2,6-naphthalenedicarboxylic acid (2,6-NDA, 172.9 mg) and CoCl₂·6H₂O (190.2 mg) were dissolved in a mixed solution containing 32-mL dimethyl formamide (DMF), 2-mL ethanol, and 2-mL deionized water. Thereafter, 1.0-mL triethylamine (TEA) was injected into the above solution mixture under vigorous stirring to form a uniform colloidal suspension. The formed suspension was sealed and further ultrasonicated for 300 min at 70 kHz and room temperature. The obtained sample was further immersed in a 0.2-mM acetic acid solution for acid etching. The resultant precipitate was washed with DMF and ethanol for three times, and dried in vacuum oven at 60 °C.

Morphologies of the as-prepared samples were obtained on a field emission scanning electron microscopy (FESEM) (Hitachi SU-8010), transmission electron microscopy (TEM) (HT7700), and high-resolution transmission electron microscopy (HR-TEM) (JEOL JEM-2001F). The crystal structures of the as-prepared samples were analyzed by X-ray powder diffraction (XRD) (Empyrean 200,895) using Cu Kα radiation. Chemical structures of the as-prepared samples were measured by X-ray photoelectron spectroscopy (XPS, Escalab250Xi) with Al Kα radiation. Raman spectra were obtained by a LabRAM HR Evolution unit. The metal content in samples was analyzed by inductively coupled plasma mass spectrometry (ICP-MS) (Vista Axial). The X-ray absorption spectroscopic (XAS) measurements of the as-prepared samples were conducted in Beijing Synchrotron Radiation Facility and Taiwan Synchrotron Radiation Facility. Quantum Design MPMS-7 superconducting quantum interference device (MPMS-VSM) magnetometer was utilized to obtain the magnetic properties of the as-prepared catalysts. The electron paramagnetic resonance (EPR) data of the as-prepared catalysts were acquired on a Bruker EMXmicro.

All electrochemical measurements were carried out by an electrochemical analyzer (CHI 760E) in a typical three-electrode cell. A saturated calomel electrode (SCE, CH Instruments) was used as the reference electrode, and a graphite rod was used as the counter electrode. The potential was converted to reversible hydrogen electrode (RHE) via a Nernst equation (E_RHE =  E_SCE × 0.244 V + 0.0591 × pH). To evaluate the OER activities, the scan rate of linear sweep voltammetry (LSV) was set as 1.0 mV s⁻¹ with the potentials from 1 to 1.8 V vs. RHE in 1.0 M KOH. Electrochemical impedance spectroscopy (EIS) was measured at 1.5 V vs. RHE with a frequency range from 10⁵ to 0.01 Hz. All polarization curves were calibrated with iR correction unless noted. Cyclic voltammetry cycles (CVs) at 0.96-1.06 V vs. RHE with the scan rates from 10 to 50 mV s⁻¹ were applied.

The PEC-OER measurements were conducted with a front-side illumination in all cases (light enters from the absorber side). A saturated calomel electrode (SCE, CH Instruments) and a Pt wire were used as reference electrode and counter electrode, respectively. The recorded potential vs. SCE (E_SCE) was converted against RHE using the Nernst equation (E_RHE =  E_SCE × 0.244 V + 0.0591 × pH). A 1.0 M potassium borate buffer, prepared by adjusting the pH of 1.0 M H₃BO₃ to 9.0 with KOH solution, was used as the electrolyte. The stability of photoanode was evaluated by measuring curves of the potential vs. time at 1.0, 2.0, and 2.5 mA cm⁻².

All spin-polarized density functional theory (DFT) calculations were performed with Vienna Ab initio Simulation Package (VASP). In this work, the generalized gradient approximation (GGA) within Perdew-Burke-Ernzerhof (PBE) method was used to describe the exchange-correlation interaction. The core electrons were replaced by the projector augmented wave (PAW) pseudopotential, and 450-eV plane-wave expansion was setup for energy cut-offs. The k-points sampling was set to 1 × 1 × 1 Monkhorst-Pack k-points grid for geometrical optimization and that of electronic property calculation was 1 × 1 × 2. The convergence threshold was set to 10^-5 eV and 0.05 eV Å⁻¹ for energy and force, respectively. Weak interaction was described by DFT-D3 method using empirical correction in Grimme’s scheme.

The (1 × 2) surface of bulk CoNDA (100) was chosen, which contains 16 Co atoms, 83 O atoms, 96 C atoms, and 64 H atoms. A vacuum slab of about 15 Å was maintained in the super-cell configuration that was large enough for the calculations. To simulate the tensile state of structure, tensile CoNDA structure was constructed by expanding the lattice parameters of the pristine CoNDA in three directions by 5%.

AE-CoNDA as cocatalyst was synthesized by ultrasonic exfoliation of a mixed solution of CoCl₂·6H₂O, 2,6-NDA, and DMF, followed by acid etching at a predetermined concentration of acetic acid. The AE-CoNDA@BiVO₄ was assembled by spraying the AE-CoNDA ink on BiVO₄, as illustrated in Fig. S1. Morphological images (Figs. S2-S5) exhibited that the AE-CoNDA and control samples uniformly coated on the nanoworm-like BiVO₄ surface. Especially, the HR-TEM image of AE-CoNDA@BiVO₄ (Fig. S2) exhibits a lattice fringe of 1.18 nm for the interlayer spacing distance of naphthalene-based MOF with transition metal nodes along the [100] direction, which assigned to AE-CoNDA. In addition, in XRD pattern (Fig. S6), the diffraction peak below 10° stands for crystalline MOF structure, indicating that the AE-CoNDA is highly crystallized. Raman spectra (Fig. S7) also show obvious characteristic peaks which assigned to naphthalene-based MOF. Above results confirmed that AE-CoNDA was highly crystallized and successfully deposited on BiVO₄ photoelectrode.

To demonstrate the catalytic performance, we first investigated PEC water oxidation property of AE-CoNDA@BiVO₄ in alkaline electrolyte under AM 1.5G irradiation, as illustrated in Fig. 1a. The optimized AE-CoNDA@BiVO₄ delivered a high photocurrent density of 4.30 mA cm⁻² at 1.23 V (Figs. 1b and S8), which was much greater than that for BiVO₄ alone (1.30 mA cm⁻² at 1.23 V) and CoNDA@BiVO₄ (3.20 mA cm⁻² at 1.23 V). In addition, the AE-CoNDA@BiVO₄ possessed a maximum incident photon-to-current conversion efficiency (IPCE) of 54.0% at 400 nm at 1.23 V (Fig. 1c), which was about 1.6 and 1.3 times higher than that of bare BiVO₄ (33.0%) and CoNDA@BiVO₄ (43.0%), respectively. AE-CoNDA@BiVO₄ showed the maximum photon-to-current efficiency (ABPE) of 1.61% at 0.6 V (Fig. 1d), which was superior to that of BiVO₄ (0.82% at 0.78 V) and CoNDA@BiVO₄ (1.12% at 0.6 V). The lower charge transfer resistance for the AE-CoNDA@BiVO₄ than that of CoNDA@BiVO₄ and BiVO₄ alone under light irradiation demonstrated that more effective interfacial charge transfer occurred at the AE-CoNDA@BiVO₄ interface (Fig. S9). Moreover, clearly, it was seen that after introduction of AE-CoNDA, both charge transfer efficiency (η_transfer) and charge transport efficiency (η_transport) of AE-CoNDA@BiVO₄ (Figs. S10-S13) increased throughout the whole potential range. The product analyses revealed that the generated photocurrent was mainly from the O₂ production on AE-CoNDA@BiVO₄ rather self-corrosion of photoelectrode (Fig. 1e). Besides, the potential on AE-CoNDA@BiVO₄ was not raised for at least 50 h to maintain photocurrent density at 2.5 mA cm⁻² (Fig. 1e), which is much durable than CoNDA@BiVO₄ under 1.5 mA cm⁻² and BiVO₄ under 1.0 mA cm⁻² (Fig. S14). Notably, the observed PEC-OER activity for AE-CoNDA@BiVO₄ was regarded as an extreme high record among all reported Co-based BiVO₄ hybrid photoanodes (Fig. 1f and Table S1). To further support the enhancement of photo-generated charge carrier transfer efficiency of BiVO₄ after integrating AE-CoNDA cocatalyst, operando-irradiated XPS measurements [35,36] were conducted (Fig. S15). The results revealed that the holes transferred from BiVO₄ to AE-CoNDA cocatalyst occurred.

To probe the crucial role of AE-CoNDA as cocatalyst, the OER performance of AE-CoNDA was further evaluated in a three-electrode cell in 1.0 M KOH solution. The control samples with different reaction conditions were also synthesized to investigate the contribution of acetic acid etching. The optimal etching time was found to be 1.0 h, and optimal concentration of acetic acid was 0.2 mM (Figs. S16-S18). As shown in Fig. 2a, the as-prepared AE-CoNDA delivered only a low overpotential of 260 mV to reach a current density of 10 mA cm⁻². Compared to the CoNDA without acid etching (360 mV), this result highlights the significant role of acid etching in enhancing the OER performance. Notably, the overpotential for the AE-CoNDA was even smaller than that of the benchmark Ir/C (330 mV) at the same current density, demonstrating the excellent OER activity for the former. The fast reaction kinetics of AE-CoNDA was further underscored by its smaller Tafel slope (62 mV dec⁻¹) than that of CoNDA with 82 mV dec⁻¹, and even commercial Ir/C with 97 mV dec⁻¹, respectively (Fig. 2b). The AE-CoNDA further exhibited a much smaller electrochemical resistance (R_ct) (Fig. S19 and Table S2) than that of the CoNDA, implying a fast charge transfer for AE-CoNDA. The ratio of inner charge to outer charge density for AE-CoNDA was calculated to be 94.5 (Figs. 2c and S20), which was much higher than that of CoNDA (52.7), as a result of the high electrochemical porosity. These results suggested that the acid etching not only increased the electrode roughness [37], but also produced abundant exposed active sites. The larger electrochemically active surface area (ECSA) of AE-CoNDA than that of CoNDA indicated the more electroactive surfaces for AE-CoNDA (Figs. S21 and S22). Moreover, the turnover frequency (TOF) of AE-CoNDA was calculated to be 18.66 h⁻¹ at an overpotential of 0.3 V, which was approximately three times higher than the corresponding value for CoNDA (5.86 h⁻¹) (Fig. 2d). Above results claimed that the increase in active surface area is not the main reason for the increase in the current density of the catalyst, but the intrinsic characteristics of the catalyst promote the catalytic activity of the catalyst, resulting in a higher current density.

To further validate the rapid charge transfer for AE-CoNDA, in situ EIS measurements were conducted (Table S2). The results showed that the R_ct value of AE-CoNDA was much lower than that of CoNDA over a wide range of potential (Fig. 2e), indicating the fast charge transfer efficiency for AE-CoNDA. Besides, the R_ct value of the AE-CoNDA became stable at low potential (1.4 V), demonstrated that the OER process was triggered at low potentials on the AE-CoNDA [38,39]. Notably, the excellent alkaline OER performance in terms of the overpotential (260 mV at 10 mA cm⁻²) and Tafel slope (62 mV dec⁻¹) for the AE-CoNDA was superior to those of almost all non-carbonized Co MOF-based OER electrocatalysts previously reported (Table S3). Moreover, the OER process on the AE-CoNDA was stable, no apparent change in the overpotential was observed over 100 h of continuous operation (Fig. 2f), which is much stable than CoNDA under same potential (Fig. S23). Especially, the XRD pattern (Fig. S24a) and Raman spectra (Fig. S24b) of AE-CoNDA after OER process remained similar curves with pristine AE-CoNDA, indicating structural stability of catalyst.

To gain the insights of superior OER activity on AE-CoNDA, we first performed XPS measurements. In the high-resolution Co 2p, O 1s, and C 1s XPS spectra (Figs. 3a, b and S25), the Co 2p XPS peaks in the AE-CoNDA displayed a higher binding energy, shifted approxiately 0.62 and 0.40 eV, while O 1s XPS peaks positively shifted 0.06 eV to higher binding energy, as the peaks of O-C = O and C-C/C = C negatively shifted to lower binding energies (Fig. S25). These results demonstrated the electrons from Co atoms within AE-CoNDA transferred to the O atoms adjacent to C atoms and ultimately located on the C atoms. To deeply investigate the fine structure of AE-CoNDA, we further conducted X-ray absorption spectroscopy (XAS). The obtained X-ray absorption near-edge structure (XANES) spectra (Fig. 3c) showed the absorption edge with an increased photon energy of Co species within the AE-CoNDA in comparison with CoNDA and Co foil, indicating a higher valence state for Co species in AE-CoNDA. Moreover, the Co L-edge XAS spectra of the AE-CoNDA (Fig. 3d) exhibited a Co absorption edge shifted to a higher photon energy relative to that of the Co foil and CoNDA, but close to that of reference CoO, suggesting the average Co valence state of + 2 for AE-CoNDA.

Figure 4a shows the corresponding Fourier transforms of extended X-ray absorption fine structure (EXAFS) of AE-CoNDA, which reveals the major peak located at 1.6 Å attributable to the nearest shell coordination of the Co-O bond. It is noteworthy that the Co-O peak of AE-CoNDA shifted toward a higher R direction by about 0.05 Å relative to that of CoNDA, indicating that the bond length of Co-O in the AE-CoNDA was stretched [35]. The stretched Co-O bond is ascribed to partial destruction of CoNDA structure and weakened interaction between Co cations and NDA ligand after acid etching. The EXAFS fitting was performed for the coordination shells in the R range of 1.0-3.0 Å. Amplitude reduction factor S₀² as determined from fitting was 0.85. The EXAFS fitting results (Table S4) validated that the coordination number of Co-O bonds was determined to be smaller than 6. Additionally, as revealed by the EPR spectra (Fig. S26), an apparent signal at g-factor value of 1.86 was observed for AE-CoNDA, whose intensity was stronger than which of CoNDA, indicating more uncoupled Co centers as defects in the framework in AE-CoNDA [40]. EXAFS fitting result and EPR results proved distorted octahedral configurations and unsaturated coordinative Co species of CoO₆ centers in AE-CoNDA.

To gain more insights into the electron configuration and spin state of AE-CoNDA with stretched Co-O bond length, the O K-edge XAS spectra were further analyzed. As shown in Fig. 4b, the O K-edge XAS spectra consisted of two characteristic peaks between 532.0 and 534.0 eV, arising from the hybridization between unoccupied O 2p and Co 3d orbitals. Clearly, the intensity of O K-edge from AE-CoNDA was enhanced after acid etching, indicating an increased unoccupied density of states and a strengthening orbital hybridization of O 2p and Co 3d orbitals. Such enhanced intensity in O K-edge spectra, accompanying with the Co L-edge peak for AE-CoNDA (Fig. 3d), suggests an electron reconfiguration occurred in both the O 2p and Co 3d orbitals.

It is important to notice that the electronic configuration of Co species primarily dominated the spin state of Co species. To deeply explore the spin state of Co species within AE-CoNDA, we further carried out the zero-field cooling (ZFC) temperature-dependent magnetic susceptibility tests with a magnetic property measurement system (MPMS-VSM) [41]. The total effective magnetic moment (μ_eff) of AE-CoNDA was obtained through χ⁻¹-T liner fitting to calculate the e_g occupancy. In Fig. 4c, the calculated μ_eff for AE-CoNDA and CoNDA was 3.667 and 1.935, respectively, and the number of d-orbital unpaired electrons can be further calculated by formula [42]. Thus, the e_g occupancy of Co species raised from approximate 0.98 to 1.84 after Co-O bond length stretched. The calculated e_g occupancy indicated more unpaired d electrons existed in the AE-CoNDA (Fig. 4d), which induced a spin state transition from an intermediate spin (e_g¹t_2g⁶) to a high spin (e_g²t_2g⁵) of AE-CoNDA [43].

To give an in-depth understanding of the relationship between the spin state of Co active sites and adsorption ability of oxygen-contained intermediates, we performed DFT calculations. Firstly, a CoNDA model with short bond length and AE-CoNDA model with long bond length were established (Fig. S27). As shown in Fig. 5a, the Co active sites adsorbed oxygen-contained intermediates of OH*, O*, and OOH* through hybridization between the O 2p and Co 3d orbitals throughout the OER process. Considering that the high spin state of Co active sites was beneficial for Co 3d and O 2p orbital hybridization to facilitate the adsorption of oxygen-contained intermediates [44], we also calculated the spin-resolved density of spin (DOS). The 3d-orbital of Co sites was splitted as dx²-y², dz², dxy, dxz, and dyz orbitals in Fig. 5b and c. As for AE-CoNDA (Fig. 5b), the state density of dz² (close to the Fermi level) was much stronger than that of CoNDA (Fig. 5c), demonstrating that more electrons transferred between different orbitals in AE-CoNDA. The overlapped area between the Co dz² orbital and O 2p orbital for AE-CoNDA was larger than that for CoNDA, indicating an enhanced interaction between the Co active sites and oxygen-contained intermediates after the Co-O bond length stretching. These results demonstrated that the bond length stretching facilitated the electron transfer from the t_2g orbitals to the e_g orbitals, thus leading to a high spin state [40].

As displayed in Fig. 5d, the AE-CoNDA model showed an accumulated charge of O center in NDA linker from Co active sites with the stretched Co-O bond length compared to control CoNDA, indicating a rapid electron transfer for AE-CoNDA [45]. The specific free energy (ΔG) for each elementary step was calculated to estimate their OER activity at the center Co sites. In Fig. 5e, it can be clearly seen that the elementary steps for the formation of OH* and transition of OOH* to O₂ desorption were down-hilled, while other reaction steps involved in the OER process for AE-CoNDA and CoNDA were up-hilled. Based on its ΔG diagram, the ΔG value for O* formation at the center Co sites decreased from 1.41 eV for CoNDA to 1.26 eV for AE-CoNDA, revealing that the spin state transition from intermediate spin to high spin led to a decrease in free energy. The application of a bias (U = 0.4 V) also caused a decrease in the ΔG of OH* to O* for AE-CoNDA compared with CoNDA (Fig. S28). Therefore, the formation of the stretched Co-O bond led to a high spin state for the Co active sites, ensuring a readily combination of the O* with the center Co sites. As a result, a reaction path from OH* to O*, which was evidenced as the rate-determined step, was significantly accelerated for AE-CoNDA (Fig. 5e).