Effect of the Gas Diffusion Layer Design on the Water Management and Cell Performance of a PEM Fuel Cell

  Citation: Martín-Alcántara, A.; González-Morán, L.; Pino, J.; Guerra, J.; Iranzo, A. Effect of the Gas Diffusion Layer Design on the Water Management and Cell Performance of a PEM Fuel Cell. Processes 2022,10, 1395. https://doi.org/10.3390/ pr10071395 Academic Editors: Alessandro D’Adamo, Stefano Fontanesi and Thomas Lauer Received: 13 June 2022 Accepted: 12 July 2022 Published: 17 July 2022 Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. Copyright: © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/). processes Article Effect of the Gas Diffusion Layer Design on the Water Management and Cell Performance of a PEM Fuel Cell Antonio Martín-Alcántara 1,* , Laura González-Morán 2, Javier Pino 2, José Guerra 2and Alfredo Iranzo 1,2 1 AICIA-Thermal Engineering Group, Camino de los Descubrimientos, s/n, 41092 Sevilla, Spain; airan[email protected] 2Thermal Engineering Group, Energy Engineering Department, University of Seville, Camino de los Descubrimientos, s/n, 41092 Sevilla, Spain; [email protected] (L.G.-M.); [email protected] (J.P.); [email protected] (J.G.) *Correspondence: [email protected] Abstract: The influence of the different properties of the gas diffusion layer (GDL) on the operation of a liquid-cooled, proton-exchange polymer electrolyte fuel cell (PEMFC) has been studied in this work. Three-dimensional numerical simulations (CFD) have been conducted to compare several commercial GDLs with different properties, analyzing their influence on the cell performance. Specifically, four GDLs (AvCarb P-75, SIGRACET 34BC, SIGRACET 34BA and Toray TGP-H-090) have been studied, two of them including a microporous layer (MPL). The effect of the MPL has been inspected by contrast of the results obtained with the same GDL, with or without MPL. Potentiostatic boundary conditions have been applied, varying the electric potential between 1.05 and 0.35 V to obtain a representative i−V curve with enough resolution. Detailed postprocessing tasks were carried out to gain a deeper understanding on the phenomena occurring within the cell for each GDL. It can be concluded from this work that a high electrical conductivity and a high permeability lead to a better fuel cell performance. On the other hand, although the presence of MPL provides lower permeability leading to a worse overall performance, it has been shown that the lack of it may result in membrane dehydration and cell degradation issues. Keywords: PEMFC; fuel cell; coolant channel; CFD; numerical model 1. Introduction Proton-exchange membrane fuel cells (PEMFCs) directly convert hydrogen and oxygen (or air) flow streams into electric energy without going through a thermodynamic direct cycle. The anode is fed with hydrogen and the cathode with oxygen, and as a part of the exothermic reaction water and heat are generated. Although oxidation is present in the process, there are no combustion mechanisms generating pollutants [ 1 ] or greenhouse gases. For this reason, fuel cells have become attractive for numerous researchers, companies and governments in the last few years, being considered a promising alternative for power generation both in stationary and portable applications in order to achieve 2030’s climate objectives [ 2 – 4 ]. Nevertheless, the main technological barriers appearing in the development of these technologies are costs and durability, both delaying the full implementation of PEMFC nowadays. PEMFCs are characterized by a polymeric membrane and successive thin porous layers. The membrane is a very thin solid electrolyte permitting the transfer of protons but not the electric current. It is the active layer where the electrochemical reaction takes place on the catalyst layer (CL). The actual working temperature of the membrane lies in the range around 40–90 ◦ C. These temperatures require a very effective CL (typically platinum or platinum alloys) on both sides of membrane for the ease of the reaction. Next to the CLs, gas diffusion layers (GDLs) are assembled to ensure a uniform distribution of the reactants flowing through the channels towards the membrane. All these components are sandwiched in between solid conductive plates with a controlled compression not Processes 2022,10, 1395. https://doi.org/10.3390/pr10071395 https://www.mdpi.com/journal/processes Processes 2022,10, 1395 2 of 16 compromising their adequate performance [ 1 ]. GDLs normally are carbon-based and consequently have a high electronic conductivity. Their pore size has been designed to maximize the H2and O2diffusion. Because at the cathode side, the reaction 1 2O2+2H2+2e−→H2O, (1) occurs, some water is mostly produced in this electrode. In addition, the temperature is frequently below 100 ◦ C, and the pressure in the range 1–3 atm, which propitiates the progressive formation of water droplets, limiting the diffusion process and partially blocking the flow of O2 . Consequently, GDLs are designed with a high hydrophobicity in order to avoid flooding caused by liquid water [ 1 ]. A general description of the GDL properties and compositions can be found in [ 1 , 5 – 7 ]. The thermal and electrical conductivity of GDLs is studied in [ 8 – 12 ], whereas their porosity and permeability properties are discussed in [11–13] . Regarding the microporous layer (MPL), a widespread opinion establishes that the addition of this thin layer may result in a detrimental effect on the PEMFC’s performance. However, various discrepancies are found in [ 14 ], and the existence of an optimal thickness range exploiting the capabilities of the MPL is analyzed in [ 15 – 17 ]. Some disagreements have been found in these works that do not contribute to clarify the issue. Furthermore, the benefits of using a MPL on the fuel cell performance not only depend on the thickness of the layer, but also on the side in which it is assembled. For instance, the work in [ 8 ] assessed different possibilities regarding the optimum MPL location. Alternatively, all the aforementioned studies accounted for different MPL configurations, but none of them considered the absence of this layer. Therefore, a sensitivity analysis on the characteristics of some commercial GDLs will be performed in this study to analyze the influence of such layer on the PEMFC’s performance. To that end, the use of computational fluid dynamics (CFD) modeling has become a crucial tool for the understanding of the fuel cell operation and the phenomena involved in it. The different mechanisms that might deteriorate the device can be identified and prevented before the assembly of a real PEMFC, avoiding unnecessary damages and their derived costs [ 18 , 19 ]. For instance, it is worth mentioning the recent work of Corda et al. [20] , which realistically analyzes the effect of the clamping pressure on the GDL’s properties, concluding that compression plays a negative role on the PEMFC’s performance, especially at high current densities. In this work, the effect of different commercial GDLs is analyzed by means of CFD simulations. A discussion on several properties, such as electrical and thermal conductivity, is provided, together with their influence on water saturation and membrane hydration. Liquid cooling is also considered in the model, where the clear relationship with the local water saturation levels within the GDL is analyzed in detail, in particular the along-the- channel evolution of the under-rib with respect to under-channel locations within the GDL. The study of the influence of the MPL is also carried out, showing clear effects on the temperature profiles, oxygen concentration, and cell performance. The paper is structured as follows. First, the description of the model, the numerical details, and the different properties of the studied GDLs will be introduced. Next, the results obtained from the simulations are shown and discussed. Some conclusions and future works will be drawn at the end of this work, remarking that higher electrical and thermal conductivity, porosity and permeability lead to a better PEMFC operation. Simultaneously, the presence of an MPL is found to lead to a worse performance, but its absence leads to serious hydration and degradation issues. Processes 2022,10, 1395 3 of 16 2. Materials and Methods The built-in PEMFC model of the software ANSYS-FLUENT [ 21 , 22 ] was used for the performance of the numerical simulations of this work with the finite volume method (FVM). Therefore, the conservation equation for a given magnitude φ can be enunciated as ZAρφV·dA =ZA Γφ∇φ·dA +ZVSφdV, (2) where A is the integration area, V is the integration volume, and Sφ is the source term involving φ . In addition to the transport equations, two potential equations are solved for the solid phase (j≡solid) and the membrane phase (j≡membrane), respectively, ∇ · σj∇ϕj+Rj=0, (3) with σ representing the electric conductivity, ϕ representing the electric potential and Rj representing the volumetric transfer current. The Stefan–Maxwell equation governs the multi-species diffusion in the GDLs [ 19 ], although the model presented in this work calculates species diffusivity from the multicomponent diffusion model in [22], i.e., Dij e f f = (1−s)rsε1.5Dij, (4) where Dij e f f is the effective diffusivity of the gas species ij , s is the liquid saturation, rs is the exponent of the pore blockage, ε is the porosity of the porous medium, and Dij is the gas species mass diffusivity computed by the full multicomponent diffusion (see [ 21 ]). The electrochemistry is modeled by means of the Butler–Volmer equations: Ran =ian re f [H2] [H2]re f !γan eαan Fηan/(RT)−e−αcat Fηan/(RT), (5) Rcat =icat re f [O2] [O2]re f !γcat eαan Fηcat/(RT)+e−αcatFηcat/(RT). (6) A saturation model approach was used for the liquid water formation and transport. The conservation equation for the liquid sis given by ∇ · ρl Ks3 µ1 dpc ds∇s=rw, (7) where the subscript l stands for the liquid phase, rw is the condensation rate, K is the permeability, µ is the viscosity and pc is the capillary pressure computed as a function of s according to the Leverett function [ 23 ]. Alternatively, the water content λ is computed following the empirical correlation of Springer et al. [24]: (λ=0.043 +17.18a−39.85a2+36a3,(a≤1), λ=14 +1.4(a−1),(a>1),(8) where ais the water activity defined in [22] as a=Pv/Psat +2s. Assuming steady-state conditions, the numerical solution of Equations (2) – (8) was obtained in the computational domain represented by the single straight-channel in Figure 1, with the main metrics of the grid summarized in Table 1. The geometry represents a single channel of a representative cell for automotive applications with corrugated bipolar plates and integrated water-cooled channels. Based on the work of Wilkinson and Vanderleeden [25] , corrugated plates are claimed to be advantageous because they reduce the overall plate thickness and lead to higher stack power densities. This particular design configuration is well suited to straight-channel continuous flow field patterns. For simplicity, one half of Processes 2022,10, 1395 4 of 16 a repeating unit was modeled. On the other hand, the physical properties of the PEMFC components are shown in Tables 2and 3. Other than these, the values adopted are the same as those in [19]. Figure 1. Schematics of the PEMFC domain: ( a ) computational mesh, ( b ) cross-sectional view of the PEMFC geometry. Table 1. Mesh properties. Metric Value Cells 10,170 Faces 318,917 Nodes 124,461 Max. Face Angle (◦)129.6 Min. Face Angle (◦)49.04 Max. Edge/Length Ratio 2602.2 Min. Edge/Length Ratio 10.7 Max. Element/Volume Ratio 3.4 Several GDLs of 285 ± 30 µ m thickness were assessed in the simulations. In particular, the commercial diffusion layers chosen for the study were the AvCarb P-75, SIGRACET 34BC, SIGRACET 34BA and the TORAY GP-H-090, with only the first two including an MPL. In order to conduct a systematic analysis on the effects of the MPL, all the computational domains were built with an MPL reserved room. If the microporous layer was actually considered for the simulation, the properties of the reserved space were introduced accordingly; otherwise, the properties of the adjacent GDL were extended to this region. Six different GDLs were studied in all, with their main properties described in Table 3. Processes 2022,10, 1395 5 of 16 Table 2. Physical and transport properties of the rest of the PEMFC components not appearing in Table 3. Other than these, the value of the properties given in [19] were adopted. Geometrical and Material Parameters Channel depth (mm) 100.0 Anode CL thickness (mm) 0.005 Cathode CL thickness (mm) 0.005 Thermal conductivity of CL (W/(m K)) 10.0 Porosity of CL 0.2 Gas permeability of CL (m2×10−13)2.0 Contact angle of CL, θc(◦)95 Surface-to-volume ratio of CL (1/m ×105)2.0 Membrane thickness (mm) 0.030 Equivalent weight of membrane (kg/kmol) 1100 Membrane density (kg/m3)1980 Thermal conductivity of membrane (W/(m K)) 2.00 Thermal conductivity of bipolar plate (W/(m K)) 100 Table 3. Physical and transport properties of GDLs and MPLs. AvCarb Sigracet Toray P-75 P-75 w/o MPL SG34BC SG34BC w/o MPL SG34BA TGP-H-090 GDL porosity (%) 85 85 75 75 83 78 MPL porosity (%) 40 85 40 75 83 78 Contact angle of GDL, θc(◦)107 [7] 107 [7] 104 [7] 104 [7] 104 [7] 138 [7] Contact angle of MPL, θc(◦)126 [7] 107 126 104 104 138 Gas permeability of GDL (m2×10−12)5.70 5.70 8.71 8.71 8.71 4.53 [7] Gas permeability of MPL (m2×10−12)0.097 5.70 0.076 8.71 8.71 4.53 Density of GDL (kg/m3)2719 2719 444.44 444.44 307.14 440 Electrical conductivity of GDL (S/m) 1250 1250 286.36 286.36 254.55 1250 Thermal conductivity of GDL (W/(m K)) 1.7 1.7 0.467 [13] 0.467 [13] 0.533 [13] 1.7 Thickness (µm) 275 275 315 315 280 280 MPL YES NO YES NO NO NO PTFE (%) YES YES 15 15 5 5 The steady-state operating conditions are described in the following (see Table 4); the species mass flow rate, temperature and relative humidity were fixed at the anode and cathode inlets. At the coolant inlet, mass flow rate and temperature were fixed. The cell voltage was prescribed for each operation point (potentiostatic conditions) so that the corresponding current density could be calculated numerically. Table 4. Summary of the operating conditions. Anode Inlet Cathode Inlet Anode Cooling Inlet Cathode Cooling Inlet Fuel mass flow (kg/s) 1 ×10−71.4 ×10−6− − Temperature (K) 343.15 343.15 343.5 343.15 Relative Humidity (%) 40 5 − − Coolant mass flow (kg/s) − − 1.1 ×10−42.9 ×10−5 The polarization curve i−V of each case was computed ranging from 1.05 to 0.35 V with a step width of 0.1 V. Subsequently, a detailed postprocessing was carried out to better understand the internal phenomena occurring in the GDLs and the overall impact on the fuel cell. The postprocessing tasks comprised the following stages: 1. Calculation of the volume-weighted average of the key variables (temperature, water content, liquid saturation and oxygen mass fraction) for each operation point for the GDLs, MPLs, CLs and membrane. This step was extended to all the cases in study; 2. Study on the evolution of the aforementioned variables along the longitudinal axis of the fuel cell, specifically at the points highlighted in the cross-sectional plane of Figure 2. The duplicity of the points in GDLs and MPLs is done to enable the study Processes 2022,10, 1395 6 of 16 on under-rib and under-channel effects. This is only done for the cases with MPL at the low and medium voltages 0.45 and 0.65 V, respectively; 3. Representation of contour plots in the membrane mid-plane (in-plane), studying the distributions of temperature, water content and current flux density for the cases with MPL and their variations at 0.45 and 0.65 V. Figure 2. PEMFC cross-section plane showing the locations for the longitudinal axis evolution of variables. The legend indicates which variables are studied at each location. It must be mentioned that no experimental data of a single straight-channel with cooling channels in a similar arrangement to that of Figure 1were identified, and therefore, a detailed model validation could not conducted. In order to overcome this drawback, best practice guidelines derived from the experience in previous CFD research works on PEMFC modeling and experimental validation [ 18 , 19 , 26 , 27 ] were used for the development of the CFD model. 3. Results The results obtained from the CFD simulations are described and commented on in this section. In Figure 3, the polarization curve i−V is constructed from the data resulting from the simulations. Most of the cases draw the typical shape of a polarization curve, starting with a steep decrease (activation region), followed by a smooth and linear decrease in the ohmic region, to finish with a suddenly decrease due to the mass losses. Notice that the more accused discrepancies between the different curves are found in the latter. Processes 2022,10, 1395 7 of 16 Figure 3. Polarization curves for each of the commercial GDLs in study, with and without MPL. In terms of the PEMFC performance, three wel-differentiated behaviors can be inferred from Figure 3. First, the case of the poorest performance (i.e., the case providing the smallest current density) corresponding to the AvCarb GDL+MPL assembly is observed. As mentioned in [ 8 , 15 – 17 ], the low i magnitude can be explained by the fact that the MPL in the cell does not fit its ideal thickness. If of arbitrary size, it has been discussed in [ 13 ] that the MPL deteriorates the electrical conductivity, consequently decreasing the PEMFC performance. Indeed, the largest difference on the current density at the very same voltage (0.45 V) is more than twice this with MPL when compared with its counterpart without MPL. Secondly, regarding the SIGRACET diffusion layer, a similar trend to the previous one can be observed, meaning that the GDL+MPL assembly generates a smaller current density than that obtained solely with the GDL. Even though the differences in their electrical conductivity are just about 11.1%, a better performance can be seen in the SIGRACET 34BC case (which demonstrates the highest electrical conductivity, of course). On the other hand, the SIGRACET 34BA has a 12.4% higher thermal conductivity with respect to the SIGRACET 34BC, but it has been found to exhibit a worse overall performance due to the lower electrical conductivity of the latter, which seems to be dominant at this point. Basically, high electrical conductivity is of relevance because it is directly related to a superior proton conduction, which leads to a better performance [ 9 ]. In addition, the comparison between both SIGRACET components reported in [ 8 , 10 , 11 ] corroborates the results in [ 12 ]: the higher the porosity is, the lower the electrical conductivity. Notwithstanding, high thermal conductivity is also necessary to achieve temperatures in the range appropriate for membrane hydration [ 28 ]. Higher PTFE quantities lead to a higher contact angle [ 11 ] and better porosity and permeability properties [ 12 ]; however, the thermal conductivity will be reduced. To conclude the analysis of Figure 3, the GDLs without the MPL will be addressed, presenting higher electrical and thermal conductivity than those formerly discussed. Furthermore, these GDLs results are similar in both cases, with the main differences lying in their porosity and permeability properties, which are, respectively, 8 and 20% higher in AvCarb’s than in TORAY’s. Carcadea et al. stated in [ 11 ] that the increase in the porosity leads to a significant improvement in the performance, as herein shown. Processes 2022,10, 1395 8 of 16 From the results in Figure 3 , the power vs. current density curve can be drawn in Figure 4 . Like in Figure 3, the aforementioned three distinct regions are displayed now in Figure 4in terms of power. From the top to the bottom, one can first identify the cases without MPL, which are of highest thermal and electrical conductivity, then the cases without MPL, which are of lower thermal and electrical conductivity, and finally the cases without MPL, which are of high electrical and thermal conductivity. It can be seen that the GDL AvCarb P-75 without MPL generates the largest power quantity (about 1.1 W/cm 2 ) among all the layers in study. In contrast, the GDL with MPL engineered by the same manufacturer is the one providing the poorest performance (around 0.6 W/cm 2 at 0.73 A/cm 2 ). Their differences are roughly given by a factor of 2 (in the best of circumstances). On the other hand, one can note that the power curve in the cases without MPL increases with the current density up to a maximum of about 2 A/cm 2 . From there on, the PEMFC performance starts to decrease. Regarding the differences in porosity and permeability of the GDLs, the impact on the power output is only a 2.5% decrease. However, the lower conductivity implies a more noticeable drop (20–15%) on the PEMFC power response. Figure 4. Comparison power vs. current density for all the studied cases. Same legend as in Figure 3. 3.1. Volume-Weighted Averages The averaged water content and the maximum and mean temperatures in the membrane are displayed against the current density in Figure 5. The water content is plotted in the main axis and temperature in the secondary. For comparison’s sake, the main differences between the SIGRACET 34BA and AvCarb P-75 are assessed, with the first GDL presenting low thermal and electrical conductivity without MPL, and the second (with MPL) with 5 and 3.5 times higher electrical and thermal conductivity. Processes 2022,10, 1395 9 of 16 Figure 5. Temperature T and water content λ vs. current density in the membrane: ( a ) SIGRACET 34BA (no MPL), (b) AvCarb P-75 (with MPL). Because the cathode reaction is exothermic, temperatures will typically increase with the current density. If more oxygen is consumed, more heat will be released. In Figure 5, the temperature growth looks linear from small to intermediate current densities, but from a given point (say i= 1.25 A/cm 2 in Figure 5a and i= 0.75 A/cm 2 in Figure 5b ), such an increase becomes apparently exponential. Alternatively, the water content increases with the current density because a larger water quantity is generated in the cathode due to a higher oxygen consumption. Similar to temperature, the initial response of the water content presents a linear relation with the current density, but from a critical point (practically coinciding with that of the temperature sudden increase), evaporation effects must be dominant so that the water level starts to decay rapidly. The differences in the electrical and thermal conductivity are related to this process, as can be clearly observed in Figure 5a,b. Due to a better proton conduction, the AvCarb GDL reaches lower temperatures than those in SIGRACET 34BA, which approximately results in a 30% higher water content of the first with respect to the latter. When the MPL is present, the PEMFC performance is enhanced because the water transport mechanisms and hydration are reinforced [ 11 ]. It must be kept in mind that membrane hydration is a decisive mechanism to maximize the PEMFC operation, avoiding the degradation of their components [ 9 ]. Rather, the high temperatures reached by the SIGRACET 34BA are driven by the lower conductivity and by a worse level of hydration, as the water content in membrane describes. These high temperatures are damaging for the lifespan of the cell components. Even though the average temperature is below 100 ◦ C (say, a degradation threshold [ 1 ]), this value is locally exceeded in about 2 K. In addition, the thermal gradient should not exceed a 5 K degradation threshold, which is surpassed by 1.5 K, leading to higher thermal stresses. Figure 6shows the liquid saturation in the cathode GDL (main axis) and the oxygen mass fraction in the cathode CL (secondary axis) against the current density. The diffusion layers SIGRACET 34BA and AvCarb P-75 without MPL were compared, with their main differences in the thermal and electric conductivity. Thus, according to both cases in Figure 6, it can be observed that the oxygen mass fraction decreases with the current density because a larger quantity of O2 needs to be consumed to work at the given operation point. The layers next to the CL present the lowest oxygen concentration level because they are closer to the reaction themselves. It can be noticed that the O2 decay is practically linear in both cases, and their concentrations are similar (see Figure 6a,b). On the other hand, even though the thermal and electric conductivity of the AvCarb P-75 GDL is several times higher than that of SIGRACET 34BA, their differences in the porosity and permeability are smaller. To specify, the permeability of SIGRACET 34BA is 35% higher, whereas the Processes 2022,10, 1395 16 of 16 14. Ozden, A.; Shahgaldi, S.; Li, X.; Hamdullahpur, F. A review of gas diffusion layers for proton exchange membrane fuel cells—With a focus on characteristics, characterization techniques, materials and designs. Prog. Energy Combust. Sci. 2019 ,74, 50–102. [CrossRef] 15. Chen, G.; Zhang, G.; Guo, L.; Liu, H. Systematic study on the functions and mechanisms of micro porous layer on water transport in proton exchange membrane fuel cells. Int. J. Hydrogen Energy 2016,41, 5063–5073. [CrossRef] 16. Ismail, M.; Borman, D.; Damjanovic, T.; Ingham, D.; Pourkashanian, M. On the through-plane permeability of microporous layer-coated gas diffusion layers used in proton exchange membrane fuel cells. Int. J. Hydrogen Energy 2011 ,36, 10392–10402. [CrossRef] 17. Park, S.; Lee, J.W.; Popov, B.N. A review of gas diffusion layer in PEM fuel cells: Materials and designs. Int. J. Hydrogen Energy 2012,37, 5850–5865. [CrossRef] 18. Iranzo, A.; Gregorio, J.M.; Boillat, P.; Rosa, F. Bipolar plate research using Computational Fluid Dynamics and neutron radiography for proton exchange membrane fuel cells. Int. J. Hydrogen Energy 2020,45, 12432–12442. [CrossRef] 19. Iranzo, A.; Munoz, M.; Rosa, F.; Pino, J. Numerical model for the performance prediction of a PEM fuel cell. Model results and experimental validation. Int. J. Hydrogen Energy 2010,35, 11533–11550. [CrossRef] 20. Corda, G.; Fontanesi, S.; d’Adamo, A. Methodology for PEMFC CFD Simulation Including the Effect of Porous Parts Compression. Int. J. Hydrogen Energy 2022,47, 14658–14673. [CrossRef] 21. ANSYS. Ansys Fluent User’s Guide; Ansys Inc.: Canonsburg, PA, USA, 2022; pp. 3099–3190. 22. ANSYS. Ansys Fluent Theory Guide; Ansys Inc.: Canonsburg, PA, USA, 2022; pp. 845–880. 23. Pasaogullari, U.; Wang, C. Liquid water transport in gas diffusion layer of polymer electrolyte fuel cells. J. Electrochem. Soc. 2004 , 151, A399. [CrossRef] 24. Springer, T.E.; Zawodzinski, T.; Gottesfeld, S. Polymer electrolyte fuel cell model. J. Electrochem. Soc. 1991 ,138, 2334. [CrossRef] 25. Wilkinson, D.; Vanderleeden, O. Serpentine flow field design. In Handbook of Fuel Cells. Fundamentals, Technology, Applications; Vielstich, W., Lamm, A., Gasteiger, H., Eds.; John Wiley and Sons: Chicester, UK, 2003; Volume 3. 26. Iranzo, A.; Muñoz, M.; Pino, J.; Rosa, F. Update on numerical model for the performance prediction of a PEM Fuel Cell. Int. J. Hydrogen Energy 2011,36, 9123–9127. [CrossRef] 27. Iranzo, A.; Boillat, P.; Rosa, F. Validation of a three dimensional PEM fuel cell CFD model using local liquid water distributions measured with neutron imaging. Int. J. Hydrogen Energy 2014,39, 7089–7099. [CrossRef] 28. Burheim, O.S.; Pharoah, J.G.; Lampert, H.; Vie, P.J.; Kjelstrup, S. Through-plane thermal conductivity of PEMFC porous transport layers. J. Fuel Cell Sci. Technol. 2011,8, 021013. [CrossRef]