Internal hydrogen-bond enhanced two-electron oxygen reduction reaction for π-d conjugated metal-organic framework to H₂O₂ synthesis

Abstract

Tailoring the electronic structure of electrocatalysts for oxygen reduction reaction (ORR) has been widely adopted to optimize their performance. However, the steric effect originating from the layered or crystal structure of a catalyst is often neglected. Herein, we demonstrate the importance of such steric effect in a one-dimensional π-d conjugated metal-organic framework with Ni-(NH)₄ nodes (Ni-BTA) for optimizing its electrocatalytic performance, where the activity and selectivity towards two-electron ORR for H₂O₂ production are largely enhanced. Theoretical simulation and in-situ characterization demonstrate the formation of inter-layer H-bonds between *OOH intermediates and -N-H groups in the adjacent top layers of the Ni-sites, enhancing the *OOH binding energy to an optimum value. Thus, the as-prepared Ni-BTA catalyst exhibits an outstanding electrocatalytic 2e⁻-ORR performances under neutral and alkaline conditions (e.g., >85% H₂O₂ selectivity from −0.1-0.4 V vs. RHE and >13.5 mol g⁻¹ h⁻¹ H₂O₂ yield in neutral electrolytes), also showing great potential on water treatment and disinfection. Here, we highlight the alternative avenues for utilizing the non-coordinated structure to regulate the catalytic performance, thus providing opportunities for the design of catalysts and beyond.

Introduction

Hydrogen peroxide (H₂O₂) is an environment-friendly oxidant and disinfectant, and approximately 6.0 million metric tons production is required annually by 2024 to meet the demand of chemical and medical industries^1,2,3. So far, over 95% of H₂O₂ is synthesized by the energy-intensive anthraquinone process, but this method suffers from serious safety risks and environmental pollutions^4,5. Fortunately, electrocatalytic synthesis of H₂O₂, especially via the two-electron oxygen reduction reaction (2e⁻-ORR), offers an alternative route to produce H₂O₂ from O₂ and H₂O, with intrinsic high-safety, green, and on-site features^6,7. In particular, the electrosynthesis of H₂O₂ in a wide pH range from alkaline to neutral conditions is especially desirable for practical applications, such as water treatment, bleaching and disinfection.

Carbon-based materials with single-atom metallic active sites (SACs) are widely utilized as ORR catalysts, and previous works have been majorly focused on tailoring the electronic structure (i.e., the d-band center of the metal centers) of the SACs to alter *OOH intermediate adsorption for achieving both high ORR activity and H₂O₂ selectivity^8,9. Strategies for this purpose include adjusting the central metal atoms¹⁰, engineering the coordination atoms in the first/second coordination spheres⁷, and creating a strong metal-support interaction (e.g., π-π interactions) by loading the SACs onto an oxide surface (e.g., Co-phthalocyanine/O-CNTs)^11,12, etc. In these works, a single-layer graphene sheet containing specific single-atom metal sites is commonly adopted as the simulation object to investigate the relevant mechanisms from a theoretical perspective. Although it is typically capable of providing a reasonable explanation for the experimental phenomena, however, the synergistic effects brought about by the layered or crystal structure of the catalysts is often ignored. Taking graphene or layered metal organic framework materials (e.g., Cu-HHTP¹³, HHTP = 2,3,6,7,10,11-hexahydroxytriphenylene) as an example, their interlayer spacing is approximately 0.3–0.4 nm, and the catalytic activity of the metal center would be obviously affected by the adjacent layer under with such a small distance besides its intralayer coordination environment, but it is rarely reported.

*OOH, the only and key intermediate during 2e⁻-ORR, could form a non-covalent hydrogen-bond (H-bond) interaction with the -NH-containing protic ionic liquid electrolyte (-N-H⁺⋯O-OH _Pt/Au) on the surface of Pt/Au catalysts, and this interfacial H-bond interaction largely enhanced the catalyst’s 4e⁻-ORR kinetics¹⁴. In addition, such interfacial H-bond could also be formed between different reaction intermediates adsorbed on the metal sites, and various organic ligands (e.g., N-methylimidazoles¹⁵, cyste-amine¹⁶) decorated on the surface of metal-based catalysts, to optimize the semi-hydrogenation, hydrogen evolution and oxidation catalytic reactions, etc. Similarly, the internal component or functional groups (e.g., -NH₂/-NH-, -OH, and -COOH) in the SACs or MOF-based catalysts, which exist in an out-of-plane manner (or in the adjacent layer), as well with an appropriate distance of ~0.3–0.5 nm to the exposed metal sites, may also possibly act as the same role of the aforementioned surface ligands to form H-bonds with the adsorbed *OOH intermediates, and subsequently enhance materials’ 2e⁻-ORR performance. Moreover, such functional groups suggest a very stable and precise structure, compared to those surface-decorated ligands, thus delivering a precise adjustment ability to the catalytic reactions while enabling the repeated formation/cleavage of the H-bond interactions for a long period of catalysis process by avoiding the dissolution of surfactants and surface ligands on the surface of catalysts. Based on these considerations, the adoption of such performance-enhance mechanism via the induced H-bond interactions, would provide an additional dimension for the design of state-of-the-art electrocatalysts for 2e⁻-ORR, which, unfortunately, has long been neglected.

Herein, a layered one-dimensional π-d conjugated metal-organic framework with Ni-(NH)₄ sites (Ni-tetra-aminobenzene, shorted as Ni-BTA) is adopted as an example to clarify the steric effect of internal H-bond on optimizing 2e⁻-ORR performance. Both theoretical simulation and in-situ characterization demonstrate the formation of H-bonds between the absorbed *OOH intermediates on Ni-sites with the top-layer -NH- groups in Ni-BTA, which, synergistically with the axial coordination structure (axial-OH) on Ni-sites, effectively enhance the *OOH binding energy close to the optimum value (~4.23 eV). Thus, the as-prepared Ni-BTA catalyst exhibits good 2e⁻-ORR performance under both neutral and alkaline conditions, including the high H₂O₂ selectivity (alkaline:>90%, neutral:>80%), high H₂O₂ yield (alkaline:>27.0 mol g⁻¹ h⁻¹, neutral:>13.5 mol g⁻¹ h⁻¹), and long stability (alkaline:>40 h, neutral:>24 h), thus endowing the electro-synthesized H₂O₂ with good dye degradation and antibacterial performance. This study demonstrates the major role of H-bonds induced by the layered structure of SACs, a typical non-coordination structure in the materials, on regulating the catalytic performance, which provides insights for the design of state-of-the-art catalysts considering into their non-coordination structure for 2e⁻-ORR and beyond.

Results

Theoretical predictions of layered Ni-BTA for ORR

Typically, the coordination configuration of the metal-based active sites, such as the metal centers¹⁰ and the non-metal ligands in their first or second coordination sphere⁷, is commonly considered when evaluating the catalytic capability of a metal-based electrocatalyst. To validate the above hypothesis of utilizing the H-bond interactions, we herein focus on investigating the potential non-coordination regulation of adjacent layers in the one-dimensional conductive metal-organic framework (1D c-MOF) for tailoring its ORR performance, which has been seldom investigated previously.

Ni-BTA, a typical 1D c-MOF with a layered structure and abundant Ni-(NH)₄ moieties, was chosen as a model object for the investigation in our study. According to the density functional theory (DFT) simulations, the Ni-center in a single Ni-BTA chain (s-Ni) shows a low *OOH adsorption ability with a high *OOH adsorption energy of 4.82 eV (ΔG_*OOH = 4.82 eV, Fig. 1a, b and Supplementary Data 1), indicating its low activity but high selectivity for 2e⁻-ORR. Replacing the Ni sites in s-Ni by other typical transition metal species (i.e., Fe, Mn, Cu, and Co) could significantly alter their *OOH adsorption ability (i.e., 3.70 eV for s-Mn, 3.73 eV for s-Fe, 4.30 eV for s-Co, and 5.05 eV for s-Cu), which could be attributed to the varying intrinsic electronic structure and oxygen affinity among different metals. Similarly, the out-of-plane coordination of the metal center in a single Ni-BTA chain or the analogs with typical functional groups, such as -OH for the simulation of the real scenario of typical fabrication in concentrated alkaline solution^17,18, could also modify the *OOH adsorption ability of metal centers (e.g., a lower ΔG_*OOH of 4.49 eV for s-OH-Ni). However, no single modification is capable of relocating the ΔG_*OOH to the top of the volcano plot for achieving the both high 2e⁻-ORR activity and selectivity.

Fig. 1: Theoretical predictions of Ni-BTA for ORR.

a The proposed Ni-(NH)₄ coordination structure with coordinated axial-OH and dual-Ni-BTA molecular layer (d-OH-Ni) in Ni-BTA (the brown, light blue, gray, pink and red colored balls represent C, N, Ni, H and O atoms, respectively). b The volcano plot depicting the adsorption energies of *OOH intermediate on different M-(NH)₄ coordination structures of Ni-BTA and other M-BTA samples (M=Mn, Fe, Co, and Cu). c Calculated *OOH adsorption energies and relative charge states, d-band center values of the nickel metal center in Ni-BTA with different Ni-(NH)₄ coordination structures. d The N-H bond length and the charge state of H atom in -NH- group of the top Ni-BTA layer before and after *OOH adsorption on the Ni-center in the bottom Ni-BTA layer. e Differential charge distribution on dual-layer Ni-(NH)₄ coordination structure with/without axial-oxygen before and after the adsorption of *OOH. The isosurface value is set to be 0.001 e Å⁻³. Yellow and cyan isosurfaces (±0.001 e Å⁻³) show the electron gain and electron loss, respectively.

When taking into account the interaction between adjacent Ni-BTA chains, which have an interlayer spacing of ~0.35 nm¹⁸, the Ni-centers in dual-layer Ni-BTA chains with/without axial -OH coordination (d-Ni and d-OH-Ni) both show enhanced *OOH adsorption capability (Fig. 1b and Supplementary Data 1), indicating the positive effect of the multi-layer structure of Ni-BTA on tailoring its ORR activity. Specifically, d-OH-Ni delivers a ΔG_*OOH of 4.21 eV, which is very close to the top of the volcano plot for 2e⁻-ORR and corresponds to a small overpotential of 0.02 eV for 2e⁻-ORR (Supplementary Fig. 1). Consequently, the Ni-(NH)₄ moieties in d-OH-Ni theoretically possess the potential to achieve high activity and selectivity for 2e⁻-ORR, simultaneously. In comparison, d-OH-Mn/Fe/Co deliver too low ΔG_*OOH values, while the d-OH-Cu shows a too high ΔG_*OOH value, making them unsuitable for achieving high-efficiency 2e⁻-ORR (Supplementary Fig. 2). Additionally, the calculated ΔG_*OOH under s-M, s-OH-M, d-M and d-OH-M configurations for Mn/Fe/Co/Cu are also provided in Supplementary Fig. 2 as comparisons to Ni-BTA. Especially, the Ni-(NH)₄-OH moieties in dual-layer Ni-BTA chains with different structures and numbers for axial-OH all suggest optimized ΔG_*OOH values that are closer to 4.23 eV than the initial s-Ni (Supplementary Fig. 3), further indicating the positive effect of layered-structure in Ni-BTA on tailoring the ORR performance.

To clarify the intrinsic mechanism of dual-layer structure of Ni-BTA chains on enhancing the adsorption capability of *OOH over Ni-(NH)₄ moieties, the dependence of ΔG_*OOH on the relative charge states, and d-band center values of the Ni sites in the analog configurations are further studied (Fig. 1c). Compared with the single-layer Ni-BTA analogs (s-Ni and s-OH-Ni), the dual-layer counterparts (d-Ni and d-OH-Ni) show a decreased charge states and increased d-band center values of the Ni sites (Supplementary Figs. 4, 5), which would result in a stronger *OOH adsorption ability^8,19. This modification of the dual-layer analogs’ electronic structures may be attributed to the π-π interactions between the adjacent layers, which has also been observed in other similar systems, such as the phthalocyanine/graphene heterostructures¹², but such small change of electronic structure should not lead to so obvious increase on the ΔG_*OOH (i.e., from 4.82 eV of s-Ni to 4.57 eV for d-Ni). To further clarify the internal mechanism, various analysis of the adsorption performance of *OOH on the Ni sites in dual-layer d-Ni and d-OH-Ni was carried out. As shown in Fig. 1d, an elongation in the -N-H bond length (1.04 and 1.03 Å vs. 1.02 Å) and a deficiency of electrons on the H atoms in -N-H groups (0.55 e⁻ vs. 0.45 e⁻ for d-Ni and 0.52 e⁻ vs. 0.48 e⁻ for d-OH-Ni) are observed on the top-layer -N-H groups after the adsorption of *OOH on the Ni sites in the bottom layer, confirming the electron transfer and interaction between the absorbed *OOH intermediates and the H atoms on the adjacent Ni-BTA chain. In addition, it is worth noting that the bond length between the bottom O atom in the adsorbed *OOH species and the H atom in the adjacent -N-H- group is ~1.9-2.0 Å, further confirming the formation of the H-bond between the *OOH intermediates and the H atoms on the adjacent Ni-BTA chain (Fig. 1e). Based on these results, it is theoretically predicted that the adsorption of *OOH on the Ni sites in one Ni-BTA chain would be further enhanced due to the formation of H-bonds between the adsorbed *OOH and the H atoms in the neighboring Ni-BTA chains, thus effectively promoting its 2e⁻-ORR catalytic performance.

To sum up, for realizing the optimal adsorption ability to *OOH intermediates for d-OH-Ni moieties in our study, the intrinsic oxygen affinity among different metals is the premise and the H-bond effect is an effective assistance, which collaboratively contribute to the optimal adsorption ability to *OOH intermediates of the Ni site in Ni-BTA and endow it huge potential on achieving high 2e⁻-ORR performance towards H₂O₂ electrosynthesis.

The experimental synthesis and characterization of Ni-BTA

As guided by the theoretical predictions, a series of M-BTA analogs (M=Ni, Mn, Fe, Co, Cu) were prepared by the coordinating polymerization of 1, 2, 4, 5-benzenetetramine (BTA) with different metal ions in the presence of ammonia. During the fabrication, deprotonated -NH₂ groups (-NH-) in BTA were firstly coordinated with Ni²⁺ or other metal ions in a square and planar manner, forming one-dimensional (1D) M-BTA chains with abundant π-π/π-d conjugated structures (Fig. 2a, b). Subsequently, the M-BTA chains assembled into a herringbone structure via interlayer stacking (Fig. 2c), as confirmed in the previous works^18,20.

Fig. 2: The synthesis and characterization of Ni-BTA.

Synthetic illustration (a), molecular (b) and crystal structure (c) of Ni-BTA (the brown, light blue, gray, pink and red colored balls represent C, N, Ni, H and O atoms, respectively). d Experimental and Pawley-refined XRD patterns of Ni-BTA. e The Ni K-edge XANES spectra, f k²-weighted EXAFS spectra of the Ni-BTA and reference samples. g The corresponding EXAFS fitting curve at R-space of Ni-BTA, with the schematic model (Ni-N₄O_1.5). TEM images (h) and the elemental mappings (i) of Ni-BTA powders. Scale bars, 50 nm (h), 400 nm (i).

The XRD pattern and the Pawley-refined results confirm the typical fishbone-like stacking structure for Ni-BTA, with three distinct peaks at 2θ = 20.5, 23.6, and 29.0° (Fig. 2d, Supplementary Fig. 6 and Supplementary Table 1), similar to the previously reported structure^18,20, which provides direct evidence on confirming the rationality of the proposed d-/d-OH-M models. Although the Mn/Fe/Co/Cu-BTAs deliver a low crystallinity, they still show similar XRD peaks with the Ni-BTA, indicating the successful preparation of the other M-BTAs (Supplementary Fig. 7). In the corresponding Fourier-transform infrared (FTIR) spectra of the M-BTA analogs (Supplementary Fig. 8), an emerging peak around 3300 cm⁻¹ could be found, which can be assigned to -N-H stretching vibration of -NH- groups. On the contrary, the BTA precursor only featured a symmetric and antisymmetric N-H stretching at 3300-3500 cm⁻¹ associated with the -NH₂ groups. In addition, the existence of -C = N species at 1500-1700 cm⁻¹ in the FTIR spectra and the metal-N (-M-N) species at 600-700 cm⁻¹ in the Raman spectra of M-BTA further confirm the formation of -C-HN-M coordination configuration in M-BTA¹⁷. This is also in alignment with the deconvolution results of the high-resolution N 1 s and C 1 s XPS spectra of all the M-BTA analogs, in which -C = N (398.3 eV) and -C-NH-M (399.6 eV, noted as -C-N-M) species coexist, again confirming the coordination between BTA and metal ions in M-BTA (Supplementary Figs. 9, 10). Meanwhile, the obvious O 1 s peaks in XPS survey scans and the formation of metal-O (M-O, ~531.2 eV) species in their high-resolution O 1 s XPS spectra should be attributed to the spontaneously adsorbed oxygen-containing species, e.g., -OH, H₂O, or O₂, in the unsaturated coordinated metal sites or nanopores in the M-BTA^17,18. In addition, inductively coupled plasma (ICP) mass spectrometry and elemental analysis (EA) results on the Ni-BTA suggest that the atomic ratio of C, N, O, H, and Ni is about 7.1: 4.4 :1.4 :8.0 :1.0 (i.e., Ni_1.0C_7.1N_4.4O_1.4H_8.0), close to the theoretical calculation results (i.e., Ni_1.0C_6.0N_4.0O_1.4H_6.0), but with extra O-atoms (~1.4:1 to Ni-atoms) in experiment (Supplementary Table 2), which provides reliable rationality for the proposed M-BTA models with the out-of-plane-OH coordination at the M-(NH)₄ sites.

Synchrotron-based X-ray absorption spectroscopy (XAS) analysis was further carried out to characterize the coordination environment of Ni-centers. The X-ray absorption near edge structure (XANES) spectrum of Ni-BTA shows an absorption edge between those of Ni-phthalocyanine complex (Ni-Pc) and NiO, indicating that the valence state of Ni species in Ni-BTA is possibly +2 (Fig. 2e). Besides, the absorption edge of Ni-BTA is higher than that of Ni-Pc, and close to the NiO, indicating the possible extra coordinate structure on Ni-(NH)₄ moieties in Ni-BTA, i.e., the out-of-plane coordinated -OH groups. Correspondingly, the extended X-ray absorption fine structure (EXAFS) spectrum of Ni-BTA exhibits a dominant peak at 1.57 Å, which is between Ni-N in Ni-Pc (1.43 Å) and Ni-O in NiO (1.69 Å) scattering paths (Fig. 2f) and may be assigned to the co-existed Ni-N/O paths in Ni-BTA. No Ni-Ni path at ~2.17 Å was detected for Ni-BTA, suggesting atomically dispersed Ni sites in Ni-BTA. Similar result could be drawn from the Wavelet transforms of Ni K-edge EXAFS data of Ni-BTA, in comparison with the Ni foil, NiO, and Ni-Pc in Supplementary Fig. 11. The quantitative least-squares EXAFS curve fitting analysis further confirms the existence of extra axial-O (using -OH as the example of axial-O groups) on the Ni sites apart from the Ni-(NH)₄ coordination configuration (Fig. 2g and Supplementary Figs. 12, 13), according to the more reasonable fitting results of Ni-N₄O/Ni-N₄O_1.5 configurations than that of Ni-N₄ configuration in theory (Supplementary Table 3). Additionally, the Ni-N₄O_1.5 configuration (i.e., Ni center coordinated by average 4 N and 1.5 O atoms) was proposed according to the element analysis and ICP results (Supplementary Table 2). On the basis of these considerations, it is reasonable to conclude that the d-OH-Ni model is more reasonable than the d-Ni and s-Ni configurations to represent the actual coordination environment of Ni centers in the as-prepared Ni-BTA sample.

Besides, all M-BTA samples show a small aggregated structure. Especially, the Ni-BTA exhibits a rod-like morphology with a diameter of ∼30 nm and a length of over 150 nm (Fig. 2h and Supplementary Fig. 14). These nanosized particles would improve the exposure of active sites and facilitate the ORR reaction kinetics. The TEM-based elemental mappings confirm the homogeneous distribution of C, N, O, and M (M=Mn, Fe, Co, Ni, or Cu) over the corresponding materials (Fig. 2i and Supplementary Fig. 15), proofing the extra oxygen coordination for all the M-BTA samples. These results thus confirm the formation of M-(NH)₄-OH moiety for the M-BTA samples, agreeing well with the proposed structure as in the theoretical simulations (Fig. 1a).

Additionally, highly amorphous Ni-BTA (a-Ni-BTA) without such interlayer stacking structure was also prepared by bubbling the air into the reactor and shortening the reaction time, which accelerates the coordination reaction and distorts the internal layered structure of Ni-BTA. As a result, almost no peaks exist for a-Ni-BTA owing to its highly disordered feature (Supplementary Fig. 16); Meanwhile, the a-Ni-BTA delivers a particle-like structure with a diameter of <5 nm, indicating its low crystallinity (Supplementary Fig. 17). In addition, to further confirm the H-bond enhanced 2e⁻-ORR activity for the catalytic sites with weak *OOH adsorption capacity (i.e., s-Cu and s-Ni), the Cu-BTA with higher crystallinity (ordered layered structure), compared to that of Cu-BTA prepared with the same method of Ni-BTA, was also prepared by lowering the nucleation rate and prolonging the growth time (Supplementary Fig. 18). The district XRD peaks of H-Cu-BTA, with similar positions as the Ni-BTA, as well the rob-like morphology, collaboratively confirm its largely enhanced ordered structure. Specifically, the effect of crystallinity for the materials on the 2e⁻-ORR performance will be discussed in the following section.

Electrocatalytic performance for 2e⁻-ORR under neutral and alkaline conditions

The 2e⁻-ORR catalytic selectivity and activity of all samples were measured on the rotating ring-disk electrode (RRDE) at 1600 rpm in O₂-saturated alkaline and neutral electrolytes, including 0.1 M KOH, 0.9 wt.% NaCl, 3.6 wt.% NaCl, and 0.1 M Na₂SO₄ aqueous solutions. The collection efficiency (N) of the RRDE was determined to be 0.37 (Supplementary Fig. 19), and the ring electrode was held at 1.2 V in the neutral media and 1.5 V vs. RHE in 0.1 M KOH solution. As shown in their linear sweep voltammetry plots (Fig. 3a, b), all the M-BTA catalysts show the onset potentials of 0.70–0.80 V vs. RHE (the potential at a disk current of −0.1 mA cm⁻²) in 0.1 M KOH, indicating their high ORR activity. Significantly, the Ni-BTA shows the largest ring current ~0.16 mA and the highest H₂O₂ selectivity of ~90% in the potential range of 0-0.6 V vs. RHE, which is higher than those of Mn-BTA ( ~ 0.08 mA, ~52%), Fe-BTA ( ~ 0.1 mA, ~62%), Co-BTA ( ~ 0.125 mA, ~70%), and Cu-BTA ( ~ 0.12 mA, ~70%), preliminarily indicating the best 2e⁻-ORR performance of Ni-BTA than others in our study. In addition, the Tafel slope of Ni-BTA, according to the LSV curve of the ring current, is calculated to be 79.9 mV dec⁻¹, which is the lowest among all the M-BTA analogs (Supplementary Fig. 20) and confirms its fastest H₂O₂ electrosynthesis reaction kinetics^21,22. Furthermore, Ni-BTA also has an ability to inhibit the undesirable H₂O₂ reduction with a negligible reaction current density (Fig. 3c and Supplementary Fig. 21)²³, and possesses the largest double-layer capacitance (indicating the largest electrochemically active surface area) for achieving high utilization of the active sites (Supplementary Fig. 22), which collaboratively contribute to its high 2e⁻-ORR catalytic performance²¹. These results are highly consistent with the theoretically predicted performance from the volcano plot based on ΔG_*OOH values on all the M-BTA analogs.

Fig. 3: Electrochemical H₂O₂ selectivity.

a LSV curves of ORR achieved in O₂-saturated 0.1 M KOH solution at 10 mV s⁻¹ (non-iR corrected), (b) the calculated H₂O₂ selectivity, (c) H₂O₂ reduction (H₂O₂RR) polarization curves of Ni-BTA and M-BTA in 0.1 M KOH with 10 mM H₂O₂ (non-iR corrected). d LSV curves of ORR in O₂-saturated 3.6 wt.% NaCl solution at 10 mV s⁻¹ (non-iR corrected) and the calculated H₂O₂ selectivity (e). f the 2e⁻-ORR performance of Ni-BTA in different neutral media, including the 0.9 wt.% NaCl, 3.6 wt.% NaCl and 0.1 M Na₂SO₄.

Similar performance trends of the catalysts are also observed in the neutral electrolytes, including 3.6 wt.% NaCl aqueous solution (simulated seawater), 0.9 wt.% NaCl aqueous solution (physiological saline), and 0.1 M Na₂SO₄ aqueous solution. Especially, in the simulated seawater, Ni-BTA still exhibits the highest H₂O₂ selectivity of >85% at the potential range of −0.1-0.4 V vs. RHE with the largest H₂O₂ electro-synthetic limiting current of the ring (j_ring > 0.12 mA) among all the catalysts (Fig. 3d, e), confirming its strong resistance for Cl⁻ poisoning. Moreover, its negligible difference in ORR catalytic activity and selectivity, upon lowering the concentration of NaCl solution from 3.6 wt.% to 0.9 wt.% or changing NaCl to Na₂SO₄ solution (Fig. 3f and Supplementary Fig. 23), confirms the comparable 2e⁻-ORR selectivity with previous reported catalysts in alkaline or neutral solutions (Supplementary Table 4), thus endowing Ni-BTA with huge potential for applications in a wide range, including in sewage treatment and bio-medical disinfection, etc.

Additionally, the low-crystallinity Ni-BTA (a-Ni-BTA) still shows a higher H₂O₂ selectivity of ~80% than that of other M-BTA, confirming the unique advantage of Ni-BTA on achieving high 2e⁻-ORR performance. However, the lower H₂O₂ selectivity and onset potential (0.70 V), as well its smaller ring current of ~0.12 mA of a-Ni-BTA than that of Ni-BTA, provide additional evidence for the ordered layer-structure on enhancing the material’s 2e⁻-ORR performance (Supplementary Fig. 24). Similarly, the high-crystallinity Cu-BTA (H-Cu-BTA) also delivers an enhanced 2e⁻-ORR performance (H₂O₂ selectivity nearly 80%) compared with the amorphous Cu-BTA, but this performance is still lower than that of Ni-BTA sample with high-crystallinity (Supplementary Fig. 25). Although the high-crystallinity Mn/Fe/Co-BTAs were not successfully prepared, it would be rational to believe that the Ni-BTA (high-crystallinity) would deliver better 2e⁻-ORR performance than the Fe/Co/Mn-BTA (either amorphous or high-crystallinity), and the reasons are as follows: in our study, the H-bond formation via the dual-layer structure is proposed to enhance the adsorption capacity to *OOH intermediates, and the Co/Fe/Mn-BTA with dual-layer structure all show too strong adsorption capacity to *OOH intermediates (<4.23 eV), ~4.05 eV for d-OH-Co, 3.95 eV for d-OH-Fe and 3.61 eV for d-OH-Mn in theory. That is, the high-crystallinity Fe/Co/Mn-BTA should deliver a lower 2e⁻-ORR performance than the counterparts with the amorphous structure. Thus, we conclude that the Ni-BTA possesses the intrinsic better 2e⁻-ORR performance than those Mn/Fe/Co/Mn-BTAs.

H₂O₂ production performance in flow cells and its application perspectives

Inspired by the high 2e⁻-ORR performance tested by RRDE, the material was fabricated into a gas-diffusing electrode (1 × 1 cm²), and its H₂O₂ production rate in practical conditions was further studied in a flow cell with direct oxygen feeding (Fig. 4a). Before the H₂O₂ accumulation test, electrochemical impedance spectroscopy (EIS) test confirms a low charge transfer resistance (~0.85 Ω and ~0.70 Ω in 1 M NaCl and 1 M KOH solutions, respectively) between the MOF-type electrocatalyst and the electrode (Supplementary Fig. 26), which would be attributed to the relatively good electron conductivity of the Ni-BTA MOFs and the close contact between the catalyst and the gas-diffusion electrode^24,25,26. And the H₂O₂ concentration in the electrolyte was quantified by the potassium oxalate colorimetric method⁵, according to the standard curve (Supplementary Fig. 27).

Fig. 4: H₂O₂ production performance in flow cells.

a Schematic diagram of the flow cell for H₂O₂ production, the LSV curves (non-iR corrected), H₂O₂ FE (b) and H₂O₂ yield rate (c) of Ni-BTA in 1 M KOH and 1 M NaCl solutions. d The long-term H₂O₂ electrosynthesis at 100 mA cm², e the corresponding H₂O₂ concentration accumulation in 1 M NaCl solution. The degradation performance for Rhodamine B in 1 M NaCl solution (f), and the disinfection performance for E. coli and S. aureus in 0.9 wt.% NaCl (g) after 2e⁻-ORR catalysis using Ni-BTA catalyst. The data points and error bars in g represent the average and standard deviation of data from triplicate parallel tests.

As shown in Fig. 4b, Ni-BTA catalyst shows high H₂O₂ Faradaic efficiencies (FE) over 80% with increasing ORR current densities until over 200 mA cm^–2, either in 1 M KOH (pH ~13.8) or 1 M NaCl (pH ~6.5) electrolyte. Correspondingly, the H₂O₂ yield rate could reach 34 mol g^–1 h^–1 at –0.4 V vs. RHE in 1 M KOH and 23.4 mol g^–1 h^–1 at –0.6 V vs. RHE in 1 M NaCl electrolyte, respectively (Fig. 4c). Additionally, it was assumed that all Ni centers were actively involved in 2e⁻-ORR and that the content of Ni in the material was set to 24.5 wt.% according to the ICP results. The turnover frequencies (TOFs) reached 1.56 s⁻¹ and 2.27 s⁻¹ in 1 M NaCl and 1 M KOH electrolytes, when the potentials were at −0.59 V vs. RHE and −0.42 V vs. RHE, respectively (Supplementary Fig. 28), indicating the high intrinsic activity of the catalyst. For the H₂O₂ accumulation test, it was firstly operated at a constant current density of 100 mA cm⁻² for 24 h in 30 mL 1 M NaCl electrolyte. A total H₂O₂ amount of 32.4 mmol was produced with a stable cell voltage of ~5.25 V (i.e., cathode potential of −0.25 V vs. RHE, Fig. 4d and Supplementary Fig. 29), corresponding to an average H₂O₂ production rate of 13.5 mol g⁻¹ h⁻¹. It is worth noting that there were two electrolyte refreshments for the cathodic tank after working until 8 h and 15 h, which was intended to protect the catalyst from being oxidized under high H₂O₂ concentration and to avoid H₂O₂ decomposition, thus ensuring the efficiency of H₂O₂ production. A recovered H₂O₂ FE from ~70% to ~85% after the electrolyte refreshment verifies its importance, and the accumulated H₂O₂ concentration in the cathodic tank all reached over 11 g L⁻¹ before the electrolyte refreshment in the three cycles (Fig. 4e).

Similarly, H₂O₂ accumulation was also assessed at a higher constant current density of 200 mA cm⁻² in 30 mL 1 M KOH electrolyte, and a high H₂O₂ concentration of 26.8 g L⁻¹ with a high H₂O₂ FE over 65% was achieved after continuous operation of 10 h without electrolyte refreshment (Supplementary Fig. 30), which is close to the concentration of commercial H₂O₂-type disinfectants (~30 g L⁻¹). In addition, when operating at a constant potential with an initial current density of 100 mA cm⁻² and in a larger electrolyte volume of 260 mL, the catalyst could still deliver a high current retention (>80%) and stable H₂O₂ FE over 90% at 0.2 V vs. RHE after working for 40 h in 1 M KOH electrolyte (Supplementary Fig. 31), further demonstrating the stability of Ni-BTA for long-term operation. To sum up, the Ni-BTA shows outstanding H₂O₂ yield and stability comparable or even superior to the most reported catalysts in alkaline and neutral solutions (Supplementary Fig. 32 and Supplementary Table 5), endowing it with significant potential for practical applications.

Additionally, the structural stability of the Ni-BTA catalyst is confirmed by the in/ex-situ XRD and/or TEM characterizations (Supplementary Figs. 33, 34), in which the three characteristic peaks as well as the rob-like morphology of Ni-BTA are well-retained during the catalytic process (after 7 h). Similarly, the almost overlapped Ni K-edge EXAFS curves of Ni-BTA before/after the catalysis indicate a relatively stable chemical and coordination structure for Ni-BTA during catalysis. And the small shift of the Ni K-edge XANES spectrum after catalysis to higher energy, referring to an increased oxidation state for the central Ni atoms, which should arise from the absorbed oxygen-containing species during the ORR process that are not totally desorbed. This thus leads to a smaller R-factor of 0.6% when adopting the Ni-N₄O₂ configuration than the Ni-N₄O_1.5 configuration (1.4%, Supplementary Fig. 35 and Supplementary Table 6) to fit the XAFS spectrum.

To better validate the feasibility of electrocatalytic H₂O₂ synthesis under practical conditions, we further assessed the material’s catalytic performance in various diluted electrolytes, including 0.9 wt.% NaCl (physiological saline), 3.6 wt.% NaCl (simulated seawater), and 0.5 M Na₂SO₄ aqueous solutions (Supplementary Fig. 36). Even in low-concentration physiological saline and simulated seawater, the ORR current density could still exceed 50 and 100 mA cm⁻², respectively, while maintaining a high Faraday efficiency of over 80% at 50 or 100 mA cm⁻² current densities (Supplementary Fig. 37). Therefore, electrocatalytic H₂O₂ synthesis using the Ni-BTA electrode in 1 M NaCl electrolyte at 100 mA cm⁻² could be directly utilized in an in-situ and UV-coupled Fenton process for rapidly removing Rhodamine B and Methylene blue, demonstrating its practical feasibility for waste-water treatment and bleaching (Fig. 4f and Supplementary Figs. 38, 39). In addition, complete eradication of S. aureus and E. coli could be achieved upon only 30 min electrocatalytic H₂O₂ synthesis at 50 mA cm⁻² in the 0.9 wt.% NaCl electrolyte (Fig. 4g and Supplementary Fig. 40), showing its huge potential in the disinfection, medical, and healthcare fields. Thus, it can be concluded that such high H₂O₂ electrosynthesis rate and efficiency of the Ni-BTA catalyst in different pH values and various electrolyte types/concentrations endow it with remarkable versatility, efficiency, and effectiveness for a series of practical H₂O₂-realated applications.

Mechanistic study from a dynamic calculation and in-situ ATR-IR spectra

To better understand the enhancement mechanism for 2e⁻-ORR electrocatalysis, the H-bond induced *OOH adsorption modification was further confirmed by the potential-dependent in-situ ATR-SEIRAS spectra in O₂-saturated 0.1 M KOH D₂O solution. As shown in Fig. 5a, with the decrease of potential from the open-current potential (OCP) to 0.7 V vs. RHE, a series of new peaks emerged and maintained a high intensity until the potential was further decreased to 0.2 V vs. RHE, indicating the occurrence of the oxygen reduction reaction. Specifically, the two distinct peaks at ~1210 cm⁻¹ and ~1480 cm⁻¹ can be assigned to the O-O stretching mode of the surface-adsorbed OOD_ad and weakly adsorbed molecular oxygen (O_2ad), respectively, and the other peaks at ~1340 cm⁻¹ should be associated with the OOD bending mode of surface-adsorbed hydroperoxide (DOOD_ad), according to the previous works²⁷ and the calculated results via computational spectroscopy (Supplementary Table 7). And there is no obvious signal assigned to Ni-OD stretching vibration absorption peak (around 2900 cm⁻¹), indicating the favorable 2e⁻-ORR pathway rather than 4e⁻-ORR pathway for the Ni-center in our catalyst. It also confirms that the exposed Ni-sites will not be covered by OD species under ORR condition, in agreement with the result concluded from the surface Pourbaix diagrams (Supplementary Fig. 41)^28,29. It is noticeable that these characteristic peaks for the typical intermediates during ORR all shift towards a high wavenumber compared to calculated results via computational spectroscopy (Supplementary Table 7), which may be induced by the effect of test conditions or solvation. In addition, the emerged peaks at 3240 cm⁻¹ could be assigned to the N-H stretching mode in the BTA ligands. Its redshift during ORR, compared with the initial position for N-H stretching at 3300 cm⁻¹ for Ni-BTA (Supplementary Fig. 8), indicating an enlarged N-H bond in the Ni-BTA during ORR process, clearly confirms the presence of hydrogen-bond interactions, i.e., Ni-N-H + ···OHO-Ni, between the N-H groups in the Ni-BTA and ORR intermediates¹⁴. These observations confirm the formation of H-bonds between the *OOH intermediate adsorbed on the Ni sites in one layer of the Ni-BTA and the N-H groups on the adjacent layers, which makes an extra contribution to enhancing the *OOH adsorption ability of the Ni sites, thus leading to the improved ORR activity while maintaining its high selectivity for 2e⁻-ORR.

Fig. 5: The insights on oxygen reduction reaction for Ni-BTA.

a Potential-dependent in-situ ATR-SEIRAS spectra of Ni-BTA in O₂-saturated 0.1 M KOH D₂O solution. b Free energy (thermodynamic barrier) profile of 2e⁻ and 4e⁻-ORR pathways at the equilibrium potentials of 0.695 V vs. RHE on different Ni-(NH)₄ configurations. c, d Free energy changes along with Ni-O bond and O-O bond cleavages for the 2e⁻ and 4e⁻-ORR pathways (c) and the corresponding atomic structure of different reaction states (d), respectively. The initial state, transition state, and final state structures are denoted as IS, TS, and FS, respectively. Light blue, brown, gray, red, and magenta spheres represent the N, C, Ni, O and H atoms, respectively.

Finally, to better understand the good 2e⁻-ORR catalytic performance of the crystalline Ni-BTA, the free energy with corrected values (Supplementary Table 8) of 2e⁻ and 4e⁻-ORR pathways at the equilibrium potentials of 0 V, 0.695 V and 1.23 V vs. RHE on different Ni-(NH)₄ configurations are plotted as Fig. 5b and Supplementary Fig. 42. When only focusing on the 2e⁻-ORR process, d-OH-Ni gives a low thermodynamic overpotential (η) of 0.02 eV, while the d-Ni, s-OH-Ni, and s-Ni deliver much larger overpotentials of 0.35 eV, 0.26 eV, and 0.60 eV, respectively. However, it should be noted that the formation of *O intermediate on the surface of d-OH-Ni presents lower free energy than that of *OOH intermediate, suggesting that 4e⁻-ORR pathway is more energetically favorable than 2e⁻-ORR pathway for the d-OH-Ni, which is inconsistent with the experimental observations were clearly shows its favorable 2e⁻-ORR pathway (Fig. 3).

Addressing this, we further calculated the reaction dynamic barriers for 2e⁻ and 4e⁻-ORR over d-OH-Ni for identifying the decisive step using a molecular dynamics method via an explicit solution model under alkaline environment^30,31. And the final solution network of the explicit solution model is obtained from a pre-molecular dynamic simulation, meanwhile achieving the stable initial state of the Ni-BTA during ORR in theory (Supplementary Fig. 43 and Supplementary Data 1). For the 2e⁻-ORR pathway, the rate-determining step was found to be the Ni-O elongation by the slow-growth approach, with a reaction barrier of 0.33 eV (Fig. 5c, d, and Supplementary Fig. 44). In contrast, the rate-determining step for 4e⁻-ORR changed to the formation of *O with a significantly higher reaction barrier of 1.21 eV. In addition, to reduce the effect of solution network variation on the barrier, we calculated the error bar on the kinetic barriers by three individual slow-growth kinetic simulations (Supplementary Figs. 45, 46 and Supplementary Data 1), and the similar results obtained from these simulations (i.e., much higher energy required to break the O-O bond than that of the Ni-O bond) confirm that d-OH-Ni would follow the energetically more favorable 2e⁻-ORR pathway, resulting in the rapid and selective formation of H₂O₂. Therefore, the explicit solution model simulation, combined with the slow-growth approach considering into the real alkaline environment, comprehensively and clearly confirms that the 2e⁻-ORR pathway is more favorable than the 4e⁻-ORR for our Ni-BTA catalyst, which is in good agreement with the experimental results.

Discussion

In conclusion, conductive Ni-BTA with layered structure and Ni-(NH)₄ nodes was prepared as the catalyst for 2e⁻-ORR toward H₂O₂ high-efficiency production. Combined theoretical simulation and in-situ ATR-SEIRAS experimental characterization collectively confirm the significant effect of the internal H-bonds formed between the*OOH intermediates and-N-H groups in the adjacent top-layer of Ni-centers on enhancing the *OOH binding energy close to the optimal value. Besides, the axial coordination structure (axial-OH) on Ni-centers induced during the preparation process under alkaline conditions also plays a synergistic effect on tailoring the catalytic performance. Thus, an exceptionally high H₂O₂ yield of up to 13.5 mol g⁻¹ h⁻¹ in a neutral medium with a satisfactory FE of ~80% and a durability of 24 h could be achieved. Besides, there was a double-enhanced H₂O₂ yield rate of over 27.0 mol g⁻¹ h⁻¹ in an alkaline medium. Such high H₂O₂ yield rates, both in alkaline and neutral media, are well-placed among the state-of-the-art catalysts for H₂O₂ electrocatalysts, thus endowing the catalyst with dye degradation and bactericidal performances. This work contributes to the future decarbonization of H₂O₂ production and highlights the major role of non-coordination structure on regulating the catalytic performance of ORR reactions and beyond.

Methods

Materials

All chemicals were utilized without further purification. 1,2,4,5-benzenetetramine tetrahydrochloride (BTA·4HCl, >97%), ammonia solution (NH₃·H₂O, GR, 25-28%), NiCl₂·6H₂O (AR, 98%), MnCl₂·4H₂O (99.99%), FeSO₄·7H₂O (AR, >99%), CoCl₂·6H₂O ( > 98%) and CuCl₂·2H₂O ( > 99.99%), KOH (AR, 95.0%) and Na₂SO₄ (AR, 99%) were purchased from Aladdin Reagents. NaCl (GR, 99.8%) was procured from Macklin. Nafion 117 membrane with a thickness of ~0.18 mm was purchased from DuPont Co, the gas-diffusion electrode (GDE, SGL-39BB) was purchased from Toray. The rotating ring disk electrode (RRDE, disk area: 0.2475 cm², ring area: 0.1866 cm², with a collection coefficient of ~0.37), the platinum foil (2×2 cm²) and Ag/AgCl electrode (R0302) were purchased from Tianjin ida Co.

Synthesis of Ni-BTA and its analogs

The Ni-BTA and the other M-BTA (M=Ni, Cu) were prepared by the coordination polymerization of BTA·4HCl and Ni²⁺ or M²⁺ ions (M=Mn, Fe, Co, Cu) under the help of ammonia. In detail, for the preparation of Ni-BTA, 0.3 mmol BTA·4HCl was dissolved in 10 mL deionized water and stirred for 10 mins, as solution 1. Then, a solution of 0.2 mmol NiCl₂·6H₂O in 10 mL deionized water was added into solution 1 dropwise and stirred for another 10 mins. Thirdly, 1.0 mL concentrated aqueous ammonia (~14 M) was added to the above mixed solution and stirred for 6 h under room temperature. Finally, the Ni-BTA powders were obtained after the powders were washed three times with deionized water and acetone, and then dried at 60 °C under vacuum. Similarly, the other M-BTA were prepared by changing the NiCl₂·6H₂O to MnCl₂·4H₂O, FeSO₄·7H₂O, CoCl₂·6H₂O and CuCl₂·2H₂O. In addition, the low-crystallinity or amorphous Ni-BTA (a-Ni-BTA) was prepared using 0.2 mmol NiCl₂·6H₂O and 0.2 mmol BTA·4HCl by bubbling the air into the aqueous solution, according to previous works¹⁸ and shorting the reaction time to 2 h.

Characterization of Ni-BTA and its analogs

The morphology and microstructure were characterized by scanning electron microscopy (SEM, S4800, Japan) and transmission electron microscopy (TEM, JEOL-200CX, Japan). The chemical structure and surface composition were collected with X-ray photoelectron spectroscopy (XPS, Escalab 250Xi, USA), X-ray absorption spectroscopy (XAFS, Shanghai Synchrotron Radiation Facility, Shanghai, China), Raman microscopy (Raman, HR800, Japan). In particular, the Athena module of the IFEFFIT software package was conducted to analyze the XAFS raw data according to standard procedures, and the EXAFS data fitting was performed on the Artemis program. In addition, in-situ ATR-SEIRAS spectra were performed on a Bruker Tensor FTIR spectrometer equipped with a MCT detector in O₂-saturated 0.1 M KOH D₂O solution. The crystal structure was confirmed by X-ray diffraction (XRD, D8 Advanced, Germany) using Cu Kα radiation, and the in/ex-situ XRD was conducted in a custom-made three-electrode electrochemical system in O₂-saturated 3.6 wt.% NaCl electrolyte at 0 V vs. RHE. The post-catalysis SEM images and XAFS were collected on the Ni-BTA catalyst after a 2-h and/or 7-h catalytic process at 0 V vs. RHE in O₂-saturated 3.6 wt.% NaCl electrolyte.

Electrochemical measurements

The electrochemical measurements were conducted at room temperature (~25 °C), and the electrolyte is prepared for immediate use if without specific note.

H₂O₂ selectivity was carried out using a CHI 760E electrochemical workstation in a three-electrode system, with graphite rod electrode and Ag/AgCl electrode as the counter electrode and the reference electrode, respectively. The rotating ring disk electrode (RRDE) composed of glassy carbon (GC) disk and Pt ring was chosen as the working electrode. And the reference electrode was calibrated through the CV test at a scan rate of 1.0 mV s^–1 in the H₂ saturated solutions, using Pt foil as both the working electrode and counter electrode³². The recorded potentials were normalized by converting to reversible hydrogen electrode (RHE) according to the Eq. 1:

$${E}_{{RHE}}={E}_{{Ag}/{AgCl}}+0.059\times {pH}+0.197$$

(1)

To prepare the catalyst ink, 1.0 mg of the obtained catalyst powders were dispersed in 950 μL of the mixture of isopropanol and water (v/v = 1:3.25) and 50 μL of 0.2 wt.% PTFE solution. 20.0 μL of the catalyst ink was drop-casted on the disk electrode of RRDE with a mass loading ~80 μg cm^-2. LSV curves were measured on the RRDE at a scan rate of 10 mV s⁻¹ and a rotating speed of 1600 rpm in O₂-saturated alkaline and neutral media, including 0.1 M KOH (pH ~13.0), 0.9 wt. % NaCl (pH ~7.2), 3.6 wt. % NaCl (pH ~6.7) and 0.1 M Na₂SO₄ (pH ~7.0) aqueous solutions, respectively. During the LSV test, the Pt ring potential was held at 1.5 V and 1.2 V vs. RHE in alkaline and neutral media, respectively. Before the LSV tests, there was an electrochemically cleaning process for the Pt ring with a scan rate of 100 mV s^-1 for 5-cycle CV curves in the same potential range of ORR to prevent the surface passivation of the Pt ring. The H₂O₂ selectivity and ORR electron transfer number are calculated using the following equations:

$${H}_{2}{O}_{2}\, \% =200\times \frac{{I}_{R}/N}{{I}_{D}+{I}_{R}/N}$$

(2)

$$n=4\times \frac{{I}_{D}}{{I}_{D}+{I}_{R}/N}$$

(3)

where I_R, N, and I_D refer to the ring current, collection efficiency, and disk current, respectively, and the collection efficiency was determined to 0.37 using the [Fe(CN)₆]³⁻/⁴⁻ redox system (Supplementary Fig. 19).

H₂O₂ production and faradaic efficiency of the catalysts were carried out in a three-electrode flow-cell, in which the gas-diffusion layer (GDE, SGL-39BB) coated with the catalysts with an active area of 1.0 × 1.0 cm², a platinum foil measuring 2×2 cm² and Ag/AgCl electrode were chosen as the working electrode, the counter electrode, and the reference electrode, respectively. The anode and cathode were separated by a Nafion 117 membrane (pre-treated with 5 wt.% H₂O₂ solution at 80 °C for, followed 0.5 M H₂SO₄ at 80 °C for 1 h). In detail, 20.0 μL of the 5 mg mL^–1 catalyst ink (5.0 mg catalyst powder in a mixture of 400 μL deionized water and 500 μL isopropanol solution with 100 μL 0.2 wt.% PTFE aqueous solution) was drop-casted on the GDE to prepare the working electrode with a catalyst loading of 0.1 mg cm^–2. Either the concentrated electrolytes, including 1 M KOH and 1 M NaCl solutions, or the diluted electrolytes, including 3.6 wt. % NaCl electrolyte (simulated seawater), 0.9 wt. % NaCl (physiological saline) and 0.1 M Na₂SO₄ were all cycled at a rate of 6 mL min^-1, using pumps on each side, with a volume of 30 or 260 mL to test the H₂O₂ production and faradaic efficiency of the catalyst. During the test, pure O₂ was purged at a rate of 100 mL min⁻¹ through a diffusion channel at the back of the GDE. The i-t test at different potentials from 0.7 to −0.6 V vs. RHE and the V-t test at different current densities (50, 100 and 200 mA cm⁻²) were both adopted to study the H₂O₂ production of the catalyst. The H₂O₂ concentration in the last electrolyte was ascertained using the colorimetric method of potassium titanium oxalate (Supplementary Fig. 27). The Faraday efficiency (FE) was only performed once and calculated according to following equation.

$${FE} \; \% \;=\frac{2{cVF}}{Q}\times 100 \% $$

(4)

where c and V are the H₂O₂ concentration and the volume of the electrolyte in the cathode (30 or 260 mL), F is the Faraday constant, and Q is the total charge passed.

Dye decomposition and disinfection test

The organic dye decomposition experiment was tested at 100 mA cm^-2 in the flow cell by electro-Fenton-like reaction under the radiation of UV light. During this process, 30 mL 1 M NaCl aqueous solution containing Methylene blue or Rhodamine B (the concentration of dyes is 60 mg L^–1) is adopted as the electrolyte, and the UV lamp was inserted into the external beaker to promote the decomposition of H₂O₂. Meanwhile, the electrolyte with dyes only under UV radiation was used as the control group. The photos and the UV-vis absorption spectra of the electrolyte after different working times were taken to clear the dye decomposition performances.

The disinfection performance of electro-synthesized H₂O₂ for E. coli (ATCC 8099) and S. aureus (ATCC 25923) was taken at a constant current density of 50 mA cm^–2 in the flow cell using 30 mL 0.9 wt.% NaCl (physiological saline) as the electrolyte. 450 uL electrolyte after different H₂O₂ accumulation tests (0.5 h and 1.0 h) were mixed with 50 uL S. aureus or E. coli suspensions (10⁸ CFU mL^–1) to obtain the appropriate diluted bacterial solution (10⁷ CFU mL^–1), then the diluted bacterial solution was smeared on LB agar plate, and the colonies were obtained after cultured 16–20 h. Meanwhile, the bacterial turbidity (OD₆₀₀) of the bacterial solution and the number of bacteria on the LB agar plate after cultivation were tested or quantified to calculate the antibacterial ratio. The Ctrl group is using pure physiological salt to replace the electrolyte after different H₂O₂ accumulation tests.

DFT Calculation

All DFT calculations were performed using the projector augmented wave (PAW) pseudopotential method for the core electrons through the Vienna ab initio simulation package (VASP)^33,34. The electron exchange and correlation energy were conducted via the generalized gradient approximation (GGA) of Perdew-Burke-Ernzerhof (PBE) functional under the energy cut-off of 400 eV^35,36. To simulate the Ni-BTA, s-Ni, d-Ni, s-OH-Ni and d-OH-Ni were placed in a box with uniform lattice constant of 7.62 Å × 20 Å × 25 Å. The k-points were set to be 3 × 1 × 1 with Gamma centered grids for all structures for geometry optimization³⁷. The convergence condition for the geometry optimization is that the maximal force on each atom is less than 0.02 eV/Å and the convergence condition for energy is 10^-6eV.

The 2e⁻-ORR mechanism for producing H₂O₂ comprises of two elementary steps (5 and 6):

$$ \ast+{O}_{2\left(g\right)}+{H}^{+}+{e}^{-}\longrightarrow*{OOH}$$

(5)

$$*{OOH}+{H}^{+}+{e}^{-}\longrightarrow {H}_{2}{O}_{2(l)}+\ast$$

(6)

The asterisk (*) denotes the active site of the catalyst.

The free energy of adsorbed species, taking into account the vibrational analysis, computed based on the computational hydrogen electrode (CHE) model, is defined in Eq. 7^38,39,40:

$$G=E+{ZPE}+{\Delta H}_{0\to T}-{TS}$$

(7)

where E is the DFT-derived total ground state energy of a system, ZPE is the zero-point energy correction, ∆H_0→T represents the enthalpy variation by temperature (Δ) means from 0 to 298.15 K in Eq. (7), and TS is the entropy correction computed with standard methods. These three parts are used to convert E into free energy (G) at 298.15 K. To estimate the reaction free energy of *OOH intermediate (∆G_*OOH) at the catalyst (*) surface in ORR procedure, and the following equation was employed as Eq. (8):

$$\Delta G\left(*{OOH}\right)=G(*{OOH})+\frac{3G({H}_{2})}{2}-G(*)-2G({H}_{2}O)-{eU}$$

(8)

The free energies (thermodynamic barrier) at U = 0.695 V or 1.23 V vs. RHE for all adsorption species are calculated as follows:

$$\Delta G\left(U,\eta \right)=\Delta G(U=0\,V)-(4.92-{neU})$$

(9)

where the n is the electron transfer number from O₂ reduced to the intermediates (i.e., n = 0 for *O₂ and n = 1 for *OOH, etc.), $\eta$ represent the thermodynamic barrier, U is the potential (U = 0.695 V or 1.23 V, respectively).

To evaluate the kinetic energy barrier of the ORR, we used a constrained ab initio molecular dynamics (AIMD) method, the “slow growth” method, to obtain the free energy profile at 300 K through different collective variable^30,31. A cut-off energy of 400 eV, a k-mesh grid of 1 × 1 × 1, a time step of 0.5 fs and a canonical ensemble (NVT) were adopted in the AIMD simulations. The whole system contained 24 H₂O molecules and one OH^– in the bulk water to model the alkaline environment. The final stabilized solution network was obtained by performing AIMD simulations of the system at 300 K and optimized the geometry structure. In this method, the value of the reaction coordinate ξ is linearly changed from the characteristic value of the initial state (IS) to that of the transition state (TS) with a velocity of transformation $\dot{\xi }$ The free energy is needed to perform a transformation from IS to TS and can be computed as (10):

$${W}_{{IS}\to {TS}}={\int }_{\xi \left({IS}\right)}^{\xi \left({TS}\right)}\frac{\partial F}{\partial \xi }\cdot {\dot{\xi}} {dt}$$

(10)

where the F is the free energy calculated at general coordinate ξ, which is evolving with t. f (r, 300 K) is the gradient of the free energy ($\frac{\partial F}{\partial \xi }$) along the limiting degrees of freedom at general coordinate ξ and 300 K, ξ means the bond length (r) or bond elongation (Δr) of Ni-O or O-O bond after the OOH species absorbed on the Ni-sites in d-OH-Ni model, which is directly exported frame-by-frame from the AIMD simulation into REPORT file. $\frac{\partial F}{\partial \xi }$ is calculated along the track of a constrained MD simulation through the SHAKE algorithm in VASP code, ${W}_{{IS}\to {TS}}$ is the free-energy difference between the IS and TS, which is obtained by the integrating process. In the limit of infinitesimally small ∂ξ, the bond lengths of the Ni-O and O-O bonds were chosen as ξ to describe the 2e^– and 4e^– processes of ORR, respectively, with ∂ξ (ultrasmall bond elongation, also referring to $\dot{\xi }{dt}$) value of 0.0005 Å in each AIMD step, and the time step in AIMD simulation is set to be 0.5 femtosecond.