Bipolar plate research using Computational Fluid Dynamics and neutron radiography for proton exchange membrane fuel cells

Depósito de Investigación de la Universidad de Sevilla https://idus.us.es/ This is an Accepted Manuscript of an article published by Elsevier in International Journal of Hydrogen Energy, Vol. 45, on April 2020, , available at: https://doi.org/10.1016/j.ijhydene.2020.02.183 © 2020 Elsevier. En idUS Licencia Creative Commons CC BY-NC-ND 1 Bipolar Plate research using Computational Fluid Dynamics and Neutron Radiography for Proton Exchange Membrane Fuel Cells Alfredo Iranzo a *, José Manuel Gregorio b, Pierre Boillat c,d, Felipe Rosa a a Thermal Engineering Group, Energy Engineering Department, School of Engineering, Universidad de Sevilla. Camino de los Descubrimientos, s/n, 41092, Sevilla, Spain b AJUSA Hydrogen Technologies, Parque Empresarial Ajusa, Calle 1, nº1. 02006 Albacete, Spain c Electrochemistry Laboratory (LEC), Paul Scherrer Institut (PSI), 5232 Villigen PSI, Switzerland d Laboratory for Neutron Scattering and Imaging (LNS), Paul Scherrer Institut (PSI), 5232 Villigen PSI, Switzerland * Corresponding author: airan[email protected] Tel.: +34-954-487471, Fax: +34-954-487247 Abstract This work presents the development of liquid-cooled industry-scale bipolar plates for improved water management in PEM Fuel Cells. The methods used for the design development are based on Computational Fluid Dynamics (CFD) modelling and simulation, and Neutron Radiography experiments to analyse liquid water distributions within the cell for different operating conditions. A novel 140 cm2 bipolar plate was designed and manufactured on 0.1 mm thick stainless steel using pre-coated strip steel. CFD modelling carried out for the novel design predicted a significant improvement in terms of cell performance, as well as a more uniform *Manuscript (Revised, clean, unmarked, FINAL version) 2 temperature distribution within the membrane. Liquid water distributions were later analysed by neutron radiography experiments, defining a set of different operating conditions (current density, stoichiometry, inlet gases dew point, and cell temperature). Electrochemical and neutron radiography results are presented for all cases and the influence of the operating conditions is discussed. Liquid water distributions within the cell are also analysed and compared against the CFD model results obtained. The influence of the gas flow configuration (reactant gases and cooling water) is clearly observable in the results. Keywords: Bipolar Plate; PEM Fuel Cell; CFD; numerical model; Neutron Radiography; water management 1. Introduction The development of Bipolar Plates (BPPs) allocates major research efforts in the field of PEM fuel cells, as BPP design is having a direct influence in the final performance of the stack [1,2] and also on durability [3,4]. The combined effect of both Bipolar Plate and Gas Diffusion Layer (GDL) must ensure a uniform distribution of reactants over the electrodes as well as an efficient liquid water removal from the GDL pores and BPP channels, so that reactants mass transport is not hindered by liquid water blockages. Development of novel flow fields for BPPs aims at avoiding mass transport limitations, so that the stack can safely and stably operate at higher current densities, which indeed results in an increase in the stack power density, thus reducing the final cost and enabling an improved and more compact system integration. Different flow field design layouts for BPPs have been proposed aiming at enhancing the final performance of PEM fuel cells, where parallel, serpentine, interdigitated or pin-type 3 layouts are the most commonly found [2,5]. The material of the BPP is also playing a fundamental role in the performance, power density, and durability due to eventual BPP corrosion [6,7]. While graphite plates of a few mm thickness can be used for stationary applications, metallic-coated plates in the range of 0.1-0.5 mm are used in industry for transport applications [8,9], given the higher power density required and the suitability for mass production by well-known processes such as stamping or hydroforming [10]. This work presents the developments carried out by AJUSA Hydrogen Technologies and University of Sevilla aimed at developing novel flow field designs of bipolar plates with improved water management for the commercial PEM fuel cell systems manufactured by AJUSA Hydrogen Technologies [11]. The scientific approach followed was based on a combination of both experimental and modelling techniques. Regarding experimental techniques, not only conventional test bench polarization curves and durability tests were carried out, but also in-operando visualization of the liquid water content and distributions by means of Neutron Imaging [12-14]. This technique was applied at the SINQ facility at Paul Scherrer Institut in Villigen, Switzerland. The NEUTRA beamline [15] was used to determine the liquid water content and distributions for different cell operating conditions representative of the real operation of the cell in the final system. Neutron radiographs of the in-operando fuel cell under different conditions and flow configurations provided an outstanding understanding of the liquid water distributions and their influence on the final cell performance. Regarding modelling and simulation, a Computational Fluid Dynamics (CFD) model was developed within the commercial software ANSYS-FLUENT [16], where the model included the cell electrochemistry, heat transfer, and multiphase effects to properly represent the liquid water generation and transport within the different cell components. In 4 addition to the basic cell components (BPPs, GDLs, electrodes and membrane), other important components for water management – such as the microporous layer (MPL) or anode and cathode water cooling channels – were also included in the model. A particular effort was made to ensure a suitable mesh generation, where the guidelines derived from previous works [17] have been applied, resulting in a computational mesh with over 10 million elements. The neutron radiographs obtained were also used to validate the CFD models developed in terms of water management. 2. Materials A novel bipolar plate design was proposed, modelled with CFD, manufactured, and tested in a single cell. The flow-field channels of both anode and cathode sides were entirely redesigned. Bipolar plates were manufactured in 0.1 mm thick 316L stainless steel (Figure 1, top), with a 20 nm thickness carbonaceous coating applied by PVD (Physical Vapour Deposition) to avoid corrosion. The resulting Bipolar plate is fulfilling the 2020 DOE targets for Bipolar Plates for automotive applications in terms of corrosion and electrical conductivity considering Interfacial Contact Resistance (ICR). The metallic bipolar plate consists of the anode and cathode plates glued together using reinforced epoxy papers (prepregs), resulting in the anode flow field, cathode flow field, coolant flow field and the respective gas and cooling manifolds. A Gore M820.15 3L-MEA was used in all experiments, with a final 5L-MEA manufactured by assembling GDLs to the framed 3L-MEA, using liquid epoxy adhesives by rolling process. The sealing gaskets were manufactured according to proprietary technology [18] (Figure 1, middle) with special liquid silicone rubber featuring a good compression set. 7L-MEA (with injected gasket) can be easily mounted and centered with the Bipolar Plates also at stack level. The liquid silicone penetrates the GDL ends, reinforcing the resistance of the active area boundaries. 5 Figure 1. Bipolar plate cross-section with 0.1 mm thick stainless steel (top); Through-plane view of MEA over Bipolar Plate featuring flow directions (middle); Anode side of the Bipolar Plate (bottom). A GDL of nominal thickness 245 m and 78% porosity was used, with Micro Porous Layer (MPL), where orthotropic properties (thermal conductivity and electric conductivity) where 6 considered in the GDL. The flow field layout of the cell is crossflow, in a top-to-bottom configuration. The cathode channels feature a vertical parallel flow field design, whereas the anode channel is a horizontal single serpentine. The cooling water flows from top-right towards bottom-left. The main flow configuration is depicted in Figure 2. The active area of the final cell is 140 cm2. Figure 2. Configuration of reactant gases and cooling water: Air (white vertical arrows); Hydrogen (red serpentine arrow); Cooling water (Blue arrow). The CFD modelling and experimental work and results for the new design is presented in the next sections. 3. Experiments and CFD simulations 3.1. Experimental tests The experimental testing carried out by AJUSA Hydrogen Technologies at the cell level revealed that cell performance was improved by more than 50 mV at 0.85 A/cm2 with respect to the previous designs due to the novel bipolar plate designed. 7 The experimental tests were carried out in a Hydrogenics 200 W test bench with a maximum load of 150 A. Measurements were carried out according to the procedures described in IEC TS 62282-7-1:2010 Fuel cell technologies - Part 7-1: Single cell test methods for polymer electrolyte fuel cell (PEFC) [19]. The operating conditions for the polarization curves corresponds to the ones defined by AJUSA Hydrogen Technologies, with constant air stoichiometry, atmospheric pressure at the cathode outlet, and an overpressure at the anode outlet. The nominal conditions for steady state operation are 0.6 V. Electrochemical tests in galvanostatic mode (constant current) were also carried out at Paul Scherrer Institut during the Neutron Radiography experiments. For each current value the cell was maintained in steady state conditions during at least 2 hours. The electrochemical performance data is presented in Table 3 and will be discussed in Section 4. 3.2. Neutron Imaging This technique was applied at the SINQ facility at Paul Scherrer Institut in Villigen, Switzerland, where the NEUTRA beamline [15] was used to determine the liquid water content and distributions for different cell operating conditions representative of the real operation of the cell in the final system. In order to avoid the interference of the cooling water in the neutron radiographs, ‘heavy’ water (D2O) was used in the experiments, as it is transparent to the neutron beam. The matrix of operating conditions is presented in Table 1, where the conditions of each test were maintained during several hours (5-8 hours) in order to ensure that steady state conditions for the water distribution were achieved. In order to resemble the real operation in the final system, an overpressure of 20-25 kPa (relative pressure) was set at the anode outlet. Cathode outlet was operating without overpressure. 8 Table 1. Operating conditions for Neutron Imaging tests Exp 01 Load 80 A Air 3.3 NL/min Pout: 0.0 kPa Tin: 60 ºC Tdew,in: 50 ºC H2 2 NL/min Pout: 20.0 kPa Tin: 26 ºC Tdew,in : 22 ºC Cooling 1 L/min Tout: 60-62 ºC Exp 02 Load 120 A Air 5.0 NL/min Pout: 0.0 kPa Tin: 60 ºC Tdew,in: 50 ºC H2 2 NL/min Pout: 20.0 kPa Tin: 26 ºC Tdew,in : 22 ºC Cooling 1 L/min Tout: 60-62 ºC Exp 03 Load 150 A Air 6.0 NL/min Pout: 0.0 kPa Tin: 60 ºC Tdew,in : 50 ºC H2 2.5 NL/min Pout: 20.0 kPa Tin: 26 ºC Tdew,in : 22 ºC Cooling 1 L/min Tout: 60-62 ºC Exp 04 Load 100 A Air 6.0 NL/min Pout: 0.0 kPa Tin: 60 ºC Tdew,in : 50 ºC H2 2.5 NL/min Pout: 20.0 kPa Tin: 26 ºC Tdew,in : 22 ºC Cooling 1 L/min Tout: 60-62 ºC Exp 05 Load 150 A Air 5.0 NL/min Pout: 0.0 kPa Tin: 50 ºC Tdew,in : 45 ºC H2 2.0 NL/min Pout: 25.0 kPa Tin: 35 ºC Tdew,in : 30 ºC Cooling 0.5 L/min Tout: 40 ºC The last experiment (Exp 05) was intended to enforce water flooding, and thus the dew point of inlet air was maintained 5 °C above the outlet temperature (ensuring a high 15 Porosity - 0.85 / 0.4 Absolute permeability m2 5.7E-12 / 9.7E-14 Catalytic layers Thermal conductivity W/mK 0.3 Electric conductivity 1/ m 280 Porosity - 0.40 Absolute permeability m2 9.7E-14 Membrane Thermal conductivity W/mK 0.13 Electric conductivity 1/ m [23] Equivalent weight kg/kmol 1100 Electrochemical and transport parameters Anode reference exchange current density A/m2 2500 Cathode reference exchange current density A/m2 2.5 Anode concentration exponent - 0.5 Anode charge transfer coefficient - 1 Cathode concentration exponent - 1 Cathode charge transfer coefficient - 1 Diffusivity H2 m2/s 8.0E-05 Diffusivity O2 m2/s 2.0E-05 Diffusivity H2O m2/s 5.0E-05 Pore blockage exponent - 2.5 Simulations were carried out in a HP Z620 workstation, where the cases were run in parallel on 10 CPUs and required 34 GB RAM. In Table 2, the parameters anode reference exchange current density and cathode reference exchange current density were used as model fitting parameters to adjust the cell voltage (in addition to the exponent n as indicated above), following the procedure established in previous works [17]. 4. Results and discussion Figure 4 shows the neutron radiographs obtained for the cell during the experimental campaign. The time-averaged radiographs corresponding to the steady-state operation of the cell are shown. The results corresponding to the five operating conditions shown in 16 Table 1 are depicted, together with the time evolution of the cell voltage during the steadystate operation of the cell. The averaged cell electrochemical data is presented in Table 3. 17 Figure 4. Time-averaged neutron radiographs obtained for the cell (left) where liquid water is marked in blue; time evolution of the cell voltage during steady-state operation (right). The full sequence of experiments is presented from Exp 01 (top) to Exp 05 (bottom). Table 3. Cell performance results obtained during the Neutron Imaging tests Current Current density Voltage Resistance Specific resistance (A) (A/cm2) (V) (mOhm) (Ohm*cm2) Exp 01 80 0.545 0.678 0.354 0.0521 Exp 02 120 0.817 0.630 0.340 0.0500 Exp 03 150 1.021 0.590 0.336 0.0494 Exp 04 100 0.681 0.655 0.353 0.0519 Exp 05 150 1.021 0.553 0.352 0.0518 The images clearly show a qualitatively constant pattern for the water distribution in all experiments, with an accumulation of liquid water in the lower area of the cell (also predicted by the CFD model as described later). The progressive increase of liquid water in the direction of the gas flow (downwards as depicted in Figure 2) is clearly a consequence of the fact that water being electrochemically produced by the cell is being transported towards the gas channels, until the gas flow becomes saturated along its path from inlet to outlet, and thus liquid water begins to accumulate both in GDL and gas channels. It is also observed that a larger water content is present in the left side of the cell, corresponding to the inlet of colder cooling water. The time evolution of the cell voltages (Figure 4, right) are relatively stable except for Exp 05 with cell flooding, where voltage peaks associated to flooding are clearly observed. It is however not feasible to directly correlate the voltage peaks with the water content in the neutron radiographs, as peaks 18 present a duration of  1 second, while the exposure time of each radiograph is 10 seconds as indicated in Section 3.2. The severe flooding of the complete cell active area for Exp 05 is also clearly observed in the corresponding radiograph in Figure 4. It is also worth to note that more water is present in the anode side of the cell rather than in the cathode side, which is visible in Figure 4 in the horizontal channels structure of the anode bipolar plate. A higher water accumulation in the anode side and even anode flooding can be observed in many fuel cell applications, and it is a consequence of several factors such as the significantly lower drag exerted by the anode gas flow and the high humidification used in the anode side for many applications [25-28]. It must be considered that typically a lower Re number is achieved in the anode side compared to the cathode side, thus leading to much lower water removal from the anode flow field [26]. The thinner membranes being used in state-of-the-art MEAs also favours the back-diffusion of water towards the anode [25,29]. In the experiments carried out a higher water accumulation in the anode side of the cell is clearly observed (Figure 4). A water balance within the membrane calculated at the end of this section with the CFD results will be used to further verify the reasons for a higher water saturation in the anode. In addition, a progressive increase in the overall liquid water content is observed from Exp 01 to Exp 03 in Figure 4, as current is being increased progressively from 80 A to 120 A and 150 A respectively. Reducing the intensity while maintaining a higher gases stoichiometry (Exp 04) is contributing to reducing the total liquid water content observed, due to the higher water removal capability of the gases. The CFD simulations carried out for the novel design predicted a significant improvement in terms of cell performance as well as a more uniform temperature distribution within the membrane, which is of paramount importance for increasing durability [3,4]. Liquid water distributions were also obtained as a result of the CFD modelling activities, and are 19 compared against the neutron imaging experimental campaign at Paul Scherrer Institut at the end of this section. Regarding the CFD simulation results, in-plane liquid water saturation (s) is depicted in Figure 5 for different operating conditions (Exp 01 to Exp 03). Figure 5. Liquid water distribution in the anode GDL (left) and cathode GDL (right) in the cell, at 80 A (Exp 01, top), 120 A (Exp 02, middle) and 150 A (Exp 03, bottom). It can be observed that as the intensity increases, the liquid water content increases reaching maximum values in the lower part of the cell (due to the fact that gases become saturated with moisture as they advance along the flow paths towards the outlet) [12]. In a side-view of the cell, a larger amount of liquid water is observed under the ribs than under the channels on both the anode and cathode sides (Figure 6), which is a well-known feature caused by the fact that lands are the coldest GDL locations due to the effect of the water cooling [30]. 20 Figure 6. Liquid water distribution in the anode (top) and cathode (bottom) “through-plane” in the cell (Exp 02; 120 A). The simulation results also show that the temperature distribution in the membrane is improved, obtaining lower gradients and a lower maximum temperature, achieving a maximum temperature of 73 ºC (at 120 A) compared to 90 ºC (at 100 A) of previous designs. Figure 7 presents the temperature, current density, and water content () in-plane distributions within the membrane. A clear relationship can be observed between the membrane water content distribution (Figure 7, bottom) and the anode and cathode GDL liquid water distribution depicted in Figure 5, with increasing amount of water content in the direction of the gas flows (downwards). The membrane is relatively well hydrated, with  values in the range 7-14, particularly in the lower section of the cell where water and vapour is becoming accumulated as the cathode gas flow progresses. 21 Figure 7. Temperature distribution in the mid-plane of the membrane in Exp 02 at 120 A (top), current density distribution (middle) and membrane water content  (bottom). The amount of water (in grams) calculated both from the experimental images and from the results of the CFD model are shown in Table 4. Equation (1) was used to calculate the experimental water thickness w in the neutron radiographs, and the area average of w was converted to water weight by multiplying it by the cell active area and by the water density. The liquid water results corresponding to the CFD simulation have been provided in Table 4 as a total in grams and as the breakdown contributions among the different cell components (anode channels, GDLs/MPLs/CLs, and cathode channels). The calculation was carried out by integrating the liquid water saturation results (variable “s” in Eq. 4 and 22 Eq. 5) in each of the volume domains corresponding to the cell components of interest. The volume-integral of the liquid saturation “s” is providing the total volume of liquid water in each cell component, and the weight of water is derived from this quantity by multiplying this by water density. Table 4. Total liquid water accumulation in the cell and cell voltage (Neutron Imaging tests and CFD model). Values corresponding to Neutron Imaging tests are time-averaged. Water NI Water CFD Voltage NI Voltage CFD (g) (g) (an ch / GDL/MPL/CL / cat ch) (V) (V) Exp 01 0.24 0.21 (0.04 / 0.16 / 0.01) 0.678 0.626 Exp 02 0.72 0.31 (0.11 / 0.19 / 0.01) 0.630 0.607 Exp 03 0.88 1.01 (0.79 / 0.21 / 0.01) 0.590 0.581 It is observed that the overall trends are reproduced correctly, although the quantitative comparison presents deviations especially at 120 A (Exp 02). It is worth noting that the CFD model allows to gain further insight into the breakdown of the liquid water content for each component, where it is observed that a higher water content in anode channels than in cathode channels, as could also be experimentally observed in the neutron radiographs (Figure 4, where the horizontal channels of the anode bipolar plate are containing a higher amount of water than the vertical cathode channels). Although all the overall trends are correctly reproduced by the model, a quantitative comparison is however revealing significant differences in the amount of water accumulated within the cell: 12.5% error at 80 A (Exp 01), 57% at 120 A (Exp 02) and 15% error at 150 A (Exp 03), where errors have been calculated as 100·(grams of water CFD - grams of water NI) / grams of water NI. 23 In order to verify the distributions of liquid water and the degree of similarity between CFD model and experimental results, a visual comparison is presented in Figure 8. Figure 8. Liquid water distribution in the cell at 80 A (Exp 01, top), 120 A (Exp 02, middle) and 150 A (Exp 03, bottom), for results obtained from CFD model (left) and time-averaged Neutron Imaging tests (right). It can be observed that the progressive water accumulation in the vertical direction (where both anode and cathode flows are flowing downwards) is reasonably well predicted by the CFD model. However, the horizontal water distribution observed in the experimental results, caused by the temperature gradient of the cooling water (flowing from left to right 24 in Figure 8) is not predicted by the CFD model. This may be caused by a non-accurate prediction of the condensation rates or by differences in the temperature field within the CFD results with respect to the real temperature field, which can be a consequence of different issues such as the values of the non-isotropic thermal conductivities. Table 4 presents also the voltages obtained experimentally during the experiments (as an average of the time-evolution presented in Figure 4) and by the CFD model. It can be observed that the model is predicting the cell voltage reasonably well, although quantitative values are presenting differences: 7.5% error at 80 A (Exp 01), 3.5% at 120 A (Exp 02) and 1.5% error at 150 A (Exp 03), where errors have been calculated as 100·(VCFD - VNI) / VNI. Either electrochemical parameters in the Butler-Volmer equations (Table 2) or the value defined in the CFD model for the BP-GDL interfacial contact resistance (3.07E-7 Ohm*m2) may be the reason for the differences obtained. Obviously, potential modelling errors such as two-phase flow treatment or mesh resolution may be also having an influence and all this will need to be further investigated for achieving an improved accuracy of the modelling results. It is widely acknowledged that multi-phase flow and water management is one of the most challenging issues when modelling low-temperature PEM fuel cells [31, 32]. For this, model validation is unequivocally an essential requirement for the development and confident use of CFD models. A validation procedure based only on the polarization IV curves is usually not fully appropriate as stated by Wang [31], as polarization curves are an integral output of the cell so that the individual contributions of each physicalelectrochemical phenomena cannot be properly validated. It is thus necessary to perform a validation procedure using local quantities when possible in order to gain a full confidence on the models developed [13]. The comparison of model results against neutron imaging experiments can be further used in future to progressively refine model developments [31, 33-35]. 31 [34] Wang, Y., Chen, K.S. Through-plane water distribution in a polymer electrolyte fuel cell: comparison of numerical prediction with neutron radiography data. Journal of the Electrochemical Society, 2010; 57 (12), pp. B1878-B1886 [35] Quan, P., Lai, M.-C. PEM fuel cell water transport - simulation and neutron imaging experiment. ECS Transactions, 2010; 26 (1), pp. 179-196 1 Bipolar Plate research using Computational Fluid Dynamics and Neutron Radiography for Proton Exchange Membrane Fuel Cells Alfredo Iranzo a *, José Manuel Gregorio b, Pierre Boillat c,d, Felipe Rosa a a Thermal Engineering Group, Energy Engineering Department, School of Engineering, Universidad de Sevilla. Camino de los Descubrimientos, s/n, 41092, Sevilla, Spain b AJUSA Hydrogen Technologies, Parque Empresarial Ajusa, Calle 1, nº1. 02006 Albacete, Spain c Electrochemistry Laboratory (LEC), Paul Scherrer Institut (PSI), 5232 Villigen PSI, Switzerland d Laboratory for Neutron Scattering and Imaging (LNS), Paul Scherrer Institut (PSI), 5232 Villigen PSI, Switzerland * Corresponding author: airan[email protected] Tel.: +34-954-487471, Fax: +34-954-487247 Abstract This work presents the development of liquid-cooled industry-scale bipolar plates for improved water management in PEM Fuel Cells. The methods used for the design development are based on Computational Fluid Dynamics (CFD) modelling and simulation, and Neutron Radiography experiments to analyse liquid water distributions within the cell for different operating conditions. A novel 140 cm2 bipolar plate was designed and manufactured on 0.1 mm thick stainless steel using pre-coated strip steel. CFD modelling carried out for the novel design predicted a significant improvement in terms of cell performance, as well as a more uniform *Manuscript (Changes Marked on Previous Version) 2 temperature distribution within the membrane. Liquid water distributions were later analysed by neutron radiography experiments, defining a set of different operating conditions (current density, stoichiometry, inlet gases dew point, and cell temperature). Electrochemical and neutron radiography results are presented for all cases and the influence of the operating conditions is discussed. Liquid water distributions within the cell are also analysed and compared against the CFD model results obtained. The influence of the gas flow configuration (reactant gases and cooling water) is clearly observable in the results. Keywords: Bipolar Plate; PEM Fuel Cell; CFD; numerical model; Neutron Radiography; water management 1. Introduction The development of Bipolar Plates (BPPs) allocates major research efforts in the field of PEM fuel cells, as BPP design is having a direct influence in the final performance of the stack [1,2] and also on durability [3,4]. The combined effect of both Bipolar Plate and Gas Diffusion Layer (GDL) must ensure a uniform distribution of reactants over the electrodes as well as an efficient liquid water removal from the GDL pores and BPP channels, so that reactants mass transport is not hindered by liquid water blockages. Development of novel flow fields for BPPs aims at avoiding mass transport limitations, so that the stack can safely and stably operate at higher current densities, which indeed results in an increase in the stack power density, thus reducing the final cost and enabling an improved and more compact system integration. Different flow field design layouts for BPPs have been proposed aiming at enhancing the final performance of PEM fuel cells, where parallel, serpentine, interdigitated or pin-type 3 layouts are the most commonly found [2,5]. The material of the BPP is also playing a fundamental role in the performance, power density, and durability due to eventual BPP corrosion [6,7]. While graphite plates of a few mm thickness can be used for stationary applications, metallic-coated plates in the range of 0.1-0.5 mm are used in industry for transport applications [8,9], given the higher power density required and the suitability for mass production by well-known processes such as stamping or hydroforming [10]. This work presents the developments carried out by AJUSA Hydrogen Technologies and University of Sevilla aimed at developing novel flow field designs of bipolar plates with improved water management for the commercial PEM fuel cell systems manufactured by AJUSA Hydrogen Technologies [11]. The scientific approach followed was based on a combination of both experimental and modelling techniques. Regarding experimental techniques, not only conventional test bench polarization curves and durability tests were carried out, but also in-operando visualization of the liquid water content and distributions by means of Neutron Imaging [12-14]. This technique was applied at the SINQ facility at Paul Scherrer Institut in Villigen, Switzerland. The NEUTRA beamline [15] was used to determine the liquid water content and distributions for different cell operating conditions representative of the real operation of the cell in the final system. Neutron radiographs of the in-operando fuel cell under different conditions and flow configurations provided an outstanding understanding of the liquid water distributions and their influence on the final cell performance. Regarding modelling and simulation, a Computational Fluid Dynamics (CFD) model was developed within the commercial software ANSYS-FLUENT [16], where the model included the cell electrochemistry, heat transfer, and multiphase effects to properly represent the liquid water generation and transport within the different cell components. In 4 addition to the basic cell components (BPPs, GDLs, electrodes and membrane), other important components for water management – such as the microporous layer (MPL) or anode and cathode water cooling channels – were also included in the model. A particular effort was made to ensure a suitable mesh generation, where the guidelines derived from previous works [17] have been applied, resulting in a computational mesh with over 10 million elements. The neutron radiographs obtained were also used to validate the CFD models developed in terms of water management. 2. Materials A novel bipolar plate design was proposed, modelled with CFD, manufactured, and tested in a single cell. The flow-field channels of both anode and cathode sides were entirely redesigned. Bipolar plates were manufactured in 0.1 mm thick 316L stainless steel (Figure 1, top), with a 20 nm thickness carbonaceous coating applied by PVD (Physical Vapour Deposition) to avoid corrosion. The resulting Bipolar plate is fulfilling the 2020 DOE targets for Bipolar Plates for automotive applications in terms of corrosion and electrical conductivity considering Interfacial Contact Resistance (ICR). The metallic bipolar plate consists of the anode and cathode plates glued together using reinforced epoxy papers (prepregs), resulting in the anode flow field, cathode flow field, coolant flow field and the respective gas and cooling manifolds. A Gore M820.15 3L-MEA was used in all experiments, with a final 5L-MEA manufactured by assembling GDLs to the framed 3L-MEA, using liquid epoxy adhesives by rolling process. The sealing gaskets were manufactured according to proprietary technology [18] (Figure 1, middle) with special liquid silicone rubber featuring a good compression set. 7L-MEA (with injected gasket) can be easily mounted and centered with the Bipolar Plates also at stack level. The liquid silicone penetrates the GDL ends, reinforcing the resistance of the active area boundaries. 5 Figure 1. Bipolar plate cross-section with 0.1 mm thick stainless steel (top); Through-plane view of MEA over Bipolar Plate featuring flow directions (middle); Anode side of the Bipolar Plate (bottom). A GDL of nominal thickness 245 m and 78% porosity was used, with Micro Porous Layer (MPL), where orthotropic properties (thermal conductivity and electric conductivity) where 6 considered in the GDL. The flow field layout of the cell is crossflow, in a top-to-bottom configuration. The cathode channels feature a vertical parallel flow field design, whereas the anode channel is a horizontal single serpentine. The cooling water flows from top-right towards bottom-left. The main flow configuration is depicted in Figure 2. The active area of the final cell is 140 cm2. Figure 2. Configuration of reactant gases and cooling water: Air (white vertical arrows); Hydrogen (red serpentine arrow); Cooling water (Blue arrow). The CFD modelling and experimental work and results for the new design is presented in the next sections. 3. Experiments and CFD simulations 3.1. Experimental tests The experimental testing carried out by AJUSA Hydrogen Technologies at the cell level revealed that cell performance was improved by more than 50 mV at 0.85 A/cm2 with respect to the previous designs due to the novel bipolar plate designed. 7 The experimental tests were carried out in a Hydrogenics 200 W test bench with a maximum load of 150 A. Measurements were carried out according to the procedures described in IEC TS 62282-7-1:2010 Fuel cell technologies - Part 7-1: Single cell test methods for polymer electrolyte fuel cell (PEFC) [19]. The operating conditions for the polarization curves corresponds to the ones defined by AJUSA Hydrogen Technologies, with constant air stoichiometry, atmospheric pressure at the cathode outlet, and an overpressure at the anode outlet. The nominal conditions for steady state operation are 0.6 V. Electrochemical tests in galvanostatic mode (constant current) were also carried out at Paul Scherrer Institut during the Neutron Radiography experiments. For each current value the cell was maintained in steady state conditions during at least 2 hours. The electrochemical performance data is presented in Table 3 and will be discussed in Section 4. 3.2. Neutron Imaging This technique was applied at the SINQ facility at Paul Scherrer Institut in Villigen, Switzerland, where the NEUTRA beamline [15] was used to determine the liquid water content and distributions for different cell operating conditions representative of the real operation of the cell in the final system. In order to avoid the interference of the cooling water in the neutron radiographs, ‘heavy’ water (D2O) was used in the experiments, as it is transparent to the neutron beam. The matrix of operating conditions is presented in Table 1, where the conditions of each test were maintained during several hours (5-8 hours) in order to ensure that steady state conditions for the water distribution were achieved. In order to resemble the real operation in the final system, an overpressure of 20-25 kPa (relative pressure) was set at the anode outlet. Cathode outlet was operating without overpressure. 8 Table 1. Operating conditions for Neutron Imaging tests Exp 01 Load 80 A Air 3.3 NL/min Pout: 0.0 kPa Tin: 60 ºC Tdew,in: 50 ºC H2 2 NL/min Pout: 20.0 kPa Tin: 26 ºC Tdew,in : 22 ºC Cooling 1 L/min Tout: 60-62 ºC Exp 02 Load 120 A Air 5.0 NL/min Pout: 0.0 kPa Tin: 60 ºC Tdew,in: 50 ºC H2 2 NL/min Pout: 20.0 kPa Tin: 26 ºC Tdew,in : 22 ºC Cooling 1 L/min Tout: 60-62 ºC Exp 03 Load 150 A Air 6.0 NL/min Pout: 0.0 kPa Tin: 60 ºC Tdew,in : 50 ºC H2 2.5 NL/min Pout: 20.0 kPa Tin: 26 ºC Tdew,in : 22 ºC Cooling 1 L/min Tout: 60-62 ºC Exp 04 Load 100 A Air 6.0 NL/min Pout: 0.0 kPa Tin: 60 ºC Tdew,in : 50 ºC H2 2.5 NL/min Pout: 20.0 kPa Tin: 26 ºC Tdew,in : 22 ºC Cooling 1 L/min Tout: 60-62 ºC Exp 05 Load 150 A Air 5.0 NL/min Pout: 0.0 kPa Tin: 50 ºC Tdew,in : 45 ºC H2 2.0 NL/min Pout: 25.0 kPa Tin: 35 ºC Tdew,in : 30 ºC Cooling 0.5 L/min Tout: 40 ºC The last experiment (Exp 05) was intended to enforce water flooding, and thus the dew point of inlet air was maintained 5 °C above the outlet temperature (ensuring a high 9 condensation). This was aimed at verifying the capabilities of the cell to withstand adverse flooding conditions. In this experiment the cell temperature was lowered from the nominal conditions (60–62°C) down to 40°C while increasing both the anode humidification and the current (above 1 A/cm2), in order to promote flooding. During all experiments, neutron imaging was conducted continuously, allowing the observation of water accumulation or removal until reaching a stable steady-state condition. The detector used consisted of a neutron sensitive scintillator screen converting the neutron image to a visible light image, which was subsequently captured by a CCD camera (Andor iKon-L, 2048x2048 pixels). The main parameters of the imaging setup are pixel size 143 m, effective resolution 300 m, and 10 s exposure time. After image correction steps (noise filtering, removal of the camera background, correction for the beam intensity fluctuations, correction for the spatial inhomogeneities of the beam, correction for the cell displacements, and removal of the contribution from the scattered neutrons), the images of the cell under operation were divided pixel-wise by the image of the cell in a reference condition without liquid water, obtained by flowing dry gases until all liquid water disappears. By doing this, only the attenuation due to liquid water remains in the images. Such images contain the neutron transmission for each pixel, which can be converted into a water thickness w by the inverted Lambert-Beer formula as described in previous works [13, 14]: (1) Where (I/I0) is the relative neutron transmission in the neutron radiographs obtained, and  is the attenuation coefficient of neutrons in liquid water, with a value of 3.5 cm-1 for the given setup at NEUTRA beamline [13-15]. The calculated liquid water thickness w can be 16 Table 1 are depicted, together with the time evolution of the cell voltage during the steadystate operation of the cell. The averaged cell electrochemical data is presented in Table 3. 17 Figure 4. Time-averaged neutron radiographs obtained for the cell (left) where liquid water is marked in blue; time evolution of the cell voltage during steady-state operation (right). The full sequence of experiments is presented from Exp 01 (top) to Exp 05 (bottom). Table 3. Cell performance results obtained during the Neutron Imaging tests Current Current density Voltage Resistance Specific resistance (A) (A/cm2) (V) (mOhm) (Ohm*cm2) Exp 01 80 0.545 0.678 0.354 0.0521 Exp 02 120 0.817 0.630 0.340 0.0500 Exp 03 150 1.021 0.590 0.336 0.0494 Exp 04 100 0.681 0.655 0.353 0.0519 Exp 05 150 1.021 0.553 0.352 0.0518 The images clearly show a qualitatively constant pattern for the water distribution in all experiments, with an accumulation of liquid water in the lower area of the cell (also predicted by the CFD model as described later). The progressive increase of liquid water in the direction of the gas flow (downwards as depicted in Figure 2) is clearly a consequence of the fact that water being electrochemically produced by the cell is being transported towards the gas channels, until the gas flow becomes saturated along its path from inlet to outlet, and thus liquid water begins to accumulate both in GDL and gas channels. It is also observed that a larger water content is present in the left side of the cell, corresponding to the inlet of colder cooling water. The time evolution of the cell voltages (Figure 4, right) are relatively stable except for Exp 05 with cell flooding, where voltage peaks associated to flooding are clearly observed. It is however not feasible to directly correlate the voltage peaks with the water content in the neutron radiographs, as peaks 18 present a duration of  1 second, while the exposure time of each radiograph is 10 seconds as indicated in Section 3.2. The severe flooding of the complete cell active area for Exp 05 is also clearly observed in the corresponding radiograph in Figure 4. It is also worth to note that more water is present in the anode side of the cell rather than in the cathode side, which is visible in Figure 4 in the horizontal channels structure of the anode bipolar plate. A higher water accumulation in the anode side and even anode flooding can be observed in many fuel cell applications, and it is a consequence of several factors such as the significantly lower drag exerted by the anode gas flow and the high humidification used in the anode side for many applications [25-28]. It must be considered that typically a lower Re number is achieved in the anode side compared to the cathode side, thus leading to much lower water removal from the anode flow field [26]. The thinner membranes being used in state-of-the-art MEAs also favours the back-diffusion of water towards the anode [25,29]. In the experiments carried out a higher water accumulation in the anode side of the cell is clearly observed (Figure 4). A water balance within the membrane calculated at the end of this section with the CFD results will be used to further verify the reasons for a higher water saturation in the anode. In addition, a progressive increase in the overall liquid water content is observed from Exp 01 to Exp 03 in Figure 4, as current is being increased progressively from 80 A to 120 A and 150 A respectively. Reducing the intensity while maintaining a higher gases stoichiometry (Exp 04) is contributing to reducing the total liquid water content observed, due to the higher water removal capability of the gases. The CFD simulations carried out for the novel design predicted a significant improvement in terms of cell performance as well as a more uniform temperature distribution within the membrane, which is of paramount importance for increasing durability [3,4]. Liquid water distributions were also obtained as a result of the CFD modelling activities, and are 19 compared against the neutron imaging experimental campaign at Paul Scherrer Institut at the end of this section. Regarding the CFD simulation results, in-plane liquid water saturation (s) is depicted in Figure 5 for different operating conditions (Exp 01 to Exp 03). Figure 5. Liquid water distribution in the anode GDL (left) and cathode GDL (right) in the cell, at 80 A (Exp 01, top), 120 A (Exp 02, middle) and 150 A (Exp 03, bottom). It can be observed that as the intensity increases, the liquid water content increases reaching maximum values in the lower part of the cell (due to the fact that gases become saturated with moisture as they advance along the flow paths towards the outlet) [12]. In a side-view of the cell, a larger amount of liquid water is observed under the ribs than under the channels on both the anode and cathode sides (Figure 6), which is a well-known feature caused by the fact that lands are the coldest GDL locations due to the effect of the water cooling [30]. 20 Figure 6. Liquid water distribution in the anode (top) and cathode (bottom) “through-plane” in the cell (Exp 02; 120 A). The simulation results also show that the temperature distribution in the membrane is improved, obtaining lower gradients and a lower maximum temperature, achieving a maximum temperature of 73 ºC (at 120 A) compared to 90 ºC (at 100 A) of previous designs. Figure 7 presents the temperature, current density, and water content () in-plane distributions within the membrane. A clear relationship can be observed between the membrane water content distribution (Figure 7, bottom) and the anode and cathode GDL liquid water distribution depicted in Figure 5, with increasing amount of water content in the direction of the gas flows (downwards). The membrane is relatively well hydrated, with  values in the range 7-14, particularly in the lower section of the cell where water and vapour is becoming accumulated as the cathode gas flow progresses. 21 Figure 7. Temperature distribution in the mid-plane of the membrane in Exp 02 at 120 A (top), current density distribution (middle) and membrane water content  (bottom). The amount of water (in grams) calculated both from the experimental images and from the results of the CFD model are shown in Table 4. Equation (1) was used to calculate the experimental water thickness w in the neutron radiographs, and the area average of w was converted to water weight by multiplying it by the cell active area and by the water density. The liquid water results corresponding to the CFD simulation have been provided in Table 4 as a total in grams and as the breakdown contributions among the different cell components (anode channels, GDLs/MPLs/CLs, and cathode channels). The calculation was carried out by integrating the liquid water saturation results (variable “s” in Eq. 4 and 22 Eq. 5) in each of the volume domains corresponding to the cell components of interest. The volume-integral of the liquid saturation “s” is providing the total volume of liquid water in each cell component, and the weight of water is derived from this quantity by multiplying this by water density. Table 4. Total liquid water accumulation in the cell and cell voltage (Neutron Imaging tests and CFD model). Values corresponding to Neutron Imaging tests are time-averaged. Water NI Water CFD Voltage NI Voltage CFD (g) (g) (an ch / GDL/MPL/CL / cat ch) (V) (V) Exp 01 0.24 0.21 (0.04 / 0.16 / 0.01) 0.678 0.626 Exp 02 0.72 0.31 (0.11 / 0.19 / 0.01) 0.630 0.607 Exp 03 0.88 1.01 (0.79 / 0.21 / 0.01) 0.590 0.581 It is observed that the overall trends are reproduced correctly, although the quantitative comparison presents deviations especially at 120 A (Exp 02). It is worth noting that the CFD model allows to gain further insight into the breakdown of the liquid water content for each component, where it is observed that a higher water content in anode channels than in cathode channels, as could also be experimentally observed in the neutron radiographs (Figure 4, where the horizontal channels of the anode bipolar plate are containing a higher amount of water than the vertical cathode channels). Although all the overall trends are correctly reproduced by the model, a quantitative comparison is however revealing significant differences in the amount of water accumulated within the cell: 12.5% error at 80 A (Exp 01), 57% at 120 A (Exp 02) and 15% error at 150 A (Exp 03), where errors have been calculated as 100·(grams of water CFD - grams of water NI) / grams of water NI. 23 In order to verify the distributions of liquid water and the degree of similarity between CFD model and experimental results, a visual comparison is presented in Figure 8. Figure 8. Liquid water distribution in the cell at 80 A (Exp 01, top), 120 A (Exp 02, middle) and 150 A (Exp 03, bottom), for results obtained from CFD model (left) and time-averaged Neutron Imaging tests (right). It can be observed that the progressive water accumulation in the vertical direction (where both anode and cathode flows are flowing downwards) is reasonably well predicted by the CFD model. However, the horizontal water distribution observed in the experimental results, caused by the temperature gradient of the cooling water (flowing from left to right 24 in Figure 8) is not predicted by the CFD model. This may be caused by a non-accurate prediction of the condensation rates or by differences in the temperature field within the CFD results with respect to the real temperature field, which can be a consequence of different issues such as the values of the non-isotropic thermal conductivities. Table 4 presents also the voltages obtained experimentally during the experiments (as an average of the time-evolution presented in Figure 4) and by the CFD model. It can be observed that the model is predicting the cell voltage reasonably well, although quantitative values are presenting differences: 7.5% error at 80 A (Exp 01), 3.5% at 120 A (Exp 02) and 1.5% error at 150 A (Exp 03), where errors have been calculated as 100·(VCFD - VNI) / VNI. Either electrochemical parameters in the Butler-Volmer equations (Table 2) or the value defined in the CFD model for the BP-GDL interfacial contact resistance (3.07E-7 Ohm*m2) may be the reason for the differences obtained. Obviously, potential modelling errors such as two-phase flow treatment or mesh resolution may be also having an influence and all this will need to be further investigated for achieving an improved accuracy of the modelling results. It is widely acknowledged that multi-phase flow and water management is one of the most challenging issues when modelling low-temperature PEM fuel cells [31, 32]. For this, model validation is unequivocally an essential requirement for the development and confident use of CFD models. A validation procedure based only on the polarization IV curves is usually not fully appropriate as stated by Wang [31], as polarization curves are an integral output of the cell so that the individual contributions of each physicalelectrochemical phenomena cannot be properly validated. It is thus necessary to perform a validation procedure using local quantities when possible in order to gain a full confidence on the models developed [13]. The comparison of model results against neutron imaging experiments can be further used in future to progressively refine model developments [31, 33-35]. 25 In order to analyse the higher anode accumulation, the results presented on Table 5 for the water balance within the membrane has been obtained from the CFD results. Table 5. Water balance in the membrane obtained from the CFD results. Exp 01 80 A Exp 02 120 A Exp 03 150 A EOD (g/s) 2.58E-02 4.12E-02 5.47E-02 BD (g/s) 3.95E-02 6.08E-02 8.17E-02 anode average 7.24 8.62 8.54 cathode average 8.54 10.55 11.12 It is known that water transport within the membrane is mainly caused by the electroosmotic drag (EOD, transporting water from anode to cathode proportionally to the cell current) [29] and back-diffusion (BD, transporting water by diffusion from cathode to anode proportionally to the water content gradient) [29]. Results in Table 5 shows that transport by back-diffusion is in all cases higher than electro-osmotic drag, thus causing a net water transport towards the anode side and promoting there a higher water accumulation. It is observed that electro-osmotic drag is progressively increasing with the current, but also back-diffusion is higher as the water content gradient also increases with the current according to the results presented in Table 5. 5. Conclusions Professor T. Nejat Veziroglu Dr. Alfredo Iranzo Miami, Florida, United States University of Sevilla Founding Editor-in-Chief Thermal Energy Engineering Group International Association for Hydrogen Energy Camino Descubrimientos s/n 41092 Sevilla, SPAIN Reference: Declaration of interest Dear Prof. Veziroglu, we are glad to submit the manuscript: “Bipolar Plate research using CFD and Neutron Radiography for Proton Exchange Membrane Fuel Cells”, to the International Journal of Hydrogen Energy. Declaration of interest: All authors declare no conflicts of interests. 'Declarations of interest: none'. Please do not hesitate to contact us for any question regarding this article. Thank you very much in advance. Alfredo Iranzo, PhD (corresponding author) *Declaration of Interest Statement