Durable and self-hydrating tungsten carbide-based composite polymer electrolyte membrane fuel cells

Proton conductivity of the polymer electrolyte membranes in fuel cells dictates their performance and requires sufficient water management. Here, we report a simple, scalable method to produce well-dispersed transition metal carbide nanoparticles. We demonstrate that these, when added as an additive to the proton exchange Nafion membrane, provide significant enhancement in power density and durability over 100 hours, surpassing both the baseline Nafion and platinum-containing recast Nafion membranes. Focused ion beam/scanning electron microscope tomography reveals the key membrane degradation mechanism. Density functional theory exposes that OH• and H• radicals adsorb more strongly from solution and reactions producing OH• are significantly more endergonic on tungsten carbide than on platinum. Consequently, tungsten carbide may be a promising catalyst in self-hydrating crossover gases while retarding desorption of and capturing free radicals formed at the cathode, resulting in enhanced membrane durability.


Supplementary Note 1. Synthesis of early transition metal carbides nanoparticles supported on carbon spheres
The conventional synthesis of carbides or nitrides entails normally through high-temperature carburization or nitridation of the metals, leading to low surface area. One of the most facile routes to high-surface-area transition metal carbides is attributed to Lee et al., 1 who developed a temperature-programmed reduction-carburization (TPRC) method to form carbides from precursor oxides under a wide range of conditions. Unfortunately, the unstable mesoporous structure limits the application of this material, especially for high temperature and high pressure reactions.
Recently, an old technique 2 has gained renewed interest, whereby biomass is hydrothermally treated in water under relatively mild conditions providing bulk, mesoporous, or nanostructured carbon materials. 3,4 Cui et al. found that the presence of metal ions effectively accelerates the hydrothermal carbonization (HTC) of starch with shorter reaction times and control of particle shape of carbon materials. 5 Sun and Li applied hydrothermal reduction to encapsulate noble metal nanoparticles into the core of carbon spheres. 6 Inspired by these works, we believe that the formation of solid carbon by HTC could lead to carbide or nitride nanoparticles through the reduction-carburization or nitridation processes, respectively. Here we present a preparation method for production of nano carbides particles dispersed on carbon materials. The synthetic strategy is shown in Figure 1, and all experimental details are summarized in the methods section.
Supplementary Figure 1 shows a top view of the as-prepared nano-WC sample, which has a smooth spherical structure with a diameter of 3-5 µm containing well-dispersed WC nanoparticles on its surface. We used focused ion beam to mill a selected region to reveal the cross-sectional morphology of individual carbon spheres. Although W signal is found across the entire sphere by EDX mapping, nanoparticles of WC are only observed on the surface of carbon spheres. Due to the low resolution of the SEM technique, bright spots on the carbon sphere surface are not necessarily individual particles of WC, since several nanoparticles closely packed on a support surface may also result in a bright spot in SEM images at low magnification. We have further carried out TEM/STEM analysis to characterize the dispersion of WC on carbon spheres, as shown in Figure 1b SEM image and EDX mapping of WC nanoparticles supported on carbon spheres cut by a focused ion beam. The sample was milled using a Ga+ ion beam with a selected region (7μm × 7μm × 5μm, lenth × width × depth) operated with an energy of 30 kV and a current of 600 pA.
Both HAADF-STEM and TEM investigations confirm that the WC nanoparticles are uniformly dispersed on the carbon sphere with a narrow size distribution. We have also conducted STEM tomography analysis of the as-prepared nano-WC sample. STEM images were recorded by tilting the sample from -65° to +55° with 1° increment. Representative 3D rendering of the reconstructed image is shown in Figure 2c, which again shows homogeneously dispersed WC nanoparticles.

Supplementary Figure 2.
Bright field TEM images of nano-WC located on a carbon sphere.
In order to confirm the crystalline structure of our as-prepared nano-WC nanoparticles, we have Ganesan and Lee reported a method of W2C preparation by heating a mixture of a resorcinolformaldehyde polymer and ammonium metatungstate. 8 Yan and Shen also reported a similar method by heating a mixture of ion-exchange resin and W precursor. 9 We believe that "graphitic coke" differs from our materials, because the WC nanoparticles of prior work are not catalytically active in their applications (ORR and methanol oxidation). We have also annealed the samples collected after the HTC step in inert gas (He) at different temperatures (Supplementary Figure   5). Apparently, none of the patterns match well those of α-WC but are similar to diffraction patterns of tetragonal WO3 structure. This indicates that the reduction carburization is a facile route to carburize the W precursor into the interstitial structure. Figure 5. Powder X-ray diffraction characterization of various materials indicated in the legend. Sample collected after the HTC step was further annealed in flowing He gas at 500, 700 and 900 °C. Standard diffraction patterns of WC (JCPDS 00-002-1055) and C (JCPDS 00-026-1076) are included as references.

Supplementary
We have also conducted thermogravimetric analysis (TGA) of the nano-WC sample under flowing air (see Supplementary Figure 6). With the assumption that all W is oxidized to WO3 and the carbon sphere is combusted, we estimate the total loading of W in nano-WC to be around 60 wt.%. We have further calculated the surface elemental concentration of nano-WC from the XPS survey spectrum, which can detect 2-5 nm in depth from the surface (Supplementary Table   1). The nano-WC contains about 52 wt.% of W on top atomic layers of WC and carbon sphere.
We estimate that about 86% of W in nano-WC is carburized near the surface during synthesis. Considering the well-known similarities of carburization, nitridation and sulfidation, 10 it is possible that by changing gas precursors in the second step of our approach, nanoparticles of nitrides and sulfides can also be formed. Further work will focus on such materials. Here PH2O is the ratio of the partial pressure of water vapor in the mixture to the equilibrium vapor pressure of water (P*H2O) at a given temperature. Figure 9. Initial fuel cell performance consisting of baseline recast Nafion membrane (black) and composite membranes incorporating nano-WC (red), commercial WC (pink), and Pt black catalysts (blue). The polarization I-V evaluation of the fuel cell was conducted and controlled by a fuel cell test station from Arbin Instruments. The H2 and O2 humidifiers were maintained at 70, 55, 41, and 14 °C while the fuel cell temperature was set to 70 °C such that the relative humidity of the inlet gases was 100, 50, 25, and 5%. Gas supply lines temperature were maintained 5 o C higher than the fuel cell temperature to prevent condensation of water vapor. Hydrogen fuel and oxygen were fed in co-flow to the fuel cell. H2 and O2 flow rates were 200 ml/min and 400 ml/min, respectively. Figure 9 shows the fuel cell performance of recast Nafion, Pt Nafion, nano-WC/Nafion and commercial WC/Nafion membranes. The Pt/Nafion membrane (blue) shows the least decrease in performance when the humidity drops from 100% RH to 5% RH. Our nano-WC/Nafion shows similar improvement as the Pt/Nafion but less than Pt black due to the lower activity of nano-WC catalyst compared to that of Pt black. However, the improvement is still significant considering the low cost of nano-WC catalyst and the positive effect on membrane durability. Recast Nafion without self-hydrating function shows the largest decrease in performance (from 1 W/cm 2 at 100% RH to 0.3 W/cm 2 at 5% RH). Supplementary Figure 10c Table 2). This degradation is due to major defects formed during the test from higher gas crossover. The accelerated durability tests were conducted according to the DOE protocol at 90 °C and 35% RH. 12 Fuel cells were first conditioned at 1A/cm 2 for 8 hours at 100% RH and 70 °C. Then the fuel cell temperature was raised to 90 °C, and the relative humidity was reduced to 35%. When the fuel cell and humidifiers reached the desired temperature, the fuel cell was switched to OCV, and the durability test started.

The degradation rate for regions i-iii in
The OCV was recorded for evaluation of durability. This test is designed to be much faster than the conventional one so that the lifespan of different membranes can be studied in laboratories, usually within 100 to 300 hours. Since the failure of Pt/Nafion membrane happens within 100 hours, we conducted tests for 100 hours.  The failure of Pt/Nafion is repeatable based on our tests on multiple samples (Supplementary Figure 11). All three samples showed similar trends and failed at ~70 hours. The slight variability in the OCV vs. time profiles is due to the necessarily random formation of pinholes through which reactant gas crosses over, leading to random drops in OCV and eventually, failure. In light of such randomness, the three profiles shown are quite similar.
Supplementary Figure 11. Accelerated fuel cell durability tests of the Nafion composite membranes with 5wt.% Pt NPs.
Supplementary Figure 12. Cross-sectional SEM image of Pt/Nafion and WC/Nafion membranes collected after 100 hours of accelerated durability testing. Figure 13. Gas crossover of recast Nafion measured (open symbols) and vacancy volume percentages estimated by the tomography of recast Nafion membranes after 100 hours durability tests (closed symbols). Gas crossover was tested by the linear scan voltammetry (LSV) method from 0-0.7V with Scan Rate of 2 mV/s on an AMETEK Versa STAT 3 station using 100% RH nitrogen and hydrogen in the working and counter electrodes, respectively. The hydrogen electrode was also used as the reference electrode. Nitrogen and hydrogen flow rates were both set to 100 mL/min. The H2 crossover from reference electrode surface to working electrode surface was then oxidized when H2 moves away from the surface and new H2 molecules come into contact with the surface of the working electrode. H2 oxidation current at 0.3 V was used to compare the H2 crossover of different membranes. Pt/Nafion showed the fastest increase of gas crossover during the durability test due to fastest degradation of the membrane. The nano-WC/nafion membrane is the most stable one with the least gas crossover.

Supplementary
Due to the large scale of composite membranes and strong beam damage to the Nafion structure, where surf is the total energy from DFT, bulk is the number of bulk units, bulk is the energy of one bulk unit, and is the area of the surface.
The surface energy of each low-index surface of WC is provided in Supplementary Table 3. Supplementary  In contrast, the potential free energy diagram for WC (100) (Supplementary Figure 14b) indicates that the lowest energy pathway to produce OH• is a co-adsorbed H* and OOH* intermediate whereby the reaction is strongly endergonic (+4.01 eV). While the thermodynamic barrier for desorption of OH• will decrease at high coverage due to lateral interactions, previous results indicate that these interaction energies are relatively mild with a pairwise O-O interaction of 0.16 eV on a Pt(100) surface. 15 Therefore, the production of OH• through the mechanism described by Yu et al. should be thermodynamically unfavorable even at high coverages on WC (100).

Supplementary Note 4. Assessment of new materials as electrode catalyst
In order to further assess the reactivity of nano-WC, we have tested our catalyst in a fuel cell as electrodes. The results (Supplementary Figure 15) show similar performance using nano-WC catalyst on the anode and cathode. The peak power density of the fuel cells with nano-WC is about 0.09 W/cm 2 . The results indicate catalytic activity of nano-WC catalyst for H2 oxidation and O2 reduction.
Supplementary Figure 15. Polarization curves of fuel cells using nano-WC catalyst as anode catalyst (black) and cathode catalyst (red). The electrode with nano-WC catalyst was prepared by air spraying nano-WC, Nafion and isopropyl Alcohol (IPA) mixture onto commercial gas diffusion media (carbon cloth with microporous layer). The loading of nano-WC is 0.62 mg/cm 2 and the loading of Nafion is 25 wt%. The fuel cells were tested with (1) home-made nano-WC electrode on the anode and commercial Pt/C electrode (0.3 mg/cm 2 ) on the cathode and (2) nano-WC electrode on the cathode and commercial Pt/C electrode (0.3 mg/cm 2 ) on the anode. The testing temperature of the cell is 70 o C, 100% RH with 200 ml/min H2 and 400 ml/min O2.