Interface-induced tensile strain stabilizes ruthenium oxide for high-performance acidic water oxidation

Catalyst synthesis and structural characterization

As illustrated in Fig. 1a, we developed a unique “bottom-up” strategy for the synthesis of catalysts (details in “Methods”). Our synthesis strategy begins with carbon dots (CDs) on the basis of the following motivations: (i) Their small size offers a high density of surface anchoring sites for metal atoms. (ii) In addition, the oxygen-rich surface groups (e.g., –OH, –COOH) on CDs enable strong chelation/coordination with metal cations, promoting uniform dispersion and stable anchoring of the metal cations on the CDs surface¹³. Such strong interactions promote interconnection of the dots and drive their self-assembly into lamellar aggregates during freeze-drying of the Ru³⁺/CDs mixture^13,41. Subsequent annealing reduction under Ar/H₂, followed by oxidative annealing in O₂, yields the final high density of the carbon-supported RuO₂ (RuO₂-C) catalyst.

Specifically, CDs were first prepared via a standard electrochemical etching method^42,43. Transmission electron microscopy (TEM) revealed uniformly dispersed particles with an average diameter of 3.24 nm, no obvious aggregation and without observable large graphene-like sheets (Fig. 1b). High-resolution TEM (HRTEM) and the corresponding intensity profile (Fig. 1c) resolve lattice fringes with a spacing of 0.21 nm, which is consistent with the (100) plane of graphite. The Fourier transform infrared (FTIR) spectrum of the CDs shows characteristic bands at 3442 and 1720 cm⁻¹, corresponding to v (O–H) and v (C = O) stretching vibrations⁴³, respectively, indicating abundant hydroxyl and carboxyl groups on the CDs surface (Supplementary Fig. 1).

Next, the as-prepared CDs were used to synthesize RuO₂-C catalysts with varying Ru loadings. Specifically, equal volumes of RuCl₃ solution were added to different volumes of the CDs dispersion to obtain a series denoted as RuO₂-C-x, where x represents the number of different contents of Ru. As shown in Supplementary Fig. 1, upon introducing RuCl₃—which itself has no characteristic infrared peaks—into the CDs solution, the FTIR spectra of the resulting RuCl₃ + CDs composite retained the v (O–H) and v (C = C) stretching vibration peaks, whereas the stretching vibration peak of the C = O bond significantly weakened or even disappeared. These suggest that Ru³⁺ may coordinate with the C = O groups on the CDs, leading to partial disruption of bonds and confirming the strong anchoring interaction between the metal ions and the CD surface.

The mixture was then freeze-dried and subjected to a two-step thermal protocol—Ar/H₂ atmosphere pre-annealing to consolidate/carbonize the CDs-derived framework, followed by a mild oxidative anneal in O₂ to form rutile RuO₂ and consolidate interfacial coupling. The O₂-annealing parameters were systematically optimized by screening temperature, duration, and oxygen partial pressure (Supplementary Figs. 2–7), revealing that the oxidative annealing process markedly enhanced acidic OER performance. Moreover, the carbon framework—formed by self-assembly of CDs after anchoring metal ions—plays a key role in regulating the spatial distribution of metal species, suppressing particle aggregation, and ultimately tuning catalytic performance^9,12,13. As summarized in Supplementary Figs. 8–15, systematic optimization of the synthesis parameters (Ru/C ratio) identified RuO₂-C-1 as the optimal sample; it exhibits distinct powder and aqueous appearances from the C-free RuO₂ reference (Supplementary Fig. 15a, b) and a Ru mass fraction of 9.1 wt% as determined by ICP-AES (Supplementary Fig. 16 and Supplementary Table 1).

The crystal structure of RuO₂-C-1 was confirmed by X-ray diffraction (XRD, Supplementary Fig. 17). The broadened diffraction peaks can be indexed to rutile RuO₂ (JCPDS 88-0322), consistent with nanosized RuO₂. Due to the low Ru content and quartz-substrate background, the (110)/(101) reflections are weak in laboratory XRD; therefore, synchrotron XRD was further performed to improve the signal-to-noise ratio and more clearly resolve the rutile features (Fig. 1d), with the (110) reflection at 2θ = 13.01° corresponding to an average tensile lattice expansion of ~1.31%. Scanning electron microscopy (SEM, Fig. 1e and Supplementary Fig. 15c) revealed the flexible, laminar morphology of RuO₂-C-1, which originates from the CDs-derived assembled carbon framework formed during freeze-drying/annealing. Atomic force microscopy (AFM, Fig. 1f, Supplementary Figs. 15d, e and 18) further confirms the formation of two-dimensional RuO₂-C-1 nanosheets with an average thickness of 1.52 ± 0.04 nm. These thin sheets expose more inner-layer Ru atoms, affording a higher density of catalytically active sites.

The TEM and HRTEM images revealed that rutile RuO₂ nanocrystallites are uniformly distributed throughout the CDs-derived carbon matrix and are interconnected without excessive aggregation. Pronounced lattice strain and dense grain boundaries are ubiquitous at interparticle junctions (Fig. 1g–i, Supplementary Figs. 15f and 19). To better visualize the existence of the RuO₂-carbon interface, we further acquired HRTEM focusing on the interfacial region (Supplementary Fig. 15g), where a strained RuO₂ nanocrystal is in direct contact with the surrounding carbon matrix. As shown in Fig. 1i and Supplementary Fig. 19, the lattice spacing is measured to be 0.3257 ± 0.0017 nm and indexed to the RuO₂ (110) plane, which is expanded relative to the C-free RuO₂ reference (0.318 nm) (Supplementary Fig. 14), supporting tensile strain in RuO₂-C-1.

Aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images further confirmed the pronounced lattice strain in RuO₂-C-1 (Fig. 1j, k and Supplementary Fig. 20). Since the atomic number of O and C is much smaller than that of Ru, the oxygen and carbon atoms are too weak to be seen⁴⁴; thus, these aberration-corrected HAADF-STEM images are primarily used to resolve the RuO₂ lattice and evaluate its strain, rather than to directly visualize the carbon atomic structure. Consistently, the exposed RuO₂ (110) plane exhibits an enlarged interplanar spacing of 0.3259 ± 0.0019 nm compared with the reference value of 0.318 nm, corresponding to ~2.5% tensile strain and local distortion (Fig. 1j, k and Supplementary Fig. 20). The Ru content measured by TEM-energy dispersive X-ray spectroscopy (TEM-EDX) is consistent with the ICP results (Supplementary Fig. 15h), and the STEM-EDX elemental mapping analysis confirms a uniform distribution of Ru, O and C signals across the entire structure (Fig. 1l and Supplementary Fig. 15i). On the basis of these observations, a model crystal structure for RuO₂-C-1 is proposed, which primarily refers to the rutile lattice with an ~2.5% tensile strain (Fig. 1m). These observations indicate that the tensile strain observed in RuO₂-C-1 is likely a result of interfacial interactions between the carbon support and RuO₂, as well as from the grain boundaries. Recent studies have shown that these types of stress‒strain can increase the catalytic activity of metals^11,21,29.

Probing the electronic structure and local structure of the catalysts

To elucidate the effects of interfacial coupling and tensile strain on the surface electronic structure and chemical state, X-ray photoelectron spectroscopy (XPS) measurements were carried out for RuO₂-C-1 and all the reference catalysts, and the Ru 3d, C 1s and O 1s spectra were obtained. For pure RuO₂, the high-resolution Ru 3d XPS spectrum is deconvoluted into the Ru(Ⅳ) doublets at 281.11 eV (3d_5/2) and 285.28 eV (3d_3/2), accompanied by shake-up satellites¹² (filled in cyan) (Fig. 2a and Supplementary Table 2). Upon the introduction of CDs, the binding energies of Ru 3d_5/2 (280.88) and 3d_3/2 (285.05) for RuO₂-C-1 were negatively shifted (~0.23 eV) compared with that of RuO₂ and an additional −0.10/−0.15/−0.20 eV shifted versus RuO₂-C-2/-3/-4, respectively (Supplementary Figs. 21–23). This negative shift possibly suggests that Ru in the RuO₂-C-1 catalyst presented a lower average Ru valence state and higher electron density than did those in RuO₂ and indicates the strong electronic interaction between Ru and the oxygen-functionalized carbon matrix in RuO₂-C-1. Moreover, as depicted in Supplementary Fig. 24, RuO₂-C-1 exhibits negatively shifted Ru 3p_3/2 and Ru 3p_1/2 peaks compared with RuO₂, consistent with the Ru 3d analysis.

Figure 2a shows the C 1s spectrum of the RuO₂-C-1, which can be deconvoluted into three components corresponding to the C–C/C = C, C–O, and C = O bonds located at 284.5, 285.9, and 288.4 eV, respectively^42,43,45. The O 1s spectra were deconvoluted into lattice oxygen, vacancy-associated oxygen, and a high-binding-energy component attributed to adsorbed oxygen species and, for RuO₂-C-1, additional contributions from oxygenated carbon groups (C–O/C = O), consistent with the C 1s results^12,21(Fig. 2b). These results confirm that oxygen-containing functional groups on the CDs serve as anchoring and coordination sites for Ru, establishing a strongly coupled interface and introducing tensile strain that underpins the structural stability and electronic modulation of the composite.

To further substantiate the interfacial interaction between RuO₂ and the carbon matrix, we conducted near-edge X-ray absorption fine structure (NEXAFS) spectroscopy analysis. As shown in Fig. 2c, the C K edge spectrum of RuO₂-C-1 presents three characteristic peaks at ~285.3, 288.4, and 292.2 eV, which are assigned to π* (C = C), π* (C = O), and σ* (C–C), respectively^46,47,48. Compared with that of the carbon (C), the π* (C = C) characteristic peak of RuO₂-C-1 displays a reduced intensity accompanied by a slight negative shift, and both the σ* (C–C) peaks also shift to low energies. Notably, the π*(C = O) resonance displays an enhanced intensity, indicating increased polarization and electronic rehybridization of carbonyl groups upon coordination with Ru species. The concurrent attenuation of π*(C = C), enhancement of π*(C = O), and low-energy shifts collectively evidence charge redistribution and bond hybridization at the carbon-RuO₂ interface, providing direct proof of strong interfacial electronic coupling and local electronic restructuring^46,47. The O K-edge XANES spectrum of RuO₂-C-1 exhibits two prominent characteristic peaks at ~531.8 and ~534.6 eV, which are assigned to transitions into O 2p states hybridized with unoccupied Ru 4d t_2g and Ru 4d e_g manifolds, respectively (Fig. 2d)³⁵. Together, these spectroscopic signatures support an oxygen-mediated interfacial coupling (Ru-O-C) through oxygen-containing functionalities, rather than direct Ru-C bond formation^49,50,51.

To elucidate the local coordination and bonding configuration of Ru in RuO₂-C-1, we performed Ru K-edge X-ray absorption spectroscopy (XAS), including X-ray absorption near-edge structure (XANES) and Fourier-transformed extended X-ray absorption fine structure (FT-EXAFS), with Ru foil and RuO₂ as references (Fig. 2e, f). As shown in Fig. 2e (inset), the absorption edge of RuO₂-C-1 is located between that of metallic Ru and RuO₂, indicating an average Ru valence intermediate between Ru (0) and Ru (IV), which is in agreement with the XPS analysis. The FT-EXAFS spectra in Fig. 2f and Supplementary Fig. 25 reveal that the Ru–O bond length in the first coordination shell of RuO₂-C-1 is slightly longer than that of standard RuO₂. The peak at 3.19 Å in RuO₂-C-1 corresponds to the backscattering of Ru-Ru in the second coordination shell, resulting in a lower vibrational amplitude than that of RuO₂, which is attributed to the smaller size of the nanoparticles^18,25. Moreover, the Ru-O peak intensity of RuO₂-C-1 is slightly weaker than that of standard RuO₂, likely due to the introduction of oxygen vacancies at the interface caused by tensile strain, which reduces the coordination number of Ru²¹. The EXAFS fitting results (Supplementary Table 3) confirm that the coordination number for Ru-O in RuO₂-C-1 is 5.8, which is lower than that in RuO₂, providing further evidence of the structural modifications.

To strengthen the shell assignment and reduce ambiguity from overlapping contributions in conventional FT-EXAFS, we further performed wavelet-transform EXAFS (WT-EXAFS) analysis (Fig. 2g–i), which maps the EXAFS intensity simultaneously in R and k space and enables separation of light-atom (Ru-O) and heavy-atom (Ru-Ru) scattering contributions⁵². In the WT-EXAFS, Ru foil is dominated by a high-k Ru–Ru feature, whereas rutile RuO₂ exhibits a distinct Ru-O contribution at lower k together with a higher-k Ru-Ru component. RuO₂-C-1 preserves these RuO₂-like Ru-O/Ru-Ru signatures but shows a more diffuse/attenuated Ru–Ru intensity and a slightly broadened Ru-O feature, consistent with weakened higher-shell order and increased local disorder/strain in the composite⁵³. These WT-EXAFS results corroborate the FT-EXAFS fitting, supporting an oxide-like Ru coordination environment in RuO₂-C−1 while highlighting strain/interface-induced local distortion relative to standard RuO₂. XAS provides compelling evidence that interfacial coupling between the carbon support and RuO₂, together with grain-boundary effects, induces tensile strain in RuO₂-C-1. This strain generates an unsaturated coordination environment and effectively tunes the local electronic state of Ru^11,21,25.

Electrochemical OER performance

To demonstrate the catalytic advantages arising from interfacial coupling and tensile strain, we assessed the OER performance of RuO₂-C-1 in a conventional three-electrode setup using O₂-saturated 0.5 M H₂SO₄. and compared with those of the homemade RuO₂ (RuO₂; Supplementary Fig. 14), commercial RuO₂ (C-RuO₂; Supplementary Fig. 26), and commercial IrO₂ (C-IrO₂; Supplementary Fig. 27). The saturated calomel electrode (SCE) and carbon rod were served as the reference and counter electrodes, respectively. The SCE was first calibrated against the reversible hydrogen electrode (RHE) before the test (Supplementary Fig. 28). The RuO₂-C-x series was screened under identical conditions to identify the optimal Ru/C ratio (Supplementary Fig. 29), establishing RuO₂-C-1 as the best-performing composition. As shown by the linear sweep voltammetry (LSV) curves presented in Fig. 3a and Supplementary Fig. 30, RuO₂-C-1 delivers the most active polarization response, requiring a low overpotential of 140 mV to reach 10 mA cm⁻², substantially lower than that of the as-prepared RuO₂ (η₁₀ = 215 mV), C-RuO₂ (η₁₀ = 270 mV), and C-IrO₂ (η₁₀ = 300 mV) (Fig. 3c). Moreover, the improved activity of RuO₂-C-1 is highly reproducible (η₁₀ = 140 ± 5 mV) (Supplementary Fig. 31).

Next, the Tafel slope analysis was used to assess the OER kinetics. As shown in Fig. 3b, c, RuO₂-C-1 has the lowest slope of 36.4 mV dec⁻¹, which is distinctly smaller than those of RuO₂ (80.8 mV dec⁻¹), C-RuO₂ (100.4 mV dec⁻¹) and C-IrO₂ (126.6 mV dec⁻¹), indicating markedly faster interfacial charge transfer and more favorable OER kinetics for RuO₂-C-1. We also summarize the different current densities of the catalysts at 1.45 V vs. RHE. Compared with the control samples, RuO₂-C-1 exhibited the highest current density of 176.1 mA cm⁻² (Fig. 3c). RuO₂-C-1 was further evaluated in 0.05 M H₂SO₄ using a H₂ reference electrode. As shown in (Supplementary Fig. 32), the catalyst delivered a highly reproducible overpotential of 148 ± 1 mV at 10 mA cm⁻² (three independent measurements), confirming that the low activity is intrinsic and not an artifact of electrolyte concentration or reference-electrode conversion. Moreover, RuO₂-C-1 delivers an overpotential of 140 mV at 10 mA cm⁻² with a Ru loading of only 25.8 μg_Ru cm⁻², showing competitive acidic OER performance among reported Ru/Ir-based catalysts (Supplementary Table 4).

To better elucidate the origin of the high OER performance of RuO₂-C-1, we first calculated the mass activity of various catalysts on the basis of the Ru metal loading for. In particular, at 1.45 V vs. RHE, the mass activity of the RuO₂-C-1 reaches 6814.1 A g_Ru⁻¹, which is 124.1, 524.2, and 920.8 times greater than that of RuO₂ (54.9 A g_Ru⁻¹), C-RuO₂ (13.0 A g_Ru⁻¹), and C-IrO₂ (7.4 A g_Ru⁻¹), respectively (Fig. 3d). We then quantified the electrochemically active surface area (ECSA) from the double-layer capacitance (C_dl). RuO₂-C-1 has the largest C_dl and ECSA among all the samples (vs. RuO₂, C-RuO₂, and C-IrO₂; Supplementary Fig. 33 and Supplementary Table 5), indicating a greater density of exposed active sites that underpins its superior OER activity. As depicted in Fig. 3e, considering the ECSA-normalized OER activity, RuO₂-C-1 delivers the highest specific activity, followed by RuO₂, C-RuO₂, and C-IrO₂. The ECSA-based TOF consistently mirrors this order (Fig. 3e), suggesting intrinsically improved OER activity on the RuO₂-C-1 catalyst because of the incorporation of carbon. In addition, we further measured the Faradaic efficiency (FE) of the acidic OER using the rotating ring disc electrode (RRDE). As shown in Supplementary Fig. 34, the RuO₂-C-1 reaches an FE of ~99.8%, verifying its efficient OER performance¹⁴.

As depicted in Supplementary Fig. 29, the overpotential first decreases and then increases as the carbon content is decreased, whereas the carbon matrix alone shows negligible OER activity. In addition, the pure CDs precursor at the same catalyst loading likewise exhibits no discernible OER activity in acidic electrolyte (Supplementary Fig. 35). To benchmark against a conventional carbon support, we also prepared a reference catalyst using Kochen Black under a similarly preparation protocol and Ru loading (RuO₂-KB). RuO₂-KB catalyst shows lower OER activity and delivers consistently lower ECSA-normalized specific activity and TOF than RuO₂-C-1 (Supplementary Figs. 36–38), supporting that the performance enhancement is closely linked to the CDs-enabled interfacial coupling and tensile-strain effects. This trend indicates that RuO₂ centers are the true active sites and that strong interfacial coupling with the carbon matrix synergistically further enhances the intrinsic activity of the Ru sites, enabling better catalytic performance even at low Ru metal loadings.

To assess acidic durability, we performed systematic chronopotentiometry (CP) tests (Fig. 3f, g). Under identical conditions, C-RuO₂ and C-IrO₂ exhibited rapid performance loss, particularly at higher current densities. C-RuO₂ showed the poorest tolerance, sustaining 10 mA cm⁻² for only 23 h and displaying a sharp voltage rise within 2 h at 100 mA cm⁻². C-IrO₂ maintained a comparatively stable CP trace for 46 h at 10 mA cm⁻², but similarly degraded after ~4 h at 100 mA cm⁻² despite IrO₂ being prone to acid resistance. The homemade RuO₂ sustained a stable CP trace for 718 h at 10 mA cm⁻², and for ~185 h at 100 mA cm⁻². In contrast, RuO₂-C-1 exhibited markedly enhanced durability, sustaining stable operation for over 1900 h at 10 mA cm⁻² and ~1000 h at 100 mA cm⁻². Figure 3h summarizes the key OER metrics for all the catalysts, highlighting the superior intrinsic activity, low Ru loading, and good acid durability of RuO₂-C-1. We also compared the stability of RuO₂-C-1 with that of state‒of‒the‒art OER catalysts and confirmed the competitive performance of RuO₂-C-1 (Supplementary Table 4).

We further carried out post-stability characterization to examine the morphology and structural evolution of RuO₂-C-1 after long-term operation at 10 mA cm⁻². As shown in Supplementary Fig. 39a, the Bragg diffraction peaks of rutile RuO₂ become more pronounced, which can be attributed to partial loss of the carbon support during durability testing. Nevertheless, SEM (Supplementary Fig. 39b) and TEM (Supplementary Fig. 39c) images reveal that the characteristic two-dimensional nanosheet morphology of RuO₂-C-1 is largely preserved after prolonged operation under harsh acidic conditions. The HRTEM image (Supplementary Fig. 39d) further confirmed that the crystalline structure of RuO₂-C-1 underwent a negligible change, and statistical analysis of the RuO₂ (110) of RuO₂-C-1 lattice space shows that the interplanar spacing remains essentially unchanged after stability testing (Supplementary Fig. 40). Furthermore, STEM-EDX elemental mapping (Supplementary Fig. 39e) confirmed that Ru, O, and C remained uniformly distributed across the nanosheets after durability testing. TEM-EDX analysis (Supplementary Fig. 39f) revealed a slight decrease in the carbon content of RuO₂-C-1, indicating the partial dissolution of the carbon scaffold into the acidic electrolyte during long-term operation. To further evaluate possible changes in the CDs-derived carbon during acidic OER. We performed post-stability C K-edge NEXAFS. The post-stability C K-edge NEXAFS shows that RuO₂-C-1 retains the main carbon resonances (π*(C = C), π*(C = O), and σ*(C–C)) with only minor lineshape changes, indicating that the CDs-derived carbon framework is largely preserved after acidic OER (Supplementary Fig. 41). While XPS analysis further verified that the oxidation states and surface electronic configuration remained unchanged (Supplementary Fig. 42).

Furthermore, to quantitatively assess possible carbon corrosion of RuO₂-C-1 under acidic OER conditions, we conducted in-situ differential electrochemical mass spectrometry (DEMS) and total organic carbon (TOC) analyses experiments. In-situ DEMS was performed to assess possible gaseous carbon corrosion of RuO₂-C-1 at 1.55 V vs. RHE in 0.5 M H₂SO₄. As shown in Supplementary Fig. 43), a pronounced O₂ signal (m/z = 32) was detected while CO₂ (m/z = 44) remained negligible, indicating minimal gaseous carbon-corrosion products under high anodic potential. Separately, TOC analysis was conducted on electrolytes collected from chronoamperometric stability tests at 1.55 V vs. RHE, revealing only a small amount of dissolved carbon corresponding to a cumulative carbon loss of ~1.04% (5 h) and ~1.61–1.65% (10–50 h) relative to the total carbon in the catalyst, with the dissolved-carbon fraction approaching a plateau after ~10 h (Supplementary Fig. 44). Together, these results suggest that carbon loss is limited and mainly occurs at the early stage, after which the carbon scaffold remains sufficiently stable to support durable operation. Although carbon is thermodynamically unstable under acidic OER potentials^22,23,25, the improved durability of RuO₂-C-1 is not attributed to “intrinsically oxidation-proof” CDs. Instead, it is proposed to arise from the CDs-derived assembled carbon framework and the strongly coupled RuO₂-carbon interface: the condensed carbon framework can reduce highly exposed reactive sites, while interface coupling redistributes interfacial charge to suppress Ru overoxidation/dissolution and mitigate locally aggressive oxidative conditions^32,54,55.

PEMWE performance

The ultimate benchmark for an acidic OER catalyst is its performance in a practical proton exchange membrane water electrolyzer (PEMWE). Motivated by the excellent three-electrode OER activity of RuO₂-C-1, we assembled a PEMWE device for operation with Milli-Q water at 80 °C (Fig. 3i and Supplementary Fig. 45). The cell employed a Nafion 212 membrane, with RuO₂-C-1 as the anode catalyst and commercial Pt/C as the cathode catalyst, giving a total platinum group metal (PGM) loading as low as 0.35 mg cm⁻² (Supplementary Table 6). Specifically, the current‒voltage (I‒V) polarization curve shows that the RuO₂-C-1-based PEMWE delivers current densities of 1 and 3 A cm⁻² at cell voltages of 1.543 and 1.833 V, respectively, showing competitive performance among recently reported acidic oxygen evolution electrocatalysts (Fig. 3j and Supplementary Table 6). More impressively, the RuO₂-C-1-based PEMWE operates stably at 1.43 V (about 0.35 A cm⁻²) for over 800 h (Fig. 3k), the postoperative morphology, structural, and Ru valance analyses further confirmed that the morphology, crystal structure, composition and surface electronic properties of RuO₂-C-1 are well preserved (Supplementary Figs. 46 and 47).

Moreover, RuO₂-C-1-based PEMWE also sustains stable operation at 1.55 V (about 1.0 A cm⁻²) for over 690 h (Supplementary Fig. 48), underscoring its robust durability under practically relevant conditions and highlighting its promise for deployment. The corresponding Ru dissolution during operation was quantified by ICP-AES (Supplementary Fig. 49). The cumulative Ru loss increased mainly during the initial stage (within the first 50 h) and then entered a slow-growth regime during prolonged operation. Consistently, the Ru dissolution rate decreased rapidly after the initial stage and remained at a low but finite level. Notably, as shown in Fig. 3j, the PEMWE performance of RuO₂-C-1 reaches the US DOE 2025 target (3.0 A cm⁻² at 1.9 V with a total PGM loading of 0.5 mg cm⁻²)⁵⁶ and closely approaches the more stringent US DOE 2026 target (3.0 A cm⁻² at 1.8 V; PGM loading of 0.5 mg cm⁻²)⁶. The RuO2-C-1-based PEMWE device exhibits exceptional structural robustness and catalytic activity, effectively overcoming the inherent durability and activity limitations of pure RuO₂ in PEMWE. This achievement underscores the promising prospects of RuO₂-C-1 for industrial-scale, sustainable hydrogen generation.

Mechanistic and theoretical understanding of the activity and stability of RuO₂-C-1

The catalyst interfacial charge distribution plays a pivotal role in governing OER performance^4,34. To further probe the charge transfer behavior at the carbon/RuO₂ interface in RuO₂-C-1, we employed the transient photoinduced voltage (TPV) measurements (Supplementary Fig. 50). The time-resolved TPV profiles upon laser excitation are shown in Supplementary Fig. 51. The raw TPV signals for RuO₂-C-1 and RuO₂ are presented in Supplementary Fig. 51a, c, respectively. Exponential fitting of the TPV decay curves (Supplementary Fig. 51b, d) reveals distinct charge-relaxation behaviors. RuO₂-C-1 exhibited a two-step decay process. The first decay component (τ₁ = 2.58 × 10^-4 s) is comparable to that of pristine RuO₂ (τ = 2.03 × 10⁻⁴ s), indicating that the fast relaxation pathway is dominated by the intrinsic charge-transport characteristics of RuO₂. In contrast, the second decay component of RuO₂-C-1 (τ₂ = 1.10 × 10⁻² s) is significantly slower, suggesting a redistribution of interfacial charges and delayed recombination. The prolonged τ₂ demonstrates that strong interactions between carbon and RuO₂ facilitate charge storage and retard charge decay at the interface⁴². Although TPV probes photo-induced charge dynamics, the prolonged decay component observed for RuO₂-C-1 reflects a slower interfacial charge relaxation process, which is interpreted as enhanced interfacial charge buffering/delocalization enabled by the strongly coupled RuO₂-carbon interface. Under anodic polarization during OER, such interfacial charge buffering can facilitate more stable electron withdrawal from Ru centers and suppress uncontrolled charge accumulation that promotes Ru overoxidation/dissolution, thereby stabilizing the RuO₂ lattice and contributing to the markedly enhanced OER durability of RuO₂-C-1⁴.

In-situ Ru K-edge XAS was employed to monitor the evolution of the oxidation state of Ru under operating OER conditions, aiming to further elucidate the mechanistic origin of the enhanced stability of RuO₂-C-1. The Ru K-edge XANES spectra exhibit a gradual positive shift in the absorption edge energy as the applied potential increases from the open-circuit potential (OCP) to 1.40 V vs. RHE, indicating a progressive increase in the Ru oxidation state (Fig. 4a). This evolution reflects potential-driven electron withdrawal from Ru sites toward adjacent atoms or adsorbed oxygen-containing intermediates⁵⁷. When the potential is further increased to 1.50 V vs. RHE—which is widely regarded as the upper stability limit for RuO₂-based catalysts in acidic media⁵⁸—the rate of the absorption edge shift with potential (d_E0/d_V) decreases markedly (inset of Fig. 4a and Supplementary Fig. 52). This behavior suggests that the Ru oxidation state tends to level off (within experimental uncertainty) at high anodic potentials³⁵, which is consistent with the TPV results. The in-situ Ru K-edge EXAFS fitting was performed to further examine the potential-dependent Ru coordination environment (Supplementary Fig. 53 and Supplementary Table 7). The Ru-O coordination number remains nearly unchanged from OCP to 1.5 V vs. RHE, indicating that the primary Ru-O coordination structure is largely preserved under acidic OER conditions. This stabilization of Ru species under high anodic potentials is expected to inhibit overoxidation and suppress Ru dissolution, thereby contributing to the excellent catalytic durability^57,59.

We further carried out in-situ Raman spectroscopy to probe the structural stability of the RuO₂-C-1 and RuO₂ catalysts during continuous potential changes from the open-circuit potential to 1.5 V vs. RHE (Fig. 4b, c and Supplementary Fig. 54). The in-situ Raman spectra reveal the nanocrystalline characteristics of both RuO₂-C-1 and RuO₂, showing two prominent vibrational peaks at approximately 436 and 596 cm^-1, which are assigned to Ru⁴⁺–O and Ru³⁺–O bonding, respectively¹⁷; the relatively broadened features are consistent with a nanocrystalline/strain-disordered RuO₂ lattice^19,60. The Raman-derived atomic ratio of Ru³⁺/Ru⁴⁺ is highly correlated with the structural integrity of Ru-based catalysts, since excessive formation or leaching of high-valence Ru species often triggers lattice destabilization and performance decay. As shown in Supplementary Fig. 55, RuO₂-C-1 has a markedly higher (I_Ru3+/I_Ru4+) intensity ratio than RuO₂ does, indicating a Ru³⁺-rich surface environment. The presence of additional Ru³⁺ species weakens the Ru–O lattice bond strength and helps prevent overoxidation of lattice oxygen, thereby improving structural resilience in acidic OER operation⁶¹.

Moreover, as the potential increases from the OCP to 1.5 V vs RHE, the Raman signal of RuO₂ gradually weakens and eventually disappears, a phenomenon that has also been reported previously^20,21. Although the attenuation may be partially influenced by optical scattering from O₂ bubble evolution, the Raman intensity of RuO₂-C-1 remains almost unchanged. RuO₂-C-1 also maintains an almost constant Raman peak position, providing strong evidence that its crystal structure remains highly stable under anodic potentials. Taken together, these findings indicate that the interfacial coupling between the carbon support and RuO₂, along with the tensile strain it induces, effectively stabilizes the catalytically active Ru species. This stabilization markedly enhances both the acidic OER durability and the structural robustness of the RuO₂-C-1 catalyst^11,21,25.

To elucidate the nature of the oxygen-containing intermediates and gain a more comprehensive mechanistic understanding, we employed in-situ attenuated total reflectance surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS). Owing to its high sensitivity to surface-bound species and ability to monitor potential-dependent intermediates, this technique enables direct probing of the binding characteristics of oxygenated species on the surfaces of the RuO₂-C-1 and RuO₂ catalysts⁶². The in-situ ATR-SEIRAS spectra of RuO₂-C-1 at various applied potentials (Fig. 4d and Supplementary Fig. 56) display a pronounced absorption band at 1213 cm⁻¹, corresponding to the O–O stretching vibration of surface-adsorbed *OOH species. The appearance and gradual growth of this band with increasing potential (1.4–1.8 V vs. RHE) indicate that the adsorbate evolution mechanism (AEM) governs O₂ formation on RuO₂-C-1¹⁹. In contrast, for pure RuO₂, the *OOH-related O–O stretching vibration emerges only at ~1.5 V and exhibits a blueshift to 1224 cm⁻¹, signifying a weakened interaction between the *OOH intermediate and the RuO₂ surface⁶³ (Fig. 4e). The lower-frequency feature on RuO₂-C-1 therefore reflects the stronger binding of *OOH relative to RuO₂, suggesting a more favorable adsorption configuration induced by the carbon-mediated tensile strain. Moreover, broad absorption features in the 3150–3850 cm⁻¹ range are assigned to O–H stretching of surface hydroxyl species, whose intensities increase monotonically with potential, reflecting the accumulation of proton-coupled oxygen intermediates during the OER⁶⁴.

To further confirm that RuO₂-C-1 follows an AEM pathway, we performed in-situ ¹⁸O isotope–assisted DEMS measurements (Fig. 4f and Supplementary Fig. 57). This technique enables direct discrimination between AEM and lattice oxygen-mediated (LOM) pathways by monitoring oxygen molecules with different isotopic compositions during the OER^{21,22,23,25,61}. In a typical experiment, the catalyst was first labeled with ¹⁸O by operating in 0.5 M H₂SO₄ prepared with H₂¹⁸O, followed by thorough rinsing with H₂¹⁶O to remove physisorbed ¹⁸O species. The labeled catalyst was subsequently transferred into 0.5 M H₂SO₄ of H₂¹⁶O electrolyte for the OER, during which the evolved oxygen was monitored in real time via in-situ DEMS²⁵.

Figure 4g displays the dominant mass signal of ³⁶O₂, accompanied by minor ³⁴O₂ and ³²O₂ signals, during three consecutive LSV scans of RuO₂-C-1 in H₂¹⁸O-containing H₂SO₄ electrolyte. The trace ³⁴O₂ signal arises from background ¹⁸O contamination and instrumental residuals rather than from the participation of lattice oxygen^21,25. The clear predominance of ³⁶O₂ demonstrates that the evolved O₂ originates exclusively from the electrolyte-derived oxygen atoms, thereby excluding the LOM. When the ¹⁸O-labeled catalyst was subsequently tested in H₂¹⁶O electrolyte, only ³²O₂ was detected, and no ³⁶O₂ signal appeared (Fig. 4h), confirming the absence of lattice oxygen involvement in the OER^22,23,25. Together with the in-situ ATR-SEIRAS results, these in-situ DEMS measurements unambiguously establish that O₂ formation on RuO₂-C-1 proceeds predominantly via the AEM pathway.

Density functional theory (DFT) calculations were carried out to elucidate the underlying mechanism of the improved OER activity and stability of RuO₂-C-1. We note that the DFT model is idealized and does not fully capture surface reconstruction, electrolyte effects, or potential-dependent intermediates. Therefore, the calculations are used mainly to provide qualitative mechanistic insight. A carbon-supported rutile RuO₂ (110) slab model with ~2.5% tensile strain was constructed (Fig. 4i and Supplementary Data 1), given the STEM and XAS characterization and the thermodynamic stability of the (110) surface^12,17. For comparison, a reference RuO₂ (110) slab without strain was also modeled (Fig. 4j). Firstly, we performed DFT-based XPS core-level binding energy shift simulations to directly probe the electronic structure modulation. The computed Ru 3d binding energy shift is 0.12 eV on the RuO₂-C-1 interface model relative to pristine RuO₂ (110) (Supplementary Fig. 58), in good agreement with the experimentally observed shift of 0.23 eV. This consistency confirms that coupling with the carbon substrate induces interfacial charge redistribution in RuO₂, consistent with the experimentally observed electronic modulation.

The DFT calculations were then carried out to investigate the full reaction pathway with Gibbs free energy profiles of the key intermediates (*OH, *O and *OOH) at the Ru active sites of RuO₂-C-1 (Fig. 4k) and unstrained RuO₂ (Supplementary Fig. 59). As shown in Fig. 4l, m, the formation of the *OOH is the rate-determining step (RDS) of both. Notably, the free-energy barrier for this step on RuO₂ is 1.98 eV at U = 0 V, corresponding to an overpotential of 0.75 V (at U = 1.23 V) (Fig. 4l and Supplementary Fig. 59). Instead, RuO₂-C-1 has a lower the free-energy barrier of 1.91 eV, corresponding to an overpotential of 0.68 V, indicating the beneficial effects of tensile strain and interfacial coupling in facilitating *OOH formation (Fig. 4m). Additional DFT calculations on RuO₂ (110) show that tensile strain reduces the rate-determining *OOH formation barrier from 0.753 eV at 0% to a minimum of 0.673–0.677 eV at 2–3%, before slightly increasing to 0.681 eV at 4% (Supplementary Fig. 60), indicating an optimal strain window of ~2–3%, consistent with the experimentally observed ~2.5% lattice expansion. The further transition state calculations (Supplementary Figs. 61–63) demonstrate that the application of a 2.5% tensile strain lowers the OER kinetic barrier by reducing the free-energy penalty for the rate-determining *OOH formation step, thereby further accelerating the reaction kinetics.

Charge‒density difference analysis was further conducted to probe the structural stability of RuO₂-C-1. The carbon framework assembled from CDs provides effective anchoring sites for RuO₂, and the computed charge‒density redistribution at the carbon‒RuO₂ interface reveals substantial charge transfer between the two components, while pure RuO₂ shows only intrinsic charge polarization within the oxide framework (Supplementary Fig. 64). This pronounced electronic reorganization confirms strong interfacial interactions, which contribute to stabilizing the RuO₂ structure under OER conditions. These theoretical observations are fully consistent with the experimental results, further validating the role of carbon-induced interfacial coupling in enhancing catalyst stability. Collectively, these results indicate that tensile strain primarily lowers the OER kinetic barrier (notably for the rate-determining OOH formation step), while interfacial coupling mainly regulates interfacial charge distribution to suppress Ru overoxidation/dissolution, and together they synergistically enhance both the acidic OER activity and durability of RuO₂-C-1.

In summary, we construct a two-dimensional carbon–RuO₂ composite (RuO₂-C-1) featuring strong interfacial coupling and ~2.5% tensile strain. These structural attributes stabilize catalytically active Ru centers, enhance intrinsic OER activity with substantially reduced Ru usage. RuO₂-C-1 delivers a low overpotential of 140 mV at 10 mA cm⁻² and operates for >1900 h in acid. In a PEMWE, it reaches 3 A cm⁻² at 1.833 V and sustains >800 h at 0.35 A cm⁻², approaching U.S. DOE 2026 target. In-situ spectroscopies, together with TPV and DFT, show that interfacial charge redistribution and strain stabilization suppress Ru overoxidation and lower the OER barrier. These insights establish interfacial strain engineering as a powerful approach for unlocking the intrinsic stability of RuO₂ and advancing practical, scalable, low-Ru PEMWE.