Theoretical Viewpoint. As shown in Fig. 2, interactions between a metal and a support with different Fermi levels (Ef) result in electron transfer until the system reaches equilibrium, inducing a BEF at the interface between the metal and the support.24,25 Given that H2O is a typical polar molecule, H2O molecules of hydrogen-bond network on catalyst surface can be easily polarized by such BEF.3,26 Polarized H2O molecules undergo rearrangement from a random to a uniform distribution, thereby increasing hydrogen-bond connectivity around catalysts and, hence, proton relay. Based on this analysis and the equivalence between Ef and Φ,27 we hypothesized that ΔΦ = |Φmetal − Φsupport| determines proton relay in the hydrogen-bond network around metal-supported catalysts. Accordingly, a high ΔΦ may enhance proton relay, enabling synergetic H2O dissociation between the metal and the support for efficient HER catalysis.
Design and Synthesis of Model Catalysts. We selected ruthenium (Ru) nanoparticles (Φ[Ru] = 4.81 eV, Supplementary Fig. 1) and cobalt phosphide (CoP) nanosheets (Φ[CoP] = 5.21 eV, Supplementary Fig. 2) to design Ru/CoP catalysts with a high ΔΦ of 0.4 eV (Fig. 3a). These catalysts should promote proton relay in their surrounding hydrogen-bond network for synergetic H2O dissociation between Ru and CoP, affording high HER activity. Subsequently, we designed control catalysts by combining Ru nanoparticles (Φ[Ru] = 4.81 eV) and cerium-doped CoP (CeCoP) nanosheets with doping levels of 1.0 (Φ[Ce0.01CoP] = 5.07 eV), 2.7 (Φ[Ce0.027CoP] = 4.85 eV) and 4.0 wt.% (Φ[Ce0.04CoP] = 4.68 eV, Supplementary Figs. 3–5) and ΔΦ values of 0.26, 0.04 and 0.13 eV for Ru/Ce0.01CoP, Ru/Ce0.027CoP and Ru/Ce0.04CoP, respectively (Fig. 3a). The smaller ΔΦ of these catalysts likely limits proton relay in their surrounding hydrogen-bond network and synergetic H2O dissociation between Ru and CeCoP, decreasing HER activity. Here, the comparison of Ru/CoP with various Ru/CeCoP model catalysts provides the opportunity to examine our view.
We synthesized Ru/CoP catalysts in a two-step process by (i) preparing CoP nanosheets and (ii) loading Ru nanoparticles on CoP nanosheets using a wet chemical method (see details in the Methods section). By inductively coupled plasma mass spectrometry (ICP-MS) analysis, we found that Ru/CoP contained ~ 1.0 wt.% total Ru (Supplementary Table 1). In addition, powder X-ray diffraction (XRD) patterns of Ru/CoP were indexed to those a CoP standard (JCPDS #29–0497) due to the low Ru loading (Supplementary Fig. 6).28 Transmission electron microscopy (TEM) images further revealed that Ru nanoparticles of ~ 1.6 nm were uniformly distributed on CoP nanosheets (Fig. 3b). High-resolution TEM images displayed two ordered lattice fringes, with distances of 2.27 and 2.82 Å matching the (100) and (011) crystal planes of Ru and CoP, respectively (Fig. 3c).29 Energy-dispersive X-ray (EDX) elemental mapping (Fig. 3d) highlighted the uniform distribution of Co and P in nanosheets and the high concentration of Ru in nanoparticles, confirming the structural features of Ru/CoP. In X-ray photoelectron spectroscopy (XPS) profiles, we identified the Ru 3d signal at 279.7 eV and Co 2p and P 2p signals at 778.9 and 129.4 eV, in line with Ru0 benchmarks and with Co and P species in CoP, respectively (Fig. 3e).29 Combined, these characterization results demonstrated that we had successfully synthesized the desired Ru/CoP catalysts. Using a similar strategy, various Ru/CeCoP catalysts were also prepared and showed similar structural features, including particle size/loading and morphology, except for the chemical composition of the CeCoP support (Supplementary Fig. 7), which enables a straightforward catalyst comparison only by considering ΔΦ.
To examine the HER activity of Ru/CoP and various Ru/CeCoP model catalysts, we recorded linear sweep voltammetry (LSV) profiles on a standard three-electrode system with an Ar-saturated 1.0 M KOH electrolyte (Fig. 4a and 4b). As expected, Ru/CoP displayed a much higher HER activity with a η10 of 6 mV and a Tafel slope of 23.1 mV dec− 1, outperforming commercial 20 wt.% Pt/C benchmarks and state-of-the-art HER electrocatalysts in alkaline media (Supplementary Table 2). Especially when normalizing to Ru weight loading, Ru/CoP achieved a record high Ru utilization activity of 66.1 A mgRu−1, considerably surpassing previously reported Ru-based electrocatalysts under similar conditions (Fig. 4c and Supplementary Table 2).
The catalytic current density of Ru/CoP did not decay over 500 h (Fig. 4d). During this period, Ru/CoP reached approximately 100% Faraday efficiency (FE) for HER. Based on the analysis of spent electrocatalysts, no structural and componential changes were observed in the long-term tests (Supplementary Fig. 8). Metal leaching was assessed by ICP-MS, showing total concentrations of leached Ru and Co ions of only 0.5 and 0.6 ppm in the electrolyte after 500 h of HER (Supplementary Fig. 9). These findings underscore the robustness of Ru/CoP as HER catalysts and their potential for industrial H2 production.
Correlation between Δ Φ and HER Activity. We correlated ΔΦ values with HER catalytic parameters (η10 and Tafel slopes) of Ru/CoP and various Ru/CeCoP model catalysts (Fig. 5a and 5b). The plots of observed η10 and ΔΦ values of these catalysts displayed a nearly linear, decreasing trend, which was expressed as η10 = 167.9–409.1 ΔΦ. The derived Tafel slopes of these catalysts also decreased linearly with ΔΦ, as expressed by the following equation: Tafel slope = 150.0–324.3 ΔΦ. These strong correlations support our hypothesis that high ΔΦ promotes synergetic H2O dissociation, increasing HER activity.
To examine this concept, the Ru/CoP with the highest ΔΦ and HER activity and the Ru/Ce0.027CoP with the lowest ΔΦ and HER activity were further analysed. Their difference in HER activity may derive from four factors: (i) Ru/CoP has more catalytic sites thanks to morphological changes; (ii) CoP serves as the primary contributor through electronic metal-support interactions; (iii) synergy effects of interfacial Ru-CoP sites are the leading contributions; (iv) proton relay in the hydrogen-bond network on Ru/CoP enables synergetic H2O dissociation between Ru and CoP.
To assess whether Ru/CoP has more catalytic sites, the electrochemical surface area (ECSA) of these catalysts was investigated in cyclic voltammetry (CV) profiles with various scan rates to calculate the double layer capacitance (Cdl).30,31 As shown in Supplementary Fig. 10, the ECSA values of Ru/CoP (915 cm2) and Ru/Ce0.027CoP (945 cm2) was similar. Thus, we ruled out the possibility that the HER activity of Ru/CoP was higher due to the increased number of catalytic sites.
Considering electronic interactions between Ru and CoP, CoP in Ru/CoP could primarily account for HER activity. To examine the validity of this hypothesis, we applied thiocyanate ion (SCN−) as a probe molecule to selectively block the surface of loaded Ru.29 If CoP was the main contributor for HER activity, such SCN− treatment should bring insignificant influences in HER activity. However, the HER activity of Ru/CoP after SCN− treatment is sharply degraded to that of bare CoP (Supplementary Fig. 11). Therefore, we also ruled out this possibility.
Assuming that the Ru-CoP interfacial sites play key roles in Ru/CoP catalysts, Ru loading should linearly correlate with the number of Ru-CoP interfacial sites and with HER activity as long as Ru size did not change with Ru loading. At various Ru loadings, the Ru size of Ru/CoP catalysts remained almost unchanged at ~ 1.62 nm (Supplementary Fig. 12). Given the nonlinear relationship between Ru loading and HER activity (Supplementary Fig. 13), such Ru-CoP interfacial sites in Ru/CoP catalysts unlikely contribute profoundly to the improved HER activity. After excluding possibilities i-iv, we concluded that synergetic H2O dissociation mediated by proton relay in hydrogen-bond networks is the most plausible explanation for the improved HER activity of Ru/CoP due to its high ΔΦ.
Experimental Evidence. If such synergetic mechanism occurs, hydrogen adsorption on Ru/CoP catalysts should substantially change. To examine hydrogen adsorption, we acquired electrochemical impedance spectra (EIS) of Ru/CoP and Ru/Ce0.027CoP catalysts at different overpotentials. The recorded Nyquist plots were simulated by a double-parallel equivalent circuit model (Supplementary Fig. 14 and Table 3).32 We analysed hydrogen adsorption on the catalyst surface based on the second parallel components (Cφ and R2), where Cφ and R2 are hydrogen adsorption pseudo-capacitance and resistance. By integrating Cφ vs. potential (η), we determined the hydrogen adsorption charge (QH) and consequently the amount of adsorbed H on the catalyst surface (Fig. 6a).32 Under the initial electrolyte conditions, the QH of Ru/CoP (6357 µC) was exponentially higher than that of Ru/Ce0.027CoP (1189 µC). After introducing dimethyl sulfoxide (DMSO) into the electrolyte to break hydrogen bonds between H2O molecules,21,33 the QH value of Ru/CoP was significantly decreased from 6357 to 1239 µC. These findings strongly suggested the successful synergetic H2O dissociation between Ru and CoP mediated by proton relay in hydrogen-bond network, which can serve as a much improved H supply path.
In addition to the amount of adsorbed H, the corresponding hydrogen adsorption kinetics should also change. By correlating log R2 vs. η and calculating the EIS-derived Tafel slopes, we assessed hydrogen adsorption kinetics (Fig. 6b).34 Under the initial electrolyte conditions, Ru/Ce0.027CoP provided two kinetic responses (a 110.2 mV dec− 1 Tafel slope at overpotentials ranging from 0 to 50 mV and a 163.0 mV dec− 1 Tafel slope at overpotentials ranging from 50 to 130 mV), again confirming proton relay failure in the hydrogen-bond network around Ru/Ce0.027CoP and the occurrence of two individual H2O dissociation processes on Ru and Ce0.027CoP. By contrast, Ru/CoP delivered only one kinetic response (Tafel slope = 47.5 mV dec− 1) in the initial electrolyte environment, albeit with two kinetic responses (a 109.8 mV dec− 1 Tafel slope at overpotentials ranging from 0 to 50 mV and a 161.1 mV dec− 1 Tafel slope at overpotentials ranging from 50 to 120 mV) in the DMSO/electrolyte environment. Such a phenomenon revealed that the individual H2O dissociation processes on Ru and CoP were integrated into a synergetic mechanism of H2O dissociation between Ru and CoP through proton relay in the hydrogen-bond network around Ru/CoP.
To gather direct evidence, we performed operando cyclic voltammetry (CV), thus monitoring OH and H intermediates in situ during H2O dissociation (Fig. 6c).35,36 As blanks, we synthesized reduced graphene oxide-supported Ru nanoparticles (Ru/rGO, Supplementary Fig. 15), CoP and Ce0.027CoP. With the initial electrolyte, Ru/rGO displayed weak OH and H characteristic peaks at 0.60 and 0.29 V vs. RHE, whereas CoP and Ce0.027CoP showed the unconspicuous characteristic OH and H peaks at 0.64 and 0.20 V vs. RHE,35,36 which could be responsible for their limited H2O dissociation capability. Ru/Ce0.027CoP showed similar peaks at 0.20, 0.29, 0.60 and 0.64 V vs. RHE. The peaks at 0.60 and 0.29 V vs. RHE corresponded to Ru − OH and Ru − H signals, whereas the peaks at 0.64 and 0.20 V vs. RHE corresponded to Ce0.027CoP−OH and Ce0.027CoP−H signals. These findings supported H2O dissociation on Ru and Ce0.027CoP in Ru/Ce0.027CoP due to failed proton relay in the hydrogen-bond network. Comparatively, Ru/CoP yielded much stronger Ru-OH and CoP-H signals at 0.60 and 0.20 V vs. RHE. These observations demonstrated synergetic H2O dissociation between Ru and CoP, in which OH of H2O on Ru was readily transferred to Ru, followed by timely proton relay from H2O on Ru to H2O on CoP in the hydrogen-bond network and facile H transfer from H2O on CoP to CoP. In the DMSO/electrolyte environment, the previously strong Ru − OH and CoP − H signals vanished, and the much weaker Ru − OH, Ru − H, CoP − OH and CoP − H signals reappeared as DMSO blocked proton relay in the hydrogen-bond network, preventing synergetic H2O dissociation between Ru and CoP and forcing separate H2O molecules to inefficiently dissociate on Ru and CoP, respectively.
To gather further evidence, we also performed operando attenuated total reflectance-surface-enhanced infrared absorption spectra (ATR-SEIRAS) at a HER potential of − 0.05 V vs. RHE to in situ monitor the signal of H intermediates on Ru/Ce0.027CoP and Ru/CoP catalysts (Fig. 6d). As shown in the references, the vibrational frequencies of adsorbed H on Ru and CoP are ~ 1850 and ~ 2000 cm-1, respectively.29,37 Under the initial electrolyte conditions, Ru/Ce0.027CoP showed very weak Ru − H and Ce0.027CoP−H signals at 1861 and 2008 cm-1, confirming the sluggish H2O dissociation on Ru and Ce0.027CoP and thus insufficient H supply. For Ru/CoP, the Ru − H signal at ~ 1850 cm-1 completely disappeared, and the CoP − H signal at 2010 cm− 1 became much stronger, suggesting a synergistically enhanced H2O dissociation between Ru and CoP, which provided enough H on CoP. With the DMSO/electrolyte, very weak Ru − H and CoP − H signals appeared again, confirming that this synergistic H2O dissociation was mediated by proton relay in the hydrogen-bond network around Ru/CoP. These findings matched our operando CV results.
Mechanistic Insights. To understand why this synergistic H2O dissociation occurred, we performed Ab initio molecular dynamics (AIMD) at room temperature (300 K) (see Methods for details on the computational simulation and model construction) to visualize H2O molecules in double-layer microenvironments around Ru/CeCoP and Ru/CoP models (Figs. 7a and 7b). On the Ru/CeCoP model, O-down H2O is located ~ 2.36 Å from the Ru side, and H-down H2O is located ~ 3.44 Å from the CeCoP side. O- and H-down H2O molecules are located closer to Ru (~ 2.20 Å) and CoP (~ 3.16 Å), respectively. Moreover, H2O molecules of the hydrogen-bond network on Ru/CoP are more uniformly arranged than those of Ru/CeCoP (Fig. 7c). This arrangement increases their density (especially of the H2O molecules closest to the catalyst surface) in the double-layer region and hence hydrogen-bond connectivity.
To experimentally rationalize the AIMD-simulated double-layer microenvironments around Ru/CeCoP and Ru/CoP, we probed the O-H stretching signal of H2O molecules on Ru/rGO, Ru/Ce0.027CoP and Ru/CoP catalysts by operando ATR-SEIRAS (Fig. 7d), which can be deconvoluted into four components. The signals located near 3200 and 3000 cm-1 represented two hydrogen-bonded and strongly hydrogen-bonded H2O molecules, with H- and O-down configurations respectively, distributed in the closest double-layer region to the catalysts.11,26,38–41 The signal located near 3400 cm-1 represented four hydrogen-bonded H2O molecules distributed in the double-layer region above that of two hydrogen-bonded and strongly hydrogen-bonded H2O molecules.18,41–44 The Ru/rGO showed a significant peak of strong hydrogen-bonded H2O molecules at 3027 cm-1, confirming the O-down H2O configuration on the Ru surface. In contrast to Ru/rGO spectra, Ru/Ce0.027CoP and Ru/CoP spectra showed elevated peaks of two hydrogen-bonded H2O molecules at 3199 and 3150 cm-1, revealing a H-down H2O configuration on CoP and Ce0.027CoP surfaces. In Ru/CoP spectra, all peaks were downshifted by 50 cm-1 in relation to the Ru/Ce0.027CoP peaks, and the peak of four hydrogen-bonded H2O molecules at 3351 cm-1 became the main peak, which indicated that hydrogen-bond connectivity around Ru/CoP was higher than that around Ru/Ce0.027CoP.43,44 Therefore, it can be seen that the salient consistency between the AIMD simulation and experimental spectroscopy results unequivocally substantiates the optimized hydrogen-bond network in the double-layer microenvironment around Ru/CoP, which facilitated OH delivery from O-down H2O to Ru, proton relay from O-down H2O on Ru to H-down H2O on CoP and H delivery from H-down H2O to CoP to realize a synergistic H2O dissociation pathway between Ru and CoP. Certainly, apart from in situ spectroscopy, the direct visualization for double-layer microenvironment are also worth exploring to provide solid verification for real results, but this is still challenging.18,41
To further unravel the occurrence of synergistic H2O dissociation and its contribution to HER, we performed a kinetic analysis combining simulations and experiments. The transition states (TSs) of all possible H2O dissociation pathways on Ru/CoP were computed (Figs. 7e). Disregarding the synergistic pathway, H2O dissociation may occur at Ru (pathway i), Ru-CoP interface sites (pathway ii) and CoP (pathway iii), with TS energy barriers of 1.13, 1.21 and 1.43 eV. Such high energy barriers suggest failed H2O dissociation at these components, explaining why we observed less catalytic contributions of these components in experiments.
To examine synergistic H2O dissociation (pathway iv), we extracted a chain of the closest three H2O molecules to the Ru/CoP surface, consisting of O-down H2O(1) on Ru, H2O(2) mediator and H-down H2O(3) on CoP, from the AIMD-simulated hydrogen-bond network. Using this approach, we simplified pathway iv as OH delivery from O-down H2O(1) to Ru, proton relay from O-down H2O(1) to H-down H2O(3) and simultaneous H delivery from H-down H2O(3) to CoP. The energy barrier of this pathway was only 0.39 eV. It indicates that H2O molecules can be dissociated through this super-efficient pathway, which yields sufficient H on the Co top sites of CoP. Subsequently, these catalytic sites with hydrogen adsorption Gibbs free energy of 0.03 eV can realize the energetically favorable hydrogen desorption45 (Fig. 7f). Eventually, the much improved HER process takes place over Ru/CoP. To experimentally support the simulated results, we compared HER activation energies (Ea) over Ru/CoP without and with DMSO in the electrolyte (Fig. 7g). The exchange current densities were recorded at different temperatures to determine Ea from Arrhenius plots (Supplementary Fig. 16).32 Ea over Ru/CoP was much lower with the initial electrolyte (21.6 kJ mol‒1) than with the DMSO/electrolyte (38.1 kJ mol‒1). This decrease in Ea provides strong evidence supporting synergistic H2O dissociation on Ru/CoP and its capability to expedite catalytic HER.
Through electronic structures, we aimed to further understand the relationship between HER and the high ΔΦ of binary components. As shown in Fig. 7h, the comparison between charge density difference mappings of Ru/CeCoP and Ru/CoP highlights that Ru/CoP has an electron-poorer region at the metal and an electron-richer region at the support than Ru/CeCoP. Such charge distribution implies a much stronger Ru-to-CoP electron transfer, which should induce a much stronger BEF at the Ru-CoP interface of Ru/CoP.24
To support this assertion, the surface potential (Vs) and surface charge density (ρ) of Ru/Ce0.027CoP and Ru/CoP were further investigated by Kelvin Probe Force Microscopy (KPFM, Fig. 7i) and Zeta potential measurements (Fig. 7j) to evaluate their BEF intensity based on the model developed by Kanata46 and previous works47–49 (see details in Methods section). The BEF intensity of Ru/CoP was 5.0 times higher than that of Ru/Ce0.027CoP (Fig. 7j). Consequently, strong BEF at the Ru-CoP interface facilitates the formation of hydrogen bonds between H2O molecules around Ru/CoP and increases hydrogen-bond connectivity, promoting proton relay in the hydrogen-bond network between Ru and CoP. Furthermore, OH species with lone pair electrons primarily adsorb at the electron-deficient region50 whereas H species with unsaturated electrons on 1s orbital are easily trapped at the electron-rich region,51 so the much stronger electron-losing event on Ru and electron-enrich event on CoP induce O-down and H-down H2O adsorption configurations and shorten the delivery distance of OH or H to Ru or CoP.
Taken together, these findings indicate that the high ΔΦ of Ru/CoP triggers significant electron transfer from Ru to CoP, explaining the strong BEF at the Ru-CoP interface, the electron-deficient Ru and the electron-rich CoP. This electronic configuration simultaneously benefits OH delivery from O-down H2O to Ru, proton relay from O-down H2O on Ru to H-down H2O on CoP in the hydrogen-bond network and H delivery from H-down H2O to CoP to realize a synergetic H2O dissociation pathway. Through this pathway, the catalytic HER kinetics was significantly expedited, resulting in high HER activity. Both experiments and simulations show that engineering the critical parameter of ΔΦ yields advanced metal-supported alkaline HER electrocatalysts.
Practical Applications. The optimized Ru/CoP was practically utilized as both cathode and anode in the anion exchange MEA electrolyzer at 60°C in a 1.0 M KOH electrolyte environment (Fig. 8a and Supplementary Fig. 17). The polarization curve of this setup displayed superior HER activity, reaching a catalytic current density (j) of up to 1.0 A cm− 2 at a cell voltage as low as 1.72 V (Supplementary Fig. 18). Over 200 hours, this j had no decay, and the FE of H2 remained at ~ 100%, eventually producing approximately 15.1 mol H2 (Fig. 8b). To the best of our knowledge, Ru/CoP outperforms the results achieved from the most of the state-of-the-art electrocatalysts reported to date under similar operating conditions (Supplementary Table 4).
To assess the total costs of this technique, we analysed its H2 production performance at different j values (Fig. 8c). Based on these results, we performed a techno-economic analysis (TEA). In this analysis, we referenced the economic models of previous studies, 36 considering an electricity price of US$0.03 kWh− 1. Details are provided in Methods section. In the j range from 0.2 to 2.8 A cm− 2, the cost of producing 1 kg H2 using this technique meets global targets for “green” H2 production (US$2.5 kgH2 −1).52 The minimum cost (US$1.86 kgH2 −1) is achieved when this setup is operated at 0.6 A cm-2 , which is lower than that of the state-of-the-art MEA electrolyzer estimated using similar TEA model (Supplementary Table 4). These findings suggest that our Ru/CoP catalyst is a more competitive choice for industrial H2 production.