Introduction

Hydrogen is a clean energy carrier with a high gravimetric energy density. Utilizing the electricity generated via renewable resources, such as wind and solar energy, for converting and storing hydrogen energy is an effective approach to address the fossil fuel crisis1,2,3. Currently, proton exchange membrane water electrolysis (PEMWE) technology is widely acknowledged as a promising and sustainable route for green hydrogen fuel production owing to its numerous merits, including high purity, high current density, low ohm resistance and perfect compatibility with renewable energy4,5. However, the anodic oxygen evolution reaction (OER) is a challenge to PEMWE, due to its intrinsic sluggishness in the four-electron transfer process6. Thus, developing efficient OER electrocatalysts is crucial for improving the efficiency of PEMWE devices and further advancing their adoption and applicability.

Over the past decades, Ru-based electrocatalysts have demonstrated tremendous potential toward OER in acidic environments owing to the presence of Ru sites with diverse coordination environment on their surfaces, which remarkably influence the adsorption/desorption energy of the intermediates7,8. Further, the intrinsic activity of Ru sites in commercial RuO2 requires an overpotential of approximately 330 mV to achieve 10 mA cm−2 current density in acidic environments. However, it suffers from severe dissolution and corrosion during OER in acidic media. Therefore, achieving simultaneous enhancement of the activity and stability of Ru-based electrocatalysts for their applicability in PEMWE is an arduous task9,10.

There are two primary OER mechanisms in acidic solutions11. The first one is the traditional adsorbate evolution mechanism (AEM). AEM exhibits an inherent linear relationship between the adsorption energy and the adsorption of *OH and *OOH active intermediates species, revealing a minimum theoretical overpotential of approximately 0.37 V, thereby constraining the advancement of efficient electrocatalysts12,13,14. Additionally, the oxidation of Ru metal atoms to high-valence soluble RuO2(OH)2 or RuO4 species under strong acidic and oxidizing environment is detrimental to their long-time stability15,16,17,18. The second mechanism is the kinetically favorable lattice oxygen mechanism (LOM). This mechanism avoids the generation of *OOH intermediates, but it uses lattice oxygen to release oxygen products and forms lattice oxygen vacancies, which disrupt the thermally stable crystal structure19,20. Based on the reaction process analysis, the theoretical limitations of AEM and structurally unstable LOM result in inferior activity and stability, thereby restricting the practical application of Ru-based electrocatalyst in acidic media.

Recently, oxide path mechanism (OPM) has been discovered for OER that enables the direct coupling of adsorbed oxygen species (*O-O*) to release O2, which effectively overcomes the scaling relationship limitation and maintains a complete thermally stable structure (Fig. 1)21,22. Previous reports claimed that the OPM in acidic is based only on the as-formed intermediates *O-O* and the theoretical simulation of the dual-active sites22,23,24. However, the exact underlying structural character of OPM, exhibiting good activity and stability, is still not entirely understood25. Uncovering the exact structural character of OPM and elucidating the exact function of each atom in the microstructure is highly desirable for artificially designing high-performance OPM catalysts in acidic solutions.

Fig. 1
figure 1

A survey of the catalytic reaction pathway and its charactersitics of RuO2, including AEM, LOM and OPM.

Full size image

In this work, we partially substitute the Ru atoms in RuO2 host with Mn atoms at the optimal composition to form a unique Mn4-δ-O-Ru4+δ microstructure over the Mn0.2RuO2. We observe a strong signal assigned to *O-O* species, a typical intermediate for OPM on Mn0.2RuO2 during the OER in an acidic electrolyte using operando synchrotron radiation Fourier transform infrared spectroscopy (SR-FTIR). In contrast, the *OOH intermediate, a typical intermediate for AEM, is observed on RuO2 with a Ru-O-Ru active center. The operando differential electrochemical mass spectrometry (DEMS) reveals the OER pathway switches after incorporating Mn atoms. Furthermore, we determine both Mn and Ru as dual-active centers for the OPM pathway by operando X-ray absorption fine spectroscopy (XAFS) and differential delta-XAFS (ΔXAFS) spectra analysis. Density functional theory (DFT) calculations reveal that the presence of Mn atoms in the Mn0.2RuO2 induce a charge redistribution in the Mn4-δ-O-Ru4+δ microstructure, which facilitates the *OH adsorption and the charge transfer in Mn0.2RuO2. This results in a lower energy barrier for the direct connecting between two O* on the neighboring Mn/Ru sites, favoring the formation of a bridge *O-O* intermediate on the Mn-O-Ru microstructure, rather than absorbing one *OH into the neighboring *O to form a *OOH intermediate. This study provides a clear understanding of the OPM pathway based on the microstructural and electronic structure changes in the material, thus giving additional insights into the manual design of OPM catalysts from a structure viewpoint.

Results

Structural characterization to identify the Mn4-δ-O-Ru4+δ pair

The Mn0.2RuO2 electrocatalyst was prepared by incorporating Mn atoms into the RuO2 host via a hydrothermal reaction and subsequent annealing process (Fig. 2a and Supplementary Fig. 1). The bulk crystal structure and morphology of the obtained Mn0.2RuO2 were characterized using X-ray diffraction (XRD), scanning electron microscopy (SEM), and transmission electron microscopy (TEM). The XRD pattern of Mn-doped RuO2 exhibits characteristic diffraction peaks identical to those of rutile p-RuO2 (PDF#40-1290), indicating the absence of manganese oxide impurities and confirming the formation of a Mn-doped RuO2 octahedral rutile phase (Fig. 2b). SEM images in Fig. 2c and Supplementary Fig. 2 reveal an array of nanorods for Mn0.2RuO2. The TEM image displays the uniformity of the Mn0.2RuO2 nanorods with a length of approximately 55 nm (Supplementary Fig. 2). The energy dispersive spectroscopy (EDS) elemental mapping images display an even distribution of Mn, Ru, and O elements in Mn0.2RuO2. From the EDS line-scanning and EDS spectra, the Mn/Ru atom ratio within the nanorods can be calculated to be approximately 1:4.6, which is close to our input ratio 1:5 (Fig. 2e and Supplementary Fig. 3). The inductively coupled plasma-mass spectrometry (ICP-MS) measurement further confirm the atom ratio of Mn: Ru is around 1:5 (Supplementary Table 1).

Fig. 2: Fine-structure characterization of Mn0.2RuO2.
figure 2

a Schematic diagram of the structure for Mn0.2RuO2. b The XRD patterns of Mn0.2RuO2 and p-RuO2. c SEM image of Mn0.2RuO2. d Atomic-scale HAADF-STEM image of Mn0.2RuO2. e EDX linear-scanning elemental distribution of Mn, Ru and O for the Mn0.2RuO2. Inset: line-scanning TEM image. f Atomic-scale HAADF-STEM image and the corresponding EDS elemental mapping images for Mn0.2RuO2. g EELS spectra of Ru M-edge, O K-edge, and Mn L-edge for Mn0.2RuO2. Inset: HADDF-STEM image. h FT k3-weighted Mn and Ru K-edge EXAFS spectra. i WT analysis of EXAFS spectra for Mn0.2RuO2 and p-RuO2.

Full size image

High-resolution TEM (HRTEM) image exhibits a lattice pattern corresponding to the (110) plane of Mn0.2RuO2 (Supplementary Fig. 4). In comparison with p-RuO2, the introduction of Mn atom does not change the morphology of the nanorods in Mn0.2RuO2 (Supplementary Figs. 5–6). Atomic level resolved high-angle annular dark-field imaging-scanning transmission electron microscopy (HAADF-STEM) and micro-structure sensitive extended XAFS (EXAFS) were performed to identify the local environment of Mn atoms within the RuO2 lattice. The HAADF-STEM images show a regular arrangement of metal atoms (Mn/Ru) and no crystal lattice of individual manganese oxides in Mn0.2RuO2 (Fig. 2d and Supplementary Fig. 7). The lattice fringes could be attributed to the (110) plane of the octahedral structure, indicating that Mn atoms are successfully substituted into the RuO2 crystal lattice26. Additionally, three random regions EDS elemental mapping images confirm the atomic-scale uniform distribution of Mn and Ru atoms (Fig. 2f and Supplementary Fig. 8). Electron energy-loss spectroscopy (EELS) in the selected area clearly exhibits Ru M-edge, O K-edge and Mn L-edge spectra, suggesting that Mn atoms are located at the original Ru positions (Fig. 2g).

Subsequently, we conducted a detailed analysis of the Fourier transform (FT) curves of the Mn and Ru K-edge EXAFS k3χ(k) functions for Mn0.2RuO2. The Mn K-edge FT-EXAFS spectra of standard MnO, Mn2O3, MnO2 and Mn0.2RuO2, as well as the Ru K-edge FT-EXAFS spectra of p-RuO2 and Mn0.2RuO2 are given in Fig. 2h and Supplementary Fig. 9. The FT-EXAFS curves of the Mn K-edge EXAFS spectra for Mn0.2RuO2 show intense peaks located at approximately 1.5 and 3.2 Å, corresponding to the Mn-O and Mn-Ru scattering, respectively. These observations closely resemble the features of Ru K-edge FT-EXAFS spectra in Mn0.2RuO2 and p-RuO2; however, they differ from the characteristics of standard manganese oxide, indicating that the introduced Mn atoms occupy the same crystal lattice environment as that of the Ru atoms, thereby preserving the symmetric arrangement of the RuO2 octahedral structure. To determine the specific coordination numbers of various paths, the FT-EXAFS spectra were reasonably fitted (Supplementary Fig. 10 and Supplementary Table 2). The best-fitting results illustrate the similar local environment of Mn and Ru atoms in Mn0.2RuO2. Furthermore, the wavelet transform (WT) EXAFS analysis provided additional insights into the local bonding environment. The Mn K-edge WT plots of Mn0.2RuO2 reveal the maximum WT value at approximately 6.0 Å−1 ascribed to the bonding between Mn and light O atoms in first shell, and the maximum WT value at approximately 8.5 Å−1 associated with the scattering between Mn and heavy Ru atoms in high shell (Fig. 2i). Consequently, combining the Mn and Ru K-edge WT plots of Mn0.2RuO2 and p-RuO2 demonstrates that the Mn atoms precisely replace the lattice positions of the Ru atoms and the local coordinate environments of Mn-O-Ru and Ru-O-Ru are nearly identical.

To determine the oxidation states of Mn and Ru atoms in the presence of Mn atoms in Mn0.2RuO2, X-ray absorption near-edge structure (XANES) and X-ray photoelectron spectra (XPS) analyzes were carried out. In the Mn K-edge XANES spectra, the absorption edge of Mn0.2RuO2 shifts marginally towards a higher photon energy compared to those of MnO and Mn2O3; however, it is lower than that of MnO2, demonstrating that the oxidation state of Mn ranges between +3 and +4 (Fig. 3a). Furthermore, The Mn K-edge XANES spectra of Mn0.2RuO2 exhibit a shape that is markedly different from the reference samples. Meanwhile, Mn0.2RuO2 displays a similar geometry structure but optimized electron structure relative to that of p-RuO2 in the O K-edge spectra, consistent with the Ru K-edge EXAFS results (Supplementary Fig. 11). Figure 3b shows a shift in the absorption edge of Ru K-edge for Mn0.2RuO2 to a higher energy position compared to that of p-RuO2, suggesting that the introduction of Mn atoms leads to an electron redistribution around Ru. This leads to the oxidation state of Ru in Mn0.2RuO2 exceeds +4. Moreover, the O 1 s XPS spectra also confirm that the oxidation state of Ru is higher in Mn0.2RuO2 than in p-RuO2 (Supplementary Fig. 12). The Mn L-edge XANES spectra of Mn0.2RuO2 reveals that the absorption peak of Mn0.2RuO2 lies between those of Mn2O3 and MnO2, indicating that the oxidation state of Mn in Mn0.2RuO2 is below +4 (Fig. 3c). Similarly, the Ru M-edge spectra display a blue shift in accordance with the results of the Ru K-edge XANES analyzes (Fig. 3d). The XPS spectra of Mn 2p in Mn0.2RuO2 further indicates that the oxidation state is below +4, denoted as Mn4-δ (Fig. 3e). In the Ru 3d and C1s XPS spectra of RuO2 and Mn0.2RuO2, two sets of doublet peaks are observed between 280 and 290 eV, corresponding to the peaks of Ru4+ 3d5/2 and 3d3/2 + C 1s, respectively. The Ru 3d and C 1s XPS peak of Mn0.2RuO2 is slightly shifted to a higher binding energy compared to that of p-RuO2, further confirming the presence of higher valence Ru species in Mn0.2RuO2, denoted as Ru4+δ (Fig. 3f). Specifically, through linear fitting of Mn and Ru K-edge XANES spectra for Mn0.2RuO2 and the reference samples, the oxidation states of Mn and Ru is estimated to be +3.55 and +4.46, respectively, in Mn0.2RuO2 (Supplementary Fig. 13). These findings provide detailed evidence that electrons are transferred from Ru atoms to neighboring Mn atoms through bridging O atoms, forming unique localized Mn4-δ-O-Ru4+δ pairs.

Fig. 3: Electronic structure characterization of Mn0.2RuO2.
figure 3

a Mn K-edge XANES spectra of Mn foil, MnO, Mn2O3, MnO2 and Mn0.2RuO2. Inset: magnified XANES spectra. b Ru K-edge XANES spectra of Ru foil, p-RuO2 and Mn0.2RuO2. Inset: magnified XANES spectra. c Mn L-edge XANES spectra of MnO, Mn2O3, MnO2 and Mn0.2RuO2. The arrow points to the high energy direction. d Ru M-edge XANES spectra of p-RuO2 and Mn0.2RuO2. The arrow points to the high energy direction. e Mn 2p XPS spectra of Mn0.2RuO2. f Ru 3d and C 1s XPS spectra of p-RuO2 and Mn0.2RuO2. The arrow points to the high binding energy direction.

Full size image

OER performance of Mn0.2RuO2 electrocatalyst

The OER measurements were used to characterize the MnxRuO2 and p-RuO2 to determine the effect of the Mn4-δ-O-Ru4+δ microstructure on the electrocatalytic performance. We synthesized samples with different Mn doping levels, namely, Mn0.1RuO2, Mn0.2RuO2, and Mn0.3RuO2, and filtered them using linear sweep voltammetry (LSV) curves with iR-corrected. Among them, Mn0.2RuO2 exhibits an overpotential of 188 mV at a current density of 10 mA cm−2, outperforming Mn0.1RuO2, Mn0.3RuO2, p-RuO2 and commercial RuO2 (referred to as com-RuO2) (Fig. 4a and Supplementary Figs. 14–15). The Tafel slope plots derived from LSV curves display that the Tafel slope of p-RuO2 is 114 mV dec−1, which is smaller than that of com-RuO2, potentially attributed to differences in particle size. Among them, Mn0.2RuO2 shows the lowest Tafel slope, suggesting that the unique Mn4-δ-O-Ru4+δ structure could improve the OER kinetics (Fig. 4b). Moreover, the mass activity of Mn0.2RuO2 is 47.7 A g−1, which surpasses those of p-RuO2 (4.2 A g−1) and com-RuO2 (1.7 A g−1) at an overpotential of 188 mV, highlighting the optimal activity of Mn0.2RuO2 under acidic OER conditions (Fig. 4c).

Fig. 4: Activity and stability measurement.
figure 4

a The LSV curves of com-RuO2, p-RuO2, and Mn0.2RuO2 with 85% iR compensation in 0.5 M H2SO4 (pH = 0) with a scan rate of 10 mV s-1 at room temperature. b Tafel slopes and (c) mass activity of com-RuO2, p-RuO2 and Mn0.2RuO2. d Cdl plots derived from the fitting CV curves at difference scan rates. e EIS Nyquist plots of com-RuO2, p-RuO2 and Mn0.2RuO2 measured at the same potential. f Comprehensive comparison of the OER performance for Mn0.2RuO2 and our reference samples. g The constant current tests at 10 mA cm-2 for com-RuO2, p-RuO2 and Mn0.2RuO2. h Comparison of Mn0.2RuO2 with previously reported Ru/Ir-based electrocatalysts in terms of overpotential and stability at 10 mA cm−2 in 0.5 M H2SO4, at room temperature.

Full size image

Next, we performed cyclic voltammetry at different scan rates to obtain double-layer capacitance (Cdl) and evaluate the electrochemical active surface area (ECSA) for OER activity (Supplementary Fig. 16)27. The Mn0.2RuO2 and p-RuO2 show a similar morphology, however, the Cdl value of Mn0.2RuO2 is nearly 4 times higher than that of p-RuO2 (Fig. 4d). To consider the intrinsic activity of each active site, we normalized the LSV curves using both the ECSA and metal mass activity. Significantly, Mn0.2RuO2 still exhibits a lower overpotential than p-RuO2, illustrating that doping Mn into the RuO2 lattice is an effective approach to enhance the intrinsic activity of each metal site (Supplementary Fig. 17)28. The BET normalized TOF also support the higher intrinsic activity of Mn0.2RuO2 (Supplementary Figs. 18–19)29. Through online gas chromatography testing, it was demonstrated that Mn0.2RuO2 achieves a high OER Faradic efficiency exceeding 99% and 98% at current density of 10 mA cm−2 and 30 mA cm-2 for OER, respectively (Supplementary Fig. 20 and Supplementary Table 3). Additionally, electrochemical impedance spectroscopy (EIS) was performed on all samples at 1.45 V versus RHE, which reveals that the Mn0.2RuO2 displays the smallest charge transfer resistance (Fig. 4e)30. Moreover, the two-probe van der Pauw method was conducted at room temperature to analyze the conductivity capability of the electrocatalysts. The electronic conductivity of Mn0.2RuO2 is 3.3 S cm−1, considerably superior to those of p-RuO2 (1.98 S cm−1) and com-RuO2 (0.52 S cm−1), inferring that an asymmetric Mn4-δ-O-Ru4+δ structure facilitates the electron transfer in the sample (Supplementary Fig. 21). Therefore, a comprehensive analysis suggests that Mn0.2RuO2 demonstrates excellent OER activity, attributed to its outstanding reaction kinetics, mass activity, charge transfer, and electrical conductivity.

The durability of the electrocatalysts in acid is another crucial performance parameter for OER, which has been evaluated using chronopotentiometry stability test. Compared to the fast potential enhancement for p-RuO2 and com-RuO2, Mn0.2RuO2 manifests almost no obvious decays over 150 h operation at 10 mA cm−2 (Fig. 4g). Furthermore, Mn0.2RuO2 exhibits only a slight increase of 50 mV in potential over a duration of 95 h at 50 mA cm−2, demonstrating significantly improved acidic stability (Supplementary Fig. 22). ICP-MS was used to obtain quantitative information on Ru and Mn dissolution during the OER process (Supplementary Table 4). Notably, introducing Mn atoms into RuO2 lattice reduces the dissolution of Ru atoms to 1.3% after a 20 h durability test, which is significantly lower than previously reported Ru-based oxides (Supplementary Fig. 23)9. The stability number (S-number) of Mn0.2RuO2 represents improvement by an order of magnitude compared to p-RuO231,32, and Mn0.2RuO2 exhibits higher S-number than other Ru/Ir-based electrocatalysts previously reported (Supplementary Fig. 24 and Supplementary Tables 5-6). These results reveal that Mn0.2RuO2 exhibits outstanding resistance to corrosion and dissolution under acidic condition, which can be attributed to the unique electron structure of the Mn4-δ-O-Ru4+δ pairs.

The structural integrity of the Mn0.2RuO2 after OER was further examined. The morphological structure of Mn0.2RuO2 maintains its initial state, while metal vacancies were observed up to three atomic layers on the surface, as shown in Supplementary Figs. 25–27. To clarify the electronic structure of Mn0.2RuO2 after OER, XPS fitted areas were employed. The calculation results reveal that the local Mn4+ domains were formed after the OER testing (Supplementary Fig. 28 and Supplementary Table 7). Analyzes of the XPS and XANES spectra reveal that the increase of oxidation states for both Mn and Ru due to these as formed metal vacancies on the surface, whereas the octahedral structure remains stable. Meanwhile, no oxygen vacancies generated after the OER process (Supplementary Figs. 28-29). Additionally, we compared the overpotential at 10 mA cm−2 and long-time stability of Mn0.2RuO2 with recently reported Ru/Ir-based oxide OER catalysts in acidic media to evaluate their performance relative to those of state-of-art OER catalysts (Fig. 4h)19,33,34,35,36,37,38,39,40,41,42. Meanwhile, we also compared the recorded overpotentials to reach a specific mass-normalized current density for the catalysts (Supplementary Fig. 30 and Supplementary Table 8). The above results demonstrate that the Mn0.2RuO2 electrocatalyst with Mn4-δ-O-Ru4+δ structure shows better activity and higher stability than that of the RuO2 with the Ru-O-Ru structure.

In addition, we also assembled a PEM electrolyzer utilizing a Nafion 115 proton exchange membrane, with p-RuO2/Mn0.2RuO2 as the anode electrocatalysts and commercial Pt/C as the cathode electrocatalysts. The current-voltage (I-V) characteristic polarization curves show that the Mn0.2RuO2 based PEM electrolyzer achieves a high current density of 1 A cm−2 at a cell voltage of 1.69 V, outperforming p-RuO2 based one, which requires a 1.90 V cell voltage to reach 1 A cm−2. Meanwhile, p-RuO2 based PEM electrolyzer lost activity within 40 h at a current of 200 mA cm−2, while the Mn0.2RuO2 based PEM electrolyzer can be maintained for 180 h, indicating the potential for practical application of the Mn0.2RuO2 electrocatalyst (Supplementary Fig. 31 and Supplementary Table 9).

Elucidating the OER mechanism for the Mn4-δ-O-Ru4+δ microstructure sample using operando characterizations

To understand the dynamic structural evolution and reaction pathway on the Mn4-δ-O-Ru4+δ microstructure, operando characterizations, including operando SR-FTIR, operando DEMS, and operando XAFS, were used to identify the key intermediates on the Mn4-δ-O-Ru4+δ microstructure and the reaction active sites during actual OER working conditions.

Operando SR-FTIR is highly sensitive to the vibrations of key reactive intermediates, providing nearly real-time visualization of the structural evolution under working potentials. Figure 5a shows the emergence of distinctive vibration frequencies at 1069 cm−1 and 1113 cm−1 with the applied potential gradually increases from 1.3–1.5 V in Mn0.2RuO2. Based on previous reports, the emergence of vibration bands approximately 1100 cm−1 can be assigned to the bridging oxygen configuration at the dual active site22,25,43,44,45,46,47. This potential-dependent SR-FTIR absorption band indicates that the key *O-O* intermediate emerges at the dual-metal active sites simultaneously, forming a characteristic M-*O-O*-M configuration, suggesting that Mn0.2RuO2 follows an OPM pathway. For comparison, operando SR-FTIR measurements were also performed on p-RuO2. Apparently, a single broad infrared vibration band at 980 cm−1 emerges with increasing applied potentials from 1.3-1.5 V. This vibration band could be attributed to the typical Ru-*OOH intermediate, illustrating that p-RuO2 tends to follow a traditional AEM pathway (Fig. 5b)47. Additionally, operando DEMS with isotope labeling was carried out to validate the oxygen evolution mechanism, utilizing 0.5 M H2SO4 H218O as the electrolyte. The primary mass signals for p-RuO2 are 34O2 (16O18O) and 36O2 (18O18O) products, while for Mn0.2RuO2 32O2 (16O16O), 34O2 (16O18O) and 36O2 (18O18O) were detected (Figs. 5c, d). It is worth noting that the presence of 32O2 originates from surface oxygen adsorbates, and the direct coupling of two 16O adsorbates results in the observation of 32O2, a phenomenon exclusively observed in Mn0.2RuO2. The signal intensity of 36O2 is two and three orders of magnitude higher than that for 32O2 and 34O2, respectively, indicating that the oxygen products primarily originate from two 18O atoms in the isotope-labeled electrolyte. The Ru M-edge and Ru L3-edge XANES spectra, which are sensitive to atomic structure, further exclude the possibility of a lattice oxygen mechanism pathway (Supplementary Fig. 32)48. These results also support the occurrence of dual-metal active sites via the OPM pathway in Mn0.2RuO2.

Fig. 5: Operando SR-FTIR, DEMS, XAFS measurements.
figure 5

a, b Operando SR-FTIR spectra for Mn0.2RuO2 (a) and p-RuO2 (b). c, d Operando DEMS signals of 32O2, 34O2 and 36O2 for (c) Mn0.2RuO2 and (d) p-RuO2. e, f Operando Mn (e) and Ru (f) K-edge XANES spectra. Inset: magnified XANES spectra, and experimental Mn (e1) and Ru K-edge (f1) ΔXANES spectra. g, h Theoretical Mn (g) and Ru (h) K-edge spectra of Mn0.2RuO2 and OO-Mn0.2RuO2 (a model of *O-O* key intermediate absorption on the Mn0.2RuO2 surface) Inset: theoretical Mn and Ru K-edge ΔXANES spectra. i Schematic of the OPM pathway for Mn0.2RuO2.

Full size image

Operando XAFS technology was conducted to explore and gain active structure evolution information, due to its structure sensitivity. The Mn K-edge XANES spectra show a gradual positive shift in the absorption edge photon energy as the applied potential increases from OCP to 1.5 V, which indicates an increase in the Mn oxidation state, suggesting a partial electron transfer from Mn atoms to neighboring atoms or absorbed oxygen-containing species under potential-driven conditions (Fig. 5e). The Mn K-edge FT-EXAFS spectra shows that the main peak corresponds to first-shell of Mn-O coordination at approximately 1.5 Å. Interestingly, as the applied potential increases, the length of Mn-O bond remains nearly unchanged, indicating that lattice oxygen hardly participates in the OER process (Supplementary Figs. 33–34 and Supplementary Table 10)4,49. In the Ru K-edge XANES spectra, the adsorption edge position progressively increases as the applied potential is continuously raised in the oxidation reaction regions, resulting in higher oxidation states during the OER process (Fig. 5f). Analogously, the first-shell Ru-O bond remains at approximately 1.5 Å distance, and the change in potential does not affect the bond length between Ru and O atoms (Supplementary Figs. 33 and 35 and Supplementary Table 11). More importantly, to exclude the contribution of bulk materials and simultaneously amplify the signal response of the surface structural changes under operating conditions, delta-XAFS (ΔXAFS) spectra were introduced50. Based on the theory of multiple scattering, we simulated Mn and Ru K-edge XANES spectra based on the structure models of the initial and adsorbed key oxygen-containing (*O-O*) intermediates (Fig. 5g–h). The comparison between experimental and theoretical Mn K-edge ΔXANES spectra exhibits similar characteristic peaks, which are also observed in the Ru K-edge ΔXANES spectra. These results indicate that the subtle changes in the Mn and Ru K-edge XANES spectra under applied reaction potentials stem from the adsorption of oxygen-containing intermediates on Mn and Ru dual-metal active sites in Mn0.2RuO2. Moreover, we also simulated other non-bridging oxygen adsorption models, such as *OOH on the Ru atom and *O on the Mn atom. The theoretical ΔXANES spectra showed that this adsorption configuration do not match well with the experimental ΔXANES spectra, excluding other adsorbate configuration (Supplementary Fig. 36). Combined with operando SR-FTIR and operando DEMS results (Fig. 5i), it is evident the synergistic effect of both Mn and Ru atoms are involved the adsorption of oxygen-containing species and forming a characteristic Mn-*O-O*-Ru configuration of Mn0.2RuO2. This conclusion provides a deeper understanding on the catalytic mechanism transition from AEM to OPM.

DFT simulation to bridge the Mn4-δ-O-Ru4+δ microstructure with the excellent OER performance via the OPM pathway

Density functional theory (DFT) calculations were performed to gain an insight into the origin of the OPM pathway and the excellent OER performance of Mn4-δ-O-Ru4+δ microstructure in Mn0.2RuO2. The structural models of p-RuO2 and Mn0.2RuO2 were constructed based on the HAADF-STEM and EXAFS results, where the exposed (110) surface forms the active crystal plane (Supplementary Figs. 37–38 and Supplementary Table 12). The atomic coordinates of the optimized computational models are provided in Supplementary Data 1. Bader charge analysis was utilized to explore the charge distribution on the p-RuO2 and Mn0.2RuO2 surfaces. The substitution of Ru with Mn atoms in Mn0.2RuO2 leads to a remarkable alteration in the positive valence of the nearest neighboring Ru atoms (Figs. 6a, b and Supplementary Fig. 39), which strongly supports the Ru 3d XPS and Ru K-edge XANES results, and the Ru atom in the Mn-O-Ru microstructure of Mn0.2RuO2 shows an oxidation state > +4. In p-RuO2 with Ru-O-Ru microstructure, when *OH was adsorbed on one Ru site, the other Ru atom exhibits a limited valence state change (Fig. 6a). However, in Mn0.2RuO2 with Mn4-δ-O-Ru4+δ microstructure, when *OH is adsorbed on the Mn site, the electron donating ability of Ru atom changes remarkably, indicating a strong electron exchange capability between Mn and Ru atoms (Fig. 6b). After the adsorption of the *OH and *O species, the Ru atom of Ru-O-Ru microstructure in p-RuO2 losses about 1.88 |e| to *OH, which reduces to 1.76 |e| in the Mn4-δ-O-Ru4+δ microstructure in Mn0.2RuO2. As a result, the over-oxidation of active Ru atoms into soluble species could be depressed in Mn0.2RuO2 (Fig. 6a–b).

Fig. 6: OER mechanism analysis.
figure 6

a, b Bader charge analysis for p-RuO2 (a) and Mn0.2RuO2 (b). c Differential charge density analysis of adsorption *OH on Ru atom (left) and Mn (right) atom for Mn0.2RuO2. d Gibbs free energy diagram for AEM path and OPM path on the surface of Mn0.2RuO2. e PDOS and d-band center of Ru 4d (−4.23) for p-RuO2, Ru 4d (−4.32) and Mn 3d (−4.61) for Mn0.2RuO2. Corresponding d-band centers are denoted by arrows. f Calculated energy for Ru demetallation of p-RuO2 and Mn0.2RuO2.

Full size image

The differential charge density and adsorption energy of the *OH species on the Ru and Mn sites were compared, which indicates that Mn atoms serve as the initial reaction sites in Mn0.2RuO2 (Fig. 6c and Supplementary Fig. 40). Hence, the unique Mn4-δ-O-Ru4+δ structure shows optimal adsorption towards *OH species than the Ru-O-Ru microstructure. As shown in Fig. 6d, both the traditional AEM and dual-sites OPM pathways were considered for Mn0.2RuO2. In the AEM pathway, the rate-limiting step is the formation of *OOH intermediate at a single metal site, which has a higher free energy change of 0.56 eV. In the OPM pathway, the rate-limiting step shifts to *O-O* coupling and O2 releasing process, which occurs on dual-sites with lower free energy of 0.44 eV, thus effectively reducing the energy barrier of the rate-limiting step. These results are consistent with the findings of operando XANES and operando SR-FTIR experiments, confirming that the OER follows the OPM pathway on the Mn4-δ-O-Ru4+δ in Mn0.2RuO2. The Gibbs free energy assessment for both AEM and OPM pathways was performed for p-RuO2 (Supplementary Fig. 41). The AEM pathway exhibits a lower energy barrier on the p-RuO2 surface with the Ru-O-Ru microstructure, which agrees with the results of operando SR-FTIR.

Additionally, the partial densities of states (PDOS) analysis reveals that introduced Mn atoms could shift the Fermi level upwards, leading to the electron transfer from Ru to the Mn atom in Mn0.2RuO2 increases the valence state and is in agreement with the XANES and XPS results (Fig. 6e). The PDOS near the Fermi level increases for Mn0.2RuO2, indicating a higher availability of electron states for electron transfer, resulting in improved electrical conductivity. Furthermore, the introduction of Mn optimizes the d-band center of Ru, which facilitates the desorption of intermediate species. Finally, the corrosion resistance of p-RuO2 and Mn0.2RuO2 was evaluated. The energy cost of Ru demetallation significantly increases from 1.68 to 3.19 eV with the substitution of Mn atoms into RuO2 lattice, demonstrating that Mn incorporation stabilizes the lattice structure (Fig. 6f). Considering the experimental dissolution of Ru, we further calculated the Gibbs free energy change of Mn0.2RuO2 during the OER process via both the AEM and OPM pathways in the presence of Ru vacancy. Although vacancies affect the adsorption step of reactants, it is worth noting that following the OPM pathway still exhibits a lower energy barrier for the formation of key intermediates (Supplementary Fig. 42). These findings illustrate that the introduction of Mn atoms enables a switch in the reaction pathway even considering the dissolution of Ru atoms during the OER process.

Discussion

In this work, we used advanced operando characterization to elucidate the reaction pathway transition from AEM to OPM owing to microstructural changes induced by substituting Ru atoms with the Mn in octahedral rutile RuO2. We observed a key intermediate, *O-O*, during the OER process in Mn0.2RuO2 with Mn4-δ-O-Ru4+δ microstructure using operando SR-FTIR, which is direct evidence for the OPM pathway. In contrast, only *OOH, a key intermediate in the AEM pathway, was observed in the p-RuO2 with Ru-O-Ru microstructure. Operando DEMS also supported the presence of dual-metal active sites in Mn0.2RuO2, while the single-metal active site in p-RuO2 during the OER process. Moreover, we identified both the Mn and Ru atoms as active sites in the Mn4-δ-O-Ru4+δ dual-site structure, and form a bridge *O-O* intermediate using operando XANFS and ΔXANES spectra. This change in the reaction pathway remarkably enhanced the OER performance of Mn0.2RuO2 in acidic solutions. Experimental data and DFT simulations inferred that this reaction pathway change was related to the microstructural change from Ru-O-Ru to Mn4-δ-O-Ru4+δ, exhibiting better electron redistribution between the Mn/Ru pair as compared to the Ru/Ru pair in RuO2. In this case, the direct connection of *O atoms on Mn/Ru atoms was considerably easier than forming *OOH species on one metallic site. Additionally, the electron redistribution in the Mn4-δ-O-Ru4+δ increased the electrical conductivity of the material, facilitating the adsorption of *OH species and the desorption of the products, which imparted better stability to Mn4-δ-O-Ru4+δ in acid as compared to that of Ru-O-Ru. Therefore, this study confirmed that the OPM pathway could not be attributed only to the distance between two nearby active host centers. The primary factor for inducing an OPM pathway was the existence of a suitable microstructure where electrons could be delocalized between two active centers. Consequently, OER catalysts could be reasonably predicted and artificially designed with suitable microstructures that preferentially follow the OPM pathway.

Methods

Chemical and materials

Ruthenium chloride anhydrous (RuCl3, AR), Manganese nitrate tetrahydrate (Mn(NO3)2·4H2O, AR), urea (AR), glucose (AR), commercial RuO2 (AR), commercial Pt/C (AR), and heavy-oxygen water (H218O, 97 atom % 18O) were purchased from Aladdin Co., Ltd. Nafion solution (AR) was purchased from Sigma-Aldrich. Sulfuric acid (AR), Hydrogen peroxide 30% aqueous solution (H2O2, AR), and ethanol (AR) were purchased from Shanghai Research Institute of Chemical Industry. Nafion 115 membrane (N115, thickness = 127 μm) was purchased from Dupont. Polytertrafluoroethylene (PTFE) membrane (porosity ≥ 50%, pore size ≤ 20 nm, thickness = 40 μm) was provided by Linglu Instruments Shanghai. All chemicals were used without further purification.

Synthesis of catalysts

In a typical procedure, 2 g urea and 10 g glucose were mixed in 10 mL deionized water. The mixture was subjected to ultrasound for 30 minutes until a homogeneous solution was obtained. Subsequently, 20 mg Mn(NO3)2·4H2O and 83 mg RuCl3 were added to the solution, and the solution was stirred for 30 minutes to obtain a uniform colloidal suspension. Then, the solution was sealed and heated at 150 °C for 10 hours, forming a porous foam. Finally, the foam was annealed in a muffle furnace at 500 oC for 15 hours to prepare the Mn0.2RuO2 nanorods. The same preparation process was also used to prepare p-RuO2, Mn0.1RuO2 and Mn0.3RuO2 with different amounts of Mn(NO3)2·4H2O added.

Characterization

Both the SEM and TEM results are measured at the Experimental Center of Engineering and Materials Science, University of Science and Technology of China. The SEM images were captured by Gemini SEM 500, while JEM-2100F was used to acquire the TEM, HRTEM, and EDS mapping images. The XRD patterns were obtained using a Philips X’Pert Pro Super diffractometer with Cu Ka radiation (λ = 1.54178 Å) in the 2θ range of 20-80o. The concentration of dissolved metal in the electrolyte of the samples were determined by ICP-MS (Plasma Quad). Electron paramagnetic resonance spectra were generated using the Japan JES-FA 200 (JEOL) spectrometer. Atomic scale HADDF-STEM images, EELS spectra, and atomic scale mapping images were obtained using the Thermo Fisher Scientific Themis Z. The XPS signals of the samples were collected using a Thermo NEXSA G2 spectrometer equipped with an Al-Kα (hυ = 1486.6 eV) X-ray source at Instrument Center for Physical Science, University of Science and Technology of China51.

Electrochemical measurements

In a typical test, a mixture consisting of 0.75 mL of deionized water, 0.25 mL of ethanol, 0.03 mL of Nafion solution, and 5 mg of sample was sonicated for 1 hour to prepare a well-dispersed catalyst ink. For the preparation of the working electrode, 5 µL of catalyst ink was uniformly deposited onto a freshly polished glassy-carbon electrode (GCE) with a diameter of 0.3 cm, corresponding to a catalyst loading of 0.34 mg cm−2. The target electrolyte solution was prepared by diluting analytical-grade sulfuric acid in a brown volumetric flask, ensuring it was freshly prepared and promptly used. The calibration reference electrode process was as follows: Two Pt wires served as the working and counter electrodes, respectively, with Ag/AgCl being calibrated as the reference electrode, and H2-saturated 0.5 M H2SO4 solution was employed as the electrolyte52. The corrected reference electrode potential value was obtained through CV cycles performed near the open circuit voltage. In the standard three-electrode system, the GCE, Pt wire, and Ag/AgCl were utilized as the working, counter, and reference electrodes, respectively. In this work, OER measurements were conducted in an O2-saturated 0.5 M H2SO4 electrolyte (pH =0) using a CHI 760E electrochemical workstation at room temperature. The measured potentials were recorded versus the reference electrode and converted to reversible hydrogen electrode (RHE) following the relationship: E (V vs. RHE) = E (Ag/AgCl) + 0.059 pH + 0.197 V. OER polarization curves were corrected with iR-compensation following the formula: Potential (V vs. RHE) = E (V vs. RHE) – iR, where i represents current and R denotes the uncompensated ohmic resistance of the electrolyte. The iR-compensation was executed through on-the-fly correction, an automatic correction mode based on positive feedback, with the compensation level set to 85% unless otherwise specified. The R value for commercial RuO2 within this system was measured to be ∼2 ohms (Supplementary Fig. 43). The Nyquist plots for all samples were measured at the same potential value with an amplitude of 5 mV and frequencies ranging from 100 kHz to 0.1 Hz. The ECSA of the catalyst was estimated by determining the double-layer capacitance (Cdl) derived from the CV curves within a non-Faradaic potential window at a series of different scan rates of 10, 20, 30, 40, and 50 mV s-1. The ECSA was proportional to the Cdl53. Stability testing was conducted at a constant current density of 10 and 50 mA cm-2 using a CHI 760E electrochemical workstation.

PEMWE measurements

The catalysts-coated membrane (CCM) process was employed to prepare the membrane electrode assembly (MEA) for the PEM electrolyzer. Before the construction of CCM, N115 membrane was treated with H2O2, deionized water, and 0.5 M H2SO4 at 80 °C for 1 hour, respectively. The treated N115 membrane was stored in deionized water. The anode catalyst ink consisted of synthesized p-RuO2 and Mn0.2RuO2, ionomer (D2020), and dispersing solution. In detail, the anode catalyst (5 mg mL−1) was added to a mixed solution of water and isopropanol (Vwater: Visopropanol = 1:3) and sonicated for 30 minutes to ensure complete wetting of the catalyst. Subsequently, 10 wt % ionomer was incorporated into the dispersion, followed by another 30 minutes to prepare the anode catalyst ink. The cathode catalyst ink (commercial Pt/C, 5 mg mL−1) was obtained by the same procedure. Afterwards, the catalysts were uniformly sprayed onto the proton exchange membrane using the ultrasonic spraying method to form the CCM for the PEM electrolyzer. Pt-coated Ti felts were used as the porous transport layer (PTL) for both the cathode and anode in the construction of the CCM electrolyzer. The effective active area of the electrolyzer was 5 cm2 (2 cm × 2.5 cm). The PEM electrolyzers were operated at 60 °C, with deionized water as the reactant at an inlet flow rate of 10 mL min−1 (recirculated by a peristaltic pump). Polarization curves were examined over the range of 0–3 A cm−2, and stability was assessed using the timed current curves at 200 mA cm−2.

XAFS measurements

Mn K-edge and Ru K-edge XAFS spectra were collected at the 1W1B and BL13SSW beamline stations of the Beijing synchrotron radiation facility (BSRF) and Shanghai synchrotron radiation facility (SSRF), respectively. The positions of the absorption edges were calibrated using Mn foil and Ru foil standard samples. Operando XAFS spectra of Mn0.2RuO2 were collected using a custom-built electrochemical cell. A thin carbon paper loaded with Mn0.2RuO2 catalyst served as the working electrode, while Pt wire and Ag/AgCl were used as the counter and reference electrodes, respectively. To prevent electrolyte leakage and ensure accurate sample detection, Kapton film was applied to the opening window of the electrochemical cell (Supplementary Fig. 44)54. During the OER process, potentials ranging from OCP to 1.5 V vs. RHE were applied under constant voltage, and the XAFS spectra were collected using fluorescence mode. The Mn K-edge and Ru K-edge EXAFS spectra were processed using the ATHENA module of the IFEFFIT software package. Additionally, the O K-edge, Ru M-edge, and Mn L-edge XANES spectra were collected at the BL12B beamline station of the National synchrotron radiation laboratory (NSRL), while the Ru L3-edge XANES spectra were obtained at the BL16U1 beamline station of the SSRF.

SR-FTIR measurements

Operando SR-FTIR experiments were conducted at the BL01B beamline station of NSRL using a custom-built cell (Supplementary Fig. 44)55,56. The p-RuO2 and Mn0.2RuO2 catalysts were loaded onto a GCE embedded in the center of the cell, while Pt wire and Ag/AgCl were served as the counter and reference electrodes, respectively. To prevent electrolyte evaporation from damaging the microscope lens, a BaF2 window was placed aboved the cell. During the OER process, potentials were applied from OCP to 1.5 V vs. RHE under constant voltage, and the SR-FTIR spectra were recorded by averaging 126 scans at a resolution of 1 cm−1.

DEMS with isotope labeling

Operando DEMS with isotope labeling measurements were carried out using the QAS 100 device (Linglu Instruments, Shanghai). The p-RuO2 and Mn0.2RuO2 catalyst ink (5 mg mL-1) were directly deposited onto the Au film sputtered onto a porous PTFE membrane, which served as the working electrode. The hydrophobic PTFE membrane allows gas flow while rejecting liquid (Supplementary Fig. 44). A saturated Ag/AgCl electrode and Pt wire were used as reference electrode and counter electrode, respectively. For the isotope labeling studied, 2 mL 0.5 M H2SO4 solution was prepared using H218O (97 atom % 18O) as the solvent. The p-RuO2 and Mn0.2RuO2 catalysts were subjected to three LSV cycles within the potential range of 1.0–1.6 V vs RHE at a scan rate of 10 mV s−1. The mass signals of different molecular weight gas products were recorded in real-time, including 32O2 (16O16O), 34O2 (16O18O), and 36O2 (18O18O). Before the electrochemical measurements, the electrolyte was purged with high-purity Ar gas to remove the dissolved oxygen.

Theoretical calculation details

First-principles calculations were carried out using the Vienna Ab-initio Simulation Package (VASP) code. To enhance calculation efficiency, projector augmented wave (PAW) pseudo-potentials were employed to treat the core electrons, and the Perdew, Burke, and Ernzernhof (PBE) functional was used for exchange-correlation corrections57,58,59,60. A plane-wave cut-off energy of 450 eV was set to expand the electron wave functions. The convergence criteria for the electronic self-consistent iteration and the maximum force on each atom were set to 10−5 eV and 0.02 eV Å−1, respectively. During the structural and electronic calculations, K points grids of 3 × 3 × 1 and 5 × 5 × 1 in the Brillouin zone were generated automatically using the Gamma center method. An effective U-J value of 4.00 eV was applied for Mn atom61,62.