A coupled modeling-experimental approach for the in-situ estimation of effective water transport parameters in AEMFC

EFCF 2025: Low-Temp. Fuel Cells, Electrolysers & H2 Processing 1 – 4 July 2025, Lucerne Switzerland https://doi.org/10.5281/zenodo.17476267 B0708 Page 1/10 B0708 A coupled modeling-experimental approach for the insitu estimation of effective water transport parameters in AEMFC Valentin Leduc* (1), Giuseppe Sdanghi (1), Josua Wimmer (2), Michael Eppler (2), Matthias Hanauer (2), Ulrich Sauter (2), Tobias Schmitt (2), Jérôme Dillet (1), Feina Xu (1), Sophie Didierjean (1), Gael Maranzana (1) (1) LEMTA, Université de Lorraine, CNRS, 54500 Vandœuvre-lès-Nancy/France; (2) Robert Bosch GmbH, Corporate Research, Robert-Bosch-Campus 1, DE 71272, Renningen/Germany; *Contact corresponding authors: www.EFCF.com/ContactRequest Abstract A detailed modelling of transfer phenomena in AEMFC is a highly challenging task. To do that, it is important to consider: i) the multi-dimensional nature of transfers, ii) the stochastic two-phase flow in porous media as well as in the flow channels, iii) the properties of the interfaces between the different layers which are not homogenous and difficult to characterize, and the coupling between heat, mass and charge transfers. In this context, we propose a coupled modeling-experimental approach consisting in the development of a 1D model whose effective transport parameters are estimated by original in situ experiments. The advantage of such an approach is that it takes into account the real geometry including the contact resistances, the multidimensional nature of both mass and heat transfer and the presence of liquid water. The drawback is that the estimated effective parameters lose some of their intrinsic character, but in the end the model is simple enough to be implemented in real time. Original experiments have been developed to decorrelate the effective diffusion coefficients in the gas phase, adsorbed phase, heat transport properties as well as the electro-osmotic drag coefficient. Simple water balances are used to estimate the parameters, based on the measurement of water flows into and out of the cell. Both the numerical model and the original experimental setup (Figure 1) built for this investigation will be presented, as well as the inverse method used for the estimation of the effective transfer parameters. The method presented in this work leads to simple models that can be used to control systems, as well as to understand the physics of coupled transfer phenomena. Figure 1 : Schematic of the experimental bench EFCF 2025: Low-Temp. Fuel Cells, Electrolysers & H2 Processing 1 – 4 July 2025, Lucerne Switzerland https://doi.org/10.5281/zenodo.17476267 B0708 Page 2/10 Introduction Water management in Anion Exchange Membrane Fuel Cells (AEMFCs) is more challenging than in Proton Exchange Membrane Fuel Cells (PEMFCs). On one hand, the cathode side may be prone to dehydration. Indeed, the oxygen reduction reaction consumes water and the electro-osmotic drag makes water pass through the membrane toward the anode side. On the other hand, the anode side may be prone to flooding. Water is produced because of the hydrogen oxidation reaction, and in addition, an increase of the relative humidity must be taken into account here due to the electro-osmotic drag, as mentioned above. Since the water concentration at the anode side is generally higher than at the cathode side, a water concentration gradient triggers a water flux from the anode to the cathode side due to diffusion. Therefore, it is essential to develop a comprehensive understanding of the water transport phenomena in AEMFCs, along with a detailed knowledge of the membrane water transport properties. In addition, it appears that enhancing the water diffusion coefficient leads to an improvement in the AEM's lifetime [1]. Several techniques exist to measure AEM water transport properties. These are based on the same techniques commonly used for PEMFCs. The most common techniques are dynamic vapor sorption (DVS), NMR, quasi elastic neutron scattering, liquid-vapor (L-V), vapor-vapor (V-V) or liquid-liquid (L-L) permeation. Wei et al. [2] used a L-V permeation set up. They found that the limiting mode of water transport was the interfacial resistance for Nafion (PEM) and Fumapem (AEM) membranes. However, the Fumapem membrane was in 𝐶𝑂3−2 form and the transport through the GDL was neglected. Luo et al. [3] used a L-V and L-L set up. They studied Aemion and Fumapem membranes at 70°C in 𝐶𝑂3−2 form, the vapor side was equilibrated at 40% of relative humidity (RH). With ex-situ measurement without GDL at different thicknesses, they decorrelate the internal and interfacial resistances. Eriksson et al. [4] were interested in the membrane Tokuyama A201 in 𝑂𝐻− form. They performed L-V, V-V experiments as well as measurements under relevant operating conditions. They estimated water transport parameters which were independent of the membrane humidification. The interfacial resistances were not considered in their study. Petrovick et al. [5] measured the water transport parameters of a Versogen and a Sustainion (AEM) and Nafion (PEM) membranes. For the AEM, different counterions were studied (carbonate, hydroxide and exchangecarbonate). They also measured the sorption curve of the PiperION membrane in the 𝑂𝐻− form at 25°C thanks to the DVS technique. Marino et al. [6] estimated the diffusion coefficient of water in the adsorbed phase for the membrane FAA-3 with NMR technique. Li et al [7] used a L-V set up to estimate the bulk coefficient and the interfacial resistance of a Tokuyama A201 membrane. However, they did not take into account the variation of RH along the channel, which is why the interfacial resistance was dependent on the flow rate. In this study, we developed a coupled experimental/modeling approach for the in-situ estimation of the water transport parameters in AEMFC. Specifically, we were able to evaluate the total water transport resistance in the adsorbed phase, as well as to decorrelate the bulk and the interface resistances. 1. Scientific Approach In this study, a vapor-vapor setup has been developed to estimate in-situ effective water transport resistances. Experimental measurements were coupled with a transport model that can calculate a water flux in the cell depending on different parameters. Inverse method was then employed in order to ensure that experimental and modelled results were consistent. This enables us to identify the parameters of interest, i.e., the effective diffusion coefficient of water in the gas phase through the GDL (𝐷𝑤,𝑔 𝑒𝑓𝑓) and the effective diffusion coefficient in EFCF 2025: Low-Temp. Fuel Cells, Electrolysers & H2 Processing 1 – 4 July 2025, Lucerne Switzerland https://doi.org/10.5281/zenodo.17476267 B0708 Page 3/10 the absorbed phase in the membrane (𝐷𝑤,𝑀 𝑒𝑓𝑓). A water flux through the cell was induced by supplying hydrogen flows of different RH to both sides of the cell. The RH of each flow supplied to the cell was controlled by two in-house built membrane humidifiers. Dew point sensors were used to measure the RH at either the inlet or the outlet of the cell. Any current was present in the cell in order to suppress the electro-osmotic drag. Therefore, we could only focus on the water transfer by diffusion. In this study, the diffusion of water through PiperION membranes has been investigated. Several thicknesses were employed to decorrelate the bulk diffusion and the interfacial resistances of the membrane from the total water transport resistance [8]. The impact of the membrane hydration on the transport properties was investigated by applying different average RH. The investigation focused on membranes in both in 𝑂𝐻− and 𝐶𝑂3−2 form. 2. Experiments/Calculations/Simulations 2.1 Experimental bench The experimental bench used in this work was equipped with flow meters (BROOKS SLA5850S) supplying the cell with hydrogen, air or nitrogen. Prior to entering the cell, the gases were humidified by two membrane humidifiers, as described elsewhere [9]. The water balance was evaluated using the RHs measured by two dew point temperature sensors (Vaisala HMP7), located at the outlets of both sides of the cell. The same sensor was used to measure also the RH of the gas flows supplied to the cell, thanks to an electro-valve which allowed the gas to bypass the cell and flow directly into the sensor. This configuration was intended to reduce the relative error on the measurement. The pressure drop across the cell was measured by employing two differential pressure sensors. The temperatures of the cell and of the humidifiers were controlled using thermal baths coupled with a Pt100 temperature sensor. To prevent water condensation, almost the entire experimental set-up was placed into a hot box. A fan and an electric heater provided uniform heat, which allowed keeping a constant temperature of 80 °C inside the hot box. The remaining tubes were also overheated at 80 °C with heating wires. In this work, a symmetrical cell of 25 cm² was employed. The cell was made of gold-coated brass, ensuring durability and stability. This material was selected for its optimal thermal and electrical conductivity. The channels were parallel. Not all the active surface area was used in the experiments. Indeed, a water concentration gradient was imposed in the cell, and the water concentration was varying along the channel. To avoid reaching equilibrium with regards to water concentration, it was essential to limit the exchange surface. As a result of a preliminary sensitivity study, the width selected for the exchange window was 5 cm, while the length was dependent on the thickness of the membrane. With regards to the materials, GDLs Freudenberg H15C14 were used in this study. The membranes were PiperION of 4 different thicknesses: 20, 40, 60 and 80µm. They were either used in their original form i.e., 𝐶𝑂3−2 or converted into 𝑂𝐻− form. MEA were obtained by coating the catalyst layer on the membranes by spray deposition. On one hand, the conversion of membranes without catalyst layers was achieved by immersing them into a 1 M KOH solution during at least 30 minutes. This process was repeated twice, then the membranes were washed using distillated water. Then, the assembly was completed in a glove box. On the other hand, MEA were converted in-situ by 20 cycling of potential between 0.5 V (300 seconds) and 0.2 V (120 seconds). 2.2 Model A 1+1D model was developed for this study. The one-dimensional transfer of water in the through plane direction through the assembly (y-direction) was considered. A co-flow configuration was adopted. The assembly consisted of two GDLs and a membrane. The EFCF 2025: Low-Temp. Fuel Cells, Electrolysers & H2 Processing 1 – 4 July 2025, Lucerne Switzerland https://doi.org/10.5281/zenodo.17476267 B0708 Page 4/10 model took into account the variation of water concentration along the channel as the boundary conditions were subject to change, resulting in a 1+1D model. The parameters estimated with this model were effective. This means that they are dependent on the material and take several factors into account, such as geometry of the flow fields, clamping force, surface state of the plate among others. Water was supplied to the cell in vapor form along with the gas, and it was transferred across the GDL. Water transport through the GDL is a combination of diffusion and convection. This phenomenon can be described by the Stefan-Maxwell equation for a binary mixture: 𝑁𝑤 𝐺𝐷𝐿 =𝑋𝑤(𝑦) (𝑁𝑤 𝐺𝐷𝐿+𝑁𝑔)−𝐷𝑤,𝑔 𝑒𝑓𝑓 𝑑 𝐶𝑤(𝑦) 𝑑𝑦 (1) Where 𝑁𝑤 𝐺𝐷𝐿 (𝑚𝑜𝑙/(𝑠.𝑚2)) is the molar flux of water through the GDL, 𝑋𝑤,𝐶𝑤 are the molar fraction and the concentration of the water vapor, respectively, 𝑁𝑔 is the molar flux of the carrying gas which is null in the through plane direction because there was no consumption of gas in the cell. Using Equation 1 and knowing that 𝐶𝑤=𝑋𝑤×𝐶𝑡𝑜𝑡, it is possible to write a differential equation with 𝑋𝑤 as unknow. Once the latter is resolved, we obtain the expression of 𝑋𝑤 : 𝑋𝑤(𝑦)=1+(𝑋𝑤 𝑐ℎ− 1 )𝑒(𝑁𝑤 𝐺𝐷𝐿 𝐷𝑤,𝑔 𝐶𝑡𝑜𝑡 𝑦 ) (2) With 𝑋𝑤 𝑐ℎ the molar fraction of water in the channel. As per Equation 2, the relative humidity at the interface between the GDL and the membrane could be calculated. Water transport across the membrane can be described using Fick's law (Equation 3). 𝑁𝑤 𝑀=− 𝐷𝑤,𝑀 𝑒𝑓𝑓 𝜌𝑑𝑟𝑦 𝐸𝑊𝑑 𝜆 𝑑𝑦 (3) With 𝐸𝑊 (𝑔/𝑚𝑜𝑙𝑆𝑂3 −) the equivalent weight of PiperION (425.5 𝑔/𝑚𝑜𝑙) and 𝜌𝑑𝑟𝑦 the dry volume weight of the ionomer (1990 𝑘𝑔/𝑚3). It is important to note that the interfacial resistances were included in the effective diffusion coefficient. Sorption curves were measured using the DVS technique to evaluate the relationship between the relative humidity and the water content, λ: 𝜆(𝑅𝐻)=20.14×𝐻𝑅3−22.75×𝐻𝑅2+12.16×𝐻𝑅+0.14 (4) By determining the water uptake at each side of the membrane and knowing 𝐷𝑤,𝑀 𝑒𝑓𝑓, the water flux through the membrane could be calculated. The variation of the water concentration along the channels was considered: 𝑑 𝑛𝑤 𝑐ℎ(𝑥) 𝑑𝑥 = −𝑏 𝑁𝑤 𝐺𝐷𝐿 (5) With 𝑏 the width of the exchange window. EFCF 2025: Low-Temp. Fuel Cells, Electrolysers & H2 Processing 1 – 4 July 2025, Lucerne Switzerland https://doi.org/10.5281/zenodo.17476267 B0708 Page 5/10 𝑋𝑤 𝑐ℎ(𝑥)= 𝑛𝑤 𝑐ℎ(𝑥) 𝑛𝑤 𝑐ℎ(𝑥)+𝑛𝑔 𝑐ℎ (6) Using this model, it was possible to estimate 𝑋𝑤 at the outlet of the cell, which can be expressed as: 𝑋𝑤 𝑐ℎ(𝑥=𝐿)=𝑃𝑠𝑎𝑡(𝑇𝑑𝑝 𝑚𝑜𝑑𝑒𝑙 𝑜𝑢𝑡 )/𝑃𝑡𝑜𝑡 (7) Therefore, the model provides the outlet dew point temperature (𝑇𝑑𝑝 𝑚𝑜𝑑𝑒𝑙 𝑜𝑢𝑡 ) in accordance with the operational parameters. Inverse methods were used to estimate the effective diffusion coefficient of water in the gas phase in the GDL (𝐷𝑤,𝑔 𝑒𝑓𝑓) and the effective diffusion coefficient in the absorbed phase in the membrane (𝐷𝑤,𝑀 𝑒𝑓𝑓). This indicates that 𝑇𝑑𝑝 𝑚𝑜𝑑𝑒𝑙 𝑜𝑢𝑡 should be equivalent to 𝑇𝑑𝑝 𝑒𝑥𝑝𝑒𝑟𝑖𝑚𝑒𝑛𝑡𝑎𝑙 𝑜𝑢𝑡 . The following least-squares sum was minimized using the fminsearch function in MATLAB. 𝐿𝑆𝑆 (𝐷𝑀 𝑒𝑓𝑓,𝐷𝐺𝐷𝐿 𝑒𝑓𝑓)= ∑[ 𝑇𝑑𝑝 𝑚𝑜𝑑𝑒𝑙 𝑜𝑢𝑡 (𝑖)−𝑇𝑑𝑝 𝑒𝑥𝑝 𝑜𝑢𝑡 (𝑖) ]² 𝑖 (8) To decorrelate these two diffusion coefficients, it is assumed that 𝐷𝑤,𝑔 𝑒𝑓𝑓 is inversely proportional to the total pressure while 𝐷𝑤,𝑀 𝑒𝑓𝑓 is assumed to be independent of the total pressure [10]. For a given average RH, 5 total pressures were investigated from 1.2 to 2 bar. 3. Results The overall water transport resistance in the adsorbed phase 𝑅𝑡𝑜𝑡 𝑎𝑏𝑠 (i.e., the transport resistance in the membrane) was estimated by using Equation 3: 𝑁𝑤 𝑀= − 𝜌𝑑𝑟𝑦 𝐸𝑊𝐷𝑤,𝑀 𝑒𝑓𝑓 ∆𝜆 𝑡𝑀𝑒𝑚𝑏 =− 𝜌𝑑𝑟𝑦 𝐸𝑊 ∆𝜆 𝑅𝑡𝑜𝑡 𝑎𝑏𝑠 (9) According to Equation 9, it was possible to estimate 𝐷𝑤,𝑀 𝑒𝑓𝑓 by introducing the thickness of the dry membrane, 𝑡𝑀𝑒𝑚𝑏. 𝑅𝑡𝑜𝑡 𝑎𝑏𝑠 is the sum of several contributions, and can be expressed as follows: 𝑅𝑡𝑜𝑡 𝑎𝑑𝑠 =𝑡𝑀𝑒𝑚𝑏 𝐷𝑏𝑢𝑙𝑘 +𝑅𝑖𝑛𝑡𝑒𝑟𝑓 1+𝑅𝑖𝑛𝑡𝑒𝑟𝑓 2 = 𝑅𝑏𝑢𝑙𝑘+𝑅𝑠𝑢𝑚 𝑖𝑛𝑡𝑒𝑟𝑓 (10) Equation 10 shows that it is possible to decorrelate the interfacial resistance (𝑅𝑖𝑛𝑡𝑒𝑟𝑓) and bulk resistance (𝑅𝑏𝑢𝑙𝑘) by employing different membrane thicknesses. The main assumption of our method is that 𝑅𝑖𝑛𝑡𝑒𝑟𝑓 does not depend on the membrane thickness. Furthermore, the method developed in this study allows estimating only the sum of the interfacial resistances at both the interfaces, 𝑅𝑖𝑛𝑡𝑒𝑟𝑓1+𝑅𝑖𝑛𝑡𝑒𝑟𝑓2. A linear function was found when plotting 𝑅𝑡𝑜𝑡 𝑎𝑑𝑠 obtained using the coupled experimental/modeling approach described above as a function of the membrane thickness. Considering Equation 10, it is deduced that the slope of the linear function is inversely proportional to 𝐷𝑏𝑢𝑙𝑘, and the intercept is equal to 𝑅𝑠𝑢𝑚 𝑖𝑛𝑡𝑒𝑟𝑓, which allows estimating both parameters. Several linear functions could be obtained when applying different average RHs across the cell (i.e., 58, 68 and 78%). EFCF 2025: Low-Temp. Fuel Cells, Electrolysers & H2 Processing 1 – 4 July 2025, Lucerne Switzerland https://doi.org/10.5281/zenodo.17476267 B0708 Page 6/10 Figure 2 : Total water transport resistance in adsorbed phase for different thicknesses of PiperION membranes (in the 𝐶𝑂2−3form ) at several RHs (70°C) All measurements were carried out at a temperature of 70°C for membranes in 𝐶𝑂2−3 form. The graph shows trend lines with their expressions. The correlation coefficient is sufficiently high to make the linear regressions relevant. It is worth noting that 𝑅𝑡𝑜𝑡 𝑎𝑑𝑠 has been estimated considering the presence of the GDLs in the assembly. When neglecting the presence of the GDL, the value of 𝑅𝑡𝑜𝑡 𝑎𝑑𝑠 would have been 30% and 40% higher than that estimated in our study in the case of 20 µm membrane. For the 80 µm membrane, the increase in 𝑅𝑡𝑜𝑡 𝑎𝑑𝑠 would have been between 7% and 18%. These observations justify the need to take 𝑅𝐺𝑎𝑠 𝑡𝑜𝑡 into account to have a better evaluation of 𝑅𝑡𝑜𝑡 𝑎𝑑𝑠. In addition, 𝑅𝐺𝑎𝑠 𝑡𝑜𝑡 does not depend neither on the membrane thickness nor on the average relative humidity. 3.1 Impact of the relative humidity on 𝑹𝒊𝒏𝒕𝒆𝒓𝒇 and 𝑹𝒃𝒖𝒍𝒌 for different membrane thicknesses Table 1 shows the values of 𝐷𝑏𝑢𝑙𝑘 and 𝑅𝑠𝑢𝑚 𝑖𝑛𝑡𝑒𝑟𝑓 evaluated from the linear regression presented above. It can be seen that 𝐷𝑏𝑢𝑙𝑘 decreases when increasing the RH in the selected range of values, whereas 𝑅𝑠𝑢𝑚 𝑖𝑛𝑡𝑒𝑟𝑓 follows an opposite behavior. It is worth noting that both parameters are not dependent on the membrane thickness, therefore they are intrinsic properties of the PiperION material. 𝑅𝐻 58% 68% 78% 𝐷𝑏𝑢𝑙𝑘 (𝑚²/𝑠) 5,8 10−10 4,8 10−10 4,3 10−10 𝑅𝑠𝑢𝑚 𝑖𝑛𝑡𝑒𝑟𝑓 (𝑠/𝑚) 42 752 61 942 117 618 EFCF 2025: Low-Temp. Fuel Cells, Electrolysers & H2 Processing 1 – 4 July 2025, Lucerne Switzerland https://doi.org/10.5281/zenodo.17476267 B0708 Page 7/10 Table 1 : Value of the bulk diffusion coefficient and the interfacial transport resistance at 70°C as a function of the RH We further investigated the contribution to the total resistance of water in the adsorbed phase of both 𝑅𝑏𝑢𝑙𝑘 and 𝑅𝑠𝑢𝑚 𝑖𝑛𝑡𝑒𝑟𝑓 , depending on the RH (varying from 28 to 78 %), and for two different membrane thicknesses, i.e., 20 and 80 µm (Figure 3). The calculation of the bulk resistance was performed in accordance with Equation 10. The membrane’s ion exchange capacity (IEC) can decrease at lower RHs [11]. This is why the experimental protocol adopted in this study first planned measurements at high RH to limit the risk of membrane damaging. Figure 3 : Bulk resistance and interfacial resistances as a function of the RH for two membranes (20µm and 80µm thick) at 70°C. For the 20 µm membrane and at the lowest RH (i.e., 28%), 𝑅𝑏𝑢𝑙𝑘 was found to account for 56% of the total transport resistance. At RH = 38%, the value of 𝑅𝑏𝑢𝑙𝑘 was almost equal to that of 𝑅𝑖𝑛𝑡𝑒𝑟𝑓. For RH > 38%, the contribution of 𝑅𝑖𝑛𝑡𝑒𝑟𝑓 become predominant. For the highest RH (i.e., 88%), the interfacial resistance is responsible for 66% of the total transport resistance. With regards to the 80 µm membrane, it was found that 𝑅𝑏𝑢𝑙𝑘 is always predominant. At the lowest RH, the bulk resistance contributes approximately 80% of the total transport resistance. The contribution of the bulk resistance decreases to 67% at the highest RH investigated. According to the results obtained, it can be concluded that 𝑅𝑖𝑛𝑡𝑒𝑟𝑓 is the main factor contributing to the total transport resistance of the 20 µm membrane, with the exception of cases where the relative humidity is below 38%. Conversely, 𝑅𝑏𝑢𝑙𝑘 was found to be the dominant factor whatever the RH in the case where a thick membrane was used. This trend can be explained by the fact that the bulk resistance increases proportionally with the thickness, while the interface resistance remains constant. EFCF 2025: Low-Temp. Fuel Cells, Electrolysers & H2 Processing 1 – 4 July 2025, Lucerne Switzerland https://doi.org/10.5281/zenodo.17476267 B0708 Page 8/10 The total transport resistance was found to decrease when increasing the RH from 28% to 38% while it increases when increasing progressively the RH towards higher values. It is therefore concluded that the drop at the low RH can be due to bulk diffusion, since the interfacial resistance remains constant for this range of RH. Despite this exception, both 𝑅𝑏𝑢𝑙𝑘 and 𝑅𝑖𝑛𝑡𝑒𝑟𝑓 were found to increase when increasing RH. This trend of 𝑅𝑖𝑛𝑡𝑒𝑟𝑓 can be explained by the membrane swelling whereas it is dimensionally constrained: the size of the membrane pores may decrease, which makes 𝑅𝑖𝑛𝑡𝑒𝑟𝑓 increasing at higher RH. An alternative hypothesis is that the model used here is too simple, resulting in effective parameters that do not evolve in line with the intrinsic parameters as a function of RH. But a similar trend was observed for a Nafion membrane [12]. Concerning the bulk resistance, it is more practical to study the bulk diffusion coefficient instead, which is independent of the thickness. The diffusion coefficient initially increases between the range of RH 28-38 %, before decreasing for higher values of RH (Figure 4). Figure 4 : Bulk diffusion coefficient of Piperion at 70°C This variation can be attributed again to the fact that the membrane swells but by being dimensionally constrained. It should be noted that the growth rate of the two resistances is not identical. It has been observed that 𝑅𝑖𝑛𝑡𝑒𝑟𝑓 increases as a function of RH at a faster rate than that for 𝑅𝑏𝑢𝑙𝑘, meaning that the contribution of 𝑅𝑖𝑛𝑡𝑒𝑟𝑓 rises significantly as the RH increased. As evidence, the contribution of 𝑅𝑖𝑛𝑡𝑒𝑟𝑓 to the total resistance is equal to 50% for RH = 38%, while it is about 16% higher for RH = 88 % when using a 20 µm membrane. For the 80 µm membrane, the contribution of 𝑅𝑖𝑛𝑡𝑒𝑟𝑓 reaches 20% for RH = 38%, while at RH = 88% it is about 33%. 3.2 Impact of a catalyst layer on the 𝑹𝒕𝒐𝒕 𝒂𝒅𝒔 The impact of a catalyst layer with different platinum loadings on the water transport properties was also investigated. For that, the membranes were converted to the 𝑂𝐻− form. The results obtained when using MEAs were compared with those obtained using only membranes without a catalyst layer. The conversion of the membrane with a catalyst layer was carried out in-situ, as mentioned above. Figure 4 represents the behavior of the total water transport resistance as a function of RH for various 20 µm membranes. The membranes used for this investigation include one with catalyst layers (MEA) loaded at 0.68 mg of platinum, another with a loading of 0.34 mg of platinum, and two without catalyst layers (two samples for reproducibility). EFCF 2025: Low-Temp. Fuel Cells, Electrolysers & H2 Processing 1 – 4 July 2025, Lucerne Switzerland https://doi.org/10.5281/zenodo.17476267 B0708 Page 9/10 Figure 5 : Total water transport resistance in the adsorbed phase for a membrane, and membrane with catalyst layer As outlined in the previous section, all the total resistances were found to increase with the RH. The MEA exhibit a higher transport resistance, which is 30% (0.34 mg) and 34% (0.68 mg) higher than that estimated for the membrane alone for RH between 58% and 78%. This difference is particularly pronounced at 88% of RH, with an increase of 𝑅𝑡𝑜𝑡 𝑎𝑑𝑠 of 43% compared to the MEA 0.34 mg and 51% compared to the MEA 0.68 mg. Moreover, the catalyst loading had also an impact on 𝑅𝑡𝑜𝑡 𝑎𝑑𝑠 at RH = 88%. Indeed, the value found for a loading of 0.68 mg was 16% higher than that at 0.34 mg. Conversely, the catalyst loading had only a negligible impact on the total resistance at lower RHs, (≈ 5% of difference), which indicates that the transport properties are almost identical for the two cases. Therefore, it can be concluded that the presence of a catalyst layer introduces an interface resistance rather than a diffusion resistance. Conclusion The water transport properties of the membrane PiperION were investigated in the present study. Specifically, the contribution of the interfacial and the bulk resistance to the total water transport resistance in the adsorbed phase was evaluated using a coupled experimental/modelling approach. This method was based on the evaluation of the water flux in the through plane direction across the assembly, which was induced by a difference in the RH between the two sides of the assembly. To do that, a vapor-vapor setup was used. Diffusions in the gas and absorbed phase were decorrelated by performing experiments with different pressures. The decorrelation of the interfacial and bulk resistance was performed using membranes with different thicknesses. Several average RH across the assembly were investigated. The interfacial resistance was found to be the predominant contributor to the total resistance when using very thin membranes, and for RH higher than 38%. On the other hand, the contribution of the bulk resistance was found to be predominant when using thick membranes, whatever the average RH installed over the cell. Both the interfacial and bulk resistances increased when increasing the RH.