Introduction

Molecular oxygen (O2) activation, the catalytic conversion of O2 into reactive oxygen species (ROS), is fundamental in biology1,2, and increasingly attractive for applications in energy, green chemistry, catalysis, environmental remediation, healthcare, and other fields3,4,5,6,7,8,9. As the most abundant and sustainable oxidant, O2 possesses natural availability, non-toxicity, benign environmental compatibility, and favorable thermodynamics (+ 1.23 V vs. SHE)10. The catalytic generation of ROS from O2 is an attractive alternative to conventional chemical oxidants such as hydrogen peroxide (H2O2)11, peroxymonosulfate (PMS)12, persulfate (PS)13, and ozone (O3)14,15, which suffer from energy-intensive production processes, significant transportation and storage challenges, and potential secondary pollution risks16. However, practical O2 activation is severely constrained by kinetic barriers of ground-state O217. These limitations originate from its stable triplet electronic ground state, where unpaired electrons in degenerate π* antibonding orbitals imposes spin-forbidden constraints on its reactivity.

To achieve efficient O2 activation, conventional strategies typically rely on external energy input (light, electricity, or thermal activation)18,19,20,21, which greatly compromise the economic viability and sustainability of O2 utilization. Recently, energy-free O2 activation systems have been developed to address these challenges. For example, dual Fe atoms anchored on MnO2 assemble into active dimers that efficiently convert O2 into reactive intermediates4, while enzyme-mimicking metal-nitrogen-carbon catalysts enable biomimetic O2 activation analogous to natural enzymatic pathways22. In addition, oxygen vacancies and multi-component transition-metal sites on catalyst surfaces can also promote O2 activation23,24,25,26. In these systems, O2 activation generally follows a stepwise mechanism: O2 is first absorbed and activated to reactive intermediates, such as superoxide radicals (·O2-) or H2O2, which are subsequently converted into reactive hydroxyl radical (·OH) or singlet oxygen (1O2) to mediate oxidative reactions4,27. However, these conventional catalytic systems face several structural and kinetic limitations that result in low ROS production rates (< 2 μM·min−1) and poor reactivity (approximately 50 times lower than O3-based systems)23,24,28, making them unsuitable for practical applications. This is because in most reported catalytic systems, the dominant active sites are typically distributed in an isolated manner23,25,29. Such isolated sites exhibit highly localized electronic states with sparse density near the Fermi level, which results in high energy barriers for electron transfer to the adsorbed O2 molecules, preventing continuous electron transfer and limiting catalytic efficiency. Furthermore, the lack of close spatial coupling between these isolated sites prevents electrons from efficiently shuttling across multiple sites during the multi-step O2 activation process (O2 → H2O2/·O2- → 1O2/·OH), leading to additive energy barriers. This structural limitation often causes the system to remain at the stage of generating low-reactivity intermediates, suppressing the formation of higher-reactivity species such as ·OH or 1O2. Meanwhile, in the absence of spatial confinement, these reactive intermediates that escape into bulk solution are short-lived and readily consumed by solvent or dissolved species, dissipating oxidative capacity30,31.

Here, we report a catalytic strategy that achieves direct molecular oxygen activation from ambient air under room-temperature, bias-free conditions, without added oxidants. This represents a conceptually distinct energy-neutral O2 activation process, in which molecular oxygen is activated under ambient conditions without the consumption of chemical oxidants or external energy input, relying solely on mild aeration to supply O2. Central to this approach is the rational design of Co-Mo dual sites with atomic-scale proximity, where d-d orbital hybridization and intermetallic electron delocalization facilitate efficient electron transfer. This design establishes short-range electron-transfer channels that drives rapid O2 activation through sustained electron donation and selectively generates 1O2 under mild conditions. Remarkably, the CMCN/air system achieves exponential enhancements in both ROS generation (17.0 μmol L-1·min−1) and reaction rate (kobs = 0.494 min−1), which are comparable to traditional chemical oxidants, while reducing operational costs by 1-2 orders of magnitude. The catalyst operates robustly across diverse water matrices, degrades a range of organic pollutants, transforms inorganic ions, and inactivates bacteria. This development is expected to establish a paradigm for sustainable O2 activation, pointing toward next-generation energy-neutral catalytic technologies.

Results

Rational design and characterization of Co-Mo dual-site catalysts

To achieve sustainable oxygen activation, we designed a nitrogen-doped carbon catalyst with atomically adjacent Co-Mo dual sites (CMCN) using ZIF-8 as a structural precursor. The N-doped carbon matrix not only stabilized the coordination of Co and Mo to create highly dispersed active sites but also served as an efficient electron-conducting bridge to enhance intermetallic charge transfer. In addition, the reverse donation of Co (3d7) electronic configuration and the single electron transfer ability of Co2+/Co3+ redox pairs, in conjunction with the electronic complementarity provided by multivalent Mo species, collectively driving efficient O2 activation. The detailed synthesis process and characterization of the samples are presented in Supplementary Fig. 1, Supplementary Method, and Supplementary Tables S1, S2.

The surface microstructure and structure of CMCN were investigated by scanning electron microscopy (SEM) and transmission electron microscopy (TEM) (Fig. 1a and Supplementary Fig. 2a, b). CMCN exhibited a dodecahedral structure with thermal collapse morphology after high-temperature pyrolysis. Energy dispersive X-ray spectroscopy (EDS) (Fig. 1b and Supplementary Fig. 2c) mappings revealed relatively uniform distributions of carbon (C, blue), nitrogen (N, green), cobalt (Co, red), and molybdenum (Mo, yellow) across the catalyst surface. The evaporation of zinc (Zn) during pyrolysis created binding sites that facilitated the incorporation of Co and Mo into the catalyst framework32.

Fig. 1: Morphology and structural characterization of CMCN.
Fig. 1: Morphology and structural characterization of CMCN.
Full size image

a TEM image of CMCN. b EDS mappings of N, Co, and Mo. c, d Normalized K-edge XANES of (c) Co and (d) Mo. e, f Fitting curves for the (e) Co and (f) Mo K-edge EXAFS spectra of CMCN (inset: theoretical model structures, the silver, dark blue, red, blue, and purple balls represent C, N, O, Co, and Mo atoms, respectively). g WT EXAFS for K-edge for Co foil, CMCN_Co, Mo foil, and CMCN_Mo.

X-ray diffraction (XRD) pattern exhibited only the (002) and (101) diffraction peaks characteristic of carbon, with no detectable characteristic peaks of metallic or metal oxide phases (Supplementary Fig. 3), which suggested that the Co and Mo species existed either as nanoscale particles or in the form of amorphous metal nitrides. Significantly, CMCN demonstrated the highest defect density compared to CN, CoCN, and MoCN, which was confirmed by Raman spectroscopy and C 1 s X-ray photoelectron spectroscopy (XPS) spectra (Supplementary Figs. 3, 4). These defect sites could serve as additional centers for O2 adsorption and activation33,34. The N 1 s XPS spectra revealed the formation of Co-Nx/Mo-Nx covalent bonds, which collectively constituted the coordination environment of active sites. Notably, compared to CoCN or MoCN, the binding energy of Co 2p3/2 in CMCN decreased by 0.2 eV while the binding energy of Mo 3d5/2 increased by 0.2 eV, respectively (Supplementary Fig. 4), indicating a redistribution of electron density between adjacent Co and Mo centers35,36,37,38,39. This atomic-level electronic modulation, driven by Co-Mo proximity, enhanced the electron density around Co sites while preserving moderate electron deficiency at Mo sites, thereby constructing a short-range, energetically favorable pathway for interfacial charge migration. This unique electronic structure optimized the adsorption and activation of oxygen intermediates, providing the fundamental basis for the subsequent stepwise activation of O2.

The atomic coordination and structure of CMCN’s Co-Mo dual sites were revealed by the synchrotron-radiation-based X-ray adsorption fine structure (XAFS) analysis, including X-ray adsorption near-edge spectroscopy (XANES) and K-edge extended XAFS (EXAFS), confirming the formation of coupled CoMoN6 centers with distinct electronic properties (Fig. 1c–f). The XANES spectra in Fig. 1c revealed that the rising edge position for CMCN was located between Co foil and CoO, indicating that the valence state of Co was between 0 and + 240. The Fourier transform EXAFS spectra (FT-EXAFS) of Co K-edge showed almost the same peak of CMCN and CoPc at 1.42 Å corresponding to the Co-N coordination (Fig. 1e). Additionally, the peak at around 2.45 Å represented the average nearest Co-metal atoms coordination. The Mo K-edge XANES spectra located between the Mo foil and MoO2, suggesting the valence state of Mo was between 0 and + 4 (Fig. 1d). The FT-EXAFS spectra of the Mo K-edge (Fig. 1f) displayed two distinct peaks: a pronounced peak at 1.38 Å (assigned to Mo-N coordination) and a peak at 2.55 Å, corresponding to the Mo-metal bond distance in CMCN41. Wavelet transform (WT) mapping further distinguished the backscattered atoms. The maximum intensity at ~ 4.2 Å−1 was associated with the Co-N bond in the Co-k space of CMCN by comparing with the CoPc sample (Fig. 1g). In the Mo-k space, the maximum intensity at ~ 6.4 Å−1 was different from the positions of the characteristic peaks in Mo foil (~ 8.6 Å−1) and MoO3 (~ 9.8 Å−1) (Supplementary Fig. 5), suggesting that the existence of Mo-N coordination. Notably, the additional intensity at ~ 6.2 Å−1 in both Mo-k and Co-k WT spectra suggested a Co-Mo scattering path, confirming chemical bonding between adjacent Co-Nx and Mo-Nx centers. To demonstrate the coordination forms of Co and Mo, the EXAFS curves were fitted and analyzed using the quantitative least squares method (Fig. 1e, f, Supplementary Fig. 6 and Supplementary Table 3). The results showed that both Co and Mo were coordinated by three N atoms on average. In addition, the second-shell fitting indicated a coordination number of one for Co-Mo, which bond distance was around 2.55 Å. These results confirmed CMCN adopted a short-range Co-Mo dual-metal center, which might have higher mass transfer efficiency and reaction activity42. Therefore, it could be inferred that the structure of CMCN was the CoMoN6 model.

Air activation and ROS generation performance

Electron paramagnetic resonance (EPR) spectroscopy was initially employed to investigate the generation of ROS under operational conditions. Using 2,2,6,6-tetramethyl-4-piperidinyloxyl (TEMP) as a spin trap, the characteristic 1:1:1 triplet signal was observed in all prepared catalysts, confirming the formation of TEMP-1O2 adducts (Fig. 2a)43. Notably, the CMCN sample displayed markedly enhanced TEMP-1O2 signal intensity compared to the other materials, demonstrating its better performance for the 1O2 generation. As shown in Fig. 2b, while distinct six-line signals corresponding to the 5,5-Dimethyl-1-pyrroline N-oxide (DMPO)-·O2- adduct were observed, no characteristic signals of DMPO-·OH adducts were detected. These results demonstrated that ·O2- and 1O2 were the primary ROS generated in the system, with no measurable production of ·OH. Furthermore, control experiments established the oxygen-dependence of this process (Fig. 2c). The 2,2,6,6-Tetramethylpiperidine 1-oxyl (TEMPO) signal significantly weakened under N2 atmosphere, indicating that the 1O2 production was strictly dependent on oxygen supply. The addition of p-benzoquinone (p-BQ, ·O2⁻ scavenger), caused significant decrease of the TEMPO signal, which demonstrated that ·O2⁻ served as the precursor for 1O2 formation.

Fig. 2: 1O2 generation performance via O2 activation.
Fig. 2: 1O2 generation performance via O2 activation.
Full size image

a TEMP-based EPR spectra for 1O2. b DMPO-based EPR spectra for ·OH and ·O2-. c TEMP-based EPR spectra for 1O2 in different conditions. d A DPBF probe experiment for CMCN. e Time-dependent concentration of 1O2 detected by DMA. f Comparison of 1O2 production rate of CMCN (short-range sites) and CoCN + MoCN (Long-range sites). g Comparison of ROS production rate with reported materials (inset: schematic representation of the O2 activation process). h Comparison of 1O2 production rate with reported materials. Error bars represent the standard deviation of the experiment in triplicate.

The generation of 1O2 was further verified by the chemical probe of 1,3-diphenylisobenzofuran (DPBF) and 9,10-dimethylanthracene (DMA). The DPBF intensity in the CMCN/air system decreased more quickly at 410 nm within 5 min than CN/air, CoCN/air, MoCN/air systems (Fig. 2d). Also, quantitative analysis using DMA revealed that the CMCN/air system generated 85.0 μM of 1O2 within 5 min44,45, corresponding to a remarkable production rate of 17.0 μM·min−1 (Supplementary Method). This performance exceeded those of CN/air (6.2 μM), CoCN/air (23.9 μM), and MoCN/air (4.0 μM) systems by factors of 13.7, 3.6, and 21.3, respectively (Fig. 2e). The observed 1O2 yield significantly exceeded the arithmetic sum of individual CoCN and MoCN components, providing evidence for the synergies between the Co and Mo sites at atomic-scale proximity. This was attributed to the short-range electron delocalization enabled by the atomic-scale proximity of adjacent Co and Mo centers, which created efficient interfacial electron pathways. These pathways optimized O2 adsorption, stabilized intermediates, and accelerated the O2 → ·O2- → 1O2 conversion. In contrast, single-metal systems lacked sufficient electronic coupling, thereby limiting O2 activation and ROS generation efficiency. To elucidate the mechanism underlying this synergistic enhancement, we designed a control experiment using an equimolar physical mixture of CoCN and MoCN (total metal loading identical to CMCN), which the calculated distance of Co and Mo in the solution was about 30 nm. Notably, this mixture produced only 18.1 μM 1O2, which was merely 21.3% of the yield of CMCN (Fig. 2f). This result conclusively confirmed that the efficient molecular oxygen activation and 1O2 generation stemmed from the atomic-scale proximity (~ 2.55 Å) of the Co-Mo dual sites, rather than from simple additive effects of the individual components.

Furthermore, we detected other possible ROS yield (·O2-, H2O2 and ·OH) during the molecular oxygen activation. First, the amount of ·O2- was quantified with using nitroblue tetrazolium chloride (NBT), considering its tetrazolium ring could be selectively reduced to blue formazan (Supplementary Fig. 7)3,44. The amount of ·O2- generated in the CMCN/air system was 20.4 μM, which was more than CN/air, CoCN/air and MoCN/air control systems, indicating that ·O2- acted as the crucial primary intermediate in 1O2 generation. More remarkably, the system exhibited exceptional product selectivity, producing only minimal amounts of other ROS (6.9 μM H2O2 and 1.4 μM ·OH, quantified by the HRP-ABTS colorimetric and benzoic acid probe methods, respectively) during the reaction (Supplementary Fig. 8). These results provided evidence that the atomically-paired Co-Mo sites catalyze 1O2 formation, while effectively suppressing the generation of solution-diffusible oxygen species. As shown in Fig. 2g, the CMCN/air system achieved superior ROS production rate among ambient air or O2 activation systems by an unprecedented 10 ~ 500-fold enhancement (Supplementary Table 4). Notably, its 1O2 yield rate was even comparable to systems requiring external energy or oxidant inputs (Fig. 2h and Supplementary Table 5).

Mechanism investigations of O2 to 1O2

In order to gain a deeper understanding of the conversion of O2 to 1O2 at the molecular level, density functional theory (DFT, Supplementary Method) calculations revealed that, unlike single-metal catalysts, the two oxygen atoms of O2 bonded separately to distinct metal centers (Co and Mo) in CMCN, forming a bridged structure (Co-O-O-Mo). This dual-metal synergy provided enhanced electron back-donation, significantly elongating the O-O bond (Fig. 3a). Also, adsorption energy calculations demonstrated that the CoMoN6 configuration in CMCN exhibited remarkably strong O2 adsorption (− 2.69 eV), substantially exceeding single-metal sites (Co-N3: − 2.03 eV; Mo-N3: − 1.71 eV), which stemmed from the cooperative electronic effects of bimetallic sites (Fig. 3b).

Fig. 3: Mechanism investigations of O2 activation.
Fig. 3: Mechanism investigations of O2 activation.
Full size image

a The local adsorption configurations of O2. The silver, dark blue, red, blue, and purple balls represent C, N, O, Co, and Mo atoms, respectively. b The Eads of O2 on CoCN, CMCN, and MoCN. c O2-TPD of CN and CMCN. d In situ Raman spectra of CMCN catalyst in O2-saturated aqueous solution.

In addition, temperature-programmed oxygen desorption (O2-TPD) profiles further confirmed the superior activation capability of CMCN, with peaks at 100–200 °C, 200–400 °C, 400–600 °C, and > 600 °C being assigned to superoxide (O2−), monatomic oxygen (O−), surface lattice oxygen (Olatt2−), and bulk lattice oxygen (bulk Olatt2−), respectively46,47. While CN showed weak peaks at 183 °C (O2−) and 361 °C (O−), CMCN displayed intense desorption features at 173 °C (O2−), 274 °C (O−), and 590 °C (minor Olatt2−), indicative of abundant reactive superoxide intermediates that that can mechanistically evolve into 1O2 through short-range electron transfer channels established by the atomic-scale Co-Mo dual sites, thereby substantially diversifying the oxygen activation pathways (Fig. 3c). Furthermore, in situ Raman spectroscopy captured the dynamic evolution of surface-bound oxygen species during O2 activation (Fig. 3d). Specifically, the broad band in the 600–800 cm−1 range was assigned to the stretching vibrations of O–O (VO-O) in oxygen intermediates formed upon initial O2 activation. Meanwhile, the additional peak around ~ 1060 cm−1 was attributed to the O-O stretching vibration of peroxo-like intermediates on the surface of CMCN10,48,49. This efficient oxygen activation was attributed to the short-range electron transport network established by atomically adjacent Co-Mo dual sites, where electronic synergy through metal-nitrogen coordination enabled rapid interfacial electron transfer, allowing adsorbed O2 to efficiently capture electrons and undergo stepwise conversion into highly reactive species.

Electron transfer and catalytic mechanism

Based on these experimental findings, we proposed a synergistic activation mechanism enabled by an atomically adjacent Co-Mo dual-site catalyst. Density of states (DOS) analysis revealed that the Co 3d-band center shifted to lower energy (− 1.174 eV in CMCN vs − 0.633 eV in CoCN), reflecting the electronic modulation induced by the introduction of Mo, which optimized O2 adsorption strength while preventing excessive binding that could poison active sites (Fig. 4a). Conversely, the Mo 4d-band moved closer to the Fermi level (− 0.208 eV in CMCN vs − 0.243 eV in MoCN), enhancing adsorption of oxygen intermediates and facilitating ·O2⁻ to 1O2 conversion (Fig. 4b). Remarkably, the d band of Co and Mo in CMCN exhibited excellent orbital alignment near the Fermi level (Fig. 4c), indicating strong d-d electronic coupling induced by the atomic-scale Co-Mo proximity. Furthermore, charge density analysis confirmed that the introduction of Mo increased electron density around Co sites (Supplementary Fig. 9), establishing short-range, efficient interfacial electron pathways. These pathways, together with the optimized Co and Mo electronic states, promoted dynamic electron injection into O2 antibonding orbitals, thereby facilitating stepwise O2 → ·O2- → 1O2 conversion. This electron donation effect provided sufficient electron sources for O2 activation, revealing that the short-range Co-Mo dual sites in CMCN established efficient electron transfer pathways, which were consistent with electrochemical impedance spectroscopy (EIS) and ultraviolet photoelectron spectroscopy (UPS) analysis (Supplementary Fig. 10).

Fig. 4: Electron transfer mechanism of CMCN.
Fig. 4: Electron transfer mechanism of CMCN.
Full size image

PDOS of (a) Co 3 d orbitals of CoCN and CMCN, (b) Mo 4 d orbitals of MoCN and CMCN. c PDOS of Co-d and Mo-d in CMCN. d Electron density difference for O2 adsorption on CoCN, CMCN, and MoCN and the corresponding charge transfer and the length of O-O bond (the brown, silver, red, blue, and purple balls represent C, N, O, Co, and Mo atoms, respectively). e, f XPS of (e) CMCN-Co and (f) CMCN-Mo with O2 in 5 min. g, h Schematic representation of the molecular oxygen activation process (the silver, dark blue, red, blue, and purple balls represent C, N, O, Co, and Mo atoms, respectively).

In addition, the O2 activation mechanism were further elucidated by differential charge density maps (Fig. 4d), which revealed a significant electron redistribution, indicating the formation of strong covalent bonds between O2 and neighboring atoms. The O-O bond was elongated to 1.44 Å in CMCN because of the effect of strong covalent interactions, while that of CoCN and MoCN was only 1.37 Å and 1.31 Å, respectively. Also, bader charge analysis confirmed enhanced electron transfer in CMCN (0.84 e) compared to CoCN (0.62 e) and MoCN (0.54 e), which was consistent with adsorption energy trends. This precise electronic modulation stemmed from atomic-scale Co-Mo synergy, where short-range electron delocalization accelerated interfacial electron transfer, dramatically promoting O2 → ·O2⁻ → 1O2 conversion kinetics. Furthermore, we conducted in-situ XPS test on CMCN to further reveal the dynamic electron transfer process during O2 activation (Fig. 4e, f). After being exposed to a saturated dissolved oxygen solution for 5 min, the Co 2p3/2 peak showed a positive binding energy shift (0.2 eV), indicating that the Co sites lost electrons during the activation process. This electron loss promoted electron transform from O2 to form ·O2-. Simultaneously, the Mo 3d5/2 peak showed a negative shift of 0.2 eV, suggesting that Mo sites gained electrons during O2 activation. Linear sweep voltammetry (LSV, Supplementary Fig. 11) further confirmed this mechanism, showing current enhancement upon air injection, demonstrating the exceptional capability of the system to activate molecular oxygen at the Co-Mo dual sites.

Traditional oxygen activation catalysts were fundamentally limited by inefficient interfacial charge transfer due to hindered electron relay between spatially separated active sites, solvent-induced dissipation of reactive intermediates, and intrinsically low ROS generation rates (Fig. 4g). The CMCN/air system overcomes these limitations through an atomic-scale Co-Mo dual-site design that creates a unique short-range of electron transfer, as depicted in Fig. 4h. Mechanistic studies revealed that Co and Mo dual-sites synergistically facilitated O2 adsorption on the surface. Co sites acted as electron donors, promoting the initial reduction of O2 to ·O2- intermediates, while the adjacent Mo sites served as electron acceptors, forming directional electron transfer channels that stabilized and further activated these intermediates. This atomic-scale electronic coupling effectively promoted the conversion of transient ·O2- into 1O2. Consequently, the localized Co-Mo orbital hybridization induced a highly delocalized electron density distribution, generating short-range, efficient interfacial electron pathways that facilitated rapid electron injection into the π* antibonding orbitals of O2. This instantaneous O2 activation resulted in a dramatically enhanced ROS generation rate, which improved by 10-500 times relative to single-metal catalysts and far exceeding previously reported systems.

Catalytic performance of the air-powered oxidation system

The atomically adjacent Co-Mo dual-site catalyst is expected to create short-range electron transfer pathways, facilitating rapid interfacial electron injection into O2 and thereby accelerating their activation into ROS for ultra-fast water purification. To evaluate the catalytic performance of CMCN in air-assisted ROS generation, ibuprofen (IBP) was selected as a probe contaminant. As shown in Fig. 5a and Supplementary Figs. 12, 13, the CMCN/air system demonstrated superior degradation performance with a kobs value of 0.494 min-1, representing 8.8-fold, 7.2-fold, and 20.6-fold enhancements over CN/air, CoCN/air, and MoCN/air systems, respectively. In addition, control experiments confirmed the critical role of O2 in this system. Under nitrogen (N2)-saturated conditions, CMCN and other samples showed minimal removal on IBP (Fig. 5b and Supplementary Fig. 14), which ruled out the effect of adsorption. Also, the self-oxidation of IBP under aeration conditions could be neglected (1.0%). Notably, CMCN attained 69.8% IBP removal using only dissolved oxygen, further improving to 98.4% with aeration. These results conclusively demonstrated that the degradation performance of CMCN primarily originated from molecular oxygen activation rather than direct adsorption or auto-oxidation processes. To further assess the performance of the CMCN/air system, comparisons of kobs values were conducted with other reported catalysts for air or O2 activation (Fig. 5c and Supplementary Table 6). The obtained results indicated that the CMCN catalyst achieved superior kobs (~ 10–103 times higher than other air activation systems) despite the lower catalyst dosage and metal contents (Co = 3.4% and Mo = 8.4% in Supplementary Fig. 15).

Fig. 5: IBP degradation performance.
Fig. 5: IBP degradation performance.
Full size image

a Degradation curves of IBP under aeration with different samples. b Degradation curves of IBP under different gas atmosphere conditions. Reaction conditions: [Catal.] = 0.2 g/L, [aeration rate] = 200 mL/min, [pH] = 5, [IBP] = 10 mg/L, [Temp.] = 20 ± 2 °C. c Comparison of the catalytic performance with other recently reported catalysts under aeration or O2. d Zeta potential and the effect of pH on degradation. e The influence of coexisting substances on IBP degradation. f The toxicity evaluation of different intermediates. Error bars represent the standard deviation of the experiment in triplicate.

Quenching experiments systematically elucidated the ROS contribution mechanism through a series of complementary investigations. The concentrations of all quenching agents were optimized and tested at three representative levels (as detailed in Supplementary Fig. 16), ensuring that the inhibition effects arose from ROS scavenging rather than from side reactions with the catalyst or substrates. The identification of 1O2 as the dominant active species was confirmed by the significantly decrease in degradation rate upon addition of TEMP scavenger, while the negligible inhibition observed with tert-butanol (TBA, k·OH = 6 × 109 M−1·s−1) excluded the involvement of ·OH. Also, the experiment with catalase (CAT) suggested that H2O2 played limited effect on IBP degradation. In addition, p-BQ (k·O2⁻ = 9.8 × 108 M−1·s–1) caused significant inhibition of IBP degradation, confirming the crucial role of ·O2⁻ in the CMCN/air system. Complementary experiments with potassium iodide (KI) confirmed the participation surface-bound ROS (ROSsurface), suggesting they might also take part in the generation pathways of 1O2. To further elucidate the critical role of N-coordinated metal sites, we conducted potassium thiocyanate (KSCN) poisoning experiments (Supplementary Fig. 17). The thiocyanate anions (SCN⁻) effectively deactivated the catalytic centers by forming labile chelates with the metal-Nx moieties, resulting in a reduction of IBP degradation efficiency to 44.4% and a corresponding decrease in kobs to 0.0196 min−1. These findings provided definitive evidence that the N-coordinated short-range Co-Mo dual sites played a pivotal role in facilitating the electronic properties essential for efficient IBP degradation.

The CMCN/air system exhibited high catalytic efficiency within the pH range of 4–8 (Fig. 5d and Supplementary Fig. 18), but its removal efficiency and kobs values slightly decreased as pH increased, which directly corroborated the proposed electron-transfer mechanism. This was because acidic conditions enhanced the O2/·O2− redox potential30,50 and synergized with the catalyst’s short-range electronic structure to facilitate efficient oxygen activation. Also, the zeta potential decreased with increasing pH, matching IBP degradation trends. Since IBP (pKa = 4.9) becomes negatively charged above pH 4.9, electrostatic repulsion reduces its adsorption on the similarly charged catalyst surface, directly limiting degradation efficiency. In addition, the system maintained robust performance in the presence of high-concentration coexisting substances (Cl-, H2PO4-, SO42-, Na+, K+, Mg2+, NH4+, dissolved organic matters (DOM)) (Fig. 5e and Supplementary Fig. 19), demonstrating the kinetic advantages conferred by the short-range Co-Mo coordination. The introduction of HCO3- and CO32- reduced IBP degradation efficiency to around 78.0%, which provided direct experimental evidence for ·O2- as the primary reaction intermediate. Furthermore, the minimal metal leaching (< 4 μg/L) underscored the structural durability imparted in CMCN/air system, while broad temperature tolerance and compatibility with various oxidants (PMS, PS, H2O2) highlighted the practical applicability of the system (Supplementary Figs. 20–22). These findings collectively affirmed that the atomic-adjacent Co-Mo dual-site structure created a uniquely efficient and stable catalytic system for environmental remediation applications. The proposed degradation pathways of IBP in the CMCN/air system were deduced by integrating liquid chromatography-mass spectrometry results (Supplementary Fig. 23 and Supplementary Table 7). In pathway I, IBP was subjected to electrophilic attack, resulting in decarboxylation to produce intermediate P1 (m/z = 162). Then, it underwent hydroxylation to generate P2 (m/z = 178), which hydroxylated to generate P3 (m/z = 192) with a ketocarbonyl group due to the lack of H atoms in the adjacent C atoms. Finally, the further hydroxylation and demethylation produced P5 (m/z = 164). In pathway II, IBP formed intermediates P6 and P7 (m/z = 222) by hydroxylation, followed by oxidation to generate intermediate P5 (m/z = 162). In pathway III, IBP was electrophilically attacked and underwent demethylation to form P10 (m/z = 178) and P11 (m/z = 150). In the end, the benzene rings of the above intermediates were cleaved into smaller molecules such as short-chain fatty acids (P12, P15-P17), aldehydes (P13), and olefinic alcohols (P14), which were further oxidized and finally mineralized to CO2 and H2O. Furthermore, toxicity analysis (T.E.S.T., Fig. 5f) indicated that almost all intermediates exhibited lower toxicity and higher biodegradability than IBP. These results suggested that the CMCN/air system demonstrated remarkable potential for water purification by efficiently degrading IBP through multiple pathways while generating less toxic and more biodegradable intermediates. Therefore, the atomically adjacent Co-Mo dual-site catalyst achieved exceptional water purification efficiency through a short-range electron transfer pathway that enabled rapid oxygen activation, and this system maintained robust activity across wide pH ranges and in complex water matrices.

Practicability and universality of CMCN/air system

Stability, performance in real-water conditions, and broad applicability are critical metrics for evaluating the practical viability of a technology. Firstly, to assess the stability and enhance the practicability of this system, the CMCN catalyst was deposited on a commercially available PVDF membrane and applied in a flow-through catalytic reactor with aeration for IBP degradation (Fig. 6a). Benefiting from the abundant percolating channels and the synergistic effect of Co and Mo within the filtration membrane, the flow-through catalytic system maintained an efficient IBP removal (> 86.2%) and a high-water flux (~ 720 L·m−2·h−1·bar−1) even after 270 min, which was comparable to membrane treatments with traditional chemical oxidants51,52,53,54. In addition, the system demonstrated robust resistance to environmental interferences. Tests conducted in various real water matrices (tap water, sewage, and brackish water) consistently showed high IBP removal efficiencies (Fig. 6b and Supplementary Table 8). Notably, the brackish water used in the tests exhibited high salinity, with major ions such as Mg2+, SO42-, and Cl- at concentrations ranging from tens of thousands to over 700,000 mg·L−1. The CMCN/air system still retained high catalytic efficiency in brackish water, demonstrating strong resistance to high salinity. This robustness was attributed to its reliance on 1O2, which was largely unaffected by high salinity, rather than the ·OH that were more susceptible to ionic interference.

Fig. 6: Evaluation of the practicability and universality of CMCN/air system.
Fig. 6: Evaluation of the practicability and universality of CMCN/air system.
Full size image

a Schematic illustration of the continuous aeration experiment and images of CMCN/PVDF membrane. b–d The universality of CMCN/air system: (b) degradation curves of IBP in actual water and the removal rate for different pollutants, reaction conditions: [Catal.] = 0.2 g/L, [aeration rate] = 200 mL/min, [pH] = 5, [Temp.] = 20 ± 2 °C. c the sulfur resource recovery, and (d) the antibacterial property. Error bars represent the standard deviation of the experiment in triplicate.

For the application universality, this system exhibited broad reactivity toward diverse environmental contaminants. In addition to demonstrating high removal efficiencies for various organic contaminants (Supplementary Fig. 24 and Supplementary Table 9), the CMCN/air system also exhibited remarkable activity in inorganic pollutants conversion. For instance, it achieved a turnover frequency (TOF) of 1.087 s-1 in sulfite (SO32-, S(IV)) oxidation, outperforming most previously reported catalysts (Fig. 6c and Supplementary Table 10). The oxidation products (CaSO4, Na2SO4, and other benign sulfates) are environmentally and recyclable, indicating potential application in wastewater treatment from flue gas desulfurization. In addition to chemical pollutants, the system exhibited high bactericidal efficiency, as demonstrated by the near-complete inactivation (96.7%) of Pseudomonas fluorescens (P. fluorescens) within 30 min (Fig. 6d), showing the system’s rapid and effective disinfection potential.

Apart from the above metrics, the economic advantages of the CMCN/air system are particularly striking when compared to conventional AOPs. As shown in Supplementary Fig. 25 and Supplementary Tables 11, 12, traditional AOPs involve significant expenses for oxidant procurement, transportation, and storage, while O3-based treatment additionally requires energy-intensive generation. In contrast, the air-based oxidation technology requires solely on simple aeration equipment, achieving 1-2 orders of magnitude reduction in operational costs while maintaining competitive treatment efficiency. These above results suggest that the CMCN/air system provides viable ideas for the development of more environmentally friendly, efficient, and energy-saving water remediation technologies, which has high performance in various fields, such as diverse organic contaminants, inorganic pollutants, and pathogenic bacteria. This multifunctional performance and robust operation across varied water matrices indicate that this system has high potential for application in actual wastewater in the future.

Discussion

In summary, this work demonstrates a breakthrough in molecular oxygen activation by rationally engineering the atomic-scale proximity dual-site catalyst, which facilitates smooth and short-range electron transfer pathways through d-d orbital hybridization and electron delocalization. This system overcomes O2 inherent kinetic constraints, which dramatically accelerate O2-to-1O2 conversion in an energy-neutral way. Remarkably, the CMCN/air system achieves superior ROS generation rates (17.0 μmol·L−1) and a reaction rate (kobs = 0.494 min−1) which are comparable to traditional chemical oxidants. The CMCN/air system operates without chemical additives or external energy input, achieving 1-2 orders of magnitude reduction in operational costs. More importantly, this system demonstrates broad applicability across diverse organic contaminants, inorganic pollutants, and pathogenic bacteria. These findings establish short-range dual-site engineering as a paradigm for sustainable and scalable O2 activation, opening pathways toward next-generation energy-neutral environmental technologies.

Methods

Catalyst preparation and optimization

We used the organic metal frame (MOF) precursor to stabilize bimetallic sites and enrich local O2 concentration through its abundant pore channels. Also, Co and Mo sites were introduced simultaneously to ensure atomic-scale proximity between active centers. During pyrolysis, the ZIF-8 organic linkers were decomposed into a nitrogen-doped carbon framework, while the volatile Zn component created a porous structure, effectively suppressing the aggregation of Co/Mo species. In the pyrolysis process of MOF precursor (Co/Mo ZIF-8) at 950 °C, the organic linkers of MOFs were thermally decomposed during high-temperature annealing to incorporate N dopants into the carbon framework. Subsequently, due to the low boiling point of Zn, it could gasify to produce defects and pores, effectively hindering the aggregation of Co and Mo species. After pyrolysis, the obtained powders were treated with 1 M HNO3 to selectively remove any residual metal nanoparticles or loosely bound metal oxides while retaining the atomically dispersed Co and Mo active sites. This acid-washing step ensures that the catalytic performance arises from well-dispersed active centers rather than from bulk metal particles. To provide a comparative perspective, different Co/Mo molar ratios (3, 6, and 9) were synthesized. The Co/Mo molar ratios of 6 was selected for in-depth study due to its optimal catalytic performance. The samples obtained after pyrolysis were recorded as CMCN-a, CMCN, and CMCN-b. The synthesis procedures of the CMCN precursor (Co/Mo ZIF-8) were as follows.

Synthesis process

Briefly, 2.21 g of 2-methylimidazole were dissolved in methanol under vigorous stirring at room temperature, which was recorded as solution A. Zn(NO3)2·6H2O, Co(NO3)2·6H2O, and Na2MoO4·2H2O (the radio of Co/Mo was 3, 6, and 9) were dissolved in methanol under vigorous stirring at room temperature, which was recorded as solution B. Then, pouring solution A into solution B quickly. After stirring for 30 min, the mixture solution was transferred to a 100 mL PTFE-lined stainless steel high-pressure vessel for hydrothermal treatment (120 °C, 4 h). Co/Mo ZIF-8 was obtained by centrifugation, washing, and drying. In addition, Co ZIF-8 and Mo ZIF-8 were prepared by a similar procedure without the addition of Na2MoO4·2H2O and Co(NO3)2·6H2O, respectively. Besides, ZIF-8 was synthesized without the introduction of Co(NO3)2·6H2O as well as Na2MoO4·2H2O.

Co/Mo ZIF-8, Co ZIF-8, Mo ZIF-8, and ZIF-8 were respectively pyrolyzed at 950 °C for 2 h with the pyrolysis heating rate of 5 °C/min under N2. After naturally cooling down to room temperature, the black powders were collected and labeled as CMCN, CoCN, MoCN, and CN, respectively. Then, put the above powders into a solution of 1 M HNO3 to remove the metal nanoparticles.

Characterizations

The crystal structure of as-prepared samples was measured by X-ray diffraction (XRD, Bruker D8 Advance diffractometer, Gobel mirror monochromated Cu Kα radiation, λ = 1.54056 Å). The chemical composition and state of the catalysts’ surface were verified by X–ray photoelectron microscopy (XPS), which was acquired on a Thermo Escalab 250 using an Al Kα X-ray source and all the binding energies were calibrated using the C 1 s peak (284.8 eV). The surface functional groups of as-prepared samples were compared by Fourier transform infrared spectroscopy (FTIR), which was measured by a Thermo Nicolet iS50 Spectrometer. The X-ray absorption spectroscopy (XAFS) study was performed at the BL08B2* of SPring-8 (8 GeV, 100 mA), Japan, in which, the X-ray beam was mono-chromatized with water-cooled Si (111) double-crystal monochromator and focused with two Rh coated focusing mirrors with the beam size of 2.0 mm in the horizontal direction and 0.5 mm in the vertical direction around sample position, to obtain X-ray adsorption fine structure (XAFS) spectra both in near and extended edge. Co foil, CoO, Co3O4, CoPc, Mo foil, and MoO3 samples were used as references. The oxygen-programmed temperature-raising desorption (O2-TPD) analyses was obtained by Micromeritics AutoChem II 2920 (USA). Scanning electron microscopy (SEM) was performed with a SIGMA 500 operated at an accelerating voltage of 10 kV. Morphological information was obtained on a Zeiss Neon 40EsV FIBSEM attached with an energy dispersive spectroscopy (EDS). Transmission electron microscopy (TEM) images were obtained with a JEOL 1400. Electron spin resonance (ESR) spectra were obtained by a Bruker EMXplus-6/1 instrument (Germany) using 5,5-dimethyl-1-pyrroline-N-oxide (DMPO) and 2,2,6,6-Tetramethylpiperidine (TEMP) as spin-trapping agents to detect ·OH, ·O2- and 1O2. The metal ions leaching and content of elements of Co and Mo of as-prepared samples were detected by inductively coupled plasma-mass spectrometry (ICP-MS, Agilent 7800, USA). The electrochemical impedance spectroscopy (EIS) and linear sweep voltammetry (LSV) were measured on the electrochemical workstation (CHI660E) with a standard three electrode system, in which the catalyst-coated carbon cloth was the working electrode, the Pt sheet was the counter electrode, and the Ag/AgCl electrode was the reference electrode. To prepare the working electrode, 5 mg of catalyst was dispersed in 0.5 mL ethanol and 0.02 mL Nafion solution. Following ultrasonic dispersion, 100 μL of the slurry was dropped onto the glassy carbon electrode and subsequently dried under a baking lamp. All electrochemical tests were conducted in 0.1 M Na2SO4 electrolyte. The work function (Φ) can be determined through testing the UPS. Specifically, the linear part of the UPS is extrapolated to the baseline, and then the following formula is applied for calculation:

$$\Phi=h\nu \,-\,({E}_{{cutoff}}\,-\,{E}_{f})$$
(1)

where hν is the photon energy set at 21.22 eV, Ecutoff is the low-energy cutoff edge, and Ef is the Fermi level.