Engineering all-organic electrocatalysts with asymmetric dual-active sites for uncommon oxygen-evolving pathway

Abstract

The performance of electrocatalytic oxygen evolution reaction is governed by molecular evolution pathways, yet their regulation remains a significant experimental and theoretical hurdle. Herein, we propose the construction of asymmetric non-metallic dual-active sites through engineering hydroxyl groups in an all-organic electrocatalyst, thereby triggering an uncommon asymmetric oxide pathway mechanism. This reaction pathway reduces the requirement of strict equilibrium on adsorption energies at dual-active sites for traditional oxide pathway mechanism, and also avoids the limitation of the linear scaling relationship for adsorbate evolution mechanism. Therefore, the proposed metal-free electrocatalyst shows competitive oxygen evolution reaction performance with an overpotential of 364.3 ± 4.5 mV at 1.0 A·cm⁻². Combining operando spectroscopic/mass spectrometry and theoretical calculations unveils that the asymmetric oxide pathway mechanism undergoes the critical step of *OH intramolecular nucleophilic attack to neighboring *O, forming *OOH with drastically reducing energy barriers. This work shows the diversity of oxygen evolution pathways in all-organic electrocatalysts and offers a viewpoint for optimizing the electrocatalytic performances of metal-free solids.

Controlling reaction pathways in the oxygen evolution reaction remains a major challenge for efficient electrocatalysis. Here, the authors report an all-organic catalyst with asymmetric dual sites that enables an unconventional oxide pathway, lowering energy barriers and improving high-current performance.

Introduction

The oxygen evolution reaction (OER) represents an essential anodic process that underlies key renewable energy conversion systems such as water electrolyzer, electrochemical CO₂ reduction device, and rechargeable metal-air battery^1–4. However, their practical application remains constrained by the inherently sluggish kinetics of OER^5,6. High-performance and scalable OER electrocatalysts are the cornerstone of developing sustainable energy. The current OER catalysts remain predominantly dependent on transition metals leveraging active d-orbital electrons^7,8. In contrast, metal-free catalysts have been underperforming due to electronic structure limitations, despite their inherent advantages in elemental abundance, cost-effectiveness, and environmental compatibility⁹. This performance deficit originates from the inherent weak p–p orbital coupling between nonmetallic active centers and oxygen intermediates, which impedes effective binding of reactive species and stabilization of transition states^10–13. The energy mismatch between the p-orbitals of nonmetallic active sites and those of oxygen-containing intermediates causes the adsorption strength deviating from the optimal value. This fundamentally limits the kinetics of the multi-proton-coupled electron transfer process and leads to a significant increase in the reaction energy barrier. Currently, achieving precise energy alignment and robust electronic interaction between p-orbitals of metal-free catalysts and oxygen intermediates remains a fundamental scientific challenge. Conventional metal-free carbon materials exhibit potential OER activity, yet their intrinsic disordered structures introduce uncertainties in active site identification, thereby complicating systematic performance optimization and mechanistic elucidation¹⁴. Covalent organic frameworks (COFs)^15–20, as well-defined all-organic electrocatalysts (AOEs), combine good stability, abundant porosity and high specific surface areas, enabling atomic-level regulation of active sites for precise p-band engineering and establishing clear structure-activity relationships^21,22. COFs provide an ideal molecular platform for developing efficient metal-free OER catalysts and probing p-orbital mediated mechanisms^23–26.

Generally, OER is governed by three principal catalytic mechanisms: the adsorbate evolution mechanism (AEM), the lattice oxygen mechanism (LOM), and the oxide path mechanism (OPM) (Fig. 1)²⁷. Traditional AEM pathway yields three intermediates (*OH, *O, and *OOH), whose adsorption energies are constrained by a linear scaling relationship, fundamentally limiting the catalytic activity²⁸. The LOM pathway directly forms O₂ through lattice oxygen-mediated *O−O* coupling, achieving faster kinetics but sacrificing structural stability through lattice oxygen loss²⁹. Notably, metal-free catalysts rarely follow the LOM pathway due to the thermodynamic challenge of extracting lattice oxygen from stable covalent atoms. These inherent limitations of AEM and LOM pose a major challenge in simultaneously delivering high activity and durability. Recent studies reveal that the OPM pathway directly couples surface-adsorbed *O species to form O₂, circumventing *OOH formation and lattice oxygen involvement, breaking the linear scaling relationship and maintaining structural integrity³⁰. However, OPM requires the dual active sites satisfying precise geometrical configurations and moderate oxygen affinity to synergistically promote *O−O* radical coupling^31,32. The current metal-free catalysts mainly rely on the AEM pathway, limiting substantial performance improvements. Therefore, regulating the catalytic mechanism is important for overcoming the bottleneck for AOE development.

Fig. 1. Schematic diagram of OER mechanism.

AEM, LOM, and OPM are previously reported, while AOPM is proposed here. The classic OPM occurs at two adjacent active sites and is mainly characterized by the generation of symmetrically adsorbed *O−O* intermediate. The AOPM requires asymmetric dual active sites and forms the *OOH intermediate through intramolecular nucleophilic attack.

Herein, we develop a metal-free MEC-T (or MEC-3, while MEC represents laboratory of molecule-based energy chemistry) with asymmetric dual-active sites by introducing polar hydroxyl groups into a benzoxazole-based COF through a molecular modularization strategy. This electrocatalyst exhibits overpotentials of 277.7 ± 2.1 mV at 10 mA·cm⁻² on a glassy carbon electrode (GCE) and 364.3 ± 4.5 mV at 1000 mA·cm⁻² on a carbon cloth (CC), along with electrocatalytic stability over 200 h at 500 mA·cm⁻², representing a competitive OER catalyst. Density functional theory (DFT) calculations elucidate that hydroxyl engineering downshifts the p-band centers of carbon sites, strengthening oxygen intermediate adsorption, thereby triggering an asymmetric oxide pathway mechanism (AOPM) with reduced reaction energy barriers. This work describes the AOPM pathway in metal-free solid systems and provides guidance for designing efficient AOEs.

Results and discussion

Characterization of catalysts

A benzoxazole-linked covalent organic framework (COF), designated as MEC-L (also called as LZU-190³³), was constructed by dynamic polymerization of 2,5-diamino-1,4-dihydroxyphenyl dihydrochloride with benzene-1,3,5-tricarbaldehyde (BT) (Fig. 2a, Supplementary Figs. 1 and 2). To precisely regulate the hydroxyl content in the framework, we synthesized a series of COFs (MEC-L₂T₁, MEC-L₁T₂, and MEC-T) with tunable hydroxyl density by partially replacing BT with trial-aldehyde phloroglucinol (TP) (Fig. 2a). The powder X-ray diffraction (PXRD) patterns of COFs are consistent with the simulations (Fig. 2b and Supplementary Figs. 3−6), verifying the successful construction of benzoxazole linkages and their structural isomorphicity³³. Fourier transform infrared (FT-IR) spectra show signature stretching vibrations at 1125 cm⁻¹, 1456 cm⁻¹, and 1640 cm⁻¹, confirming the formation of benzoxazole, while the C−O vibration at 1080 cm⁻¹ corresponds to phenolic hydroxyl groups (Fig. 2c)³⁴. The framework structure is further corroborated by solid-state ¹³C cross-polarization magic angle spinning (CP-MAS) nuclear magnetic resonance (NMR) profiles and Raman spectra (Fig. 2d and Supplementary Fig. 7)³⁵. These COFs exhibit good thermal stabilities (Supplementary Fig. 8), while their Brunauer-Emmett-Teller (BET) specific surface areas decrease with increasing hydroxyl density (Supplementary Fig. 9). Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images reveal their nanosheet-like morphology (Fig. 2e and Supplementary Figs. 10−13). High-resolution TEM (HRTEM) images show that they exhibit clear lattice fringes with a spacing of 0.335 nm, consistent with the (001) crystal face. Elemental mapping images assigned through energy dispersive X-ray (EDX) spectroscopy affirm the uniform dispersion of C, N, and O throughout the COF matrices (Fig. 2e and Supplementary Figs. 14−16). Atomic force microscopy (AFM) measurements unveil the nanosheet characteristics of these materials, with average thicknesses of 3.2 ± 0.2 nm (MEC-T), 5.2 ± 1.0 nm (MEC-L₁T₂), 4.8 ± 0.6 nm (MEC-L₂T₁), and 5.7 ± 0.1 nm (MEC-L) (Fig. 2f and Supplementary Figs. 17−20). The water contact angle of MEC-T decreases to 14.8^o in comparison with MEC-L, indicating that hydroxyl modification enhances its hydrophilicity (Fig. 2g and Supplementary Fig. 21). Elemental analysis uncovers that the increasing hydroxyl density in COFs leads to elevated oxygen contents, while keeping constant nitrogen amounts (Fig. 3a).

Fig. 2. Synthesis and structural characterization.

a Illustration of COF electrocatalysts with different amounts of OH modifications. b PXRD pattern and the Rietveld refinement of MEC-T. c FT-IR spectra of COF electrocatalysts (inset: structure of the oxazole unit). d Solid-state ¹³C CP-MAS NMR spectra of MEC-T and MEC-L (inset: structural units of MEC-T and MEC-L). e TEM and EDX elemental characterization of MEC-T. f AFM image of MEC-T (inset: thickness measurement). g Static contact images of water droplets on MEC-L and MEC-T. Source data are provided as a Source Data file.

Fig. 3. Chemical and electronic structure analysis.

a N and O elemental contents of COF electrocatalysts. Error bars denote the standard deviation based on three repeated trials. b NEXAFS spectra of COFs at the C K-edge, N K-edge and O K-edge. c XPS spectra of COFs at the C 1s and O 1s. d Proposed mechanism of electronic structure modulation by hydroxyl functionalization between MEC-T and MEC-L. The deep red shadow indicates that a greater electron deficiency at the carbon atoms. Source data are provided as a Source Data file.

We employed synchrotron-based soft X-ray absorption near-edge structure (XANES) to analyze the chemical bonding structures of COFs. The C K-edge XANES spectra exhibit characteristic peaks classified as π*(C=C), π*(C=N), and π*(C−O−C/C−OH) at 285.0, 286.9, and 288.1 eV, respectively (Fig. 3b)³⁶. The N K-edge XANES spectra exhibit absorption peaks at 397.0 and 399.5 eV (Fig. 3b), originating from transitions to unoccupied π* (C=N) and π* (C−N) orbitals of the benzoxazole unit, respectively. The intensity of the weak peak at 398.0 eV increases with hydroxyl group content, likely due to partial protonation of nitrogen atoms induced by intramolecular hydrogen bonding. Besides, the O K-edge XANES spectrum of MEC-T (Fig. 3b) shows a distinct increasing absorption assigned to π*(C−OH) at 530.5 eV³⁷, confirming the presence of functionalized hydroxyl groups. Notably, compared with the strong conjugation in MEC-L, the electron-withdrawing effect of hydroxyl groups in MEC-T significantly reduces the intensity of π*(C=C) and π*(C=N) characteristic peaks in the corresponding XANES spectra (Fig. 3b), which could be attributed to electron transfer from the benzoxazole unit to the hydroxyl groups, thereby diminishing the extent of electron conjugation and delocalization. The bonding configuration and electron distribution of COFs were resolved via X-ray photoelectron spectroscopy (XPS). The analysis indicates an increase in oxygen content from 26.4% for MEC-L to 33.8% for MEC-T (Supplementary Fig. 22), being consistent with the elemental analysis results. The high-resolution O 1s spectra exhibit a gradual intensification of the peak at 532.0 eV with increasing hydroxyl density, corresponding to the C−OH (Fig. 3c)^38,39. Deconvolution of the C 1s spectra resolves three distinct components, with binding energies characteristic of C=C/C−C (284.6 eV), C−O/C=N (286.2 eV), and C=O (287.8 eV) bonds, respectively (Fig. 3c)^40–42. Notably, MEC-T presents an upshifted peak of ca. 0.3 eV compared with that of MEC-L, implying that introducing hydroxyl groups reduces the electron density of carbon atoms. The high-resolution N 1 s spectra show characteristic peaks at 399.0 and 400.0 eV, originating from C=N and C−N in the benzoxazole unit (Supplementary Fig. 23), respectively^43–45. These results demonstrate that hydroxyl functionalization can effectively regulate the electronic structures of carbon atoms in the COFs (Fig. 3d).

Electrocatalytic performance evaluation

The OER performances of COFs were examined employing a GCE in 1.0 M KOH and compared with commercial IrO₂ benchmark. From the linear sweep voltammetry (LSV) tests, MEC-T possesses optimal electrocatalytic activity with an overpotential at 10 mA·cm⁻² (η₁₀) of 277.7 ± 2.1 mV (Fig. 4a,b and Supplementary Figs. 24 and 25), outperforming those of MEC-L₁T₂ (η₁₀ = 287.3 ± 2.1 mV), MEC-L₂T₁ (η₁₀ = 293.7 ± 1.2 mV), MEC-L (η₁₀ = 386.0 ± 1.0 mV) and IrO₂ (η₁₀ = 344.3 ± 7.8 mV). Moreover, MEC-T possesses the most favorable reaction kinetics, as evidenced by its lowest Tafel slope (45.3 ± 3.2 mV·dec⁻¹) among the test samples (Fig. 4b and Supplementary Fig. 26). The turnover frequency (TOF) value of MEC-T at an overpotential of 300 mV is 11.79 ± 1.3 s⁻¹ (Supplementary Fig. 27), which is 26 times higher than that of MEC-L (TOF = 0.35 ± 0.02 s⁻¹). The Faraday efficiency of MEC-T for OER is 99.68% (Supplementary Fig. 28), representing that almost all electrons contribute to produce O₂. Moreover, electrochemical active surface area (ECSA) assessment reveals that MEC-T possesses the highest density of accessible active sites (Supplementary Figs. 29 and 30). Electrochemical impedance spectroscopy (EIS) demonstrates that hydroxyl functionalization substantially improves charge transfer efficiency (Supplementary Fig. 31). Arrhenius profiles based on variable-temperature LSV measurements uncover that MEC-T exhibits a significantly lower apparent activation energy (E_a = 37 ± 1.3 kJ·mol⁻¹) relative to MEC-L (E_a = 105.7 ± 1.4 kJ·mol⁻¹) (Fig. 4c and Supplementary Fig. 32), indicating the much favorable kinetics of the former. Electrochemical characterization verifies that the hydroxyl functionalization strategy can significantly enhance the OER performance of this benzoxazole-based catalyst. The catalytic activity of MEC-T can be further improved by growing it on carbon nanotubes (CNTs) (Supplementary Figs. 33 and 34). Notably, the OER activity of MEC-T is competitive among advanced electrocatalysts (Supplementary Fig. 35 and Table S1 and S2).

Fig. 4. Electrocatalytic OER activity and stability.

a Polarization curves for COF electrocatalysts and IrO₂ on GCE acquired at 5 mV·s⁻¹ in 1.0 M KOH (pH = 13.6 ± 0.2) at 25 ^oC. E: potential. J: current density. b Overpotentials (η) at 10 mA·cm⁻² and Tafel slopes for COF electrocatalysts and IrO₂ on GCE. c Arrhenius plots of MEC-T and MEC-L. j₀: exchange current density. E_a: activation energy. d LSV curves of MEC-T and IrO₂ on CC acquired at 5 mV·s⁻¹ in 1.0 M KOH. e Δη/Δlog|j| ratios across current densities. (inset: comparison of the overpotentials at different current densities of MEC-T and IrO₂). f The AEMWE device activities of MEC-T versus commercial benchmarks (inset: AEM electrolyzer photograph). Polarization curves were obtained on CC acquired at 5 mV·s⁻¹ in 1.0 M KOH at 25 °C without iR-compensation. g Galvanostatic measurement of MEC-T at 0.5 A·cm⁻² on CC in 1.0 M KOH at 25 °C without iR-compensation. h Durability of the AEM electrolyzer at 0.5 A·cm⁻² on CC in 1.0 M KOH at 25 °C with initial and final cell voltages of 1.81 V and 1.95 V, respectively. Error bars denote the standard deviation based on three repeated trials. Source data are provided as a Source Data file.

We further evaluated the OER performance of MEC-T under ampere-level current densities using carbon cloth (CC) as a conductive substrate. The fabricated MEC-T/CC electrode exhibits overpotentials of 326.0 ± 2.0 and 364.3 ± 4.5 mV at 0.5 and 1.0 A·cm⁻², which are lower than those of the benchmark IrO₂/CC (Fig. 4d, Supplementary Figs. 36 and 37). LSV results show that CC exhibits very poor OER activity (Supplementary Fig. 38), indicating that the catalytic performance is attributable to the MEC-T catalyst. The Δη/Δlog|j| ratio characterizes the kinetic response of overpotential (η) to increasing current density (j), where MEC-T demonstrates significantly lower values than those of IrO₂, indicating its enhanced intrinsic OER activity (Fig. 4e). MEC-T retains its catalytic activity after 50,000 cyclic voltammetry (CV) scans with negligible polarization current decay (Supplementary Fig. 39). Galvanostatic measurements further confirm that MEC-T delivers stable current densities of 0.1–1.0 A·cm⁻² and operates robustly for over 200 h at 0.5 A·cm⁻² with a voltage fluctuation under 2.6% (Fig. 4g and Supplementary Fig. 40). Notably, the cumulative charge transfer (Q) calculated as the product of operational duration and current density is one of the highest among metal-free catalysts (Supplementary Fig. 35). Multiple characterizations after OER indicate that the MEC-T catalyst still maintains an intact morphological structure and stable chemical composition (Supplementary Figs. 41 − 45). Based on its favorable OER performance, we constructed an anion exchange membrane (AEM) electrolyzer with Pt/C and MEC-T serving as the cathode and anode, respectively. The MEC-T||Pt/C configuration outperforms the IrO₂||Pt/C benchmark for overall water splitting, achieving cell voltages of 1.75 ± 0.02 and 1.92 ± 0.02 V at 0.5 and 1.0 A·cm⁻² (Fig. 4f and Supplementary Fig. 46). Moreover, the AEM electrolyzer demonstrates stable operation for 260 h at 0.5 A·cm⁻², while keeping a small voltage fluctuation within 7.8% (Fig. 4h).

Elucidating the reaction pathway

Isotope-labeled operando differential electrochemical mass spectrometry (DEMS) measurements were employed to explore the OER pathway. The initial DEMS measurements were conducted in H₂¹⁸O electrolyte while monitoring the mass signals of evolved oxygen species. During each CV cycle, the MEC-T catalyst produced ³⁶O₂, ³⁴O₂ and ³²O₂ (Fig. 5a), whereas MEC-L produced only ³⁶O₂ and ³⁴O₂ (Supplementary Fig. 47a). The detection of ³²O₂ indicates the coupling of two adsorbed ¹⁶O species on the MEC-T catalyst surface⁴⁶. Subsequently, the above catalysts were rinsed with H₂¹⁶O, and a second set of DEMS experiments was performed in H₂¹⁶O electrolyte. MEC-T continued to generate ³⁶O₂, ³⁴O₂, and ³²O₂ in each CV cycle (Supplementary Fig. 48), whereas MEC-L yielded only ³⁴O₂ and ³²O₂ (Supplementary Fig. 47b). The small quantity of persistent ³⁶O₂ evolution indicates the direct coupling of residual surface ¹⁸O adsorbates on MEC-T. These findings demonstrate that MEC-T facilitates OER by coupling surface-adsorbed oxygen intermediates, suggesting that the process likely proceeds via the OPM or LOM rather than the conventional AEM pathway observed in MEC-L. The kinetic isotope effect (KIE) of COFs was explored through evaluating their OER activities in 1.0 M KOH/H₂O and 1.0 M KOD/D₂O. The KIE values of MEC-T are significantly lower than those of MEC-L (Fig. 5b,c), suggesting a pathway difference and the characteristics of OPM or LOM^47,48. The pH-dependent OER activity assessment can provide further evidence for the reaction pathway. At 1.57 V versus reversible hydrogen electrode (RHE), the proton reaction order ( ρ^H = ∂log ( J)/∂pH) of MEC-T is 0.78 (Fig. 5d and Supplementary Fig. 49), which is much higher than that of MEC-L (−0.025), indicating a strong pH dependence. The results suggest that MEC-T involves a non-synergistic proton-coupled electron transfer (PCET) process and likely follows the OPM or LOM, whereas MEC-L exhibits characteristics consistent with the cooperative PCET process in the AEM^49,50. The tetramethylammonium cation (TMA⁺) can combine with the peroxide (O₂²⁻) intermediate produced via the LOM pathway, thereby inhibiting OER activity. However, TMA⁺ probe experiments show that introducing TMA⁺ fails to affect the activities of MEC-T and MEC-L (Fig. 5e), thus ruling out the LOM pathway. By the way, there is actually no so-called lattice oxygen in the all-organic electrocatalysts.

Fig. 5. Revealing the oxygen-evolving pathway.

a DEMS signals for O₂ products of MEC-T using H₂¹⁸O as the electrolyte during CV scans. b LSV curves of MEC-T and MEC-L in 1.0 M KOH∙H₂O and 1.0 M KOD ∙ D₂O on GCE acquired at 5 mV·s⁻¹. J: current density. c Kinetic isotope effect (KIE) plots. d The proton reaction orders of MEC-T and MEC-L. e LSV curves of MEC-T and MEC-L with or without TMA⁺ on GCE acquired at 5 mV·s⁻¹ in 1.0 M KOH. f Potential-dependent operando ATR-SEIRAS for MEC-T in 1.0 M KOH ∙ H₂O. g Potential-dependent operando ATR-SEIRAS for MEC-T in 1.0 M KOD ∙ D₂O solution. The color bars represent normalized signal intensity, with blue and red indicating low and high absorption intensities. Source data are provided as a Source Data file.

Additionally, the real-time evolution of critical reaction intermediates was monitored by potential-dependent operando attenuated-total-reflection surface enhanced infrared adsorption spectroscopy (ATR-SEIRAS). Unexpectedly, we failed to find the characteristic *OO* signal for the classic OPM pathway from the operando spectra of MEC-T, yet instead a characteristic *OOH vibration band at 1233 cm⁻¹ (Fig. 5f and Supplementary Figs. 50 and 51)^7,51,52. Notably, the intensity of the *OOH characteristic peak in MEC-T was higher than that in MEC-L following the AEM pathway (Supplementary Fig. 52), suggesting that MEC-T is more favorable for *OOH generation or adsorption. Under the AEM and AOPM pathways, the theoretical spectral positions of *OOH are very close (Supplementary Fig. 53), without significant differences, suggesting that their final adsorption states at the carbon sites are virtually indistinguishable. Furthermore, isotope-labeled potential-dependent operando ATR-SEIRAS reveal that MEC-T exhibits a characteristic *OOD vibration band at 1213 cm⁻¹ (Fig. 5g and Supplementary Fig. 54), verifying the reliability of this reaction intermediate. These observations indicate that the OER mechanism on MEC-T differs from both conventional OPM and AEM pathways. We designate this mechanism as AOPM, which is characterized by the formation of OOH intermediate through an intramolecular asymmetrically nucleophilic coupling.

Insights into OER mechanism

In view of the above, DFT calculations were undertaken to unravel the catalytic mechanisms of the proposed electrocatalysts. Supplementary Data 1 contains the atomic coordinates of the optimized models. Firstly, we plotted the OER free energy profiles of the potential active sites on MEC-T and MEC-L via a classical AEM pathway (Supplementary Figs. 55−58). For MEC-T, both the hydroxyl group and the O site on the oxazole ring are thermodynamically unfavorable for initial OH adsorption, speculating them as inactive sites (Supplementary Fig. 59 and 60). Although the N site can adsorb OH, its ΔG for RDS is as high as 1.894 eV, suggesting that it may be less suitable as an effective active site (Fig. 6a). Under this pathway, the C_c and C_a sites are the plausible active centers for MEC-T and MEC-L, with corresponding free energy changes (ΔG) of 1.701 and 1.700 eV for the rate-determining step (RDS), respectively (Fig. 6a and Supplementary Fig. 61). When following the OPM pathway, the favorable dual-active sites for MEC-T and MEC-L are likely two symmetrical C_b sites on para position of the benzene ring. On this carbon atom site, the ΔG for *OH generation by MEC-L is 2.121 eV (Supplementary Figs. 62 and 63), while that of MEC-T is only 1.382 eV, supporting more favorable *OH adsorption on the latter. However, MEC-T exhibits a much higher transition state (TS) energy barrier (3.090 eV) for O₂ formation compared with that of MEC-L (1.826 eV) (Fig. 6c,d and Supplementary Fig. 64), suggesting that its excessively strong (*O + *O) adsorption may inhibit the desorption process. Therefore, it is inferred that direct *O−O* coupling is unfavorable at these symmetrical dual C_b sites on MEC-T. Finally, we calculated the free energy changes for MEC-T and MEC-L based on the AOPM pathway consisting of five elementary steps. For MEC-T, OH⁻ adsorbs onto the C_b active site and loses an electron to form *OH (Supplementary Fig. 65). Subsequently, the adsorbed *OH undergoes proton-coupled electron transfer (PCET), producing *O. Then, a second OH⁻ adsorbs at the C_a site and loses an electron, forming adjacent *O and *OH pairs. The interatomic distance of 2.240 Å between C_a and C_b sites on MEC-T may facilitate the intramolecular nucleophilic attack by *OH (Fig. 6c). Thus, the *OH adsorbed on C_a site undergoes intramolecular migration to the *O on C_b site, generating *OOH (Fig. 6d) with a TS energy barrier of 1.490 eV as the RDS (Fig. 6c). Finally, the *OOH releases O₂ via a PCET process. For comparison, such TS energy barrier increases to 2.627 or 2.301 eV on the dual C_b + C_b’ or C_a + N sites, which could be attributed to the stronger adsorption of OH at these sites for preventing the desorption (Supplementary Figs. 66−69). Additionally, the ΔG for RDS of the C_a + C_c dual-site is 1.603 eV, which is also higher than that of C_a + C_b. Accordingly, the asymmetric C_a + C_b dual-site configuration shows optimal thermodynamic and kinetic advantages under the AOPM pathway, exhibiting significantly lower energy barriers compared with both AEM and OPM pathways (Supplementary Figs. 70 and 71). For MEC-L, the *OH formation step in AOPM remains energetically unfavorable with a high ΔG of 2.121 eV (Fig. 6c and Supplementary Fig. 72), suggesting that it may follow the AEM pathway. Notably, the identification of active sites relies primarily on theoretical calculations and is inevitably speculative.

Fig. 6. Probing the mechanism for OER.

a Free energy changes (ΔG) of the rate-determining step (RDS) at distinct carbon sites and nitrogen site on MEC-T following the AEM pathway (inset: structure of MEC-T and the spatial distribution of its potential active sites). b ΔG of the RDS at potential active sites on MEC-T following the AOPM pathway (inset: structure of MEC-T and the spatial distribution of its potential active sites). c Gibbs free energy plots for MEC-L and MEC-T following the AOPM and OPM pathways (inset: structure of MEC-T and the interatomic distance between C_a and C_b sites). d Optimized structures for principal intermediates and transition state (TS) of OPM (up) and AOPM (down) on MEC-T. Color codes: black, C; red, O; blue, N; white, H. e Methanol-modulated LSV curves (without iR compensations) for MEC-L and MEC-T on GCE acquired at 5 mV·s⁻¹ in 1.0 M KOH. The discrepancy between the methanol-added (red lines) and neat methanol (blue lines) curves (filled area) is ascribed to the MOR. f The distribution of p-orbital PDOS for C_a and C_b atoms on MEC-L and MEC-T, respectively. Their corresponding p-band centers were labeled, ɛ_p: p-band center. g Differential charge density plot of MEC-T. The blue and red regions correspond to electron accumulation and depletion after introducing of oxygen groups on the benzene rings, respectively. h Average Bader charge of C atoms for MEC-T. i Average Bader charge of C atoms for MEC-L. Source data are provided as a Source Data file.

Methanol oxidation reaction (MOR) experiments illustrate that the difference between MOR and OER activities of MEC-T is remarkably larger than that of MEC-L, confirming that MEC-T has a stronger *OH adsorption ability (Fig. 6e)^53,54. Besides, partial density of states (PDOS) of carbon p-orbitals in MEC-T and MEC-L were analyzed. Compared with MEC-L, the more distal positioning of the carbon p-band center from the Fermi level in MEC-T yields more electron-deficient sites (Fig. 6f, Supplementary Figs. 73 and 74), which strengthens their adsorption affinity of oxygen intermediates^55,56. Moreover, differential charge density plot also reveals a depletion of electron density on the carbon atoms in MEC-T after oxygen group introduction, being agreement with PDOS analysis and the experimental results (Fig. 6g). The Bader charges of the carbon atoms in MEC-T and MEC-L are +0.470 |e| and +0.309 |e|, respectively, indicating that the hydroxyl groups contribute to the electron deficiency of the carbon atoms (Fig. 6h,i and Supplementary Fig. 75). Additionally, the oxygen atom in the benzoxazole can also significantly enhance the positive charge on the carbon atoms (Supplementary Fig. 76). Therefore, hydroxyl functionalization could promote a complete shift of the catalytic pathway from AEM to AOPM by re-adjust the intermediate adsorption at carbon active sites. Based on experimental results and DFT calculations, the most reasonable active centers are proposed as the C_a and C_b dual sites in MEC-T. However, the possibility of direct participation of N/O atoms has not been absolutely ruled out. The main role of N/O atoms is likely to modulate the electronic structure of the carbon sites, although they also undergo significant electronic perturbations.

In summary, we present a molecular modularization strategy to boost the OER performance of AOEs by precisely tailoring hydroxyl groups on the organic frameworks. The optimized all-organic MEC-T shows an overpotential of 326.0 ± 2.0 mV at 0.5 A·cm⁻², while maintaining catalytic stability for over 200 h, representing a competitive metal-free electrocatalysts. Multiple operando characterizations combined with theoretical calculations demonstrate that hydroxyl functionalization alters the electronic structure of carbon sites, revolutionizing the reaction pathway. The identified AOPM requires asymmetric dual active sites and forms the *OOH intermediate through intramolecular nucleophilic attack, demonstrating favorable thermodynamic and kinetic advantages over both the single-active-site dependent AEM and symmetric dual-active-site mediated OPM. This work provides a perspective on developing efficient AOEs by engineering reaction pathways. Nevertheless, the universality and predictability of AOPM need to be validated across a broader range of material systems in the future.

Methods

Materials

2,4,6-triformylphloroglucinol (97%), 1,3,5-triformylbenzene (96%), tetrahydrofuran (THF, AR, 99%), tetramethylammonium chloride (≥99%), isopropyl alcohol (≥99%), carbon nanotubes (CNTs, ca. 50 μm in length, 8-15 nm in diameter), 2,5-diamino-1,4-benzenediol dihydrochloride (97%) and N, N-dimethylformamide (DMF, 99%) were purchased from Aladdin. IrO₂ (Ir ≥85.5%, 99.99% metals basis), benzimidazole (96%), N-methyl pyrrolidone (NMP, 99%), potassium hydroxide (KOH, 99.99%), mesitylene (98%, water ≤ 50 ppm), and Nafion 117 solution (ca. 5% in water/lower aliphatic alcohol mixture) were purchased from Energy Chemical. The carbon cloth (CC, Phychemi, model WOS1009) was sequentially cleaned with acetone and ethanol, followed by refluxing in nitric acid at 100 °C for 2 h, neutralized by rinsing, and vacuum-dried at 60 °C. D₂O (99.9%) was supplied by Anergy Reagents Co., Ltd. A Milli-Q IQ7003 system was applied to produce ultrapure water having a resistivity of ≥18 MΩ·cm. Unless otherwise specified, all electrodes used in this work were supplied by Gaoshi Ruilian Science and Technology Co., Ltd.

Catalyst synthesis

Synthesis of MEC-L, MEC-L₂T₁, MEC-L₁T₂, and MEC-T. A Pyrex tube was charged with benzene-1,3,5-tricarboxaldehyde (BT, 0.335 mmol), 2,5-diamino-1,4-benzenediol dihydrochloride (0.50 mmol), benzimidazole (1.50 mmol), mesitylene (2.25 mL), and NMP (2.25 mL). After uniform sonication, the tubes were placed in liquid nitrogen for rapid freezing, followed by three freeze-extract-thaw cycles for degassing. Finally, they were sealed and heated in a constant-temperature reaction oven at 185 °C for 5 days. After completion, the solid was isolated via centrifugation, followed by sequential washing with acetone and THF. Residual monomers and oligomers were removed by Soxhlet extraction using DMF as the solvent for 48 h. The resulting powder was subjected to methanol washing and lyophilization, yielding MEC-L. Similarly, MEC-L₂T₁, MEC-L₁T₂, and MEC-T were synthesized by replacing the monomer benzene-1,3,5-tricarboxaldehyde with trial-aldehyde phloroglucinol (TP) at molar substitution ratios of 33%, 67%, and 100%, respectively. MEC-T@CNT was prepared by adding 30 mg of CNTs to the reaction system used to synthesize MEC-T.

Catalyst characterizations

A Rigaku MiniFlex600 diffractometer was employed to acquire powder X-ray diffraction (PXRD) data. A JEM2100F transmission electron microscope (TEM) and a Zeiss EVO 10 scanning electron microscope (SEM) were applied for morphological characterization. The catalyst thickness was assessed via Bruker Dimension ICON atomic force microscope (AFM). An Escalab 250Xi instrument was employed to perform X-ray photoelectron spectroscopy (XPS) measurements. The Nicolet 6700 Flex (Thermo Fisher) was applied to acquire Fourier transform infrared (FT-IR) spectra. Further mechanistic investigation was conducted via operando ATR-SEIRAS on a Nicolet iS50 FT-IR platform (Supplementary Fig. 50). The test was carried out in 1.0 M KOH·H₂O or KOD·D₂O electrolyte by applying potentials between 1.1 and 1.5 V versus RHE. Thermal stability assessment was performed via thermogravimetric analysis (TGA, TA Q50 instrument) under N₂. A Bruker AM-400 nuclear magnetic resonance (NMR) spectrometer was employed to acquire the solid-state ¹³C cross-polarization magic angle spinning (CP-MAS) NMR spectra. N₂ sorption isotherms were acquired on a Micrometrics ASAP 2020 Plus HD 88 analyzer, with samples pretreated by degassing under high vacuum at 200 °C for 12 h. The soft X-ray absorption spectra at the C-K, N-K and O-K edges were measured at TLS 20 A of the National Synchrotron Radiation Research Center in Hsinchu using fluorescence yield mode. A QAS 100 differential electrochemical mass spectrometer (Supplementary Fig. 77), supplied by Shanghai LingLu Instruments Company, was used for isotope-resolved operando DEMS experiments. The DEMS system consists of a constant-pressure electrochemical cell and a high-vacuum mass spectrometer, capable of real-time tracking of gas-phase products via pressure-driven sampling. The three-electrode system comprises a counter electrode (Pt wire), a working electrode, and a reference electrode (Ag/AgCl). These electrodes are all immersed in 1.0 M KOH saturated with argon. The working electrode preparation procedure is as follows: a thin gold film was deposited onto a porous polytetrafluoroethylene membrane (pore size ≤20 nm, porosity ≥50%) using magnetron sputtering, then the catalyst is drop-cast. First, cyclic voltammetry (CV) testing of the electrocatalyst was performed in 1.0 M KOH prepared with H₂¹⁶O. The gaseous products were monitored in real time with the potential window set to 1.1–1.7 V versus RHE at 5 mV·s⁻¹. After thorough rinsing the isotope-labeled catalyst with H₂¹⁶O, subsequent CV scans were recorded in 1.0 M KOH prepared with H₂¹⁶O, followed by gas product analysis.

Electrochemical measurements

Electrocatalytic testing was conducted in 1.0 M KOH electrolyte using a CHI760E workstation equipped with a three-electrode setup. The measurement data were automatically acquired from the instruments. Origin software was used to process data and generate charts. Preparation and storage of electrolyte: before each experiment, we dissolved 11.2 g of solid KOH with ultrapure water, diluted it to a total volume of 200 mL to prepare a fresh 1.0 M KOH electrolyte (pH = 13.6 ± 0.2), and stored it in a clean polyethylene bottle. The electrochemical setup featured a counter electrode (platinum plate) and a reference electrode (Hg/HgO). The calibration of the Hg/HgO reference electrode was performed with respect to the Ag/AgCl electrode. Specifically, the open-circuit potential is recorded in 1.0 M KOH with saturated argon for 1 h. Working electrodes were produced via depositing the catalyst on a 5 mm diameter glassy carbon electrode (GCE) or a 0.2 cm² CC substrate. To improve the conductivity of COFs, Ketjenblack was employed as a conductive additive. Catalyst ink was acquired via ultrasonicating the catalyst powder (3 mg) with Ketjenblack conductive additive (1 mg) in a solvent blend of isopropanol (500 μL), water (100 μL), and Nafion binder (5 wt%, 50 μL). The resulting homogeneous suspension was drop-cast in volumes of 10 µL and 60 µL onto GCE and CC substrates, respectively. The catalyst loadings on GCE and CC were 0.23 mg·cm⁻² and 1.38 mg·cm⁻². The anion exchange membrane (AEM) electrolyzer was equipped with an anion exchange membrane (Fumasep FAA-3-PK-130) with a thickness of 130 μm and a size of 2.0 × 2.0 cm. Prior to use, the membrane was pre-treated by immersion in 1.0 M KOH for one day. The anode was a CC coated with IrO₂ or MEC-T, and the cathode was a CC coated with Pt/C. The mass loading of all catalysts was fixed at 1.38 mg·cm⁻². The assembled modules were installed in the AEM flow cell. The apparatus is continuously supplied with 1.0 M KOH and operated at 25 °C. Prior to performance evaluation, the working electrode was conditioned through repeated CV cycles to reach a steady-state CV profile. Linear sweep voltammetry (LSV) measurements were conducted at 5 mV·s⁻¹, applying 95% iR compensation to GCE and 85% to CC. Electrochemical impedance spectroscopy (EIS) spectra were acquired at 1.50 V versus RHE in an O₂-saturated electrolyte from 0.01 to 10⁵Hz. ZView software was employed to fit the test data and derive the related parameters. The long-term durability was evaluated by chronoamperometric measurements.

The calculation formula for catalyst loading is as follows:

$$m_{\mathrm{catalyst}}=C\times V_{\mathrm{dispensed}}\times \frac{m_{\mathrm{catalyst},\mathrm{powder}}}{m_{\mathrm{catalyst},\mathrm{powder}}+m_{\mathrm{additive}}}$$ 1

$$\mathrm{Catalystloading}=\frac{m_{\mathrm{catalyst}}}{A_{\mathrm{geo}}}$$ 2

Where m_catalyst is the catalyst loading on the electrode; C is the total solid concentration of the catalyst ink; V_dispensed denotes the volume of ink drop-cast onto the electrode; A_geo represents the effective geometric area of the working electrode.

We determined the electrochemical active surface area (ECSA) via the double-layer capacitance (C_dl) obtained at scan rates of 20 − 120 mV·s⁻¹ within a non-Faradaic window of 1.0−1.1 V versus RHE.

ECSA values were derived from the equation:

$$\mathrm{ECSA}=\frac{C_{\mathrm{dl}}}{C_{\mathrm{s}}}$$ 3

A specific capacitance (C_s) value of 0.04 mF. cm^-2 was adopted, consistent with typical values reported in the literature.

The RHE scale was used for potential calibration, employing the formula:

$$E_{\mathrm{RHE}}=E+E_{\mathrm{Hg}/\mathrm{HgO}}+0.0591\times \mathrm{pH}$$ 4

The potential of the Hg/HgO reference electrode is 0.098 V vs. standard hydrogen electrode at 25 °C in 1.0 M KOH solution.

We calculated the Faradaic efficiency (FE) according to the equation:

$$\mathrm{FE}=\frac{I_{\mathrm{ring}}}{\epsilon \times I_{\mathrm{disk}}}$$ 5

Where I_ring denotes the Pt ring electrode current collected at 0.4 V versus RHE; ε represents the device’s oxygen collection efficiency, valued at approximately 0.37; I_disk represents the constant disk current. The Faraday efficiency was measured only once.

We determined the apparent activation energy (E_a) by applying the Arrhenius formula:

$$\frac{\mathrm{d}\left(\log j_0\right)}{\mathrm{d}\left(1/T\right)}=-\frac{E_{\mathrm{a},\mathrm{app}}}{2.303R}$$ 6

The exchange current density (j₀) in the formula, sourced from the Tafel plot, was measured based on the LSV curves of the GCE-supported catalysts at 20, 35, 50, and 65 °C. T represents the temperature in Kelvin, and R is the gas constant.

The following equation was employed to determine the catalysts’ relative turnover frequency (TOF):

$$\mathrm{TOF}=\frac{J\times N_A}{n\times F\times \tau }$$ 7

n designates the electron transfer number, which equals 4 in the context of OER; J denotes the current density, and its value used for TOF calculation was obtained at 1.53 V versus RHE; N_A is the Avogadro number, with a value of 6.022 × 10²³; F denotes the Faraday constant; τ refers to the count of active sites of OER, determined from the following equation¹²:

$$\tau =\frac{\mathrm{area}\mathrm{of}\mathrm{the}{C}_{\mathrm{dl}}\mathrm{curve}}{\mathrm{scan}\mathrm{rate}\times 1.601\times {10}^{-19}}$$ 8

The constant used in the formula is the elementary charge, and its unit is the coulomb (C).

pH-dependent reaction order

To determine the pH-dependent reaction order for OER, electrochemical measurements on the catalysts were performed in KOH solutions with pH values of 13.0, 13.5 and 14.0. The pH-dependent reaction order was then calculated using the following formula:

$${\rho }^{\mathrm{RHE}}=\frac{\partial \log \left(j\right)}{\partial \mathrm{p}H}$$ 9

Kinetic isotope effect (KIE)

KIE were studied via electrochemical methods and the experiments were carried out in 1.0 M KOH aqueous solution and 1.0 M KOD D₂O solution. The overpotentials were determined by the formula:

$${\eta }^{{\mathrm{H}}_2\mathrm{O}}=E_{\mathrm{RHE}}-1\mathrm{.229V}$$ 10

$${\eta }^{{\mathrm{D}}_2\mathrm{O}}=E_{\mathrm{RDE}}-1\mathrm{.262V}$$ 11

The KIE was defined as the formula:

$$\mathrm{KIE}={∣\frac{j_{{\mathrm{H}}_2\mathrm{O}}}{j_{{\mathrm{D}}_2\mathrm{O}}}∣}_{\eta }$$ 12

The current densities in H₂O and D₂O should be compared at the same overpotential.

DFT calculations

In Materials Studio, the Cambridge Sequential Total Energy Package (CASTEP) module was utilized to conduct DFT calculations, including geometric optimization, electronic properties, and energy relaxation. The simulations employed spin-polarized DFT based on a plane-wave pseudopotential (PWP) approach. Exchange–correlation functions were addressed through the generalized gradient approximation (GGA) adopting the Perdew–Burke–Ernzerhof (PBE) functional. Structural minimization was accomplished using the Broyden–Fletcher–Goldfarb–Shanno (BFGS) algorithm. The final localization to the transition state is achieved by executing a Linear Synchronous Transition (LST) and subsequently performing multiple conjugate gradient minimization with Quadratic Synchronous Transition (QST) maximization. The plane-wave kinetic energy cutoff was fixed at 500 eV. For geometry optimization, the self-consistent field (SCF) convergence threshold was fixed at 1 × 10⁻⁶ eV/atom, with hybrid density applied during electronic minimization. The convergence threshold was adjusted as follows: total energy ≤1  × 10⁻⁵ eV·atom⁻¹, stress ≤0.05 GPa, maximum force ≤0.03 eV·Å⁻¹, and maximum atomic displacement ≤0.001 Å.

Author contributions

C.T.H. conceived the research, planned its methodology, and supervised the project. L.H.Y. and L.M.C. conducted the experimental work and data analysis. Z.H. and S.C.H. conducted XANES measurements. C.T.H. carried out theoretical calculations. C.T.H., L.M.C. and Z. H. drafted the manuscript. X.F.Z., L.D.W., J.Y. and Y. D. participated in discussions and provided critical feedback. All authors discussed and commented on the manuscript.

Peer review

Peer review information

Nature Communications thanks Yizhong Lu, Yongping Zheng and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

C.T.H. discloses support for the research of this work from the National Natural Science Foundation of China [grant numbers 22375079, 22161021], the Jiangxi Provincial Natural Science Foundation [grant numbers 20224ACB214001, 20242BAB26033] and the Jiangxi provincial government [grant number 0224/09039001]. L.M.C. discloses support for publication of this work from the National Natural Science Foundation of China [grant numbers 22379057, 22265012] and the Jiangxi Provincial Natural Science Foundation [grant numbers 20232ACB213009, 20242BAB26046, 20243BCE51069]. All other authors declare no relevant funding.

Data availability

All data are available in this published article and its Supplementary Information and Source Data files. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-75304-1.