Thermal neutron radiography of a passive proton exchange membrane fuel cell for portable hydrogen energy systems

Journal of Power Sources 480 (2020) 228668 Available online 6 September 2020 0378-7753/© 2020 Elsevier B.V. All rights reserved. Perspective Thermal neutron radiography of a passive proton exchange membrane fuel cell for portable hydrogen energy systems Antonio M. Chaparro a , * , P. Ferreira-Aparicio a , M. Antonia Folgado a , Rico Hübscher b , Carsten Lange b , Norbert Weber c a Dep. of Energy, CIEMAT, Avda. Complutense, 40, 28040 Madrid, Spain b TU Dresden, Faculty of Mechanical Science and Engineering, Institute of Power Engineering, 01062 Dresden, Germany c Institut für Fluiddynamik. Helmholtz-Zentrum Dresden-Rossendorf e.V., Bautzner Landstr. 400, 01328 Dresden, Germany HIGHLIGHTS GRAPHICAL ABSTRACT •New passive air-breathing PEMFC for hydrogen portable applications tested. •In-operando water accumulation in the cell followed with thermal neutron imaging. •Liquid water profiles in cross section dictate the cell response. •Cell in vertical position improves the steady-state power by 15%. •Horizontal position with cathode on top causes anode flooding and cell failure. ARTICLE INFO Keywords: PEMFC Hydrogen Neutron radiography Water transport Portable fuel cell ABSTRACT A proton exchange membrane fuel cell (PEMFC) for portable applications is studied with thermal neutron radiography. The PEMFC operates under passive conditions, with air-breathing cathode and dead-end anode supplied with static ambient air and dry hydrogen, respectively. A columnar cathodic plate favors the mobility of water drops over the cathode surface and their elimination. Thermal neutron images show liquid water build-up during operation of the cell in vertical and horizontal position, i.e. aligned parallel and perpendicular to the gravity field, respectively. Polarization curves and impedance spectroscopy show orientation dependent cell response that can be related with the water profiles. In vertical position, the elimination of drops sliding over the cathode surface as well as natural convection favor lower water contents in the cathode and improve oxygen transport. The vertical cell can be operated for hours in ambient conditions, providing steady power densities above 100 mW cm −2 . In horizontal position, natural forces are less effective for water removal leading to 17% decrease in peak power density. The horizontal position is especially adverse if the upper electrode is the cathode, because anode flooding causes cell failure after production of a small amount of water (5 mg cm −2 ). * Corresponding author. Dep. of Energy, CIEMAT, Avda. Complutense 40, 28040, Madrid, Spain. E-mail address: [email protected] (A.M. Chaparro). Contents lists available at ScienceDirect Journal of Power Sources journal homepage: www.elsevier.com/locate/jpowsour https://doi.org/10.1016/j.jpowsour.2020.228668 Received 2 June 2020; Received in revised form 13 July 2020; Accepted 14 July 2020 Journal of Power Sources 480 (2020) 228668 2 1. Introduction Over the last two decades, the development of portable hydrogen fuel cell systems has attained sufficient maturity for low power and portable applications which benefit from their high power density and safety [1–7]. The particular operation conditions of portable devices dictate changes in the design of the cells targeting autonomy, compactness, and light-weight power production requirements. In addition, they may have to operate very close to the user, which should not compromise personal comfort and safety, e.g. by emission of heat, liquid water, moving parts, or noise [8]. A passive air-breathing cathode is most suitable for a portable proton-exchange membrane fuel cell (PEMFC) working with ambient air and passive rejection of water [9–12]. No applied convective force is required for oxygen and water transport: only natural forces originating from concentration and heat gradients, water evaporation-condensation, and gravity drag of water drops ensure efficient mass transfer. Portability of a PEMFC is improved if using a passive anode in front of the air-breathing cathode. This special anode operates in dead-end mode in a static atmosphere of dry hydrogen without periodical purging, which improves the efficiency and autonomy of the power system. For continuous passive operation, the flooding and drying-out of cathode and anode must be prevented, which depend on a delicate equilibrium between different operation parameters [13,14]. Optimizing passive water rejection means incorporating components such as superhydrophobic layers that repeal the liquid water from porous electrodes [15], or a hydrophilic membrane to allow water permeation from the anode [16], which considerably improve the water management and operation of the passive PEMFC. A technique that provides most useful information of water dynamics in a PEMFC is neutron radiography. Neutrons have a high sensitivity to liquid water inside porous structures and larger cavities. High lateral and temporal resolutions can be attained, with pixel sizes as low as 30 μ m and image acquisition times of 5 s. In-operando studies have been carried out with passive air-breathing PEMFC [10,17] and conventional cells [18–23]. In the latter, water accumulation is observed within the channels of the flow-field plates for different designs and operation conditions. Siegel et al. [18] showed water accumulation and voltage drop in the anode and cathode of a cell fed with dry hydrogen and periodical purging. At high current densities, neutron images show intense water accumulation under the ribs of the flow-field plates leading to transport losses parallel and perpendicular to the cell plane depending on the cell compression [19,22]. The influence of the gas flow configuration in a water cooled commercial stack was studied by Iranzo et al. by combining neutron radiography with localized temperature and current density measurements as well as modeling [23]. Neutron imaging has also been used for the observation of water inside the gas diffusion layer by the dark field technique with a grating interferometer [24]. Only few studies with neutron imaging have been carried out on passive air-breathing cells. Weiland et al. studied cells changing the configuration of the cathodic plate, particularly the number and size of openings [10]. They found accumulation of the water produced by the cell only in the cathode, but not in the anode, and attributed voltage losses to the obstruction of the openings. Coz et al. [17] studied water management in a planar air-breathing stack fabricated on printed circuit board, and delivering up to 150 mW cmcm −2 ; their cell design showed heterogeneous water accumulation in the anode and cathode leading to 50% loss in performance after 10 h of operation. In addition to neutron radiography, other techniques used for in-operando liquid water studies of PEMFCs are X-ray radiography [25–30], droplet microvisualization inside optically transparent cells [31], and nuclear magnetic resonance [32], some of them requiring important changes in cell setup or configuration that may alter the operation characteristics [33]. In this work, thermal neutron radiography is used for the inoperando study of a fully passive PEMFC for portable applications. Thermal neutrons, produced by nuclear fission and a moderator, are used for the first time with this research objective. The PEMFC design is based on the concept depicted in Fig. 1, characterized by a passive anode and cathode able to operate under static hydrogen and air atmospheres [34,35]. Neutron imaging experiments have been performed keeping the original design almost unmodified, except for the use of aluminum anodic plates, which are almost transparent to thermal neutrons, instead of plastic plates. Cells are studied in vertical and horizontal orientation to observe how natural forces influence passive water transport and performance. Results are complemented with polarization curves and impedance spectroscopy. 2. Experimental 2.1. Passive PEMFC description Characteristics of the passive PEMFC design are shown in Fig. 1a, and the scheme of operation in Fig. 1b. The cell has a circular design with an active area of 7.1 cm 2 and optimized anode and cathode architectures for passive operation. The circular design favors homogeneous cell response and hydrogen gas tightness which are particularly important for the passive operation [35]. The anodic plate is a perforated disk made of aluminum alloy (SIMAGALTOK 82 EN AW 6082, 5 mm thickness, 5 cm diameter) that allocates a gas diffusion electrode (GDE) (ELAT GDE LT250EWALTSI, BASF, 0.25 mg Pt cm −2 ) and an Au covered Ni grid (Dexmet) current collector. For gas tightness, the collector grid is framed with a 3D printed PLA and sealed with silicone gaskets. A hydrophilic membrane (Nafion 117, Ion Power Inc.) is placed at the back of the perforated anodic plate to improve passive liquid water removal from the anode (Fig. 2b). The cathode is open to air, with a columnar plate (adapted from an aluminum pin heatsink, Fischer Elektronik), a grid current collector (Au covered Ni grid from Dexmet) and a gas diffusion electrode (W1S1009, FuelCellsEtc. 0.30 mg Pt cm −2 ). Using somewhat larger Pt loading in the cathode than in anode assures larger hydrophilicity of the cathode which we have found beneficial for the water management in a passive air-breathing cell. Finally, a Nafion 212NR membrane (Ion Power Inc.) is used as electrolyte. All cell elements, except the cathodic plate, are fixed with eight stainless steel bolts (M3) and nuts at 1 N m torque; the cathodic plate is tightened independently to the cell body with another set of nuts at 0.5 N m torque. 2.2. Electrochemical characterization The cell response was studied with polarization curves and impedance spectroscopy in three cell orientations: vertical, horizontal with cathode up, and horizontal with cathode down. The cells were operated inside a climatic chamber (Dycometal), and supplied with static hydrogen (Air Liquide, 99.999%, 0.5 bar g ) in the anode, and static ambient air at 22 ◦C and 30% RH, in the cathode. Polarization curves were obtained potentiostatically (Autolab 30 N with 10 A current booster), by stepping the cell voltage from open circuit to 0.3 V at 10 mV s −1 and 2.4 mV steps. Before the acquisition of the curves, the cells were brought to and kept in a steady-state at 0.5 V for about 30 min, until constant current and temperature were attained. During the measurement of the polarization curve, the ventilator of the climatic chamber was switched-off to allow for fully static ambient air conditions. The temperature of the cells was monitored during measurements by means of a thermocouple put in contact with the cathodic plate. The impedance spectroscopy measurements were carried out potentiostatically (sine wave, V ac =10 mV RMS ) from 20 kHz to 0.1 Hz (50 points), with the cells placed inside the climatic chamber and ventilator switched-off during measurements. Previous to the impedance acquisition, the cell was polarized at the predetermined potential until its temperature and current response were stabilized. The impedance spectra were analyzed by fitting to an equivalent circuit using Autolab software (Nova). Kronig-Kramer tests, checking stability and linearity conditions, were satisfactorily passed for all spectra. A.M. Chaparro et al. Journal of Power Sources 480 (2020) 228668 3 2.3. Neutron radiography 2.3.1. Thermal neutron source and imaging set-up Neutron imaging of an in-operando fuel cell with thermal neutrons was performed for the first time at the training and research reactor AKR-2 of the Technische Universit¨ at Dresden [36]. The AKR-2 is a very low flux, thermal, homogeneous, solid moderated zero-power reactor with a maximum continuous power of 2 W. Nuclear fission is mainly caused by thermal neutrons, i.e. neutrons with the same kinetic energy as the particles of ambient materials. The nuclear fuel and moderator material (solid polyethylene) are distributed homogeneously in the form of fuel plates. The radial channel 7 was used for best combination of exposure time and effective beam divergence, with an absolute thermal neutron flux of Φ =2300 n cm −2 s −1 (±5%) [36]. The raw data of the neutron radiographs were taken with an imaging system containing an Andor iKon-M 934 (1024 ×1024 pixels) camera cooled down to −70 ◦C combined with a light-tight DNIDS box provided by Paul-Scherrer-Institute, a Nikkor AF-S 50 mm 1:1,4G lens and a 200 μ m thick 6 LiF/ZnS(Ag) scintillator from Tritec. The effective pixel size has been estimated to be about 0.133 mm, which results in a spatial resolution of 0.266 mm that may have additional influences like scattering terms. See Supplementary Information (S1.1) for an experimental test of the spatial resolution. More details of the neutron radiography installation and imaging set-up will be presented in a forthcoming publication. 2.3.2. Fuel cell installation and operation The passive PEMFC is fixed to an aluminum holder in front of the radial channel with the possibility for vertical and horizontal cell positions, i.e. main plane parallel and perpendicular to the gravity field, respectively. The vertical cell was studied under frontal and lateral irradiation. During the acquisition of the neutron radiography the load current is controlled (Keithley 2460 Sourcemeter), while the cell is supplied with dry hydrogen at 0.5 bar g static pressure from a metalhydride canister (Horizon Hydrostick Pro, 1 g H 2 ), and static ambient air (22 ◦C, 30% RH). The cell voltage and temperature are continuously recorded using a Keithley 2701 Multimeter with a 7700 multiplexer card. Figs. S1a and b of the Supplementary Information show photographs of the set-up. Four cell configurations have been radiographed: vertical cell with frontal (1), and lateral (2) incidence and horizontal cell with anode on top (3) and cathode on top (4). Fig. S1b shows the cell in configuration (1). 2.3.3. Image acquisition and processing All neutron images were corrected for the dark-field and the flatfield. The dark-field correction consists of eliminating the pixel bias and the time-dependent dark current noise. With this aim, a series of images is taken with the camera shutter permanently closed. The recorded images are averaged resulting in the dark-field master image (DF). Correction for the neutron beam profile is handled via flat-fielding by taking a series of images without any object in-between the neutron source and the detector and averaging them (FF). The final image (IM) is obtained from the corrected neutron images (NI) according to: IM =NI −DF FF −DFmFF (1) where mFF is the calculated median (middle value separating the higher half from the lower half pixel sets) of the FF image. Averaging of a series of sub-exposure images, including DF and FF, was always performed by the median pixel of the series and without a 2D-Median Filter. Neutron images were taken in-operando at different current demands. Before each experiment, the cell was dried by thorough spraying of compressed air over the cathode surface and through the anode chamber. The dried cell was then mounted in the holder and the dry-cell reference (IM dry ) image was obtained from five sub-exposure images after correction (Eq. (1)). Then, the cell was started-up by increasing progressively the current demand in steps, whilst acquiring neutron images simultaneously. The acquisition time for a single neutron image was 202 s (180 s sub-exposure and about 22 s readout time). Images of the generated liquid water (IM W ) were obtained from the corrected fuel cell images (IM) by dividing by the reference, ‘dry-cell’, image (IM dry ), and taking the natural logarithm as: Fig. 1. a) Exploded view of the passive PEMFC components; b) scheme of the passive operation. Fig. 2. Polarization curves and power density of the passive PEMFC in three different orientations. Cell supplied with dry H 2 in dead-end mode and static ambient air. Pt loadings of anode/cathode: 0.25/0.30 mg cm −2 . The red lines correspond to the best fit of the polarization curves by the theoretical formula (Appendix 1). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) A.M. Chaparro et al. Journal of Power Sources 480 (2020) 228668 4 IMW=log(IM IMdry)(2) To improve the signal-to-noise ratio and match the exposure time with IM dry , moving averages of five consecutive images were computed. All digital image processing was carried out with the publicly available program ImageJ. Additional information of the neutron imaging characteristics is given in Supplementary Information (S1), including the analysis of gray scale as a function of water produced in the cell and the spatial resolution of the images. 3. Results 3.1. Polarization curves and impedance spectroscopy 3.1.1. Polarization curves Polarization curves of the passive PEMFC are shown in Fig. 2 for three cell orientations: vertical, horizontal cathode-down, and horizontal cathode-up. The curves were taken after three cycles from open circuit to 0.3 V at 22 ◦C and 30% RH under ambient static air, while feeding the dead-end anode with static dry H 2 (pressure 0.5 bar g ). The temperature of the cells during polarization curve acquisition was within 25–27 ◦C (see Supplementary Information, S2, Fig. S4), which is a result of its heat generation and dissipation to the surroundings. The curves show the best performance for the vertical cell with a maximum power density of 90 mW cm −2 , compared to 75 mW cm −2 for the horizontal cell. Little difference is observed between the cathode-up or -down configuration in the horizontal cell. However, the cathode-up configuration leads to cell failure in constant current experiments (see next section). Polarization curves have been analyzed with a model for passive airbreathing PEMFC, which accounts for changes in oxygen diffusivity inside the catalyst layer and the gas diffusion layer of the PEMFC cathode [37]. All details of the model, assumptions, and calculations are given in Appendix 1. The analysis of the curves shows oxygen diffusivity in the cathodic catalyst layer changing with cell orientation from 8.0⋅10 −4 cm 2 s −1 for the vertical cell to 6.7⋅10 −4 cm 2 s −1 and 6.5⋅10 −4 cm 2 s −1 for the horizontal cells with cathode-down and cathode-up, respectively. In the gas diffusion layers, the calculated oxygen diffusivities are 145⋅10 −3 cm 2 s −1 for the vertical cell, and 109⋅10 −3 cm 2 s −1 and 111⋅10 −3 cm 2 s −1 for the horizontal cells with cathode-down and cathode-up, respectively. The diffusivity values are gathered in Table 1. The oxygen diffusivity changes within the porous layers can be related with changes in their water saturation (s) by using the expression [38]: Deff =Dbulkf( ε )g(s)(3) where D eff and D bulk are the effective diffusivity and the open space diffusivity, respectively, ε is the porosity, f( ε ) is the ratio of the dry diffusivity of the film and D bulk , and g(s) is the ratio of D eff and the dry diffusivity of the film. In the literature different expressions for f( ε ) and g (s) have been proposed for the catalyst layer and the gas diffusion layer of a PEMFC electrode [38–44]. Here we are employing the expressions of García-Salaberri et al. [38] for the gas diffusion layer, and by Zhen and Kim [44] for the catalyst layer, resulting in the water saturations given in Table 1. The resulting water saturation values in Table 1 show a 35% larger water content in the gas diffusion layer of the horizontal cell, while the catalyst layer saturation does not change much with cell orientation. It is remarkable that the two horizontal positions, with cathode-up and -down, present similar water saturation conditions in the layers. The analysis of the polarization curves shows that the cell orientation changes water saturation in the cathode mostly in its gas diffusion layer, which reduces oxygen diffusivity and leads to a 17% decrease in peak power density for the horizontal with respect to the vertical cell. 3.1.2. Impedance spectroscopy Impedance spectroscopy provides more information on transport limitations arising from the change in cell orientation. Impedance results are shown in Fig. S5 of Supplementary Information (S3) in the form of plots of the imaginary part as a function of the real part of the impedance (Nyquist plots) at two cell voltages and for the same cell orientations as in Fig. 2. A flattened single semicircle is observed in all cases, which is due to the cathodic impedance of the cell [46]. Analysis of the spectra was carried out by fitting to an equivalent circuit (inset Fig. S5a) with a serial resistance (R s ) that accounts for high frequency ohmic losses, i.e. the ionic resistance of membrane and electrodes and electronic resistances, and a parallel combination of a resistor (R 1 ) and constant phase element (Y 1 , n) that accounts for the low frequency losses. The resulting values of the fitting are given in Table 2 together with the steady-state cell temperature during the measurements. They show a 40% increase in R 1 , the low frequency resistor, by changing from vertical to horizontal cell position at 0.3 V, and a minor increase at 0.7 V (6%). This component of the equivalent circuit reflects the limitations for current generation by slow processes, which for a PEMFC under normal operation conditions correspond either to oxygen reduction in the cathodic catalyst layer, and/or to the oxygen transport impedance influenced by the presence of liquid water in the gas diffusion layer or the flow-field channels [46]. Since oxygen reduction kinetics accelerates at low cell voltage because of the favored charge transfer with increasing overvoltage, while the reverse trend is observed in Table 2, it can be inferred that R 1 is mostly determined by mass transport losses in the cathode, that increase when changing the passive PEMFC from vertical to horizontal position. Table 1 Diffusivities of the cathodic gas diffusion layer (D eff,GDL ) and catalyst layer (D eff, CL ) obtained from the analysis of the polarization curves (Appendix 1), and water saturations (s avg,GDL , s avg,CL ) calculated from Eq. (3) using the expressions in foot notes (1) and (2), with D bulk =D O2-N2 =0.20 cm 2 s −1 (1 atm, 20 ◦C) [45], and the porosity values ε GDL =0.75 and ε CL =0.35, for the GDL and CL, respectively. Cell position Vertical Horizontal cathode-down Horizontal cathode-up D eff,GDL /10 −3 cm 2 s −1 145 109 111 s avg,GDL (a) 0.38 0.59 0.58 D eff,CL /10 −4 cm 2 s −1 8.0 6.7 6.5 s avg,CL (b) 0.71 0.73 0.73 a f( ε ) = ε 3.5 ; g(s) =(1-s) 2.15 (ref. [36]). b f( ε ) =1.07 ε 1.75 ; g(s) =(1-s) 3 (ref. [42]). Table 2 Results of the impedance spectroscopy analysis (Fig. S5). Cell Orientation Vertical Horizontal cathode down Horizontal cathode up Cell voltage/ V 0.7 0.3 0.7 0.3 0.7 0.3 Cell temp./ ◦C (a) 29 34 26 30 26 30 R s /mOhm cm 2 550 (±4) 503 (±5) 595 (±4) 551 (±5) 589 (±4) 539 (±6) R 1 /mOhm cm 2 1753 (±14) 2121 (±35) 1866 (±14) 2962 (±35) 1894 (±14) 3032 (±42) Y 1 /mF s n−1 cm −2 25 (±1) 45 (±2) 22 (±1) 32 (±1) 23 (±1) 32 (±1) n 0.72 (±0.01) 0.65 (±0.01) 0.73 (±0.01) 0.70 (±0.01) 0.73 (±0.01) 0.70 (±0.01) C 1 */mF cm −2 (b) 7.5 12.8 6.7 11.4 7.1 12.1 χ 2 0.013 0.028 0.011 0.014 0.009 0.022 a Steady-state cell temperature attained before data acquisition by self-heating (ambient conditions: 22 ◦C, 30% RH). b Effective capacitance for time constant dispersion normal to the electrode surface [47,48]: C 1 * =Y 1 1/n ∙ R 1 (1/n−1) . A.M. Chaparro et al. Journal of Power Sources 480 (2020) 228668 5 Additional information from the impedance spectroscopy can be obtained from the other parameters in Table 2. The constant phase element, which reflects the charge storage capacitance of carbon and platinum in the cathodic catalyst layer, shows little variation with cell orientation. In fact, the associated effective capacitance (C 1 *) calculated from Y 1 , n and R 1 (Table 2) [47,48], which is proportional to the electrochemically accessible electrode area, remains almost unaltered by changing the cell orientation. Finally, a slight increase by 10% in the serial resistance (R s ) for the horizontal cells must be attributed to the 3–4 ◦C of lower self-heating steady-state temperature attained by the cell. More insights into liquid water behavior inside the passive PEMFC can be gained with neutron radiography. 3.2. Neutron radiography Neutron radiography was carried out on the passive PEMFC in vertical and horizontal orientations. For the vertical cell, images were acquired with frontal and lateral incidence. The H 2 inlet port position (up or down) was also tested in the vertical cell to determine any possible influence of the H 2 flow inside the anodic chamber on the water distribution. Images of the horizontal cell were obtained under lateral neutron incidence with cathode-up and cathode-down orientations. Fig. 3 shows neutron radiographs of a cell in vertical position and under frontal incidence with H 2 feeding through the down-port (Fig. 3a) and up-port (Fig. 3b). The detected water appears as a spotted white contrast inside the active area of the cell. The amounts of water generated at the acquisition of the images are indicated in mg cm −2 . The evolution of cell parameters (current, voltage, power, energy, water generation, and temperature) during the acquisition of the images is provided in Figs. S6 and S7 in Supplementary Information (S4). The neutron images of Fig. 3 show liquid water accumulations preferentially centered in the active area of the cell with a slightly higher concentration in the upper half. The plotted distribution of water resembles the array of columns of the cathodic plate, which shows enhanced condensation over the colder areas of the cathode. The small deviation of water condensation towards the upper half of the active area can be explained by the combined effects of gravity and natural convection. The water droplets located in the lower half will leave the cell faster, also entrained by larger drops sliding down from the upper half of the cell. The natural convection contributes to this asymmetry, since ambient air rising along the hot fuel cell will first “dry” the lower part of the cell thus reaching the upper half with a high vapor-content. The effect of natural forces on water distribution and passive cell performance is further discussed in the following section. Finally, the position of the inlet port does not appear to provoke a significant difference in liquid water distribution. Images of the vertical cell under lateral neutron exposure are shown in Fig. 4 with the two different H 2 port locations (see Supplementary Information (S4) Figs. S8 and S9 for cell parameter evolution during the acquisition). As in the frontal incidence images (c.f. Fig. 3), the lateral images and gray scale profiles reflect preferential water accumulation in the central part of the active area with a somewhat larger accumulation in the upper half. The water concentration profiles over the cell cross section are shown in Fig. 5, obtained from Fig. 4 by averaging gray levels over vertical slices covering the active area section. The profiles show a preferential water accumulation around the columns of the cathodic plate, reflecting the passive water removal from the cathode surface by the more hydrophilic and colder cathodic plate. Some water accumulation is observed in the anodic current collector grid (grids are plotted as vertical dotted lines in Fig. 5) especially in Fig. 5b with H 2 feeding from the upper port. The MEA region, right in the center of the profile, shows an apparent lower concentration which must reflect the decrease in absolute water volume concentration inside the highly saturated porous media under operation (c.f. Table 1). The hydrophilic anodic membrane placed behind the anode grid to provide passive water removal shows minor water accumulation, which can be interpreted as an effective permeation. A steady-state profile is attained after about 70 mg cm −2 of water generation, which corresponds with the stabilization of cell voltage and temperature at a current density above 0.2 A cm −2 and a stable power production (Supplementary Information (S4), Fig. 3. Thermal neutron radiography under frontal incidence of the passive PEMFC in vertical position with H 2 inlet from (a) down-port and (b) up-port. The left image is the dry reference image (IM dry , Eq. (2)). Numbers indicate the amount of water produced by the cell in mg cm −2 . The active area of the cell, frame and H 2 ports are drawn for visual guidance. On the right: pixel value profiles taken over the center diameter of the active area of the rightmost images. A.M. Chaparro et al. Journal of Power Sources 480 (2020) 228668 6 Figs. S8 and S9). A different situation occurs when the cell is in horizontal position (Fig. 6). Two cases have been analyzed, cathode-up (Fig. 6a) and cathode-down (Fig. 6b). (See Supplementary Information (S4), Figs. S10 and S11 for cell parameters evolution during images acquisition). The cathode-up configuration gives rise to a voltage drop and cell failure shortly after generating 5 mg cm −2 of water. As a result, no water accumulation can be observed in the neutron images with cathode-up. The cell is returned to full operative state after a short anode purging. With the cathode-down, however, the cell operates without failure; plots of the gray intensity over the active area (Fig. 6c) show initially a rather uniform water generation profile (8 mg cm −2 water production) followed by preferential accumulation of large drops within the central region of the cathodic plate after 60 mg cm −2 of water production, when a steady-state cell voltage and power generation are attained. Furthermore, random oscillations in the cell voltage are observed in Fig. S11 with the amplitude increasing with the current density, which reflect a large effect on cell response by the water drops emerging and moving over the surface of the cathode, most probably by changing the oxygen accessibility. All this shows that the response of the horizontal cell is more sensitive to water dynamics over the cathode surface than in vertical position (cf. Figs. S6-S11). The cross sectional water profiles extracted from Fig. 6 are plotted in Fig. 7. The profile in Fig. 7 shows little water generation and accumulation for the cathode-up configuration because of the early failure of this cell after producing only 5 mg cm −2 water. With the cathode-down, no failure is observed and a water profile evolves in the cathode and anode apparently in a similar way as for the vertical cell (c.f. Fig. 5) in spite of their significant differences in the polarization curves (Fig. 2) and impedance spectroscopy (Fig. S5). A closer comparison is carried out in the next section, pointing out the essential differences between the water profiles to explain the different cell responses. 4. Discussion 4.1. Power generation and passive PEMFC orientation Polarization curves and impedance spectroscopy show an important influence of the cell orientation on the performance of the passive PEMFC. The studies in the literature discussing the effect of the cell orientation on air breathing PEMFC performance arrive at different results. Jang et al. [49] report an increase in the cell performance in vertical position using a cell with vertical slits in the cathodic plate. Obeisun et al. [50] studied an air breathing cell with long horizontal openings in the cathodic plate and observed that cell orientation had a little effect which was apparent after a prolonged operation time. The different results must be attributed to different designs used for the cathodic plate in these two works. A plate that allows droplet mobility over the cathode surface, as in the design by Jang et al. [49], will make the cell more sensitive to orientation and favor the vertical position, compared with the design of Obeisun et al. [50] which is more prone to an accumulation of static water drops on the cathode surface that are preferentially removed by evaporation. In addition to the shape of the openings, most determining for the air-breathing performance are the thickness of the perforated cathodic plate that limits water drop movement and cell performance [51], and the gas diffusion properties of the Fig. 4. Thermal neutron radiography images, and gray scale profiles under lateral incidence in a passive PEMFC operating in vertical position with (a) H 2 inlet from down-port, and (b) H 2 inlet from up-port. Numbers indicate the amount of water produced by the cell in mg cm −2 . Cathode plate (white bar), anode plate (empty bar) and H 2 ports are drawn for visual guidance. The left image is the dry reference image (IM dry , Eq. (2)). Pixel value profiles taken over the cathode surface are shown on the right of each image. A.M. Chaparro et al. Journal of Power Sources 480 (2020) 228668 7 electrode [52]. The cell orientation has been analyzed with theoretical models of air-breathing cells showing that their thermal dissipation and power response are favored in vertical position by the effect of natural convective heat flow [53,54]. The columnar plate of the design in Fig. 1 allows for a high mobility of water drops over the cathode surface, which is a situation similar to the design of Jang [49]. This property leads to lower mass transport resistance and larger peak power density when the cell is in vertical position (Fig. 2). The obtained water contents and mass transport resistances in Tables 1 and 2, respectively, reflect the orientation dependence of the two principal natural forces for the transport of liquid water over the cathode surface, namely gravity and natural convection. The gravitational sliding of water over the electrode surface in vertical position is important when single drops are large enough to overcome adhesion by surface tension effects, which pin them to the cathode surface, current collector grid and the columnar cathodic plate. Such water droplet removal is facilitated by hydrophobic elements that lower the adhesive forces. Natural convection caused by the “hot” fuel cell and oxygen/vapor concentration gradients [55] might even have more significant impact than gravitational dragging on liquid water removal. This phenomenon has been modelled by Ismail et al. [41] and was Fig. 5. Gray intensity vs. cross sectional distance in a passive PEMFC operating in vertical position with H 2 inlet from (a) down-port, and (b) up-port, taken from neutron images in Fig. 4. Gray areas indicate the main components of the cell. Fig. 6. Thermal neutron radiography images under lateral incidence in a passive PEMFC operating in horizontal position with (a) cathode-up, and (b) cathode-down. Numbers indicate the amount of water produced by the cell in mg cm −2 . Cathode plate (white bar), anode plate (empty bar), and H 2 ports are drawn for visual guidance. The left image is the dry reference image (IM dry , Eq. (2)). c) Pixel value profile over the cathode surface in b. Fig. 7. Gray intensity vs. cross sectional distance in a passive PEMFC operating in horizontal position, with (a) cathode-up and (b) cathode-down. Gray areas indicate the main components of the cell. A.M. Chaparro et al. Journal of Power Sources 480 (2020) 228668 8 experimentally observed by Fabian et al. [56] and Coz et al. [17]. For vertically aligned cells, a cold fluid particle rising in front of the hot cathode surface will be heated more and more as it rises, draining air from the surroundings and forming a strong upwards directed jet that enhances water evaporation. The horizontally aligned fuel cell, however, resembles the classical Rayleigh-B´ enard instability problem, where an air layer is heated from below, but natural convection will hardly reach the cathode surface. As a result, the horizontal cell does not benefit from improved dragging of water and natural convection over the cathode surface. The performance of a passive air-breathing PEMFC with a high water mobility over the cathode surface is undoubtedly much more dependent on orientation than that of a conventional cell, where formation of small droplets on the back surface of the electrode and their dragging in the channels by the gas stream minimizes the effects of gravity and natural convection [57]. 4.2. Water distribution in the passive PEMFC and cell performance Water distribution profiles obtained from neutron radiography explain the differences in cell performance from vertical to horizontal positions. In spite of the similarity in water profiles of the vertical and horizontal cathode-down positions observed in Figs. 5b and 7b, a closer view, almost in the limit of the resolution of the technique, allows to see differences determinant for cell performance (Fig. 8). Fig. 8a shows the steady-state water profiles in the MEA region for a vertical and a horizontal cathode-down cell. The steep rise in water profile from the cathodic side of the MEA towards the cathode surface and columnar plate reflects a larger amount of water inside the cathode of the horizontal cell, which is in qualitative accordance with the results of water saturation obtained from the analysis of the polarization curves (Table 1). On the anode side, however, the water profiles are similar for both orientations. A change in water distribution from cathode to anode and its dependence on cell orientation reflect different nature of forces acting for water transport. In the cathode, water transport is dominated by gravity and natural convection over the surface, which are very dependent on cell orientation. In the anode, water is driven by diffusion through the membrane and capillarity in the porous electrode, which is rather insensitive to cell orientation. In fluid-dynamic terms, the passive air-breathing PEMFC is characterized by a different Bond number in the cathode and anode (the Bond or E¨ otv¨ os number describes the relative importance of gravitational forces and surface tension). Whereas the anode accomplishes Bo ≪ 1, as in a conventional PEMFC electrode [57], the air-breathing cathode resembles a situation where Bo >1. The neutron imaging allows to explain the conditions for cell failure observed with horizontal cell position and cathode-up after production of 5 mg cm −2 water. Such an amount of water occupies approximately 20% of the open pore volume of the cathode (35⋅10 −3 cm 3 ). Its strong and fast deleterious effect can only be explained by the flooding of a small, but critical, part of the MEA like the anodic catalyst layer. This situation is inferred from a comparison of the neutron water profiles for the different cell orientations in the MEA region and after production of 5 mg cm −2 of water (Fig. 8b), showing that the horizontal cell with cathode-up is characterized by a larger water concentration in the anode than in the cathode side, which differs from the other cell configurations. The configuration with the cathode on top increases the flow of water to the anode, causing the flooding of its catalyst layer, which is critical for cell operation. Most probably, the low solubility of hydrogen in water favors an immediate failure. 5. Conclusions The main conclusions obtained from the experimentation and analysis shown in this work are: - Thermal neutron radiography at the low-flux nuclear reactor AKR-2 (2 W) is able to resolve build-up of liquid water in the cathode and anode of a passive air-breathing PEMFC. - Liquid water accumulation is predominant on the cathode surface, with differences depending on the cell orientation that determine cell response. - The response of the cell shows 17% larger peak power density in the vertical with respect to the horizontal position. - Polarization curve analysis and impedance spectroscopy show that the horizontal cell presents larger mass transport losses due to higher water contents in the cathodic gas diffusion layer. Neutron imaging confirms more important water accumulation in the cathodic electrode of the horizontal cell. - Cathodic water saturation and performance changes with cell orientation are attributed to two natural forces: gravity and natural convection. The action of these natural forces is favored by the columnar cathodic plate used, which confers high mobility to water drops over the cathode surface and improves cell response in vertical position. - Cell failure occurs in horizontal position with the cathode up. In this position, back-diffusion and gravity operate in the same direction leading to fast flooding of the anodic catalyst layer. CRediT authorship contribution statement Antonio M. Chaparro: Conceptualization, Methodology, Formal analysis, Investigation, Writing - original draft, Visualization, Project Fig. 8. Comparing steady-state water profiles, after production of 85 mg cm −2 H 2 O (a) and after production of 5 mg cm −2 H 2 O (b) for the indicated cell orientations. A.M. Chaparro et al. Journal of Power Sources 480 (2020) 228668 9 administration. P. Ferreira-Aparicio: Conceptualization, Resources, Project administration. M. Antonia Folgado: Methodology, Validation, Formal analysis, Investigation. Rico Hübscher: Methodology, Software, Formal analysis, Investigation, Writing - original draft, Visualization. Carsten Lange: Methodology, Resources, Project administration. Norbert Weber: Conceptualization, Methodology, Software, Resources, Investigation, Writing - original draft, Visualization. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgments This work was supported by the Ministerio de Economía y Competitividad of Spain and Fondo Europeo de Desarrollo Regional (FEDER), Project E-LIG-E, ENE2015-70417-P (MINECO/FEDER), Ministerio de Ciencia, Innovaci´ on y Universidades of Spain, Project ELHYPORT (PID2019-110896RB-I00), and a posdoc fellowship of the German Academic Exchange Service (DAAD). Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi.org/10.1016/j.jpowsour.2020.228668. Appendix 1 The polarization curve for a passive air-breathing PEMFC which accounts for oxygen diffusivities in the cathodic electrode is described in Ref. [37]. The latter is based on a one dimensional model for a flooded agglomerate porous catalyst layer, which considers both charge transfer kinetics and oxygen diffusion losses within the pores. The model assumes macroscopically homogeneous electrode layers, the gas diffusion layer and the catalyst layer, with uniform overpotential distribution over the electrode thickness; the cathode kinetics is governed by first order oxygen reduction with Tafel behavior. The polarization curve obeys the general expression in Eq. (A.1) with three main potential losses, electrochemical ( η ele ), ohmic ( η ohm ), and mass transport ( η mt ): V=E− η ele − η ohm − η mt (A.1) Expressions for each overpotential term are determined as a function of current density. The electrochemical overpotential ( η ele ) is due to the electrochemical reaction in the cathodic catalyst layer (no anodic polarization is considered). To calculate its relation with current density and oxygen diffusivity, the following expression is used: DCL d2c(x) dx2=i(x) nF (A.2) Here, c(x) is the local oxygen concentration as a function of depth (x, see Eq. A.4) in the catalyst layer, i(x) the local current density, D CL is the effective diffusion coefficient of oxygen in the catalyst layer, n (=4) the number of exchanged electrons per oxygen molecule, and F (=96,485 C mol −1 ) the Faraday constant. The local current density (i(x)) depends on local oxygen concentration, electrochemical overpotential, and other kinetics parameters according to: i(x)=c(x) c0Aii0exp( η ele α nF RT ),(A.3) where A i (cm 2 cm −3 ) is the internal electrochemical area of the catalyst layer, i 0 (A cm −2 ) the exchange current density, and α the transfer coefficient. Notice that positive signs are used for the overpotentials (as in Eq. A.1) and for the reduction current; therefore the exponential term in Eq. A.3 is also positive. Eq. A.2 is solved with boundary conditions at the membrane/CL (Eq. A.4a) and CL/GDL (Eq. A.4b) interfaces: x =L CL : dc/dx =0 (A.4a) x =0 : c =c 0 , (A.4b) where L CL is the thickness of the catalyst layer, and c 0 is the oxygen concentration at the CL/GDL interface. Substitution of Eq. A.3 into A.2 and integration gives for the concentration profile: c(x)=c0(exp(2LCL  a √)+exp(2x a √) 1+exp(2LCL  a √))exp(−x a √),(A.5) where: a=Aii0 nFDCLc0exp( η ele α n F RT )(A.6) The parameter 1/a 1/2 is the penetration depth of the electrochemical reaction within the catalyst layer. The current profile (i(x)) and total current density (i t ) are given by: i(x)=nF DCLa c(x)(A.7) A.M. Chaparro et al.