Designing neighboring-site activation of single atom via tunnel ions for boosting acidic oxygen evolution

Abstract

Realizing an efficient turnover frequency in the acidic oxygen evolution reaction by modifying the reaction configuration is crucial in designing high-performance single-atom catalysts. Here, we report a “single atom–double site” concept, which involves an activatable inert manganese atom redox chemistry in a single-atom Ru-Mn dual-site platform with tunnel Ni ions as the trigger. In contrast to conventional single-atom catalysts, the proposed configuration allows direct intramolecular oxygen coupling driven by the Ni ions intercalation effect, bypassing the secondary deprotonation step instead of the kinetically sluggish adsorbate evolution mechanism. The strong bonding of Ni ions activates the inert manganese terminal groups and inhibits the cross-site disproportionation process inherent in the Mn scaffolding, which is crucial to ensure the dual-site platform. As a result, the single-atom Ru-Ni-Mn octahedral molecular sieves catalyst delivers a low overpotential, adequate mass activity and good stability.

Introduction

Proton exchange membrane (PEM) electrolyzes operating in acidic media are promising and sustainable for the large-scale production of clean hydrogen (H₂) fuels^1,2. However, the sluggish kinetics of the oxygen evolution reaction (OER) process, involving multistep proton and electron transfer, pose a significant challenge to practical efficiency^3,4,5. Despite massive recent advances in discovering robust durability and high activity for acidic OER, developing high-efficiency and cost-effective electrocatalysts remains challenging due to a limited selection of catalysts (such as Ru and Ir metal/oxide) with high costs and low earth abundance. Single-atom catalysts (SACs) have recently emerged as promising catalysts for maximizing atom utilization^6,7,8. Nevertheless, individual active atoms determine the single adsorption mode (end-on adsorption), which follows the conventional adsorbate evolution mechanism (AEM) and employs a single adsorption-dissociation relationship to explain the catalytic behavior. Specifically, the water molecule is deprotonated at the SAC to form OH* and O* species. The terminal O* evolves with the second H₂O to form OOH* and eventually deprotonates, leading to the evolution of oxygen^9,10. The presence of a scaling relationship that linearly correlates binding energies of the intermediates reduces the degrees of freedom available for catalyst optimization, resulting in an overdependence on the turnover frequency of active atoms.

In general, the structure with dual active sites typically gives rise to the tuning and optimization of multi-intermediate, such as diatomic oxygen mechanism (DOM) used for coupling the O–O bond as predicted by a Yeager model (bridge adsorption)¹¹. The DOM path allows direct O*–O* radical coupling without generating peroxide ions, thereby breaking the scaling relationship, and realizing a low-energy barrier^12,13. However, the DOM mechanism imposes more stringent requirements on metal active sites, including geometric configuration and catalytic path. For instance, the appropriate atomic distances, dual active sites with matched catalytic processes, and support without dynamic reconfiguration are expected to promote O–O radical coupling¹⁴. Most dual atomic catalysts to date utilize carbon as supporting substrates to adjust the appropriate atomic distance, but this approach faces challenges in direct application to acidic oxygen evolution due to poor stability in harsh liquid-phase electrolytes^12,15. Meanwhile, acid-resistant metal oxides, constrained by their fixed lattice arrangements, face challenges in achieving highly symmetric configurations of single atoms (SA) with appropriate spatial distances. In contrast to molecular catalysts with fixed spatial configurations, single atom embedded in metal oxides exhibit random distribution and variable spacing, making it difficult to form bridge adsorption configurations.

From another perspective, the ligand metal ions adjacent to the SA serve as potential active sites, enabling intramolecular O–O terminal formation during the OER polarization process. However, the supports are typically inert and cannot complete the entire catalytic cycle as a single site. How to stimulate the redox capacity of neighboring inert metal atoms to match the rate of active SA still binds the hypothesis. It is important to note that this process involves a catalytic cycle comparable to the action of SA, rather than the conventional detachment of auxiliary intermediates. Drawing inspiration from the oxo-bridged Mn₄Ca cluster of oxygenic photosynthesis¹⁶, manganese oxide octahedral molecular sieves (OMS) have been chosen as a promising framework for the intramolecular formation of O–O bonds. Unlike the conventional MnO₂, OMS is a non-stoichiometric oxide with a crystal structure like a nanocage that is internally hollow and acts as a reservoir for ions^17,18,19. This class of inorganic systems can trap a wide variety of ions, imparting novel physical and chemical properties. If the identity of the inserted ion can be controlled (such as alkali metal), the activity of the material towards proton-coupled electron transfer reactions can be enhanced by breaking the pH-dependent scaling relationships of the OER. Therefore, it is reasonable to assume that cavity ions can perturb the electronic rearrangement of SA and neighboring metal, activating the inert atoms to match the activity of the SA. In this scenario, a symmetric system with SA, referred to as the “single atom–double site” configuration, would be the ideal state to realize the direct intramolecular oxygen coupling that we are seeking.

In this study, we present a material that involves a dual-site Ru–Mn pair by elaborately loading atomic ruthenium on OMS equipped with tunnel Ni ions to activate inert Mn units. Ni ions are selectively doped at tunnel sites in the OMS to prevent a cross-site disproportionation for Mn sites. The introduction of Ni ions contributes to the activation of SA neighboring inert manganese atoms, creating a new Lewis active site that facilitates the intramolecular coupling of terminal oxygens. The resultant Ru_(SA)-NiOMS catalysts demonstrate dual-site-like kinetics towards the OER, exhibiting improved OER performance with an overpotential of 160 mV and good PEM stability, displaying negligible activity decay after 100 h at 100 mA cm^–2. Operando analysis suggests that the strong bonding of nickel ions can modify a partially electronic state along a single-site Ru–Mn pair, allowing the OH* intermediate to desorb over the Mn site, thereby addressing the rate-limiting step mismatch problem. This research underscores the significance of modulating the reaction configuration through tunnel ions and provides a novel perspective for achieving a complete transition from inertness to activity in the development of OER single-atom electrocatalysts.

Results

Theoretical predictions on the structure-performance relationship

The density functional theory (DFT) calculations were performed to predict the DOM possibility of tunnel modification strategy to guide catalyst design. The OMS models with different coordination structures were constructed, wherein Ni ions exist in the tunnel and are connected to eight [MnO₆] octahedra, forming a confined system. On this basis, partial Mn atoms are replaced by Ru to connect the Mn center for creating the Mn–O–Ru framework. In such a configuration, the units consisting of symmetric site-oxygen-site exhibit appropriate oxygen-oxygen coupling distances like the oxo-bridged Mn₄Ca cluster in the oxygen-evolving complex of photosystem (Fig. 1a–c and Supplementary Figs. 1–4). It is emphasized that active surfaces can suffer from blockage due to the adsorption of O*, OH*, or H₂O*. The matching of the evolution potentials for the dual sites is also crucial for the DOM mechanism, especially in the coupling between the highly active Ru and inert Mn.

Fig. 1: Mechanistic prediction of intermediates on the catalyst surface.

a Simulative structural model of symmetric dual active sites. b The illustration of the end-on model and bridge model. c Schematic structure models of Ru_(SA)-NiOMS, Ru_(SA)-OMS, NiOMS, and OMS from the top view. Surface Pourbaix diagrams of d Ru_(SA)-NiOMS, e Ru_(SA)-OMS, f NiOMS, and g OMS. The thermodynamically stable states of the surface under relevant reversible hydrogen electrode (RHE) and pH values are highlighted by green (for O*), blue (for OH*), and orange (for H₂O*), respectively. h Computed H₂O*, OH*, and O* adsorption free energies on catalyst surfaces. i Energy levels of theoretical redox potential (U_R) and theoretical coupling potential (U_L). The U_R is defined as the electrode potential of H₂O* deprotonation, and the U_L is defined as the limiting electrode potential needed to remove the surface OH* species for the theoretical formation of intramolecular O–O bonds.

Given the above considerations, surface Pourbaix diagrams were constructed to describe the deprotonation process by plotting surface terminations’ equilibrium potential as a pH function (Fig. 1d–g). As the electrode potential increases, water and hydroxyl molecules readily adsorb on the basal plane, and deprotonation is progressively favored on these surfaces (Fig. 1h, i). Specifically, the hydroxylated surface is the most stable structure at low potentials for the Mn and Ru sites of Ru_(SA)-NiOMS. As the potential increases, the adsorbed hydroxyls are oxidized further to O*, and the oxidation potentials on the dual sites are nearly identical. The introduced Ni species bring more electronic states for the orbitals, thus balancing the proton and electron transport in the active region of the dual sites. The nearly simultaneous deprotonation process opens the possibility of rate-matching for oxygen-oxygen coupling. Using the potential difference (U_R-U_L) as a descriptor, the surface deprotonation capacity can be easily estimated²⁰. It can be observed that the calculated U_R-U_L values of the Mn site of Ru_(SA)-NiOMS are smaller than those of OMS and exhibit a close degree to that of the Ru sites, suggesting a synchronized surface deprotonation ability under reaction conditions. The thermodynamic standard potentials are also consistent with the experimental results. Two redox features are observed in the cyclic sweep, corresponding to the successive deprotonation (Supplementary Fig. 5). In contrast, the Mn sites are terminated by O* at a higher potential (U_RHE > 1.33 V) for OMS, indicating an inhibited surface deprotonation capacity. Thus, it is challenging for OMS to form interconnected dual-site with the Ru SA. Therefore, introducing Ni ions to modify the Ru-Mn distance and intermediates coverage could activate the DOM pathway and thus increase the intrinsic activity of the acidic OER.

Preparation and characterization of Ru_(SA)-OMS and Ru_(SA)-NiOMS

To realize the above assumptions, SA catalysts with and without tunnel ions were synthesized through a cation exchange method (see the methods for details). Due to low doping of Ru, the X-ray diffraction (XRD) patterns show indistinguishable changes compared to pristine OMS, proving the formation of Ru SA (Supplementary Fig. 6 and Supplementary Table 1). Regular nanorods without any discernible metallic particles can be observed from the TEM images (Fig. 2a, b and Supplementary Fig. 7). The detailed atomic configuration of Ru_(SA)-OMS and Ru_(SA)-NiOMS were studied using HAADF-STEM. The basic OMS structure is resolved, where Mn is consistent with the [001] projection of the structure by the perfect structure model (Fig. 2c–e). Some extra atom columns are lining up between Mn columns, as shown by the green spheres, indicating that the Ni ions are statically inserted into the 2 × 2 tunnels of the OMS (Fig. 2d). The occupancy variation of the Mn column varies significantly, as observed in the green circle-marked brightness portion and the weak rest of the same column. Thus, Ru is atomically dispersed in Mn arrays, as further verified by the intensity profile along the black line, where the contrast is proportional to the square of the atomic number (Supplementary Figs. 8–10). Energy dispersive spectroscopic elemental mapping (EDS) confirmed a uniform dispersion of Ru dopant in catalysts (Fig. 2f).

Fig. 2: Atomic structure and elemental distribution of Ru_(SA)-OMS and Ru_(SA)-NiOMS.

TEM image of as-synthesized a Ru_(SA)-OMS and b Ru_(SA)-NiOMS nanorods. Atomic-resolution aberration-corrected high-angle annular dark-field (HAADF-STEM) images of the c Ru_(SA)-OMS and d Ru_(SA)-NiOMS nanorods viewed sideways along the [110] direction. The Violet spot indicates Mn atoms, green represents Ni, and orange indicates Ru atoms. e Perspective view of cryptomelane OMS along [001] direction. Ni ions are surrounded by eight [MnO₆] octahedra, forming a tunnel that extends along the [001] direction. Rotating along the a-axis presents an atomic model structure of the original STEM image. f HAADF-TEM image and element mapping images of Ni, Mn, O and Ru. FT-EXAFS spectra of the g Ru and h Mn K edge (a.u.: arbitrary unit). i Crystal orbital Hamilton population (COHP) analysis on the adjacent Mn, Ru, Ni sites of Ru_(SA)-OMS and Ru_(SA)-NiOMS, and the corresponding integrated crystal orbital Hamiltonian population (ICOHP) values.

The formation of Ru SA is further determined by Ru K-edge extended X-ray adsorption fine structure (EXAFS) spectroscopy. Ru_(SA)-OMS and Ru_(SA)-NiOMS deliver a prominent peak at ~1.45 Å from the Ru–O contribution and a relatively weak peak at ~2.68 Å from the Ru–Mn (Fig. 2g). There is no rise at ~2.32 Å from the metal Ru–Ru contribution and ~3.20 Å from the ruthenium(IV) oxide, confirming the single-atom character²¹. Moreover, the normalized X-ray absorption near-edge spectroscopy (XANES) spectra at the Ru K edge inferred that the Ru^IV atom forms the coordination structure of Ru–O–Mn based on the position of the absorption edge and white line peak (Supplementary Figs. 11–13). EXAFS spectra of the Mn K edge further probed the local bonding environment of the Mn species. EXAFS results demonstrate an approximately 5% Mn–Mn bond length shrinkage in Ru_(SA)-NiOMS (2.44 Å) compared with Ru_(SA)-OMS (2.56 Å), suggesting that the covalency contraction originates from the net effect of tunnel ions (Fig. 2h and Supplementary Figs. 14–18 and Supplementary Table 2). In other words, NiOMS provides an entirely different anchoring environment compared to OMS for Ru single sites, and the accommodation of Ni into the former leads to shortened Mn–Mn bonding length that is inclined to activate ligand oxygen^22,23. It is worth noting that Ru–Mn bonds have indistinguishable differences in either electronic structure or local coordination environment with or without Ni ions, primarily due to the low doping content of Ru (Supplementary Fig. 19).

Although introducing additional tunnel ions seems to have little effect on the bond length of Mn–O, the regulated bond strength enables moderate adsorption of *OH due to the charge redistribution, bringing the Mn–O closer to the Ru–O bonding strength. Therefore, the crystal orbital Hamilton population (COHP) further analyzed the bonding strength on the active centers of metal dimers (Fig. 2i)²⁴. The Ni ions weakly interact with three adjacent coordinated oxygen atoms, further causing the redistribution of electrons in the Mn-O and Ru-O bonds. The ICOHP value of the Ni-O bond is much lower than that of the Mn-O and Ru-O bonds, indicating a weak covalent interaction between Ni and the framework structure. In contrast to Ru_(SA)-OMS, the bonding state of Mn of Ru_(SA)-NiOMS gradually moves away from the Fermi energy level and is accompanied by a decrease in antibonding orbital filling, which contributes to an increase in the adsorption strength of oxygen intermediate²⁵. The bonding strength was qualified to gain quantitative insight into the net effect of tunnel ions by integrating the COHP up to the Fermi level. After the introduction of Ni, the ICOHP value of the Ru-O bond decreases from 2.542 to 2.536 eV, while the ICOHP value of the Mn-O bond increases from 2.006 to 2.325 eV. This results in the Mn-O and Ru-O bonds exhibiting similar bond strengths, leading to similar adsorption and desorption behaviors of intermediates (Fig. 1d), thereby achieving rate-matching. Interestingly, a good linear correlation exists between ΔG_{(*H2O, *OH, *O)} values and ICOHP. This finding reveals that the Ni ions can readily tune electronic structure by introducing new bonds qualitatively, which is correlated with the observed intrinsic activity.

Oxygen evolution driven by Ru–Mn pairs

The OER activities of Ru_(SA)-OMS and Ru_(SA)-NiOMS were evaluated in a conventional three-electrode setup with O₂-saturated 0.5 M H₂SO₄ aqueous solution and RuO₂ for comparison (Supplementary Figs. 20, 21). Ru_(SA)-NiOMS exhibits the most catalytically active polarization tendency, as the anode current density increases sharply from a potential of 1.34 V (Fig. 3a). The Ru_(SA)-NiOMS shows an overpotential of 160 mV at the geometric current density of 10 mA cm⁻², which far exceeds most of the currently reported SA catalysts (Supplementary Table 3). The electrochemical behavior during the redox transition was recorded to investigate the proton transfer phenomenon. The lack of Ru_(SA)-NiOMS hysteresis in the CV indicates that the polarization current is controlled by electron transfer kinetics (Supplementary Fig. 5). Thus, the coverage of adsorbed intermediates species was in equilibrium with precursor²⁶. The kinetics of the redox transitions could be assessed by comparison with the standard rate constant, which was defined from the steady-state Tafel²⁷. The slope of Ru_(SA)-NiOMS was found to be 32.78 mV dec⁻¹, implying that the rate-determining step (RDS) of OER was the rearrangement of chemical bonds with a pair adjacent terminal oxygen (M=O*) (Fig. 3b). In contrast, Ru_(SA)-OMS exhibits different reaction mechanisms such as the nucleophilic attack of H₂O molecule, resulting in a mediocre Tafel slope (68.13 mV dec⁻¹) due to the potential dependence of the intermediate coverage. Additionally, Ru_(SA)-NiOMS gives the highest mass activity of 18,966 A g⁻¹, which is higher than Ru_(SA)-OMS and reported catalysts, respectively (Fig. 3c and Supplementary Figs. 22, 23 and Supplementary Tables 4, 5).

Fig. 3: Acidic OER activity and stability investigations on RDE and PEM device.

a Polarization curves (current normalized by geometric area) of the as-synthesized catalysts, with RuO₂ as a comparison. b Actual steady-state Tafel plot constructed from OER current densities sampled from steady-state CA responses. c A comparison of mass activity and overpotential. d Chronopotentiometric curve obtained with constant current densities of 10 mA cm⁻². HAADF-TEM image of e Ru_(SA)-OMS and f Ru_(SA)-NiOMS after OER test. g Dependence of the Ru and Mn molar concentration in the electrolyte on the OER reaction process, inset: dissolved Ru mass percent in the electrolyte of RuO₂, Ru_(SA)-OMS and Ru_(SA)-NiOMS after 300 h tests. h The chronopotentiometry curve of Ru_(SA)-NiOMS catalysts operated at 100 mA cm⁻² in the PEM electrolyzer (25 °C ± 1 °C) with commercial Pt/C cathode catalyst. Insert in h show the photograph and schematic of the PEM electrolyzer device.

The overpotential of Ru_(SA)-NiOMS was almost constant throughout the continuous operation of 300 h, which is better than that of Ru_(SA)-OMS and commercial RuO₂ (Fig. 3d). In addition, the Faradaic efficiency is performed, revealing that the measured current primarily originated from water oxidation (Supplementary Fig. 24). The used catalysts were further characterized using HRTEM. The Ru_(SA)-NiOMS nanorods maintained the original tunnel structure with an interlayer distance of 6.91 Å, which is characteristic of a (110) plane of Ni ions intercalated OMS. Time-dependent inductively coupled plasma optical emission spectroscopy (ICP-OES) results show that less than 0.1% Ru and 0.03% Mn in the Ru_(SA)-NiOMS was dissolved into the electrolyte after OER stability testing. Notably, the surface of Ru_(SA)-OMS was covered with distinctive nanoparticles and an amorphous layer, which was dissolution redeposition phenomenon driven by the inherent instability of Mn³⁺ that occurred during the electrooxidation (Fig. 3e–g)^28,29. The Ru_(SA)-OMS were dissolved in the electrolyte, and the dissolved Mn ions were redeposited back onto the surface layer at the cost of the partial leaching of Ru. The unstable dynamic system prohibits Ru atoms from coming into oxo-oxo coupling with symmetric metal sites and is accompanied by irreversible ruthenium dissolution. In brief, Ru_(SA)-NiOMS does not suffer from a dynamic dissolution-deposition process and maintains a stable crystal structure. Finally, to investigate the practical application potential of Ru_(SA)-NiOMS, the PEM electrolyzer using commercial Pt/C as the cathode catalyst for HER and a Nafion proton exchange membrane was constructed (Fig. 3h and Supplementary Figs. 25, 26). Ru_(SA)-NiOMS catalyst showed improved PEM durability, with only a 0.01 V cell voltage increase after 100 h in our initial trial. The post-test structural characterization also verified that Ru_(SA)-NiOMS exhibits good activity and stability in acidic media, demonstrating great potential for future practical applications (Supplementary Figs. 27–31 and Supplementary Table 6).

Access to steady-state molecular platform for DOM path

In situ X-ray absorption spectroscopy (XAS) was employed to verify the robust molecular platform for DOM further and to reveal the local environments of the core metals^30,31,32. XANES spectra at the Mn K-edge of standard Mn oxides and Mn foil were determined to estimate the Mn valence state. With increasing Mn valence, the absorption edge shifts linearly to higher energies. The XANES spectra of the Mn K-edge collected in situ on Ru_(SA)-OMS films changed significantly at different applied potentials (Fig. 4a, b). At open circuit voltage (OCV), the spectra immensely resembled that of the air state, while at 0.80 V vs RHE, the spectra matched well with the Mn^2.4+O₂ reference that contains Mn²⁺. The process can be attributed to the surface active Mn³⁺ with a high spin state (t_2g³e_g) is rapidly consumed by disproportionation to form Mn²⁺ and Mn⁴⁺, resulting in no net charges passing across the electrode (Supplementary Fig. 32)²⁸. The valence state was initially Mn^2.4+ at 0.80 V, then gradually increased with potential to Mn^4.3+ at 1.60 V, and settled at Mn^4.3+ after OCV. The Ru_(SA)-OMS was reduced below the potential of ∼0.80 V vs RHE and oxidized above 1.40 V vs RHE. The disproportionation of Mn³⁺ in Ru_(SA)-OMS is potential-dependent, as evidenced by the fact that the disproportionation of Mn³⁺ is completely inhibited as the bias potential is increased to 1.40 V, which allows the high-valent manganese to remain in the lattice. In addition, the concentration of ionic species in solution shows a sharp fluctuating change (Supplementary Fig. 33). Therefore, the above data supported that the potential-dependent disproportionation with the electrooxidation deposition process represents a plausible mechanism for the observed changes in surface dynamic reconfiguration. For Ru_(SA)-NiOMS, an identical measurement was performed for alternating oxidation at 1.60 V and reduction at 0.80 V (Fig. 4e, f), where only the valence state at 0.8–1.2 V (pre-OER process) varied with the potential, and then remained in the steady state in agreement with the OCV. The characteristic suggests that Ru_(SA)-NiOMS avoided the instability of the disproportionation process and was also not over-oxidized during the OER process. In addition, a stable and reversible Mn redox was found throughout the cycle with reproducible changes, indicating fast redox kinetics. In comparison, the valence state hysteresis changes for Ru_(SA)-OMS are attributed to poor electronic conductivity and fast solvation kinetics³³.

Fig. 4: Operando X-ray absorption spectroscopy (XAS) analysis.

Operando a XANES and c EXAFS spectra of Ru_(SA)-OMS at the Mn K-edge recorded in flowing O₂-saturated 0.5 M H₂SO₄. The positive peak shift indicated the increase in valence. b Changes in the normalized absorbance of Mn in the operando K-edge XANES spectra of Ru_(SA)-OMS. d In situ investigations of Mn ligand changes with applied potential in Ru_(SA)-OMS. In situ e XANES and g EXAFS spectra of Ru_(SA)-NiOMS at the Mn K-edge cycled from 0.80 V to 1.60 V vs. RHE and back to OCV. Changes in the f normalized absorbance of Mn in the operando K-edge XANES spectra and h intensity under peak Mn–Mn during cycling for Ru_(SA)-NiOMS. i Energetic profiles for Ru_(SA)-OMS and Ru_(SA)-NiOMS transition from αto δ-phase, where Ni cations are present initially in the tunnel cavity. j Oxidation State evolution and self-consistent structure of Ru-Mn pairs involved in the electrooxidation of water to oxygen in acid condition.

Combining EXAFS spectra analysis, a more complete understanding of Mn–O hybridization as a function of potential can be yielded. EXAFS analysis of the Ru_(SA)-OMS at the Mn-K edge shows the typical Mn–O shell (1.41 Å) and di-μ-oxo bridging Mn–Mn shell (2.51 Å) at OCV (Fig. 4c, d and Supplementary Fig. 34a)³⁴. Before the OER, the coordination number of Mn–O in the first coordination shell decreases sharply, and its apparent distance fluctuates in the range of 0.05 Å distance. With a further increase in potential, a peak at a distance of about 3.0 Å exhibits typical fingerprints of a mono-μ-oxo-bridged Mn–Mn structure, indicating a more disordered structure. A reasonable explanation is that Mn–Mn could break in some regions under OER conditions because of the disproportionation of Mn–O and anisotropic tension. Notably, the significant structural change only occurs at high potentials, reversible upon potential removal. The evidence implies that the dissolved and redeposited surfaces are considered the active form of Ru_(SA)-OMS and the di-μ-oxo-bridged MnO₆ sites have no opportunity to function as a platform for DOM. For Ru_(SA)-NiOMS, the apparent distance change of the Mn–Mn and Mn–O paths is negligible, indicating the steady-state molecular platform (Fig. 4g, h and Supplementary Fig. 34b). Thus, introducing tunnel Ni ions to stabilize the easily disproportionable Mn–Mn molecular platform seems feasible.

The atom-level mechanism of the phase transition from the tunnel to the layer is further explained. The phase transition barriers of Ru_(SA)-NiOMS are consistent with the Operando XAS study (Fig. 4i and Supplementary Figs. 35–37). The present observation indicates that Mn³⁺ is inconstant relative to the disproportionation into Mn²⁺ and Mn⁴⁺ in acidic solutions. The dissociated surface further suffered electrochemical oxidation at a high potential and is caught in a dissolution-deposition (DI-DE) cycle (Fig. 4j). The disproportionation phenomenon is entirely suppressed on Ru_(SA)-NiOMS, which enables the Mn³⁺ to remain in the OMS. To summarize, Ni ions are effective tunnel fillers to control the kinetics in structure transition, which enhances the phase transition barrier and preserves polyhedron connectivity by inhibiting the disproportionation reaction.

Chemical recognition of peroxo-like species from the DOM

The capacitive behavior originating from the redox feature is potential-dependent, manifested in the charge storage mechanism^35,36. The intensity of white lines shows that the Ru oxidation state remains constant in Operando XAS, which implies that the charge accumulation and dissipation mainly rely on the Mn valence change (Supplementary Fig. 38). We conducted in situ EIS to analyze the charge accumulation process of the Mn sites during OER (Fig. 5a and Supplementary Figs. 39–41). From the complex capacitance analysis, the increase of the real capacitance of Ru_(SA)-NiOMS indicated that charge was accumulated by the Mn valence state change below 1.35 V vs. RHE. Above 1.35 V, the real capacitance decreased, implying that the accumulated charge was consumed to supply the OER. Also, a dissipative peak emerged in the imaginary capacitance plots at the low frequency, representing charge dissipation on the catalytic surface. In this regard, Ru_(SA)-NiOMS shows an improved efficiency and energy equilibrium for charge accumulation and dissipation compared to Ru_(SA)-OMS due to the Mn³⁺ being rapidly consumed by disproportionation, leaving no net charge passing through the electrode. The in situ attenuated total reflectance surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS) was performed under acidic water oxidation conditions to probe the OER mechanism. An evident absorption band at a vibration frequency of 1023 cm⁻¹ was observed for Ru_(SA)-OMS, which corresponds to the O–O–H bending, indicating that the formation of *OOH involved in the AEM-type OER was the rate-determining step (RDS), as the low initial content of *OOH (Fig. 5b)³⁷. Notably, typical linearly bonded superoxo species (metal–O–O) consistent with oxygen bridges in the DOM mechanism were detected, and bare characterization at 1023 cm⁻¹ was observed to negate AEM (Fig. 5c)³⁸. The observation is consistent with theoretical predictions of the DOM pathway in which oxygen coupling is the rate-limiting step.

Fig. 5: Identification of acidic OER mechanism.

a Complex capacitance plots (C′: the real capacitance represents the ability of the system to store electrical energy and C″: the imaginary capacitance indicates losses in the form of energy dissipation.) as a function of potential for Ru_(SA)-OMS and Ru_(SA)-NiOMS at a low frequency (10⁻¹Hz). Operando attenuated total reflectance surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS) recorded in the potential range of 0.90–2.40 V versus RHE for b Ru_(SA)-OMS and c Ru_(SA)-NiOMS. d Schematic illustration of the operando electrochemical mass spectrometry (DEMS). DEMS signals of ³²O₂, ³⁴O₂, and ³⁶O₂ from the gaseous products for ¹⁸O-labeled e Ru_(SA)-OMS and f Ru_(SA)-NiOMS.

The isotope-labeled operando differential electrochemical mass spectrometry (DEMS) was conducted to prove the DOM mechanism further. As depicted in Fig. 5d–f, Ru_(SA)-OMS produced ³²O₂ and ³⁴O₂ as reflected by the mass signals, and the ³⁶O₂ product was not distinguishable from the noise in the baseline, while the ³⁶O₂ signals were discernible on Ru_(SA)-NiOMS as the OER proceeds through the DOM pathway. Therefore, the DEMS data support the notion that O–O bond formation occurs by intramolecular coupling between oxygen atoms ligated to Ru and Mn at the Ru_(SA)-NiOMS surface (Supplementary Figs. 42–44)^39,40. In addition, the OER activity was not significantly changed over pH units, confirming the violation of the AEM pathway. Indeed, the Tafel plots at fixed potentials yielded a zeroth-order dependence on proton concentration in acidic conditions (Supplementary Fig. 45). As each redox corresponds to a single electron transfer, such a constant ratio excludes the combined effects of multiple processes, such as the simultaneous contribution of redox changes from oxygen ligand LOM (lattice oxygen oxidation mechanism) as previously suggested^41,42. In short, the operando analysis and the ex situ experiment prove that the OER proceeds via the DOM mechanism in Ru_(SA)-NiOMS.

Kinetics and mechanistic studies

The mechanistic divergence of OMS, Ru_(SA)-OMS, and Ru_(SA)-NiOMS might be understood by considering the difference in electronic properties of metal atoms. Ru incorporation essentially decreases the redox potential of M–OH to M=O, enabling catalysis at a much lower overpotential. The second metal (tunnel ions) promotes the O–O bond formation in the RDS as well. Due to the molecular nature of the dual-site catalysts, we compared their mechanisms to those of single-site catalysts. For OMS, the Tafel slopes greater than 120 mV dec⁻¹ are classified as quasi-infinite and consistent with an initial turnover-limiting chemical step⁴³. Thus, the reaction has an RDS involving one electron transfer. According to the XAS data and literature, the oxidation state of Mn in the resting state is lower than +4⁴³. Terminal Mn–OH groups, not terminal Mn=O, are favored in acidic solutions. Thus, OER formation would have to be accompanied by proton-coupled electron transfer (PCET) to the electrode, which an infinite Tafel slope cannot generally accommodate. Nevertheless, the disproportionation is likely coupled with a cross-site proton transfer on the surface, which limits the process rate due to the trouble required to access adjacent Mn^IV=O sites (Fig. 6a). In other words, four adjacent Mn^III are necessary to generate a pair of neighboring Mn^IV sites. The probability of sampling a reaction coordinate with four consecutive Mn^III sites is probably small⁴³.

Fig. 6: Proposed mechanism for the acidic OER.

a The turnover-limiting cross-site proton-coupled disproportionation mechanism scheme for OMS. b The conventional acid-base mechanism scheme for Ru_(SA)-OMS. c The oxo-oxo direct coupling mechanism scheme for Ru_(SA)-NiOMS. d The free-energy description scheme of the two reaction pathways (AEM and DOM) for O–O bond formation described experimentally are shown, with the distributed configurations of the crucial intermediates represented as the Gaussian distribution. The low-energy-barrier pathway produces surface -OO- species by oxo-oxo coupling, whereas the high-energy-barrier creates surface -OOH species by nucleophilic attack. The free energy (ΔG) diagrams of e AEM and f DOM for Ru_(SA)-OMS and Ru_(SA)-NiOMS.

The Ru_(SA)-OMS exhibits a Tafel slope of 68.13 mV dec⁻¹, indicating that the OER process has a quasi-equilibrium step (PES) involving one-electron transfer, followed by a chemical RDS. According to the XAS and electrochemical data, the active sites were identified as Ru atoms instead of Mn, and the oxidation state of Ru in the resting state is +4 (Supplementary Fig. 38). Thus, Ru^IV=O should be the resting surface termination state. Besides, the isotope detection of Ru_(SA)-OMS is slight and independent of the applied potential, indicating lattice oxygen is not involved in OER reaction. According to these data, the formation of Ru^IV=O from Ru^III–OH is the PES, and the nucleophilic attack of water to Ru^IV=O from Ru^III–OOH is the RDS (Supplementary Figs. 46, 47). The remaining steps in the catalytic cycle can be inferred from spectroscopic evidence of Ru^III–OOH. A similar step was proposed, consistent with what is commonly offered for OER (Fig. 6b). In this mechanism, a neighboring Mn site may serve as an acceptor for electrons or protons to accelerate reactions under certain circumstances⁴⁴. However, there is no bimetallic cooperation mode in Ru_(SA)-OMS because the potential of the first-row transition metal Mn–O units to produce O=Mn–O units is higher than the potential for Ru=O to generate OER (Fig. 6e and Supplementary Fig. 48). This assertion can be gleaned from the evidence that the Mn=O terminal group appears after the potential at which OER begins to occur.

The combination of a peroxo dimer (O=Ru^IV–O–Mn^IV=O) was proposed for Ru_(SA)-NiOMS, and the path is confirmed as it observes a Tafel slope (32.78 mV dec^–1) consistent with the value given for the coupled terminal oxidation mechanism (Fig. 6c). The proposed mechanism for O–O bond formation relies on the high valent M=O group. In situ XAS of Mn confirms that the catalyst resides in a Mn^IV resting state. These Mn^IV centers localized at the edges can be considered to have significant Ru^IV oxyl radical character, a direct consequence of the electronic intervention embodied by the Ni ions. Specifically, in the acidic condition, the resting state of the Mn^III edge site undergoes a turnover-limiting cross-site proton-coupled disproportionation for the pristine Mn unit. The unfavorable disproportionation state leads to a slow Mn–OH deprotonation step, to the extent requiring high polarization to produce Mn^IV=O terminal recombination to release O₂. The Mn=O end-groups are also challenging to match with the reaction potential of Ru=O simultaneously, due to the structural instability caused by solvation-redeposition resulting from disproportionation. Fortunately, the tunnel nickel ions serve as electron-supplying and receiving units to activate the faster deprotonation of Mn in a single point and ensure structural stability through the new bonding in conjunction with the strengthened Mn–O bonding, leading to bypassing of the proton-coupled disproportionation process (Fig. 6f and Supplementary Figs. 49, 50). Indeed, The distributed reactant configurations of the Mn site at resting state determine which transition state is reached (Fig. 6d).

In summary, we presented the concept of “single atom–double site” to design a neighborhood-activated single-atom catalyst. This can be achieved by introducing tunnel Ni ions to modify the reaction configuration of Ru-Mn pairs. The reaction configuration change induced by the intercalation can endow with the thermodynamically breakthrough in an inert Mn site, as revealed by in situ ATR-SEIRAS and DEMS characterization. The behavior of the Mn site is distinct from the conventional use of auxiliary intermediates detachment, but rather a catalytic cycle similar to the role of SA and behaves as DOM. It is further concluded that the DOM mechanism highly depends on subtle factors, including the appropriate distances between Ru–Mn pairs, synchronized terminal oxygen state, and a reconstruction-free support. The obtained Ru_(SA)-NiOMS is capable of continuous operation for more than 100 h in PEM without current decay and achieves an excellent overpotential (160 mV) and mass activity (18,966 A g⁻¹). These dual-site catalysts offer an attractive direct intramolecular oxygen coupling platform by bridging the active SA and activated metal sites.

Methods

Preparation of manganese oxide octahedral molecular sieves (OMS)

KMnO₄ (1.8 mmol) and MnSO₄·H₂O (0.6 mmol) were added into a 50 mL Teflon-lined autoclave containing 35 mL deionized water under 1 h magnetically magnetic stirring to give a uniform slurry. Then, the reactor was heated to 180 °C and maintained for 12 h. The obtained product was immersed in a 1 M HNO₃ solution for 8 days and changed every 2 days. After that, the product was washed with copious amounts of deionized water to remove free ions adsorbed onto precipitates and dried at 60 °C.

Preparation of manganese oxide octahedral molecular sieves with Ni tunnel ions (NiOMS)

Similar to OMS, KMnO₄ (1.8 mmol), MnSO₄·H₂O (0.6 mmol), and NiCl₂·6H₂O (0.2 mmol) were transferred to a 50 mL Teflon-lined stainless-steel autoclave and heated at 180 °C for 12 h. The obtained product was then placed in a 1 M NiCl₂·6H₂O solution to exchange needless K ions with Ni ions. After soaking in solution for 8 h, the manganese oxide was placed in a fresh solution of NiCl₂ for another 8 h. This ion exchange process was repeated for three exchanges, resulting in a complete transformation of the NiOMS.

Preparation of Ru_(SA)-OMS and Ru_(SA)-NiOMS

In this, 50 mg of OMS or NiOMS was stirred with 5 mM RuCl₃·xH₂O aqueous solution for 12 h at 60 °C, washed by centrifugation, dried under vacuum and calcined at 200 °C for 90 min in the air to obtain Ru_(SA)-OMS and Ru_(SA)-NiOMS. Ruthenium atoms were obtained by chemical substitution of Mn atoms on the oxide supports. Therefore, the ruthenium atoms are uniformly distributed on the surface of the catalyst. Mass loading a of the Ru element was determined by ICP, and the relevant data are shown in Supplementary Table 1. For comparison, a series of Ru_(SA)-NiOMS-X (X refers to the concentration of Ni elements) was also synthesized as described above. As a representative, Ru_(SA)-NiOMS in the following refers to the best performing catalyst Ru_(SA)-NiOMS(2) unless otherwise stated.

Physicochemical characterizations

The morphologies of catalysts were observed by SEM (SEM, JEOL JSM-6700 F), HR-TEM (TEM-EDX, Philips Tecnai F20, 200 kV), and aberration-corrected transmission electron microscope (ACTEM, JEOL JEM-ARM200F). The elemental compositions were analyzed by ICP (ICP-OES, inductively coupled plasma optical emission spectroscopy). The crystal structure of the samples was characterized by X-ray powder diffraction (XRD, Bruker D8 Discover) with Cu Kα radiation (λ = 1.5418 Å). The chemical valence state and surface atomic ratio were collected by X-ray photoelectron spectroscopy (XPS, ESCALAB 250Xi). The extended X-ray absorption fine structure (EXAFS) was measured at Taiwan Photon Source (TPS) beamline, 44A Quick-scanning X-ray absorption spectroscopy (XAS), in National Synchrotron Radiation Research Center (NSRRC), Hsinchu, Taiwan. Resonance Raman spectra were conducted on a confocal Raman microscope (Invia Reflex) with 532 nm wavelength at the sample surface.

Electrochemical in situ ATR-SEIRAS experiments

ATR-SEIRAS measurements were performed by a Nicolet iS50 FT-IR spectrometer with a liquid nitrogen-cooled MCT detector and a fixed-angle IR optical path. The spectral resolution of the measurements was 8 cm⁻¹, and 32 interferograms were added for each spectrum. The working electrode was prepared in two steps by chemically depositing an ultra-thin Au film on a silicon crystal to enhance the IR signal and conduct electrons. Then, the catalyst slurry was dropped onto the Au surface with a loading of 0.1 mg cm⁻² (catalyst slurry ratio: 7 mg catalyst, 3 mg carbon black dispersed in 1 mL ethanol and 50 μL Nafion added (sonication for 30 min)). The prepared working electrodes were mounted in an electrochemical three-electrode cell with Ag/AgCl as the reference electrode, Pt foil as the counter electrode, and the Ar-saturated 0.5 M H₂SO₄ was used as the electrolyte for the OER reaction. All measurements were performed using linear scanning voltammetry (LSV) to analyze the OER reaction intermediates at different potentials.

In situ XAFS measurements

In situ X-ray absorption spectroscopy, including XANES (X-ray absorption near-edge structure) and EXAFS (extended X-ray absorption fine structure) at Ru K-edge, were collected in total-fluorescence-yield mode using a silicon drift detector in BL-44A at National Synchrotron Radiation Research Center (NSRRC), Taiwan. The measurement in a typical three-electrode setup as the same condition in the electrochemical characterization case was performed in a specially designed Teflon container with a window sealed by Kepton tape. Subtracting the baseline of pre-edge and normalizing that of post-edge obtained the spectra. EXAFS analysis was conducted using the Fourier transform on k2-weighted EXAFS oscillations. All EXAFS spectra are presented without phase correction.

Electrochemical measurement

The electrochemical characterizations were conducted in a three-electrode system with a Pt plate and the saturated Hg/Hg₂SO₄ electrode used as the counter and reference electrodes. The measurements were carried out on an Autolab electrochemical workstation (Autolab Instrument) at 25 °C ± 1 °C. The active catalyst (8 mg), carbon black (2 mg) and binder (Nafion) were mixed, and isopropyl alcohol was used as the solvent. The viscous slurry was uniformly coated on glassy carbon electrode and dried under infrared light. The measured potentials versus Hg/Hg₂SO₄ were converted to the reversible hydrogen electrode (vs. RHE) using the following equation: E_RHE = E_Hg/Hg2SO4 + 0.198 V + 0.0656 pH. The potentials were corrected through a manual post-correction approach according to the formula: E = E_applied - iR, where i is the current flowing through the cell, and R is the ohmic resistance of the cell. In situ electrochemical impedance spectroscopy (EIS) measurements were carried out at the specified potential from 1.10 V to 1.60 V (vs. RHE) in the frequency range of 0.01–100 kHz with an AC amplitude of 5 mV. The double-layer capacitance (C_dl) was obtained by collecting CV curves with 20 to 100 mV s⁻¹ scan rates.

Calculation of the specific current density per electrochemically active surface area (ECSA)

The electrochemical double-layer capacitance (C_dl) was determined by measuring the capacitive current associated with double-layer charging from the scan-rate dependence of cyclic voltammetry stripping. The C_dl was estimated by plotting the ∆j = (j_a − j_c), where j_c and j_a are the cathodic and anodic current densities, respectively, against the scan rate, in which the slope was twice that of C_dl.

Calculation methods

All DFT calculations were performed using the Vienna Ab initio Simulation Package (VASP)⁴⁵. The projector augmented wave (PAW)⁴⁶ pseudopotential with the PBE⁴⁷ generalized gradient approximation (GGA) exchange correlation function was utilized in the computations. All energetics of metal oxides were calculated using the DFT with the Hubbard-U framework (DFT+U) to account for strongly localized d-electrons for Mn. The Hubbard-U correction terms were at U_eff(Mn) = 3.9 eV as obtained via linear response theory. The cutoff energy of the plane waves basis set was 500 eV, and a Monkhorst-Pack mesh of 3 × 3 × 1 was used in K‐sampling. All structures were spin-polarized, and all atoms were fully relaxed with the energy convergence tolerance of 10⁻⁵eV per atom, and the final force on each atom was <0.05 eV Å⁻¹.

The adsorption energy of reaction intermediates can be computed using the following equation:

$${\Delta {{{\rm{G}}}}}_{{{{\rm{ads}}}}}={{{{\rm{E}}}}}_{{{{\rm{ads}}}}}-{{{{\rm{E}}}}}_{*}+\Delta {{{{\rm{E}}}}}_{{{{\rm{ZPE}}}}}-{{{\rm{T}}}}\Delta {{{\rm{S}}}}$$

(1)

Where ads are intermediates (OH*, O*, H₂O*), and (${{{{\rm{E}}}}}_{{{{\rm{ads}}}}}-{{{{\rm{E}}}}}_{*}$) is the binding energy, $\Delta {{{{\rm{E}}}}}_{{{{\rm{ZPE}}}}}$ is the zero-point energy change, $\Delta {{{\rm{S}}}}$ is the entropy change. In this work, the values of $\Delta {{{{\rm{E}}}}}_{{{{\rm{ZPE}}}}}$ and $\Delta {{{\rm{S}}}}$ were obtained by vibration frequency calculation.

The Gibbs free energy of the reaction steps can be calculated by the following equations:

$$ \ast+{{OH}}^{-}-{{OH}}^{*}+{e}^{-}$$

(2)

$${{OH}}^{*}+{{OH}}^{-}-{O}^{*}+{H}_{2}O+{e}^{-}$$

(3)

$${O}^{*}+{{OH}}^{-}-{{OOH}}^{*}+{e}^{-}$$

(4)

$${{OOH}}^{*}+{{OH}}^{-}- \ast+{O}_{2}+{H}_{2}O+{e}^{-}$$

(5)

The transition state searches are performed using the Dimer method in the VTST package. The final force on each atom was <0.5 eV Å⁻¹. The TS search is conducted by using the climbing-image nudged elastic band (CI-NEB) method to generate initial guess geometries, followed by the dimer method to converge to the saddle points.