Introduction

Understanding how the electrocatalyst/electrolyte interface reforms under operating conditions1,2,3,4 can offer mechanistic insight that allows tracking of the catalytically active motif and enlighten a path towards the development of active and stable electrocatalysts. The nature of such nanoscale interfaces is likely heterogeneous and transforms strongly based on electrochemical operation. To date, however, probing such a dynamic process has been challenging because the surface represents only a tiny fraction of a bulk electrocatalyst, and heterogeneous solid/liquid nanojunctions are nontrivial to characterize in situ5,6. Electrochemical water splitting for emerging alternative fuels has been considered one of the most feasible tactics to meet the challenge of decarbonization. Identification of the key intermediate state during the oxygen-evolution reaction (OER)7,8,9, which is critical to water splitting and CO2 reduction reactions, remains elusive. The dynamic evolution of irreversible surface reconstruction on an electrocatalyst during OER and the deciphering of the reaction mechanisms are of utmost importance6,10,11.

Engineering the coordination environment of a metal center is of fundamental importance in heterogeneous catalysis12,13. In particular, modulating the metal–oxygen bonding environment at the electrocatalyst surface offers an effective path toward enhancing the interfacial reactivity. The surface geometric construction and electronic regulation are found to be two decisive factors for the intrinsic improvement of electrocatalytic performance14. Accordingly, the first-row 3d-transition metal (TM) oxides have recently been regarded as promising candidates for OER, owing to their earth abundance and fascinating electronic properties derived from the crystal-field theory. Exemplified by a TM element, the valence variation of a TM plays a pivotal role in catalyzing water oxidation. It is commonly recognized that the high-valent TM species and oxyhydroxides can engender a high reactivity towards OER15. Accompanying over-oxidation, the strong hybridization between TM 3d and O 2p facilitates the interaction of electrocatalysts and oxygen-related adsorbates16, leading to structural disorder and electrochemical irreversibility, which makes tracking technologically challenging. For this reason, it is not straightforward to unravel the nature of the highly covalent bonds in over-oxidized species during OER.

Cu-based materials have been proven to be state-of-the-art catalysts for CO2 reduction17,18 and photocatalytic water reduction19. So far the OER performance measured with Cu-based electrocatalysts have, however, fallen far short of a satisfactory standard; in particular, they underperform relative to other late 3d-TM (e.g., Co and Ni) oxides20,21. As the cause of this poor performance has not been the focus of extensive studies during the past decade, seeking an inspiring strategy to accelerate OER kinetics on Cu-based electrocatalysts and unveiling their active sites are very challenging but urgent tasks.

Herein, the rational regulation of the active Cu center with unsaturated coordination (denoted H-Cu2O) via facile hydrogenation is described. Specifically, the substitutional hydrogen induces a surface geometric rearrangement of copper(I) oxide to generate the distinctive oxygen nonstoichiometry, resulting in exceptional performance and durability. Many operando spectroscopic tools, including grazing-angle X-ray scattering (GAXS), quick X-ray absorption (quick-XAS), soft X-ray absorption (soft-XAS), Raman spectra and electrochemical impedance spectroscopy (EIS), were utilized to uncover that the Zhang-Rice singlet state22,23,24,25 is unexpectedly observed to participate directly in OER-cycle superconductors. For the first time, a Zhang-Rice singlet state is proposed to trigger electrocatalytic oxygen release. Notably, such a high-valent CuO4 geometry containing Cu3+ with d9L (L: an O 2p hole) charged character of Cu d-electrons is a paramount key to dominate the oxygen-evolution step of the OER rather than a CuOOH species, showing a fundamental difference from the conventional scheme.

Results

Catalyst characterization

A scanning-electron-microscope (SEM) image of the H-Cu2O catalyst displays a rod-like morphology (Supplementary Fig. 1). This H-Cu2O catalyst had disordered surfaces, of which the thickness of the disordered outer layer surrounding a crystalline core was about 6 nm as depicted in illustrated in Fig. 1a. A transmission-electron-microscope (TEM) image and the corresponding fast-Fourier-transform (FFT) patterns of the H-Cu2O catalyst further validate the crystalline core covered with a thin structurally disordered layer, induced by the hydrogenation (Supplementary Fig. 2). Secondary-ion mass-spectrometry (SIMS) depth profiles of the H-Cu2O catalyst illustrate a clear introduction of hydrogen content in the surface of thickness ca. 6–8 nm (Supplementary Fig. 3). To understand deeply the disordered surface, multiple techniques, including X-ray photoelectron spectra (XPS) and low-grazing-angle quick-XAS, were employed to characterize the samples as prepared. The Cu 2p3/2 XPS spectra show a main feature at ∼932.4 eV26, corresponding to the chemical state of Cu1+ or Cu0 (Supplementary Fig. 4a). The presence of Cu1+ for H-Cu2O catalyst was confirmed by the dominant signal at 916.8 eV in the Cu LMM spectra (Supplementary Fig. 4b)26. As shown in Fig. 1b, soft-XAS at the Cu-L3 edge were employed to probe directly the Cu 3d electronic properties of the H-Cu2O catalysts. The main feature at 933.4 eV in Cu2O is assigned to the final state 2p3/2p6d10s1 from initial state 3d10 for Cu1+27. Clearly, both of Cu2O and H-Cu2O confirm the presence of Cu1+ surface. Since Cu1+ is fully occupied with ten electrons in 3d, the broad and weak peak at 933.4 is assigned to the unoccupied of 4s1 states27. Relative to pristine Cu2O, the decreased intensity at 933.4 eV indicates that the unoccupied states of 4s1 decrease in H-Cu2O. This result shows slight charge localization induced by hydrogenation, likely leading to more rapid charge transfer for the OER. X-ray absorption near-edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) were further recorded to reveal the chemical state and local structure of the H-Cu2O catalysts. Note that there is no obvious energy shift between Cu2O and H-Cu2O in the Cu K-edge XANES spectra, exhibiting the primary Cu1+ features (Supplementary Figs. 5, 6). The result agrees with the Cu LMM Auger spectra (Supplementary Fig. 4b). The Fourier-transformed k3-weighted EXAFS spectra (Fig. 1c), in which two main features at ~1.5 Å and 2.7 Å correspond to the scattering paths of the nearest oxygen (Cu−O) and the secondary copper atoms (Cu−Cu), testify that most coordination sites of the unsaturated Cu centers appear in the H-Cu2O catalysts. Based on these observations, we believe that the H dopant could reform the local electronic configuration and atomic arrangement of bonded Cu and adjacent O atoms and consequently enhance the localization capacity of charges on Cu atoms. In theory, the bonding strength of oxygen intermediates could depend on the degree of the filling in the antibonding states28. The enhanced localization capacity of charges on Cu could contribute to much filling of the antibonding states, resulting in weak adsorption of oxygen intermediates. In contrast, the less filling of the antibonding states would result in the strong adsorption of oxygen intermediates. According to OER volcano plot in metal oxides, the binding strength with oxygen intermediates should be neither too strong nor too weak29. For p-type Cu2O, the highest occupied d-state is quite closer to the Fermi level, resulting in the less filling of antibonding states and the stronger adsorption of oxygen intermediates. Since the adsorption of oxygen on Cu2O is too strong that restricts OER activity, the H-Cu2O is effective for filling much of the antibonding states with weak intermediates adsorption, and thus achieving a better activity30.

Fig. 1: Structural characterization and OER performance of H-Cu2O catalysts.
figure 1

a TEM image of H-Cu2O. b Cu L3-edge XANES spectra of Cu2O, H-Cu2O and references (ref.). c Cu K-edge EXAFS spectra of Cu2O and H-Cu2O. d OER polarization curves of Cu2O, H-Cu2O, Cu(OH)2, and Cu foam in 1.0 M KOH solution with 90% iR-correction. e The corresponding Tafel plots of all catalysts. f The long-term electrochemical stability of H-Cu2O was measured at current density 10 mA cm−2 (without iR-correction). The catalyst loadings for Cu2O and H-Cu2O were 11 mg cm−2 and 10.8 mg cm−2, respectively.

Full size image

Electrochemical properties toward OER activity

The OER catalytic performance of the H-Cu2O catalysts was evaluated using linear-sweep voltammetry (LSV) in KOH solution (1 M) at scan rate 1 mV s–1. With a definite coordination-unsaturated structure, the H-Cu2O model catalysts exhibit a small overpotential 263 mV, small Tafel slope 91 mV dec–1 and high durability over 45 h at a current density 10 mA cm–2 for OER in alkaline media (Fig. 1d–f). Such an exceptional performance is superior to those of previously reported 3d TM-oxide catalysts (Supplementary Fig. 7). To well evaluate the stability of H-Cu2O, the long-term durability at 10 mA cm–2 and 100 mA cm–2 are tested (Supplementary Figs. 22, 25 and 28). Evidently, the results show negligible decay of H-Cu2O after OER-100 hours of operation with 93% Faradaic efficiency, indicating good stability of the reconstructed surface. Apart from the overpotential and Tafel slope, the electrochemically active surface area (ECSA), which is estimated from the double-layer capacitances (Cdl), is another significant controlling factor for the intrinsic activity of catalysts. The Cdl value of H-Cu2O is 13.7% greater than for Cu2O, indicating that hydrogenation certainly augments the number of active sites, which is beneficial for the OER (Supplementary Fig. 8c). To exclude the contribution of larger ECSA for OER performance, a histogram of specific activity of the catalysts with error bar at 1.6 V vs. RHE is given (Supplementary Fig. 8d), reflecting good intrinsic electrocatalytic activity in H-Cu2O catalysts. We conclude that the improved OER performance of the H-Cu2O catalysts is attributed to not only the increased ECSA but also the enhanced intrinsic activity owing to the enriched electron densities and coordinatively unsaturated Cu centers. In addition, the H-Cu2O catalysts disclose a small resistance of charge transfer relative to Cu2O as extracted from EIS analyses (Supplementary Fig. 9). To probe the OER kinetics and the properties of the catalyst/electrolyte interfaces, we exploited operando EIS measurements (Supplementary Fig. 10). As displayed in the 3D contour Bode plots during OER, the phase-angle relaxation at the low-frequency region (100–101 Hz) is closely related to the charge transfer at catalyst/electrolyte interfaces. When the applied potential increase, the phase-angle of H-Cu2O at low-frequency region decreases quickly relative to Cu2O, indicating the superfast OER kinetics due to the low-coordinated environment of Cu atoms.

Identification of the active site

To capture the dynamic structural reconstruction or transformation of the catalysts, we implemented operando GAXS and operando quick-XAS during the OER operation. To obtain a more intuitive impression, a customized operando liquid electrochemical cell was designed as depicted in Fig. 2a. Figure 2b, c presents 2D contour plots of the color-coded scattering intensities as a function of applied potential for the catalysts recorded with operando GAXS at 18 keV of synchrotron X-ray. Stages I, II, and III are assigned to hydroxylation, oxygen evolution and after OER (potential off), respectively. The main characteristic scattering signals are attributed to Cu2O (111) and Cu2O (200) facets throughout the entire potential range. In region II, the formation of a Cu(OH)2 phase on the surface of catalysts is captured in both Cu2O and H-Cu2O catalysts. Such Cu(OH)2 phases well retain their situation under a potential-off condition, indicating an irreversible surface reformation from Cu2O to Cu(OH)2 (Fig. 2d). Relative to pure Cu2O, stage II at a much smaller potential (below 1.5 V) for H-Cu2O catalysts is observed, inferring more rapid deprotonation on the reconstructed surface of Cu2O catalysts.

Fig. 2: Operando GAXS and quick-XAS characterizations for the OER.
figure 2

a Schematic illustration of the operando liquid electrochemical cell for the GAXS and quick-XAS apparatus. Operando GAXS of (b) pure Cu2O and (c) H-Cu2O catalysts. d Proposed structural reconstruction during OER. Operando Fourier-transformed EXAFS of (e) pure Cu2O and (f) H-Cu2O catalysts (stages I, II and III are assigned to hydroxylation, oxygen evolution and after OER (potential off), respectively).

Full size image

Operando Cu K-edge XANES spectra of the H-Cu2O catalysts and pristine Cu2O under the OER are provided in Supplementary Fig. 11. The edge energy of H-Cu2O at open-circuit potential (OCP) is identical to that of pure Cu2O, confirming the nature of Cu1+. Upon increasing the anodic potentials, the absorption edge of Cu K-edge XANES spectra for H-Cu2O rapidly shifts to greater energies relative to pure Cu2O, supporting a more rapid deprotonation on the reconstructed surface of the H-Cu2O catalysts31,32. To probe the electronic structure of the catalysts during OER process, we recorded operando Cu K-edge XANES spectra. The absorption edge (the first-derivative signal) denoted with a dotted line in Fig. 3a, b, shows a smaller energy shift with increased applied voltage, indicating an increased average Cu valence state. A sharp shift was located at a smaller potential (1.4 V) for H-Cu2O catalysts vs. 1.6 V for Cu2O. Besides, the appearance of a slight shoulder at ∼8985 eV on the rising K-edge XAS at applied voltage 1.5 V for H-Cu2O and 1.7 V for Cu2O, respectively. The feature can be assigned to four-coordinated square–planar geometry33. On this basis, the oxidation state of the copper site at stage II might be assigned as +3, showing the electrochemically driven conversion of Cu(OH)2 into CuO4 geometry probably. We recorded also the operando Cu K-edge k3-weighted EXAFS spectra, which are sensitive to the local crystal structures of an absorbing metal ion. The fitted profiles of Fourier-transformed EXAFS spectra (FT-EXAFS) as results are presented in Supplementary Figs. 12 and 13 and Supplementary Tables 1 and 2, respectively. As shown in Fig. 2e, f, the corresponding 3D FT patterns along with 2D contour plots reveal a remarkable two-step dynamic structural evolution during electrochemical operation. The detailed coordination environment involving the relative coordination numbers and bond lengths of Cu centers is summarized quantitatively in Fig. 3c and Supplementary Fig. 14. Before the oxygen evolution (i.e., region I), the coordination numbers (ΔN/NOCP) and bond lengths (ΔR/ROCP) of Cu−O for H-Cu2O catalysts sharply increased when the applied potential was greater than OCP. For H-Cu2O catalysts, the largest ΔN/NOCP and ΔR/ROCP were obtained at the onset of OER at 1.4 V, because of a phase transition from Cu2O to Cu(OH)2 (Fig. 3c). In contrast, ΔN/NOCP and ΔR/ROCP of pure Cu2O slowly increased with applied potential up to 1.6 V (Fig. 3d). Note that the hydrogenation treatment is an effective route to accelerate Cu pre-oxidation and surface reconstruction during OER. As the applied potential increased further to 1.6 V (i.e., region II), ΔN/NOCP and ΔR/ROCP of the catalysts decreased gradually, likely implying the electrochemical oxidation of Cu(OH)2 to form an over-oxidized Cu species during oxygen evolution (Supplementary Fig. 14). This is also accordant with the operando XANES spectra (Fig. 3a, b). Such structural reconstruction of Cu species for H-Cu2O catalysts is also evident from the wavelet transform (WT) analyses of EXAFS spectra as depicted in Fig. 3e. Two dashed lines at 5.6 and 7.7 Å–1 are defined as the strongest oscillation amplitudes of Cu–O and Cu–Cu bonds. The yellow region represents the FT magnitude of the Cu–O and Cu–Cu bonds in H-Cu2O during OER and two references located about 1.8 and 2.7 Å. In the OCP state, the coordination environment of Cu in H-Cu2O was confirmed as Cu2O because of a similar pattern. On switching on a voltage, a phase transition from H-Cu2O toward Cu(OH)2 occurred (stage I).

Fig. 3: Understanding the correlation between OER activity and local structure transformation.
figure 3

a, b 3D patterns of operando Cu K-edge XANES of H-Cu2O and Cu2O catalysts. c, d Structural coherence change of Cu–O in EXAFS coordination number (N) and bond length (R) of H-Cu2O and Cu2O catalysts under an applied potential relative to the OCP state. e Comparison of Cu K-edge WT-EXAFS recorded for H-Cu2O, standard references (Ref.) and catalytic materials at OCP, 1.4, 1.5 V and after OER (stages I, II, and III are assigned to hydroxylation, oxygen evolution and after OER (potential off), respectively).

Full size image

Upon further increased applied voltage from 1.4 V, Cu(OH)2 was further transformed to a operando-generated copper species, which has different patterns for Cu–O and Cu–Cu bonds in stage II. In addition, the Cu–Cu bond shows a different amplitude distribution between stages I and II in the k-space from 7.7 to 6.1 Å–1. After OER, the pattern of the Cu–Cu bond shifted to a larger k value and a larger Cu–Cu bond distance; phase III was formed. The surface chemistry of H-Cu2O catalysts was hence notably altered by dynamic surface reconstruction during OER. This effect indicates that the reconstruction facilitated by hydrogenation is the vital key to evolve oxygen effectively.

The electronic structures of the real active sites in H-Cu2O catalysts during the OER were well assessed using operando soft-XAS (Fig. 4) at the Cu L3-edge and O K-edge as known from previous studies on the high-temperature superconductors (high-Tc) Cu superconductors34,35 and even more complicated Cu oxide36,37. The major challenge is to separate the liquid from the ultrahigh vacuum for OER experiments. Figure 4a shows an operando flow cell with an ultrathin Au@Si3N4 membrane window to separate the liquid cell from the ultrahigh vacuum condition. Figure 4b shows the Cu-L3 XAS of H-Cu2O in regions I–III. At the beginning of stage I (OCP state), we see a broad feature at 933.4 eV (black circles) for an initial Cu1+ (3d10) state. A sharp feature at 931.3 eV appeared at applied voltage 1.4 V in region II (green circles) and is assigned to Cu2+ (2p53d10 final state from 3d9 initial state) from Cu(OH)2. Most interestingly, a high broad shoulder feature at 932.4 eV appeared under the OER condition, which is well known in the observed La1–xSrxCuO4 (Supplementary Fig. 15) corresponding to the doped hole for Cu3+ state with 2p53d10L final state34,35,38 (L denotes a hole in the O 2p states). This completely agrees with the operando XANES results (Fig. 3a, b). The spectral weight of this final state is weak as doped holes are located mainly at the O 2p state19,38. Upon switching off the applied voltage, this Cu3+ state immediately vanished and only Cu2+ remained. As the doped holes are located mainly at the O 2p states, the quantitative content of the related state Cu3+ is expected to be observed from the O-K XAS spectra. It is well known that for a charge-transfer system with increased valence state, the pre-edge features shift to lower energy, and their spectral weight increases35,39.

Fig. 4: Operando soft-XAS characterization of high-valent Cu as an active site during OER.
figure 4

a Schematic illustration of operando soft-XAS setup under ultrahigh vacuum condition. b Cu L3-edge and c O K-edge of H-Cu2O. d Electronic spin states and orbital physics of Cu site in OER. e Schematic illustration of copper geometry site interaction with oxygen-related adsorbate during oxygen evolution.

Full size image

As shown in Fig. 4c, a strong pre-edge peak at 530.3 eV occurs under an applied voltage below 1.4 V, which is assigned to the transitions to the upper Hubbard band from the O 1 s core level (corresponding to the Cu2+ state for simplicity). Under OER conditions (stage II) a feature below 529 eV was observed, which is attributable to transitions from O1s to the doped hole states constructed by the strong O 2p-Cu 3d hybridization so-called Zhang-Rice singlet state or Cu3+ state35. A similar feature was recently observed also for cuprate superconductor Ba2CuO4–y, wherein the local octahedron is in an exceptionally compressed version40. Both the Cu-L3 and the O-K XAS spectra hence demonstrate the existence of a Zhang-Rice singlet state or Cu3+ state under OER conditions. The Cu3+ (3d9L) states are split into two poorly resolved peaks at 528.17 eV and 529 eV. A similar splitting was observed in Sr14Cu24O41 originating from a different local environment of oxygen ions36,41 (Supplementary Fig. 16c). Note that the Zhang-Rice singlet state disappeared upon switching off the applied voltage, which means that this real OER active species cannot be observed in experiments ex situ.

In addition to the Cu3+ species, a feature at 531.4 eV appears, which can be assigned to a CuOx(OH)y-related intermediate1. After switching off the applied voltage, those features disappear. To obtain the detailed spectral weight of the Zhang-Rice singlet-state-related spectral weight under OER conditions, we analyzed the O-K XAS (black circles) after subtracting an edge jump (black line in Fig. 4c) in the same way as used for high-Tc Cu superconductors34 as shown in Supplementary Figs. 16a and 17 with the O-K XAS of La2–xSrxCuO434 (Supplementary Fig. 16b) for comparison. The relative spectral weight of Cu3+ species corresponds to x = 0.1 in La2–xSrxCuO4.

Proposed catalytic mechanism

Based on all above operando X-ray spectroscopic data, a schematic electronic structure at 1.4 V and under OER conditions is depicted in Fig. 4d. As a first step, Cu2O with Cu1+ (3d10) state transfers gradually to Cu(OH)2 below 1.5 V in the region I. The Cu2+ ion in the Cu(OH)2 phase has a five-coordinated square-pyramidal geometry. On further increasing the applied voltage, part of Cu2+ was transferred to Cu3+ (3d9L) in CuO4 with four-coordinated square–planar geometry in stage II. After switching off the applied voltage, the Cu3+ (3d9L) species returned to the Cu(OH)2 state. With the valence state of the late TM ion increased by one unit, the charge transfer energy decreased by 3–4 eV and even became a negative value. The O 2p character gained weight above EF and shifted near EF42,43, which means that the covalence between Cu 3d and O 2p increased from Cu1+ to Cu2+ and further to a Cu3+ state. Previous work indicated that the catalytic activity of TM oxides could be enhanced on increasing the covalence between TM 3d and oxygen 2p orbitals6. Figure 4e shows the oxygen generation from the surface of H-Cu2O at the Cu3+ (3d9L) state, which had a square–planar geometry with a bare Cu3+ as an active site. The CV analyses for catalysts at scan rate of 5 mV s–1 are supplemented to support the in situ generation of Cu3+ species as shown in Supplementary Fig. 26. During the oxidation process, only one anodic peak appeared at low potential and was attributed to the conversion of Cu(I) into Cu(II). Coincidentally, the anodic peak of Cu(II)/Cu(III) oxidation at the high- potential region highly overlapped the large OER current response. Even so, in the subsequent reduction process on the reverse potential scan, the broad cathodic peak was observed and resulted from the reductive transformation of Cu(III) to Cu(II).

To explore further the unique geometry sites, we recorded operando Raman spectra of H-Cu2O in KOH (1.0 M). The Raman spectra in Supplementary Fig. 18 show a comparison of H-Cu2O and references. As shown in Supplementary Fig. 19a, the spectra also exhibited two major parts, stage I (hydroxylation below 1.5 V) and II (oxygen evolution above 1.5 V). The H-Cu2O phase transition was gradual from Cu2O to Cu(OH)2 with increasing potential in stage I. Three peaks appeared at 292.2, 490.1, and 3557.4 cm–1, of which the latter corresponds well to the stretching vibration of O–H in Cu(OH)2 reference. In stage II, a peak at 603.9 cm–1 appeared, due mainly to the framework vibration of Cu–O in CuO444. In the deuterium water experiment (Supplementary Fig. 19b), the peak at 490.1 cm–1 obviously shifted to 480.1 cm–1 with a shift ratio of 97.9%, which further proves that the 480.1 cm–1 peak is attributed to the Cu–OH structure. Furthermore, the redshift of 3557.4 cm–1 to 2628.8 cm–1 verifies the existence of the O–H structure. It must be clarified that there is no obvious redshift at 605.1 cm–1 in the deuterium experiment, indicating that there is no H in this structure. In addition, the potential is also relevant to the appearance and disappearance at of the peak at 605 cm–1 (Supplementary Fig. 20). The structure CuO4 disappeared quickly when the potential was cut off. It demonstrates that CuO4 is the active structure; this result is consistent with previous X-ray measurements. A continuous phase transition can also be demonstrated in the TEM images of H-Cu2O catalyst as shown in (Supplementary Fig. 21). After the OER catalytic process, the H-Cu2O crystal structure consisted of a Cu2O core (circle I) and a Cu(OH)2 shell (circle II) (Supplementary Fig. 21a). The crystal planes related to the Cu2O core and Cu(OH)2 shell were also identified in SAED patterns, which is consistent with previous GAXS patterns (Supplementary Fig. 21b, c). The average depth (~30 nm) of the surface layer of outer Cu(OH)2 was further observed on an enlarged scale of TEM images (Supplementary Fig. 21d–f), but the reaction depth (~30 nm) is much larger than a disorder layer (~6 nm) as such reconstruction occurred also for a pure Cu2O phase. Another interesting finding is the further generation of an amorphous region on the catalyst surface. The amorphous region has putative active sites that are tuned back from the oxygen-related species during the OER. These operando X-ray measurements and TEM results indicate that the structural disorder led to the rapid surface reconstruction at H-Cu2O in OER31,45. Taking into account all complementary information from multiple operando experiments including GAXS, quick-XAS, soft-XAS, Raman, and EIS, the active-site configuration and reaction cycle are proposed in Fig. 5, taking advantage of the fact that OH− is preferentially adsorbed on the region of the coordinative unsaturation, and additionally the intrinsic tendency of O 2p to favor the delocalization of local electrons for a noticeable transformation of the pre-designed metal–oxygen bonding environment. The impact of the hydrogenation on H-Cu2O is observed herein to generate the coordinatively unsaturated Cu centers with strong charge localization, which allows the Cu cations to be easily over-oxidized, thereby leading to a facile transformation to high-valent CuO4 geometry44. It is noticeable that we report on operando spectroscopic observations of the Zhang-Rice singlet state responsible for managing the oxygen-evolution step in the form of high-valent Cu3+ with d9L configurations as revealed by soft-XAS. To the best of our knowledge, a Zhang-Rice singlet state serving as active center has not yet been observed for OER electrocatalysts to date. Our results provide direct evidence of high-valent CuO4 sites, rather than oxyhydroxide species, as the key intermediate state of the pre-equilibrium step on H-Cu2O for oxygen evolution. Furthermore, as shown in Supplementary Fig. S23, the H-Cu2O catalyst exhibits pH-dependent OER activity, implying that non-concerted proton-electron transfers may participate in catalyzing the OER46. Under alkaline conditions, electrochemically driven deprotonation results in the intramolecular hydroxyl nucleophilic attack pathway where the adjacent OH− attacks CuO4 to form the O–O bond. Therefore, a pH-dependent nucleophilic attack pathway for O–O bond formation might be presented as shown in Supplementary Fig. S24.

Fig. 5: Proposed OER mechanism for H-Cu2O.
figure 5

Dynamic configuration of active sites during OER.

Full size image

As demonstrated for H-Cu2O electrocatalysts, several operando spectral methods allowed us to disentangle the dynamic restructuring during OER at nanoscopic solid/liquid interfaces. This report is the first of the unexpected observation of the Zhang-Rice physics responsible for managing the oxygen-evolution step in the form of high-valent Cu3+ with d9L charge character. Specifically, maximizing the hybridization between Cu(3d) and O(2p) states with the additional appearance of a ligand hole in O 2p orbitals favors oxygen-evolution catalysis involved in the OER cycle. Notably, the nature of the definite Zhang-Rice singlet state corroborates the high-valent CuO4 geometry with four-coordinated square–planar geometry that served as the key intermediate state in the pre-equilibrium step. As far as we are aware, this unusual observation is in sharp contrast with the commonly proposed scheme, predicting an oxyhydroxide species as the active center for OER electrocatalysts. Our work emphasizes that the charge and spin states of TM oxides would be essential to catalyze oxygen evolution during water oxidation.

Methods

Preparation of pristine Cu2O and H-Cu2O electrodes

The electrodeposition of Cu2O was implemented in a two-electrode configuration, including working (Cu foam) and counter (platinum foil) electrodes. The Cu foam was cleaned with sequential ultrasonic treatments in hydrochloric acid (1 M) and deionized water for 10 min each. The surficial Cu(OH)2 layer on the Cu foam was formed by the subsequent anodization in a NaOH solution (1 M). The electrodeposition condition was performed under a constant current density 4.5 mA cm–2 at 18 °C until the potential 1.6 V was reached. These Cu2O electrodes with surficial Cu(OH)2 layers were rinsed with copious water, dried in air and annealed at 500 °C (heating rate 5 °C min–1) in a tubular furnace with a flowing nitrogen stream for 5 h.

Preparation of H-Cu2O electrodes

In a typical synthetic procedure of a hydrogenation treatment, the H-Cu2O electrode was set up based on pristine Cu2O on applying gaseous H2 (99.999%) as the hydrogen source. First, the pristine Cu2O electrode was placed in a tubular silica furnace and kept under vacuum for 1 h. The furnace was then filled with hydrogen near room temperature. After annealing at 100 °C for 36 h under 1.5 bar H2 atmosphere we obtained H-Cu2O electrodes. The catalyst loadings for Cu2O and H-Cu2O were 11 mg cm−2 and 10.8 mg cm−2, respectively.

Characterization

The morphology of the catalysts was characterized using SEM (JEOL, JSM-6700F) and TEM (JEOL, ARM-200FTH). The high-resolution XPS were measured at TLS beamline BL-24A of the National Synchrotron Radiation Research Center (NSRRC), Taiwan. The XPS measurements were performed under ultrahigh vacuum condition (<10–6 bar). The binding energies of collected spectra were calibrated to Au 4f7/2 of 84 eV for comparison. The XANES measurements were performed at Taiwan beamline BL-12XU in Spring-8, Japan. Athena software was used to process the acquired EXAFS data. The soft-XAS was collected in fluorescence mode at TLS beamline BL-11A of NSRRC, Taiwan.

Electrochemical measurements

All electrochemical measurements were made out at room temperature in a typical three-electrode system with an electrochemical potentiostat (CHI 6278E, CHI Instruments). The counter and reference electrodes were platinum foil and Hg/HgO, respectively. All applied potentials were calibrated to a reversible hydrogen electrode (RHE, ERHE) for comparison, as shown in Equation. (1) ERHE = ERef + 0.059 pH + EExp. (ERHE (V): the potential of reversible hydrogen electrode; ERef (V): the potential of reference electrode; EExp (V): the potential of working electrode). For OER measurements, cathodic linear-sweep voltammetry with scan rate 1 mV s−1 was performed in KOH (1.0 M, pH ~13.6). The 90% iR compensation was applied for OER test by using the automated iR-correction function of the potentiostat. Gas chromatography (GC) was used to determine the Faradaic efficiency.

The ECSA was estimated from the electrochemical double layers of the catalyst surface using Equations. (2) Cdl = iv/(dE/dt) and (3) ECSA = Cdl/CS. (where Cdl is the double-layer capacitance and dE/dt is the scan rate.) The potential windows were measured from 1.1 to 1.2 V versus RHE. The geometric areas of the electrodes were calculated on dividing the general specific capacitances of Cs = 40 µF cm−2 for TM oxide in alkaline solution47.

Operando EIS measurements

The EIS were performed in a frequency range from 10−1 to 104 Hz with a small AC amplitude, 10 mV, under applied potential range from 1 to 2 V versus RHE.

Operando GAXS measurements

The GAXS was performed at beamline BL12-B2 of SPring-8 in Hyogo, Japan. The GAXS patterns were collected with a large Debye–Scherrer camera. The scattering angle was aligned to the Bragg peak of standard CeO2 powder (SRM 674b). The incident grazing angle of the X-ray was set at 1°; the X-ray energy was 15 keV (λ = 0.82656 Å). Operando GAXS measurements were performed in a self-assembled Teflon cell sealed with a Kapton tape window (2 × 2 cm2) that was similar to the three-electrode electrochemical condition. The applied potential on the electrode was measured from 1.0 to 2.0 V versus RHE with CHI Instruments. The incident X-ray beam was transmitted through the Kapton window and the electrolyte to collect the surficial GAXS pattern of the electrode.

Operando quick-XAS measurements

The Cu K-edge XAS were measured at the TPS beamline BL-44A in NSRRC, Taiwan. Operando quick-XAS measurements were performed in the aforementioned self-assembled Teflon cell. The X-ray beam was transmitted through the Kapton tape and electrolyte and reached the detector for XAS collection in the transmission mode.

Operando soft-XAS measurements

The Cu L3-edge XAS measurements were performed at the photoemission end-station at beamline BL-11A in NSRRC, Taiwan. Operando soft-XAS were also recorded in a three-electrode setup with the previous self-assembled cell. A gold-covered Si3N4 window was in contact with copper wires as the working electrode; Hg/HgO and platinum wires were respectively used as reference and counter electrodes. The catalyst powders were dispersed in ethanol with Nafion solution (20 μL, 5%, Sigma-Aldrich), and then sonicated for 10 min. The catalyst ink was drop-cast onto the gold-covered Si3N4 window. The X-ray beam was transmitted through the Si3N4 window and reached the detector for soft-XAS spectra collection in the fluorescence mode.