Ammonia mediated electrostatic adsorption synthesis of paired subnanoclusters with enhanced hydrogen oxidation performance Authors: Xinwei Li, Xin Zhang, Changhong Zhan, Tao Gan, Yuchao Wang, Shangyi Ma, Xiaoqing Huang, Ali Han, Gang Liu Published: 2026 ABSTRACT AbstractSub-nanoclusters (sub-NCs) hold promise for the alkaline hydrogen oxidation reaction yet struggle with precise atomic dispersion and structural stability. Here, we report an ammonia-mediated electrostatic adsorption strategy that enables the synthesis of tunable paired metal-metal oxide sub-NCs, exemplified by PtSARuRh-MOx (SA represents single-atom sites, M = Cr, Mo, V, Ti, Ta, Zr). Electrostatic interactions between cationic PtRuRhδ+ and oxo-/fluoro-anions (e.g., CrO42−, MoO42-, VO3−, TiF62-, TaF72− and ZrF62−) control phase pairing during pyrolysis. The optimized PtSARuRh-CrOx demonstrates competitive mass activity (13.81 A mg−1, 58-fold that of commercial Pt/C) and delivers a peak power density of 1.43 W cm−2 in an alkaline membrane fuel cell, with stable operation exceeding 200 h. In situ spectroscopy and simulations reveal that the synergistic Ru/Rh sites, electronically modulated by the paired CrOx, collectively optimize H*/OH* adsorption and weaken CO* binding. This work establishes a generalizable synthetic route for the precise construction of paired sub-NCs for advanced energy conversion applications.Subject terms: Electrocatalysis, Electrocatalysis, Fuel cells Introduction Hydrogen-based energy technologies have garnered considerable interest for their potential to mitigate environmental impacts and diversify the global energy portfolio1–3. Among these, anion exchange membrane fuel cells (AEMFCs) have emerged as a promising alternative to proton exchange membrane fuel cells (PEMFCs), benefiting from advances in stable anion-exchange membranes4 and the feasibility of using non-platinum group metal (non-PGM) catalysts at the cathode5,6. Nevertheless, the performance of AEMFCs at the anode remains limited by the sluggish kinetics of the hydrogen oxidation reaction (HOR)7–9, which still predominantly relies on Pt-based catalysts10,11. Particularly, commercial Pt-based catalysts not only show poor CO tolerance but also exhibit intrinsically low HOR activity in alkaline media—a behavior widely linked to the suboptimal adsorption characteristics of hydrogen intermediates (Had) and hydroxyl species (OHad)12,13. Alloying Pt with other metallic elements represents an effective strategy to modulate the electronic structure and optimize the binding strengths of Had and OHad for enhanced HOR performance14. Nonetheless, the widespread application of current Pt-based catalysts is hindered by inadequate activity, limited durability, and high cost15. Therefore, the simultaneous and precise regulation of Had and OHad adsorption energetics to create highly active, stable, and CO-tolerant HOR electrocatalysts is critically important—yet continues to pose a fundamental and practical challenge. Recent advances in sub-nanoclusters (sub-NCs, < 2 nm) have demonstrated their potential in electrocatalysis, outperforming single-atom catalysts (SACs) and nanoparticles (NPs) in multiple reaction system16–19. The advantages of sub-NCs stem from their high atomic utilization efficiency20,21, tunable intermetallic electronic interactions that optimize reactant/intermediate adsorption configurations22, and synergistic metal-support effects that collectively lower reaction barriers23,24. However, the enhanced activity of sub-NCs is often accompanied by compromised inherent instability caused by high surface energy and aggregation tendencies during synthesis and operation25–27. Drawing inspiration from oxygen-tolerant hydrogenases that rely on strongly coupled auxiliary clusters for stability under oxidative stress28,29, the construction of paired metal oxide sub-NCs utilizes strong interfacial coupling to concurrently stabilize the metal NCs, facilitate interfacial electron modulation, and create cooperative active sites, thereby offering a complete solution to overcome existing catalytic limitations. However, the precise synthesis of such structures remains a formidable challenge, requiring atomic-level control over cluster composition, inter-cluster distance, and coordination environments—largely beyond the capabilities of conventional wet-chemical or pyrolysis-based methods30,31. Here, we present an ammonia-mediated electrostatic adsorption (AMEA) strategy as a versatile and universally applicable platform for synthesizing paired metal-metal oxide sub-NCs. In contrast to conventional wet-chemical methods that often rely on random adsorption, the AMEA approach leverages precisely controlled electrostatic interactions between metal and metal oxide precursors. Notably, this strategy allows the incorporation of over five different metal-metal oxide combinations into sub-NCs smaller than 2 nm, as exemplified by the PtSARuRh-MOx series (M = Cr, Mo, V, Ti, Ta, Zr). Synthetic mechanistic studies reveal that ammonia plays a critical role in promoting electrostatic attraction between cationic PtRuRhOxδ+ and oxo-/fluoro-anionic precursors. The optimized PtSARuRh-CrOx catalyst exhibits favorable HOR performance, achieving current densities of 1.90 A cm−2 (H2-O2) and 1.49 A cm−2 (H2-air, CO2-free) at the automotive-relevant voltage of 0.65 V in a AEMFC assembly with an anode catalyst loading of 50 μgPGM cm−2. Furthermore, the AEMFC device can stably work under the current density of 0.5 A cm−2 for over 200 h, underscoring the high activity and durability of PtSARuRh-CrOx. In-situ electrochemical spectra, combined with computational studies, show that the synergistic Ru/Rh sites enable a favorable optimization of H* and OH* intermediate adsorption, which accounts for the improved HOR kinetics. Results Ammonia Mediated Electrostatic Adsorption (AMEA) strategy Paired PtSARuRh-MOx NCs (M = Cr, Mo, V, Ti, Ta and Zr) were engineered on carbon nanotube substrate employing an ammonia-mediated electrostatic adsorption (AMEA) strategy and subsequent pyrolysis, as illustrated in Fig. 1a. The key to the precise control lies in the special dual functionality of ammonia (NH3·H2O), serving simultaneously as: (1) an initiator for alkaline polydopamine (PDA) formation (pH~9.0)32 (Figs. S1) and (2) a redox-active molecular mediator that orchestrates the precise electrostatic assembly of complementary PtRuRhOxδ+/Mδ− (Mδ− = CrO42−, MoO42−, VO3−, TiF62−, TaF72−, ZrF62−) precursors into well-defined PtSARuRh-MOx. By virtue of its special design, the AMEA strategy exhibits high versatility, providing a generalizable platform for the fabrication of diverse electrocatalytic materials with precisely controlled compositions, as further elaborated in the Experimental Section. In the initial step of the AMEA process, Ru3+/Rh3+ cations are preferentially adsorbed onto the negatively charged CNT-O support (Fig. S2) through electrostatic interactions, serving as anchor sites for the subsequent adsorption of anionic [PtCl6]2− and Mδ−. The introduction of NH3·H2O then triggers the formation of a positively charged NH3-coordinated Pt complex within the amorphous RuRhOxδ+ domains, leading to a structure denoted as PtRuRhOxδ+ for the metal NCs precursors. The positive charge of PtRuRhOxδ+ thereby enables the effective accumulation of additional ammonia-inert Mδ− ions (Fig. S3). AMEA synthesis process. a Schematic illustration of AMEA strategy. b UV-vis spectra of H2PtCl6 in H2O/ethanol mixed solvent (9:4 v/v) with and without ammonia addition; c liquid MS of H2PtCl6 in H2O/ethanol mixed solvent (9:4 v/v) with and without ammonia addition, alongside that of Pt(NH3)4Cl2. d Photographic comparison of various ionic species in H2O/ethanol mixed solvent (9:4 v/v) before and after ammonia addition. e–g Pt 4 f (e), Cl 2p (f) and N 1 s (g) XPS spectra of PtRuRhOxδ+ and commercial-Pt(NH3)4Cl2. h Zeta potentials of RuRhOxδ+, PtRuRhOxδ+ and anions-modified PtRuRhOxδ+. Source data for Fig. 1 are provided as a Source Data file. To verify the AMEA strategy, a control experiment was conducted by mixing equivalent amounts of metal precursors and NH3·H2O in an ethanol/water solvent, deliberately omitting dopamine and CNT-O to exclude their interference. As shown in Fig. 1b, the attenuation of the characteristic [PtCl6]2− absorption band at ~260 nm upon addition of NH3·H2O to H2PtCl6 suggested a gradual conversion of the Pt coordination environment from Cl- to NH333. The formation of NH3-coordinated Pt complexes from H2PtCl6 was verified by quadrupole time-of-flight mass spectrometer (Q-TOF MS) (Fig. 1c). Upon addition of NH3·H2O, new peaks corresponding to species like [Pt(NH3)4Cl2·3CH3CN·H2O + H]+ and [Pt2Cl3(CH3CN)4·CH3CN]+ emerged. These signals, obtained with a CH3CN/H2O (10:90) eluent, were consistent with a commercial Pt(NH3)4Cl2 reference and were not observed in the pure H2PtCl6. Concurrently, a distinct color change and the emergence of precipitation (Fig. 1d) provided clear evidence for the NH3·H2O-mediated oxidative transformation of Ru3+/Rh3+ ions into oxide species (denoted as RuOₓδ+/RhOₓδ+). The structure evolution was further corroborated by the Raman spectra analysis in Fig. S4 and X-ray diffraction (XRD) patterns in Fig. S5, which collectively verified the characteristics of amorphous PtRuRhOₓδ+. The transformation from Ru3+/Rh3+/[PtCl6]2− ions to PtRuRhOxδ+ was further verified by X-ray photoelectron spectroscopy (XPS) analysis in Fig. 1e−1g and Fig. S6, which demonstrated the coexistence of Pt, Rh, Ru, N, Cl, and O. Notably, the Pt 4 f XPS spectrum exhibited distinct peaks at 73.1 eV and 75.2 eV, corresponding to Pt2+ and Pt4+, respectively. These features are consistent with Pt-N or Pt-Cl species, likely originating from [Pt(NH3)xCly]δ+ complexes34–36, and definitively confirm the conversion of anionic [PtCl6]2- into cationic NH3-coordinated Pt2+/Pt4+ complex within the PtRuRhOₓδ⁺ matrix. Zeta potential measurements were further carried out to provide quantitative validation of the AMEA mechanism through systematic charge characterization. The zeta potential results in Fig. 1h and reveal that PtRuRhOxδ+ exhibited a substantially enhanced positive surface charge (+42.3 mV) relative to control samples (RuRhOxδ⁺, + 14.0 mV; RuOxδ⁺, + 7.4 mV; RhOxδ⁺, + 21.0 mV), conclusively demonstrating the effective integration of cationic NH3-coordinated Pt2+/Pt4+ complex into RuRhOₓδ+ domains. Fig. S8 further demonstrates the universal applicability of this approach, where a significant reduction in zeta potential confirmed effective electrostatic neutralization between PtRuRhOxδ+ and various anions (MoO42−, VO3−, TiF62−, TaF72− and ZrF62−). Additionally, the formation of amorphous PtRuRhOxδ+ in these analogous systems was verified by Raman and XRD spectra (Figs. S9, S10). Thus, the AMEA strategy enables precise control over the PtRuRhOxδ+/Mδ− precursor, which subsequently yields the well-defined, paired PtSARuRh-MOx NCs upon pyrolysis at 600 °C under H2/Ar, as confirmed by the XRD pattern in Figs. S11a, S11b. Further comparative analysis in Fig. S11c revealed that pyrolysis of PtRuRhOxδ+/CrO42− precursor without PDA encapsulation could lead to the formation of aggregated RuRh alloy and Cr2O3 NPs. Characterization High-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) image in Fig. 2a revealed abundant sub-NCs for paired PtSARuRh (red dashed circles) and CrOx (yellow dashed circles). The successful formation of intimately paired PtSARuRh and CrOx at the sub-nanoscale is further supported by energy-dispersive X-ray spectroscopy (EDS) elemental mapping, which shows clear spatial overlap of Pt, Ru, Rh, Cr, and O signals in Fig. 2b and other sample sections (Fig. S12). According to the detailed statistical data analysis in Fig. S12, the proportion of paired PtSARuRh and CrOx NCs reached as high as 82%. High-resolution electron energy loss spectroscopy (EELS) line-scan analysis was employed to characterize the interfacial electronic structure, revealing well-defined elemental segregation across the heterointerface. The spectral profiles in Fig. 2c demonstrate pronounced Ru/Rh signal enhancement within PtSARuRh-rich region, while adjacent domains exhibit characteristic Cr/O spectral signatures consistent with CrOx-rich region. A red shift of 1.5 eV in the Cr L-edge spectrum within CrOx regions indicated a lower oxidation state, potentially arising from electron transfer from PtSARuRh to Cr species37. In contrast to the PtSARuRh-CrOx, isolated PtSARuRh and CrOx NCs were formed using CrCl3 as the chromium source. Zeta potential and Raman spectroscopy analysis in Fig. S13 unequivocally revealed the NH3·H2O-mediated conversion of CrCl3 to cationic CrOxδ⁺ species. Critically, this conversion establishes a fundamental electrostatic contrast: whereas the PtRuRhOxδ⁺/CrO42− precursor maintains favorable electrostatic attraction between PtRuRhOxδ⁺ and CrO42−, the PtRuRhOxδ⁺/CrOxδ⁺ analogue exhibits pronounced intercomponent repulsion. The isolated PtSARuRh and CrOx NCs were clearly observed in the reference PtSARuRh/CrOx sample, and the proportion of paired metal-metal oxide clusters was lower (~15%, adapted from Figs. S14, S15) than that in PtSARuRh-CrOx (~82%). Guiding by the AMEA mechanism, well-defined analogous systems were successfully prepared, including PtSARuRh-MoOx, PtSARuRh-VOx, PtSARuRh-TiOx, PtSARuRh-TaOx, and PtSARuRh-ZrOx in Fig. 2d. As shown in Fig. 2e, the obtained NCs have an average size below 2 nm, confirming the sub-nanoscale precision of the AMEA-guided synthesis. In contrast, the PtSARuRh/Cr2O3-tris control sample, synthesized using the well-established tris-buffer method for PDA formation38–40, exhibited a markedly less defined interface (Fig. S16). The discrete Cr2O3 NPs were clearly observable, in consistent with the Cr2O3 NPs peak in Fig. S17e. Additionally, supplementary characterization of the control samples without metal oxide NCs was also conducted. Partial single-atom sites were observed in the PtSARuRh NCs (Fig. S18), and the EELS analysis (Fig. S19) verified the co-localization of Pt, Ru, and Rh. The low-resolution TEM images of PtRu-CrOx and PtRh-CrOx control samples were shown in Fig. S20, indicating the average sizes of 1.76 and 1.78 nm, respectively. Moreover, inductively coupled plasma atomic emission spectroscopy (ICP-AES) was probed to determine the metal composition of PtSARuRh-CrOx and control samples, as displayed in Table S1. Structural characterizations. a HAADF-STEM image of PtSARuRh-CrOx. The red circles represent PtSARuRh NCs, and the yellow circles represent CrOx NCs. b STEM-EDS elemental mapping images of (a). c Scanning path in the HAADF-STEM image for line scanning EELS spectra of PtSARuRh-CrOx and EELS spectra obtained from the line scan. d HAADF-STEM image of PtSARuRh-MoOx, PtSARuRh-VOx, PtSARuRh-TiOx, PtSARuRh-TaOx, and PtSARuRh-ZrOx. e The corresponding clusters sizes of PtSARuRh-CrOx and reference samples. Source data for Fig. 2 are provided as a Source Data file. Fine Structure XPS characterization of PtSARuRh-CrOx and reference samples (Figs. S21–S23) confirmed that the Pt, and Rh were predominantly in the metallic and oxidized species on the surface of N-doped CNT. Furthermore, the Cr in PtSARuRh-CrOx was found to be in a higher Cr oxidation state compared to its PtSARuRh/CrOx counterpart. Notably, no discernible F signal was observed in the final samples derived from fluorine-containing precursors (e.g., TiF62−, TaF72−, ZrF62−), as confirmed by XPS in Fig. S23. The formation of TiOx, TaOx, and ZrOx species after pyrolysis indicates that the precursors underwent complete conversion to their respective oxides, thus demonstrating the universality of the AMEA strategy for synthesizing diverse metal-metal oxides sub-NCs. The electronic properties and local coordination environments of the metal species (Pt, Ru, Rh and Cr) in PtSARuRh-CrOx were further elucidated using X-ray absorption fine structure (XAFS) spectroscopy. As shown in Fig. 3a, the Pt L3-edge X-ray absorption near-edge structure (XANES) spectra positioned the white-line intensities of PtSARuRh-CrOx and PtSARuRh/CrOx between the of Pt foil and PtO2, aligning with XPS results in Figure S22a. The Pt chemical states in PtSARuRh-CrOx and PtSARuRh/CrOx exhibited nearly identical XPS spectra, with a slight negative shift compared to PtSARuRh, indicative of enhanced surface oxidation. This finding was corroborated by the stronger Pt-N/C bonding intensity observed in the Fourier-transform extended X-ray absorption fine structure (FT-EXAFS) spectrum of PtSARuRh (Fig. 3b). The EXAFS fitting results in Figs. S24–S29 identified a dominant coordination scattering path at 2.69 Å attributable to Pt-Ru/Rh bonds rather than Pt-Pt (2.76 Å) (Table S2), confirming Pt exists as single atoms coordinated with Ru/Rh in NCs. The Pt-Ru/Rh coordination features were further demonstrated by the Pt L3-edge WT-EXAFS in Fig. S30. Fine structural characterizations. a, b Normalized Pt L-edge XANES spectra (a) and Fourier transform of k3-weighted EXAFS spectra (b) of different samples. c, d Normalized Ru K-edge XANES spectra (c) and Fourier transform of k3-weighted EXAFS spectra (d) of different samples. e, f Normalized Rh K-edge XANES spectra (e) and Fourier transform of k3-weighted EXAFS spectra (f) of different samples. g Wavelet transforms of k3-weighte Ru K-edge EXAFS spectra of different samples and references. Source data for Fig. 3 are provided as a Source Data file. Figures 3c, 3e present the Ru/Rh K-edge XANES spectra and the oxidation states of Ru/Rh in PtSARuRh-CrOx lied between those of their metallic and corresponding oxide forms,‌ consistent with the XPS results in Fig. S22b. Compared to PtSARuRh, both PtSARuRh-CrOx and PtSARuRh/CrOx exhibit an upshift in the Ru absorption edge, indicative of a higher Ru oxidation state than PtSARuRh. The coordination structures of Ru and Rh in various samples were systematically characterized, as illustrated in Fig. 3d and 3f, respectively. Detailed fitting parameters and corresponding spectra are comprehensively presented in Figures S31–S42 and Tables S3, S4, providing quantitative evidence for the distinct coordination environments. As revealed by the EXAFS fitting results in Fig. 3d, the Ru-M (M=Pt/Ru/Rh) bond length in PtSARuRh is notably longer than that in PtSARuRh-CrOx and PtSARuRh/CrOx. This elongation, concurrent with an intensified Ru-O bonding, suggests an enhanced oxidation state of Ru in PtSARuRh. Consistently, the Ru-M coordination numbers decrease from 4.8 in PtSARuRh-CrOx to 3.8 in PtSARuRh/CrOx. Under the same conditions, the Rh-M coordination numbers remain similar, at 3.4 and 3.5 for PtSARuRh-CrOx and PtSARuRh/CrOx, respectively. WT analysis of Ru K-edge (Fig. 3g) and Rh K-edge EXAFS data (Fig. S43) reveals analogous NCs coordination environments for both elements, aligning with the nearly complete overlap of Ru and Rh signals in STEM-EDX mapping. The Cr K-edge XANES and Cr-related EXAFS fitting results were shown in Figs. S44–S48 and Table S5. Based on the linear relationship between absorption edge energy and Cr oxidation state41, PtSARuRh-CrOx exhibited a higher average Cr oxidation state than PtSARuRh/CrOx (Fig. S49a), consistent with the XPS results in Fig. S22c. Furthermore, a comparison of the Cr K-edge XANES spectra for samples with and without PtSARuRh (Fig. S49b, c) showed that the Cr oxidation states in PtSARuRh-CrOx and PtSARuRh/CrOx were lower than those in their counterparts CrOx-Cr3+ and CrOx-CrO42−, respectively. This indicates electron transfer from the metal clusters to Cr, in agreement with the EELS results in Fig. 2c. FT-EXAFS spectra (Fig. S44b) displayed characteristic coordination paths of both Cr foil and Cr2O3 in amorphous CrOx NCs. Notably, the second-shell Cr-Cr coordination path exhibited a slightly longer bond distance than Cr-Cr in Cr foil and Cr2O3, reflecting the disordered nature of CrOx in both PtSARuRh-CrOx and PtSARuRh/CrOx samples42. The Cr-O coordination number decreased from 4.1 in PtSARuRh-CrOx to 3.5 in PtSARuRh/CrOx (Table S5), with increased oxygen coordination potentially contributing to the elevated Cr oxidation state. EXAFS curve-fitting revealed distinct Cr-Cr coordination paths between PtSARuRh-CrOx and PtSARuRh/CrOx, indicating divergent coordination environments. This discrepancy was further corroborated by WT analysis of Cr K-edge EXAFS data (Fig. S50). The maximum WT intensities for Cr-O (7.3 Å−1) and Cr-O-Cr (6.8 Å−1) coordination in K space of PtSARuRh-CrOx differed significantly from those in PtSARuRh/CrOx (Cr-O: 6.5 Å−1; Cr-O-Cr: 7.1 Å−1), reconfirming distinct Cr coordination environments. The presence of metal oxides clusters in each reference sample was further confirmed by FT-EXAFS and EDS mapping, as shown in Fig. S51. Electrochemical HOR and AEMFCs Performance The HOR activities of PtSARuRh-CrOx and reference samples were evaluated in H2-saturated 0.1 M KOH using a standard three-electrode system with a rotating disk electrode (RDE). As shown in Fig. 4a, 4b, PtSARuRh-CrOx exhibited the highest HOR performance, with its anodic current increasing rapidly with overpotential, outperforming commercial Pt/C-40%, PtRu/C-60% catalysts and other reference samples. The preferable performance of PtSARuRh-CrOx to PtSARuRh/CrOx, PtSARuRh, PtSARuRh/Cr2O3-tris (Fig. S52a), PtRu-CrOx (Fig. S52b), and PtRh-CrOx (Fig. S52c) highlights the critical role of CrOx and PtSARuRh NCs incorporation to enhance the HOR catalytic activity. The higher proportion of paired metal-metal oxide clusters in PtSARuRh-CrOx endows it with higher performance relative to PtSARuRh/CrOx, highlighting the correlation between the paired structure and the catalytic activity. Negligible HOR activity were observed in PtSA and PtSA-CrOx control samples in Fig. S53, indicting synergistic effects between all metallic components in the active catalyst. In addition, the prepared carbon-based control samples (Fig. S54) exhibited only marginal HOR activity, underscoring the essential role of the active sites within the metal-metal oxide cluster. The amount of NH3·H2O and CrO42− used in the synthesis has also been optimized and was found to play an important role in reaching the optimal HOR activity, as detailed in Figs. S55 and S56, respectively. Figure 4c presents the HOR polarization curves of PtSARuRh-CrOx at different rotation rates. The koutecky-Levith (K-L) plot yields a slope of 4.66 cm2 mA−1 s−1/2 (Fig. S57a), which aligns closely with the theoretical value (4.87 cm2 mA−1 s−1/2) for a two-electron transfer process43. The absence of oxidation currents in N2-saturated 0.1 M KOH (Fig. S57b) confirms that the measured currents originate exclusively from the HOR process. The conclusion further corroborated by the K-L plots of other reference samples (Fig. S58). Figure 4d shows the kinetic current density calculated by K-L equation, along with the curve fitted using the Butler-Volmer (B-V) equation, both reflecting the same performance trend observed in the linear sweep voltammetry (LSV) profiles. For quantitative comparison of intrinsic HOR activity, the kinetic current density (jk) and exchange current density (j0) were normalized by the electrochemical surface area (ECSA) and noble metal loading, yielding the specific exchange current density (j0,s) and specific kinetic current density (jk,s and jk,m). The ECSA and j0 of PtSARuRh-CrOx and reference samples were determined via copper underpotential deposition (Cu-UPD) stripping (Fig. S59a–e), and B-V curve fitting, respectively. The ECSA-normalized current density further underscores the preferable performance of PtSARuRh-CrOx compared to the reference samples (Fig. S59f). At an overpotential of 50 mV, the mass activities (jk,m) of PtSARuRh-CrOx reached 13.81 A mg−1, surpassing Pt/C (0.24 A mg−1) by 58-fold, PtRu/C (0.27 A mg−1) by 50.5-fold, PtSARuRh/CrOx (1.79 A mg−1) by 7.7-fold, and PtSARuRh (0.59 A mg−1) by 23-fold (Fig. 4e). Moreover, the specific kinetic current activity (jk,s) of PtSARuRh-CrOx attained 30.61 mA cm−2, significantly exceeding those of Pt/C (0.22 mA cm−2), PtSARuRh/CrOx (3.53 mA cm−2), and PtSARuRh (0.8 mA cm−2) (Table S6), underscoring its preferable kinetic activity. Furthermore, the ECSA-normalized j0,s values followed a consistent trend, with PtSARuRh-CrOx achieving an competitive j0,s of 1.06 mA cm−2 (Fig. 4e). ‌ Notably, PtSARuRh-CrOx presents competitive j0,s, jk,m and jk,s values (Table S6). HOR activity and AEMFC performance. a, b Polarization curves of PtSARuRh-CrOx and reference catalysts recorded in H2-saturated 0.1 M KOH electrolyte (pH 12.9 ± 0.1) at a scan rate of 5 mV s−1 and a rotating speed of 1600 rpm. The resistance in 0.1 M KOH was measured to be 42.3 ± 0.7 Ω. c Polarization curves of PtSARuRh-CrOx at different rotation rates. d Tafel plots of different catalysts and the Butler-Volmer fitting curves. e The ECSA normalized exchange current density (jo,s) and the mass normalized kinetic current density (jk,m) at 50 mV, with vertical error bars indicating standard deviations from three independent replicate measurements. f The chronoamperometry test at 100 mV (vs. RHE) of different catalysts in H2 saturated 0.1 M KOH. g, h Fuel cell performance of H2-O2 (g) and H2-Air (CO2-free, h) AEMFCs with commercial Pt/C (0.1 mgpt cm−2) and PtSARuRh-CrOx (0.05 mgPGM cm−2) at the anode and Pt/C (0.4 mgpt cm−2) at the cathode. Anode and cathode humidifiers were set at 90 °C and 92 °C, respectively (cell temperature: 92 °C). H2 and O2 or air (CO2-free) flow rates were set to 1.0 L min−1. i Long-term stability test of the H2-O2 fuel cell at 0.5 A cm−2 using the PtSARuRh-CrOx as the anode catalyst. Test conditions: cell temperature at 80 °C, H2 flow rate at 0.5 L min−1 and O2 flow rate at 0.5 L min−1, symmetric backpressures at 100 kPa. Source data for Fig. 4 are provided as a Source Data file. The stability of the PtSARuRh-CrOx catalyst for the HOR was systematically evaluated by employing a combination of electrochemical and physicochemical characterization techniques. PtSARuRh-CrOx demonstrated satisfactory stability in chronoamperometry tests, retaining 88% of its initial current density after 180 min at 0.1 V overpotential in H2-saturated KOH (Fig. 4f)—significantly higher than its counterparts. The robust activity was further confirmed in an extended 20 h test on carbon paper, which showed only a 10% performance decay (Fig. S60). The maintained structural integrity of the PtSARuRh-CrOx catalyst after stability testing was collectively attested by complementary techniques. HAADF-STEM images (Fig. S61) revealed no significant metal aggregation or particle growth. EXAFS analysis (Figs. S62-S64) demonstrated that the local coordination environments of Pt, Ru, Rh, and Cr were well preserved, while XANES and XPS spectra (Fig. S65) suggested only minimal changes in the oxidation states of Ru and Cr. The PtSARuRh-CrOx catalyst also exhibited good CO tolerance, as shown in Fig. S66, in H2-saturated 0.1 M KOH containing 1000 ppm CO, it retained 74% of its initial current density after 2500 s, significantly exceeding Pt/C (35%). LSV comparisons in the presence and absence of CO further highlighted its robustness: both PtSARuRh-CrOx and PtSARuRh/CrOx showed negligible activity loss, in stark contrast to the 31.4% decay observed for Pt/C (Fig. S67). To probe the interfacial reaction kinetics, potentiostatic electrochemical impedance spectroscopy (EIS) was performed and analyzed using a modified Armstrong-Henderson equivalent circuit (Fig. S68). PtSARuRh-CrOx exhibited the lowest charge transfer resistance (Rct) and hydrogen adsorption resistance (Rp) among all samples (Table S7), indicating reduced kinetic barriers for hydrogen adsorption/desorption and accelerated charge transfer processes. Leveraging the favorable intrinsic properties of PtSARuRh-CrOx NCs, its application potential in AEMFCs was demonstrated by coupling the HOR with the cathode ORR. The membrane electrode assembly (MEA) was fabricated by spraying-coating PtSARuRh-CrOx (0.05 mgPGM cm−2) onto one side of an anion-exchange membrane as the anode and commercial Pt/C (0.4 mgPt cm−2) onto the opposite side as the cathode. For comparison, a reference MEA was also prepared using Pt/C (0.1 mgPt cm−2) as the anode and Pt/C (0.4 mgPt cm−2) as the cathode. Figures 4g, 4h present the polarization curves of AEMFCs employing PtSARuRh-CrOx and Pt/C as the anode catalyst under H2-O2 and H2-air (CO2-free) conditions, respectively. Under H2-O2 operation (Fig. 4g), the PtSARuRh-CrOx-based AEMFC exhibited markedly competitive performance, achieving a peak power density (PPD) of 1430 mW cm−2 at a current density of 2.8 A cm−2. This value significantly outperforms that of Pt/C-based AEMFC, which reached a PPD of only 865 mW cm−2 at 2.3 A cm−2. Furthermore, PtSARuRh-CrOx also demonstrated clear advantages over PtSARuRh/CrOx and PtSARuRh control samples, as shown in Fig. S69. The performance advantage of PtSARuRh-CrOx was similarly pronounced under H2-air (CO2-free) conditions (Fig. 4h), where it delivered a peak power density of 1176 mW cm−2 at 2.4 A cm−2, markedly surpassing the 629 mW cm−2 achieved by Pt/C at 1.5 A cm−2. Notably, at the automotive-relevant voltage of 0.65 V42,44, the current densities delivered by the PtSARuRh-CrOx-based AEMFC were markedly higher than those of the Pt/C-based device: it achieved 1.90 A cm−2 in H2-O2 and 1.49 A cm−2 in H2-air (CO2-free), whereas the Pt/C-based AEMFC managed merely 0.83 A cm−2 and 0.64 A cm−2, respectively. It is important to emphasize that the desirable performance of PtSARuRh-CrOx was achieved with a low anode loading of just 0.05 mgPGM cm−2, further underscoring the intrinsic high activity of PtSARuRh-CrOx in AEMFCs performance. Consequently, the fuel cell employing PtSARuRh-CrOx achieved specific peak power densities of 28.6 W mgPGM−1 under H2-O2 and 23.5 W mgPGM−1 under H2-air (CO2-free) conditions, showing competitive performance for PGM-based AEMFCs (Fig. S70 and Table S8). Notably, the fuel cell with a low anode catalyst loading (0.05 mgPGM cm−2) exhibits only a 14.3% voltage decay after operating at a constant current density of 0.5 A cm−2 for more than 200 h under H2/O2 testing conditions (Fig. 4i). Mechanism Understanding Attenuated total reflection in situ Fourier transform infrared (ATR-FTIR) spectroscopy (Fig. S71a) was employed to investigate the CO tolerance mechanism of PtSARuRh-CrOx. In 1000 ppm CO/H2-saturated KOH, the characteristic band at 2077 cm−1 observed for PtSARuRh-CrOx (Fig. 5a) was attributed to Ru/Rh-H* species45 rather than CO adsorption, as confirmed by its persistence under pure H2 conditions (Fig. S72). Distinct bands at 1470, 1390, and 1310 cm−1 correspond to CO32− species46,47, suggesting adsorbed CO may undergo oxidation by OH* intermediates. In contrast, no CO32− or Ru/Rh-H signals were detected for PtSARuRh (Fig. 5b) and PtSARuRh/CrOx (Fig. S73) under the same conditions, while PtSARuRh/CrOx exhibited a linear CO adsorption (CO*L) band at 1980 cm-144, 46. Meanwhile, it displays a more intense M-OH* band for PtSARuRh-CrOx than PtSARuRh/CrOx and PtSARuRh via in situ Raman spectroscopy (Fig. S71b) under the same overpotential (Fig. 5c and Fig. S74), indicating the much preferable OH* adsorption ability of PtSARuRh-CrOx. The interfacial water structure during the alkaline HOR was investigated by analyzing the O-H stretching region (3000–3700 cm−1) in Fig. S75a, b. Spectra deconvolution identified three distinct components: symmetric hydrogen-bonded (HB) water (~3200 cm−1), asymmetric HB water (~3400 cm−1)—both representing strong HB networks and isolated water (~3600 cm−1) with weak HB interactions48–50. A higher proportion of strongly hydrogen-bonded water is known to facilitate proton-coupled electron transfer (PCET) by reducing the associated energy barrier, owing to more efficient proton transport within the interfacial hydrogen-bonding network50,51. As shown in Fig. S75c, PtSARuRh-CrOx exhibits a significantly greater content of strongly hydrogen-bonded water compared to PtSARuRh/CrOx. This enhanced water structuring is critical for promoting hydrogen transfer kinetics and contributes directly to the alkaline HOR activity of PtSARuRh-CrOx. Mechanism investigation. a, b In situ ATR-FTIR electrochemical spectra of (a) PtSARuRh-CrOx and (b) PtSARuRh in 0.1 M KOH of 1000 ppm CO/H2. c In situ Raman electrochemical spectra of PtSARuRh-CrOx and PtSARuRh. d, e In situ DEMS electrochemical spectra signal of (d) m/z = 17 and (e) m/z = 1 over PtSARuRh-CrOx, PtSARuRh/CrOx and PtSARuRh. f CV curves of catalysts collected in N2-saturated KOH solution without iR compensation; g CO-stripping measurements of samples in 0.1 M KOH without iR compensation. h H* adsorption energy on Ru site for PtSARuRh and PtSARuRh-CrOx cluster models. The inset shows the optimized adsorption configurations of H* at different adsorption sites. i OH* adsorption energy on Rh site for PtSARuRh and PtSARuRh-CrOx cluster models. The inset shows the optimized adsorption configurations of OH* at different adsorption sites. j The PDOS of Ru 4 d orbital for PtSARuRh-CrOx and PtSARuRh models. k The PDOS of Rh 4 d orbital for PtSARuRh-CrOx and PtSARuRh models. l The differential charge density distributions between PtSARuRh cluster and CrOx cluster as well as Bader charge analysis. Yellow and olive represent increased and decreased charge distributions, respectively. Source data for Fig. 5 are provided as a Source Data file. Differential electrochemical mass spectrometry (DEMS) tracked reaction intermediates through time-resolved m/z signals during potential sweeps using a home-made H-cell electrolyzer, as shown in Fig. S71c. Distinct m/z 17 (OH, Fig. 5d) and m/z 1 (H, Fig. 5e) signals confirmed OH* and adsorbed H* participation in HOR, with PtSARuRh-CrOx showing the most significant signal variation at +250 mV, indicative of more rapid kinetics than PtSARuRh/CrOx and PtSARuRh and matching well with the Tafel slops in Fig. 4d. Cyclic voltammetry (CV) measurements in Fig. 5f reveal that PtSARuRh-CrOx exhibits a weaker hydrogen binding energy (HBE) compared to both PtSARuRh/CrOx and PtSARuRh, which contributes to its enhanced HOR activity. CO stripping voltammetry was employed to evaluate the hydroxyl binding energy (OHBE). As shown in Fig. 5g, the onset potential for CO oxidation on PtSARuRh-CrOx is 0.56 V vs. RHE, which is 90 mV lower than that on PtSARuRh/CrOx (0.65 V), indicating a significantly facilitated OH adsorption process enabled by the paired CrOx structure. Combined with in situ ATR-FTIR and Raman analysis, these results confirm that the competitive HOR activity and CO tolerance of PtSARuRh-CrOx originate from the optimized balance between HBE and OHBE. The CO oxidation mechanism involves two critical steps: the favorable OHBE promotes the formation of surface OH* species, which then react with neighboring adsorbed CO* via the Langmuir–Hinshelwood mechanism, leading to efficient CO2 desorption52,53. Density functional theory (DFT) calculations were carried out to explore the possible origin of the enhanced HOR activity in PtSARuRh-CrOx. The atomic coordinates of all optimized computational models are provided in Supplementary Data 1. Based on the optimized structures (Fig. S76), HBE analysis on Ru sites (Fig. 5h) suggests a weaker hydrogen adsorption for PtSARuRh-CrOx, resulting from a downshifted Ru d-band center (−1.739 eV vs. −1.622 eV in PtSARuRh; Fig. 5j). This shift increases the electron population in antibonding orbitals, thereby destabilizing H* binding. In parallel, OHBE on Rh sites shows enhanced OH* adsorption (Fig. 5i), attributable to an upshifted Rh d-band center (−1.633 eV in PtSARuRh-CrOx vs. −1.647 eV in PtSARuRh; Fig. 5k) that reduces antibonding occupancy and strengthens OH* interaction. These computed adsorption energy trends are consistent with the electrochemical behavior observed in Figs. 5f-5g, supporting the role of CrOx in electronically modulating the catalytic sites. This electronic redistribution is supported by Bader charge analysis, which indicates net electron transfer to the CrOx cluster (Fig. 5l), in agreement with earlier EELS (Fig. 2c) and XANES (Figs. S49b–S49c) findings. Furthermore, we computed the energy evolution profiles of PtSARuRh-CrOx and PtSARuRh along the HOR pathway, evaluating three plausible mechanisms for PtSARuRh-CrOx (Fig. S77). Both Pathway 1 and Pathway 2 occur at the Ru/Rh sites, differing only in the allocation of adsorption sites: Pathway 1 involves H adsorption on Ru and OH adsorption on Rh, while this configuration is reversed in Pathway 2. Notably, both pathways share an identical minimum energy barrier of 0.19 eV, which is relatively lower than the barrier of 0.68 eV calculated for PtSARuRh (Fig. S78), consistent with the observed activity trend. In contrast, Pathway 3, where H adsorbs on Ru (as in Pathway 1) but OH binds to an interfacial Cr atom, exhibits a comparatively higher barrier of 0.85 eV. This stark contrast implies that CrOx does not directly participate as an active site, but rather enhances the HOR kinetics by electronically modulating the adsorption properties of the adjacent Ru and Rh sites. A general AMEA strategy for constructing paired metal-metal oxide sub-NCs was developed to enable the precise fabrication of diverse metal-metal oxide sub-NCs (e.g., PtSARuRh-MOx, M = Cr, Mo, V, etc.), overcoming the limitations of conventional methods. The optimized PtSARuRh-CrOx electrocatalyst demonstrates a mass activity of 13.81 A mg−1 and a specific activity of 30.61 mA cm−2 at 50 mV overpotential, far exceeding commercial Pt/C and PtRu/C catalysts. In a practical AEMFC with a low PGM loading of 50 μg cm−2, it yields a high peak power density of 1.43 W cm−2 (28.6 W mgPGM−1) and maintains stable operation for 200 h at 0.5 A cm−2. Mechanistic studies through in-situ spectroscopy and theoretical calculations confirm that the synergistic Ru/Rh sites within the structure optimally adjust the adsorption of H* and OH* intermediates. This work not only presents a highly active and stable electrocatalyst but also establishes a general synthetic paradigm for the precise design of multi-metallic sub-NCs, establishing viable avenues for advanced catalyst development in energy conversion technologies. Methods Experimental Section Chemicals and Materials. Carbon nanotube (CNT, >98%, Timesnano), ethanol (C2H6O, AR, Sinopharm Chemical), dopamine hydrochloride (C8H11NO2·HCl, 98%, Innochem), ammonia solution (25%-30%, Sinopharm Chemical), nitric acid (65%-68%, Sinopharm Chemical), RuCl3·xH2O (99%, Energy Chemical), H2Cl6Pt·6H2O (99%, innochem), RhCl3·3H2O (98%, innochem), CrCl3 (99%, Sinopharm Chemical), K2CrO4 (99%, Sinopharm Chemical), Na2MoO4 (99%, Sinopharm Chemical), NaVO3 (99%, Sinopharm Chemical), (NH4)2TiF6 (99%, Sinopharm Chemical), KOH (99%, Sinopharm Chemical), K2TaF7 (99%, Sinopharm Chemical), Na2ZrF6 (99%, Aladdin), isopropanol (C3H8O, 99.7%, Sinopharm Chemical), Nafion D-521 dispersion (5% w/w in water and 1-propanol) (Alfa Aesar). PtRu/C (60 wt.%) and Pt/C (40 wt.%) were obtained from Johnson Matthey (JM). Ultrapure water (Millipore, 16 MΩ cm−1) was used throughout all of the experiments. Synthesis of CNT-O Substrates. The pristine carbon nanotubes (CNTs) were immersed in concentrated nitric acid within a three-necked flask, ensuring complete submersion of the material. The reaction system was heated at 80 °C in an oil bath for 24 hours. After cooling, the mixture was subjected to repeated washing with deionized water through decantation until the rinse water became neutral (pH ≈ 7). The resulting oxidized carbon nanotubes (CNT-O) were then collected and dried under vacuum at 60 °C for 12 h. Synthesis of PtSARuRh-CrOx. A mixture of 100 mg CNT-O was dispersed in a solution containing 90 mL deionized water and 40 mL ethanol under ultrasonication for 30 min. To this suspension, 20.3 mg RuCl3·xH2O, 18.75 mg H2PtCl6·6H2O and 20.4 mg RhCl3·3H2O were sequentially added under continuous stirring for 10 mins. Subsequently, 37.3 mg K2CrO4 were sequentially added under continuous stirring for 10 min. Then, 500 mg dopamine hydrochloride dissolved in 10 mL deionized water was introduced, followed by the dropwise addition of 0.75 mL ammonia solution (28 wt%). The reaction mixture was stirred vigorously for 3 h. The resulting PtRuRhOxδ+/CrO42− precursor was collected by vacuum filtration, washed thoroughly with deionized water three times, and dried overnight in a vacuum oven at 60 °C to obtain the precursor. For thermal treatment, the PtRuRhOxδ+/CrO42− precursor was loaded into a tube furnace and annealed under a 5% H2/Ar atmosphere. The temperature was raised to 600°C at a heating rate of 5 °C/min and maintained for 2 h to yield the final catalyst, denoted as PtSARuRh-CrOx. Synthesis of PtSARuRh/CrOx, PtSARuRh-MoOx, PtSARuRh-VOx, PtSARuRh-TiOx, PtSARuRh-TaOx, PtSARuRh-ZrOx. The synthesis of PtSARuRh/CrOx, PtSARuRh-MoOx, PtSARuRh-VOx, PtSARuRh-TiOx, PtSARuRh-TaOx and PtSARuRh-ZrOx followed a protocol identical to that of PtSARuRh-CrOx, except that K2CrO4 was replaced with CrCl3 (51.2 mg), Na2MoO4 (46.5 mg), NaVO3 (23.4 mg), (NH4)2TiF6 (15 mg), K2TaF7 (20 mg), Na2ZrF6 (12.5 mg), respectively. Synthesis of PtSARuRh/Cr2O3-tris. The synthesis of PtSARuRh/Cr2O3-tris followed a protocol identical to that of PtSARuRh-CrOx, except that 0.75 ml NH3·H2O was replaced with tris-buffer (15 mL, 0.546 M). Synthesis of PtRu-CrOx, PtRh-CrOx. The synthesis of PtRu-CrOx and PtRh-CrOx followed a protocol identical to that of PtSARuRh-CrOx, without adding RhCl3·3H2O or RuCl3·xH2O, respectively. Synthesis of PtSARuRh. The synthesis of PtSARuRh followed a protocol identical to that of PtSARuRh-CrOx, without adding K2CrO4. Synthesis of CrOx-CrO42−, CrOx-Cr3+. The synthesis of CrOx-CrO42− followed a protocol identical to that of PtSARuRh-CrOx without adding noble metal precursors. The synthesis of CrOx-Cr3+ followed a protocol identical to that of CrOx-Cr2O42−, except that K2CrO4 was replaced with CrCl3. Synthesis of PtSA-CrOx. The synthesis of PtSA-CrOx followed a protocol identical to that of PtSARuRh-CrOx, with the omission of RuCl3·xH2O and RhCl3·3H2O. Characterizations The crystalline structures of as-prepared samples were characterized using a Rigaku D/max 2400 diffractometer. XRD datasets were recorded with Cu Kα radiation (λ = 1.5406 Å), where the 2θ scanning range was set from 10° to 90° at a scan speed of 10°/min. A FEI Titan Themis 60-300 X-FEG microscope integrated with a probe spherical aberration corrector served as the testing platform to capture HRTEM micrographs, collect EDS elemental information and record HAADF-STEM images. The surface chemical states of samples were explored by XPS measurements on a Thermo Scientific ESCA Lab 250Xi device with monochromatic Al Kα radiation (1486.6 eV). All acquired binding energy values were normalized referring to the adventitious carbon C 1 s peak at 284.8 eV. The quantitative content of metallic elements in various samples was measured via an Icap6300 inductively coupled plasma optical emission spectrometer (ICP-OES). XAS measurements were completed at BL1W1B (Beijing Synchrotron Radiation Facility) and BL11B (Shanghai Synchrotron Radiation Facility). Post-processing of acquired XAS raw data was finished with Athena and Artemis software, and wavelet transformation diagrams were generated by Fortran codes. To identify transient key intermediates in the reaction system, LC-MS tests were carried out on an Agilent 6545 LC/Q-TOF mass spectrometer fitted with an auto syringe pump, and all mass data were collected under positive electrospray ionization mode. Electrochemical measurements Electrochemical measurements were carried out on a CHI 760E electrochemical workstation via a standard three-electrode configuration. A glassy carbon rotating disk electrode (GCE, diameter of 5 mm) served as the working electrode, a graphite rod acted as the counter electrode, and an Ag/AgCl electrode was used as the reference electrode. Platinum sheets were adopted for potential calibration, acting as both working and counter electrodes in H2-saturated 0.1 M KOH alkaline electrolyte. Cyclic voltammetry tests were conducted at a scanning rate of 1 mV s−1, and the thermodynamic equilibrium potential was determined by calculating the average value of zero-current crossing potential against the Ag/AgCl reference electrode. All measured potential readings were standardized to the reversible hydrogen electrode (RHE) scale in accordance with the following conversion formula: Prior to testing, the glassy carbon electrode was sequentially polished with 300 nm and 50 nm alumina slurry to obtain a smooth mirror surface. The alkaline electrolyte was freshly prepared by dissolving 1.122 g of KOH pellets into ultrapure water and diluting to a final volume of 200 mL, with the steady pH maintained at 12.9 ± 0.1 after full dissolution. The electrolyte was used immediately without long-term storage to avoid carbon dioxide contamination. To prepare the catalyst ink, 2 mg of catalyst powder was dispersed into a mixed solution composed of 375 μL isopropanol, 100 μL deionized water and 25 μL 5 wt. % Nafion aqueous dispersion. The mixture was subjected to ice-bath ultrasonication for 30 min to form a homogeneous suspension. Afterwards, 7.5 μL of the well-dispersed ink was drop-cast onto the polished glassy carbon surface. For PtSARuRh-CrOx and contrast catalysts, the nominal precious metal loading on the electrode was controlled at 12 μg cm−2. Based on the ink concentration of 4 mg mL−2 and the noble metal proportion of 7.79% in catalyst, the actual loaded PGM mass on the 0.19625 cm2 GCE was calculated as 2.337 μg, and the electrode was dried naturally at room temperature before measurement. The commercial Pt/C and PtRu/C benchmarks were loaded at 40 μgPGM cm−2 and 50 μgPGM cm−2 respectively. HOR performance tests were implemented in H2-saturated 0.1 M KOH electrolyte at a rotating speed of 1600 rpm and a linear sweep rate of 5 mV s−1, with 95% iR compensation applied for all polarization curves. The uncompensated cell resistance was measured via the built-in iR compensation module of CHI760E workstation, giving a stable resistance value of 42.3 ± 0.7 Ω in 0.1 M KOH solution. Chronoamperometry measurement was conducted at a constant potential of 0.1 V vs. RHE to evaluate catalytic stability, and electrochemical impedance spectroscopy (EIS) was tested at 0.01 V vs. RHE with an AC amplitude of 5 mV over the frequency range from 100 kHz to 0.01 Hz. The kinetic current density (jk) of each electrocatalyst was calculated based on the Koutecky-Levich relationship: where j and jd represent the measured current density and diffusion-limited current density, respectively; B is the Levich constant, c0 refers to the solubility of H2 (7.33×10-4 mol L-1), and ν is the electrode rotating angular velocity. The Levich constant B was calculated by: in which n is the electron transfer number of HOR, F is the Faraday constant (96485 C mol−1), D is the diffusion coefficient of dissolved H2 (7.33×10−4 mol L−1), and ν denotes the kinematic viscosity of electrolyte (1.01×10−2 cm2s−1).The exchange current density (j0) was derived by fitting kinetic current data with the Butler-Volmer equation: where α is the charge transfer coefficient, R is the universal gas constant (8.314 J mol−1K−1), T is the test temperature (298.15 K), and η represents the overpotential of hydrogen oxidation. The electrochemical active surface area (ECSA) was quantified by the copper underpotential deposition (Cu-UPD) stripping method. Firstly, cyclic voltammetry scanning was carried out in N2-saturated 0.5 M H2SO4 at 20 mV s−1 to obtain stable background curves. Subsequently, Cu underpotential deposition was performed at 0.02 V (vs. Ag/AgCl) for 200 s in 0.5 M H2SO4 containing 5 mM CuSO4. Linear sweep voltammetry from 0 V (vs. Ag/AgCl) at 20 mV s−1 was then recorded to collect Cu stripping signals. The ECSA value was acquired by integrating the charge density of Cu-UPD stripping peak after deducting double-layer current, and normalized to the theoretical monolayer Cu adsorption charge of 420 μC cm−2 via the formula: For CO stripping measurement, the working electrode was held at 0.1 V vs. RHE in CO-saturated 0.1 M KOH for 10 min to complete CO adsorption on catalyst surface. Afterwards, the electrode was transferred to N2-saturated 0.1 M KOH to purge residual free CO in electrolyte. Cyclic voltammetry was then conducted in the potential window of 0–1.0 V vs. RHE at a scan rate of 20 mV s−1 to record CO oxidation stripping curves. AEMFC assembly and testing The synthesized PtSARuRh-CrOx or commercial Pt/C-40% served as the anode catalyst with loadings of 0.05 mgPGM cm−2 (PtSARuRh-CrOx) and 0.1 mgPGM cm−2 (Pt/Ccom), respectively. Commercial Pt/C (60 wt% Pt, Johnson Matthey) was employed as the cathode catalyst with a loading of 0.4 mgPt cm−2. The catalyst ink was prepared by ultrasonically dispersing the catalyst and ionomer (PiperION, 5 wt%) in a mixture of water and isopropyl alcohol (1:9 v/v) for 1 h. After peeling off the protective film, the ink was spray-coated onto both sides of a 20 μm-thick PiperION A20 membrane (Versogen) to prepare a catalyst-coated membrane (CCM) with an active area of 5 cm2. The CCM was immersed in 1 M KOH solution for 12 h (with solution replacement every 6 h), followed by thorough rinsing with deionized water to remove excess KOH. The rinsed CCM was assembled into a complete AEMFC with fluorinated ethylene propylene (FEP) gaskets, gas diffusion layers (SGL29BC), graphite bipolar plates containing 5 cm2 flow fields, and metallic current collectors. Fuel cell testing was performed using a Scribner 850e test station equipped with a back-pressure module. Hydrogen and oxygen/air (CO2-free) flow rates were set to 1.0 L min−1 with 200 kPa back pressure. For fuel cell performance tests, the anode and cathode humidifiers were set at 90 °C and 92 °C (cell temperature: 92 °C). For long-term stability tests, both humidifiers and the cell were maintained at 80 °C. In situ electrospectra measurements In situ Raman measurements. In situ Raman characterization was performed using a LabSpec 5 Raman spectrometer equipped with a custom-designed electrochemical in situ Raman cell accessory, employing a 633 nm laser excitation wavelength. Measurements were conducted under HOR conditions within a potential window of 0–0.3 VRHE. The electrolyte consisted of H2-saturated 0.1 M KOH aqueous solution. Electrochemical control was achieved using a CHI 760E electrochemical workstation, with Ag/AgCl electrode and graphite rod serving as the reference and counter electrodes, respectively. In-situ Attenuated Total Reflection Fourier-transform Infrared Spectroscopy (ATR-FTIRS). In situ ATR-FTIRS measurements were performed using a Nicolet iS50 spectrometer equipped with a liquid nitrogen-cooled MCT detector and a silicon attenuated total reflection (UATR) accessory. Experiments were conducted in 1000 ppm CO/H2-saturated 0.1 M KOH aqueous solution. Electrochemical control was implemented through a CHI 760E electrochemical workstation, with the catalyst drop-cast on a gold-coated silicon prism serving as the working electrode. Ag/AgCl electrode and graphite rod were employed as the reference and counter electrodes, respectively. Background spectra were continuously acquired prior to each measurement until signal stabilization was achieved. In situ DEMS Analyses. In situ differential electrochemical mass spectrometry (DEMS) analysis was conducted using an OAS100 system (Shanghai Linglu Co., Ltd.) in H2-saturated 0.1 M KOH aqueous solution. The DEMS measurement adopted a Linglu PTFE membrane, which has a pore size of < 20 nm, a thickness of 40 μm, and exhibits uniform pore distribution. Electrochemical measurements were performed in a home-made H-cell electrolyzer with a CHI 760E electrochemical workstation, employing Ag/AgCl electrode and platinum wire as the reference and counter electrodes, respectively. Intermittent oxidative potentials (+ 25 to + 650 mV vs. RHE) were applied during the testing protocol while maintaining continuous in situ DEMS monitoring. Upon stabilization of mass spectrometric signals to baseline levels, subsequent testing cycles were initiated under identical conditions. The DEMS measurements systematically recorded ion current signals at five distinct applied potentials. Calculation methods All spin-polarized density functional theory (DFT) computations for this system were conducted using the Vienna ab initio simulation package (VASP)54. The projector-augmented wave (PAW) method described the ionic cores, while the Perdew-Burke-Ernzerhof (PBE)55,56 functional within the generalized gradient approximation (GGA) accounted for exchange-correlation effects. A plane-wave basis set cutoff energy of 450 eV was employed, and van der Waals interactions were incorporated via the DFT-D3 method55. Geometry optimizations utilized a Gamma-centered k-point mesh for Brillouin zone sampling, with convergence thresholds set to 0.02 eV Å−1 for residual forces and 1 × 10−5 eV for total energy. Prior to free energy computations, the Ru22Rh18Pt2/Cr10O30 model was constructed based on ab initio molecular dynamics (AIMD) simulations conducted at 300 K. The AIMD simulations employed the Andersen thermostat57 and utilized the NVE ensemble. A time step of 2 fs was set for each simulation run, and equilibration of the systems was achieved over 2000 simulation steps. Free energy calculations employed the computational hydrogen electrode (CHE) model developed by Nørskov and colleagues58. Within this model, the free energy of a proton-electron pair (H++e−) is equated to half the free energy of gaseous hydrogen 1/2 H2 (g) minus the term 0.0592 × pH. All calculations were conducted at a constant potential of 0 V versus RHE. The free energy change (∆G) was computed using the following expression: Here, ∆E represents the electronic energy difference, ∆ZPE denotes the zero-point energy (ZPE) change, and ∆S signifies the entropy change at temperature T = 298.15 K. Entropy (S) and zero-point energy values for gaseous molecules were sourced from the NIST Chemistry WebBook (NIST55)59 database. Notably, DFT adopts idealized structural models and ignore some interfacial factors, thus the calculated results are limited to theoretically predicted trends. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Transparent Peer Review file Source data Source Data Acknowledgements We thank the BL11B station in Shanghai Synchrotron Radiation Facility (SSRF) and BL1W1B station in Beijing Synchrotron Radiation Facility for experimental support. Author contributions X.L. was primarily responsible for catalyst preparation, characterization, data analysis and manuscript drafting. X.Z. mainly conducted theoretical calculations on the catalytic properties of the materials. C.Z. carried out part of the MEA testing. T.G. conducted partial XAS measurements. Y.W. was mainly involved in catalyst synthesis. S.M. focused on constructing computational models for the materials. X.H. assisted in manuscript revision. A.H. and G.L. jointly supervised the project design, provided guidance on experimental protocols, and contributed to manuscript writing and revision. Peer review Peer review information Nature Communications thanks Wei Luo, Sung Jong Yoo and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available. Funding G.L. discloses support for the research of this work from the National Natural Science Foundation of China (52425201) and the New Cornerstone Science Foundation through the XPLORER PRIZE. A.H. discloses support for the research of this work from the National Natural Science Foundation of China (52572276). Data availability The datasets generated and analyzed during the current study are available in the main text and supplementary information files. Source data are provided with this paper. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Ali Han, Email: alhan@imr.ac.cn. Gang Liu, Email: gangliu@imr.ac.cn. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-74322-3. References Associated Data Supplementary Materials Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Transparent Peer Review file Source Data Data Availability Statement The datasets generated and analyzed during the current study are available in the main text and supplementary information files. Source data are provided with this paper.