Introduction

Atomically dispersed metallic materials are of significant interests in various applications because of the maximum atom-utilization efficiency, unique electronic structure and tunable coordination environment1,2,3,4. Multimetallic, multi-scale atomic materials comprising multiple types of single atom (SA) or/and atomic cluster (AC) species, such as bimetallic dual-atom, medium-/high-entropy SA and heterogeneous SA-AC composites, can significantly expand the scientific and application potentials beyond the horizons of monometallic counterparts5,6,7. In recent years, strategies like metal-organic framework pyrolysis8,9, atomic layer deposition10,11 and atom self-assembly12,13,14 etc. were developed for synthesizing multimetallic atomically dispersed materials, which are of pivotal significance in chemical (e.g. electrocatalysis9,15, thermocatalysis16,17, organic synthesis18,19), physical (e.g. magnetics20, optics21,22, electronics23) and environmental (e.g. water treatment24,25, plastic degradation26) application areas.

Despite the recent surge of researches on multimetallic SA/AC composite based materials, significantly challenges still remain. First, simultaneously constructing high free-energy multimetallic SA and AC raises higher difficulty as compared to simply stabilizing monotype SA or AC moiety27,28. Apart from some highly sophisticated methodology29, in most cases the principles of synthesizing monometallic SA/AC cannot be extrapolated to another metal element. Previous studies routinely employed specie-by-specie synthesis processes for respectively fabricating each specie9,30,31, which are cumbersome and non-universal. Second, as higher multimetallic compositional space opened (i.e. ≥3 metal elements), the difficulties for precise and controllable synthesis increase exponentially. Currently there is a knowledge gap in how to integrate numerous SA or/and AC species of different elements and varied physical/chemical properties on one solid substrate. Third, the inherent synergy between symbiotic adjacent metal species in multimetallic SA/AC composite is of vital importance in determining the apparent functionality. The controllable method of creating intrinsic interaction among diverse multimetallic SA/AC is highly desired yet there is still no such a precedent in previous studies. Therefore, developing an efficient and general synthesis strategy for multimetallic SA/AC composite is urgently demanded and could bring a new repertoire of structure-property mechanisms and potential capabilities.

Hydrogen with merits of high mass energy density and cleanness is an ideal energy carrier to address global fossil energy shortage and environmental challenges. Electrocatalytic hydrogen evolution powered by renewable electricity is recognized as an eco-friendly way to produce green hydrogen32,33. Noble metals deliver attractive electrocatalytic activity and durability towards hydrogen evolution reaction (HER) yet suffer from low abundance and high costs. Therefore, the concept of multimetallic atomically dispersed material is highly promising for fabricating high-performance hydrogen electrocatalysts with minimal noble metal usage34,35,36.

Herein, we developed a general energy-selective-clustering methodology to fabricate a large library of carbon supported bi-/multi-metallic SA/AC materials. Driven by the discrepancy in cohesive energy of metals, various metal ions exhibit differentiated clustering tendencies on a nitrogen functionalized carbon nanobox (NCB) thereby achieving the symbiosis of multimetallic single atoms or/and atomic clusters with native inter-site interaction. A library comprising 23 types of bimetallic SA/AC composites, and 17 types of trimetallic, quinary-metallic, septenary-metallic SA/AC composites was fabricated involving 17 metal elements. Bimetallic MSARuAC (M = Fe/Co/Ni) were selected to demonstrate the high value of the library for hydrogen electrocatalysis utility. Moreover, through general composition modification, high-performance catalysts of FeSAPtAC and CrSAIrAC were efficiently developed for oxygen reduction and evolution reaction electrocatalysis, respectively. Our work will open new space for fabricating multimetallic atomically dispersed materials with significant potential for a wide range of applications.

Results

Mechanism of the energy-selective-clustering methodology

The formation process of metallic atomic aggregates (AA) from isolated metal precursors on a solid substrate is demonstrated in Fig. 1a. Essentially, the clustering of isolated metal atoms is dominated by the inherent metal cohesion and metal-support interaction (MSI)37,38,39. From the perspective of thermodynamics, the driving force of clustering is the cohesive energy (Ec) due to the lower free energy of aggregated bulk metal than that of isolated metal atoms. The resistance of clustering is the metal-substrate binding energy (Eb) because an isolated metal atom needs to overcome Eb for escaping from the original location before it can couple with another individual atom or incorporate into a multi-atom cluster40. Therefore, the clustering tendency of metal atoms is determined by cohesive energy and the binding energy (ΔE = Ec − Eb)41,42,43. As demonstrated in Fig. 1a, considering two elements of isolated metal atoms (M1, M2) on one substrate, the high temperature sintering can induce the metal with higher clustering tendency to form atomic cluster (AC). Simultaneously, the metal with lower cluster tendency would form more stable single atom (SA) species.

Fig. 1: Cohesive energy controlled selective-clustering mechanism of metals.
figure 1

a Mechanism interpretation for the energy controlled clustering of two metals to respectively form SA or AC. b HAADF-STEM images and corresponding EDS mapping demonstrating the evolution of Ru and Ni species on NCB upon sintering. c The k3-weighted Fourier transform Ni and Ru EXAFS spectra of the product (red profiles) in b and standard metal foils (yellow profiles). d Schematic of multimetallic atomic aggregate (AA = SA/AC) composites formation based on energy-selective-clustering mechanism.

Full size image

Based on this energy controlled selective clustering process, it is feasible to fabricate bimetallic composite of M1SA-M2SA, M1SA-M2AC or M1AC-M2AC by exploiting the different clustering tendencies of metals. Bimetallic pair of Ni and Ru was chosen to verified this principle. The substrate employed is a nitrogen functionalized carbon nanobox (NCB) prepared by chemical vapor deposition method (Fig. S1). NCB is selected on account of the robust cube morphology (Fig. S2a), which provide spacious substrates to load metal atoms and exert minimum geometric shielding on the metal ripening. Besides, NCB possesses unique graphitic tissue and surface chemistry. The NCB shells are built by long parallel graphene multilayers (Fig. S2b), suggesting the high graphitization degree and favorable electronic conductivity. The analyses of nitrogen structures on NCB surface are demonstrated in Fig. S3.

The precursor for sintering is prepared by room temperature wet-impregnation method, during which process NCB was immersed into Ni/Ru salt solutions for adsorbing the hydrated Ni/Ru ions. This scenario is demonstrated by the schematic in Fig. S4. Density functional theory (DFT) calculations were conducted to evaluate the clustering tendency of adsorbed Ru and Ni ions on NCB. First, the cohesive energies (Ec) for Ru and Ni are calculated to be −7.021 eV and −4.788 eV, respectively. Regarding to the binding energy (Eb), due to the solvation structures of hydrated Ni/Ru ions, the metal ions cannot form direct covalent bonds to the substrate therefore the adsorption of hydrated Ni/Ru ions on NCB should be weak. As expected, the calculated binding energies (Eb) are −0.754 eV and −0.680 eV for Ru and Ni, and the Eb values corresponding to three types of nitrogen modified substrates in NCB are very close (Fig. S5). Basically, the absolute values of Eb are much lower than Ec for Ru and Ni, therefore the clustering tendency of Ni/Ru ions should be dominated by the cohesive energy. On this basis, Ru has higher clustering tendency than Ni due to the much larger cohesive energy. The discrepancy in Ec will result in the selective clustering of Ru and the formation of NiSARuAC composite on NCB. This mechanism is validated by the following comprehensive experimental results.

As shown in Fig. 1b, high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image and corresponding energy dispersive X-ray spectroscopy (EDS) mapping prove the isolated atom configuration of Ni/Ru in the precursor (25 °C). Intriguingly, Ni and Ru demonstrated different clustering behaviors in the following sintering. At 100 °C, Ru atoms agglomerated into sub-nanometer atomic clusters (RuAC). On the contrary, Ni atoms formed stable single atom species (NiSA). The higher temperature of 300 °C and 500 °C further increased the RuAC size yet still cannot induce the clustering of NiSA. The EDS mapping demonstrates well-defined RuAC and monatomic NiSA uniformly distributing on NCB. Higher sintering temperatures of 600 °C and 700 °C induced further ripening of RuAC but cannot change the monatomic Fe/Co/Ni species (Figs. S6–8).

The NiSA and RuAC configurations are further verified by synchrotron extended X-ray absorption fine structure (EXAFS) spectra (Fig. 1c). Ni K-edge EXAFS spectrum displays only a pronounced nearest-neighbor peaks at ca. 1.5 Å ascribed to Ni–N bond, whereas the Ni-Ni bond is totally absent. On the contrary, Ru K-edge EXAFS spectrum displays the characteristic peak of Ru-Ru at ca. 2.4 Å, suggesting the clustering of Ru atoms into multi-atom aggregate. In-situ TEM technique was employed to further verify this selective clustering. Figure S9 displays time-series TEM images of one fixed region collected from in-situ heating TEM video (Supplementary Movie 1). With increasing temperature, isolated dark spots gradually appeared on the margins of NCBs, which can be all identified as Ru clusters by electron energy loss spectroscopy (EELS) mapping. Meanwhile, no Ni cluster or particle can be observed. The locations of Ru clusters are not as uniform as those in the specimen prepared in tube furnace (Fig. S10), which is probably due to the different heating conditions in in-situ TEM specimen holder and in tube furnace. Nonetheless, the selective-clustering mechanism is still valid.

Apart from bimetallic system, we propose that the energy-selective-clustering mechanism can be expanded and be also applicable to multimetallic systems (M1,2,······,n, n ≥ 3). As shown in Fig. 1d, hydrated M1,2,······,n ions anchored on NCB can be readily prepared by wet-impregnation method. M1,2,······,n ions will experience different ripening processes due to the different clustering tendencies, which are in thermodynamics determined by the cohesive energy (Ec) and binding energy (Eb)43,44. In addition to Ni/Ru, we investigated the other various metal elements in terms of cohesive energy, adsorption configuration and corresponding adsorption energies. As shown in Fig. S11, due to the water solvation effects, the interactions between hydrated metal ions and NCB are in the level of weak physical adsorption (Eb = ~ 0.7 eV)45,46. More importantly, because the central metal ions were largely shielded by the solvation shell, the Eb values for different metals are quite close (σ2 (Eb) = 0.006 eV2). By contrast, the absolute values of cohesive energies (Ec) of these metals (Table S1) are much higher than Eb, and discrepancy among different metals is also much more pronounced (σ2 (Ec) = 3.148 eV2). Therefore, the clustering tendency of M1,2,······,n on NCB is essentially dominated by their corresponding cohesive energies (Ec). Due to the discrepancy in cohesive energy, different metal ions will exhibit selective-clustering upon sintering and the symbiosis of multimetallic atomic aggregates (M1AA-M2AA-…-MnAA, AA = SA or AC) can be achieved. The symbiotic multimetallic AAs are spatially adjacent, which will probably induce strong native interaction and synergistic effect.

Synthesis and characterization of bi-/multi-metallic SA/AC composite library

Figure 2a is a cohesive energy based element periodic table displaying the absolute values of cohesive energy in the vertical axis47. Since the clustering tendency of hydrated metal ions on NCB is dominated by cohesive energy, guided by this element periodic table, we utilize the energy-selective-clustering methodology to fabricate a library of NCB supported bi-/multi-metallic SA/AC composites (Table S2). For the bimetallic systems, we selected 18 bimetallic pairs ascribed to three categories: i) three M1-M2 bimetallic pairs with cohesive energies both lower than Ru (Fe-Ni, Co-Ni, Ni-Pd pairs); ii) nine M-Ru bimetallic pairs with cohesive energies of M lower than Ru (M = Ti, Cr, Mn, Fe, Co, Ni, Cu, Zn, Pd); iii) six M-Ru bimetallic pairs with cohesive energy of M higher than Ru (M = V, Mo, Rh, W, Ir, Pt). Sintering conditions were kept identical in order to achieve rigorous side-by-side comparison. Figure 2b–d show the HAADF-STEM, and EDS mapping/line scans of the obtained bimetallic SA/AC composites. For the three type-i M1-M2 bimetallic pairs, no metallic AC or nanoparticles can be observed in HAADF-STEM images. The metals exist as single atoms on NCB. The EDS mappings demonstrate the uniform distribution of Fe, Co, Ni, Pd elements. Therefore, dual-single-atom composites of FeSA-NiSA, CoSA-NiSA, NiSA-PdSA were obtained (Fig. 2b). For nine type-ii M-Ru bimetal pairs, clear Ru clusters can be observed both in HAADF-STEM images and EDS mappings for all specimens. On the contrary, the Ti, Cr, Mn, Fe, Co, Ni, Cu, Zn, Pd elements with lower cohesive energy display no clue of clustering. Therefore, composites of MSA-RuAC were prepared. For six type-iii M-Ru bimetallic pairs, atomic clusters of both M element and Ru appear for all the samples. The adjacent MAC and RuAC can be clearly detected by the EDS line scans, indicating the construction of six types of MAC-RuAC composites (Fig. 2c). To further demonstrate the synthesis generality, five more Ru-free bimetallic materials (MnSA-PtAC, CrSA-RhAC, FeSA-WAC, CoSA-MoAC, NiSA-NbAC) were also fabricated (Fig. S12).

Fig. 2: Synthesis and characterization of bi-/multi-metallic SA/AC composites.
figure 2

a Element periodic table based on cohesive energy values. HAADF-STEM, and EDS mapping/line scans of b bimetallic SA-SA composites, c bimetallic AC-AC composites and d bimetallic SA-AC composites based on different metal element pairs. HAADF-STEM, EDS mapping/line scans of e trimetallic SA/AC, quinary-metallic SA/AC, and f representative septenary-metallic SA/AC composites. Unmarked scale bars are 2 nm.

Full size image

By virtue of the generality for multimetallic system, the energy-selective-clustering based synthesis methodology is also applicable to the construction of trimetallic and medium/high-entropy multimetallic SA/AC composites. Figures 2e, f and  S13 demonstrate the as-synthesized trimetallic, quinary-metallic and septenary-metallic SA/AC composites. Employing relatively low cohesive energy metals, trimetallic and quinary-metallic single atom materials including FeSA-CoSA-NiSA and FeSA-CoSA-NiSA-CrSA-MnSA were fabricated, as proved by the HAADF-STEM images. As metals with relatively higher cohesive energy were added, trimetallic and quinary-metallic SA-AC composites (e.g. FeSA-CoSA-RuAC and FeSA-CoSA-NiSA-RuAC-PtAC) can be obtained. The EDS line scans on randomly selected bright clusters recognized the RuAC and PtAC. In our study, the energy-selective-clustering method can efficiently synthesize high-entropy septenary-metallic single atom material, e.g. FeSA-CoSA-NiSA-CrSA-MnSA-CuSA-PdSA. The loading of each metal is fairly high leading to the total loading of 20.49 wt% (Table S3), which is the highest loading of high-entropy single atom materials reported to date2,48,49. The synthesis of high-entropy single atom materials is highly challenging thus rarely reported. Analogue to trimetallic and quinary-metallic counterparts, septenary-metallic SA-AC composites (e.g. FeSA-CoSA-NiSA-CrSA-MnSA-RuAC-PtAC, ZnSA-FeSA-CrSA-TiSA-MnSA-IrAC-VAC, ZnSA-FeSA-CoSA-TiSA-NiSA-NbAC-WAC and PtAC-VAC-RuAC-MoAC-NbAC-IrAC-RhAC) can be obtained by adopting high cohesive energy metals. Therefore, our method provides an excellent platform for studying atomically dispersed multimetallic SA/AC materials with tunable configurational mixing entropy.

We conducted comprehensive characterizations on the aforementioned materials. The metal elements and corresponding atomic configurations in all the as-synthesized bi-/multi-metallic SA/AC materials are summarized in Fig. 3a, which clearly demonstrates the underlying correlation between atomic configuration and the cohesive energy of metals. According to the atomic configurations of these elements in bimetallic materials, the M1AC-M2AC, M1SA-M2SA, M1SA-M2AC zones can be defined, which are in strict conformity with the relevant values of cohesive energy. Apart from bimetallic materials, the atomic configurations of 17 metal elements can be further verified in trimetallic, quinary-metallic, septenary-metallic materials, as indicated by the marks of triangle, pentagram and heptagram in Fig. 3a. For instance, regarding the Fe-Pt pair in the M1SA-M2AC zone, the configurations of Fe and Pt are verified as FeSA and PtAC in bimetallic, trimetallic, quinary-metallic and septenary-metallic materials. Overall, for the 17 elements involved (Fig. S14), based on the NCB substrate and current synthesis conditions, the metal elements with relatively lower cohesive energies (Ti, Co, Fe, Ni, Cr, Mn, Pd, Cu, Zn) formed stable SA, and the metal elements with relatively higher cohesive energies (V, Pt, Rh, Mo, Ru, Ir, Nb, W) underwent various extents of clustering and yielded AC, as displayed in Fig. 3b. Therefore, the general principle can be extracted that the clustering tendency of metal on NCB substrate demonstrates well-defined positive correlation with the cohesive energy.

Fig. 3: Correlation between cohesive energy and atomic configurations of various metal elements.
figure 3

a The summary of as-synthesized atomically dispersed bimetallic and multimetallic materials. b The cohesive energies of the metal elements involved in the synthesis and the corresponding SA or AC configuration in the obtained bi-/multi-metallic materials.

Full size image

Essentially different from the previous studies, our methodology for the first time validates the controllable design and synthesis of atomically precise bi-/multi-metallic materials guided by the correlation between target atomic configuration and inherent energy character of metals. This approach demonstrated excellent element and configuration generality. Moreover, through the modulation of synthetic factors including carbon substrate traits and the sintering conditions, the boundary between SA elements and AC elements could be effectively adjusted thereby further expanding the composition and configuration spaces of atomically dispersed multimetallic materials.

Fine structures of representative bimetallic (FeSA/CoSA/NiSA)RuAC materials

Bimetallic MSARuAC@NCB (M = Fe, Co, Ni) were selected for understanding the fine structures and electrocatalysis utility of bi-/multi-metallic SA/AC composite library. The HAADF-STEM images (Fig. 4a) and EDS line scans (Fig. S15) verified the coexistence of RuAC and isolated (Fe/Co/Ni)SA. The atomic mappings in Fig. 4a visually demonstrate the distributions of Ru clusters and monatomic Fe/Co/Ni, as well as the spatial overlapping of RuAC and FeSA/CoSA/NiSA species in these materials50,51. To investigate the native interaction between adjacent MSA and RuAC, the electronic structures were investigated by X-ray photoelectron spectroscopy (XPS). Figure 4b shows the Ru 3p spectra with spin-orbit split Ru 3p3/2 and Ru 3p1/2 peaks. Compared to MSA free RuAC@NCB, the Ru 3p positions for MSARuAC@NCB display shift of around 0.3 eV to the lower binding energy, suggesting the less electron donating from RuAC to MSA-modified NCB than that to MSA-free NCB substrate. Different positions of Ru 3p indicate the different oxidation states of RuAC in MSARuAC@NCB, suggesting the varying degrees of interaction between RuAC and MSA-modified substrates.

Fig. 4: Electronic structures and atomic configurations of MSARuAC@NCB.
figure 4

a HAADF-STEM images and atomic mapping of FeSARuAC@NCB, CoSARuAC@NCB and NiSARuAC@NCB. b High-resolution XPS spectra of Ru 3p for MSARuAC@NCB and RuAC@NCB. c The normalized Ru K-edge XANES spectra and e the k3-weighted Fourier transform EXAFS spectra of MSARuAC@NCB and the references. Insets: models for spectra fitting. d The normalized Fe/Co/Ni K-edge XANES spectra and f the k3-weighted Fourier transform EXAFS spectra in R space of MSARuAC@NCB and the references.

Full size image

The electronic structures are further analyzed by X-ray absorption near edge structure (XANES) spectra. Per Fig. 4c, the pre-edge spectra of Ru in MSARuAC@NCB are close to Ru foil, indicating the metallic state of RuAC with enriched electron density. Moreover, the intensities of the white lines of Ru K-edge are different among MSARuAC@NCB, suggesting the different changes in charge density around RuAC, which is in line with the XPS results. This phenomenon reveals the intrinsic electronic interaction between the symbiotic MSA and RuAC in the composites. The XANES of Fe K-edge, Co K-edge and Ni K-edge (Fig. 4d) verify the positive valence states of single atom Fe, Co and Ni in MSA-N-C moieties.

The atomic configurations of MSARuAC composites were resolved by synchrotron EXAFS spectra. In Fig. 4e, two characteristic peaks locate at ca. 2.4 Å and 1.5 Å being attributed to Ru-Ru and Ru-N bonding, respectively, which suggest the presence of nitrogen involved coordination environments for RuAC52. Wavelet transform (WT) analysis also identified the coexistence of Ru-N and Ru-Ru paths (Fig. S16). Regarding to the Fe/Co/Ni species in MSARuAC@NCB, the EXAFS spectra of Fe/Co/Ni in R space display the pronounced nearest-neighbor peaks at ca. 1.5 Å (M-N) and the absence of M-M peaks (Fig. 4f). The WT-EXAFS also display only one maximum intensity at nearly 4.5 Å−1 being ascribed to the M-N bond, suggesting the formation of single Fe/Co/Ni atoms (Fig. S17). Guided by the EXAFS analysis, we constructed theoretical models to simulate the real catalysts (atomic coordinates of models listed in Supplementary Data 1). We conducted comprehensive DFT calculations and ab initio molecular dynamics (AIMD) simulations to determine the most reasonable models. The detailed screening processes are demonstrated in supporting information (Figs. S18–23 and Supplementary Movies 2–19). Figure 4e insets demonstrate the adopted models after screening, which consist of 10 atoms Ru cluster (Ru10) isolated by MSA-N-C substrates. The EXAFS spectra of MSARuAC@NCB can be well fitted based on these three models (Figs. 4e,  S24 and Table S4).

Hydrogen evolution performance and electrolyzer application

DFT calculations were conducted to find the potential M1SAM2AC sites that could deliver high catalytic activity toward hydrogen evolution reaction (HER). Ten different M1SAM2AC models were considered. As demonstrated in Figs. S25–30, NiSARuAC, FeSARuAC, CoSARuAC were screened via theoretical energy criterion. The HER performance of as-synthesized FeSA/CoSA/NiSA@NCB specimens were evaluated in 0.5 M H2SO4 and 1 M KOH electrolytes, respectively. The monometallic baselines of RuAC@NCB, FeSA/CoSA/NiSA@NCB were also assessed for comparison. As shown in Fig. 5a, all MSARuAC@NCB catalysts deliver excellent electrocatalytic HER activity with ultralow overpotentials at 10 mA cm−2, superior to commercial Pt/C. Pure NCB delivered negligible catalytic current, proving that the metal species contribute to the HER activity. Notably, MSARuAC@NCB exhibit alkaline HER activities in the order of CoSARuAC@NCB (8 mV) > NiSARuAC@NCB (10 mV) > FeSARuAC@NCB (17 mV). Both FeSA/CoSA/NiSA@NCB and RuAC@NCB demonstrate much inferior HER activities than MSARuAC@NCB (Figs. 5a, S31), suggesting the underlying synergy mechanism for MSARuAC composite in enhancing electrocatalytic performance.

Fig. 5: HER performance of MSARuAC@NCB and practical electrolyzer demonstration.
figure 5

LSV curves (non-iR corrected) of MSARuAC@NCB, RuAC@NCB, NCB and Pt/C in a 1 M KOH and b 0.5 M H2SO4. LSV curves (with 85%-iR corrected) of MSARuAC@NCB and 20% Pt/C in c 1 M KOH and d 0.5 M H2SO4 at industrial current density. Comparison of e ƞ10 and f ƞ1000 in 1 M KOH and 0.5 M H2SO4 for MSARuAC@NCB with the state-of-the-art HER electrocatalysts. Values were taken from references (Supplementary Tables 6 and 7). g Polarization curves operated at 25 °C and 55 °C of tested AWE. h Polarization curves of tested PEMWE operated at 25 °C and 55 °C. i Chronopotentiometry curve for AEM operating at 1 A cm−2. Inset: photograph of the AEMWE device.

Full size image

The Tafel plots provide profound insight of intrinsic HER kinetic process. As shown in Fig. S32, CoSARuAC@NCB, NiSARuAC@NCB and FeSARuAC@NCB respectively display Tafel slope values of 20.7, 23.3 and 34.3 mV dec−1, which are distinctly lower than that of Pt/C (45.2 mV dec−1). The Tafel slope values indicate that the HER processes of MSARuAC@NCB comply with the Volmer-Tafel mechanism53. The charge transfer resistance of MSARuAC@NCB are much smaller than that of RuAC@NCB (Fig. S33), suggesting that the incorporation of FeSA/CoSA/NiSA species effectively improves the interfacial electron transfer kinetics. CoSARuAC@NCB exhibits the most facile charge transfer, which is highly conducive to the best HER activity.

For acidic HER, NiSARuAC@NCB, FeSARuAC@NCB and CoSARuAC@NCB demonstrate η10 values of 47, 61 and 75 mV (Fig. 5b). Of note, the rank in acidic HER activity for MSARuAC@NCB catalysts is different from that in alkaline HER activity, which is reflective of the different reaction mechanism under different pH conditions. All the bimetallic composite based catalysts are superior to the monometallic candidates for acidic HER (Fig. S34). The Tafel slope of NiSARuAC@NCB is 46.1 mV dec−1 (Fig. S35), indicating the reaction follows the Volmer-Heyrovsky mechanism. The value is distinctly lower than those of FeSARuAC@NCB, CoSARuAC@NCB and commercial Pt/C, suggesting the faster kinetic property of the former. According to the Nyquist plots (Fig. S36), the comparison of charge transfer resistance for different catalysts agrees with the trend of acidic HER activity. The intrinsic activities of catalysts were further reflected by turnover frequency (TOF) value. The TOF of bimetallic (FeSA/CoSA/NiSA)RuAC are about an order of magnitude higher than that of RuAC@NCB at -0.05 V vs. RHE in both alkaline and acidic media (Fig. S37), highlighting the significantly higher catalytic activity of bimetallic MSARuAC as compared to monometallic RuAC.

The pragmatic evaluation of MSARuAC@NCB for practical HER were conducted at industrial-level current densities (Fig. 5c, d). Remarkably, the optimal MSARuAC@NCB catalyst can readily reach high current densities of 1.2 A cm−2 at low overpotentials of 188 and 210 mV in alkaline and acidic media, respectively. Commercial Pt/C failed to work at such high current density in both alkaline and acidic media given the exaggerated overpotential required. Per Fig. S38, all MSARuAC@NCB catalysts demonstrate much higher mass activities than that of commercial Pt/C, which suggests the great potential for MSARuAC@NCB as highly cost-effective catalysts. Figure 5e and f compare the overpotentials of MSARuAC@NCB at 10 and 1000 mA cm−2 (η10 and η1000) to the state-of-the-art electrocatalysts in previously published literatures, respectively. The pH-universal catalysts that were reported working at high current density is quite rare by far (Tables S6, 7). Among them, MSARuAC@NCB deliver the lowest η10 and η1000, suggesting the superiority of MSARuAC@NCB for pH-universal HER electrocatalysis.

Long-term chronopotentiometry tests were performed to evaluate the catalytic durability. As demonstrated in Fig. S39, the potentials of MSARuAC@NCB remained stable after 100 hours operation at 10 mA cm−2. The increases in potentials are only 21 mV for CoSARuAC@NCB in 0.5 M H2SO4 and 48 mV for NiSARuAC@NCB in 1.0 M KOH during 100 h. In contrast, the potentials of RuAC@NCB baseline significantly increased by 64 and 120 mV in alkaline and acid media in only 50 hours. This result strongly proves the critical role of the interaction between MSA and RuAC in stabilizing the structure of the active sites and maintaining the high HER activity.

The application potentials of MSARuAC@NCB were further verified in practical alkaline water electrolyzer (AWE), proton exchange membrane water electrolyzer (PEMWE) and anion exchange membrane water electrolyzer (AEMWE). The AWE employed CoSARuAC@NCB cathode and commercial RuO2 anode (Fig. S40) needs a cell voltage of 1.63 V to achieve 500 mA cm−2 and 2.07 V to achieve 2 A cm−2 at 25 °C (Fig. 5g). At an elevated temperature of 55 °C, industrial level current densities of 2.2 A cm−2 and 3.7 A cm−2 can be obtained at 2 V and 2.25 V, respectively (Fig. 5g). The performance is among the highest compared to Pt/C anode based counterpart and other previously reported AWE devices (Table S8). Per Fig. 5h, the PEMWE for simulating industrial condition of acidic water electrolysis employing NiSARuAC@NCB cathode and IrO2 anode (Fig. S41) displays current densities of 1 A cm−2 at 2 V (25 °C) and 2.5 A cm−2 at 2.4 V (55 °C). Both electrolyzers demonstrate excellent catalytic durability, which delivered slow degradation for up to 350 hours (Figs. S42, 43). The CoSARuAC@NCB based AEMWE (Fig. S44) exhibits much lower voltage polarization as compared to commercial Pt/C based device (Fig. S45). Remarkably, the AEMWE electrolyzer can operate stably at industrial-level current density of 1 A cm−2 for 1000 hours (Fig. 5i).

Electrocatalysis mechanisms derived from MSARuAC composites

DFT calculations were performed to understand the underlying electrocatalytic mechanism. The first step is to determine the true active sites in MSARuAC composites. For alkaline HER, the H2O adsorption is the primary decisive step because the proton supply is mainly from the H2O dissociation in alkaline medium54. Therefore, the active site should be the position in the MSARuAC composite that has the strongest adsorption towards H2O. According to Fig. 6a, the H2O adsorption energy on RuAC sites for all MSARuAC composites are distinctly higher than those on FeSA/CoSA/NiSA sites (atomic coordinates of models listed in Supplementary Data 1). Therefore, RuAC should be the active sites (Fig. 6b). Figures 6c and S46 demonstrate the steric position of the alkaline HER active site, as well as the pronounced charge transfer between the H2O adsorbate and the active site. Regarding to the active site in acidic medium, the adsorption of both H and H2O should be considered. According to the calculated adsorption energies in Figs. 6a and S47, the H adsorption on FeSA/CoSA/NiSA sites are stronger than the H adsorption on RuAC sites. In addition, the H adsorption is much stronger than the H2O adsorption on FeSA/CoSA/NiSA, which greatly accelerate proton supply to the MSA site. Therefore, the active site for acidic HER should be on the FeSA/CoSA/NiSA species. Figures 6c and S48 exhibit the configuration of H adsorbed active site and corresponding differential charge density.

Fig. 6: HER mechanism for MSARuAC@NCB and extensions of bimetallic materials in other electrocatalysis systems.
figure 6

a Comparison of H2O and H adsorption energies. b Schematic of H2O and H adsorption behaviors in alkaline and acidic media. c Differential charge density in NiSARuAC@NCB model. d Schematic of the HER mechanism of MSARuAC@NCB. e Gibbs free energy diagrams for alkaline HER. f Gibbs free energy diagrams for acidic HER. g Relationship between η10 in alkaline HER and OH* desorption energy barrier. h Relationship between η10 in acidic HER and ΔGH*. i The principle of composition and configuration modulation towards SA/AC bimetallic sites for various electrocatalytic systems. j ORR polarization curves of FeSAPtAC@NCB and PtAC@NCB in 0.1 M KOH electrolyte. k OER polarization curves of CrSAIrAC@NCB and IrAC@NCB in 0.5 M H2SO4.

Full size image

Combining the experimental observation and theoretical calculation, the scenario of alkaline/acidic HER on MSARuAC composites is demonstrated in Fig. 6d. Due to the different adsorption capabilities towards H2O and H, the MSARuAC composites have dual catalytic active sites, which is ascribed to MSA and RuAC in acidic and alkaline solutions. The decoupling of active sites enables MSARuAC@NCB performing well for both acidic and alkaline HER. Moreover, the intrinsic electronic interaction between the symbiotic MSA and RuAC generates inter-site synergy effect on the hydrogen evolution kinetics enhancement, which endows the bimetallic composites with far superior acidic/alkaline HER activity and durability than monometallic SA or AC.

On the basis of the active sites identification, the diverse structure-property relationships should be further established. For alkaline HER, the catalysts follow the order of CoSARuAC@NCB>NiSARuAC@NCB>FeSARuAC@NCB in terms of both η10 and mass activity (Fig. S49a). Whereas in 0.5 M H2SO4, the HER performance rank becomes NiSARuAC@NCB>FeSARuAC@NCB>CoSARuAC@NCB (Fig. S49b). This phenomenon is expected considering the decoupled active sites in different pH solutions and the different HER reaction pathways. According to Fig. 6e, all MSARuAC@NCB catalysts are energetically favorable for H2O adsorption and dissociation processes during alkaline HER (Volmer step). The OH− desorption is endothermic for all MSARuAC@NCB catalysts and identified as the rate-determining step (RDS). In-situ attenuated total reflection surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS) shown in Fig. S50 experimentally verified the OH- desorption as RDS. Basically, the facile OH− desorption kinetics under alkaline environment alleviates the site poison caused by the strong OH− adsorption and guarantees fresh active sites for continuous HER catalysis. As shown in Fig. 6g, the decrease of OH− desorption energy barrier agrees well with the trend of η10 decease. In especial, CoSARuAC@NCB delivered the lowest energy barrier (0.556 eV) in this RDS, which endows CoSARuAC@NCB with the exceptionally low η10 of 8 mV. As for the acidic HER, a moderate ΔGH* close to 0 eV is desired55,56,57. NiSARuAC@NCB exhibits the most desirable ΔGH* of -0.33 eV, which is more favorable than that of FeSARuAC@NCB (–0.53 eV) and CoSARuAC@NCB (–0.68 eV) (Fig. 6f). Moreover, the acidic HER activities (η10) of MSARuAC@NCB catalysts present a good linear relationship with the calculated ΔGH*, which matches well with the Sabatier principle (Fig. 6h)57,58.

As demonstrated in Fig. 6i, under the principle of cohesive energy discrepancy driving SA/AC symbiosis, the library of atomically dispersed multimetallic materials can be further expanded. Benefited from the excellent universality, our synthesis methodology has significant potentials of expansion to other various electrocatalytic systems by designing and modulating the composition and configuration of atomically dispersed bi-/multi-metallic materials. Taking the bimetallic material as an example, for a target electrocatalytic reaction, metal element (M1) with high intrinsic catalytic activity can be selected and M1SA/AC specie can be constructed. During the symbiosis process, M2SA/AC species with inherent interaction with M1SA/AC can be simultaneously introduced, which can effectively modulate the electronic structure of M1SA/AC and optimize the intermediate adsorption/desorption behavior thereby yielding the best catalytic performance. Moreover, the ultra-high flexibility of switching M2SA/AC species is highly favorable for control experiment designing, providing excellent platforms for the electrocatalysis mechanism exploration.

To demonstrate the viability, we showcase additional three important electrocatalytic systems of alkaline oxygen reduction reaction (ORR), acidic oxygen evolution reaction (OER), and carbon dioxide reduction reaction (CO2RR). For alkaline ORR, intrinsic high-active Pt was selected as the main element. Various secondary metal element was coupled with Pt clusters, yielding a set of MSAPtAC@NCB materials (Fig. S51a). The MSA species can effectively modulate the electronic structure of PtAC, as evidenced by the d-band center calculation results (Fig. S51b, c). As shown in Fig. 6j, we rapidly obtained the high-performance ORR catalyst of FeSAPtAC@NCB with the superior half-wave potential of 0.94 V. The ORR activities and d-band center of the catalysts exhibit classic volcano correlation with FeSAPtAC@NCB at the vertex (Fig. S52). An analogous strategy of catalyst development is also applicable for the acidic OER by conducting the control group of MSAIrAC@NCB (Fig. S53). As demonstrated in Figs. 6k and S54, CrSAIrAC@NCB at the vertex of the volcano delivered the lowest η10 of 238 mV. Figures S55,56 and Tables S9, 10 displayed a comprehensive comparison of FeSAPtAC@NCB and CrSAIrAC@NCB with the state-of-the-art oxygen electrocatalysts, highlighting their leading positions of catalytic performances in the field. Regarding CO2RR system, monatomic CuSA can be selected as the primary catalytically active species. The introduction of ancillary MSA can significantly increase the CO production selectivity by 22.5 times comparing the performances of NiSACuSA@NCB and CuSA@NCB (Fig. S57). The above results prove the great potential of our universal synthesis methodology for creating novel composite materials applied in extensive energy electrocatalysis areas.

Discussion

In summary, we develop a highly efficient and general methodology to synthesize a library of multimetallic single atom or/and atomic cluster composites. Because of the discrepancy in cohesive energy, multiple metal monomers follow energy-selective-clustering mechanism on nitrogen-functionalized carbon substrates, leading to the symbiosis of multimetallic single atoms or/and atomic clusters. Employing this mechanism, a total 40 types of bimetallic, trimetallic, quinary-metallic, and septenary-metallic SA/AC composites from 17 metal elements were constructed on NCB, including ultra-high loading septenary high-entropy single metal atom material that were barely synthesized in the field. The multimetallic SA/AC composites demonstrate great potential for electrocatalysis application. As a demonstrator, the model catalysts of MSARuAC@NCB exhibit superior electrocatalytic HER performances in both alkaline and acidic media and endow practical electrolyzers with top-level activity and durability. Controllable element modulations towards bimetallic M1SAM2AC@NCB created high-performance ORR and OER catalysts, which exemplified the significant potential of our synthesis methodology for versatile electrocatalysis systems. This work may inspire new researches for exploration and discovery of atomically dispersed metallic materials with ultra-complex elemental composition and mixing entropy.

Methods

Chemicals

Magnesium carbonate hydroxide pentahydrate (C4H2Mg5O14·5H2O, 98%) was purchased from Alfa Aesar (China) Chemicals Co., Ltd. Potassium hydroxide (KOH, 95%), acetonitrile (C2H3N, 99.5%), bis(2,4-pentanedionato)manganese(II) (C10H14MnO4, 99%), chromium(III) acetylacetonate (C15H21CrO6, 99.99%), rhodium trichloride (RhCl3, 98%), Ammoniumniobate(V)oxalatehydrate (C2H5NNbO4, 98%) were purchased from Meryer Chemical Technology Co., Ltd. Ruthenium(III) chloride (RuCl3, Ru 45-55%), vanadyl acetylacetonate (C15H21O6V, 97%), titanium diisopropoxide bis(acetylacetonate) (C16H28O6Ti), tungsten(VI) chloride (WCl6, 99%), Iridium(III) chloride (IrCl3, 99.8%), molybdenyl acetylacetonate (C10H14MoO6, 97%), chloroplatinic acid hexahydrate were purchased from HEOWNS. Cobalt(II) chloride hexahydrate (CoCl2·6H2O, 99.99%), nickel(II) chloride hexahydrate (NiCl2·6H2O, 99.99%), iron(III) chloride hexahydrate (FeCl3·6H2O, 99%), copper(II) chloride dihydrate (CuCl2·2H2O, 99%), zinc chloride (ZnCl2, 99.999%) and palladium(II) chloride (PdCl2, 98%) were purchased from Macklin (Shanghai) Biochemical Technology Co., Ltd. Nafion was purchased from Sigma-Aldrich. Nickel foam (NF) was offered by Shanghai Keqi Technical Service Studio. Polyethersulfone (PES) membrane was obtained from Hangzhou Cobetter Filter Equipment Co., Ltd. Deionized water was used to prepare all solutions and electrolytes. Nafion 117 and Pt-coated Ti mesh substrate were purchased from Sinero. All chemicals are of analytical purity and used without further purification.

Synthesis of nitrogen-functionalized carbon nano box (NCB)

NCB was prepared by a chemical vapor deposition (CVD) method. In the typical synthesis process, 2.0 g 4MgCO3·Mg(OH)2·5H2O was placed in a porcelain boat and heated to 900 °C in a tube furnace with a heating rate of 5 °C min−1 under 60 sccm N2 flow. A gas-washing bottle containing acetonitrile was connected in the N2 up-stream and the flow rate was reduced to 20 sccm. The CVD process lasted for 3 h at 900 °C, and afterwards the gas-washing bottle was removed. The product powders were washed with 6 M HCl at room temperature for 24 h and dried in a blast oven at 60 °C to obtain nitrogen-functionalized carbon nanobox (NCB).

Synthesis of MSARuAC@NCB (M=Ni, Ti, Cr, Mn, Fe, Co, Cu, Zn, Pd)

In the typical synthesis process, 2.37 mg NiCl2·6H2O (or 3.64 mg C16H28O6Ti, 3.49 mg C15H21CrO6, 2.53 mg C10H14MnO4, 2.70 mg FeCl3·6H2O, 2.38 mg CoCl2·6H2O, 1.70 mg CuCl2·2H2O, 1.36 mg ZnCl2, 1.77 mg PdCl2), 4.14 mg RuCl3 and 20 mg NCB were dissolved in 10 mL deionized water, and then stirred at room temperature for 24 h. The mixture was then filtered and freeze-dried. The obtained solid was calcined at 500 °C for 1 h under Ar atmosphere with a heating rate of 10 °C min−1.

Synthesis of MACRuAC@NCB (M=V, Mo, Rh, W, Ir, Pt)

In the typical synthesis process, 2.65 mg C10H14O5V (or 3.26 mg C10H14MoO6, 2.09 mg RhCl3, 3.96 mg WCl6, 2.98 mg IrCl3, 5.17 mg H2PtCl6), 4.14 mg RuCl3 and 20 mg NCB were dissolved in 10 mL deionized water and then stirred at room temperature for 24 h. The mixture was then filtered and freeze-dried. The obtained solid was calcined at 500 °C for 1 h under Ar atmosphere with a heating rate of 10 °C min−1.

Synthesis of FeSANiSA@NCB, CoSANiSA@NCB and NiSAPdSA@NCB

In the typical synthesis process, 2.37 mg NiCl2·6H2O, 2.70 mg FeCl3·6H2O (or 2.38 mg CoCl2·6H2O, 2.37 mg NiCl2·6H2O, or 1.77 mg PdCl2, 2.37 mg NiCl2·6H2O) and 20 mg NCB was dissolved in 10 mL deionized water and then stirred at room temperature for 24 h. The mixture was then filtered and freeze-dried. The obtained solid was calcined at 500 °C for 1 h under Ar atmosphere with a heating rate of 10 °C min−1.

Synthesis of RuAC@NCB, FeSA@NCB, CoSA@NCB, NiSA@NCB

The synthesis processes of RuAC@NCB and FeSA/CoSA/NiSA@NCB specimens are similar to that of NiSARuAC@NCB except that RuCl3, FeCl3·6H2O, CoCl2·6H2O, NiCl2·6H2O were not added.

Synthesis of FeSACoSANiSA@NCB

2.70 mg FeCl3·6H2O, 2.38 mg CoCl2·6H2O, 2.37 mg NiCl2·6H2O, and 10 mg NCB were dissolved in 10 mL deionized water, and then stirred at room temperature for 24 h. The mixture was then filtered and freeze-dried. The obtained solid was calcined at 500 °C for 1 h under Ar atmosphere with a heating rate of 10 °C min−1.

Synthesis of FeSACoSANiSACrSAMnSA@NCB

2.70 mg FeCl3·6H2O, 2.38 mg CoCl2·6H2O, 2.37 mg NiCl2·6H2O, 3.49 mg C15H21CrO6, 2.53 mg C10H14MnO4, and 10 mg NCB were dissolved in 10 mL deionized water, and then stirred at room temperature for 24 h. The mixture was then filtered and freeze-dried. The obtained solid was calcined at 500 °C for 1 h under Ar atmosphere with a heating rate of 10 °C min−1.

Synthesis of FeSACoSANiSACrSAMnSACuSAPdSA@NCB

2.70 mg FeCl3·6H2O, 2.38 mg CoCl2·6H2O, 2.37 mg NiCl2·6H2O, 3.49 mg C15H21CrO6, 2.53 mg C10H14MnO4, 1.70 mg CuCl2·2H2O, 1.77 mg PdCl2, and 10 mg NCB were dissolved in 10 mL deionized water, and then stirred at room temperature for 24 h. The mixture was then filtered and freeze-dried. The obtained solid was calcined at 500 °C for 1 h under Ar atmosphere with a heating rate of 10 °C min−1.

The syntheses of other bi-/multi-metallic atomically dispersed materials follow the same steps by using the appropriate metal precursor, heating rate, sintering time and sintering temperature (Table S2).

Material characterizations

Field emission scanning electron microscopy (SEM) measurement was conducted using a JSM-7800F from JEOL. Transmission electron microscopy (TEM) was conducted using a JEM-2100F from JEOL which operated at 200 kV. High-angle annular dark-field (HAADF)-STEM was conducted using a JEM-ARM200F from JEOL. The in-situ heating TEM was conducted using a JEM-ARM300F from JEOL. X-ray diffraction (XRD) was tested on a Bruker-D8 Advanced X-ray Diffractometer using Cu Kα radiation (λ = 0.15406 nm). XPS measurements were conducted using an X-ray photoelectron spectrometer (Axis Supra, Kratos). Inductively coupled plasma-mass spectrometry (ICP-MS) was conducted using Agilent 5110. X-ray absorption spectroscopy (XAS) spectra were performed in the Shanghai Synchrotron Radiation Facility (SSRF). The acquired X-ray absorption near-edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) data were analyzed by the ATHENA module of the IFEFFIT software package59. The EXAFS fitting data was collected using the ARTEMIS module implemented in the IFEFFIT software package. The wavelet transforms (WT) were obtained by the Fortran-based HAMA code60.

Electrochemical measurements

Electrochemical measurements of HER

Electrochemical performances were assessed via an electrochemical workstation (CHI660E, Shanghai, China) in a three-electrode cell system at room temperature. Catalyst-coated carbon cloth, graphite rod, and saturated calomel electrode (SCE) were used as working, counter and reference electrodes, respectively. The electrolytes employed are N2-saturated 1 M KOH (pH = 13.7 ± 0.1) and 0.5 M H2SO4 (pH = 0.4 ± 0.1) aqueous solution for alkaline and acidic HER, respectively. The electrolyte is synthesized before the electrochemical testing and stored sealed at room temperature. The working electrodes were prepared as follows: uniform suspension containing 5 mg catalyst, 965 μL isopropyl alcohol, and 35 μL 5 wt% Nafion solution were obtained by ultrasonic mixing for about 30 min. Afterward, 200 μL catalyst ink was coated onto the 1 × 1 cm2 carbon cloth, which was dried in air. The typical catalyst loading is 1 mg cm−2 on carbon cloth, yielding a noble metal mass loading of ca. 0.40 mgRu cm−2 loading for all samples. The potential measured is referenced to the reference electrode, necessitating conversion to the potential relative to the reversible hydrogen electrode (RHE). Platinum wire electrodes served as both the working and counter electrodes during calibration in an H2-saturated electrolyte. The zero current potential was determined through cyclic voltammetry (CV) measurements.

$${E}_{{{\rm{RHE}}}}={E}_{{{\rm{SCE}}}}+{E}^{{{{\rm{\theta }}}}}_{{{\rm{SCE}}}}+0.059{{\rm{pH}}}$$
(1)

Cyclic voltammetry (CV) was conducted in the potential range of 0.168 to –0.532 V versus RHE at a sweep rate of 100 mV s−1 for 20 cycles for catalyst activation. Linear sweep voltammetry (LSV) was conducted in the same potential range of CV measurements at a scan rate of 5 mV s−1. Tafel slope was calculated with according to the LSV profiles. CV was also performed in the scan rate range of 10 to 70 mV s−1 to calculate the double-layer capacitance (Cdl). Electrochemical impedance spectroscopy (EIS) was tested by applying an AC voltage with 5 mV amplitude in a frequency range from 100 KHz to 0.01 Hz. For the durability evaluation, the chronopotentiometry were obtained with a constant current density of 10 mA cm−2. Each data has undergone repeatability testing, mitigating sources of random error through stringent test specifications. The data demonstrate high accuracy and reproducibility, falling within experimental error margins. All measurements were performed without iR compensation if not explicitly specified.

Electrochemical measurements in the AWE device

Electrochemical performances for practical alkaline water splitting were tested in the zero-gap alkaline water electrolyzer (AWE). The electrochemical measurements were performed on the IV2010 DC power supply. The prepared CoSARuAC@NCB and commercial RuO2 catalysts were used as cathode and anode with a loading of 1 mg cm−2. The mass loading of Ru on the cathode and anode is approximately 0.4 and 0.76 mgRu cm−2, respectively. For comparison, noble metal catalysts Pt/C and RuO2 were also used as cathode and anode under the same conditions. The mass loading of Pt on the cathode is approximately 0.20 mgPt cm−2, and the mass loading of Ru on the cathode is approximately 0.76 mgRu cm−2. To prepare the anode and cathode ink, catalysts were dispersed to a mixture of isopropanol and distilled water. Then, Nafion® solution (5 wt%) was added. After ultrasonicated for at least 1 h in a low-temperature water bath, a uniform catalyst ink can be obtained. Porous polyethersulfone (PES) has a thickness of 0.12 mm and the pore diameter is 0.22 μm, which was used as the separator. The active area of the electrode is measured to be 1×1 cm2. 6 M KOH solution was used as both anolyte and catholyte. The circulation of the electrolyte was enabled by the peristaltic pump at 80 rpm and the inner diameter of the hose was 3 mm. No iR compensation was applied. For the activity evaluation, polarization curves were obtained at 25 °C and 55 °C. For the durability evaluation, the chronopotentiometric curves were obtained with a constant current density of 1.5 A cm−2 at 25 °C.

Electrochemical measurements in the AEMWE device

Electrochemical performances for practical alkaline water splitting were tested in an anion exchange membrane water electrolyzer (AEMWE). The electrochemical measurements were performed on the IV2010 DC power supply. The prepared slurries of the cathodic and anodic catalysts were firstly air-sprayed onto porous carbon paper and Ni foam gas diffusion layers (GDLs), respectively. Subsequently, the catalyst-coated GDLs were pressed with an anion exchange membrane (Fuma® FAA-3-50, thickness 130μm) to assemble into the membrane electrode assembly (MEA). The active area of the electrode is measured to be 1 × 1 cm2. The mass loading of Ru on the cathode and anode is ca 0.40 and 0.76 mgRu cm−2, respectively. Note that except for catalysts, other experimental conditions including the assembly techniques of the device and the testing parameters kept identical. The anion exchange membrane was immersed into 1 M KOH solution for at least 24 h prior to being used to exchange Cl− into OH−. 1 M KOH solution was used as both anolyte and catholyte. For the activity evaluation, polarization curves were obtained at 25 °C. The stability of the AEMWE was evaluated by chronopotentiometry test at a current density of 1 A cm−2 at a water temperature of 25 °C.

Electrochemical measurements in the PEMWE device

For the PEMWE, home-made IrO2 was used as the anode catalyst for the oxygen evolution reaction (OER) and NiSARuAC@NCB was used as the cathode HER catalyst. Nafion 117 served as the cation exchange membrane (DuPont, thickness 183 µm, N117). The catalyst inks were directly air-sprayed on both sides of the Nafion 117 membrane, with an IrO2 loading of 2 mg cm−2 (1.71 mgIr cm−2) for the anode and a NiSARuAC@NCB loading of 1 mg cm−2 (0.40 mgRu cm−2) for the cathode. The titanium felt with a thickness of 0.25 mm were used as gas diffusion layers in both the anode and cathode. The resulting CCMs were then pressed between the anode and cathode GDLs to form the MEA. The active area of the electrode is measured to be 1 × 1 cm2. During measurement, water was continuously pumped through the PEMWE. All voltages measured in PEMWEs were obtained without iR correction. The electrochemical measurements were performed on the IV2010 DC power supply.

The calculation for turnover frequency (TOF)

The TOF value was calculated based on the estimated number of active sites:

$${{\rm{TOF}}}=\frac{{{\rm{Number}}}\, {{\rm{of}}}\, {{\rm{hydroge}}}{{\rm{n}}}\, {{\rm{turnover}}}}{{{\rm{Number}}}\; {{\rm{of}}}\, {{\rm{active}}}\, {{\rm{sites}}}}$$
(2)
$${{\rm{Number}}}\; {{\rm{of}}}\; {{\rm{hydrogen}}}\; {{\rm{turnover}}}({{\rm{mol}}}/{{\rm{s}}})=\frac{{\mbox{j}}({\mbox{A}})}{2{\mbox{F}}}$$
(3)

Where j is current, the number 2 means two electrons per mol of H2 evolution, and F is a Faraday constant of 96,485.3 C. We assume that all Ru atoms in the catalysts are active for HER. The numbers of Ru atoms number in RuAC@NCB and MSARuAC@NCB catalysts were calculated from the Ru molar mass and the mass loading on the carbon cloth. The weight contents of various metals in the catalysts were determined by ICP-MS as shown in Table S5.

Calculation Method

All the calculations were performed using the density functional theory (DFT) within the Vienna Ab initio Software Package (VASP 5.4.4) code, employing the Perdew–Burke–Ernzerhof (PBE) generalized gradient approximation and the projected augmented wave (PAW) method61,62,63,64. The cutoff energy of the plane-wave was fixed at 400 eV. Electronic self-consistent iterations were set with a force of 10−5 eV and 0.01 eV/Å. To prevent interactions between periodic images, a 15 Å vacuum layer was added. During the model construction process, we selected Ru10 clusters and placed them on the modified carbon-based substrate FeSA/CoSA/NiSA-N4. The Brillouin zone of the surface unit cell was performed with 3 × 3 × 1 Monkhorst–Pack grid varying k-point sampling for catalyst optimizations65.

For calculations of isolated metal atoms in a vacuum, a 9 × 10 × 11 Å broken symmetry cell was utilized to guarantee the correct occupancy of degenerate orbitals. All theoretical calculations were non-spin-polarized, except for the isolated metal atom calculations and magnetic Fe, Ni, Mn, and Co. Cohesive energy (Ec) is represented as the energy variation per atom according to the following formula:

$${E}_{\rm{c}}=\frac{{{\mbox{E}}}_{{\mbox{bulk}}}}{{\mbox{N}}}-{E}_{\rm{at}}$$
(4)

where Eat represents the energy of the isolated metal atom within a vacuum, Ebulk signifies the energy of the bulk unit cell comprising N atoms.

The adsorption energy (Eads) of the surface species is determined by

$${E}_{{{\rm{ads}}}}={E}_{{{\rm{total}}}}-{E}_{{{\rm{surface}}}}-{E}_{{{\rm{species}}}}$$
(5)

where Etotal denotes the overall energy of the adsorbed species in conjunction with the catalyst surface, Esurface signifies the energy of the bare surface, and Especies represents the energy of the species in the gaseous phase.

The binding energy (Eb) is determined by:

$${E}_{{{\rm{b}}}}={E}_{{{\rm{AB}}}}-{E}_{{{\rm{A}}}}-{E}_{{{\rm{B}}}}$$
(6)

where EAB stands for the total energy of the adsorption structure involving the entire metal solvation, EA refers to the energy of metal-solvated structures within an empty crystal, and EB indicates the energy of the substrate.

The free energies of adsorbates at 298.15 K were estimated according to the harmonic approximation. The entropy was determined by applying the following equation:

$${{\rm{S}}}({{\rm{T}}})={k}_{\rm{B}}\sum _{{\mbox{i}}}^{{{\rm{harm}}}\, {{\rm{DOF}}}}\left[\frac{{\varepsilon }_{{{\rm{i}}}}}{{{\mbox{k}}}_{{\mbox{B}}}{\mbox{T}}\left({{\mbox{e}}}^{{\varepsilon }_{{{\rm{i}}}}/{{\mbox{k}}}_{{\mbox{B}}}{\mbox{T}}}-1\right)}-{\mathrm{ln}}\left(1-{{\mbox{e}}}^{{-\varepsilon }_{{{\rm{i}}}}/{{\mbox{k}}}_{{\mbox{B}}}{\mbox{T}}}\right)\right]$$
(7)

where kB is Boltzmann’s constant; DOF is the number of harmonic energies (εi) considered in the summation. The free energies of gas phase species are adjusted as follows:

$${{\rm{G}}}_{{\rm{g}}}({{\rm{T}}})={E}_{{{\rm{elec}}}}+{E}_{{{\rm{ZPE}}}}+\int {{\mbox{C}}}_{{\mbox{P}}} \, {{{\rm{dt}}}}-{{\rm{TS}}}({{\rm{T}}})$$
(8)

Where Cp is the specific heat capacity of a gas. The parameters are acquired from the NIST database (https://doi.org/10.18434/T4D303). We employed 0.035 bar of gas-phase water as the benchmark reference state for water due to its equilibrium with liquid water at 300 K under this specific pressure condition66,67.

Ab initio molecular dynamics (AIMD) simulation

AIMD simulations based on geometrically optimized structures were conducted at 298.15 K with a time step of 1 fs. The Nose-Hoover thermostat was employed to maintain a constant temperature. Following continuous 10 ps AIMD simulations, thermodynamically stable states were achieved.