Preprint of "Investigation of Fuel Cell Behavior with Different Catalyst Loadings at Varying Humidities and Temperatures"

Investigation of Fuel Cell Behavior with Different Catalyst Loadings at Varying Humidities and Temperatures 1 2 Yurii V. Yakovlevz, Yevheniia V. Lobko, Miquel Gamón Rodríguez, Alina Madalina Darabut, Lucinda Blanco 3 Redondo, Iva Matolínová 4 Charles University, Faculty of Mathematics and Physics, Department of Surface and Plasma Science, V 5 Holešovičkách 2, 18000 Prague 8, Czech Republic 6 z yurii.yako[email protected]uni.cz 7 Abstract 8 Proton exchange membrane fuel cells, which are suitable for operation at low humidities, are important for the practi9 cal application of fuel cell-based systems. In this work, we study the performance and durability of Proton Exchange 10 Membrane (PEM) fuel cells with varying Pt loadings (ranging from 0.2 to 1 mg cm-2) and ionomer carbon ratios (0.6 11 and 0.8) in cathode catalyst layers under different operating humidities and temperatures. It was found that perfor12 mance is affected by relative humidity, reaching a maximum at 75%RH for loading <0.4 mgPt cm-2 and at 50%RH for 13 loading > 0.8 mgPt cm-2. An increase in the I/C ratio to 0.8 was found to impair fuel cell performance at high humidity 14 levels while enhancing performance and durability under low humidity operation. 15 16 1. Introduction 17 Proton Exchange Membrane (PEM) fuel cells have emerged as a promising alternative to traditional power sources 18 due to their high energy conversion efficiency, reduced emissions, and potential for clean energy generation [1]. How19 ever, one of the challenges associated with PEM fuel cells is their sensitivity to operating conditions, particularly at 20 low humidity. In such an environment, especially at elevated temperatures, rapid dehydration of the ionomer phase 21 and membrane dramatically reduces proton conductivity and, as a result, the overall performance [2,3]. In a dry envi22 ronment, low proton conductivity in the catalyst layer renders the majority of catalyst particles inactive [4]. In addi23 tion, membranes demonstrate high ohmic overpotential and low fuel cell performance at high current densities. More24 over, the durability of PEM fuel cells is significantly affected at low humidity levels [5]. 25 External humidifiers are required to control the humidity levels and fuel performance [6]. However, the addition of 26 humidifiers increases the cost and weight of the system and may be appropriate for stationary or high-power mobile 27 systems. In contrast, medium-power solutions such as portable power generators or unmanned aerial vehicles often 28 use open-cathode or air-breathing designs [7,8]. These designs face challenges in achieving proper water management 29 and often exhibit poor performance due to dehydration. 30 Improving the water retention properties of the membrane [9], gas diffusion [10-12], microporous [13,14], and cata31 lyst layer [13,15–17], among others, can improve fuel cell performance at low humidity. Alternatively, adjusting the 32 catalyst layer loading and thickness may also impact the performance. For example, increasing Pt loading improves 33 reaction kinetics, as observed in a study on air-breathing microbial fuel cells, where increasing Pt loading from 0.2 to 34 2 mg cm-2 boosted fuel cell performance [18]. However, high loadings and thicker catalyst layers can impede mass 35 transport, potentially affecting performance, as noted in [19]. Therefore, determining the optimal catalyst loading is 36 necessary to achieve the best fuel cell performance [20]. While much of the recent research focus has been on achiev37 ing ultra-low Pt loadings (≤0.1 mgPt cm⁻²) [21-23], the durability and stable performance required for commercial 38 and heavy-duty applications still rely on more moderate loadings ≥0.2 mgPt cm⁻² [24], due to their higher reliability 39 [25, 26]. The fundamental interactions between water management, ionomer content, and catalyst loading in these 40 practically relevant systems are not fully understood. Our study addresses this specific knowledge gap. 41 Thus, the objective of this study is to systematically investigate and map the interactions between these key parame42 ters. By varying the cathode catalyst loading (from 0.2 to 1.0 mgPt cm⁻²), the ionomer-to-carbon (I/C) ratio, the operat43 ing humidity and temperature, we aim to elucidate the complex trade-offs between reaction kinetics, mass transport, 44 and water management. The goal is to establish clear design principles that connect catalyst layer composition to op45 timal performance and durability, providing a crucial guide for the development of reliable fuel cell systems. 46 2. Experimental 47 2.1. Catalyst ink preparation. 48 For catalyst ink preparation, 250 mg of catalyst powder (Pt40%@Vulcan) was prewetted with 2 ml of acetone in the 49 glass vial to prevent a catalytic burning of isopropanol vapours. Then, 1.8 ml of the ionomer dispersion (D-521, 5% 50 PFSA dispersion) was poured in, followed by dilution in the mixture of acetone (8 ml) and isopropanol (10 ml). The 51 prepared mixture was homogenised using a horn-type ultrasonicator (HD-3100, Bandeling SonoPulse) for 20 min 52 with a 5 s on/off working cycle. To prevent the dispersion overheating the vial was cooled with an ice bath. The cata53 lyst ink composition was optimized for the preparation of catalyst layers with an ionomer-to-carbon (I/C) ratio of 0.6. 54 Formulation of the catalyst ink with I/C = 0.8 differed in the increased amount of ionomer dispersion, 2.4 ml, with the 55 rest of the routine same. 56 2.2 CCM preparation 57 Catalyst-coated membranes were prepared by using ultrasonic spraying (ExactaCoat, Sono-Tek) of the catalyst ink 58 onto the cation exchange membrane (Fumapem FS-715-RFS, 15 μm of thickness). Fast solvent evaporation from the 59 catalyst layer was guaranteed by securing the membrane on the preheated plate to 65 °C. The spraying procedure was 60 optimised to increase the catalyst loading by 0.05 mgPt cm-2 every cycle. Catalyst loading for the cathode catalyst lay61 er (CCL) had values of 0.2, 0.4, 0.8, and 1.0 mgPt cm-2. Anode catalyst layer loading was the same for all samples and 62 equal to 0.3 mgPt cm-2. The desired loading was achieved by controlling the number of spraying cycles. A constant 63 geometric area of 4 cm2 was ensured by using a cover mask with a precut 2x2 cm2 window. Prepared catalyst-coated 64 membranes (CCMs) were stored in the box, and no additional pretreatment was used before measurements. 65 2.3 Fuel cell measurements. 66 The prepared CCMs were sandwiched between two pieces of gas diffusion layer (GDL) with a micro-porous layer 67 (H24C5, Freudenberg) with an area of 4 cm2 each. This membrane electrode assembly (MEA) was sealed in the 68 graphite cell with a single serpentine flow field. The testing equipment allows constant temperature and humidity op69 eration. Cell temperature was set at 40, 50, 60, and 70 °C; the relative humidity of gases was controlled by setting the 70 proper dew point of the bubbler humidifier and balanced at 100, 75, 50, and 25%. Experiments in a totally dry envi71 ronment were conducted using a humidifier bypass valve. Conditioning of the fuel cell at each set of temperature and 72 humidity is performed at a constant gas flow of 50 sccm for 6 hours. 73 Measurements of fuel cell polarization curves were done after the break-in process described in [19]. Polarisation 74 curves were measured in the galvanostatic mode with 5 mA cm-2 steps every 10 s with simultaneous voltage record75 ing, using a potentiostat (PT2005; Kolibrik.net). The fuel cell was fed by hydrogen and filtered air with gas-to-current 76 stoichiometric ratios of 1.2 and 5, respectively. Polarization curves were repeated at least 5 times, and average values 77 of peak power density and current density at a voltage of 0.6 V were used in the work. Cyclic voltammetry (CV) was 78 measured using H2 /N2 feed with a flow rate of 40 sccm. The anode electrode, owing to its small overpotential, was 79 used as the counter and reference electrode, due to its low overpotential, while the cathode was configured as a work80 ing electrode. Voltammograms were recorded at dV/dt rates of 100, 50, and 20 mV s-1 in the potential range from 0.08 81 to 1 V; measurements were repeated 15 times at each speed. For curves recorded at 50 mV s-1, a total charge of the 82 hydrogen desorption region was defined for curves recorded at 50 mV s-1, where the upper limit of integration was 83 0.4 V, and linear extrapolation of the double layer region to the lower potentials was used as a baseline. Double-layer 84 capacitance was retrieved from the slope of the current at 0.45 V vs. scan rate. Data used to calculate the total charge 85 and double layer current calculation were sampled from the 3rd, 6th, 9th, and 12th scans and then averaged. 86 Potentiostatic impedance spectroscopy (PEIS) was measured in the frequency range of 100 kHz – 100 mHz with an 87 excitation voltage amplitude of 5 mV and polarization at 0.45 V. Measurements were performed using H2/N2 gas feed 88 with flow rates of 50/100 sccm and 100 % of relative humidity for the anode and cathode sides, respectively. Both CV 89 and PEIS were measured using the potentiostat SP-150 (BioLogic). 90 91 3. Results and discussion 92 Cyclic voltammetry can be used for monitoring the electrochemically active surface of platinum catalyst and total 93 electrode/electrolyte interface area, measured by integration of HUPD region and double layer capacitance (Cdl), re94 spectively. During CV measurements no additional water is formed making it more straightforward to achieve equilib95 rium conditions. 96 97 Figure 1 shows the total HUPD charge and the double-layer capacitance as a function of the catalyst loading for catalyst 98 layers with a fixed ionomer-to-carbon ratio of 0.6. As can be seen from the figure, both the HUPD charge (Fig. 1a) and 99 the Cdl (Fig. 1b) increase linearly with catalyst loading. This result indicates that the amount of catalyst active sites, as 100 well as the electrode/electrolyte interface area, increases linearly with catalyst loading in the range of 0.2 – 1.0 101 mgPt cm-2 with slopes (at 100%RH) of 84.7 mC mgPt-1. This, however, is not the case for catalyst particles with low Pt 102 concentration (Pt20%@Vulcan), where linearity can be broken at concentrations above 0.17 mgPt cm-2 due to a thicker 103 catalyst layer [28]. Linear dependency for HUPD charge and Cdl is observed in the relative humidity range from 100% 104 to 25%. However, the values of HUPD charge (unlike Cdl) decrease with humidity for all catalyst loadings in the men105 tioned range of humidities. Compared to HUPD values at 100 RH%, values at 25 RH% decreased by 30-50% depend106 ing on the catalyst loading. The decrease of HUPD charge at low humidity correlates with recent findings[29], and 107 could be a result of the ionomer shrinkage [30] as well as of less developed water channels, which, according to 108 the recent MD study [31], play a significant role in proton transport. Values of Cdl have, however, lesser depend109 ency on humidity, which was also observed in [32]. Although reduced hydration of the catalyst layer could affect 110 double-layer capacitance in a similar way as HUPD charge, the higher sensitivity of the latter may indicate more 111 complicated effects of humidity on hydrogen adsorption. 112 The effect of relative humidity can be further discussed in terms of the dependence of ionomer water content and 113 electrochemical active surface area (ECSA) at different platinum loadings. An equilibrium water content, which 114 is calculated from the relative humidity values (see eq. S2), in the ionomer has a dramatic effect on the ECSA 115 values, as was previously observed in [32,33]. This effect is more pronounced in catalyst layers with lower ion116 omer content (5-30%wt.) and almost vanishes at high content (50%wt.) [32]. The empirical relationship between 117 water content and ECSA can be written in the following form [33]: 118 𝐸𝐶𝑆𝐴(𝜆)=𝐸𝐶𝑆𝐴0(1−𝑒−𝑎𝜆) (1), 119 where ECSA0 is the ECSA value at 100%RH, a – constant, λ – equilibrium water content. 120 Fitting of the experimental data by Eq. 1 (Fig.2a) shows similar behavior for all studied catalyst loadings. The 121 exponent factor, a, is in the range from 0.32 to 0.44, which is similar to the value of 0.32 (at 30 wt%) in [33]. 122 However, contrary to the study [33], where the parameter a had a positive correlation with Pt loading, in a wider range 123 of loadings, we found its value fluctuating. It is worth noting that the correlation may be hidden by measurement 124 error, and a more delicate experiment is required. 125 The temperature dependency of ECSA has been previously observed in a number of publications [34-36]. The 126 decrease in the apparent value of ECSA can be explained by an increased rate of hydrogen desorption; thus, the max127 imum value of the hydrogen monolayer coverage at room temperature decreases from 0.77 to lower values [37]. 128 Overall, an Arrhenius-type equation can be employed to describe such changes [33]: 129 𝐸𝐶𝑆𝐴(𝑇)=𝐸𝐶𝑆𝐴0,𝑟𝑒𝑓(𝑒𝐸𝐴 𝑅(1 𝑇−1 𝑇𝑟𝑒𝑓)) (2), 130 where ECSA0, ref is the ECSA value at Tref – reference temperature (298K for fitting purposes), EA is an activa131 tion energy term, R = 8.31 J mol-1 K-1 is the gas constant. 132 133 As can be seen in Figure 2b, the experimental data can be fitted well using Eq. 2. at different humidities. The analysis 134 of the data only includes measurements at 100 and 75%RH due to experimental limitations in setting low levels of 135 both humidity and temperature. The fitting parameter, EA, which combines information on H-Pt adsorption energy and 136 H-H repulsion, gets values of 2.77 and 3.67 kJ mol-1 for I/C ratios of 0.8 and 0.6, respectively. These values are lower 137 than those calculated in [33], which may be related to the different types of supports used in studies. It should be not138 ed that supported catalysts show deviations from thermodynamically predicted temperature dependencies [34] (see 139 also Fig. S1, eq. S3). 140 However, the overall effect of variation of humidity and temperature on the catalyst activity and, hence, fuel cell per141 formance is different. While lowering the humidity level could imply a real decrease in the number of catalyst active 142 sites (as indicated by the ECSA drop) and ultimately fuel cell performance, temperature variations, on the other hand, 143 could alter apparent values of ECSA only, rendering the temperature dependency of this parameter inadequate for per144 formance estimation. 145 Increasing catalyst loading and, therefore, CL thickness prolongs proton pathways to the membrane. To assess the 146 influence of loading on the proton conductivity in CL impedance spectroscopy was employed. 147 148 As can be seen from the Nyquist plot (Fig. 3), all studied MEAs have the same high-frequency resistance (HFR), 149 which corresponds to the resistance of the membrane. The same HFR for different catalyst loading, which also in150 cludes the contact resistance between the membrane and CL, indicates the absence of CL delamination at higher load151 ing as well as the similarity of testing conditions. The impedance characteristics of a fuel cell measured with N2 sup152 ply at the cathode typically have linear regions, and the real part of resistance at the transition between two regions 153 can be used for the calculation of CL proton resistance [38]. The calculated proton resistance is shown in the inset of 154 Figure 3 and appears to be a linear function of catalyst loading. The linear dependency of proton resistance in the cata155 lyst layer on loading may indicate the fractal dimension of the proton-conducting network close to 1. This allows us to 156 conclude that the proton-conducting channels in the catalyst layer are rather straight. 157 The overall performance depends on the efficiency of the proton, electron, and mass transport as well as on several 158 active sites. Despite the plethora of parameters with complicated interdependencies defining final performance, it can 159 be easily characterised by measuring polarisation curves. Typically, the maximum power density is used for quick 160 performance assessment, while performance characterization at 0.6 V is more practical. 161 162 Figure 4 maps the performance of fuel cells with different catalyst loading operating at different humidities (detailed 163 polarization and power density curves presented in Figure S2). As can be seen from the figure, the performance of fuel 164 cells, irrespective of the catalyst loading, strongly depends on the operating humidity. Thus, all the studied systems 165 demonstrated a decline in performance when working at lower humidities. Such a result is expected due to the fact of 166 progressive ionomer dehydration in the drier environment. The proton conductivity of the ionomer decreases with the 167 degree of hydration, leading to an increase in ohmic losses. However, the performance of the fuel cell at 100%RH 168 does not reach peak values (shown by “stars” in Fig. 4). Instead, peak power density can be reached at relative humid169 ities between 75 and 50% for all catalyst loadings. Moreover, with an increase in catalyst loading, the peak perfor170 mance gravitates to lower relative humidity values. Assuming that the thickness of the catalyst layer is proportional to 171 the catalyst loading, such behavior may be related to water management effects. In this regard, at reduced humidity, 172 when saturation by water vapors is low, the formation of condensed water droplets is less favorable. This, in turn, im173 proves gas diffusion in the tortuous CL pores, while the condensed water droplets may reduce or block gas transport. 174 At lower humidities (25% and 0%RH), however, a drop of proton conductivity upon ionomer dehydration overrides a 175 benign effect of water droplets evaporation on the fuel cell performance. As a result, MEAs with all catalyst loadings 176 demonstrate a reduced performance. 177 An increase in catalyst loading leads to several counteracting phenomena. On the one hand, a linear increase in the 178 number of catalyst sites with loading (as shown in Fig. 1a) facilitates ORR reaction kinetics and so the total perfor179 mance. On the other hand, proton transfer resistance also increases with loading, which could deteriorate the perfor180 mance. Moreover, efficient mass transport at higher loadings is hindered, leading to performance limitations [39,40]. 181 Therefore, the best performance can be reached in MEA with an optimal concentration of catalyst. Examination of 182 Figure 4 reveals that MEA with a loading of 0.4 mgPt cm-2 has the best performance, indicating that trade-offs between 183 many parameters are established. This corresponds to the optimal value of 0.5 mgPt cm-2 [41] demonstrated in the lit184 erature. To show a more detailed picture of environmental effects on fuel cell performance, the peak power density 185 was measured at different working temperatures and relative humidities. 186 187 As can be seen from Figure 5, fuel cell performance is sensitive to both temperature and humidity (detailed polariza188 tion and power density curves presented in Figures S3 and S4). Looking at the data measured under fully hydrated 189 conditions (100 %RH), it can be concluded that increasing temperature has a positive effect on fuel cell performance 190 as the peak power density increases from 0.79 to 0.92 W cm-2 as the temperature increases from 40 to 70 oC. This re191 sult is not surprising as the increase in temperature facilitates better proton mobility and higher catalyst activity. 192 Moreover, the improved gas permeability of thin ionomer films covering catalyst particles and better gas diffusion in 193 nanosized pores also contribute to better performance [39,40]. However, when the relative humidity decreases, water 194 retention in the catalyst layer becomes more difficult; thus, proton transport drops dramatically. Therefore, peak per195 formance at lower relative humidities tends to shift to lower temperatures where retention of water is easier and con196 ductivity of protons can be maintained at sufficient levels. 197 This behavior, however, changes dramatically when the catalyst layer has higher ionomer content. Increasing the ion198 omer-to-carbon (I/C) ratio to 0.8 results in worse fuel cell performance as compared to that of an I/C ratio of 0.6. Such 199 performance deterioration can be explained by mass transport problems in the catalyst layer as ionomer forms thicker 200 films over catalyst particles and blocks pores [42]. However, MEAs with an I/C ratio of 0.8 show lesser variation of 201 performance with humidity at a given temperature. Some slight improvement at intermediate humidities (25-50 %RH) 202 may also be related to the removal of water excess and CL pores de-blockage [43]. 203 It can be concluded that samples with higher ionomer content better retain water and support a reasonable level of 204 proton conductivity, influencing overall fuel cell performance. Water retention can be extremely important when a fuel 205 cell operates at low humidity. This helps not only to improve fuel cell performance but also to prevent its deterioration 206 in the long run. It is important to note that an improvement of water retention is, however, counterintuitive, provided 207 hydrophobicity increases at the ionomer loadings of around 30 wt% [44]. 208 Operation of the fuel cell at low humidity may result in the degradation of the ionomer part. As was shown previously, 209 operation at low humidity provokes the formation of hydrogen peroxide, which provokes ionomer decomposition 210 [45]. Therefore, it is practically important to study ways of mitigating the performance degradation for fuel cells 211 working in low-humidity environments. To prevent dehydration in the catalyst layer addition of ionomer may be a 212 sound strategy due to the hydrophilic treatment of the catalyst layer. 213 Fuel cells with catalyst layers with I/C ratios of 0.6 and 0.8 were tested at constant load in the potentiostatic mode. 214 Starting operation with fully hydrated MEA, the fuel cell was fed with totally dry gasses for 20 hours. Therefore, the 215 fuel cell was operating most of the time in the self-humidification regime. Although the hydration of ionomer will 216 depend on the produced current [46], we have chosen voltage control mode to approach realistic conditions. Such 217 measurements were repeated three times, and the current density was recorded as a function of time and presented in 218 Figure 6. As can be seen in the figure, at the beginning of every 20-hour cycle, fuel cell performance is higher. This 219 higher performance is observed for both I/C ratios and is related to higher hydration of the ionomer and corresponding 220 lesser proton resistance. As a fuel cell operates with dry gases, its performance gradually decreases. Interestingly, sys221 tems with lower ionomer content (I/C=0.6) show a continuous decrease of current density, whereas ones with an I/C 222 ratio of 0.8 after 5 hours of operation reach a stable operation. Moreover, apparent degradation can be observed for 223 systems with I/C=0.6 as every new cycle started and finished at smaller currents than the previous. This contrasts with 224 a system with I/C=0.8, where performance during three cycles shows small variation. To understand the reason for the 225 performance variation after every cycle, we measured the ECSA of fully hydrated samples, which corresponds to the 226 number of active sites. 227 Figure 7 shows the changes in ECSA as the test in a dry environment progresses. The initial ECSA of the system. As 228 can be seen from the figure, in both cases, ECSA values decrease over time. We assume that the decrease in the appar229 ent ECSA values is not attributed to the catalyst degradation [38], but rather to the destruction of ionomer channels 230 due to dry operation. In other words, more catalyst particles failed to fulfil the conditions of the triple phase boundary 231 num fuel cells, International Journal of Hydrogen Energy. 44 (2019) 6245–6256. 379 [49] Y.V. Yakovlev, J. Nováková, P. Kúš, T.N. Dinhová, I. Matolínová, V. Matolín, Highly developed nanostructur380 ing of polymer-electrolyte membrane supported catalysts for hydrogen fuel cell application, Journal of Power Sources. 381 439 (2019) 227084. 382 383 Figures of the manuscript: 384 Figure 1. HUPD charge (a) double layer capacitance at different humidities calculated from cyclic voltammograms for various catalyst loadings measured at 70oC. 385 Figure 2. ECSA dependencies on ionomer equilibrium water content (a) and temperature (b). 386 a) b) b) a) Figure 3. Impedance spectroscopy of MEA with the cathode catalyst loading in the range from 0.2 to 1.0 mgPt cm-2 at 70 °C and 100%RH. 387 Figure 4. Peak power density (a) and current density at 0.6 V (b) as a function of catalyst loading and relative humidity at a temperature of 70 °C. The points of the maximum performance at the given catalyst loading are indicated by stars. 388 a) )b Figure 5. Peak performance of fuel cells with cathode catalyst loading of 0.4 mgPt cm-2 and I/C ratios of 0.6 and 0.8 as a function of temperature and humidity. 389 Figure 6. Fuel cell performance was measured at 70 °C, 0 %RH, constant voltage of 0.6 V for samples with I/C ratios of 0.6 and 0.8. 390 391 Figure 7. ECSA values measured for systems with I/C ratios of 0.6 and 0.8 during dry test protocol at 70 °C and 100%RH. 392