Unusual double ligand holes as catalytic active sites in LiNiO₂

Abstract

Designing efficient catalyst for the oxygen evolution reaction (OER) is of importance for energy conversion devices. The anionic redox allows formation of O-O bonds and offers higher OER activity than the conventional metal sites. Here, we successfully prepare LiNiO₂ with a dominant 3d⁸L configuration (L is a hole at O 2p) under high oxygen pressure, and achieve a double ligand holes 3d⁸L² under OER since one electron removal occurs at O 2p orbitals for Ni^III oxides. LiNiO₂ exhibits super-efficient OER activity among LiMO₂, RMO₃ (M = transition metal, R = rare earth) and other unary 3d catalysts. Multiple in situ/operando spectroscopies reveal Ni^III→Ni^IV transition together with Li-removal during OER. Our theory indicates that Ni^IV (3d⁸L²) leads to direct O-O coupling between lattice oxygen and *O intermediates accelerating the OER activity. These findings highlight a new way to design the lattice oxygen redox with enough ligand holes created in OER process.

Introduction

The electrocatalysis of the oxygen evolution reaction (OER) is at the core of many energy conversion devices, such as water and CO₂ electrolysers, and metal-air batteries^1,2,3. Because of the sluggish kinetics of OER, the exploration of OER electrocatalysts with high intrinsic activity plays a decisive role in determining the device performance, and thus the fundamental understanding of OER mechanisms is essential^4,5,6,7. The conventional adsorbate evolution mechanism, involving multiple adsorbed intermediates with highly correlated adsorption strengths, exhibits a minimum theoretical overpotential of about 0.37 V^8,9. This limitation has been circumvented by a recently identified mechanism^10,11—the lattice oxygen oxidation mechanism—wherein the oxygen ligands are electrochemically activated, coupled directly with *O intermediates, and released from the lattice matrix⁴. Lattice-oxygen-activated electrocatalysts, such as 3d late-transition-metal (TM) oxides with high metal oxidation states, universally feature a high TM-O covalency with a strong orbital hybridization between TM 3d and O 2p orbitals, which enables intramolecular electron transfer from the oxygen ligands to the TM cations, leaving ligand holes for lattice oxygen activation^4,10. Such an anionic redox mechanism is well documented in lithium-ion-battery (LIB) cathodes¹² and thermal oxidative catalysis¹³.

Ni^III oxides are intrinsically anionic-redox-active. Here we use the notation Ni^III rather than Ni³⁺ as a reminder that the formally trivalent Ni ion is in a highly covalent situation where holes in the oxygen 2p ligands play a significant role. Ni^III oxides such as RNiO₃, RSrNiO₄ (R = rare earth element), and LiNiO₂ are in the negative charge-transfer energy regime (Δ < 0, Fig. 1a), leading to 3d⁸L (L denotes a ligand hole located on a molecular orbital formed by the O 2p orbitals) as the dominant ground state configuration, instead of the conventional 3d⁷ (Δ > 0, Fig. 1b)^{14,15,16,17,18,19}. It is well known that as the Ni^III oxidized to Ni^IV, the electron removal in such charge transfer system should happen in the O 2p band rather than the Ni 3d band²⁰, resulting in a 3d ⁸L² configuration, namely double O 2p hole states (Fig. 1a).

Fig. 1: Schematic single-electron removal spectra of Ni^III oxides with a formal 3d⁷ filling.

a Δ < 0 and b Δ > 0 scenarios. Charge-transfer energy Δ: energy cost for electron transferring from the O 2p band to the Ni 3d band, and U: energy cost for electron transferring from the occupied 3d band to the unoccupied 3d band. The hybridization between Ni 3d and O 2p is neglected for simplicity. The Fermi level (E_F) is shown as a vertical dash black line and the O 2p band is used as the energy reference considering the position of E_F is arbitrary in the band gap.

Based on the lattice oxygen oxidation mechanism, we proposed that a catalyst featuring double ligand hole states would endow high OER activity. To test our prediction, LiNiO₂ is employed as the precatalyst for the double ligand hole states considering that it has an edge-shared Ni-O-Ni network, which is more stable than the corner-shared network in RNiO₃²¹ and RSrNiO₄²². In addition, as a well-known cathode material of LIB, LiNiO₂ can be easily delithiated and oxidized to form Ni^IV ions^23,24, although this oxidizing process is rarely observed during OER in aqueous electrolytes. The stabilization of the highly active Ni^IV has been recently attributed to the high-entropy charge-glass-like state of LiNiO₂¹⁷. The electronic structure of LiNiO₂ is found to have many nearly degenerate states composed of [NiO₆] octahedra with 3d⁸, 3d⁸L and 3d⁸L² configurations¹⁷. The LiNiO₂ system can fluctuate between these states such that the hole states on the oxygen ligands are stabilized by the entropy term of the free energy.

In this study, a pristine Ni²⁺-free LiNiO₂ sample was prepared using high-pressure O₂, and its structure was comprehensively characterized using multiple complementary techniques, including microscopic techniques, X-ray diffraction (XRD), and X-ray absorption spectroscopy (XAS) at the Ni K, Ni L_3,2, and O K edges. The electrocatalytic properties of LiNiO₂ was robustly assessed using rotating ring disk electrodes (RRDE). The delithiation, oxidation and resulting structural transition of LiNiO₂ during OER were probed using operando XRD, XAS and Raman spectroscopies. Combined analysis of the operando results, density function theory (DFT) calculation and in situ differential electrochemical mass spectrometry (DEMS) reveals that the enhanced OER activity of LiNiO₂ is originated from the structurally stabilized Ni^IV with a 3d⁸L² configuration.

Results

Structural characterization of LiNiO₂

The crystalline phase and morphology of the as-prepared LiNiO₂ were characterized using synchrotron X-ray diffraction (XRD) and transmission electron microscopy (TEM) (Fig. 2). The XRD pattern of LiNiO₂ is well-matched with the standard pattern of R-3m LiNiO₂ (JCPDS No. 09-0063) without any noticeable crystalline impurities (Fig. 2a and Supplementary Fig. 1). The sharp diffraction peaks, and the well separation of (006) and (10-2) and of (10-8) and (110), indicate a well-crystalized layered structure with a low degree of Li⁺/Ni²⁺ intermixing (Supplementary Fig. 1)^23,25. Rietveld refinement was carried out to obtain the crystal structural parameters of the as-prepared LiNiO₂ (Fig. 2a), yielding a good fit for the R-3m structure and ~3.9% Ni occupancy in the [LiO₂] layers, demonstrating the high-purity of the sample. The single crystalline nature of the as-prepared LiNiO₂ is suggested by TEM techniques (Fig. 2b–d). A typical selected-area electron diffraction (SAED) pattern of LiNiO₂ shows clear diffraction spots that can be indexed to the R-3m structure of LiNiO₂ along the [010] zone axis (Fig. 2b). High-resolution TEM (HRTEM) images of the thin parts of LiNiO₂ display uniform lattice fringes of (003) and (104) of the R-3m structure, respectively (Fig. 2c and Supplementary Fig. 2). Aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images were also recorded to directly visualize the Ni atomic arrangement along the edges, which well matches the (003) lattice plane of R-3m LiNiO₂ (Fig. 2d). TEM, scanning electron microscopy (SEM) and atomic force microscopy (AFM) suggest that the general morphology of the as-prepared LiNiO₂ is nanoflakes with 200 − 400 nm in the lateral size and ~10 nm in the layer thickness (Supplementary Fig. 3 and Fig. 2e). LiNiO₂-raw, a LiNiO₂ sample prepared without the second annealing step under high-pressure O₂, shares a similar morphology with the LiNiO₂ (Supplementary Fig. 4).

Fig. 2: Structural characterization of the as-prepared LiNiO₂.

a A synchrotron XRD pattern (18 keV incident beam energy, λ = ~0.6888 Å) and a Rietveld refinement fit with the refined structural parameters listed in Supplementary Table 1, along with the structural model of R-3m LiNiO₂ which consists of alternate [LiO₂] and [NiO₂] layers formed by edge-sharing [LiO₆] or [NiO₆] units. A set of extra diffraction peaks are from graphitic carbon of carbon paper and two broad diffraction peaks at 12° and 18° are from the sample holder. b A typical SAED pattern of LiNiO₂ along a [010] zone axis of the R-3m structure, c a HRTEM image and its fast Fourier transform pattern, showing the (003) lattice fringes, and d an AC-HAADF-STEM image projected perpendicular to the (003) lattice plane. e An AFM image and the corresponding thickness profile across the horizontal red line.

The local coordination environment of transition metal ions can be obtained by extended X-ray absorption fine structure (EXAFS) spectra at the corresponding K edges^26,27. The Ni local coordination of LiNiO₂ resemble that of NiO and Ni(OH)₂, with two dominant peaks at ~1.8 and ~2.8 Å, which are attributed to Ni–O and Ni–Ni scattering paths within and between the edge-sharing [NiO₆] units, respectively (Fig. 3a and Supplementary Fig. 5). For these peaks, LiNiO₂ shows a shorter radial distance than those of Ni(OH)₂ and NiO, supporting the presence of a shorter Ni-O bond length due to the higher formal oxidation state of Ni.

Fig. 3: XAS characterisation of the as-prepared LiNiO₂.

a Fourier transforms of k³-weigted EXAFS spectra of LiNiO₂, LiCoO₂ (R-3m) and NiO, and b XANES spectra of LiNiO₂ and Ni references (NiO, Ni(OH)₂, and KNiIO₆) at the Ni K edge. The amplitude of LiNiO₂ EXAFS spectra is adjusted for a clear comparison, and the Fourier transforms were phase-corrected. TEY-XAS spectra of LiNiO₂ c at the Ni L₃ edge and d at the O K edge, along with NiO and γ-NiOOH as the Ni and O references.

From an ionic picture, low spin Ni^III ions adopt a (t_2g)⁶(e_g)¹ configuration in an octahedral crystal field and are Jahn-Teller (JT)-active, so that one can expect a strong local distortion of the [NiO₆] octahedra to remove the degeneracy of the e_g orbitals. There are local distortions of the [Ni^IIIO₆] units here, as evidenced by the broader first coordination shell at 1.7 A and a larger disorder factor (σ²) of LiNiO₂ by fitting R-space EXAFS spectra, as compared with that of the analogous R-3m LiCoO₂ (Co^III, (t_2g)⁶(e_g)⁰, Fig. 3a, Supplementary Fig. 6, and Table 1). Two distortion mechanisms are proposed. One is the conventional local JT distortion, as suggested by previous EXAFS studies of LiNiO₂^28,29, with four short Ni-O (~1.920 Å) and two long Ni-O (~2.069 Å) bonds in the [Ni^IIIO₆] octahedra (Table 1 and Supplementary Fig. 7a). However, pure JT distortion fails to explain the neutron-pair distribution function and electronic structure of LiNiO₂; therefore, a [NiO₆]-size disproportionation (SD) mechanism was proposed, where the number of Ni−O_short and Ni−O_long is equal¹⁷ (Table 1 and Supplementary Fig. 7b). Based on these two possible structural models, EXAFS fitting using two Ni−O coordination shells was performed. Both models gave comparable but better fits of the EXAFS data than that using only one Ni–O scattering path (Table 1 and Supplementary Fig. 7c), supporting the presence of distorted [Ni^IIIO₆] in pristine LiNiO₂. To distinguish JT distortion from SD, double cluster simulation would be needed to simulate the Ni-L_2,3 XAS spectrum of LiNiO₂¹⁶, which is beyond the scope of this study.

Table 1 Structural parameters of LiNiO₂, LiCoO₂, and NiO, obtained from EXAFS fitting^a

The oxidation state of the as-prepared LiNiO₂ was assessed by X-ray absorption near edge structure (XANES) spectroscopy (Fig. 3b). At the Ni K edge, the energy position of LiNiO₂ locates in-between those of Ni²⁺ (NiO and Ni(OH)₂) and Ni^IV references (KNiIO₆), suggesting that the Ni of the LiNiO₂ is in the intermediate oxidation state between Ni²⁺ and Ni^IV. Assuming that the formal oxidation state of Ni is proportional to the energy at 0.7 edge jump height, the average Ni oxidation state of LiNiO₂ can be quantified as ~3.0 (Supplementary Fig. 9), in a good agreement with the nominal value.

The detailed electronic structure at the surface region can be obtained by total electron yield soft X-ray absorption spectra (TEY-sXAS) at the L₃ edges. The multiplet spectral feature and the energy position at the L₃ edges of 3d elements are highly sensitive valence state and local environments^30,31,32,33. Figure 3c shows the Ni L₃ edge sXAS of LiNiO₂, along with those of NiO and γ-NiOOH as the Ni²⁺ and Ni^III references, respectively. Both the multiplet spectral features and energy positions of the Ni-L₃ edge are very similar to those of γ-NiOOH, indicating that the Ni of LiNiO₂ is in the formal +3 oxidation state. The dominant peak for both γ-NiOOH and LiNiO₂ locates at ~855 eV, much higher than that of NiO at ~853 eV, and further inspection reveals that the low-energy peak of LiNiO₂ locates ~0.3 eV higher than the dominant peak of NiO in Fig. 3c. The low-energy peak is an intrinsic spectral feature of Ni^III as found in theoretical and experimental XAS^15,16 and high-quality single crystal LaNiO₃ and NdSrNiO₄^21,22 In addition, Ni^III (3d⁸L) in our LiNiO₂ is supported by theoretical simulation of the Ni L_3,2 XAS in Supplementary Fig. 10a, where the LiNiO₂ spectra can be nicely reproduced by the full multiplet cluster calculation with a nominal 3d⁸L configuration. Our LiNiO₂ is free from Ni²⁺ impurity that often observed in the literature (Supplementary Fig. 10b).

The Ni valence state and Ni-O covalency were further studied by the O-K sXAS. The pre-edge peak below 533 eV can be assigned to the unoccupied 3d e_g hybridised with O 2p orbitals. An increase in the valence state of the 3d element correlates with a shift of the pre-edge to lower energy and an enhancement of the spectral intensity due to the strong covalence^21,34,35. The O-K sXAS of LiNiO₂ is dominated by a strong peak at 528.2 eV without obvious contribution of a peak at 531.9 eV found from NiO (Fig. 3d). Thus, TEY-sXAS at both the Ni L₃ edges and the O K edge suggest that the surface Ni oxidation state of LiNiO₂ is Ni^III and free from Ni²⁺ impurities.

Electrochemical measurements

The OER activity of LiNiO₂ was studied using rotating ring disk electrodes (RRDE, disk: glassy carbon, ring: Pt) at 1600 rpm. The potential of the Pt ring was held at 0.42 V throughout the measurements, so that O₂ once generated on the disk can be collected and electrochemically reduced on the ring (Fig. 4a and Supplementary Fig. 11), which enables the determination of OER onset potentials. The OER activity of LiNiO₂ was estimated using linear sweep voltammetry (LSV) in Ar-saturated and Fe-free 1 M KOH and compared with the benchmark IrO₂ (99.9%) catalyst and a NiOOH reference. The LiNiO₂ shows the best OER geometric activity (Fig. 4b, current density normalized by disk electrode area) and intrinsic activity (Supplementary Fig. 13, current density normalized by electrochemical surface area), with the potential to achieve 10 mA cm_geo⁻² (E@10 mA cm_geo⁻², ~1.545 V) lower than those of IrO₂ (~1.558 V) and NiOOH (~1.753 V). In addition, LiNiO₂ shares a Tafel slope (53 mV dec⁻¹) similar to IrO₂ (54 mV dec⁻¹) and distinguishes itself from conventional NiOOH (96 mV dec⁻¹, Fig. 4c), indicating rapid OER kinetics of LiNiO₂. Compared to the reported overpotential values of LiMO₂ (M = Co and Fe), and RMO₃ (R = Rare earth elements, M = Fe, Co and Ni) and other unary 3d transition metal catalysts, our LiNiO₂ requires much less overpotential to achieve certain current density (Fig. 4d), demonstrating the unique electrochemical properties to facilitate OER. Amongst the previously reported LiNiO₂ and even Fe-doped LiNiO₂, the as-prepared sample exhibit superior E@10 mA cm⁻² and Tafel slopes, highlighting the excellent activity of pristine LiNiO₂ (Supplementary Table 2).

Fig. 4: Electrochemical measurements and surface chemistry of LiNiO₂.

a Schematic illustration of a rotating ring disk electrode (RRDE, ring: Pt, disk: glassy carbon). b Linear sweep voltammograms and c Tafel plots of LiNiO₂, NiO-derived NiOOH, IrO₂ (99.9% metal basis, Adamas-beta) and carbon (Vulcan XC-72R) in Ar-saturated 1 M KOH solution, along with the corresponding Pt ring current shown in the inset (the onset potential of OER indicated by arrows). d OER activity comparison of the as-prepared LiNiO₂ to LiMO₂ (M = Co and Fe), and RMO₃ (R = Rare earth elements, M = Fe, Co and Ni) and other unary 3d transition metal catalysts (Supplementary Table 3). e Stability test of LiNiO₂ in 1 M KOH at 10 mA cm⁻² (on carbon paper) and (the inset of e) linear sweep voltammograms of LiNiO₂ before and after the stability test. f Window-opening cyclic voltammograms of LiNiO₂ in Fe-free 1 M KOH with stepwise decreasing the lower potential limits, showing the evolution of redox peaks (magnified in the corresponding insets) and the deterioration of OER current. The voltammograms were collected under 1600 rpm with a sweep rate of 5 mV s⁻¹, and IR drops of the voltammograms were determined by EIS prior to the measurements and compensated. g Operando Raman spectra of LiNiO₂ at 1.56 V with and without being cycled through Ni²⁺/Ni^III redox.

The stoichiometric structure of LiNiO₂ is essential to achieve high OER activity. LiNiO₂-raw was also characterized and compared with the pristine LiNiO₂, as the high-pressure or high-flow-rate O₂ during synthesis has been found to be essential for stabilizing Ni^III and minimizing oxygen vacancies in LiNiO₂ and LiNiO₂-based cathode materials^36,37. The LiNiO₂-raw shares similar morphology with the LiNiO₂ (Supplementary Fig. 4) but has ~4.6% oxygen vacancies (vs. ~0.4% in LiNiO₂) as characterized by XRD refinement (Supplementary Fig. 14 and Supplementary Table 4), and a significant fraction of Ni²⁺ impurities as shown in Ni-K XAS (Supplementary Fig. 15) and surface sensitive TEY-sXAS at the Ni L₃ and O K edges (Supplementary Fig. 16 and 17). We found that the off-stoichiometric structure of LiNiO₂-raw yields reduced OER activity (Supplementary Fig. 18).

From the ring current of the RRDE, the OER onset potentials of LiNiO₂, IrO₂ and NiOOH were estimated as ~1.437, ~1.438 and ~1.495 V, respectively (Fig. 4b and Supplementary Fig. 11), indicating that LiNiO₂ activates the oxidation of water at a much lower overpotential than NiOOH. The OER activity of NiOOH is consistent with a common understanding that Ni (oxy)hydroxides without Fe doping and without the presence of Fe in the electrolyte are poor OER electrocatalysts^38,39,40. Thus, from perspectives of overpotentials, Tafel slopes and onset potentials, the electrocatalytic properties of LiNiO₂ for OER is as good as the state-of-the-art IrO₂ and outperforms the standard NiOOH. The stability of LiNiO₂ was measured using chronopotentiometry at 10 mA cm⁻² (Fig. 4e). LiNiO₂ largely maintained its initial potential, which gradually increased by ~10 mV for the first 18 h, and then remained constant for the subsequent 30 h. The high stability of LiNiO₂ is also indicated by the little activity deterioration in LSV before and after the 48 h stability test (the inset of Fig. 4e). One may be interested in the stability of LiNiO₂ as an OER electrocatalyst, as LiNiO₂ is known to suffer from severe capacity degradation as a cathode material for LIB. The fast capacity fading is generally observed in high voltage region >4.1 V vs. Li⁺/Li, resulting from a large content of delithiation in Li_xNiO₂ (x < 0.3) and in turn structural collapse^23,41,42. If the highest cut-off voltage would be limited to 4.1 V vs. Li⁺/Li, the capacity of LiNiO₂ could retain >95% of its initial capacity after >100 cycles^41,43. However, operation potentials for our OER are below 1.8 V vs. RHE which corresponds to only 4.03 V vs. Li⁺/Li in battery operation.

The interconversion of Ni²⁺-based (hydr)oxides to NiOOH is usually indicated by Ni²⁺/Ni^III redox peaks in the potential window of 1.2–1.4 V (vs. RHE); therefore, the formation of NiOOH from LiNiO₂ was investigated using window-opening cyclic voltammetry. As shown in Fig. 4f, the lower potential limit decreased stepwise from 1.41 V (used in Figs. 4a) to 1.11 V, close to the values generally used in previous studies^{44,45,46,47,48,49}. When the lower potential limit was set to 1.41 V, the OER current is optimal and remains stable over multiple CV cycles. However, decreasing the limit to ~1.21 and ~1.11 V deteriorates the OER current significantly, accompanied by the appearance and gradual increase of redox peaks at ~1.36 (anodic) and ~1.26 V (cathodic), indicating the formation of redox-active but OER-inactive phases on the surface. A LiNiO₂ sample, cycled through the Ni²⁺/Ni^III redox, was characterized using operando Raman spectroscopy at 1.56 V (Fig. 4g). Its Raman spectrum shows clear differences in the position and relative intensity of the Raman bands and a considerable increase in the overall Raman scattering intensity, as compared to pristine LiNiO₂ at the same potential (for more details, see the operando Raman section below). All these spectral features can be indexed to β-NiOOH⁵⁰, indicating the structural transformation of OER-inactive NiOOH species from LiNiO₂ after being cycled through Ni²⁺/Ni^III redox. Window-opening voltammetry was also carried out in unpurified 1 M KOH (Supplementary Fig. 19), wherein, despite the appearance of the Ni²⁺/Ni^III redox peaks in the close potential range, the OER current remains stable even when the lower potential limit reached ~1.11 V. This indicates that Fe impurities significantly improve the performance of NiOOH, in agreement with the common understanding of Ni-based OER electrocatlaysts^38,39,40. Therefore, the voltammetric analysis shows that the OER activity and electrochemical properties of LiNiO₂ are impacted by the lower potential limits and presence of Fe impurities in the electrolyte. When undergoing the Ni²⁺/Ni^III redox, the surface structure of LiNiO₂ transforms and degrades into NiOOH, which has a poor OER activity in the absence of Fe. In contrast, the pristine and structurally unaltered LiNiO₂, whose real active species has been covered and neglected in previous LiNiO₂ OER studies^{44,45,47,48,49}, exhibits better intrinsic OER activity.

Operando XRD, XAS, and Raman studies

From the aforementioned structural and electrochemical measurements, the pristine nature of LiNiO₂ has been characterised in terms of crystalline phase, local coordination environment, oxidation state and electronic structure, and the superior OER activity of LiNiO₂ to the analogous NiOOH is confirmed and studied in detail. However, the active species of LiNiO₂, which can no longer be attributed to NiOOH, remains unclear. In addition, LiNiO₂, a typical cathode material for lithium-ion batteries, can be delithiated to increase the Ni oxidation state and induce structural transition;^24,46 therefore, delithiation is expected to become comparable or even easier during the OER in KOH aqueous solution. To establish the relationship between the structure of LiNiO₂ and its OER activity, operando synchrotron-based XRD and XAS, and Raman spectroscopy experiments were carried out.

Figure 5a shows the operando synchrotron XRD of LiNiO₂ in 1 M KOH in the potential range of 1.41–1.81 V. The patterns consist of three parts: an amorphous background from cell body and window materials (~12° and 18°), diffraction peaks of graphitic carbon from carbon paper (indicated by arrows), and diffraction patterns of LiNiO₂ and its delithiated derivatives. Using the carbon peak at ~11.8° as the internal reference for intensity and peak position (Supplementary Fig. 20), the overall LiNiO₂ pattern largely remains intact under the applied potentials, and another set of pattern emerges with potentials in around 2θ values at the expense of LiNiO₂ pattern intensity when the applied potential is greater than 1.46 V, coinciding with the onset of OER. Whilst (00 l) diffraction peaks of LiNiO₂ show shoulders at lower 2θ values, the (h0l) peaks show shoulders at higher 2θ values, which suggests the formation of another hexagonal phase from the parent LiNiO₂ with interlayer expansion and intralayer contraction for the [NiO₂] layers²⁴. Based on the peak position, the new phase can be indexed to layered Li_xNiO₂ (x = 0.45–0.65, Supplementary Fig. 21)⁵¹, and other Li_xNiO₂ polymorphs and NiOOH phases were not found. The LiNiO₂ delithiation during OER is also supported by increased Li concentration of KOH electrolyte and Ni/Li molar ratio of the sample (Supplementary Table 5). Considering that both the delithiation and OER initiate from the surface, the delithiated Li_xNiO₂ can be considered as the active species of OER.

Fig. 5: LiNiO₂ delithiation under OER conditions.

a Operando XRD patterns and b Operando Ni K edge XANES spectra of LiNiO₂ under applied potentials of 1.41–1.81 V in 1 M KOH solution. Diffraction peaks in low 2θ ranges and the pre-edge peak were magnified in the corresponding insets. A set of extra peaks are from graphitic carbon and broad diffraction peaks at 12° and 18° from the sample holder.

The oxidation of LiNiO₂ accompanying with delithiation was tracked using operando XANES measurements at the Ni K edge. No discernible change in XANES spectra was observed when the electrode was immersed in 1 M KOH (Supplementary Fig. 22). Figure 5b shows the XANES spectra in the potential range of 1.41–1.81 V. The rising edge shifts towards higher energy with the applied potentials, indicating that the oxidation of Ni^III coincides with the onset of OER. Using the relationship between the energy at 0.7 edge jump height and the Ni formal oxidation state obtained from ex-situ data (Supplementary Fig. 9), the average Ni oxidation state of LiNiO₂ was plotted as a function of the applied potentials (Fig. 6c). The oxidation state is ~3.0 at 1.41 V, sharply increases to ~3.3 at 1.56 V, gradually at potentials >1.56 V and reaches a plateau of ~3.4. In addition, the pre-edge peak intensity was found to increase with applied potentials (the inset of Fig. 5b), corresponding to the decreased symmetry of [NiO₆] octahedra (see EXAFS analysis below). Combining the potential-dependent oxidation state of Ni with the electrochemistry of LiNiO₂ reveals a positive correlation between the number of Ni^IV ions and OER current (Fig. 6c), suggesting that Ni^IV in delithiated Li_xNiO₂ can be considered as (part of) the OER active centre.

Fig. 6: Local coordination of LiNiO₂ under OER conditions.

a, b Fourier transforms of operando Ni K edge EXAFS spectra of LiNiO₂ under applied potentials of 1.41–1.81 V in 1 M KOH solution, with the Ni–O region and the Ni–Ni region magnified to better show the potential-dependent behaviour. The Fourier transformation was carried out over a k range of 2.5 − 12.4 Å⁻¹, and the Fourier transforms were phase-corrected. c Average oxidation state, N(Ni–O) and R(Ni–O) of LiNiO₂ as a function of applied potentials. The error bars represent the uncertainty of the fitted parameters. d Operando Raman spectra of LiNiO₂, showing the E_g and A_1g vibration bands, where oxygen atoms of [NiO₆] units move oppositely along adjacent O-layers and symmetrically along the c-axis, respectively.

The dominance of Ni^IV with double ligand hole state (3d⁸L²) is supported by simulation of the experimental sXAS at the Ni-L_3,2 edge of LiNiO₂ after OER taken with surface sensitive TEY mode (black circles) using the full multiplet cluster calculation which includes full intra-atomic multiplet interaction, crystal field interaction and covalence. We found the experimental Ni-L_3,2 sXAS of LiNiO₂ after OER can be well reproduced by the sum (black line) of 75% 3d⁸L² (blue line) and 25% 3d⁸L (red line) configurations (Supplementary Fig. 23b) giving rise to Ni^3.75+ valence state.

Potential-dependent changes of LiNiO₂-raw in XANES are much weaker than those of pure LiNiO₂ (Supplementary Fig. 24). The average oxidation state of Ni of LiNiO₂-raw grows much slower with applied potentials than that of LiNiO₂, and reaches only ~3.2 at 1.81 V. The impeded Ni oxidation in LiNiO₂-raw implies that the formation of Ni^IV and double ligand hole state in LiNiO₂-raw is limited. For the LiNiO₂-raw sample, the XRD patterns show increasing fraction of NiOOH phase with applied potentials, while the R-3m structures is largely maintained in LiNiO₂ without discernible formation of any NiOOH phases (Supplementary Fig. 25). It should be noted that the nanoscale thickness of our LiNiO₂ is a prerequisite to observe significant spectral changes in operando XRD and operando XAS, because the OER usually occurs within several nanometres from the surface. Thus, our results indicates that observation of Ni^IV with double ligand holes state is directly related to both our pure and nano-sheet pristine LiNiO₂ sample.

Potential-dependent changes in the local coordination environment of Ni were probed using operando EXAFS (Fig. 6a, b). Similar to the ex-situ data, the Ni coordination shell is dominated by Ni–O and Ni–Ni scattering paths. The amplitude of Ni–O at ~1.45 Å increases with the applied potentials from 1.41 to 1.81 V (Fig. 6b). Because the Ni first-shell coordination in pristine LiNiO₂ is saturated with six oxygen atoms, the coordination numbers (CN) of Ni-O cannot increase further, which alternatively implies the Ni-O bonds become increasingly ordered with the applied potentials. The lowered Ni-O distortion is also suggested by the decreased Ni-O peak border at the high R with increased amplitude (Fig. 6b) and by the increased frequency of Ni-O oscillation in the real part of the Fourier transforms (Supplementary Fig. 26). Quantitative structural information was extracted via EXAFS fitting (Supplementary Fig. 27 and Supplementary Table 6), where the parameter σ²(Ni–O) represents the distortion in the [NiO₆] octahedra. The value of σ²(Ni–O) decreases as a function of the applied potential from 0.012 Å⁻² at 1.41 V to 0.008 Å⁻² at 1.81 V, supporting that the LiNiO₂ evolves from the oxidation of SD-active Ni^III atoms (t_2g⁶e_g¹) to SD-inactive Ni^IV atoms (t_2g⁶) under OER. In addition, the CN(Ni–O) decreases from 5.8 ± 0.6 at 1.46 V to 5.0 ± 0.4 at 1.81 V, indicating the formation of oxygen vacancies during OER^52,53,54, which have been calculated to greatly reduce the OER overpotential by enhancing intermediate adsorption on the Ni sites^53,54. In addition, the increase in the relative spectral intensity of the weak pre-edge peak at 8335.4 eV in the Ni-K XANES as a function of applied potentials (the inset of Fig. 5b) implies a reduced local symmetry of Ni ion, which generally suggests a decrease in CN⁵³. Moreover, our DEMS measurements and DFT calculation indicate that lattice oxygen involves in OER process (see below). In this lattice oxygen mechanism, oxygen vacancies should be created.

In addition to the amplitude, the contour maps of Ni–O shows a gradually increasing contribution in the corresponding lower R regions (Fig. 6b), corresponding to a shift toward lower R in the real part of FT-EXAFS (Supplementary Fig. 26). A shrinking Ni–O bond length from 1.955 Å at 1.41 V to 1.909 Å at 1.81 V was determined via EXAFS fitting (Fig. 6c, Supplementary Fig. 27 and Supplementary Table 6), supporting the presence of a high valence state of Ni^IV under OER conditions. To summarise, by combining the fast operando Ni-K XANES and EXAFS spectra with the OER onset potential, Ni^IV and oxygen vacancies in delithiated Li_xNiO₂ were identified during OER, and attributed to the high OER activity.

Operando Raman spectroscopy was further employed to track the intermediate species of LiNiO₂ under the high applied potentials larger than 1.4 V. Different from the above Raman experiments in Fig. 4g, the LiNiO₂ now was not forced to undergo Ni²⁺/Ni^III redox, but directly applied to 1.46 V. The spectral profile of LiNiO₂ Raman in Fig. 6d only exhibits slightly red-shifted bands (E_g: from 472 cm⁻¹ to 462 cm⁻¹; A_1g: from 546 cm⁻¹ to 544 cm⁻¹) with increasing applied potentials during the OER process, in agreement with the formation of delithiated LiNiO₂⁵⁵. This supports the results from the above operando XRD and XAS. From comparison between Fig. 6d and Fig. 4g we can conclude that the crystal structure, intermediates and the OER activity (Fig. 4f) of LiNiO₂ strongly depend on the lowest potential applied.

The presence and formation of γ-NiOOH have been regarded as the origin of OER activity in many Ni-based oxides^46,53,56. However, according to our operando Raman spectra (Fig. 6d), the active species of LiNiO₂ could not be assigned to γ-NiOOH as per the band position and relative intensity (for LiNiO₂, A_1g: ~546 cm⁻¹ and I(E_g)/I(A_1g)=0.5; for γ-NiOOH, A_1g: 558 − 560 cm⁻¹ and I(E_g)/I(A_1g) = 1.5 − 1.7)^40,57,58,59. In addition, a large expansion in the interlayer spacing between [NiO₂] layers, caused by the transformation from LiNiO₂ (4.72 Å) to γ-NiOOH (≥7 Å), was not observed in our operando XRD data. In addition, post-mortem XRD and XAS indicate that the structure of Li_xNiO₂ remains mainly intact after OER (Supplementary Fig. 29 and 30). Moreover, the surface and bulk AC-STEM images were employed to study the structural transformation of LiNiO₂ after OER^60,61 (Supplementary Fig. 31), and both show the atomic arrangement perpendicular to the (018) lattice plane of LiNiO₂, supporting the surface structure of LiNiO₂ remains mainly intact after OER. Thus, using operando X-ray diffraction/absorption and Raman spectroscopies and post-mortem characterizations, the active species were identified as the delithiated yet structurally intact LiNiO₂ with oxygen vacancies.

DFT calculation and DEMS for OER mechanisms

First-principles DFT calculations were employed to investigate the underlying mechanism of the enhanced OER activity of delithiated LiNiO₂ with respect to the reaction steps. Similar to our previous work^11,62,63, three scenarios were considered for the OER mechanisms involving four proton-electron transfer steps: metal-site adsorbate evolution (MAE) mechanism, lattice-oxygen-vacancy-site (LOV) mechanism, and metal-and lattice-oxygen-vacancy-site (MLOV) mechanism. In these mechanisms, the adsorbate sites of the reaction steps are different: only the Ni sites in MAE, oxygen sites in LOV, and both in MLOV (Fig. 7a). We built models for perfect LiNiO₂ and delithiated Li_xNiO₂ to investigate the effects of the unsaturated Ni^IV state generated during OER. The (102) plane of LiNiO₂ with a 50% lithium vacancy was used to model the surface reaction pathways on delithiated Li_xNiO₂ (x = 0.5, Supplementary Fig. 32), as the delithiation and oxidation of LiNiO₂ and the concomitant OER all start from the edge^25,52.

Fig. 7: The OER mechanism of LiNiO₂.

a Schematic 4-step reaction pathways of MAE, LOV, and MLOV OER mechanisms. Free energy of each OER step of b Li_0.5NiO₂ and LiNiO₂ at U_RHE = 1.23 V. The estimated thermodynamic overpotentials of each potential-determining step are labelled. c MSCV and faradic CV curves recorded in ¹⁸O-enriched 0.1 M KOH electrolyte prepared using H₂¹⁸O (99% ¹⁸O), showing the mass ion current (i_MS) of ¹⁸O₂, ¹⁶O¹⁸O and ¹⁶O₂ along with the faradaic current (i_F) and the applied potential E from the first CV scan. (The inset of c) Comparison of the measured i_MS(¹⁶O¹⁸O) curve and the theoretically expected i_MS(¹⁶O¹⁸O) curve purely from electrolyte oxidation, based on the ¹⁶O abundance of 1.18% in the electrolyte (Supplementary Equation S27). d The evolution of measured ¹⁸O abundance of evolved oxygen (from total DEMS charge) for ¹⁸O-labled LiNiO₂ with CV cycles in H₂¹⁶O-based 0.1 M KOH electrolyte. The expected ¹⁸O fractions are based on the natural isotope abundance and on the measured value from DEMS experiment of unlabelled LiNiO₂ in the ¹⁶O-based electrolyte.

The Gibbs free energy differences (ΔG) of each OER step via the three OER mechanisms for Li_0.5NiO₂ and LiNiO₂ are shown in Fig. 7b, respectively. First, regardless of the OER pathway, the potential-limiting step of Li_0.5NiO₂ possess lower overpotential than that of LiNiO₂, in agreement with the operando measurement results indicating delithiated Li_xNiO₂ as the actual active phase for OER. Second, among the three OER scenarios for Li_0.5NiO₂ the MLOV route shows the lowest overpotential (0.69 V), suggesting that the OER via MLOV prevails on the delithiated Li_0.5NiO₂, and that the potential-limiting step is the adsorption of *OH (step 3). Third, for the OER on LiNiO₂, the MLOV route also prevails, but the potential-limiting step switches to the formation of *OO intermediates and oxygen vacancies (step 1).

The transition of the potential-limiting step of LiNiO₂ before and after delithiation can be attributed to the double ligand hole state induced by the presence of Ni^IV ions in Li_0.5NiO₂. In the electronic structure of octahedral-symmetric [NiO₆] system, the charge-transfer energy from the O 2p to Ni 3d orbitals decreases with the increased valence state of the Ni ion^4,20,52. The LiNiO₂ system is already in the negative charge-transfer regime, and consequently, the Ni-O interaction is highly covalent^14,17, in which Ni^III adopts an average configuration close to 3d⁸L (rather than the formal 3d⁷), leaving a ligand hole residing on the O 2p orbitals. The covalency increases further when the Ni^III is (partly) oxidised to Ni^IV via LiNiO₂ delithiation, and the resulting Ni^IV was endowed with an average configuration of 3d⁸L², that is, a double ligand hole state in O 2p. The high unoccupied density state with the O 2p ligand character in Li_xNiO₂ activates lattice oxygen, favouring direct coupling with oxygenated intermediates and the formation of oxygen vacancies (i.e. step 1 of the MLOV route), which are the potential-limiting step of LiNiO₂. The prediction of oxygen vacancies formation during OER is in good agreement with the experimental observation of operando XAS (Fig. 6c). Thus, density-functional theory indicates that the double ligand holes created under OER condition easily activates lattice oxygen, accelerating the formation of O-O binding and oxygen vacancies, and the presence of double ligand hole states is closely correlated to the high OER activity of the delithiated Li_xNiO₂.

To confirm the participation of lattice oxygen during OER, we performed ¹⁸O-isotope labelling experiments using in situ differential electrochemical mass spectrometry (DEMS). The experiments were first carried out on unlabelled LiNiO₂ in an ¹⁸O-enriched KOH electrolyte (prepared using 99% H₂¹⁸O) over six consecutive CV cycles from 1.4 V to 1.8 V. Figure 7c shows the first CV and simultaneous mass spectrometric cyclic voltammogram (MSCV) in the time domain. In the MSCV curves, the mass signal m/z = 34 of ¹⁶O¹⁸O, which is mainly originated from the coupling of ¹⁶O of the LiNiO₂ lattice and ¹⁸O of the electrolyte, shows significant ion current, with ³⁶O₂ as the dominant molecular oxygen isotope and ³²O₂ within the noise level. The measured ¹⁶O¹⁸O mass signal was also compared with the theoretically expected profile of the oxidation of the ¹⁸O-enriched electrolyte (1.18% ¹⁶O abundance, the inset of Fig. 7c), revealing a significant excess of the ¹⁶O¹⁸O ion charge. This evidences that the lattice oxygen plays a direct role in forming O₂ during OER^4,10,64. Furthermore, after the lattice ¹⁶O was partially replaced by ¹⁸O from the electrolyte, the resulting ¹⁸O-labelled LiNiO₂ was cycled in a H₂¹⁶O-based KOH electrolyte for another six consecutive cycles to study the reverse oxygen isotope replacement (Supplementary Fig. 33). The first-cycle data reveal a nearly two-fold ¹⁸O abundance than the expected ¹⁸O fractions from the natural isotope abundance and from the measured value of unlabelled LiNiO₂ in the same O¹⁶-based electrolyte (Supplementary Fig. 34 and Supplementary Equation S28). The ¹⁸O isotope excess gradually decreases with the number of cycles Fig. 7d), indicating the consumption of ¹⁸O on the labelled LiNiO₂ in forming O₂. Therefore, the two-way isotope labelling experiments confirm the participation of LiNiO₂ lattice oxygen in the OER, supporting the predicted MLOV mechanism and activated lattice-oxygen in the presence of double ligand holes.

Fe³⁺ impurity effects

It is widely accepted that the incorporation of Fe³⁺ into Ni (oxy)hydroxides promote OER activity by forming synergetic active centres of oxygen-bridged Ni and Fe sites^56,65, and the activity can be further enhanced via structural ordering, as recently proposed by Lee et al.⁵⁸. Structurally intact Li_xNiO₂ with stabilised double ligand holes could be a better platform for forming such synergetic centres than the usually amorphous NiOOH.

Figure 8a compares the linear sweep voltammograms of LiNiO₂ titrated with Fe³⁺ in purified 1 M KOH. Similar to the reported Ni (oxy)hydroxides, the OER activity of LiNiO₂ is strongly correlated with the Fe³⁺ concentration in terms of E(@10 mA cm⁻²) and Tafel slopes (Fig. 8a–c): (1) E(@10 mA cm⁻²) decreases with the Fe³⁺ concentration and saturated at 1.500 V in the presence of 100 µM Fe³⁺ (45 mV lower than that in the Fe-free electrolyte, Figs. 8a), and (2) the Tafel slope plummets from 57 to 43 mV dec⁻¹ in the presence of only 1 µM Fe³⁺, and largely remains constant up to a Fe³⁺ concentration of 200 µM (Fig. 8b). The decreased Tafel slope at such a low Fe³⁺ concentration indicates that OER is switched to a Fe-involving electrocatalytic site with higher activity, and the number of such new active sites increases with Fe³⁺ concentration, resulting in Fe-concentration-dependent E(@10 mA cm⁻²). For comparison, NiOOH was also measured in 1 M KOH containing 100 µM Fe³⁺. Although the overall activity is highly promoted as expected (Supplementary Fig. 35), NiOOH-Fe³⁺ shows 36 mV higher than LiNiO₂-Fe³⁺ in E(@10 mA cm⁻²) and 25 mV in OER onset potential (Fig. 8d), proving LiNiO₂ an excellent platform for forming Ni-Fe active sites. In addition, the presence of high-valent Ni in Ni-Fe active sites is also found to promote electrocatalytic durability³⁰. The promotional mechanism and underlying role of double ligand holes in Ni-Fe lithiated oxide systems are also of interest and related studies are currently underway.

Fig. 8: Promotional effects of Fe³⁺ on the OER activity of LiNiO₂.

a LSVs and b Tafel plots of LiNiO₂ in Fe-free 1 M KOH with the addition of Fe³⁺ solutions (prepared from ultrapure water and Fe(NO₃)₃ ∙ 9H₂O, 99.999%-Fe, PURATREM, Strem). c Potentials at 10 mA cm⁻² and Tafel slopes as a function of Fe³⁺ concentrations. d Comparison of the Fe³⁺ promotional effects on LiNiO₂ and NiOOH. Note that 100 µM Fe³⁺ reaches the saturated concentration of Fe(OH)₄^- (the predominant Fe³⁺ species in alkaline solution)⁷⁰.

In conclusion, a comprehensive picture correlating electrochemical properties and operando spectroscopies is present to uncover the pristine activity, actual active species, and reaction centres of high-entropy LiNiO₂ for OER. Pristine LiNiO₂ with single-crystalline R-3m structure and with pure Ni^III was prepared and characterized unambiguously using electron microscopy, XRD, and XAS. The best and real OER activity of LiNiO₂ is achieved when the electrochemical redox of Ni²⁺/Ni^III and the introduction of Fe impurities are prevented, and the activity is found to be comparable with that of the state-of-the-art IrO₂ and much higher than that of the conventional NiOOH in terms of overpotentials, Tafel slopes and OER onset potentials., Operando XRD, XAS, and Raman spectra were combined to probe the active species and reaction centre. The in situ delithiation, Ni^III oxidation and oxygen-vacancy formation in LiNiO₂ under OER were found to coincide with OER for the first time, and the active species was then identified as delithiated yet structurally intact Li_xNiO₂ with high-valent Ni^IV and double ligand holes, which is also supported by the prevalence of the MLOV OER mechanism from DFT calculation and by the activated lattice oxygen from in situ DEMS measurements. This unique Li_xNiO₂ active sites were also found to be superior to those of NiOOH for the formation of a synergetic electrocatalyst with Fe³⁺. This study elucidates the electrocatalysis of lithiated metal oxides in OER and highlights the importance of considerate electrochemical measurements and complementary operando techniques in understanding the pristine electrochemistry and actual active species.

Methods

Preparation of LiNiO₂

The pristine LiNiO₂ was synthesized using a solid-state reaction method with two annealing steps. NiO and LiOH were grinded and mixed by a mortar and pestle for 30 min, and transferred into a tubular furnace. A 15% excess of the Li precursor was added to compensate for the evaporation of Li₂O and/or Li₂O₂ during annealing. After the furnace was purged with O₂ (100 sccm) for 15 min, under such O₂ flow the temperature was increased to 750 °C with a ramping rate of 5 °C min⁻¹ and held at this temperature for 72 h. The second annealing step was carried out at ~100 bar O₂ pressure with cooling naturally to room temperature.

Structural characterisation

Lab-based X-ray diffraction (XRD) patterns were obtained using a Bruker D8 Advance diffractometer with Cu K_α radiation (λ = 1.5418 Å) and a LYNXEYE detector, and diffraction peaks from Cu K_α2 was stripped off using Jade 6 software. Transmission electron microscopy (TEM) measurements, including low-magnification TEM, high-resolution TEM (HRTEM) and selected area electron diffraction (SAED), were carried out on a FEI Tecnai G2 F20 S-TWIN electron microscope. Scanning electron microscopy (SEM) measurements were performed on a Zeiss Crossbeam 540. High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images were recorded using an aberration-corrected FEI Themis Z STEM. The microscope was operated with an accelerating voltage of 60 kV and the HAADF detector angle range was 50–200 mrad. Note that the accelerating voltage of STEM was selected as 60 kV to prevent beam damage, which is reported to induce Ni displacement and oxygen release on LiNiO₂ after long exposure to electron irradiation⁶⁶. X-ray absorption spectra (XAS) at the Ni K edge were measured in the transmission mode at TPS 44 A beamline of National Synchrotron Radiation Research Centre (NSRRC, Hsinchu), and using an in-house lab-based X-ray absorption spectrometer (Supplementary Fig. 36). The electron storage ring of the TPS was operated at 3.0 GeV with a ring current of ~500 mA. The measured spectra were processed and analysed using ATHENA and ARTEMIS (Demeter software package)⁶⁷. Soft XAS (sXAS) at the Ni L_3,2 edge and the O K edge were collected at TLS 11 A and 20 A beamlines of the NSRRC. The Ni L edge data of γ-NiOOH and KNiIO₆ were measured in fluorescence mode and inverse partial fluorescence yield mode, respectively, to reduce the spectral distortion of surface Ni²⁺.

Electrochemical measurements

Electrochemical measurements were performed on a standard three-electrode setup connected to a PGSTAT302N potentiostat (Metrohm Autolab, equipped with FRA32M and BA modules). A Pt gauze was used as the counter electrode, and a Hg|HgO electrode (in 1 M KOH) as the reference electrode. The potential measured using the Hg|HgO reference electrode, E(V vs. Hg|HgO, 1 M KOH), was converted into RHE scale in 1 M KOH, E(V vs. RHE), using the following equation,

E(V vs. RHE) = E(V vs. Hg|HgO, 1 M KOH) + 0.0977 V + 0.05916 pH (V) (1), where the pH of the purified 1 M KOH is measured as 13.73 at 25 °C. The values of potential in the text are thus referenced against RHE, unless stated otherwise. The contact and solution resistance were determined using electrochemical impedance spectroscopy under open circuit potential (OCP) before each voltammetry measurement, and the voltammograms were corrected by IR drops.

The working electrode was prepared by drop-casting 10 μL of a catalyst ink onto a polished rotating ring-disk electrode (RRDE, with a glassy carbon disk and a Pt ring) and drying under ambient condition to obtain an even film. The catalyst ink was prepared by dispersing 5 mg of each sample and 5 mg carbon (Vulcan XC-72R) into 1 mL isopropanol alcohol (IPA) aqueous solution (25 vol% of IPA) containing 40 μL 5 wt% Nafion solution.

Linear sweep voltammograms and cyclic voltammograms (CV) were collected using a RRDE configuration with 5 mV s⁻¹ in Ar-saturated 1 M purified KOH solution at 25 °C. The voltammograms were collected at 1600 rpm to promptly remove the generated O₂, and the disk potential was swept from the corresponding OCP of each sample, while the ring potential was held at −0.5 V (vs. Hg|HgO, 1 M KOH) to electrochemically reduce and detect the O₂ evolved from the disk. For the long-term stability test, the working electrode was prepared using carbon paper (Toray TGP-H-060) with the same catalyst loading and ink recipe as the RRDE, and the stability was evaluated using chronopotentiometry at 10 mA cm⁻² for 48 h, before and after which LSV curves were measured with 5 mV s⁻¹. The OER activity of NiOOH and IrO₂ (99.9%, Adamas-beta) was measured for comparison. The NiOOH was prepared by potential cycling NiO nanopowder (50 nm, 99.9% metal basis, Aladdin) in 1 M KOH between 1.01 V and 1.66 V for at least 40 cycles until the Ni²⁺/Ni^III redox couple at ~1.32 V was fully developed (Supplementary Fig. 37).

To avoid the inference of Fe impurities on the measured OER activity, the electrochemical cell was cleaned by concentrated H₂SO₄ overnight and boiled with ultrapure H₂O (18.2 MΩ cm) for at least 3 times to remove residual H₂SO₄ and other impurities, and the KOH solution was purified using Trotochaud’s method³⁸. Briefly, 1 M KOH solution was first prepared from semiconductor grade KOH (99.9996% metal basis, 4 ppm total metal impurity excluding Na⁺, Aladdin) and ultrapure water. Into a H₂SO₄-rinsed 50 mL polypropylene centrifuge tube, ~2 g of Ni(NO₃)₂·6H₂O (99.9985%, PURATREM, STREM) were dissolved in ~4 mL ultrapure H₂O and mixed with 20 mL 1 M KOH. The resulting Ni(OH)₂ suspension was shaken and centrifuged, and the supernatant was decanted. The Ni(OH)₂ was then washed 3 times with ~20 mL ultrapure H₂O and ~2 mL 1 M KOH, and transferred to a H₂SO₄-rinsed 100 mL polypropylene bottle. The bottle was filled with 75 mL 1 M KOH for purification, and the freshly prepared and washed Ni(OH)₂ was redispersed and shaken for 10 min, followed by resting overnight.

Operando and online characterisation

Operando synchrotron XRD patterns were collected at TLS 01C2 beamlines of the NSRRC. The XRD experiments were performed at an incident beam energy of 18 keV using a mar345 image plate area detector. The diffraction intensity and the peak shape were calibrated with a CeO₂ reference sample (SRM 674b). Operando Quick-XAS measurements at Ni K were carried out in the transmission mode at TPS 44 A beamline of the NSRRC, and 120 XAS spectra collected within 1 min were merged to achieve high data quality. Operando Raman spectra were collected using a Horiba LabRAM HR Evolution confocal Raman microscope with an excitation wavelength of 473 nm. The spectra resolution is ~1 cm⁻¹, and the spectral shifts were calibrated routinely using silicon wafer (520.7 cm⁻¹). The operando Raman experiments were performed in a homemade spectro-electrochemistry cell, consisting of a Au disk working electrode (5 mm), a Pt wire counter electrode and a Ag|AgCl (saturated KCl) reference electrode. Online differential electrochemical mass spectrometry (DEMS) measurements were carried out using a QAS-100 equipped with a PrismaPro quadrupole mass spectrometer (Linglu Instruments). The O¹⁸-labeled LiNiO₂ was prepared by cycling the unlabelled one to the OER range (e.g., 1.8 V) in ¹⁸O-enriched electrolyte, and before the experiment in the ¹⁶O-based electrolyte, rinsing and drying steps were conducted to remove free ¹⁸O species.

Computation details

The present calculations employ the Vienna ab initio Simulation Package (VASP) implementation of density functional theory (DFT) in conjunction with the projector augmented wave (PAW) formalism⁶⁸. The H 1s¹, Li 2s¹, O 2s²2p⁴, and Ni 4s²3d⁸ states are treated as valence electrons. The electronic wave functions were expanded in plane waves using an energy cutoff of 520 eV, and the force and energy convergence criteria were set to 0.02 eV Å⁻¹ and 10⁻⁵ eV respectively. To overcome the self-interaction error of the exchange correlation functional, we employed the Hubbard U model for describing the strong correlation of the localized Ni 3d states, and set the value of U_eff = (U − J) to 6.40 eV according to previous work⁴⁶. The reciprocal space was sampled using a 4×4×1 Monkhorst–Pack k-point mesh for DFT + U calculations. To simulate the valence states of Ni in real catalytic environment, we built the Li_0.5NiO₂ system containing 50% lithium vacancy. The periodic slab models of LiNiO₂ and Li_0.5NiO₂ system with 9 Ni sites per surface, with two layers at the bottom fixed during the relaxation. To prevent spurious interactions, the thickness of vacuum spacing was ~15 Å in the z-direction. We calculated the Gibbs free-energy differences (∆G) using the computational hydrogen electrode model under standard conditions⁶⁹, with the voltage applied U_RHE = 0 and 1.23 V. The calculation details regarding ∆G can be found in the Supplementary Information.