Metal-support interaction boosts the stability of Ni-based electrocatalysts for alkaline hydrogen oxidation

Abstract

Ni-based hydrogen oxidation reaction (HOR) electrocatalysts are promising anode materials for the anion exchange membrane fuel cells (AEMFCs), but their application is hindered by their inherent instability for practical operations. Here, we report a TiO₂ supported Ni₄Mo (Ni₄Mo/TiO₂) catalyst that can effectively catalyze HOR in alkaline electrolyte with a mass activity of 10.1 ± 0.9 A g⁻¹_Ni and remain active even up to 1.2 V. The Ni₄Mo/TiO₂ anode AEMFC delivers a peak power density of 520 mW cm⁻² and durability at 400 mA cm⁻² for nearly 100 h. The origin for the enhanced activity and stability is attributed to the down-shifted d band center, caused by the efficient charge transfer from TiO₂ to Ni. The modulated electronic structure weakens the binding strength of oxygen species, rendering a high stability. The Ni₄Mo/TiO₂ has achieved greatly improved stability both in half cell and single AEMFC tests, and made a step forward for feasibility of efficient and durable AEMFCs.

Introduction

Hydrogen-oxygen fuel cells with high energy efficiency and durability permit a sustainable energy system based on solar or electrical hydrogen converted from renewable energy resources¹. With the fast development of low-cost oxygen reduction electrocatalysts at the cathode and alkane chain based anion exchange membranes^2,3, enhancing the efficiency and durability of the anodic hydrogen oxidation reaction (HOR) electrocatalysts becomes particularly vital to the anion exchange membrane fuel cells (AEMFCs), as the HOR activity on the best Pt catalyst is two orders of magnitude slower in alkaline than in acidic medium^4,5. Higher Pt catalyst loading would thus offset the cost merit brought by the cathode electrocatalysts and the membranes.

Enhancing the HOR activity of electrocatalysts is typically achieved by tuning the electronic structure of materials by alloying Pt with foreign metals. Pt alloys (PtRu, PtFe, PtCo, PtNi and PtCu) show greatly enhanced HOR activity compared to Pt, which is attributed to the optimized hydrogen binding energy (HBE)^6,7,8 or potentially together with the optimized surface hydroxyl adsorption energy^9,10,11. Similar strategy is also applied to developing non-precious HOR catalysts such as Ni to completely replace Pt group metals. In the few early attempts to improve the HOR activity of Ni, electrodeposited NiMo and CoNiMo thin films show superior activities to Ni, which is attributed to the optimized HBE introduced by the electronic effects between Ni and Mo/CoMo¹². Recently, a group of Ni-based bimetallic nanoparticles, which are industry-relevant, have been developed for the HOR in base, showing much improved HOR activity with respect to Ni^13,14. Except for tuning HBE on Ni, researchers also ascribe the improved HOR activity partially to the optimized surface hydroxyl group binding strength on foreign metals, which may facilitate the combination of -H and surface -‍OH group through the bi-functional mechanism¹³. Despite the debates in understanding the HOR mechanism in base, alloying achieves success in enhancing the overall HOR activities. However, non-precious metal catalysts often fail to meet the stability requirement. For example, the HOR performance on Ni-based catalysts usually dramatically drops at ~0.1 V versus the reversible hydrogen electrode (RHE) due to the Ni surface passivation^12,15, while study shows that HOR catalysts should be able to remain stable up to at least 0.3 V for a substantial power density under practical operating conditions¹⁶. More severely, the anode potential would be driven up to ~0.8–0.9 V during the start-up and shut-down (SUSD) cycles or upon the H₂ starvation events during fuel cell operations¹⁷. Recently, the amorphous Ni₅₂Mo₁₃Nb₃₅ metallic glass was reported with the deactivation potential (the potential at which the HOR current starts to decrease) substantially increased to 0.8 V. Yet, the AEMFC using the Ni₅₂Mo₁₃Nb₃₅ anode still shows a significant output cell voltage decay (~38%, from 0.74 to 0.46 V at 200 mA cm⁻²) in 50 h¹⁸. So far, none of the current Ni-based electrocatalysts can survive the harsh anodic conditions. Therefore, with achieved acceptable activity, efforts need to be geared to improve the stability of non-precious electrocatalysts.

The deactivation of HOR on Ni-based electrocatalysts is accompanied with Ni surface oxidation^19,20,21, which passivates the catalyst surface and limits the electrochemically active window. Tuning the electronic structure of Ni-based materials for a weaker binding strength of Ni towards O/OH is therefore the most applied strategy. Ni_5.2WCu_2.2 ternary alloy²² and phase-separated Mo-Ni alloy (PS-MoNi)²³ have exhibited elevated deactivation potential to ~0.3 V. Covering a protective layer outside Ni nanoparticles such as a few-layer hexagonal boron nitride (h-BN) could also prevent Ni from oxidation, which was speculated to originate from the weakened binding affinity towards oxygen species²⁴. However, these negligible improvements in stability are not sufficient for practical AEMFC operation, and the stability issue is still the key problem to solve.

The interaction between the metal catalyst and the support also plays a vital role in tuning the reactivity of metal catalysts. Study shows that the interfacial charge transfer from Ni₃N to the carbon support can move the deactivation potential from 0.16 V for Ni₃N to 0.26 V for Ni₃N/C²⁵. Great improvements have been made on Pt group metals using metal oxide as the support. TiO₂ partially encapsulated Pt shows extraordinary HOR activity above 1.0 V, wherein Pt alone will be oxidized and lose its HOR activity²⁶. Using TiO₂ to support Ru (Ru@TiO₂) moves the deactivation potential from 0.2 V (Ru) to 0.9 V (Ru@TiO₂)²⁷. The interaction between Pt/Ru and TiO₂, known as the metal-support interaction (MSI), enables an efficient charge transfer from TiO₂ to metals as TiO₂ is intrinsically an electron-rich semiconductor (negative semiconductor)²⁸. Correspondingly, it is reasonable to speculate that the charge transfer introduced by the MSI may also exist between Ni-based non-precious metals and TiO₂. Here, we report that the MSI between Ni₄Mo and the TiO₂ support boosts the durability for HOR up to 1.2 V, enabling a continuous AEMFC power output at 400 mA cm⁻² for nearly 100 h, which makes a step forward for the automotive applications of the AEMFCs.

Results

HOR catalytic performance of Ni₄Mo/TiO₂

We prepared Ni₄Mo/TiO₂ catalysts through annealing the pre-mixed NiMo hydroxide precursor and the TiO₂ support in H₂ at 400 °C (see Methods). Optimization was made by adjusting the molar ratio between TiO₂ and Ni₄Mo according to their HOR performance (Supplementary Fig. S1 and Supplementary Table S1). The HOR catalytic performance of Ni₄Mo and Ni₄Mo/TiO₂ catalysts was evaluated using the rotating disk electrode (RDE) method. Figure 1a shows the positive-going sweeps of the cyclic voltammograms (CVs) of Ni₄Mo and the best performing Ni₄Mo/TiO₂ in both H₂ and N₂-saturated 0.1 M NaOH. Ni₄Mo reaches the limiting current density of 2.65 mA cm⁻²_geo at only 85 mV overpotential, showing a very high HOR activity. Yet, the HOR current quickly drops at 0.2 V, and eventually tracks the CV curve collected in N₂, which strongly suggests that the surface oxidation of Ni₄Mo blocks the active surface areas and prohibits the hydrogen oxidation. In contrast, while Ni₄Mo/TiO₂ shows a similar HOR limiting current density of 2.23 mA cm⁻²_geo at 90 mV overpotential, it can catalyze the HOR even up to 1.0 V, with a mere 10% limiting current density decay. There is continuous HOR current at even ~1.4 V before oxygen starts to evolve (Supplementary Fig. S2). The difference in the polarization curves in H₂ and N₂-‍saturated electrolytes confirms that the anodic current in the presence of H₂ in the full potential window is indeed originated from the H₂ oxidation. It is also clearly seen from the CV curves collected in N₂-saturated electrolyte that Ni₄Mo/TiO₂ hardly exhibits any features associated with the surface oxidation as Ni₄Mo. Further chronoamperometry measurements verify that Ni₄Mo/TiO₂ exhibits a stable HOR current at as high as 1.2 V for 8000 s without noticeable decay (Fig. 1b and Supplementary Fig. S3). In comparison, Ni₄Mo loses 87% of the HOR current when changing the holding potential from 0.2 to 0.3 V, and demonstrates an unacceptable low activity at 0.3 V (Fig. 1b).

Supplementary Fig. S4 shows the HOR polarization curves of Ni₄Mo and Ni₄Mo/TiO₂ at different rotation speeds. Koutecky–Levich plots at 0.05 V exhibit a linear relationship between the inverses of i and ω^1/2, with the slopes being 5.38 and 5.15 cm² mA⁻¹ s^−1/2 for Ni₄Mo and Ni₄Mo/TiO₂ (insets of Supplementary Fig. S4c, d). These values match reasonably well with the theoretical value of 4.87 cm² mA⁻¹ s^−1/2 for the 2 e⁻ HOR⁵, and are also in close agreement with the previous study²⁹. The exchange current density (i₀) was extracted by fitting the kinetic current to the Butler–Volmer equation (Supplementary Fig. S5). The mass activity (i_{0,m,298 K}) was then obtained by normalizing i₀ to the Ni mass, as shown in Supplementary Table S1 and Supplementary Fig. S6. Ni₄Mo has a i_{0,m,298 K} of 9.6 ± 0.5 A g⁻¹_Ni, in reasonably good agreement with previously reported values (6.8 A g⁻¹_Ni in ref. ¹³ and 14.1 A g⁻¹_Ni in ref. ¹⁴). The i_{0,m,298 K} of Ni₄Mo/TiO₂ first increases, and then decreases with increasing Ti/Ni ratio. At low Ti/Ni ratios (Ti/Ni < 0.4), Ni₄Mo/TiO₂ follows similar HOR behavior as Ni₄Mo that it starts to deactivate at ~0.2 V (Supplementary Fig. S1), despite the slightly higher mass activities. The best performance of Ni₄Mo/TiO₂ is achieved at Ti/Ni = 0.42, wherein Ni₄Mo/TiO₂ exhibits a similar i_{0,m,298 K} (10.1 ± 0.9 A g⁻¹_Ni) as Ni₄Mo, but a more stable HOR current above 1.0 V. When Ti/Ni ratio continues to increase (Ti/Ni > 0.5), in spite of its improved stability, Ni₄Mo/TiO₂ loses its mass activity significantly, which is most likely due to the lost active surface area by the TiO₂ coverage. The HOR activation energies (E_a) on Ni₄Mo and Ni₄Mo/TiO₂, determined from the Arrhenius plots (Supplementary Fig. S7), are 15.9 kJ mol⁻¹ and 19.5 kJ mol⁻¹ respectively, matching well with 18.6 kJ mol⁻¹ for Ni₄Mo reported previously¹³. Both Ni₄Mo and Ni₄Mo/TiO₂ demonstrate decreased E_a with respect to metallic Ni (30.0 kJ mol⁻¹) and partially oxidized Ni (26.0 kJ mol⁻¹)³⁰.

Figure 1c and Supplementary Table S2 summarize recent important progress in the development of Ni-based non-precious metal electrocatalysts for the alkaline HOR. Despite previous efforts in improving the HOR activity by one order of magnitude, achievements in enhancing the deactivation potential remains very limited. With the acceptable high mass activity, the Ni₄Mo/TiO₂ reaches a new deactivation potential of 1.2 V for the alkaline HOR Fig. 1.

Fig. 1: HOR performance of Ni₄Mo/TiO₂.

a Positive-going sweeps of the cyclic voltammograms of Ni₄Mo and Ni₄Mo/TiO₂ recorded in H₂ and N₂-‍saturated 0.1 M NaOH at 1600 r.p.m with a scanning rate of 0.5 mV s⁻¹. The potentials are iR-corrected; b chronoamperometry curves of Ni₄Mo and Ni₄Mo/TiO₂ at constant potentials in H₂-saturated 0.1 M NaOH at 1600 ‍r.p.m. The potentials are not iR-corrected; and c deactivation potentials and mass activities of Ni-based non-precious metal electrocatalysts for the alkaline HOR. Details are listed in Supplementary Table S2. Note 1: The Ni loadings are 477 and 376 μg_Ni cm⁻²_geo for Ni₄Mo and Ni₄Mo/TiO₂. Note 2: The data of CoNiMo¹² and 4.3%N-Ni⁷⁹ were calculated based on the original data. Note 3: The mass activities of bulky PS-MoNi²³ and Ni₅₂Mo₁₃Nb₃₅¹⁸ are unavailable. Here we set them at 1 A g⁻¹ for comparison.

AEMFC performance of the Ni₄Mo/TiO₂ anode catalyst

Encouraged by the excellent intrinsic HOR performance of the Ni₄Mo/TiO₂ catalyst in the RDE measurements, the assembled single cell test was conducted for AEMFC performance. The Ni₄Mo and Ni₄Mo/TiO₂ were used as the anode catalysts, the commercial Pt/C was employed as the cathode catalyst, and QAPPT was applied as both anion exchange membrane and ionomer to fabricate the membrane electrode assembly (MEA). Figure 2a shows the cell voltage and power density of the cells using Ni₄Mo and Ni₄Mo/TiO₂ as the anode catalysts. Ni₄Mo approaches a peak power density (PPD) of 188 mW cm⁻², which is prominent in Ni-based non-precious metals and is much higher than the NiMo/KB reported previously (120 mW cm⁻²)¹⁵. Strikingly, Ni₄Mo/TiO₂ boosts the PPD to 520 mW cm⁻², among the best AEMFC performances that have been reported under similar conditions (see Supplementary Table S3 for a summary of AEMFC performance of non-precious metal anode catalysts). In addition, Ni₄Mo can only approach a current density of 275 mA cm⁻² (at 0.685 V), and fails to operate beyond this point because of the Ni oxidation induced deactivation. However, Ni₄Mo/TiO₂ is able to deliver a much higher current density of 900 mA cm⁻² (at 0.543 V), indicating that Ni₄Mo/TiO₂ is more resistant to oxidation and remains active at a higher polarization potential. The long-term durability of Ni₄Mo/TiO₂ was evaluated at a large current density of 400 mA cm⁻² (Fig. 2b), and the cell exhibits a stable operation for nearly 100 h. Gao et al. have demonstrated that the Ni@CN_x anode AEMFC exhibits a stable performance at 200 mA cm⁻² for 100 h, which has been considered as a groundbreaking achievement at a large current density in AEMFCs³¹. The Ni₄Mo/TiO₂ anode further elevates the durability to a larger current density of 400 mA cm⁻² for nearly 100 h (also see Supplementary Table S3 for Ni-based anode AEMFC durability). Moreover, the Ni₄Mo/TiO₂ anode can operate at 0.65 V (typical fuel cell operating potential for the automotive applications) for more than 80 h (Fig. 2b), paving the way for progress in the automotive applications of AEMFCs. Furthermore, the cells using Ni₄Mo and Ni₄Mo/TiO₂ as the anode catalysts were discharged consecutively for three times. While the PPD of Ni₄Mo is severely degraded from 188 mW cm⁻² (1st cycle) to 171 mW cm⁻² (2nd cycle) and 77 mW cm⁻² (3rd cycle), as shown in Fig. 2c, Ni₄Mo/TiO₂ demonstrates a mere decrease of 7% in PPD from 520 mW cm⁻² (1st cycle) to 480 mW cm⁻² (2nd cycle), and retains the performance in the following cycle (Fig. 2d).

Fig. 2: H₂-O₂ AEMFC performance.

a Polarization and power density curves; b AEMFC durability of Ni₄Mo/TiO₂ at 400 mA cm⁻²; and c, d polarization and power density curves of Ni₄Mo and Ni₄Mo/TiO₂ AEMFCs discharged for three times. AEMFC performance test conditions (a, c and d): cell temperature at 80 °C under H₂ and O₂ condition with a backpressure of 0.2 MPa for the anode and cathode, H₂ and O₂ humidified at 80 °C (100% RH) supplied to the anode and cathode compartments with a flow rate of 1000 sccm. AEMFC durability test conditions (b): H₂ and O₂ flow rate of 300 sccm and 500 sccm respectively, under otherwise identical conditions. The cell voltages were recorded with no iR-correction. The Ni loadings are 1.35 mg_Ni ‍cm⁻² for Ni₄Mo and Ni₄Mo/TiO₂ at the anode, and the Pt loading is 0.4 mg_Pt cm⁻² for Pt/C at the cathode.

Crystallographic and Morphologic Characteristics of Ni₄Mo/TiO₂

The X-ray diffraction (XRD) pattern in Fig. 3a shows that the synthesized Ni and Mo alloy exhibits a Ni₄Mo phase structure (JCPDS 65-5480), consistent with the ICP-OES results (Supplementary Table S1). The TiO₂ distinctly presents both the anatase phase (JCPDS 89-4921) and the rutile phase (JCPDS 89-0552). The high intensity diffraction peaks at 25.3° and 37.8°, indexed to the anatase TiO₂ (101) and (004) crystal planes, suggest that the anatase phase is the major composition in the pristine TiO₂. Ni₄Mo/TiO₂ exhibits crystallographic features associated with both Ni₄Mo and TiO₂. Increasing the Ti/Ni ratio leads to the decreased relative peak intensity of Ni₄Mo to TiO₂ (Supplementary Fig. S8), corresponding to the gradually decreased Ni₄Mo loading on TiO₂. The transmission electron microscopy (TEM) image shows an interconnected particle morphology of Ni₄Mo with the particle size of about 7.5 nm (Fig. 3b and Supplementary Fig. S9). Aberration-corrected high-angle annular dark-field scanning transmission electron microscope (HAADF-STEM) image of Ni₄Mo (Fig. 3c) shows the 0.210 and 0.202 nm-lattice fringes, corresponding to the (121) and (220) planes of Ni₄Mo. The corresponding fast Fourier transform (FFT) pattern demonstrate the tetragonal Ni₄Mo crystalline phase with Ni₄Mo (121), (220) and (330) planes (Fig. 3d). Elemental mappings using energy dispersive X-ray spectroscopy (EDS) further reveal the compositional distributions of Ni and Mo elements, confirming the formation of a uniform alloy (Supplementary Fig. S10). The pristine TiO₂ has a rectangular plate morphology with the 0.354 and 0.237 nm-lattice fringes corresponding to the (101) and (004) planes of the anatase phase (Supplementary Fig. S11). TEM and HAADF-STEM images of Ni₄Mo/TiO₂ show a uniform distribution of spherical Ni₄Mo particles with a mean size of 7.6 nm on the TiO₂ support (Fig. 3e–g and Supplementary Fig. S12), maintaining the crystallographic characteristics as unsupported Ni₄Mo with the lattice fringe interplanar spacings of 0.210 and 0.202 nm for Ni₄Mo (121) and (220) planes. Besides, the lattice fringe with a spacing of 0.237 nm for the TiO₂ support is consistent with the pristine TiO₂, corresponding to the (004) plane. This structure was further identified by the FFT images (Fig. 3h, i). The inverse FFT image in Fig. 3j also show the interface between TiO₂ and Ni₄Mo, further demonstrating that Ni₄Mo is supported on the TiO₂.

Fig. 3: Physical characterizations.

a XRD patterns of Ni₄Mo, Ni₄Mo/TiO₂ and TiO₂; b TEM image, c HAADF-STEM image, and d corresponding fast Fourier transform (FFT) pattern of Ni₄Mo; e TEM image, f HR-TEM image, g HAADF-STEM image, h and i corresponding FFT patterns, and j inverse FFT pattern of Ni₄Mo/TiO₂.

MSI in Ni₄Mo/TiO₂

Raman spectroscopy study of TiO₂ shows six Raman-active modes at around 140 cm⁻¹ (E_g), 195 cm⁻¹ (E_g), 393 cm⁻¹ (B_1g), 511 cm⁻¹ (A_1g + B_1g) and 635 cm⁻¹ (E_g) (Supplementary Fig. S13a), in good agreement with Raman spectrum of the anatase TiO₂³². No detectable Raman shift or peak broadening is observed for TiO₂ treated at 400 °C in H₂ (TiO₂-H₂-400 in Supplementary Fig. S13b), indicating that there are not enough oxygen vacancies being introduced, as they will cause the blue-shift and peak broadening in the Raman spectra of TiO₂^33,34. When Ni₄Mo is decorated on the TiO₂ support, there appears to be distinct blue-shift in the vibrational mode of E_g from 140 cm⁻¹ (TiO₂) to 150 cm⁻¹ (Ni₄Mo/TiO₂) and peak broadening (Supplementary Fig. S13b), similar to that for Au or Ag decorated TiO₂ in a previous study³⁵. The blue-shift and peak broadening are most likely due to a compressive strain^36,37, possibly introduced by lattice mismatch between the metal particles and the TiO₂ support. The Raman results, together with the aforementioned electron microscopy study, confirm an interfacial interaction between the Ni₄Mo particles and the TiO₂ support.

Figure 4a shows that X-ray photoemission spectrum (XPS) of the Ni 2p_3/2 level of Ni₄Mo contains four peaks, which can be assigned to Ni⁰ (852.7 eV), Ni²⁺ (856.1 eV) and their two satellites (858.9 eV for Ni⁰ satellite and 861.6 eV for Ni²⁺ satellite)^38,39,40. Notably, both of the Ni⁰ and Ni²⁺ peaks of Ni₄Mo/TiO₂ at all Ti/Ni ratios shift negatively to lower binding energies (BE) (Fig. 4a and Supplementary Fig. S14a). XPS spectra of Mo 3d level reveal a complex Mo 3d_3/2 and Mo 3d_5/2 doublets, with the 3d_5/2 peaks centered at 228.0, 229.0, 230.0 and 232.2 eV, corresponding to Mo⁰, Mo⁴⁺, Mo⁵⁺ and Mo⁶⁺ respectively (Supplementary Fig. S14b)^41,42,43. These peaks do not shift when loading Ni₄Mo onto TiO₂ at all Ti/Ni ratios. Ti 2p doublets of TiO₂ show two peaks centered at 458.6 and 464.3 eV, assigned to Ti⁴⁺ 2p_3/2 and Ti⁴⁺ 2p_1/2 respectively⁴⁴, and the Ti⁴⁺ in Ni₄Mo/TiO₂ displays the positive core level shift to 458.9 eV (Ti⁴⁺ 2p_3/2) and 464.6 eV (Ti⁴⁺ 2p_1/2), indicating that Ti is in an electron-deficient status with a higher valence state (Fig. 4b). The O 1s spectrum for TiO₂ is de-convoluted to two peaks at 529.9 and 531.2 eV (Fig. 4c), associated with the lattice oxygen in TiO₂ (Ti-O) and the surface hydroxyl group (H-O)^27,44. The O 1s spectrum of Ti-O for Ni₄Mo/TiO₂ also exhibits a positive BE shift by ~0.2 eV. Note that Ni₄Mo/TiO₂ experienced high temperature reduction in H₂ at 400 °C; both of the Ti 2p and O 1s spectra of the bare TiO₂ treated under the same condition (TiO₂-H₂-400) were also collected, and no significant difference was observed (Supplementary Fig. S15). At all Ti/Ni ratios, both of the Ti 2p and O 1s peaks shift to higher BE. The X-ray absorption fine structure spectroscopy (XAFS) is used to further probe the impact of the TiO₂ support on the chemical environment of Ni. Notably, the absorption edge of Ni₄Mo/TiO₂ displays a slight shift toward the lower photon energy relative to Ni₄Mo (Fig. 4d, left inset), indicating the electron enrichment on Ni atoms in Ni₄Mo/TiO₂. The white line absorption intensity is also weaker than that of Ni₄Mo (Fig. 4d, right inset), signifying the lower Ni valence state in Ni₄Mo/TiO₂ (Supplementary Fig. S16). In addition, as shown in the Fourier-transform of Ni K-edge extended X-ray absorption fine structure (EXAFS) (Fig. 4e), the intensity of the peak at 2.0 Å, assigned to Ni-Ni/Ni-Mo coordination of Ni₄Mo/TiO₂, is higher than that of Ni₄Mo, demonstrating an increased Ni coordination number, which is speculated to originate from the interaction with the TiO₂ support. XPS and XAFS results strongly suggest that there exists electronic interaction between Ni₄Mo and the TiO₂ support through the charge transfer from TiO₂ to Ni.

Fig. 4: Metal-support interaction in Ni₄Mo/TiO₂.

a Ni 2p_3/2, b Ti 2p and c O 1s level X-ray photoemission spectra; d normalized Ni K-edge X-ray absorption spectra (the left inset is the magnified near-edge, and the right inset is the white line); e Fourier-transform of Ni K-edge EXAFS spectra; and f ultra-violet photoemission spectra of Ni₄Mo and Ni₄Mo/TiO₂.

Work function values determined from ultra-violet photoemission spectroscopy (UPS) measurements also support the charge transfer from TiO₂ (3.6 eV) to Ni₄Mo (4.3 eV) (Supplementary Fig. S17). To collectively elucidate the electronic interaction between Ni₄Mo and TiO₂, the d band density of states was further investigated by UPS as shown in Fig. 4f and Supplementary Fig. S18. All samples illustrate intense emission between −3 and −10 eV below the Fermi level (ε_F), related to O 2p orbitals^44,45. TiO₂ and TiO₂-H₂-400 do not show density of states at the ε_F, but a very weak emission at −0.7 to −0.9 eV, which probably originates from the Ti³⁺ defect states⁴⁴. The bands located at 0 to −3 eV are attributed to the Ni 3d states^46,47. The Ni 3d band centers are roughly estimated to be −1.31 eV and −1.59 eV for Ni₄Mo and Ni₄Mo/TiO₂ respectively, according to Eq. (1) (see Methods)^48,49,50. Clearly, the centroid d band of Ni₄Mo/TiO₂ is regulated by TiO₂, and down shifts away from the ε_F compared to Ni₄Mo. Considering that the UPS measurement only probes the occupied d states, we later performed theoretical density functional theory (DFT) calculations to identify the density of states (DOS) including the unoccupied states in the next section.

Robust structure of Ni₄Mo/TiO₂ in electrochemical measurements

Post-reaction characterizations were performed to study the structural stability of the Ni₄Mo/TiO₂ catalyst. After long-term stability test on the RDE, Ni₄Mo/TiO₂ still shows an interconnected particle morphology (TEM and HR-TEM in Supplementary Figs. S19a, b) with the Ni₄Mo phase structure well maintained (selected-area electron diffraction, SAED in Supplementary Fig. S19c), and the Ni, Mo, Ti and O elements also have a uniform spatial distribution (EDS mappings in Supplementary Fig. S19d–h). XPS measurements similarly demonstrate inconspicuous change in the Ni 2p_3/2, Mo 3d, Ti 2p and O 1s spectra after long-term stability test (Supplementary Fig. S20). ICP-MS results suggest that Ni and Ti dissolutions are negligible in 0.1 M NaOH. Yet, Mo is found to dissolve from the top surface of Ni₄Mo particles (Supplementary Table S4), in good agreement with previously reported Mo dissolution on NiMo alloys^51,52,53,54. The TiO₂ support does not exhibit an inhibitive effect on Mo dissolution. However, Mo dissolution does not seem to deteriorate the HOR performance, as evidenced by the stable HOR current in the long term CA test (Fig. 1b).

After 100 h durability test in the AEMFC setup, the crystal, morphological, compositional, and electronic structures of Ni₄Mo/TiO₂ also exhibit negligible change (XRD, TEM, HR-TEM, SAED, EDS and XPS in Supplementary Figs. S21–23). The multiple post-reaction characterizations clearly demonstrate the structural robustness of Ni₄Mo/TiO₂, which enables its high HOR performance and durability in both RDE and AEMFC tests.

Proposed mechanism for the enhanced stability of Ni₄Mo/TiO₂

The above results strongly suggest that the MSI between Ni₄Mo and the TiO₂ support modulate the electronic structure of Ni₄Mo, causing a down-shifted d band center, which would weaken the adsorption strength of simple intermediate/molecule on Ni₄Mo, such as H, O and CO, according to the d band theory⁵⁵. Parallel hydrogen-temperature-programmed desorption (H₂-TPD) results clearly demonstrate a lower desorption temperature on Ni₄Mo/TiO₂ than that on Ni₄Mo (Supplementary Fig. S24), signifying a weakened hydrogen binding strength on Ni₄Mo/TiO₂. In line with the reduced hydrogen binding strength, although it is difficult to obtain the specific activity due to the uncertainty in the electrochemical surface area (ECSA) determination, it is reasonable to argue that Ni₄Mo/TiO₂ would exhibit a higher specific activity than Ni₄Mo, as Ni₄Mo/TiO₂ has a similar Ni₄Mo particle size and mass activity, but smaller ECSA due to the coverage of TiO₂. Remarkably, oxygen-temperature-programmed oxidation (O₂-TPO) results exhibit a much higher oxidation temperature of Ni₄Mo/TiO₂ than Ni₄Mo (Supplementary Fig. S25), indicating a stronger oxidation resistance of Ni₄Mo/TiO₂, also in agreement with the hypothesis.

Quasi in situ electrochemical XPS and in situ electrochemical Raman experiments were performed to explore the electronic structures and surface conditions of the Ni₄Mo and Ni₄Mo/TiO₂ catalysts during the HOR process. After potential cycling between −0.1 to 0.2 V for a few cycles to active the surface, the Ni 2p_3/2 and Ti 2p level XPS spectra were taken at the open circuit potential (OCP) and oxidizing potentials up to 1.2 V (Supplementary Figs. S26 and S27). The Ni 2p_3/2 spectrum (Supplementary Fig. S26a) of Ni₄Mo collected at OCP shifts by −0.5 eV relative to the ex situ experimental data as the Ni atoms are negatively charged during the surface activation process, and can be described as the electron-enriched Ni metal (Ni^δ−)⁵⁶. The BE of Ni⁰ (Ni^δ−) 2p_3/2 in both Ni₄Mo and Ni₄Mo/TiO₂ stays constant at all potentials, while it is negatively shifted by 0.2-0.3 eV in Ni₄Mo/TiO₂ with respect to that in Ni₄Mo regardless of the applied potentials (Fig. 5a and Supplementary Fig. S26a, b). The Ni²⁺ content estimated from XPS data is 21% in Ni₄Mo at the initial OCP, and dramatically increases to 44% at 1.2 V, while the Ni²⁺ content is substantially reduced in Ni₄Mo/TiO₂ (21% at OCP to 33% at 1.2 V, Supplementary Fig. S26c). The correlation between the HOR polarization curves and the Ni⁰ (Ni^δ−) percentages shows that the severely decreased current on Ni₄Mo at 0.3 V matches relatively well with the decreased Ni⁰ (Ni^δ−) content (Supplementary Fig. S26d), indicating that Ni₄Mo deactivation is originated from Ni oxidation. The Ti⁴⁺ 2p_3/2 spectra demonstrate a ~ 0.2 eV positive BE shift from OCP to 1.2 V (Fig. 5b and Supplementary Fig. S27a). Interestingly, a new peak signal in Ti 2p spectra of Ni₄Mo/TiO₂ emerges at the BE lower than Ti⁴⁺ 2p_3/2 after the surface activation, which can be assigned to the reductive Ti³⁺ 2p_3/2 species⁴⁴. The Ti³⁺ 2p_3/2 peak shifts to higher BE (456.5 eV at OCP to 457.2 eV at 1.2 V), and decreases in percentage (34% at OCP to 22% at 1.2 V, Supplementary Fig. S27b) with increasing applied potential. Ti³⁺ has been verified to boost the efficient charge transfer from TiO₂ to metals by decreasing the work function of TiO₂ bulk^57,58, and thus may also play an important part in constructing the MSI of Ni₄Mo/TiO₂. The quasi in situ XPS experiments further confirm the existence of the charge transfer from TiO₂ to Ni₄Mo under the electrochemical conditions.

Fig. 5: Mechanistic analyses.

a, b Quasi in situ electrochemical XPS spectra, and c, d in situ electrochemical Raman spectra of Ni₄Mo and Ni₄Mo/TiO₂ collected at the selected potentials in 0.1 M NaOH during the HOR. The potentials are iR-corrected.

The in situ Raman spectra were also collected after the surface activation (Fig. 5c, d). The bands at 310 and 893 cm⁻¹, which could be respectively assigned to the Mo = O stretching mode and bending mode in the MoO₄²⁻ tetrahedron (see Supplementary Table S5 for the band assignment), exist in the full potential window investigated from the OCP to 1.2 V for both Ni₄Mo and Ni₄Mo/TiO₂, owing to the oxidation of Mo in Ni₄Mo to MoO₄²⁻⁵². The peak at 483 cm⁻¹ is indexed to the symmetric stretching mode of bridging Mo-O-Mo bond in Mo₂O₇²⁻, formed from the MoO₄²⁻ dimerization^52,59. Furthermore, the band at 460 cm⁻¹ starting from 0.1 V and the one at 939 cm⁻¹ starting from 1.0 V on Ni₄Mo (Fig. 5c), attributed to the Ni-OH symmetric stretching mode of Ni(OH)₂^60,61,62,63 and Mo-O stretching mode of NiMoO₄^59,64 respectively, indicate the formation of hydroxide and oxide of Ni and Mo during the HOR from 0.1 to 1.2 V. Interestingly, these peaks associated with Ni-OH or Mo-O vibrations disappear on Ni₄Mo/TiO₂, and only when the potential reaches 1.0 V is the Ni-OH vibration marginally visible (Fig. 5d), revealing that most of the Ni surface atoms likely exist in the form of Ni⁰ (Ni^δ−). These experimental results clearly suggest that Ni₄Mo/TiO₂ has great resistance to oxidation to endure the harsh oxidative conditions at high anodic potentials. This high oxidation resistance is likely due to the strong electronic modulation between Ni₄Mo and TiO₂, which renders a down-shifted d band center, and in turn weakened OH adsorption energy. DFT calculations also demonstrate a similar trend (Supplementary Fig. S28), in line with the hypothesis.

There have been debates as to the role of the surface hydroxyl species in the alkaline hydrogen oxidation on Pt group metal-based electrocatalysts, wherein the surface OH may (bi-functional mechanism⁹) or may not (HBE mechanism^4,65) directly participate in the HOR process. However, on Ni-based non-precious materials, it is rather difficult to differentiate the role of the surface OH, as Ni-based materials usually exhibit complicated surface conditions in a broad potential regime, and there is by far no experimental evidence regarding the direct participation of surface OH in the HOR yet. In our opinion, the surface OH most likely acts as the spectator, which deteriorates the HOR activity by blocking the active surface area. This speculation is supported by the observations that the HOR current decrease on Ni₄Mo and the metal surface oxidization occur concurrently (Fig. 1a and Supplementary Fig. S2), and the appearance of Raman bands associated with Ni(OH)₂ and NiMoO₆ in the Raman spectra of Ni₄Mo. Therefore, it becomes very straightforward to weaken the OH binding strength in designing the HOR catalysts to enhance their anti-oxidation ability. By introducing the MSI using TiO₂ as the support, the d band center of Ni₄Mo is downwards shifted, causing a weakened binding strength to surface O or OH. It is clearly seen from Supplementary Fig. S29, the CV of Ni₄Mo/TiO₂ is mainly composed of the capacitive current, demonstrating much mitigated surface oxidation. Therefore, we hypothesize that the surface OH plays the role as the blocking species on Ni₄Mo, and attribute the improved anti-oxidation ability of Ni₄Mo/TiO₂ to its much weakened OH adsorption strength.

CO-tolerance of Ni₄Mo/TiO₂

The modulated d band of Ni₄Mo by the TiO₂ support not only weakens the OH binding strength, but may also reduce the CO binding energy⁵⁵. As shown in Supplementary Fig. S30, the CO-stripping peak on Ni₄Mo/TiO₂ moves negatively to 0.50 V in comparison with 0.70 V on Ni₄Mo. Moreover, Ni₄Mo/TiO₂ exhibits HOR activity with a negligible decay in the presence of 2000 p.p.m CO, compared to a ~14% decrease in the HOR limiting current density of Ni₄Mo (Supplementary Fig. S31a). Chronoamperometry measurements at 0.2 V also demonstrate that the HOR current on Ni₄Mo/TiO₂ in the presence of 2000 p.p.m CO only decays 14% after 8000 s, much improved than a ~40% decay on Ni₄Mo (Supplementary Fig. S31b). The satisfying CO-tolerant capability of Ni₄Mo/TiO₂ is attributed to its weakened CO binding strength, which originates from the modulated electronic structure of Ni₄Mo by the TiO₂ support.

Applicability of the MSI to other Ni-based electrocatalyst

We further investigated whether the MSI between the metal catalyst and the TiO₂ support can be applied to other Ni-based electrocatalysts for the enhanced stability. Despite its excellent HOR activity, Ni₂W also loses the HOR activity at potentials higher than 0.2 V, in line with the surface oxidation in this potential regime (Fig. 6a). Optimized Ni₂W/TiO₂ (Ti/Ni = 0.46) not only exhibits good HOR activity at a low overpotential, but maintains the activity at a very small degradation rate (Fig. 6a and Supplementary Fig. S32, also see Methods, Supplementary Table S6, Supplementary Figs. S33–36 for detailed information on the synthesis and material structure). Although W is also found to slightly leach from the surface of Ni₂W nanoparticles in the alkaline electrolyte under the experimental conditions (Supplementary Table S7), long-term stability tests confirm that Ni₂W/TiO₂ can remain stable activity towards hydrogen oxidation at 1.2 V (Fig. 6b and Supplementary Fig. S37). This finding highlights the strategic importance of the MSI in tailoring the electronic structure of Ni-based electrocatalysts to elevate the HOR stability in alkaline electrolytes.

Fig. 6: HOR performance of Ni₂W/TiO₂.

a Positive-going sweeps of the cyclic voltammograms of Ni₂W and Ni₂W/TiO₂ recorded in H₂ and N₂-‍saturated 0.1 M NaOH at 1600 r.p.m with a scanning rate of 0.5 mV s⁻¹. The potentials are iR-corrected; and b chronoamperometry curves of Ni₂W and Ni₂W/TiO₂ in H₂-saturated 0.1 M NaOH at 1600 r.p.m. The potentials are not iR-corrected. The Ni loadings are 349 and 312 μg_Ni cm⁻²_geo for Ni₂W and Ni₂W/TiO₂.

In conclusion, by applying the MSI between NiM alloys and the TiO₂ support, we successfully synthesized Ni₄Mo/TiO₂ and Ni₂W/TiO₂ electrocatalysts with enhanced HOR stability. The catalysts not only maintain good HOR mass activities (10.1 ± 0.9 A g⁻¹_Ni for Ni₄Mo/TiO₂ and 6.8 ± 0.2 A g⁻¹_Ni for Ni₂W/TiO₂), but also show stable HOR current at as high as 1.2 V. AEMFC tests, using Ni₄Mo/TiO₂ as the anode catalyst, demonstrate a good performance with a peak power density of 520 mW cm⁻² and durability at 400 mA cm⁻² for nearly 100 h. Detailed structural and electronic analyses confirm the existence of the MSI through charge transfer from the TiO₂ support to NiM metals. Modulated electronic structure of NiM weakens the adsorption strength of H, O/OH and CO on NiM surfaces, rendering a high anti-oxidation ability of NiM/TiO₂. These results highlight the significance of the MSI in improving the catalyst stability, and evidence the intriguing promise of Ni₄Mo/TiO₂ as an efficient and robust anode catalyst, making a step forward in the applications of the AEMFC technology.

Methods

Material synthesis

The Ni₄Mo/TiO₂ catalysts were synthesized through a three-step method. Firstly, the NiMo hydroxide precursor was prepared by a solvothermal method¹⁴. In brief, 735 mg Ni(NO₃)₂·6H₂O (98%, Sinopharm) and 81 mg (NH₄)₆Mo₇O₂₄·4H₂O (99%, Alfa Aesar) were dissolved in 2.5 mL ultra-pure H₂O (18.2 MΩ cm, Millipore), followed by addition of 12.5 mL ethylene glycol (99%, Innochem) and 1 mL NH₃·H₂O (18% NH₃ basis, Sigma-Aldrich). After stirring for 2 h, the solution was transferred into a Teflon-lined stainless steel autoclave and heated at 190 °C for 1 h. When cooling down to the room temperature, the synthesized green powder was washed using ethyl alcohol (99.5%, Innochem) and ultra-pure H₂O (1:1 in volume) for five times, and collected by centrifugation. Secondly, six batches of the green precipitate were mixed with 100, 67, 50, 33, 25 and 20 mg TiO₂ (Aeroxide® P25, Acros Organic) respectively, and then dispersed in 50 mL ethyl alcohol by magnetically stirring for 20 h. Subsequently, the mixture was collected by vacuum filtration, and grounded in a mortar for 20 min after drying at 50 °C in vacuum (2.13 × 10⁴ Pa). Finally, the collected substance was heated up to 400 °C at 3 °C min⁻¹ in a reductive atmosphere (H₂/N₂ = 1:5 in volume) and remained at 400 °C for 1 h to obtain the Ni₄Mo/TiO₂ catalysts. Ni₄Mo was prepared using the same method without adding TiO₂ in the second step.

Ni₂W and Ni₂W/TiO₂ catalysts were synthesized by the same procedure, except that the initial Ni and W precursors were 735 mg Ni(NO₃)₂·6H₂O (98%, Sinopharm) and 177.5 mg (NH₄)₂WO₄ (99.99%, Alfa Aesar), and the final heat-treatment temperature was 500 °C.

Physical characterizations

Electron microscopy

High-resolution transmission electron microscopy (HR-TEM) measurement was conducted on the FEI Talos 200X and JEOL JEM-2100F transmission electron microscopes at the accelerating voltage of 200 kV. Aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) measurement was carried out on an atomic-resolution analytical microscope (thermalfisher scientific titan themsis Z) equipped with a probe spherical aberration corrector at an acceleration voltage of 300 kV. Energy dispersive X-ray spectroscopy was collected on the Oxford Instruments X-Max 80 T. The sample was made through dispersing the catalysts in ethyl alcohol by sonication. The dispersion was then dropped on an ultrathin carbon grid, and dried in air for few minutes.

Powder X-ray diffraction (XRD)

XRD measurement was performed on a D8 Advance X-ray diffractometer (Bruker) using Cu Kα radiation (λ = 0.15418 nm) at 40 kV and 40 mA. The data were collected with 2θ ranged from 10° to 90° at a scanning rate of 10° min⁻¹.

Raman spectroscopy

Raman experiment was carried out with a confocal microscope Raman spectrometer (DXR3, Thermo Scientific). The excitation wavelength was 532 nm.

X-ray photoemission spectroscopy (XPS)

The states of Ni, Mo, Ti and O in the catalysts were examined on a ThermoFischer ESCALAB 250Xi X-ray photoelectron spectrometer equipped with Al X-ray source (Al Kα, 1486.6 eV). All spectra were processed using the Shirley background correction, and calibrated with the C 1s component at 284.8 eV. The Gaussian–Lorentzian line shape was adopted to fit the spectra.

X-ray absorption fine structure spectroscopy (XAFS)

The XAFS measurements were performed at 1W1B station in Beijing Synchrotron Radiation Facility, operated at 2.5 GeV with a maximum current of 250 mA. The spectra were collected in the fluorescence mode using a Lytle detector. Samples were pelletized with diameter of 8 mm and thickness of 1 mm using the PVDF powder as the binder. The acquired Ni K-edge extended X-ray absorption fine structure (EXAFS) data were processed according to the standard procedures using the Athena and Artemis implemented in the IFEFFIT software packages⁶⁶. The EXAFS spectra were subtracted by the post-edge background from the overall absorption, and normalized with respect to the edge-jump step. The χ(k) data were Fourier transformed to real (R) space using a hanning window (dk = 1.0 Å⁻¹) to separate the EXAFS contributions from different coordination shells. Least-squares curve fitting was performed using the Artemis module of IFEFFIT software packages to obtain the quantitative structural parameters. In the fitting, the amplitude reduction factor S₀² was fixed, and the internal atomic distances R, Debye-Waller factor σ², and the edge-energy shift ΔE₀ were allowed to run freely.

Ultraviolet photoemission spectroscopy (UPS)

Surface electronic structure of the catalysts was investigated on the PHI5000 VersaProbe III electron spectrometer (Scanning ESCA Microprobe) at UV photon energy of 21.2 eV (He I) under ultra-high vacuum (4 ×10⁻⁶ Pa). The total energy resolution was 0.10 eV. Shirley background was subtracted as described in the previous study⁴⁸. The band between −3 eV and −10 eV centered at −4.9 eV for the Ni₄Mo sample was assigned to the O 2p states due to inevitable oxidation upon exposure to air⁴⁵, while the band in the same window centered at −6.0 eV for the Ni₄Mo/TiO₂ sample was O 2p states mainly originated from TiO₂⁴⁴. The bands ranged from 0 to −3 eV in both samples were assigned to the Ni 3d photoemission⁴⁶. The d band center energy relative to the Fermi level was calculated based on the following equation^48,49,50

$${\varepsilon }_{{{\rm{d}}}}=\int N(\varepsilon )\varepsilon {{{\rm{d}}}} \varepsilon \Big/\int N(\varepsilon ) {{{\rm{d}}}}\varepsilon$$

(1)

where N(ε) is the density of states, and ε is the energy of states.

For the work function calculation, the valence band spectra were also collected at the same spectrometer with a − 5 V bias voltage applied to the sample. The work function ($\varphi$) could be calculated from the following equation⁶⁷

$$\varphi=h\nu -{E}_{{{\rm{cutoff}}}}$$

(2)

where hν is the energy of the UV photon (He I, 21.2 eV), and E_cutoff is the energy of the secondary-electron cutoff.

H₂-temperature programmed desorption (H₂-TPD)

Surface property towards H adsorption/desorption of the catalyst was measured using a ChemBET instrument (Quantachrome). The signal was determined by a thermal conductivity detector (TCD). 100 mg catalyst was placed in a quartz tube, and heated from the room temperature up to 400 °C at a ramping rate of 10 °C min⁻¹ with flowing He. The sample was kept at 400 °C for 30 min, and then cooled down to the room temperature under He-flow. Subsequently, H₂ was introduced into the tube for adsorption until a stable TCD signal. After removing H₂ with flowing He, the sample was heated up to 400 °C with a ramping rate of 10 °C min⁻¹.

O₂-temperature programmed oxidation (O₂-TPO)

Surface property towards binding strength of oxygen-related species was measured on the same instrument as H₂-TPD. 100 mg catalyst was placed in a quartz tube, and heated from the room temperature up to 400 °C at a ramping rate of 10 °C min⁻¹ with flowing He. The sample was kept at 400 °C for 30 min, and then cooled down to the room temperature under He-flow. Subsequently, the sample was gradually oxidized from the room temperature to 400 °C at a ramping rate of 10 °C min⁻¹ in the presence of 3% O₂ in He.

Inductively coupled plasma-optical emission spectrometry (ICP-OES)

The chemical contents of Ni, Mo and Ti were examined by ICP-OES on Agilent-730-OES. In brief, the catalyst was mixed with 5 mL HNO₃ (67%, Sinopharm), 1 mL HF (40%, Sinopharm) and 1 mL HCl (37%, Sinopharm) in a Teflon-lined stainless steel autoclave, followed by heating at 180 °C for 8 h. After cooling down to the room temperature, the solution was transferred to a 25-mL volumetric flask, and diluted with ultra-pure H₂O to the metered volume for ICP measurement.

Inductively coupled plasma mass spectrometry (ICP-MS)

ICP-MS (Aglient-7700) was used to detect the Ni, Mo, W and Ti dissolution in the alkaline electrolyte after electrochemical measurements.

Electrochemical measurement

All electrochemical measurements were performed on the electrochemical workstation (VSP-300, Biologic). The temperature was set at 25 ± 0.5 °C, unless otherwise emphasized. The catalyst ink was prepared by dispersing and sonicating the catalyst in a mixture of 750 μL isopropanol (99.5%, Innochem), 200 μL ultra-pure H₂O and 50 μL Nafion (5 wt%, Sigma-Aldrich) with a final concentration of 6-8 mg_Ni mL⁻¹. 10 μL catalyst ink was deposited on a glassy carbon electrode (5 mm in diameter, Tianjin Aida), which was pre-polished with 50 nm alumina slurry (99.0%, Tianjin Aida), and dried in air at the room temperature, resulting in the Ni loading of ~400 μg_Ni cm⁻²_geo (see Supplementary Tables S1 and S6 for the specific loadings). The catalyst thin film electrode was then mounted onto a rotator (Pine Instrument), serving as the working electrode. A KCl-saturated calomel electrode (SCE, Tianjin Aida) and a graphite rod (spectral purity, Tianjin Aida) were used as the reference and counter electrodes, respectively. All potentials reported in this paper were referenced to the reversible hydrogen electrode (RHE), which was calibrated by measuring the HER/HOR on a Pt disk (5 mm in diameter, Pine Instrument) in H₂-saturated 0.1 M NaOH (99.99% metal trace, Sigma-Aldrich).

Cyclic voltammetry measurement

Surface properties of the as-prepared electrocatalysts were investigated using the cyclic voltammetry (CV) method from −0.3 V to 1.8 V versus RHE with a rotating speed of 1600 r.p.m and a scanning rate of 20 mV s⁻¹ in N₂-saturated 0.1 M NaOH.

Rotating disk electrode measurement

HOR performances of the as-prepared electrocatalysts were examined using the RDE technique. The catalyst thin film electrodes were first activated by CV in H₂-saturated 0.1 M NaOH between −0.1 and 0.2 V versus RHE at 20 mV s⁻¹ for several cycles until a steady polarization curve was obtained. The HOR polarization curves were then collected with a rotating speed of 400, 900, 1600 and 2500 r.p.m and a scanning rate of 0.5 mV s⁻¹ to minimize the capacitive charge contribution.

The HOR kinetic current (i_K) was calculated based on the Koutecky–Levich equation⁶⁸

$$\frac{1}{i}=\frac{1}{{i}_{{{{{\rm{K}}}}}}}+\frac{1}{{i}_{{{{{\rm{D}}}}}}}$$

(3)

where i is the measured current, and i_D is the diffusion limited current.

The HOR exchange current (i₀) was obtained subsequently by fitting i_K to the Butler–Volmer equation⁶⁹

$${{i}}_{{{{{\rm{K}}}}}}={{i}}_{0}\left({{e}}^{\frac{{{{{{{\rm{\alpha }}}}}}}_{{{{a}}}}{F}}{{RT}}{\eta }}-{{e}}^{\frac{-(1-{{{{{{\rm{\alpha }}}}}}}_{{{{a}}}}){F}}{{RT}}{\eta }}\right)$$

(4)

where i₀ is the HOR exchange current, α_a is transfer coefficient for the HOR, F is the Faraday constant (96485 C mol⁻¹), R is the universal gas constant (8.314 J mol⁻¹ K⁻¹), T is the temperature, and η is the overpotential. α_a was between ~0.1 and ~0.4.

Activation energy (E_a) of the HOR was measured at different temperatures from 2 °C to 30 °C. The E_a can be calculated from the Arrhenius equation

$$\log {i}_{0}=\frac{-{E}_{{{{a}}}}}{{{{{\mathrm{ln}}}}}10\times R\times T}{{{{\rm{+const.}}}}}$$

(5)

where i₀ (mA cm⁻²) is the exchange current density at different temperatures, E_a is the activation energy (J mol⁻¹), R is the universal gas constant (8.314 J mol⁻¹ K⁻¹), and T is the temperature (K).

Chronoamperometry measurement

Chronoamperometry (CA) experiment was taken at a constant potential for 8000 s in H₂/N₂-saturated 0.1 M NaOH with a rotating speed of 1600 r.p.m for stability test.

CO stripping and CO tolerance measurement

CO stripping was conducted using the CA and CV methods. The potential was held at 0.1 V versus RHE for 30 min, during which time the electrolyte was purged with CO, allowing for complete CO adsorption, followed by 20 min-purging with N₂ to remove the remaining CO in the electrolyte. Then CO stripping was performed by taking CV from 0 to 1.2 V versus RHE at a scanning rate of 20 mV s⁻¹ for 2 cycles. CO tolerance experiment was performed using the RDE technique and CA method in H₂-saturated 0.1 M NaOH at 0.2 V versus RHE in the presence of 2000 p.p.m CO.

Electrochemical impedance spectroscopy (EIS) measurement

Solution resistance was obtained after each RDE test by EIS measurement. The EIS spectra were taken at 0 V vs. open circuit potential (OCP) with a 10 mV voltage perturbation, and the frequency was from 100 mHz to 200 kHz. The real part of the resistance at 1 kHz was taken as the solution resistance (R ≈ 40 Ω).

In situ electrochemical Raman spectroscopy test

The in situ electrochemical Raman test was performed on a Renishaw Via Raman microscope with a 532 nm excitation wavelength and a 50× objective. The sample ink was prepared by dispersing the catalyst in a mixture of ultra-pure H₂O, isopropanol and Nafion, and then dropped on a piece of carbon paper (CP, 15 × 15 mm²), serving as the working electrode. Prior to collecting the Raman data, the Ni₄Mo and Ni₄Mo/TiO₂ working electrodes were scanned between −0.1 and 0.2 V versus RHE for 10 cycles in H₂-saturated 0.1 M NaOH in a home-made Raman cell (Supplementary Fig. S38), equipped with a platinum counter electrode and a SCE reference electrode. The solution resistance was ~26 Ω for the in situ Raman setup. The working electrode was held at each potential (OCP and 0–1.5 V versus RHE at an interval of 0.1 V) for 2 min for the Raman spectra collection. The laser intensity was set to be 10% with a collection time of 50 s for each spectrum.

Quasi in situ electrochemical X-ray Photoemission Spectroscopy test

Quasi in situ electrochemical XPS measurement was carried out on the ThermoFischer ESCALAB 250Xi instrument by applying a monochromatic Al Kα X-ray source (1486.8 eV) operating at 12.5 kV and 16 mA under ultra-high vacuum (8 × 10⁻¹⁰ Pa). The total energy resolution was 0.10 eV. The catalyst ink was prepared by dispersing the catalyst in a mixture of ultra-pure H₂O, ethyl alcohol and Nafion, and then dropped on a glassy carbon (GC) electrode (4 mm in diameter), serving as the working electrode. Before collecting the XPS data, the Ni₄Mo and Ni₄Mo/TiO₂ working electrodes were scanned between −0.1 and 0.2 V versus RHE for 10 cycles in H₂/N₂ mixture-saturated 0.1 M NaOH in a home-made cell (Supplementary Fig. S39), equipped with a platinum counter electrode and a SCE reference electrode. The solution resistance was ~10 Ω for the quasi in situ XPS setup. The working electrode was first held at each potential (OCP, 0, 0.2, 0.3, 0.4, 0.5, 0.6, 0.8, 1.0 and 1.2 V versus RHE) for 2 min, then vacuumed in the preparation chamber, and finally transferred to the test chamber for XPS spectrum collection without exposure to the air. All spectra were processed using the Shirley background correction, and calibrated with the C 1s component at 284.8 eV. The Gaussian–Lorentzian line shape was adopted to fit the spectra.

Membrane-electrode assembly and AEMFCs test

The as-synthesized Ni₄Mo or Ni₄Mo/TiO₂ was used as the anode catalyst and the commercial Pt/C (60 wt%, Johnson-Matthey) was used as the cathode catalyst. The self-designed QAPPT [quaternary ammonia poly (N-methyl-piperidine-co-p-terphenyl)] was applied as both anion exchange membrane and ionomer in the electrodes⁷⁰. For the anode catalyst, the ink was prepared by adding Ni₄Mo or Ni₄Mo/TiO₂, Vulcan XC-72 carbon and 20 mg mL⁻¹ ionomer (16 wt%) solution into isopropanol and then sonicated for 30 min. The weight ratio of Ni₄Mo or Ni₄Mo/TiO₂ to Vulcan XC-72 carbon was 4:1. For the cathode catalyst, Pt/C and the 20 wt% ionomer solution was mixed in isopropanol and then sonicated for 15 min. To make the catalyst-coated membrane (CCM), the catalyst ink was sprayed onto the membrane (25 μm), which was heated at 70 °C to remove the isopropanol. The anode and cathode catalyst loadings were 1.35 mg_Ni cm⁻² and 0.4 mg_Pt cm⁻² respectively, and the catalyst sprayed area was fixed to 4 cm². In order to exchange the anion of the membrane and ionomer to OH⁻, the CCM was soaked in 1 M KOH at 60 °C for 24 h, and meanwhile hydrogen was purged into the KOH solution to prevent the catalyst oxidation. Subsequently, the CCM was washed with ultra-pure H₂O for several times to remove the excess KOH. The membrane electrode assembly (MEA) was assembled by placing the CCM between two pieces of carbon paper (AvCarb GDS3250) used as gas diffusion layer without hot-pressing. The AEMFC performance test was conducted on an 850E Multi Range fuel cell test station (Scribner Associates, USA). The test was operated at 80 °C under H₂ and O₂ condition with a backpressure of 0.2 MPa on the anode and cathode. H₂ and O₂ humidified at 80 °C (100% RH) were supplied to the anode and cathode compartments with a flow rate of 1000 sccm. The durability test was conducted with the flow rate of H₂ and O₂ being 300 and 500 sccm under otherwise identical conditions.

Density functional theory calculations

The density functional theory (DFT) calculations were carried out using the Vienna Ab-initio Simulation Package (VASP)^71,72 with the projector augmented wave (PAW) method⁷³ and the Perdew-Burke-Ernzerhof (PBE)^74,75 exchange-correlation functional. Dispersion interactions were described using the DFT-D3 method proposed by Grimme^76,77. The hetero-structure was created by the vaspkit code⁷⁸. A 1 × 1 × 1 k-point grid with a cutoff energy of 450 eV was used for the slab optimization and the density of states (DOS) calculation. Convergence was reached when the change in energy per electron step was less than 1 × 10⁻⁵ eV. The conjugate gradient algorithm was used to relax all optimizable atoms until the force applied on them was less than 0.02 eV Å⁻¹ for the slabs.

The anatase TiO₂ (100) surface and Ni₄Mo (211) surface were chosen for a better lattice match to create the hetero-structure using the slab model with a vacuum layer along the z-axis set to 15 Å. Two layers of TiO₂ (100) $(2\sqrt{5}\times \sqrt{7})$ surface were used as the substrate, and frozen during the optimization procedure, while the Ni₄Mo (211) $(\sqrt{10}\times 3)$ surface with a thickness of 6.15 Å above the TiO₂ substrate was cut by $\frac{1}{3}$ along the y-axis to expose the interface of the hetero-structure, which was allowed for optimization during the whole calculation. The lattice mismatch for this hetero-structure was kept as low as 0.011%.

Adsorbates were added on the interface of the hetero-structure. The binding energy of H atom and OH species was calculated using the flowing formula:

$${E}_{{{{{\rm{H}}}}} \, {{{{\rm{adsorb}}}}}}={E}_{{{{{\rm{slab}}}}}+{{{{\rm{H}}}}}}-{E}_{{{{{\rm{slab}}}}}}-\frac{1}{2}{E}_{{{{{\rm{H}}}}}2}$$

(6)

$${E}_{{{{{\rm{OH}}}}} \, {{{{\rm{adsorb}}}}}}={E}_{{{{{\rm{slab}}}}}+{{{{\rm{OH}}}}}}{-E}_{{{{{\rm{slab}}}}}}-{E}_{{{{{\rm{OH}}}}}}$$

(7)

Where ${E}_{{{{{\rm{H}}}}} \, {{{{\rm{adsorb}}}}}}$ and ${E}_{{{{{\rm{OH}}}}} \, {{{{\rm{adsorb}}}}}}$ are the adsorption energies of H atom and OH species, ${E}_{{{{{\rm{slab}}}}}+{{{{\rm{H}}}}}}$ and ${E}_{{{{{\rm{slab}}}}}+{{{{\rm{OH}}}}}}$ are the total energies of the compositions, ${E}_{{{{{\rm{H}}}}}2}$ and ${E}_{{{{{\rm{OH}}}}}}$ are the energies of hydrogen molecule and OH species, and ${E}_{{{{{\rm{slab}}}}}}$ is the energy of the slab.