Membrane Interlayer with Pt Recombination Particles for Reduction of the Anodic Hydrogen Content in PEM Water Electrolysis

1Laboratory for MEMS Applications, IMTEK Department of Microsystems Engineering, University of Freiburg, 79110 Freiburg, Germany 2Institute of Electric Power Systems, Leibniz Universität Hannover, 30167 Hanover, Germany 3Hahn-Schickard, 79110 Freiburg, Germany 4Helmholtz-Institute Erlangen-Nürnberg for Renewable Energy (IEK-11), Forschungszentrum Jülich GmbH, 91058 Erlangen, Germany 5Department of Chemical and Biological Engineering, Friedrich-Alexander-Universität Erlangen-Nürnberg, 91058 Erlangen, Germany

which is very low in comparison to data without any mitigation strategies that show a H 2 in O 2 content of 2.5 vol.% for only 20 bar cathode pressure at 0.5 A cm −2 . 12 With Pt in the anode catalyst layer and a platinized PTL Ito et al. 14 showed a low H 2 in O 2 content of 1 vol.% at 0.3 A cm −2 and 10 bar differential pressure for a N115 based MEA, whereas without mitigation strategies the H 2 in O 2 content is around 2.25 vol.% at identical operation conditions despite a thicker membrane (N117). 12 The Pt within the anode catalyst layer might not only force a recombination reaction of H 2 and O 2 , it is also possible that the permeated H 2 gas is oxidized to protons and will migrate back to the cathode (hydrogen pump). This was shown by Schalenbach et al., 15 using a 3rd electrode within the membrane. This also improves the faradaic efficiency. However, the use of further electrodes within a PEMWE makes the system more complex.
If the recombination catalyst is placed within the anodic CL of an electrolysis cell the Pt is exposed to high potentials > 1.23 V, thus also the long-term stability of recombination catalysts regarding dissolution or passivation due to oxide formation could be a problem.
To reduce the H 2 in O 2 content while avoiding the exposure of Pt to high potentials, the Pt nanoparticles can be incorporated directly into the membrane. This concept has already been successfully applied for polymer electrolyte membrane fuel cells e.g. [16][17][18][19][20][21][22][23] In fuel cells the recombination leads to self-humidification of the membrane and therefore good performance at dry operation conditions, which is an indirect proof of the working principle of the recombination of permeating H 2 and O 2 to water.
Bessarabov and Kruger 24 implemented this approach within freestanding, electrodeless membranes and the group of Gubler 25 in a PEMWE with a reinforced PFSA membrane, both showing successfully a reduction of the H 2 in O 2 content within the anode compartment. However, this was shown in oral presentations without further investigations and detailed descriptions.
Therefore, in this work the use of a recombination catalyst in form of Pt nanoparticles directly introduced into the PEM via spray coating is proposed. The effect of Pt interlayers on the H 2 in O 2 content and cell performance in comparison to reference materials was investigated. Additionally, long-term tests were performed in which a constant current of 1 A cm −2 was held for a duration of 10 days, to investigate the stability of the Pt interlayer.

Materials and Methods
Membrane electrode assembly.-Single-sided catalyst coated membranes (CCMs) from HIAT gGmbH were used as base material for the electrolysis cells with an active area of 25 cm 2 . On the anode side a Nafion N115 membrane with 2 mg cm −2 iridium black (Ir) as catalyst layer (CL) and on the cathode side a Nafion NR-212 membrane with 1 mg cm −2 Pt on advanced carbon as CL were used. As shown schematically in Fig. 1, the interlayers were fabricated by spray coating (Fig. 1 (1)) onto the membrane side of one of the single-sided CCMs and sandwiched between both membrane sides ( Fig. 1 (2)).
In order to prepare the interlayers, the Pt nanoparticles (Sigma-Aldrich, particle size < 50 nm), Nafion dispersion (D2020, Ion-power) and isopropanol were mixed in a weight ratio of 0.00114/1.43/10. A spray coater from Sono-Tek with a line spacing of 1.5 mm and a syringe pump rate of 0.1 mL min −1 was used to apply the interlayers in 45 consecutive spraying cycles with a transversal speed of 30 mm s −1 . The spray coater was equipped with a 48 kHz ultrasonic nozzle. To determine the Pt loading, the MEAs were weighed before and after spray coating and the Pt content was determined by the weight to weight ratio of Pt and Nafion in the dispersion. Two MEAs with Pt interlayers (Int 1 and Int 2) were made to test the interlayers with respect to repeatability and reproducibility.
As reference a conventional N117 membrane (Ref. 1) and the same half-sided-CCMs (N115 + NR-212) as for the Int 1 and Int 2 membrane electrode assembly (MEA) without depositing an interlayer (Ref. 2) with equal catalyst layers were used.
Physical characterization.-After testing, the Int 1 MEA was embedded in Araldite 502 epoxy resin (54.8 wt% Araldite 502, 42.7 wt% Dodecenylsuccinic anhydride, 2.5 wt% Dimethylbenzylamine) and cured for 12 hours at 60 • C. The sample was prepared for imaging by wet ultra-thin sectioning using a RMC Boeckeler PowerTome ultramicrotome equipped with a Diatome ultra 45 • diamond knife. Subsequently, the remaining blockface of the sample was imaged with a Scanning Electron Microscope (SEM, FEI 250 Quanta FEG with vCD back scatter detector) in "Low Vac" measurement mode at 111 Pa water pressure. ImageJ with the plug-in local thickness was used to determine the interlayer thickness.
For comparison with a pristine Pt interlayer the Pt nanoparticles/Nafion dispersion was sprayed onto a Nafion N115 membrane during the same spraying process as for the MEA preparation and prepared and imaged with the same procedure as described above.
Electrochemical setup.-Test station.-Electrochemical measurements were carried out at a Greenlight E100 test station, which controlled the anodic DI water feed to 50 g min −1 and the feed temperature to 80 • C. Both half cells were operated at ambient pressure conditions and the product gases were cooled down and separated from the DI water.
Cell.-The test station was equipped with a liquid cooled baltic quick connect system. The cell insert had an active area of 25 cm 2 and gold coated parallel flow fields. The cell temperature was kept at 80 • C by a circulation thermostat. The baltic frame was applied with a nitrogen pressure of 4.5 bar, which led to a calculated clamping pressure of 1.4 MPa on the active area.
Sintered titanium fibers with diameters of 20 μm, porosity of 0.56 and thickness of 1 mm (2GDL40-1,00, Bekaert) and a Toray carbon paper with a porosity of 0.78 and a thickness of 280 μm (TGP-H-090, Fuel cell store) were used as porous transport layers on the anode side and on the cathode side, respectively. Power supply and impedance spectroscopy.-The polarization curves, electrochemical impedance spectroscopy (EIS) and high frequency resistance (HFR) were measured by a ModuLab XM ECS electrochemical test system (Solartron analytical). The ModuLab was equipped with a XM FRA card, a XM PSTAT card and an external 100 A booster.
For the long-term experiments and the H 2 in O 2 measurements the current was supplied by an Ametek Sorensen XG 6-220 power supply (current accuracy ± 0.2% of reading). In the following the single measurement techniques are described in detail.

Test sequences.-Each
Polarization curve + HFR.-The polarization curves were measured galvanostatically with logarithmic steps from 0.001 to 1 A cm −2 and then in steps of 0.2 A cm −2 to 2 A cm −2 . Each step was held for 10 s with a sample rate of 5 Hz.
HFR measurements were performed between each current step. Therefore, short galvanostatic EIS measurements were implemented with frequencies from 50 to 0.1 kHz. Root mean square (RMS) of the sine waves were set to 10% of the applied DC current. The HFR was determined as the impedance values without imaginary parts multiplied by the active area.
Electrochemical impedance spectroscopy.-Galvanostatic EIS measurements were conducted at 11 different DC current densities from 0.05 to 2 A cm −2 . The frequencies of the sinusoidal signal were changed from 100 kHz to 0.1 Hz with RMS of 10% of the applied DC current.
Hydrogen in oxygen content.-The H 2 in O 2 content before and after the long-term experiment was measured every three minutes with the micro-GC for 9 different current densities from 0.1 to 2 A cm −2 . Each current density was held for several hours until the H 2 content reached a constant value (steady state). Because of the small gas production rates at low current densities these steps required more time to reach a constant value than at higher current densities. This measurement sequence required approximately one day. During the long-term experiments the H 2 gas content was measured every 5 minutes.

Results and Discussion
Pt interlayer.-The final loading of solid material (Nafion and Pt) after spraying amounted to 4.67 mg cm −2 which corresponds to a Pt loading of 0.02 mg cm −2 and an additional interlayer thickness of about (26 ± 1) μm as can be seen in the cross-section of the Int 1 MEA shown in Fig. 2. The Pt interlayer was spray-coated onto the cathode half-sided CCM (Pt/C + NR-212) and sandwiched with the anode half-sided CCM (Ir + N115). No noticeable differences were observed in SEM images of cross-sections of pristine and used MEA. This can be caused by the membrane-membrane interface developing some cavities, which influence the gas permeation through the membrane or might arise from membrane defects occurring during cell assembly, since handling two half-CCMs is more likely to cause pinholes or electrical shorts.

Reduction of H
When introducing a Pt interlayer in between the N115 and NR-212 membranes (Int 1 and Int 2), the H 2 in O 2 content decreases significantly for all current densities. For current densities >0.5 A cm −2 it is below the detection limit of 0.005 vol.%. The increase of the membrane thickness by (26 ± 1) μm due to the recombination interlayer is not sufficient to explain the strong reduction of the H 2 in O 2 content. Theoretically, the H 2 in O 2 content for a (26 ± 1) μm thicker membrane should be reduced by a factor of 180 μm/206 μm. Consequently, the H 2 in O 2 reduction is mainly caused by the recombination of permeating H 2 and O 2 to water within the interlayer. Thus enough   26 at a temperature of 80 • C and assuming a water content of 22 water molecules per sulfonic acid group 27 this yields an increase in the area normed membrane resistance of (13 ± 1) m cm 2 . For Int 2 this is in the range of the measured increase of 14 m cm 2 . However, for Int 1 the measured increase is much higher (33 m cm 2 ).
Watanabe et. al 23 and Shichun et al. 17 came to the result that introducing Pt nanoparticles into their Nafion membranes in fuel cells with comparable or even higher loading does not affect the HFR. Therefore, some additional lamination issues or also the nanoparticle size and distribution across the membrane play an important role and should be subject of optimization.
As can be seen from the similar trend of the HFR curve versus current density of the Ref. 2 MEA and the HFR curves of the MEAs with Pt interlayer (Int 1 and Int 2), the increase in HFR with current density can be related to sandwiching of the MEAs. Consequently, this increasing HFR with current density might be solved by an improved setup, e.g. higher clamping pressures, hot pressing of the MEAs with interlayer or different fabrication processes. In comparison to the sandwiched MEAs, the Ref. 1 MEA shows an almost constant HFR.
The two insets in Fig. 3c show impedance spectra at a current density of 0.5 A cm −2 (left) and 1 A cm −2 (right). While the HFRs differ, the kinetic semicircles are comparable for all MEAs, especially for small current densities. Thus the performance of the electrodes is not affected by the Pt interlayers.
Long-term performance. -Fig. 4a (Int 2). This clearly shows the beneficial effect of introducing a recombination catalyst into the membrane.
However, the increase of H 2 in O 2 by time, observed for the Int 1 and Int 2 shows that with time not all permeating H 2 molecules are recombined with O 2 molecules to water. A reason for this might be that the activity of the recombination catalyst decrease with time, due to e.g. dissolution, washing out or agglomeration. However, all this is unlikely for the duration of 245 h, and is also supported by Fig. 2, which shows Int 1 after testing. Another possibility is that the H 2 crossover increases in a higher extent than O 2 does (as it can be seen for the reference MEAs), so that at the position of the interlayer (closer to cathode than anode) there is not enough O 2 present to enable a full recombination of all permeated H 2 .
In Fig. 4b the cell voltage is depicted for the same duration of 245 hours. After the first strong increase in cell voltage, which is characteristic for such long-term experiments (≈ 10 hours), 28  Particularly noticeable in Fig. 5a is that the H 2 in O 2 content is higher for the EOT measurements at higher current densities. For the lower current densities the H 2 content is nearly the same as before. For this reason, thinning or increases in permeability of the membrane cannot explain this behavior, since then the H 2 content would increase for all current densities.
The increase in H 2 in O 2 content for the Ref. 1 MEA at EOT and current densities higher than 0.5 A cm −2 can only be explained by a disproportionate, strong increase of H 2 crossover at higher current densities. It is assumed that due to the degradation during the longterm experiments the mass transfer within the cathode was affected. Higher mass transfer limitations can lead to higher dissolved H 2 concentration (supersaturation), 11 which leads to a higher H 2 crossover. This assumption is supported by the large semicircles at low frequencies of the impedance spectra that evolve especially at higher current densities, e.g. at 1 A cm −2 for the EOT measurements (Fig. 5c), right inset). These low frequencies impedance data are often related to mass transport phenomena e.g. 32,33 The disproportionate, strong increase of H 2 crossover can also be seen for the EOT measurement of the Int 1 MEA. At higher current densities (> 1 A cm −2 ) small H 2 in O 2 contents can be measured, whereas at low current densities still no H 2 reaches the anode. Thus, at low current densities enough O 2 species for the recombination reaction exist at the interlayer and for higher current densities the H 2 crossover increases in higher extent as the O 2 crossover. Consequently, not all of the H 2 crossover can be recombined because of missing O 2 . Fig. 5b shows the polarization curves before and after the longterm runs for the Ref. 1 and Int 1 MEAs. The performance of the Ref. 1 MEA is drastically reduced. Particularly, this is caused by an increased HFR (Fig. 5c). In addition to the increase in HFR, the remaining losses e.g. kinetic and transport losses also increase, which can be seen by the larger semicircles of the Nyquist plots (insets of Fig. 5c) at 0.5 and 1 A cm −2 . Furthermore, the polarization curves of the Int 1 MEA are close together. The EOT curve is for current densities > 1 A cm −2 even slightly better than the initial BOT curve and is mainly a result of the decrease in HFR (s. Fig. 5c). This might be explained by a reduction of the contact resistance between interlayer and membrane due to laminating. However, very similar to the Ref. 1 MEA the remaining losses (kinetic, transport) are also increased after the long-term experiment.

Conclusions
In this work the effect of integrating Pt nanoparticles as a recombination catalyst between a NR-212 and a N115 membrane of a PEM water electrolysis cell was shown. Already a Pt loading of 0.02 mg cm −2 mitigates the H 2 content of the anode and thus this strategy proved to overcome safety issues. Since at the beginning of the test no H 2 could be detected within the anode, the loading of platinum could even be reduced.
When holding the MEAs at a constant current of 1 A cm −2 , no H 2 was detected for over 70 hours. After this time only a small amount of H 2 appeared, much lower than of the reference MEAs without recombination interlayer. The cell degradation by cell voltage during the constant current hold was not noticeably affected by the Pt interlayers.
Characterization of the MEAs after 245 hours of constant current hold revealed a strong increase for the H 2 in O 2 content in the high current density region (> 1 A cm −2 ). MEAs in which this feature was pronounced also showed a large second semicircle in the impedance spectra at low frequencies that suggest mass transport limitations. However, the reason of this degradation is not fully understood so far and needs further investigation.
The fact that at current densities below 1 A cm −2 MEAs with Pt interlayer show no H 2 in O 2 even after 245 hours of constant current hold proves that the Pt was still active and worked as recombination catalyst. However, at higher current densities it is assumed that not enough O 2 was present to recombine all permeating H 2 . This disproportionately can be explained by the stronger increase of H 2 crossover through the NR-212 membrane attached to the cathode, in comparison to the O 2 crossover through the N115 membrane, on the anodic side of the Pt interlayer. Thus better recombination could be reached by optimizing the position of the interlayer or a better distribution of the Pt particles.
Additionally, depending on the H 2 and O 2 permeation rates the ideal position of the interlayer varies. For instance, operation at differential pressure shifts the optimal position of the interlayer toward the lower pressure side. A theoretical examination of the optimal position is provided in the Appendix.
The interlayer not only favors recombination of H 2 and O 2 inside the membrane but also reduces O 2 which permeates into the cathode. If Pt is used as the cathodic catalyst, a recombination catalyst within the membrane can increase the faradaic efficiency by reducing the amount of permeating O 2 which can recombine with H 2 within the cathode. Additionally, also related degradation mechanisms might be reduced.
If another cathode catalyst material is used, e.g. a cheap platinum group metals free catalyst, which is not active for the recombination of permeated oxygen with hydrogen within the cathode, than the produced hydrogen can be contaminated with O 2 in H 2 . 34 Consequently, a recombination interlayer can increase the purity of the produced H 2 and also reduce the safety issues related to O 2 within H 2 .
One disadvantage of the interlayer of this work is the increase of ohmic losses (HFR). However, as seen on the example of fuel cells where despite Pt additives no increase in HFR was detected, 17,23 this could be solved by improving the interlayer morphology and the membrane-membrane interface, e.g. by hot pressing prior to assembly or increasing the clamping pressure. Furthermore, a Pt recombination layer can be used to reduce HFR losses by employing thinner membranes without increasing the H 2 in O 2 content.

Appendix: Theoretical Examination of Ideal Position of Recombination Interlayer
In the presented contribution the H 2 in O 2 content could be significantly reduced by implementing Pt particles as a recombination interlayer (Int) inside the polymer electrolyte membrane (PEM). At the Pt particles H 2 and O 2 recombine to H 2 O. Here we present a short theoretical examination where inside a PEM this interlayer should be placed in order to recombine all of the H 2 and O 2 .
The calculations for this are based on the following assumptions: (i) the activity of the interlayer regarding the recombination of H 2 and O 2 to H 2 O is very high. Consequently, all gases are recombined directly at the recombination interlayer. Hence the dissolved gas concentration at the interlayer is close to zero, (ii) the thickness of the recombination interlayer is negligible compared to the membrane thickness δ Int δ mem and (iii) the diffusion of H 2 and O 2 can be described with Ficks first law.
The dimensionless position of the interlayer is given by ξ and is defined by Eq. A1 where x is the position of the interlayer and δ mem the membrane thickness. The through plane direction is toward the anode, consequently at ξ = 0 is the cathode and at ξ = 1 is the anode side of the PEM. The diffusion path length, which H 2 has to permeate from the cathode side until reaching the interlayer therefore is given by ξδ mem . The diffusion path length O 2 has to permeate from the anode side toward the interlayer is given by (1 − ξ) δ mem . Including these diffusion path lengths and the previously stated assumptions into Ficks first law of diffusion, we get the following Eqs. [A4] Figure A1 shows the results of the derived Eq. A4 as ideal dimensionless position of the recombination interlayer ξ plotted versus the cathode pressure (up to 30 bar) for different anode pressures. It can be seen, that for balanced pressure operation (green line) the ideal position of the interlayer is at 0.5 · δ mem distance from the cathode side. This is due to the ratio of 2 of the hydrogen and the oxygen permeability. The position changes for different ratios, e.g. for a ratio of 2.3 as used by Schalenbach et al. 3  for balanced pressure would be at 0.54 · δ mem distance from the cathode side. In any case at differential pressure operation with p c > p a , the interlayer should be positioned closer to the anode side (gray area). In the more unusual situation, when the cathode pressure is lower than the anode pressure p c < p a , the position of the recombination interlayer should be chosen closer to the cathode side.