley of death” between academic research and industrial application
hence gold standard in PEMWE. The actual catalyst content is still [59].
higher than it would be necessary with optimal catalyst utilization, so The participating institutions of this contribution evaluated the
numerous research activities have aimed at catalyst reduction while method and the system, defined the reference cell components, and
keeping the performance approximately unchanged. [19,20,22–31] As developed a harmonized testing procedure. The used test cell is
especially iridium is scarce and expensive novel catalyst concepts as currently available through Fraunhofer ISE. The cell components, like
supported catalysts [23,25,39] or more advanced structures, such as the catalyst coated membranes (CCMs) and PTLs, which together form the
Sr2CaIrO6 mixed oxide, are of great interest. Sr2CaIrO6 possesses sacri­ membrane electrode assembly (MEA), are commercially available. The
ficial components, such as Ca or Sr, which leach out during operation overall objective of this work is to present this potential reference
and cause a strong reconstruction of the catalyst layer (CL) [40]. measurement method, the reference hardware system and give the PEM
On the other side of the interface, the PTLs and their contact with the water electrolysis community expectation values for performance,
CLs are decisive for performance and durability [11,22,37,40]. reproducibility, and variation to other groups, that can be expected.
Increasing current density causes significant performance loss due to
dry-out and oxygen saturation. [22] Therefore, modifying the structure 2. Experimental setup
and composition of the PTL has become another area of interest for the
researching community. Various treatments have been studied, such as To understand the choices made in the process of developing a
the effect of titanium hydride formation on performance and durability reference protocol for PEM electrolysis, a good understanding of the
or classical etching to remove surface passivation [37]. Similarly, laser underlying principles is essential. Describing these underlying principles
ablation has been used to improve the contact between the patterned comprehensively and to a level of detail they deserve goes beyond of this
PTL and the lightly loaded CL [41]. For an improved CL-PTL interface, article. We strongly encourage the reader to consult some of the excel­
these two components can no longer be developed independently, an lent descriptions of fundamental processes already available in the
approach that is only in its infancy so far. In any case, there is a tendency literature [6,9,60–65]. Together with the measurement protocol to be
to reduce the thickness of the PTL. used, the experimental hardware setup is one of the key challenges in a
While the above mentioned advanced systems show individual suc­ round robin benchmarking exercise. One of the decisions to be taken is
cesses, a unifying element connecting these results is missing [42] as cell which hardware needs to be identical for all partners and which hard­
hardware [21,26,28,43–49], cell components [18,27,28,49,50], condi­ ware differences need to be accepted. It is obvious, that the test cell as
tioning procedures [16,28,51,52] and testing protocols [16,52–56] well as the components need to be harmonized, as differences can have a
differ considerably. As a result, it is still nearly impossible to compara­ high impact on the measured results. For the test bench hardware, it
tively evaluate results from different working groups on PEM water needs to be accepted that this is and will always be to a certain extent
electrolysis, despite the steadily increasing interest and the rapidly different between different groups.
growing number of publications on this topic. Tomić et al. [52] address
this issue by conducting a comparative analysis and critical review of 2.1. Test cell
different proposed degradation protocols, which were experimentally
validated to provide a better basic understanding. They also confirmed For this benchmarking activity a 4 cm2 test cell previously developed
the urgent need for harmonized testing protocols, especially for the by Fraunhofer ISE within a project financed by the Federal Ministry of
initial period, called conditioning. Moreover, various national funding Education and Research of Germany in the framework of Power-MEE
agencies as the European Commission, the Department of Energy (DOE) (03SF0536E) was employed. As shown in Fig. 1a, the cell is housed in
in the USA or the German Federal Ministry of Education and Research a compression cage which contains a centered screw for applying a
have also recognized this shortcoming and do support the development variable compression force to the cell. The applied forced is monitored
of harmonized test protocols, to enable better benchmarking of project with an integrated force sensor (K-K14/N410-G25 sensor, Lorenz Mes­
results. However, these efforts are still in their infancy and comparable stechnik GmbH, Germany) during assembly, start up and operation
results are not yet available or published. Recently, the European Joint (Fig. 1a). The anode and cathode compartments of the cell feature Pol­
Research Centre (JRC) has described detailed individual procedures yetheretherketone (PEEK) frames which hold the PTL and determine
which can be used to set up a measurement protocol with harmonized PTL compression (Fig. 1b). To seal the cell, Fluoroelastomer (FKM)
terminology [54,57,58] for all main low temperature technologies of based flat sheet sealing (Freudenberg, IceCube, 60 FC-FKM 200) is used.
water electrolysis. Among other things, certain test conditions, design of The parallel flow field of the cell is shown in Fig. 1b. It consists of 10
hardware, number and positions of sensors, their measurement accuracy channels with a width and depth of 1.0 mm. The channels are separated
and analysis routines for important characteristic values are proposed. from each other by 9 land sections which have a width of 0.9 mm. The
This very comprehensive conception is based on the experience of the entire flow field is encased by a 0.9 mm land section. To run round robin
participating industry and research institutions but can only be under­ tests with PTLs that provide reproducibility on an industrial standard,
stood as a starting point for harmonization, since application in practice laser cut titanium felts with 1 mm thickness from Bekaert (2GDL40–1,0),
is lacking so far and no comparative measurements have been carried sputtered with platinum on both sides (more information in Section 2.2)
out or published yet. A detailed, internationally applied, and harmo­ were used.
nized reference measurement protocol is still not published, nor has it The thickness tolerances of both, the PTLs and PEEK frames reached
been critically experimentally evaluated. up to +60 μm. This significant variation was mitigated by determining
With the continued work of our benchmarking group within the Task the exact thickness of the PTL and pairing it with a specific pocket depth
30 of the AFC TCP, we aim to further the discussion on harmonization of of the PEEK frame (Fig. 1c). For all cells operated in this work, the
testing protocols and hardware. For improvement, a significant effort overstand of the PTL over the height of the measured pocket depth was
should be placed on (i) identifying an open-source cell hardware that adjusted to 70 μm ± 10 μm. A detailed description of the adjustment
can serve as a reference system, and (ii) defining test conditions and process is given in the supplementary data. A similar approach was
protocols. The first will enable to connect results from various published by Martin et al. [43]. To warranty small and reproducible

2
T. Lickert et al. Applied Energy 352 (2023) 121898

Fig. 1. Compression cage with test cell in it (a), details of a single half cell (b) and schematic of the test cell setup (c).

contact resistances between the titanium base material of the flow-field E-5 bar (10 mTorr). The sputtering power setpoint was 20 W yielding a
and the PTL, a gold coating on top of a thin contact layer of platinum for current of about 70 mA. A sputtering break-in time of two minutes was
better adhesion was applied. For electrical contacting to current lead allowed before opening the shutter to commence deposition. The same
and voltage sense cables, each flow field block contains holes for a 2 mm process was repeated for each sample side. The deposition time was 15
and a 4 mm banana plug connector. min to yield a Pt loading of approximately 0.1 mg*cm− 2 as measured by
An additional 2.1 mm hole is provided to position a temperature x-ray fluorescence (XRF) spectroscopy (Fischer XDV-SDD).
sensor at a position close to the electrochemical reaction in 4 mm dis­ CCMs: A commercial three-layer CCM from Greenerity, using a
tance to the lower edge of the flow field, so that temperature changes Nafion® 115 membrane, a Pt-based cathode and an IrO2-based anode
can be detected as fast as possible. The fluidic connections are realized was used. Within the actual measurements, fully coated CCM sheets
by a G1/8 in. thread with a length of 7 mm. were used, cut to size, and distributed between the labs. In the cell setup,
the flat sheet sealings of cathode as well as anode were hence in direct
2.2. Cell components contact with the catalyst layers of the CCM. While this might cause some
longtime effects in electrolysis stacks on the industrial scale, no issues
This section includes the components and materials used to collect were observed with this approach throughout the measurements re­
the data for the herein described study. A description of state-of-the-art ported in this paper. For reproducibility and contact resistance reasons,
materials and components with latest references are described in Section the cells were assembled with dry CCMs, and the humidification and
1. swelling of the latter happened during the conditioning phase.
PTLs: The PTL materials consist of titanium felts with 1 mm thickness For details on the material preparation and cell assembly, the reader
from Bekaert (2GDL40–1,0), coated inhouse on both sides at Lab 3 and is referred to the supplementary data.
used on the anode side as well as the cathode side. The titanium felts
were cut to size using a MiniLase™ Auto Door laser marking system to fit 2.3. Test benches
the sample compartment. After removing burrs from the cut PTL sam­
ples and smoothing of the edges, the PTL materials were ultrasonically As mentioned in the test cell section, harmonization efforts con­
cleaned: in acetone, isopropanol, ethanol, and finally in deionized (DI-) cerning hardware are limited to the test cell and test cell components
water for 10 min each. Self-heating of the ultrasonic resonator leads to (PTL, CCM, coating). Harmonization of test benches across the wider
increasing temperatures during cleaning, but always remained below community is rather hard to realize and hence was deliberately not
60 ◦ C. At the end of cleaning, the materials were rinsed in DI-water and attempted within this study. It was rather the intention to check, how
allowed to air dry (more details in the supplementary data). Next, the and to what extend harmonization can be achieved with conceptionally
titanium felt PTLs were sputter-coated with platinum on both sides. This different test benches and lab environments. Two conceptually different
Pt coating was applied using a custom sputtering system in Lab 3. After test benches are depicted at the beginning of the supplementary data.
the samples were placed in the sputter coater, the chamber was evacu­ Test benches can be set up simply from scratch with standard piping
ated to a base pressure of ~6.7E-9 bar (~5E-6 Torr). Argon gas (UHP and process control. The advantage of these test benches is that they can
99.999% Ar gas) was supplied at approximately 15 sccm (standard cubic easily be adapted to a specific operation mode. Test benches that can be
centimeter per minute) to establish a setpoint chamber pressure of ~1.3 bought are usually more potent, but can be complex to operate and the

3
T. Lickert et al. Applied Energy 352 (2023) 121898

option of adaption is more limited. In that respect, some differences


within the test bench setup were accepted on purpose for this contri­
bution to reflect the reality in the scientific community. Major differ­
ences are:

• Water feed: Lab 1 and Lab 2 recirculate DI-water once it passed


through the cell flowing through an ion exchanger resin to stabilize
water conductivity below the required value (< 1 μS*cm− 1). Lab 3
uses the so-called single pass mode. DI-water passes through the cell
and flows to the drain after leaving the test cell.
• Water conductivity: At Lab 2, the conductivity of the DI-water is
continuously controlled by sensors inside the gas-water separators,
whereas at Lab 1 it is only checked before and after each test. At Lab
3 conductivity is defined by controlling the input water stream.
• Pressure control and differential pressure: One of the labs is located
at higher altitude, and therefore the environment is at lower pres­
sure. Backpressure valves are used to increase pressure values to sea
level. At another lab, located almost on sea level, a slight over­
pressure at the cathode is needed for process control reasons and
hence a small differential pressure is generated. At the third lab, both
sides are open to the environment and the lab is located almost at sea
level. Fig. 2. Positioning of the six temperature sensors in the two independent half-
• Electronics: All three partners use different electronic loads and cell loops.
different frequency response analyzers (FRAs). Resulting differences
in e.g. cable inductivities, measurement precision or simply the ne­ • Variation 03: Anode inlet (T3); Cathode inlet (T4)
cessity of changing cables during tests to be able to use the devices • Variation 04: Anode outlet (T5); Cathode outlet (T6)
makes it a generally underestimated influence on measurement data.
The test protocol that was developed and is published with this On the first glance, Variation 01 would be considered as the most
contribution copes with this by introducing so called down times. A reproducible set-up, as the position of the temperature sensor inside the
more detailed description of the protocol is explained in Section 3. A cell would be defined to the highest level of harmonization: It would be
description of the electronic hardware that was used is given in identical for all partners. However, not for every test bench the inte­
Section 2.3.2. gration of a Pt100 temperature sensor to be used as temperature control
• Dynamic of heat source: Across the three labs, different heat sources point can easily be done. In fact, Lab 2 had to integrate the temperature
were used. At Lab 1, two thermostats with individual water baths, sensor of the test bench into their cell and for Lab 3, it wasn’t possible to
pumps and heating loops are used and are connected to heat ex­ integrate all the 6 sensors. As it was realized, how complicated the usage
changers that transfer the heat to the fluidically decoupled reactant of Variation 01 might turn out in the laboratory, additional variations
water circle. The other two labs use resistive heating solutions with temperature sensors being implemented close to inlet/outlet on the
directly applied to the reactant water tube. At Lab 2, a metallic heat anode/cathode side were tested. Lab 3 recorded Variation 02 for all
block, which the tube is integrated in, is heated electrically. At Lab 3, measurements, herein as a control case for the other labs that varied
a resistive heat tape is wrapped around the tubing and heated elec­ their temperature regulation points.
trically. Thermal dynamics are therefore higher for at Lab 2 and Lab
3. 2.3.2. Electronic hardware
Electronic devices might even be more diverse than test benches
throughout the community. Results from an internal survey conducted
2.3.1. Thermal management of anode and cathode by Fraunhofer ISE, during an impedance workshop, revealed that ten
The temperature has a major effect on the performance of an elec­ different suppliers for impedance hardware cover a large part of the
trolysis cell and even small temperature changes can have an impact. community (based on the limited number of participants at the work­
Therefore, precise control of the temperature inside the cell is manda­ shop). For direct current (DC) power supply, the number is probably
tory for reproducibility. As the test cell used in this study doesn’t provide even higher. Due to this variety, and the differences that come along
the option of external heating (e.g. by a separate heating loop or a heat with hardware differences, e.g. test bench operation or data acquisition,
patch on the cell body), preheated fed water is used for temperature a measurement protocol that is as universal as possible, should cope for
control. As the Joule heat generated inside the cell and the amount of this, to be applicable by as many groups as possible. For the herein
thermal losses (i.e. thermal radiation of the cell to the surrounding air) described electrochemical tests, three different electronic setups are
can vary depending on the current density, a feedback loop based on the used and listed in Table 1.
measured temperature of the cell has been implemented at all three labs.
For most of the current densities, there is a temperature gradient across 3. Test protocol / methodology of characterizing test cells
the test cell, such that the position of the temperature sensor is an
important question that needed to be addressed for establishing a Unlike the field of PEM fuel cells, where a measurement protocol is
reproducible measurement protocol. Ideally, the herein presented setup available based on work coordinated by the DOE [66,67], no such pre­
should have 6 temperature sensors (Fig. 2) of which one on the anode cise and validated protocol exists for PEM water electrolysis. The JRC
and one on the cathode side needs to be chosen as temperature regu­ has compiled comprehensive documents on measurement procedures
lation point. Of the nine possible combinations, the following four var­ and terminology [54,57,58] which have been collated from the experi­
iations were tested: ence of several industrial partners and research institutions. As part of
the work, a test procedure for a stepwise ascending and descending
• Variation 01: Anode internal (T1); Cathode internal (T2) polarization curve in galvanostatic mode is presented in [57] but the
• Variation 02: Anode internal (T1); Cathode inlet (T4)

4
T. Lickert et al. Applied Energy 352 (2023) 121898

Table 1 3.1. Impact of non-harmonizable hardware on the test protocol


List of the used electronic devices by the three contributing partners.
Device Lab 1 Lab 2 Lab 3 The testing protocol used is based on the experience of the partici­
pating institutions and includes several compromises to perform the
DC power supply Zahner PP241 Sorensen XTR HP 6031A
6–110 same measurements. In the practical testing, as already in our first
Voltage Zahner Sorensen XTR Keithley Model 2000 Round Robin tests [3], the temperature control proved once more to be
measurement Potentiostat 6–110 one of the key issues. Analysis went so far as to check the influence of the
(PP241) positioning of the temperature sensor that controls the water circuit
Impedance device Zahner Zennium Biologic HPC- Autolab Metrohm
Pro 1005 PGSTAT 302 N
temperature on anode/cathode side either at the inlet pipe, the outlet
Current booster Power 100A Biologic 20 A Autolab pipe or inside the cell (see Section 2.3.1). A second major impact orig­
Potentiostat HCP-1005 Metrohm booster inates from the used electrical hardware (see Section 2.3.2): While some
(Zahner PP241) booster electrical devices can measure both, polarization curves as well as
Software In-house HyWARE II Multiscan
electrochemical impedance spectroscopy (EIS), others use difference
(electronic developed (Greenlight (Grandalytics)
control & data LabView VI Innovation) devices for these measurements. Hence in the test protocol, these mea­
recording) surements had to be decoupled to allow one of the participating partners
Software (process HyWARE II In-house developed to change the hardware by hand. Also, the heating and cooling rate to
control (Greenlight LabView VI for back adjust to the two different temperatures are different and depend on the
Innovation) pressure valves
test bench and/or the lab environment (exposure to sunlight or air ex­
change rates). To address the latter two points and to keep the protocols
protocol allows a considerable degree of flexibility, prone to different synchronized at the same time, an extra downtime of one hour was
ways of adaption, potentially leading to different results. Hence, it could added in several parts of the test protocol, see Section 3.2).
be considered only as a starting point for harmonization and – even more
important - its practical application and comparative measurements are 3.2. Test protocol description
lacking so far. A fully harmonized and tested reference measurement
protocol is still not published. It should be emphasized, that setting up The formal steps for the test protocol used are shown in Fig. 3.
such a general test protocol is a tremendous task and is not provided in Although not usually discussed in detail, preparation is a critical part. It
the actual publication, either. What we do provide is a detailed, includes selecting the PTL (which needs to fit the pocket depth of the test
harmonized protocol that has survived the practical trial in three indi­ cell, see Section 2.1), flattening the surface and/or edges with sand­
vidual groups, coordinating their work within the framework of their paper, and properly cleaning the surface and test cell. To ensure the
partly non-harmonizable test hardware, meaning test bench and elec­ highest level of reproducibility for each cell that was assembled, a fresh
tronics (see Section 2). set of PTLs, on anodic and cathodic side, was used to have comparable
states of the PTLs at beginning-of-test. For some CCMs, it is recom­
mended to pre-swell them in water before putting them into the cell, for
some others, this can lead to an excessive wrinkling which could even

Fig. 3. (a) General test script with iterated Characterization/Re-conditioning loop and (b) individual steps within the Characterization.

5
T. Lickert et al. Applied Energy 352 (2023) 121898

lead to delamination of the catalyst layer [29,68]. In the actual case – measurements, it was possible to compare the influence of the repeti­
and unlike earlier attempts – the cell was assembled using a dry, not pre- tions themselves on the one hand and the different temperature control
swollen CCM. By this approach, the membrane predominantly swells in settings on the other hand, see results Section 4.3.
the axial direction and performance data indicate that no delamination Finally, the two steps of a current leakage test and a conductivity test
is occurring inside the cell. In the next step, the cell is heated up to T = of the used water are re-run to check whether there have been any
80 ◦ C and the temperature was held for t = 60 min. Afterwards, the changes during the test protocol or not. While two labs have the
conductivity of the used water is tested (for those labs that do not have convenient ability to continuously monitor the water quality (in the test
an in-line check of the water conductivity) to avoid performance dif­ bench or via the house DI-water supply monitoring system), Lab 1 is
ferences due to contaminations. Values that must be achieved can be lacking this ability and can only confirm the water quality during the run
found in the detailed description of the protocol in the supplementary in an a posteriori measurement.
data. Next, a current leakage test is performed at three different volt­ The detailed measurement protocol is available in the supplementary
ages: 0.8 V, 1.0 V and 1.2 V. This is done, to determine, that there is no data.
short circuit, e.g. caused by one of the PTL-fibers or burrs from the
cutting process of the PTLs, that could penetrate the CCM. After these 3.3. Analysis of impedance data
two quality control steps, the first conditioning is started with two gal­
vanostatic steps for 30 min each at 0.2 A/cm2 and 1 A/cm2, respectively By fitting the measured impedance spectroscopic data to an equiv­
and a potentiostatic step for 15 h at 1.7 V. With the first, galvanostatic alent circuit model, additional insights into the individual loss mecha­
steps of the conditioning, we followed the suggestion of the CCM pro­ nisms can be obtained [45,69–73] In Fig. 4, an example of the python fit
ducer. The following potentiostatic mode allows to control any free is given for data from Lab 1 (MEA 2, T = 60 ◦ C, i = 1 A*cm− 2) is given. In
energy driven surface change mechanisms. We, however, identified that the actual case, an equivalent circuit with a linear sequence of an ohmic
when following this procedure, the CCMs are not fully conditioned at the resistance Rm and two parallel arrangements of a constant phase
beginning of the test. Hence the extension of the time at 1.7 V. element (CPE1/CPE2) and an ohmic resistance (R1/R2) was chosen and
The tested temperatures, pressure and water flow rate for the per­ realized by an in-house developed python code. For Lab 2, an additional
formance and reproducibility evaluation, has been chosen to cover state- inductance (La) was included to be able to correctly analyze the data
of-the art and future operating conditions in industrial PEM water (depicted in the graphs of Fig. 5). The fits were reviewed for their
electrolysis systems. For the temperatures, 60 ◦ C is widely used in the informative value using a chi-square (or χ2) test (statistical method to
industry and therefore usually used as a reference temperature [54], quantify the difference between data and the fit) for their real- and
80 ◦ C is aspired for the next years [4]. The pressure is chosen to meet the imaginary part for all frequencies, which are lower than 2% for all
requirements for most of the scientific groups. The flow rate was chosen spectra fits (χ2 < 1% for 93% of all frequency points). The upper part (a)
according to the channel structure of the used test cell to suppress mass shows the Nyquist plot of the data points (blue points, raw data), the
transport limitation with the used MEA and PTL combination. individual RC-elements (high frequency part ZHF in red, low frequency
The characterization phase begins with the measurement of the po­ part ZLF in green,) and Zsum (purple curve). The data fit is overlapping
larization curve at 80 ◦ C - the temperature, the conditioning tempera­ with Zsum and therefore not seen in Fig. 4. The lower part (b) shows the
ture, such that no temperature adjustment is necessary between these residuals, meaning the deviation between the fit and data points for both
two steps. First, the different measurement points of the polarization real (blue, ΔRe)- and imaginary parts (orange, ΔIm) for all frequencies.
curve are measured in the order of an increasing current density ("up- The main reason for an impedance analysis within this contribution
curve"). Afterwards, the same points are measured, but with opposite is that it’s needed for the iR-correction of polarization data. Herein, high
order, i.e. in the order of a decreasing current density ("down-curve"). frequency resistances (HFRs) were extracted from the respective Nyquist
After this step is finished, the temperature needs to be lowered to 60 ◦ C, plots for representative current densities (i = 0.5; 1,0; 2,0 and 2,5
hence a cooling down time is required, the duration of which depends on A*cm− 2) by fitting the data using an in-house developed python code
the test bench. To mitigate this test bench dependent difference of the from Lab 1. As limitations of the electronics (sampling precision at low
colling down time, an additional stabilization phase (1 h at 1.7 V) of one current density) led to insufficient noise levels for the lowest measured
hour was added, after which the polarization curve is measured.
With the same rationalization (no change in temperature to the
previous step), the EIS spectra were first measured at 60 ◦ C. Before this
was possible, Lab 2 had to manually adjust the electronic hardware
connections. To compensate for potential effects this voltage/current
disruption might have on the EIS spectra to be measured, another sta­
bilization phase (1 h at 1.7 V) was run. This was done at all sites to keep
the protocol as harmonized as possible. The EIS spectra (also in the later
step at 80 ◦ C) were measured at five different current density points (0.1;
0.5; 1.0; 2.0 and 2.5 A*cm− 2). After the EIS measurements at 60 ◦ C were
obtained, the temperature was increased to 80 ◦ C, with a stabilization
phase (1 h at 1.7 V) to compensate for the different heating rates at the
different sites. As last step of the characterization phase, the EIS spectra
at 80 ◦ C are obtained at the same five current densities as before at 60 ◦ C.
With this, the EIS measurements at both temperatures are finished.
The characterization phase is followed by a re-conditioning for 10 h
at 1.7 V. The practical reason for this phase was to ensure that there
wouldn’t be any need for an adjustment of the working hours for the
manual changing of the equipment at Lab 2, which is crucial in general,
Fig. 4. (a) Example of Nyquist plots with individual contributions for high
to make such a protocol usable for everyday lab work.
frequency (red line) and low frequency elements (green line), measurement
The two phases of characterization and reconditioning were repeated data (blue points) and the sum of both (purple line). (b) Corresponding dif­
in total 4 times. Depending on the laboratory, this was either done with ferences of data compared to measurement points for real (blue) and imaginary
different temperature control settings (Lab 1 and Lab 2) or with un­ part (yellow). (For interpretation of the references to colour in this figure
changed temperature control settings (Lab 3). Based on these repetitive legend, the reader is referred to the web version of this article.)

6
T. Lickert et al. Applied Energy 352 (2023) 121898

Fig. 5. (a) Polarization curves in uncorrected state (filled symbols) and iR-corrected state (half-filled symbols) for 60 ◦ C and 80 ◦ C and all three labs. (b)Corre­
sponding HFRs, extracted from impedance data and the used equivalent circuits.

current density of 0.1 A*cm− 2, the fit and therefore the HFR for 0.1 comparison across the labs. Lab specific observations and general data
A*cm− 2 was not included in the further analysis. The resulting iR- differences must be considered here to understand the data comparison
corrected graphs allows the distinction between ohmic and non-ohmic across the labs. In the top part of Fig. 5, the polarization curves for a cell
effects. assembly are shown at T1 = 60 ◦ C and T2 = 80 ◦ C together with the iR-
corrected curves. At the bottom part, the HFRs for the measured current
4. Experimental results and discussion densities are depicted together with the equivalent circuits that were
used to fit the impedance spectra. At the first glance, the measurements
In the results section, two different types of reproducibilities will be show a high reproducibility, but it must be accepted that for the un­
discussed, the “in-house” and the “across-sites” reproducibility. The in- corrected cell voltage data, already the in-house comparison (separate
house reproducibility is a measure for the ability at each institution to consideration of each lab) of the polarization data shows differences in
reproducibly create the same measurement conditions and hence the order of several mV. In general, Lab 2 shows lower differences
obtaining the same results for the same type of measurements at the throughout the entire current density range, than Lab 1 and Lab 3. This
specific lab. Especially quality issues with the used materials, or inac­ is also true for the HFR values. As differences decrease after iR-
curacies in cell or test bench handling would result in a poor in-house correction, they mainly originate from ohmic effects. But as differ­
reproducibility. A good in-house reproducibility is easier to obtain, ences remain also in the iR-corrected data, kinetic effects must also play
and standard deviation will be lower than for an across-sites reproduc­ a role as the origin, but to a smaller extent than the ohmic effects, see
ibility as for the across-sites reproducibility, the measurement condi­ also the more detailed analysis in the sections below. The polarization
tions in the cell must be identical at each of the different labs. This is curves in Fig. 5 show high linearity above i = 0.5 A*cm− 2 for all labs,
much harder to achieve and took an iterative process to address the non- resulting in the assumption of negligible mass transport losses (MTL).
harmonizable differences originating from different experimental setups
(test bench, electronics, laboratory infrastructure) and different working 4.1.1. Results from impedance analysis
methods to ensure the necessary control parameter were sufficiently For all labs, the impedance spectra for the different measurements
controlled at each of the labs. For simplicity reasons, in-house and described above recorded at 60 ◦ C and different current densities are
across-sites reproducibility were obtained from the same data set presented in Fig. 6. The Nyquist plots are used to give the reader an
generated by the test script shown in Fig. 3, which was run for three impression of the in-house differences and the differences across the labs
individual cell assemblies at each lab. For Lab 1 and Lab 3, one of the cell that were observed. To make the details easier to recognize, the spectra
assemblies was taken out of the comparative analysis as test bench in­ for low (a) and high current densities (b) are plotted in different figures
terruptions occurred during the measurements, prohibiting the inclu­ with different scales.
sion of this data in the analysis here. Therefore, two cell assemblies for When analyzing the spectra for high current densities (see Section
Lab 1 (labelled “MEA 1” and “MEA 2”) and Lab 3 (labelled “MEA 1” and 3.3 for details) the extracted HFRs show an in-house and across lab shift
“MEA 3” and three repetitions for Lab 2 (labelled “MEA 1”, “MEA 2” and (see Fig. 6). All labs show a decreasing HFR with increasing current
“MEA 3”) are discussed in the following. For easier reading and com­ density with impedance differences between 1 and 5 mΩ*cm2 (for the
parison, the data is colour coded throughout the entire publication. The tested current densities of 0.5 and 2.5 A*cm− 2). This is commonly
data for Lab 1 are shown in green, for Lab 2 in red and for Lab 3 in blue. attributed to a heating effect on the anodic side by the overpotential, and
a subsequent heating up of the membrane which results in a higher
4.1. In house reproducibility protonic conductivity [65]. Lab 1 and Lab 3 show the steepest slopes
(4.5–5 mΩ*cm2 respectively) and the highest deviation between MEAs
The in-house reproducibility is seen as the basis for the later (5–7 mΩ*cm2 for both labs) whereas Lab 2: shows lower slope values

7
T. Lickert et al. Applied Energy 352 (2023) 121898

Fig. 6. Impedance data at 60 ◦ C in the Nyquist representation for all three labs, showing (a) data at high current density (1–2.5 A*cm− 2) or (b) data for low current
density (0.1–0.5 A*cm− 2).

(2–4 mΩ*cm2) and lower deviation between the MEAs (1–3 mΩ*cm2). be addressed. In the high frequency area, there is a deviation from a
Although this contribution is not intended to analyze impedance spectra perfect semicircle and the graphs show a linear section, clearest visible
to a comprehensive extend, there is a deviation from the textbook in the graphs of the low current densities. This can be assigned to the
behavior for some of the measurements observed, which should shortly limited protonic conductivity in the electrode layer which can be

Fig. 7. Averaged HFR at different temperatures, differrent current densities and the different Labs as a measure for the ohmic resistances.

8
T. Lickert et al. Applied Energy 352 (2023) 121898

described by the so called “transmission line model” [35]. Or in a


simpler approach, by a second parallel arrangement of a constant phase
element (CPE) and an ohmic resistance (R), which was used in the
presented work, as the effect was not so clear and therefore assumed to
be of minor importance for comparable HFR’s. In addition, there is a
difference in the expression of the respective inductivities. Lab 2 seems
to have higher cable inductivities as spectra go further down below the
real axis towards higher positive values. These differences, and the
resulting changes, are considered by the fit of the spectra [74]. Another
noteworthy behavior can be seen at low frequencies but high current
densities (above 1 A*cm− 2). For these measurements, a second capaci­
tive element appears for MEA 2 at Lab 2 and MEA 1 & 3 at Lab 3. (seen
best in Fig. 6 left part). Lab 1 data does not show the effect. Usually, this
second capacitive element is attributed to mass transport losses if the
effect increases with current density [24]. This mass transport loss is
commonly considered to be caused by inhibition of product gas removal
– mainly on the anode side [30,38,49,73]. To summarize: Lab 1 shows
the lowest ohmic resistances, Lab 3 the highest with the values of Lab 2
in between. This holds true for both 60 ◦ C and 80 ◦ C and representative
mean values for each temperature, lab and current density are given in Fig. 9. Cell voltage differences from iR-corrected polarization data for each lab
Fig. 7. Differences in performance and the individual losses are more at 60 ◦ C and 80 ◦ C.
comprehensively visible than in the polarization curves.
the across-sites reproducibility to the same value as the in-house
4.1.2. Uncorrected and iR-corrected voltage deviations reproducibility.
In this section, the in-house spread of the measured polarization Also noticeable is the fact that Lab 1 and Lab 3 show similar and
curves for both, uncorrected and iR-corrected voltages, is analyzed. higher values than Lab 2, whose reproducibility, however, is higher,
From the Polarization curves in uncorrected and in iR-corrected state, especially at higher current densities and for the lower temperature. This
the highest differences in cell voltage ΔUmax were calculated at each means, with the herein described test setup and cell components, it is
current density point for each lab for two temperatures (60 ◦ C and possible to reach an in-house reproducibility of 5 mV and lower, which
80 ◦ C). Fig. 8 shows the deviations for the uncorrected performance are extremely precise measurements. As 95% of all data points are below
measurements and Fig. 9 the deviations for the iR-corrected values. ΔUih = 10 mV, all measurements that obtain results with significantly
Fig. 8 shows the uncorrected voltage deviation for all labs at 60 ◦ C higher differences, the implementation precision (the protocol run)
and 80 ◦ C. For 60 ◦ C, the highest deviation for all current densities is must be checked carefully.
found for Lab 3 (light blue curve in Fig. 8). For 80 ◦ C, the highest de­ In case of the iR-corrected values differences are in the same range as
viation depends on the current density and for the small current density for the uncorrected values. It is interesting to see, however, that the
as well as the 2.5 A*cm2 is found at Lab 3 (dark blue curve in Fig. 8) and deviations for Lab 2 are the highest, now, while they were the lowest for
for values between 0.5 A*cm2 and 2 A*cm2 is found at Lab 2 (dark green the uncorrected values. Values of up to ΔUih,corr = 11 mV are quantified.
curve, partially hidden in Fig. 8). Fig. 8 can be regarded as one of the 96% of all data points are below 10 mV, 91% are below 7 mV. It is
major outcomes of the presented contribution. It shows differences in noteworthy that the remaining iR-corrected deviations are similarly
uncorrected cell voltage of ΔUmax(60 ◦ C) from 0 to 12 mV and of high as the uncorrected ones. This suggests that influences leading to
ΔUmax(80 ◦ C) from 0 to 11 mV. differences, can counteract each other. Taking one influence out (in the
These values must be taken as the reference value for the across-sites actual case the ohmic losses), the remaining differences (from kinetic
comparison, as the best achievement one can hope for is to bring down losses and maybe upcoming mass transport losses) can still be as high.
Moreover, it is noticeable that differences in kinetics are lower for

Fig. 8. Cell voltage differences from uncorrected polarization data for each lab
at 60 ◦ C and 80 ◦ C. Fig. 10. iR-corrected polarization data for each lab at 60 ◦ C and 80 ◦ C.

9
T. Lickert et al. Applied Energy 352 (2023) 121898

higher temperatures. This can be seen clearly in Fig. 10, showing the iR-
corrected polarization curves for all labs at T = 60 ◦ C and 80 ◦ C.
Furthermore, Lab 1 shows the highest iR-corrected voltages, Lab 3 the
lowest and Lab 2 is in between. This order compensates the differences
in the HFRs (ohmic losses), which show values in reverse order (see
Fig. 7).

4.2. Across-sites reproducibility

The above discussed in-house reproducibility is the necessary basis


to understand data comparisons of different institutions. The main
objective of this publication is the quantification of the voltage differ­
ences that can be expected when different institutions compare data,
even if the same test protocol and the same materials were used (to the
possible extent). In the following, the differences between these in­
stitutions (called lab differences, ΔUlabs) are analyzed based on the same
polarization and impedance data. First, the differences in transient
behavior during conditioning are shown and discussed, followed by the
polarization data and the calculated lab differences in uncorrected and
iR-corrected state. Fig. 11 shows the transient behavior of all cell as­ Fig. 12. Mean polarization data out of an up-down curve pair of data for each
semblies during the last step of conditioning (potentiostatic step at U = lab at 60 ◦ C and 80 ◦ C.
1.7 V). As a MEA is only used once, the abbreviation “MEA” when
referring to measured results is used as synonym for “cell assembly” difference of ΔUlabs = 20 mV. A clear increase at T = 80 ◦ C for values
throughout this article. In general, a lab specific transient behavior can above a current density of i = 1 A*cm− 2 can be observed. The lowest in-
be observed: At Lab 1 currents decrease with time, whereas at Lab 3, house deviation is found for Lab 1, the highest deviation for Lab 3. This
current increase, both with different, lab specific, slopes. The behavior is, on the one hand, a reflection of the different ohmic resistances (see
at Lab 2 is different in such a way that the current density first drops and Fig. 7) and, on the other hand, an increasing mass transport loss for Lab 2
then significantly increases after roughly five hours and after that flat­ and Lab 3 whereas at Lab 1 this effect is not observed. This is reflected by
tens off, showing the highest changes in current densities of all Labs. A the low frequency end of the impedance spectra at high current densities
current density increase of 302 mA*cm− 2 is seen, whereas a current (see the impedance analysis section and especially Fig. 6).
density decrease of 129 mA*cm− 2 is observed for Lab 1. At the end of the The same data analysis process was done for the iR-corrected curves
conditioning procedure differences are getting smaller and curves from shown in Fig. 10. Like for the in-house differences, the values almost
all three lab converge towards a current density window of 106 spread over the same range. Voltage differences (ΔUabs,corr) between 9
mA*cm− 2, which is – especially taking the similarity of the polarization mV and 26 mV are quantified (see Fig. 14). The fact that the values are
curves reported in Section 4.1 into account – still surprisingly high. For similarly high, shows again that the losses can counteract each other. In
all institutions that want to compare their data to the herein presented general, lab differences are smaller at higher temperatures. But this
data, it is highly recommended to check if their data lies within this might change for higher current densities, which were not investigated
expectation window (grey bar), and make sure to take the Vi’s out of the herein.
second run. Otherwise, comparability to the data reported here cannot
be guaranteed.
4.3. Influence of thermal management
Fig. 12 shows mean polarization curves from up and down curves for
each lab. For quantification of the voltage differences, again, the delta
Temperature generally shows a high impact on any overpotential
between the highest and lowest voltage points for each current density
and effect in a PEM water electrolysis cell. Therefore, regulation and
was taken (from all three labs). These voltage differences are depicted in
Fig. 13. The differences increased compared to the in-house differences
(Fig. 8) by almost a factor of two. 96% of all data points are below a

Fig. 11. Transient current density signals at the potentiostatic step (last step) Fig. 13. Cell voltage differences when the uncorrected polarization data for all
during conditioning for each lab at 80 ◦ C. labs are unified into one data set for 60 ◦ C and 80 ◦ C, respectivily.

10
T. Lickert et al. Applied Energy 352 (2023) 121898

Fig. 8). This means, the changed T-variation is not increasing the voltage
differences above the expected reproducibility level originating from
other influences like reproducibility from the assembling procedure of
the cell.
By comparing two different temperature control variations at Lab 2
in Fig. 16, it was noticed, that the first runs, show significantly higher
voltage values then the other three variations. This was observed
whether the temperature control was changed (MEA 1) or left un­
changed (MEA 2).
For MEA 1, at i = 2.5 A*cm2, a significantly higher difference be­
tween the first and the second run was obtained (ΔU = 48 mV) while a
similar value was obtained for MEA 2, which shows it is not T-control
specific. The same effect (ΔU = 57 mV) was also observed at Lab 3, see
Fig. 16b. This behavior indicates that the conditioning process of the
CCM is not yet fully finished during the measurements of run 1 in Lab 2
and Lab 3. It is, however, in its extend a lab specific phenomenon, as Lab
2 and Lab 3 show a rather large change after the first curve, whereas the
change is significantly smaller for Lab 1. This suggests that even with a
precise test protocol, differences remain in and after the presented
Fig. 14. Cell voltage differences when the corrected polarization data for all conditioning period, as could already been seen in Fig. 11. This means
labs are unified into one data set for 60 ◦ C and 80 ◦ C, respectivily. on the one hand, that conditioning is not fully finished and on the other
hand, remaining differences in the entire testing process significantly
control of the cell temperature is crucial. This can be done in various manifest themselves in the conditioning signals. This is regarded as one
ways. Temperature sensors can be installed differently into fittings, with of the key learnings from the presented work, meaning that the condi­
different distances to the cell and at different places in the piping or even tioning procedure will be investigated in some more detail in the future.
inside the cell itself. To check the influences and the differences in
performance resulting from these degrees of freedom, 4 variations of
thermal control were tested according to the descriptions in chapter
2.3.1. Also, a suggestion of a definition of cell temperature for the pre­
sented test cell is given in this chapter.
In Fig. 15 two representative cell assemblies (MEA 1 and MEA 2) at
Lab 1 are compared. MEA 1 was tested using four different ways of
thermal control (temperature control variation, T-var. 01–04, see also
Section 3.2) for each run. For the second cell assembly (MEA 2), all four
runs were tested with the same variation (T-var. 01). From the graphs, it
can be seen, that the thermal management used shows minor differences
in the electrolysis performance (ΔU = 7 mV @ i = 2.5 A*cm− 2 for the
first runs). The difference for the second runs is slightly higher (ΔU = 10
mV). These values are close to, but below the differences observed when
T-variation is unchanged and defined as in-house reproducibility (see

Fig. 16. Polarization curves obtained (a) at Lab 2 for different temperature
Fig. 15. Polarization curves obtained at Lab 1 for different temperature control control variations (light red for T-var 01–04/dark red for T-var 03 for all runs)
variations. For MEA 1 (dark green), a different temperature control setting was and (b) at Lab 3 for T-var 02 for all runs (light/dark blue). Despite differences in
used for each run, whereas for MEA 2 (light green), the same temperature the temperature control between the different MEAs, the first run always shows
controls setting was used for all four runs. (For interpretation of the references a higher voltage than runs 2–4, which is highlighted in the insets. (For inter­
to colour in this figure legend, the reader is referred to the web version of pretation of the references to colour in this figure legend, the reader is referred
this article.) to the web version of this article.)

11
T. Lickert et al. Applied Energy 352 (2023) 121898

As mentioned earlier, this was the reason only the second runs (2nd Vi- the test bench ventilation, coverage of the cell as well as the three
curves) are taken into consideration for the comparison of Vi-curves at dimensional arrangements of hardware in the test bench. The observed
different sites. The authors conclude that the presented variations of spread of the temperature depends on the current density and decreases
thermal control (variations 01–04, see Section 3) do not differ much with an increasing amount of internal waste heat (which is partially
from each other in terms of performance (if the presented test cell is compensating the heat radiation of the cell). This is best observed for the
used). The calculated differences at i = 2.5 A/cm2 originating from cathode inlet temperature (red) and cathode out temperature (violet) of
different thermal control variations shows values of ≈ 3 mV for those Lab 1 and Lab 2 in Fig. 17: While the red curve remains stable at around
labs that tested the different variations. The influence might be higher 80 ◦ C, the violet curve changes for about the same amount (~1 K). As a
for other test cells and other temperature control configurations. representative cell temperature, the authors suggest to take the arith­
To give the reader and the user of the test protocol an impression of metic mean value out of the three anodic sensors signals (inlet, outlet
the quality of thermal management at the participating labs, tempera­ and internal, see Fig. 2) This value is called Tcell (depicted in black in
ture signals during the Vi-curves at 80 ◦ C for each lab with up- and Fig. 17). Even though small differences in the transient behavior of Tcell
down-curves for specific MEA’s are depicted in Fig. 17. From the figure, can be seen, the values are close to the set temperature in all cases and
differences are observable even with the same temperature control very close to the temperature measured by the internal sensor on the
(variation 02). A rather high spread between the sensors of up to 2 K anode side. For comparability reasons, this value should be within ±1 K
between inlet and outlet is observed for Lab 1 and Lab 3. whereas the around the set temperature to get to comparable data.
spread is slightly lower for Lab 2.
As current densities are the same, the generated waste heat by the 5. Summary
electrolysis process is the same at each lab. The amount of heat trans­
ported into the cell by the feed water can differ slightly as it can be seen The comparison of the experiments conducted at different in­
be the inlet temperatures in Fig. 17 (anode inlet in blue, cathode inlet in stitutions presented in this article show that a great deal of effort is
red). Also, the passive heat radiation from the cell to the surrounding required to obtain reproducible performance results across the different
can be different as each lab has a specific lab environment. It can be participating sites. The use of identical test cells and identical materials
influenced by several parameters, e.g. lab temperature in combination of is a necessary but by no means sufficient condition.
Polarization curves obtained with the test protocol are presented for
the selected reference setup of test cell and components and can be used
for reproducibility purposes. The contributing authors suggest the pre­
sented test cell, test protocol, and measurement strategy to be adopted
by the community to allow accurate comparison between laboratories.
To achieve good across-site reproducibility, accurate application of the
test protocol is crucial. Throughout more than two years of conducting
this comparative study, any test interruption and deviations from the
protocol had a massive impact on the performance and impedance data.
The thermal control in general is crucial for high reproducibility. The
specific implementation of the tested thermal management and the
tested settings, however, showed only a few millivolts of deviation. The
mean values presented in this article and listed in the SI therefore, and as
it is the most realistic case in scientific reality, include results from
different thermal control variations. In addition, the specific and
reproducible conditioning procedure of the CCM before starting the
actual performance measurements proved to be highly important. This
holds true despite unexpectedly large performance deviations
throughout the conditioning phase. Results suggest, that despite a rather
long conditioning phase (two steps at constant current, one step at
constant voltage, in total 16 h), the conditioning of the CCM was not yet
finished for some of the measurements. The conditioning procedure is a
topic that needs further investigations and adoption to increase
comparability further.
When following the measurement protocol, an across-site repro­
ducibility of <20 mV can be obtained for uncorrected as well as iR-
corrected values for current densities up to 2.5 A*cm− 2. This repro­
ducibility can potentially be higher for a higher degree of harmonization
of the test benches. In-house reproducibility of around 10 mV or better
(< 5 mV for one of the partners) represents the upper limit that could
potentially be achieved with the presented test cell and cell components.
Having ruled out many harmonizable influences, mainly non-
harmonizable test bench differences and production tolerances of the
cell components (PTL, MEA, frames, sealings) remain as main causes for
the observed performance differences.
Future work needs to include simplification of the measurement
script and a change in cell components to more industrially relevant
Fig. 17. Temperature signals (available sensors at each lab) over time for all
threes labs (Lab 1 in green on top, Lab 2 in red in the middle and Lab 3 in blue components, such as thinner titanium PTLs with protective coatings on
at the bottom). Up-curves are depicted (left) as well as down-curves (right). the anode side and carbon-based PTLs on the cathode side. Also, a
Representative cell temperature Tcell is depicted in black. (For interpretation of publicly available data base, which can be further updated with data
the references to colour in this figure legend, the reader is referred to the web from the community would be beneficial to increase future compara­
version of this article.) bility. Based on the collated data, the effects of operating conditions on

12
T. Lickert et al. Applied Energy 352 (2023) 121898

cell degradation and lifetime (AST protocol development) can and References
should be investigated ultimately.
[1] Odenweller A, Ueckerdt F, Nemet GF, Jensterle M, Luderer G. Probabilistic
feasibility space of scaling up green hydrogen supply. Nat Energy 2022;7(9):
CRediT authorship contribution statement 854–65. https://doi.org/10.1038/s41560-022-01097-4.
[2] Chatenet M, Pollet BG, Dekel DR, Dionigi F, Deseure J, Millet P, et al. Water
Thomas Lickert: Supervision, Conceptualization, Writing - original electrolysis: from textbook knowledge to the latest scientific strategies and
industrial developments. Chem Soc Rev 2022;51(11):4583–762. https://doi.org/
draft, Writing - review & editing. Stefanie Fischer: Investigation, Data 10.1039/D0CS01079K.
curation. James L. Young: Investigation, Formal analysis. Selina Klose: [3] Grigoriev SA, Fateev VN, Bessarabov DG, Millet P. Current status, research trends,
Investigation. Irene Franzetti: Investigation, Formal analysis. Daniel and challenges in water electrolysis science and technology. Int J Hydrogen Energy
2020;45(49):26036–58. https://doi.org/10.1016/j.ijhydene.2020.03.109.
Hahn: Investigation, Formal analysis. Zhenye Kang: Investigation, [4] Dolci F, Gryc K, Eynard U, Georgakaki A, Letout S, Kuokkanen A, et al. Water
Formal analysis. Meital Shviro: Supervision, Project administration. electrolysis and hydrogen in the European Union. 2022.
Fabian Scheepers: Supervision, Project administration. Marcelo [5] Gül Timur, Turk Dave, Blanco Herib, Cazzola Pierpaolo, Fukui Hiroyuki. Tae-Yoon
Kim et al. The future of hydrogen: Seizing today’s opportunities. Report prepared
Carmo: Project administration. Tom Smolinka: Project administration,
by the IEA. 2019.
Funding acquisition, Conceptualization. Guido Bender: Supervision, [6] Carmo M, Fritz DL, Mergel J, Stolten D. A comprehensive review on PEM water
Project administration, Funding acquisition, Conceptualization. Sebas­ electrolysis. Int J Hydrogen Energy 2013;38(12):4901–34. https://doi.org/
tian Metz: Supervision, Project administration, Conceptualization, 10.1016/j.ijhydene.2013.01.151.
[7] Shiva Kumar S, Himabindu V. Hydrogen production by PEM water electrolysis – a
Funding acquisition, Writing - original draft, Writing - review & editing. review. Mater Sci Energy Technol 2019;2(3):442–54. https://doi.org/10.1016/j.
mset.2019.03.002.
Declaration of Competing Interest [8] Ayers K. High efficiency PEM water electrolysis: enabled by advanced catalysts,
membranes, and processes. Curr Opin Chem Eng 2021;33:100719. https://doi.org/
10.1016/j.coche.2021.100719.
The authors declare that they have no known competing financial [9] Smolinka T, Garche J, editors. Electrochemical power sources: Fundamentals,
interests or personal relationships that could have appeared to influence systems, and applications. Elsevier; 2022.
[10] Thomassen MS, Reksten AH, Barnett AO, Khoza T, Ayers K. Chapter 6 - PEM water
the work reported in this paper. electrolysis. In: Smolinka T, Garche J, editors. Electrochemical power sources:
Fundamentals, systems, and applications. Elsevier; 2022. p. 199–228.
Data availability [11] Alia SM, Stariha S, Borup RL. Electrolyzer durability at low catalyst loading and
with dynamic operation. J Electrochem Soc 2019;166(15):F1164. https://doi.org/
10.1149/2.0231915jes.
Data will be made available on request. [12] Wirkert FJ, Roth J, Jagalski S, Neuhaus P, Rost U, Brodmann M. A modular design
approach for PEM electrolyser systems with homogeneous operation conditions
and highly efficient heat management. Int J Hydrogen Energy 2020;45(2):
Acknowledgements
1226–35. https://doi.org/10.1016/j.ijhydene.2019.03.185.
[13] Moreno Soriano R, Rojas N, Nieto E, de Guadalupe González-Huerta R, Sandoval-
The authors would like to thank Dieter Hellert and Thomas Favet Pineda JM. Influence of the gasket materials on the clamping pressure distribution
from the company Bekaert as well as Jan Byrknes and Armin Bayer from in a PEM water electrolyzer: bolt torques and operation mode in pre-conditioning.
Int J Hydrogen Energy 2021;46(51):25944–53. https://doi.org/10.1016/j.
the company Greenerity for the good collaboration and for their ijhydene.2021.03.076.
permission to publish the results obtained with their products. We would [14] Fan Z, Yu H, Jiang G, Yao D, Sun S, Chi J, et al. Low precious metal loading porous
like to thank Nikolai Utsch (Forschungszentrum Jülich) for his knowl­ transport layer coating and anode catalyst layer for proton exchange membrane
water electrolysis. Int J Hydrogen Energy 2022;47(44):18963–71. https://doi.org/
edgeable contributions during the revision of this article. Scientific 10.1016/j.ijhydene.2022.04.114.
discussions with experts from the Advanced Fuel Cells (AFC) Technology [15] Salehmin MNI, Husaini T, Goh J, Sulong AB. High-pressure PEM water electrolyser:
Collaboration Programme (TCP), Task 30 part of the Technology Plat­ a review on challenges and mitigation strategies towards green and low-cost
hydrogen production. Energ Conver Manage 2022;268:115985. https://doi.org/
form of the International Energy Agency (IEA), are gratefully acknowl­ 10.1016/j.enconman.2022.115985.
edged. The Advanced Fuel Cells Technology Collaboration Programme [16] Bender G, Carmo M, Smolinka T, Gago A, Danilovic N, Mueller M, et al. Initial
covered the publication costs. Fraunhofer ISE acknowledges the funding approaches in benchmarking and round robin testing for proton exchange
membrane water electrolyzers. Int J Hydrogen Energy 2019;44(18):9174–87.
of the Bundesministerium für Bildung und Forschung (BMBF) of the https://doi.org/10.1016/j.ijhydene.2019.02.074.
project POWER-MEE (03SF0536D), in which the used test cell was [17] Stähler M, Stähler A, Scheepers F, Carmo M, Stolten D. A completely slot die coated
originally developed and the project HyThroughGen (03HY108B), membrane electrode assembly. Int J Hydrogen Energy 2019;44(14):7053–8.
https://doi.org/10.1016/j.ijhydene.2019.02.016.
which contributed to the development of the test protocol. This work
[18] Ahmed KW, Jang MJ, Park MG, Chen Z, Fowler M. Effect of components and
was authored in part by the National Renewable Energy Laboratory, operating conditions on the performance of PEM electrolyzers: a review.
operated by Alliance for Sustainable Energy, LLC, for the U.S. Depart­ Electrochem 2022;3(4):581–612. https://doi.org/10.3390/electrochem3040040.
ment of Energy (DOE) under Contract No. DE-AC36-08GO28308. [19] Bernt M, Schramm C, Schröter J, Gebauer C, Byrknes J, Eickes C, et al. Effect of the
IrO x conductivity on the anode electrode/porous transport layer interfacial
Funding provided by the U.S. Department of Energy Office of Energy resistance in PEM water electrolyzers. J Electrochem Soc 2021;168(8):84513.
Efficiency and Renewable Energy (EERE) Hydrogen and Fuel Cell [20] Jiang G, Yu H, Li Y, Yao D, Chi J, Sun S, et al. Low-loading and highly stable
Technologies Office (HFTO), Award No. DE-EE0008836. The views membrane electrode based on an Ir@WOxNR ordered array for PEM water
electrolysis. ACS Appl Mater Interfaces 2021;13(13):15073–82. https://doi.org/
expressed in the article do not necessarily represent the views of the DOE 10.1021/acsami.0c20791.
or the U.S. Government. The U.S. Government retains and the publisher, [21] Liu C, Wippermann K, Rasinski M, Suo Y, Shviro M, Carmo M, et al. Constructing a
by accepting the article for publication, acknowledges that the U.S. multifunctional interface between membrane and porous transport layer for water
electrolyzers. ACS Appl Mater Interfaces 2021;13(14):16182–96.
Government retains a nonexclusive, paid-up, irrevocable, worldwide [22] Weber CC, Schuler T, de Bruycker R, Gubler L, Büchi FN, de Angelis S. On the role
license to publish or reproduce the published form of this work, or allow of porous transport layer thickness in polymer electrolyte water electrolysis.
others to do so, for U.S. Government purposes. J Power Sources Adv 2022;15:100095. https://doi.org/10.1016/j.
powera.2022.100095.
[23] van Pham C, Escalera-López D, Mayrhofer K, Cherevko S, Thiele S. Essentials of
Appendix A. Supplementary data high performance water electrolyzers – from catalyst layer materials to electrode
engineering. Adv Energy Mater 2021;11(44):2101998. https://doi.org/10.1002/
aenm.202101998.
Additional details of the used measurement protocol as well as the
[24] Giesbrecht PK, Freund MS. Investigation of water oxidation at IrO2 electrodes in
mean polarization data (uncorrected and iR-corrected) for all three labs nafion-based membrane electrode assemblies using impedance spectroscopy and
at 60 ◦ C and 80 ◦ C are provided online as supplementary data at https:// distribution of relaxation times analysis. J Phys Chem C 2022;126(42):17844–61.
doi.org/10.1016/j.apenergy.2023.121898. https://doi.org/10.1021/acs.jpcc.2c05104.
[25] Gollasch M, Schmeling J, Harms C, Wark M. Comparative analysis of synthesis
routes for antimony-doped tin oxide-supported iridium and iridium oxide catalysts

13
T. Lickert et al. Applied Energy 352 (2023) 121898

for OER in PEM water electrolysis. Adv Mater Interfaces 2023;10(15):2300036. component evaluation test. Int J Hydrogen Energy 2021. https://doi.org/10.1016/
https://doi.org/10.1002/admi.202300036. j.ijhydene.2021.08.213.
[26] Laube A, Hofer A, Ressel S, Chica A, Bachmann J, Struckmann T. PEM water [49] Lickert T, Kiermaier ML, Bromberger K, Ghinaiya J, Metz S, Fallisch A, et al. On the
electrolysis cells with catalyst coating by atomic layer deposition. Int J Hydrogen influence of the anodic porous transport layer on PEM electrolysis performance at
Energy 2021;46(79):38972–82. high current densities. Int J Hydrogen Energy 2020;45(11):6047–58. https://doi.
[27] Bernt M, Hartig-Weiß A, Tovini MF, El-Sayed HA, Schramm C, Schröter J, et al. org/10.1016/j.ijhydene.2019.12.204.
Current challenges in catalyst development for PEM water electrolyzers. Chem Ing [50] Young JL, Kang Z, Ganci F, Madachy S, Bender G. PEM electrolyzer
Tech 2020;92(1–2):31–9. https://doi.org/10.1002/cite.201900101. characterization with carbon-based hardware and material sets. Electrochem
[28] Bernt M, Gasteiger HA. Influence of ionomer content in IrO 2 /TiO 2 electrodes on Commun 2021;124:106941. https://doi.org/10.1016/j.elecom.2021.106941.
PEM water electrolyzer performance. J Electrochem Soc 2016;163(11):F3179–89. [51] Wang W, Li K, Ding L, Yu S, Xie Z, Cullen DA, et al. Exploring the impacts of
https://doi.org/10.1149/2.0231611jes. conditioning on proton exchange membrane electrolyzers by in situ visualization
[29] Yang G, Yu S, Mo J, Li Y, Kang Z, Bender G, et al. Impacts of catalyst nanolayers on and electrochemistry characterization. ACS Appl Mater Interfaces 2022;14(7):
water permeation and swelling of polymer electrolyte membranes. J Power Sources 9002–12. https://doi.org/10.1021/acsami.1c21849.
2020;448:227582. https://doi.org/10.1016/j.jpowsour.2019.227582. [52] Tomić AZ, Pivac I, Barbir F. A review of testing procedures for proton exchange
[30] Dedigama I, Angeli P, Ayers K, Robinson JB, Shearing PR, Tsaoulidis D, et al. In situ membrane electrolyzer degradation. J Power Sources 2023;557:232569. https://
diagnostic techniques for characterisation of polymer electrolyte membrane water doi.org/10.1016/j.jpowsour.2022.232569.
electrolysers – flow visualisation and electrochemical impedance spectroscopy. Int [53] Kuhnert E, Hacker V, Bodner M, Subramanian P. A review of accelerated stress
J Hydrogen Energy 2014;39(9):4468–82. https://doi.org/10.1016/j. tests for enhancing MEA durability in PEM water electrolysis cells. Int J Energy Res
ijhydene.2014.01.026. 2023;2023:3183108. https://doi.org/10.1155/2023/3183108.
[31] Yu H, Danilovic N, Wang Y, Willis W, Poozhikunnath A, Bonville L, et al. Nano-size [54] G. Tsotridis AP. EU harmonised protocols for testing of low temperature water
IrOx catalyst of high activity and stability in PEM water electrolyzer with ultra-low electrolysers. 2021. https://doi.org/10.2760/58880.
iridium loading. Appl Catal Environ 2018;239:133–46. https://doi.org/10.1016/j. [55] Frensch SH, Fouda-Onana F, Serre G, Thoby D, Araya SS, Kær SK. Influence of the
apcatb.2018.07.064. operation mode on PEM water electrolysis degradation. Int J Hydrogen Energy
[32] Liu C, Shviro M, Gago AS, Zaccarine SF, Bender G, Gazdzicki P, et al. Exploring the 2019;44(57):29889–98. https://doi.org/10.1016/j.ijhydene.2019.09.169.
interface of skin-layered titanium fibers for electrochemical water splitting. Adv [56] Kang Z, Schuler T, Chen Y, Wang M, Zhang F-Y, Bender G. Effects of interfacial
Energy Mater 2021;11(8):2002926. https://doi.org/10.1002/aenm.202002926. contact under different operating conditions in proton exchange membrane water
[33] Liu L, Chen W, Li Y. An overview of the proton conductivity of nafion membranes electrolysis. Electrochim Acta 2022;429:140942. https://doi.org/10.1016/j.
through a statistical analysis. J Membr Sci 2016;504:1–9. https://doi.org/ electacta.2022.140942.
10.1016/j.memsci.2015.12.065. [57] Malkow T, Pilenga A, Tsotridis G, De Marco G. EU harmonised polarisation curve
[34] Liu C, Carmo M, Bender G, Everwand A, Lickert T, Young JL, et al. Performance test method for low-temperature water electrolysis. Luxembourg: Publications
enhancement of PEM electrolyzers through iridium-coated titanium porous Office of the European Union; 2018.
transport layers. Electrochem Commun 2018;97:96–9. https://doi.org/10.1016/j. [58] Malkow KT, Pilenga A, Blagoeva D. EU harmonised terminology for hydrogen
elecom.2018.10.021. generated by electrolysis: An open and comprehensive compendium. 2021.
[35] Lettenmeier P, Kolb S, Sata N, Fallisch A, Zielke L, Thiele S, et al. Comprehensive https://doi.org/10.2760/732809.
investigation of novel pore-graded gas diffusion layers for high-performance and [59] Siegmund D, Metz S, Peinecke V, Warner TE, Cremers C, Grevé A, et al. Crossing
cost-effective proton exchange membrane electrolyzers. Energ Environ Sci 2017;10 the valley of death: from fundamental to applied research in electrolysis. JACS Au
(12):2521–33. https://doi.org/10.1039/C7EE01240C. 2021;1(5):527–35. https://doi.org/10.1021/jacsau.1c00092.
[36] Georg Andreas, Lickert Thomas, Smolinka Tom, Zhang Xiaoling, Cooke Kevin. [60] Zhang L, Zhao H, Wilkinson DP, Sun X, Zhang J. Electrochemical water
Corrosion protective coatings for bipolar plates and current collectors in PEM electrolysis: Fundamentals and technologies. CRC Press; 2020.
electrolysers: Novel materials and system designs for low cost, efficient and durable [61] Petrovic S, Kurzweil P, Garche J. Electrochemical energy storage: Batteries, fuel
PEM electrolysers. In: Second international workshop on durability and cells, and hydrogen technologies. McGraw-Hill Education; 2022.
degradation issues on; 2023. [62] Metz S, Smolinka T, Bernäcker CI, Loos S, Rauscher T, Röntzsch L, et al. Producing
[37] Bautkinova T, Utsch N, Bystron T, Lhotka M, Kohoutkova M, Shviro M, et al. hydrogen through electrolysis and other processes. In: Neugebauer R, editor.
Introducing titanium hydride on porous transport layer for more energy efficient Hydrogen Technologies. Cham: Springer International Publishing; 2023.
water electrolysis with proton exchange membrane. J Power Sources 2023;565: p. 203–52.
232913. https://doi.org/10.1016/j.jpowsour.2023.232913. [63] Scott K. Electrochemical methods for hydrogen production. Royal Society of
[38] Rakousky C, Keeley GP, Wippermann K, Carmo M, Stolten D. The stability Chemistry; 2019.
challenge on the pathway to high-current-density polymer electrolyte membrane [64] Millet P, Mbemba N, Grigoriev SA, Fateev VN, Aukauloo A, Etiévant C.
water electrolyzers. Electrochim Acta 2018;278:324–31. https://doi.org/10.1016/ Electrochemical performances of PEM water electrolysis cells and perspectives. Int
j.electacta.2018.04.154. J Hydrogen Energy 2011;36(6):4134–42. https://doi.org/10.1016/j.
[39] Möckl M, Ernst MF, Kornherr M, Allebrod F, Bernt M, Byrknes J, et al. Durability ijhydene.2010.06.105.
testing of low-iridium PEM water electrolysis membrane electrode assemblies. [65] Ito H, Maeda T, Nakano A, Takenaka H. Properties of nafion membranes under
J Electrochem Soc 2022;169(6):64505. https://doi.org/10.1149/1945-7111/ PEM water electrolysis conditions. Int J Hydrogen Energy 2011;36(17):10527–40.
ac6d14. https://doi.org/10.1016/j.ijhydene.2011.05.127.
[40] Torrero J, Morawietz T, García Sanchez D, Galyamin D, Retuerto M, Martin- [66] Fuel cell technologies office multi-year research, development, and demonstration
Diaconescu V, et al. High performance and durable anode with 10-fold reduction of plan: Section 3.4 fuel cells. 2016.
iridium loading for proton exchange membrane water electrolysis. Adv Energy [67] Garland N, Benjamin T, Kopasz J. DOE fuel cell program: durability technical
Mater 2023;13(23):2204169. https://doi.org/10.1002/aenm.202204169. targets and testing protocols. ECS Trans 2007;11(1):923. https://doi.org/10.1149/
[41] Lee JK, Schuler T, Bender G, Sabharwal M, Peng X, Weber AZ, et al. Interfacial 1.2781004.
engineering via laser ablation for high-performing PEM water electrolysis. Appl [68] Shi Y, Lu Z, Guo L, Yan C. Fabrication of membrane electrode assemblies by direct
Energy 2023;336:120853. https://doi.org/10.1016/j.apenergy.2023.120853. spray catalyst on water swollen Nafion membrane for PEM water electrolysis. Int J
[42] Ayers K, Danilovic N, Harrison K, Xu H. PEM electrolysis, a forerunner for clean Hydrogen Energy 2017;42(42):26183–91. https://doi.org/10.1016/j.
hydrogen. Electrochem Soc Interface 2021;30(4):67. https://doi.org/10.1149/2. ijhydene.2017.08.205.
F16214IF. [69] Garcia-Navarro JC, Schulze M, Friedrich KA. Measuring and modeling mass
[43] Martin A, Trinke P, Stähler M, Stähler A, Scheepers F, Bensmann B, et al. The effect transport losses in proton exchange membrane water electrolyzers using
of cell compression and cathode pressure on hydrogen crossover in PEM water electrochemical impedance spectroscopy. J Power Sources 2019;431:189–204.
electrolysis. J Electrochem Soc 2022;169(1):14502. https://doi.org/10.1149/ https://doi.org/10.1016/j.jpowsour.2019.05.027.
1945-7111/ac4459. [70] Siracusano S, Trocino S, Briguglio N, Baglio V, Aricò AS. Electrochemical
[44] Lohoff AS, Poggemann L, Carmo M, Müller M, Stolten D. Enabling high throughput impedance spectroscopy as a diagnostic tool in polymer electrolyte membrane
screening of polymer electrolyte membrane (PEM) water electrolysis components electrolysis. Materials 2018;11(8). https://doi.org/10.3390/ma11081368.
via miniature test cells. J Electrochem Soc 2016;163(11):F3153–7. https://doi.org/ [71] Suermann M, Schmidt TJ, Büchi FN. Cell performance determining parameters in
10.1149/2.0211611jes. high pressure water electrolysis. Electrochim Acta 2016;211:989–97. https://doi.
[45] Rozain C, Millet P. Electrochemical characterization of polymer electrolyte org/10.1016/j.electacta.2016.06.120.
membrane water electrolysis cells. Electrochim Acta 2014;131:160–7. https://doi. [72] Yuan X-Z, Song C, Wang H, Zhang J. Electrochemical impedance spectroscopy in
org/10.1016/j.electacta.2014.01.099. PEM fuel cells. London: Springer London; 2010.
[46] Stiber S, Balzer H, Wierhake A, Wirkert FJ, Roth J, Rost U, et al. Porous transport [73] Lee CH, Banerjee R, Arbabi F, Hinebaugh J, Bazylak A. Porous transport layer
layers for proton exchange membrane electrolysis under extreme conditions of related mass transport losses in polymer electrolyte membrane electrolysis: A
current density, temperature, and pressure. Adv Energy Mater 2021;11(33): review. In: ASME 2016 14th International conference on nanochannels,
2100630. https://doi.org/10.1002/aenm.202100630. microchannels, and minichannels collocated with the ASME 2016 heat transfer
[47] Majasan JO, Cho JI, Dedigama I, Tsaoulidis D, Shearing P, Brett DJ. Two-phase Summer conference and the ASME 2016 fluids engineering division Summer
flow behaviour and performance of polymer electrolyte membrane electrolysers: meeting; Sunday; 2016. V001T07A003.
electrochemical and optical characterisation. Int J Hydrogen Energy 2018;43(33): [74] Franzetti Irene, Chan Ai-Lin, Pushkarev Artem, Metz Sebastian. External effects on
15659–72. https://doi.org/10.1016/j.ijhydene.2018.07.003. the high frequency EIS response of a PEM electrolysis cell. Istanbul Congress
[48] Nagasawa K, Ishida T, Kashiwagi H, Sano Y, Mitsushima S. Design and Center, Istanbul, Turkey. In: Proceedings of WHEC-2022, 23rd world hydrogen
characterization of compact proton exchange membrane water electrolyzer for energy conference; 2022. #391.

14

You might also like