Proton exchange membrane fuel cells powered with both CO and H₂

Significance

We report a class of IrRu-N-C catalysts, with Ir and Ru single atoms uniformly populated in nitrogen–carbon composites. The catalyst not only represents an example of high-efficiency single-atom catalysis toward H₂ electrooxidation, but also exhibits excellent CO electrooxidation reaction (COOR) behavior. The traditionally invincible CO electrooxidation process occurs easily on this catalyst, with COOR initiates at nearly 0 mV versus reversible hydrogen electrode at ambient temperature. This endows the catalyst with superb CO antipoisoning property in proton exchange membrane fuel cell at ultralow noble metal loading. Combining theoretical calculation and physical probing techniques, we find that the approximated Ir and Ru single atomic sites act in synergy to confer the catalysts with this excellent CO conversion behavior.

Abstract

The CO electrooxidation is long considered invincible in the proton exchange membrane fuel cell (PEMFC), where even a trace level of CO in H₂ seriously poisons the anode catalysts and leads to huge performance decay. Here, we describe a class of atomically dispersed IrRu-N-C anode catalysts capable of oxidizing CO, H₂, or a combination of the two. With a small amount of metal (24 μg_metal⋅cm⁻²) used in the anode, the H₂ fuel cell performs its peak power density at 1.43 W⋅cm⁻². When operating with pure CO, this catalyst exhibits its maximum current density at 800 mA⋅cm⁻², while the Pt/C-based cell ceases to work. We attribute this exceptional catalytic behavior to the interplay between Ir and Ru single-atom centers, where the two sites act in synergy to favorably decompose H₂O and to further facilitate CO activation. These findings open up an avenue to conquer the formidable poisoning issue of PEMFCs.

The proton exchange membrane fuel cell (PEMFC) is one of the key enabling technologies for the transition to the upcoming hydrogen economy (1–9). However, PEMFCs have their Achilles’ heel, i.e., they are easily poisoned by carbon monoxide (10–12). The platinum catalysts, being used exclusively to drive the fuel cell anode, bind to CO preferentially (1 ppm CO in H₂ causes >90% surface blockage) and cease to catalyze hydrogen oxidation due to site blockage (13–16). This feature of Pt makes the use of cheap crude H₂ (∼US$1.5⋅kg⁻¹) from steam reforming unrealistic, and the fuel cell vehicles (FCVs) are now fed exclusively with pure or purified H₂ at ∼10 times higher price (∼US$13 to16⋅kg⁻¹).

To prompt the widespread application of PEMFCs, the anode catalysts should possess certain antipoisoning behavior (17–21), as the H₂ quality may not always be fully assured in each time of refilling. The exploration for less CO-sensitive anode catalysts has been around since the emergence of the PEMFC technology (2, 3, 5). Various Pt-based alloys were explored to alleviate the CO poisoning through the Langmuir–Hinshelwood reaction mechanism (21, 22). However, impractically high voltage losses (0.2 to 0.5 V) are still evident in the presence of trace level CO (10 to 1,000 ppm), owning to the intrinsically too strong CO adsorption on the current metallic Pt catalysts. Therefore, designing catalytic sites with reduced CO adsorption energy is highly desirable for tackling such a problem. Meanwhile, as water is also the reactant for CO electrooxidation, specific catalytic sites capable of activating water at a sufficiently low potential is demanded (23, 24). However, these stringent requirements have not been met by the currently available catalysts yet. For instance, PtRu/C, the best CO tolerant catalysts to date, requires a high Pt usage (normally >0.4 mg⋅cm⁻²) to afford for a certain cell performance in a CO presence (25, 26), which runs in the opposite direction to the cost reduction demand of the technique.

Herein, we report a class of IrRu-N-C catalysts, with Ir and Ru single atoms densely and uniformly populated in nitrogen–carbon composites. The catalyst represents an example of high-efficiency single-atom catalysis in an H₂ fuel cell with practical power density. Meanwhile, the Ir and Ru single sites act in synergy to fast catalyze the CO electrooxidation reaction (COOR), where both CO and H₂O are sufficiently activated at a low potential. Therefore, the IrRu-N-C catalyst at ultralow metal loading exhibits even better antipoisoning behavior (10 to 1,000 ppm CO) than the PtRu/C catalysts that is 17 times higher in mass loading (0.4 mg⋅cm⁻²). Our results thus suggest a promising bimetal center design strategy for producing active and antipoisoning hydrogen oxidation reaction (HOR) catalysts for PEMFC anodes.

Results

Structural Characterization.

We fabricated the catalysts through pyrolyzing (950 °C) zeolite imidazolate frameworks 8 (ZIF-8) trapped with ruthenium acetylacetonate and iridium acetylacetonate (Fig. 1A), i.e., the host guest strategy (27–29). The metal loading in the final catalysts was systematically tuned to obtain optimized catalysis behavior (Ir, 0 to 1.1 wt%, Ru, 0 to 0.6 wt%, SI Appendix, Table 1), with selected samples hereafter denoted as Ir-N-C (1.1 wt%), IrRu-N-C (Ir and Ru both at 0.6 wt%), and Ru-N-C (0.6 wt%). The Ir(acac)₃ and Ru(acac)₃ molecules were in situ trapped in the cages formed by the assembly of Zn²⁺ ions and 2-methyl imidazole. Using aberration-corrected high-annular dark-field scanning transmission electron microscopy (HAADF-STEM), the Ir and Ru are shown to be atomically confined in the micropores of ZIF-8, as indicated by the isolated starry spots in the substrate (Fig. 1 B and C). The polyhedron morphology (SI Appendix, Fig. 1) and well-maintained crystalline texture (Fig. 1D and SI Appendix, Fig. 2A) of the IrRu@ZIF-8 in comparison with pristine ZIF-8 verifies the host−guest structure of IrRu@ZIF-8. The following pyrolysis leads to sample carbonization, as indicated by the X-ray diffraction (XRD) patterns (Fig. 1D and SI Appendix, Fig. 2B), where only two broad peaks located at 25° and 44° assignable to the characteristic carbon (002) and (100) diffractions are noticed in all carbonized samples. N₂ adsorption/desorption results (Fig. 1F and SI Appendix, Fig. 3 and Table 2) show that the IrRu-N-C, Ir-N-C, and Ru-N-C catalysts all represent surface area and pore structure similar to the N-C (derived from ZIF-8) catalyst. Energy-dispersive spectrum elemental mappings indicate metal species, nitrogen, and carbon are still homogeneously distributed throughout the sample (SI Appendix, Figs. 4–8). Using HAADF-STEM (Fig. 1E), the Ir and Ru sites are revealed to maintain their atomic dispersion status (named IrRu-N-C) after carbonization, with densely planted bright dots noticed due to Z-contrast. Small amount of tiny particles are also noticed in the samples, which are revealed as Ru metallic entities (SI Appendix, Fig. 6).

Fig. 1.

The physical characterizations. (A) Schematics of IrRu-N-C catalysts. (B) The HAADF-STEM image of IrRu@ZIF-8, indicating the single-atom dispersion nature of Ir and Ru in the ZIF-8 precursor. (C) Energy-dispersive X-ray elemental mapping of IrRu@ZIF-8. (D) XRD patterns of precursors (IrRu@ZIF-8 and ZIF-8) and catalysts (IrRu-N-C and N-C). The broad peaks at around 20 to 30° and 40 to 50° can be assigned to the amorphous carbon. (E) HAADF-STEM image of IrRu-N-C, indicating the predominant atomic dispersion of Ir and Ru in the final catalyst. (F) Typical N₂ adsorption–desorption isotherms of IrRu@ZIF-8 and IrRu-N-C.

To further unveil the coordination and electronic nature of Ir and Ru in IrRu-N-C, we then resorted to extended X-ray absorption fine structure (EXAFS) and X-ray absorption near edge structure (XANES) spectroscopy (Fig. 2 A–F). First, Ir L₃-edge EXAFS results support the single-atom nature of Ir sites in the catalysts (Ir-N-C and IrRu-N-C). The phase uncorrected Fourier transform EXAFS (FT-EXAFS) results show a single prominent peak located at 1.5 Å ascribable to Ir-N/O back scattering (Fig. 2 A–C), with a metallic Ir-Ir scattering path (2.4 Å) being absent. Least-squares FT-EXAFS curve-fitting analysis shows an average coordination number (CN) of 5.1 and a bond length at 2.02 Å (SI Appendix, Table 3) of the Ir-N/O. Second, the Ru K-edge FT-EXAFS results corroborate the coexistence of single-atom Ru species (Ru-N/O path at 1.5 Å) and Ru metallic particles (Ru-Ru path at 2.5 Å) (Fig. 2E and SI Appendix, Table 4). The Ru fraction engaged as metallic particles is calculated as 19% (SI Appendix, section 1.13), with isolated Ru revealed as the dominant structure in the sample (81%). The fitting analysis shows an average CN at 5.3 and the bond length at 2.03 Å of the Ru-N/O path. From here, we confirm that Ir and Ru single atom sites dominant the sample, as further supported by the XAFS wavelet transform (WT) analysis in k-space resolution (SI Appendix, Figs. 9 and 10). Meanwhile, the overall CN for Ir-N/O and Ru-N/O at around 5 implies H_xO-IrN₄ and H_xO-RuN₄ as the most possible site configuration. Third, Ir L₃-edge XANES white line intensity (Fig. 2 B and C), signifying the transition from occupied 2p to empty 5d states, indicates an average extrapolated valence at +3 (5d⁶6s⁰) (30–32), consistent with XPS results (SI Appendix, Figs. 11–13 and Table 5). Ru K-edge XANES profiles (Fig. 2D) reveal that Ru is also positively charged on average, with edge signals located between Ru foil and RuO₂. These results are in line with the EXAFS analysis, where the metal centers are in cationic states resembling those of the metal porphyrins.

Fig. 2.

Structural analyses. (A) XANES of Ir L₃-edge spectra; the inset is used to confirm the Ir valence by L₃-edge area integration. (B) Fourier transform of k²-weighted Ir L₃-edge EXAFS data. (C) Experimental and best-fitting IrRu-N-C and Ir-N-C results of the Ir L₃-edge EXAFS. (D) XANES of Ru K-edge spectra; the inset is used to evaluate the Ru valence. (E) Fourier transform of k²-weighted Ru K-edge EXAFS data. (F) Experimental and best-fitting IrRu-N-C and Ir-N-C of the Ru K-edge EXAFS results.

Electro-Catalysis Behavior in Acidic Electrolytes.

We then tested the HOR performance of the catalysts, using the best HOR catalysts to date (i.e., the commercial Pt/C and Ir/C catalysts) as benchmarks (SI Appendix, Fig. 14). First, the Pt/C catalyst exhibits typical HOR behavior, with kinetic behavior and diffusion-limiting current resembling those in the calculation results from the Levich equation (SI Appendix, section 1.14) and literature (33, 34). The lower diffusion-limiting current on Ir/C than on Pt/C may arise from the surface oxidation of Ir and thereby the lowered true active site density. Second, the IrRu-N-C catalyst represents HOR kinetic behavior resembling the Pt/C electrode, demonstrating their highly efficient HOR catalysis nature. A higher diffusion-limiting current is evidenced on the IrRu-N-C electrode (i.e., 3.6 mA⋅cm⁻² versus the calculated 2.93 mA⋅cm⁻²) from the Levich equation. This result is well reproduced across samples, with an SD at only 6.2% (SI Appendix, Fig. 15). This phenomenon has also been reported previously (35, 36) and can be attributed to the formation of a porous three-dimensional thick electrode due to the high overall loading of the catalyst (0.51 mg⋅cm⁻² catalyst and 6.0 μg⋅cm⁻² metal loading). Notably, the thickness of IrRu-N-C electrode film is 28 μm according to three-dimensional profile evaluation (SI Appendix, Fig. 16), far exceeding the Pt/C catalyst film at 3.7 μm, and thereby results in a diffusion-limiting current exceeding the theoretical value calculated from the Levich equation. Third, neither Ru-N-C nor N-C catalyst show distinct HOR catalytic behavior, while Ir-N-C shows inferior HOR catalytic activity, thus suggesting the Ir center as the catalyst site and Ru as the promoter. Through HOR performance comparison, we show that IrRu-N-C exhibits comparable HOR behavior to the-state-of-the-art Pt/C catalyst, at a much lower metal loading (6 μg_metal⋅cm⁻² versus 20 μg_metal⋅cm⁻²).

Having achieved satisfactory HOR performance, we then turned to investigate the carbon monoxide oxidation behavior of all of the catalysts in the CO-saturated electrolyte, as it will largely determine the antipoisoning feature of the catalysts (Fig. 3A and SI Appendix, Fig. 17). Expectedly, the Pt/C catalysts do not catalyze CO oxidation until the electrode potential reaches 0.8 V versus reversible hydrogen electrode (RHE), in line with the literature (25, 26). On the commercial PtRu/C, which is recognized as the best CO-tolerating anode catalyst, the CO electrooxidation reaction (COOR) initiates at ∼0.5V, in good agreement with literature (13–16). To our surprise, however, the IrRu-N-C exhibits excellent COOR activity, which reaches its half-wave potential (E_1/2) at 50 mV versus the RHE and diffusion-limiting current at ∼100 mV, respectively. The CO to CO₂ conversion is further verified by online mass spectrometry (MS). As shown in Fig. 3B and SI Appendix, Figs. 18–20, the concurrent mass spectrometric signal at m/z = 44, corresponding to CO₂, is unambiguously shown with the electrode potential set at 25, 50, 100, and 200 mV versus RHE for 5 s in the CO-saturated electrolyte, respectively, corroborating the low potential COOR feature of our catalyst. The overall charge transfer number for the COOR is calculated at ∼2.00 using the Levich equation (SI Appendix, section 1.14) in the diffusion-limiting current region, corresponding to the complete conversion from CO to CO₂ (Eq. 1). As water also participates in the reaction, this result implies that both CO and H₂O are activated on the IrRu-N-C sample at a sufficiently low potential, thereby endowing the catalyst with this excellent CO oxidation behavior. We then carried out further experiments to distinguish the function of Ir from Ru in activating CO and H₂O. The Ru-N-C does not catalyze CO electrooxidation at all, while Ir-N-C shows a much inferior CO electrooxidation activity (E_1/2 at 0.31 V versus RHE) in the absence of Ru sites (Fig. 3A and SI Appendix, Fig. 19). We thus deduce that Ir atoms are the COOR catalytic center, while Ru may work cooperatively to facilitate H₂O dissociation or boost the CO oxidation, as further discussed in the theoretical calculation part.

$$\text{CO}+{\text{H}}_2\text{O}={\text{CO}}_2+{\text{2H}}^{+}+{\text{2e}}^{-}$$ [1]

Fig. 3.

H₂/CO electrooxidation and fuel cell test results of the different catalysts. (A) COOR linear sweep voltammetry (LSV) curves with the scanning rate at 20 mV⋅s⁻¹ and rotating speed at 1,600 rpm on RDE. (B) The IrRu-N-C catalysts’ signals of CO₂ in a thin-layer online MS cell in the presence of saturated CO in 0.1 M HClO₄ at different potential. (C) H₂-O₂ PEMFC cell performance of different anode catalysts at 95 °C with 200 kPa back pressure and 100% RH with metal loading 24 μg_metal⋅cm⁻² (IrRu-N-C and Ir-N-C), 0.1 mg_metal⋅cm⁻² (Pt/C), and 0.4 mg_metal⋅cm⁻² (PtRu/C), respectively. (D) H₂-Air PEMFC cell performance of 24 μg_metal⋅cm⁻² IrRu-N-C and 0.4 mg_metal⋅cm⁻² (PtRu/C) at 80 °C with 100 kPa back pressure and 100% RH. (E) H₂-x ppm (x = 10, 50, 100, 1,000) CO-Air PEMFC maximum peak power density of 24 μg_metal⋅cm⁻² IrRu-N-C and 0.4 mg_metal⋅cm⁻² (PtRu/C) at 80 °C with 50 kPa back pressure and 100% RH. (F) TOF of IrRu-N-C at different temperatures at CO atmosphere.

We then tested the performance of the IrRu-N-C catalysts in PEMFC single cells with or without CO presence under varied testing conditions (Fig. 3 C–E and SI Appendix, Fig. 21 and Table 6), using commercial Pt/C and PtRu/C as the reference. Firstly, the HOR performance of the IrRu-N-C catalyst is validated by running the fuel cell under H₂-O₂ mode. Specifically, the cell using IrRu-N-C (24 μg _metal⋅cm⁻²) anode exhibits peak power density at 1.43 W⋅cm⁻², comparable to the Pt/C and PtRu/C catalyst at a much higher metal loading (1.61 W⋅cm⁻² and 1.31 W⋅cm⁻², 0.1 mg _metal⋅cm⁻² and 0.4 mg⋅cm⁻², respectively) at the same operating conditions (95 °C, 100% relative humidity [RH], and 200 kPa back pressure), as shown in Fig. 3C. The maximum current density reaches 4.8 A⋅cm⁻², indicating that the H₂ electrooxidation occurs satisfactorily fast on IrRu-N-C at ultralow metal loading. The Ru-N-C, however, is not an active anode catalyst in the fuel cell device, which is consistent with the electrochemical test (SI Appendix, Fig. 14). Secondly, the single-cell performances of IrRu-N-C- and PtRu/C-based anodes were compared systematically under both H₂-O₂ and H₂-Air mode at varied testing conditions, as shown in SI Appendix, Table 6. Notably, in spite of that the IrRu-N-C anode is only one-seventeenth in the metal loading of that of the PtRu/C (24 μg_metal⋅cm⁻² versus 0.4 mg⋅cm⁻²), it exhibits superior performance than the counterpart PtRu/C at all testing conditions. For instance, at US Department of Energy operating conditions (H₂-Air mode, 80 °C, 100% RH, and 50 to 200 kPa back pressures) (37), our IrRu-N-C catalyst represents peak power densities at 592 to 787 mW⋅cm⁻², which is higher than the PtRu/C catalysts (578 to 736 mW⋅cm⁻²) at a much higher anode loading, as shown in SI Appendix, Table 6 and Fig. 3D. Taking into account the significantly lowered metal loading, our catalyst is ∼18 times in mass activity that of the PtRu/C. Lastly and most importantly, under exposure to varied content of CO (10 to 1,000 ppm), the excellent antipoisoning feature of IrRu-N-C is further elucidated. Notably, when tested under H₂-Air mode in CO presence(80 °C, 100% RH, and 50 kPa back pressure), the PtRu/C catalyst (0.4 mg⋅cm⁻²) represents peak power densities varying between 514 and 279 mW⋅cm⁻² (CO concentration between 10 and 1,000 ppm), similar to those reported in the literature (26, 38–41). However, despite of the significantly lowered metal loading, our IrRu-N-C catalyst represents higher peak power densities at all CO concentrations (i.e., 566 through 320 mW⋅cm⁻²) (Fig. 3E, and SI Appendix, Fig. 22). These results unambiguously suggest the better CO tolerance of our catalyst, which is 18.4 to 19.1 times in mass activity that of PtRu/C, the catalysts in the CO presence. Therefore, the IrRu-N-C represents an antipoisoning feature that has not been possible with the previously reported PEMFC catalysts (SI Appendix, Table 7).

Having achieved high CO tolerance in the PEMFC single cell, we further run the fuel cell under a more challenging condition (i.e., feeding the cell exclusively with humidified CO as fuel) to see the difference between the varied catalysts (SI Appendix, Fig. 23). As expected, the Pt/C and Ir/C anodes completely cease to work in pure CO, due to the full coverage of CO on the Pt and Ir nanoparticles. Even on the PtRu/C catalysts, a peak power density at only 30 mW⋅cm⁻² is achieved, due to the great difficulty and the high overpotential in CO oxidation on the PtRu alloy. Using our IrRu-N-C anode, however, the single cell exhibits its peak power density at 166 mW⋅cm⁻², with maximum COOR current density reaching 800 mA⋅cm⁻². Assuming all the Ir atoms are active, the CO turnover frequency (TOF) is calculated at 68.6 s⁻¹ (Fig. 3F), 1 to 2 orders of magnitude higher than the best CO chemical and electrochemical oxidizing catalysts in the same temperature range (12, 23, 24, 42). We again attribute this excellent COOR feature to the interplay between Ir and Ru, as the cell based on Ru-N-C is null and the one with Ir-N-C exhibits a much inferior performance (41 mW⋅cm⁻²).

Catalytic Stability Evaluation.

Catalytic stability is another prominent concern for catalysts used in fuel cell applications. We first investigated the durability of the catalyst using rotating disk electrode (RDE), where the potential was cycled between 0 and 0.4 V versus RHE, in H_2- or CO-purged 0.1 M HClO₄ electrolyte for 10,000 cycles. As shown in SI Appendix, Fig. 24, no obvious decay in either the kinetic region or in the diffusion-limiting current was observed in the H₂-purged condition after the accelerated durability tests. In the COOR test, the E_1/2 only slightly increases from 44 to 51 mV, with the diffusion-limiting current staying almost unchanged. We then collected the IrRu-N-C catalysts for the HAADF-STEM to investigate the morphology after the RDE test. As shown in SI Appendix, Fig. 25, the single-atom dispersed Ir and Ru remain intact after the durability test, suggesting the robustness of the catalyst structure.

We further carried out the stability test of the catalyst in PEMFC, under constant voltage mode at 0.65 V under 100 kPa H₂-O₂, combined with start-up/shut-down cycles (13 h a day) and N₂ purging during intervals. An overall 100 h tests were carried out in H₂ mode. Interestingly, in H₂ mode, a current decay is observed from the beginning to the end of the test each day (i.e., from ∼1.0 A⋅cm⁻² to ∼0.7 A⋅cm⁻²). However, the performance is recovered to its initial value after restarting the cell in the cycle following. The possible reason associated with this phenomenon is that the anode microporous structure is blocked by the in situ generated water during fuel cell operation. The performance recovery was observed due to the evaporation of H₂O during N₂ purging and cell resting period. At the end of the 100 h test, the fuel cell almost retained its initial behavior, as noted in SI Appendix, Fig. 26, demonstrating the excellent durability of the catalyst. Furthermore, we also tested the cell behavior of the CO-PEMFC with constant voltage hold at 0.65 V under 100 kPa CO-O₂. The IrRu-N-C catalyst showed only a 21.1% decay in current density after 140 h of continuous operation, indicating a sustainable stability in the CO-PEMFC. We then further validated the superiority of the IrRu-N-C catalysts toward HOR and COOR in PEMFC after the prolonged constant voltage test. As shown in SI Appendix, Fig. 27, the peak power densities of the H₂-PEMFC and CO-PEMFC were only slightly dropped by 8.5% and 9.6%, respectively. Combining the above results together, the IrRu-N-C is shown to exhibit good structure stability to withstand the PEMFC operating condition.

Enhancement Mechanism.

We next combined the theoretical calculations with experimental tests to reveal the underlying origin of the antipoisoning feature of IrRu-N-C. As discussed above, water activation is a prerequisite for COOR; hence we first studied the decomposition process of H₂O based on density functional theory (DFT) calculations. To begin with, it is found that single IrN₄ and RuN₄ sites are incapable of water activation due to the overly high energy barrier (Fig. 4A) for the formation of adsorbed OH* through O-H bond cleavage (i.e., 1.92 eV for IrN₄ sites and 1.35 eV for RuN₄ sites, respectively). The ineffectiveness toward water activation makes the subsequent COOR unlikely to occur on a single site. Thus, given the strong synergistic effect between Ir and Ru in COOR according to experimental observation, we deduce that two single sites in close approach may work cooperatively to complete the reaction. Therefore, a MN₄-MN₄ (M = Ir or Ru) structure is adopted to model the two adjacent metal sites. Notably, all dual sites dissociate water exothermically with low energy barriers or even barrierless (SI Appendix, Table 8), thus conferring the catalysts with abundant OH functional groups, in line with EXAFS fitting results. Next, in order to probe into the adsorption and catalytic center for CO, we carried out temperature programmed desorption (CO-TPD) and diffuse reflectance infrared Fourier transform spectroscopy (CO-DRIFT) analysis. The results show that both Ir and Ru bind to CO at appreciable binding strength (Fig. 4B), with CO linearly bonded atop the site of the single centers according to CO-DRIFT (adsorption peak at 1,990 to 2,026 cm⁻¹, Fig. 4C) (43–45). The CO-TPD results (Fig. 4B) illustrate a slightly stronger CO adsorption on Ir-N-C (716 °C) versus Ru-N-C (612 °C), making CO more likely to adsorb on the Ir active center when the two sites are in close approach. We thus select this adsorption configuration to further model COOR process. Under this adsorption configuration, the adsorbed CO* (on Ir) and OH* (on Ru) combine at a fast pace to form COOH*, as it is an exothermic step (−0.04 eV) with the energy barrier at only 0.49 eV (Fig. 4D). The final deprotonation process to form CO₂ is also thermodynamically favored (−0.14 eV) with an even lower energy barrier at 0.30 eV. The synergy between RuN₄ and IrN₄ in water dissociation and CO oxidation gives this catalyst exceptional CO electrooxidation activity and antipoisoning feature, consistent with our experimental observation.

Fig. 4.

COOR reaction paths and CO adsorption status on the catalysts. (A) Energy profiles for H₂O dissociation on IrN₄ and RuN₄ surfaces. (B) CO-TPD for the as-prepared catalysts. (C) CO-DRIFT for the as-prepared catalysts. (D) Energy profiles for the entire CO oxidation process on IrN₄-RuN₄ and Ir(111) catalysts. The x axis shows the reaction intermediates and transition states (TSs); the y axis shows the relative energy of each state. Ir, Ru, N, C, O, and H atoms are shown in olive, dark green, blue, gray, red, and white, respectively.

In order to elucidate the synergistic effect on the CO resistant feature, two homonuclear models of RuN₄-RuN₄ and IrN₄-IrN₄ were also calculated. The RuN₄-RuN₄ site, although effective toward water dissociation and CO adsorption, is completely ineffective toward COOR as the CO* + OH* step is a strong endothermic reaction (1.42 eV) unlikely to occur thermodynamically, in good agreement with our experimental observations. In contrast, the IrN₄-IrN₄ site seems effective toward COOR at the beginning; however, it has a strong tendency to sequentially dissociate water with energy barrier at only 0.56 eV, rendering the Ir sites easily covered by excessive OH (SI Appendix, Fig. 28). The axial OH not only reduces the active sites for CO adsorption but also makes the sequential CO* + OH* step thermally endothermic (0.8 eV) and hardly occurs. This explains the reason of low CO oxidation activity on the Ir-N-C catalyst in the experiment. On metallic Ir(111), where CO binds to Ir sites strongly (CO-TPD at 784 °C), both water dissociation (energy barrier at 0.89 eV) and the COOH* to CO₂ (1.19 eV) conversion demonstrate high energy barriers, with the former process also being thermodynamically unfavorable (endothermic 0.31 eV). Therefore, the Ir/C is much less effective for H₂ oxidation in the presence of CO impurity than IrRu-N-C.

Discussion

We have developed a PEMFC anode catalyst that comprises Ru and Ir single-atom sites. This catalyst shows superb catalytic performance when using H₂, CO, or a combination of the two as fuels. The poisonous CO turns into a reactive fuel due to the interplay between approximated Ir and Ru sites, as the water dissociation and CO oxidation becomes extremely easy due to the coordination between the two sites. This catalyst paves the way toward a more robust PEMFC anode with a high antipoisoning feature.

Materials and Methods

Experimental details on synthesis and characterization of IrRu-N-C catalyst, including structure and electrochemical characterization, and the IrRu-N-C–based PEMFC measurements, and DFT calculation details on models are provided in SI Appendix.

Data Availability

All study data are included in the article and/or SI Appendix.