Domino effect enables dynamic electronic structure evolution for stable and efficient water purification

Abstract

The purification of water contaminated by persistent organic pollutants remains a critical global challenge, largely due to the intrinsic trade-off between high catalytic activity and long-term stability in advanced oxidation processes. Despite the tremendous potential of single-atom catalysis, its traditional static design strategies cannot adapt to dynamic and multi-step reactions. Here, we break the limitation of this static design by reporting a domino catalysis within a heteronuclear MgCo dual-single-atom catalyst. In this system, the Mg site selectively photoreduces O₂ to H₂O₂ under light irradiation. This in-situ generated H₂O₂ and its derived surface hydroxyl groups act as dynamic modulators, progressively tailoring the electronic structure of the adjacent Co site. This on-demand electronic refinement optimally lowers the energy barriers for subsequent peroxymonosulfate activation and radical desorption, thereby driving efficient pollutant degradation while intrinsically stabilizing the Co center against leaching. Experiments and theoretical calculations elucidate that this seamless and spatially ordered cascade avoids competitive adsorption and maximizes efficiency. The system exhibits the highest normalized rate constant of 595.3 min⁻¹ M⁻¹ among various refractory contaminants over thirty regeneration cycles, outperforming the most state-of-the-art catalytic systems. This work introduces a dynamic and self-evolving catalytic approach that breaks the traditional activity-stability trade-off, suggesting a potential pathway toward effective and sustainable water purification technologies.

The authors develop a domino catalytic system that dynamically reconstructs active sites’ electronic structure, lowering reaction barriers and inhibiting metal leaching for efficient and durable water purification.

Introduction

Increasing water pollution by persistent organic contaminants poses a serious threat to global ecological security and human health¹. Advanced oxidation processes (AOPs), particularly Fenton-like reactions capable of generating oxidative reactive oxygen species, stand as pivotal technologies for water decontamination². Among various intensification strategies, coupling these reactions with inexhaustible solar energy presents a sustainable avenue³. However, the widespread implementation of such photo-enhanced Fenton-like systems is critically hampered by the fundamental and persistent trade-off between simultaneously achieving high catalytic activity and long-term stability⁴. High activity demands rapid redox cycling of the metal center, a process often accelerated by photo-generated charge carriers⁵. Yet, this frequently triggers irreversible metal leaching or deactivation, compromising stability and creating an intrinsic contradiction that limits practical application⁶.

Recent advances in single-atom catalysis (SACs), with its maximized atomic utilization efficiency and tailorable electronic structures, offer the hope for optimizing this activity-stability nexus⁷. Conventional SACs design strategies predominantly focus on static electronic modulation via pre-engineering the coordination environment⁸. While effective to a degree, this static approach often fails under dynamic reaction conditions, especially in complex multi-step AOPs⁹. A more desirable yet significantly underdeveloped paradigm is the in-situ, dynamic, and progressive modulation of the active site’s electronic state during the reaction sequence¹⁰. Such dynamic tuning could optimally adapt the catalyst to each elementary step, enhancing both activity and durability¹¹. The primary obstacle lies in the spatial and temporal decoupling of multiple reaction steps¹². Reactants, intermediates, and modulators often compete for adsorption on a single site, leading to mutual quenching or separation across distant sites, which drastically reduces energy and atom utilization efficiency¹³.

Herein, we conceptualize and propose a domino effect within the heteronuclear dual single atom catalysts (DSACs) to masterfully resolve this dilemma¹⁴. Mimicking a falling domino chain where an event automatically triggers the next, our system orchestrates a seamless, spatially ordered cascade of reactions¹⁵. We construct a catalyst featuring paired but functionally differentiated magnesium and cobalt single atoms¹⁶. Under light irradiation, the magnesium site initiates the cascade by selectively photoreducing O₂ to H₂O₂¹⁷. This in-situ generated H₂O₂ then acts not merely as a precursor, but as the dynamic modulator whose derived surface hydroxyl groups progressively tailor the electronic structure of the adjacent cobalt site. This on-demand electronic refinement optimally lowers the energy barriers for subsequent peroxymonosulfate activation and radical desorption, driving efficient pollutant degradation while inherently stabilizing the cobalt center against leaching. Through degradation experiments, in-situ spectroscopic characterization, and theoretical calculations, we elucidate this domino-triggered dynamic electronic modulation process and its contribution to enhancing catalytic performance and durability. Furthermore, life cycle assessment confirms the environmental and economic merits of this system. This work transcends conventional static catalyst design, introducing a dynamic and self-evolving catalytic paradigm that breaks the activity-stability trade-off, thereby suggesting a potential pathway toward effective and sustainable water purification technologies.

Results

Characterizations of dual single atom catalysts

The heteronuclear magnesium cobalt dual single atom catalysts (MgCo/CN DSACs) were synthesized by a simple hydrothermal method (Fig. 1a). The MgCo/CN DSACs had a sparse and three-dimensional nanosheet structure in scanning electron microscope (SEM) (Supplementary Fig. 1). The aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (AC HAADF-STEM) images with both high and low magnifications (Fig. 1b, Supplementary Figs. 2 and 3) exhibited the bright dual single atoms in yellow dashed ellipses, indicating that the dual single atom structure was successfully synthesized. The energy-dispersive X-ray elemental spectra (EDS) mapping showed the coexistence of C, N, Mg and Co elements. The diffraction peak and functional group of CN were observed in the X-ray diffraction (XRD) patterns (Fig. 1c) and Fourier transform infrared (FT-IR) spectra (Fig. 1d)¹⁸, whereas no diffraction peaks of cobalt oxides were evident in the former, nor were any signals of their functional groups detected in the latter. The Co and Mg K-edge X-ray absorption near-edge structure (XANES) spectra (Fig. 1e, h) showed that the absorption edge of Co lay between those of Co foil and CoO, while that of Mg lay between those of Mg foil and MgO, but closer to Mg foil. This indicated that the oxidation states of both Co and Mg in the 4MgCo/CN DSACs ranged between 0 and +2, yet they differed from each other. The linear combination fitting (LCF) analysis (Supplementary Figs. 3 and 4, Supplementary Tables 1 and 2) was performed to quantify the proportions of different Co and Mg species with R-factor of 0.0078 and 0.0185, respectively. The proportions of Co⁰, Co²⁺ and Co³⁺ were 25.1%, 32.4% and 42.5%, respectively, while those of Mg⁰ and Mg²⁺ were 77.6% and 22.4%, respectively. Based on the LCF results, the average oxidation states of Co and Mg in the 4MgCo/CN DSACs were estimated to be 1.923 and 0.448, respectively (Fig. 1f, i). Fourier transformed (FT) k³-weighted X-ray absorption fine structure (EXAFS) spectra of 4MgCo/CN DSACs (Supplementary Figs. 5 and 6), the peaks corresponding to Co–N at 1.48 Å and Mg–N at 1.68 Å were observed^19,20. No peaks attributable to Co–Co, Co–O, Mg–Mg and Mg–O bonds were detected, suggesting the absence of metallic or oxide clusters. Additionally, no Mg–Co bond was observed in the EXAFS spectra of 4MgCo/CN DSACs, demonstrating that no direct bonding between Mg and Co atom. The formation energy of 4MgCo/CN without the Mg–Co bond was −0.75 eV, which was more negative than those of 4MgCo/CN with the Mg–Co bond (1.15 eV), CoCo/CN without (4.59 eV) and with (4.38 eV) the Co–Co bond, and MgMg/CN without (−0.02 eV) and with (2.42 eV) the Mg–Mg bond. This indicated that the 4MgCo/CN without the Mg–Co bond configuration was thermodynamically more favorable to form (Supplementary Fig. 7). Based on the FT-EXAFS fitting results in R-space (Fig. 1g, j, Supplementary Tables 3 and 4), it was deduced that the coordination configurations of 4MgCo/CN DSACs were identified as Mg–N₄ and Co–N₄. We applied the kernel method of frequency statistics combined with Gauss fitting to analyze the interatomic distances. We collected the interatomic distance from more than 200 atoms in STEM images, all of which were below 11.91 Å, a threshold distance for non‑interacting dual single atoms, as determined by net charge difference calculation (Supplementary Figs. 8 and 9). Quantitative intensity profile analysis (Supplementary Figs. 2 and 10) exhibited that the predominant configuration consisted of MgCo dual single atoms, with the highest prevalence of interatomic distance ranging from 6.41 Å to 6.53 Å. This distance matched well with the 6.48 Å atomic distance of MgCo in the simulated MgCo/CN model, where Mg and Co adopt isolated N₄ coordination without Mg–Co bond. The same estimation method has been reported in previous work²¹. The spin states of the catalysts were characterized by electron paramagnetic resonance (EPR) spectra (Supplementary Fig. 11) and temperature-dependent inverse magnetic susceptibility measurements (Supplementary Fig. 12). These results collectively confirmed the successful synthesis of MgCo/CN DSACs featuring Mg–N₄ and Co–N₄ configurations without Mg–Co bond.

Fig. 1. Structural characterization of the MgCo/CN DSACs.

a Synthesis of MgCo/CN DSACs. b Aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (AC-HAADF-STEM) image and energy-dispersive X-ray elemental spectra (EDS) mapping of 4MgCo/CN DSACs. Yellow circles represent the locations of the dual single atoms dispersed on the CN. Scale bar: 2 nm. The unedited aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (AC-HAADF-STEM) is added into the Source Data file. Image is representative of three independent measurements with similar results. The unedited AC-HAADF-STEM image is provided in Source Data. c X-ray diffraction (XRD) patterns and d Fourier transform infrared (FT-IR) spectra of catalysts, respectively. e Normalized Co K-edge X-ray absorption near-edge structure (XANES) spectra, f corresponding oxidation state of Co species and g the fitting curves of Co K-edge extended X-ray absorption fine structure (EXAFS) spectra in 4MgCo/CN DSACs. The functional form and fitting parameters of f are exhibited in note for Supplementary Table 1. h Normalized Mg K-edge X-ray absorption near-edge structure (XANES) spectra, i corresponding oxidation state of Mg species and j the fitting curves of Mg K-edge extended X-ray absorption fine structure (EXAFS) spectra in 4MgCo/CN DSACs. The functional form and fitting parameters of i are exhibited in note for Supplementary Table 2. The above data is representative of three independent measurements with similar results. Source data are provided as a Source Data file.

Breaking trade-off of activity and stability for water decontamination

The 2MgCo/CN DSACs degraded 82.0% of carbamazepine (CBZ) in the PMS/vis system, which was much higher than those achieved by CN (39.8%), 4Mg/CN SACs (69.0%) and 4Co/CN SACs (69.2%), respectively. The CBZ degradation efficiency first increased and then decreased with increasing Mg and Co ratio from 20% to 60% and the optimal degradation efficiency (92.7%) was obtained by 4MgCo/CN DSACs (Supplementary Fig. 13). 4MgCo/CN DSACs had 53.2% of total organic carbon (TOC) removal efficiency, which was much higher than those of CN (18.6%), 4Mg/CN (33.5%) and 4Co/CN (34.1%), respectively (Supplementary Fig. 14). The degradation rate constant towards various contaminants, including sulfamethoxazole (SMX), CBZ, tetracycline hydrochloride (TCH), 4-chlorophenol (4-CP) and bisphenol A (BPA) in the 4MgCo/CN DSACs/PMS/vis system with 0.1 mM PMS was superior to most of those in previous reports (Fig. 2a, Supplementary Table 5)^22–45. The CBZ degradation efficiency of 4MgCo/CN DSACs remained almost unchanged over 30-time cycles in the PMS/vis system, while the degradation efficiency of 4Co/CN SACs on CBZ decreased by 36.7% over 30-time cycles in the PMS/vis system (Fig. 2b). The Co dissolution concentration in the 4MgCo/CN DSACs/PMS/vis system (0.43 mg/L) was below the emission standard of GB 25467-2010 (1.0 mg/L) in China, which was also much lower than that in the 4Co/CN SACs/PMS/vis system (1.39 mg/L) after 30-time cycles^46,47. These results illustrated that both CBZ degradation activity and stability were superior in the 4MgCo/CN DSACs/PMS/vis system. The methanol (MeOH, 100.0 mM), isopropyl alcohol (IPA, 100.0 mM), methyl phenyl sulfoxide (PMSO, 0.2 mM), β-carotene (β-caro, 0.2 mM), sodium fluoride (NaF, 2.0 mM), catalase (CAT, 200.0 units/mL) and sodium oxalate (Na₂C₂O₄, 1.0 mM) were selected as scavengers for •SO₄⁻, •OH, high-valent metal-oxo, ¹O₂, OH_surf, H₂O₂ and h⁺, respectively^48–51. The 3.5%, 33.6%, 52.5%, 67.7%, 14.7%, 10.4% and 85.4% of CBZ were degraded after the addition of MeOH, IPA, PMSO, β-caro, NaF, CAT and Na₂C₂O₄, respectively, indicating that •SO₄⁻ and OH_surf derived from H₂O₂ were major active species in the 4MgCo/CN DSACs/PMS/vis system (Fig. 2d). This was different from the reaction dominated by only •SO₄⁻ in the 4Co/CN SACs/PMS/vis system (Supplementary Fig. 15). The electron spin resonance (ESR) spectroscopy also illustrated that the signals of •SO₄⁻-DMPO and •OH-DMPO by 4MgCo/CN DSACs were higher than those by 4Co/CN SACs (Supplementary Fig. 16). In the Fenton-like system, 50.6%, 70.5% and 70.2% of CBZ were degraded by β-caro, NaF and CAT, respectively, while in the photocatalytic system, 13.7%, 8.9% and 9.6% of CBZ were degraded by β-caro, NaF and CAT, respectively. The excellent photoadsorption of catalysts was observed in UV-Vis diffuse reflectance spectroscopy (DRS) (Supplementary Fig. 17). These results further demonstrated that visible light contributed to the in-situ generation of H₂O₂ for the dominance of OH_surf in the contaminant removal.

Fig. 2. Catalytic activity and stability for water decontamination.

a Comparison of the kinetics of various contaminants by the state-of-the-art catalysts in peroxymonosulfate (PMS) activation system. (SMX sulfamethoxazole, CBZ carbamazepine, TCH tetracycline hydrochloride, 4-CP 4-chlorophenol, BPA bisphenol A). b Degradation of carbamazepine (CBZ) by 4MgCo/CN DSACs over 30-time cycles in the peroxymonosulfate (PMS)/visible light (vis) system. c Ratio of Co²⁺ and Co³⁺ for 4Co/CN SACs and 4MgCo/CN DSACs under different conditions. (cat. catalyst, PMS peroxymonosulfate, vis visible light). d Removal of carbamazepine (CBZ) by 4MgCo/CN DSACs under different systems after the addition of scavengers. (MeOH methanol, IPA isopropyl alcohol, PMSO methyl phenyl sulfoxide, β-caro β-carotene, NaF sodium fluoride, CAT catalase, Na₂C₂O₄ sodium oxalate, PMS peroxymonosulfate, vis visible light). e Removal of carbamazepine (CBZ) by 4MgCo/CN DSACs after the addition of catalase (CAT), sodium fluoride (NaF) and methanol (MeOH) with different dosages in the PMS/vis system, respectively. f Removal of carbamazepine (CBZ) by 4Co/CN SACs after the addition of catalase (CAT), sodium fluoride (NaF) and methanol (MeOH) with different dosages in the PMS/vis system, respectively. The p value is less than or equal to 0.05. The functional form and fitting parameters for e, f are shown in Supplementary Table 6. g Active species contribution under different reaction systems mediated by typical DSACs and SACs. (PMS: peroxymonosulfate, vis: visible light). Experimental conditions: [catalyst] = 0.01 g/L, [PMS] = 0.1 mM, [contaminants] = 10 mg/L. Date for a, c, g are representative of three independent measurements with similar results. Data for b, d, e, f are presented as mean values ± SD from three independent experiments (n = 3). Source data are provided as a Source Data file.

The relationship among degradation efficiencies with the dosages of CAT, NaF and MeOH was further estimated. The correlation constant R² of 0.991 between degradation efficiency and MeOH dosage was far higher than the others, meaning that OH_surf secondary from H₂O₂ might be the indirect active species and •SO₄⁻ might be the direct active species in the 4MgCo/CN DSACs/PMS/vis system (Fig. 2e). In contrast, the correlation constants of R² for the relationship among degradation efficiencies of CAT dosage, NaF dosage and MeOH dosage in the 4Co/CN SACs/PMS/vis system were all no more than 0.732. This indicated that •SO₄⁻, H₂O₂ and OH_surf were indirect active species in the 4Co/CN SACs/PMS/vis system (Fig. 2f). The detailed fitting information was shown in Supplementary Table 6. The ratio of Co²⁺ in 4Co/CN SACs increased from 41% to 42% and the ratio of Co²⁺ in 4MgCo/CN DSACs increased from 57% to 59% after visible light irradiation, which could be attributed to the fact that Co³⁺ was reduced into Co²⁺ by photogenerated electrons. The ratio of Co²⁺ for 4Co/CN SACs decreased from 41% to 39% after the reaction in the PMS/vis system due to the redox from PMS to •SO₄⁻, whereas the ratio of Co²⁺ increased from 57% to 69% in the 4MgCo/CN DSACs/PMS/vis system (Fig. 2c). The major contributions of •SO₄⁻ and OH_surf were also observed in the MgMn/CN DSACs/PMS/vis and MgFe/CN DSACs/PMS/vis systems, which differed from those in photo-Fenton-like system mediated by SACs and Fenton-like system mediated by DSACs and SACs, respectively (Fig. 2g, Supplementary Table 7).

Domino effect for peroxymonosulfate activation

The 4MgCo/CN DSACs activated 87.2% of PMS, which was higher than that by 4Co/CN SACs (66.6%) (Fig. 3a). Furthermore, in the in-situ Raman spectra of 4MgCo/CN DSACs, a new peak appeared at 815.7 cm⁻¹ after a 3-min reaction, and its intensity became more pronounced after 5 min, whereas this peak was absent in the in-situ Raman spectra of 4Co/CN SACs (Supplementary Fig. 18 and 19)⁵². This indicated that the interaction between MgCo dual atoms contributed to PMS activation. The PMS activation efficiency by 4MgCo/CN DSACs increased from 45.2% to 77.3% when exogenous H₂O₂ was raised from 0.0 mM to 3.97 mM (Supplementary Fig. 20). The larger ring current of 4MgCo/CN DSACs compared to that of 4Co/CN SACs mainly originated from the higher amount of H₂O₂ generated by the former (Fig. 3b). Both 4MgCo/CN DSACs and 4Co/CN SACs had an n value closer to 2 (Fig. 3c). The selectivity of 4MgCo/CN DSACs (86.1%) was higher than that of 4Co/CN SACs (46.1%), indicating that dual single atom sites could increase the 2e⁻ selectivity for in-situ H₂O₂ generation (Fig. 3d). No H₂O₂ was generated under the dark condition, despite oxygen aeration. However, 2.26 mM and 1.87 mM of H₂O₂ were generated within the first 4 min by both catalysts, and then activated to 0.39 mM and 0.93 mM, respectively, by 4MgCo/CN DSACs and 4Co/CN SACs in the PMS/vis/CBZ system (Fig. 3e and Supplementary Fig. 21). The H₂O₂ generation for these two catalysts was attributed to their conduction band potentials being higher than that of O₂/H₂O₂ (Supplementary Fig. 22). The photogenerated electrons could reduce oxygen into H₂O₂ under light irradiation. The H₂O₂ production increased from 1.14 mM to 3.97 mM as the dosage of 4MgCo/CN DSACs increased from 0.005 to 0.05 g/L in the PMS/vis/CBZ system (Supplementary Fig. 23). Supplementary Fig. 24 showed that H₂O₂ was gradually generated up to 4.38 mM in the PMS/vis/H₂O system within 24 min, which was higher than those in the PMS/vis/CBZ system (0.39 mM), the vis/CBZ system (0.03 mM), and the vis/H₂O system (0.23 mM), respectively. These results indicated that visible light was essential for H₂O₂ generation, that CBZ could consume part of the in-situ generated H₂O₂, and that PMS facilitated the photoreduction of H₂O₂. As shown in Supplementary Figs. 25 and 26, the H₂O₂ concentration decreased rapidly after introducing nitrogen (N₂), indicating that O₂ was essential for H₂O₂ generation. Upon addition of the electron-trapping agent potassium bromate (KBrO₃), the H₂O₂ concentration dropped sharply, demonstrating that the two-electron oxygen reduction reaction was the primary pathway for photocatalytic H₂O₂ production. The H₂O₂ concentration decreased by 94.1% after adding p-benzoquinone (p-BQ), suggesting that •O₂⁻ was a key intermediate in the photoreaction. The negligible H₂O₂ concentration observed upon addition of tert-butyl alcohol (TBA) indicated that •OH was not an intermediate in H₂O₂ production. In contrast, adding ethanol (EtOH) and benzyl alcohol (BnOH) significantly increased H₂O₂ production, mainly because these alcohols capture photogenerated holes and supply sufficient protons, thereby facilitating oxygen reduction for H₂O₂ generation. These results indicated that CBZ did not affect the types of key intermediates in the in-situ photocatalytic generation of H₂O₂. The Gibbs free energy for the two-electron oxygen reduction to H₂O₂ in the presence of PMS by 4MgCo/CN DSACs was negative (Supplementary Fig. 27), indicating that H₂O₂ generation was thermodynamically spontaneous. Supplementary Fig. 28 exhibited the time-resolved in-situ DRIFTS profiles of 4MgCo/CN DSACs. In the 4MgCo/CN DSACs/PMS/vis system, the characteristic O–O stretching vibrations were detected for surface-adsorbed species, including O₂* (965 cm^–1), •O₂^–* (1151 cm^–1), •OOH* (1205 cm^–1), and HOOH* (1369 cm^–1)⁵³. The intensities of these peaks progressively increased with reaction time from 0 to 24 min.

Fig. 3. Domino effect from H₂O₂ generation to peroxymonosulfate activation.

a PMS activation by different catalysts. b Rotating ring disk electrode (RRDE) polarization curves for 4MgCo/CN DSACs and 4Co/CN SACs are acquired in an O₂-saturated 0.1 M KOH electrolyte at 1600 rpm, with ring currents and disk currents measured simultaneously. The insert is the enlarged graph with the potential from −1.6 V to −1.3 V. c The average electron transfer number (n) and d H₂O₂ selectivity varied systematically with applied potential. e H₂O₂ generation concentration with different reaction time by 4MgCo/CN DSACs and 4Co/CN SACs in the peroxymonosulfate (PMS)/visible light (vis)/carbamazepine (CBZ) system, respectively. Gray represents the adsorption-desorption process under the dark condition. f The correlation plot of carbamazepine degradation rate constant with various possible influencing factors. A, B, C, D and E represented mass of Co, photoadsorption, crystallinity, OH_surf contribution ratio and Co spin state. g Isotopic mass spectra of surface hydroxyl radical produced by 4MgCo/CN DSACs and 4Co/CN SACs in the PMS/vis system with the ratio of H₂O¹⁶ to H₂O¹⁸ (1:1). h Degradation of benzoic acid (BA) by 4MgCo/CN DSACs with different sodium fluoride (NaF) dosages. i Co ion percentage ratio of 4MgCo/CN DSACs and 4Co/CN SACs under different conditions, respectively. (CAT: catalase, NaF: sodium fluoride) j electrochemical impedance spectroscopy (EIS) and k electrochemical active surface area (ECSA) of 4MgCo/CN DSACs and 4Co/CN SACs under different conditions, respectively. (CAT catalase, NaF sodium fluoride). The confidence interval and predicting interval are 95%, respectively. The functional form and fitting parameters are shown in Supplementary Table 8. Date for b, c, d, g, i, j, k are representative of three independent measurements with similar results. Data for a, e, h are presented as mean values ± SD from three independent experiments (n = 3). Source data are provided as a Source Data file.

In the ESR spectra (Supplementary Fig. 29), a distinct characteristic signal of •O₂^– was observed under light irradiation, whereas the signal was barely detectable in the dark. Collectively, these results demonstrated that the dominant route for H₂O₂ generation was the two-electron oxygen reduction. A correlation plot of the CBZ degradation rate constant within the first 4 min (k₁) and that from 4 min to 24 min (k₂) was generated against various possible influencing factors, including Co mass (A), photoadsorption (B), crystallinity (C), OH_surf contribution ratio (D) and Co spin state (E). This result illustrated that the degradation of CBZ was mainly related to OH_surf from in-situ generated H₂O₂ (p ≤ 0.05) (Fig. 3f), indicating that the photoadsorption and OH_surf contribution ratio were crucial factors for k₁ and k₂, respectively. Given that visible light was used to drive photoreduction of oxygen to H₂O₂, which served as a crucial intermediate for OH_surf. it could be deduced that CBZ degradation was primarily related to OH_surf from in-situ generated H₂O₂.

Almost no H₂¹⁸O-labeled p-HBA peak (m/z = 137.1) was observed in the 4MgCo/CN DSACs/PMS/vis system, indicating that OH_surf originated from in-situ H₂O₂ generated from oxygen in the air rather than from oxygen in PMS or H₂O. However, the H₂¹⁸O-labeled p-HBA peak (m/z = 139.1) was found in the 4Co/CN SACs/PMS/vis system⁵⁴. This indicated that •OH mainly came from the reaction between •SO₄⁻ and H₂O in the 4Co/CN SACs/PMS/vis system (Fig. 3g and Supplementary Fig. 30). The degradation efficiency of BA decreased with increasing NaF concentration, and it was remarkably reduced from 64.2% to 15.7% after the addition of 9.0 mM NaF (Fig. 3h). The activation efficiency of PMS was also reduced from 87.2% to 27.9% with increasing NaF concentration from 0 to 9 mM (Supplementary Fig. 31). The Co²⁺ ratio in 4Co/CN SACs was enhanced from 33% to 54% and 37% by CAT and NaF, respectively. However, the Co²⁺ ratio in 4MgCo/CN DSACs was reduced from 57% to 42% and 40% by CAT and NaF, respectively. These results illustrated that OH_surf from H₂O₂ contributed to the reduction of Co³⁺ to Co²⁺ in the 4MgCo/CN DSACs/PMS/vis system, which differed from that in the 4Co/CN SACs/PMS/vis system (Fig. 3i). This was supported by the increased radius in electrochemical impedance spectroscopy (EIS) (Fig. 3j) and the reduced C_dl (5.26 mF cm⁻²) in electrochemical active surface area (ECSA) from cyclic voltammetry (CV) profiles⁵⁵ (Fig. 3k, Supplementary Fig. 32 and Supplementary Table 8) observed for 4MgCo/CN DSACs after the addition of NaF and CAT.

Dynamic electronic structure evolution from oxidation reaction

To trace the effect of in-situ generated OH_surf on the conversion from PMS to •SO₄⁻, Gibbs free energy (ΔG) curves (Fig. 4a) were evaluated. The inclined adsorption of H₂O₂ and PMS on the Mg and Co sites with the adsorption energies of −3.70 and −10.62 eV was confirmed, respectively (Supplementary Fig. 33). The activation from PMS to •SO₄⁻ by MgCo/CN DSACs@OH_surf was spontaneous, while it needed to overcome an energy barrier for MgCo/CN DSACs@H₂O₂ and MgCo/CN DSACs, respectively. The adsorption of ^*SO₄ on MgCo/CN DSACs@OH_surf-PMS (−7.24 eV) was much weaker than that by MgCo/CN DSACs (−10.62 eV), indicating that OH_surf contributed to ^*SO₄ desorption. The activation energy barrier of PMS by MgCo/CN DSACs@H₂O₂ (0.80 eV) was smaller than that by MgCo/CN DSACs (1.95 eV). The spontaneous generation of OH_surf from H₂O₂ confirmed that OH_surf was more easily generated from H₂O₂. This contributed to PMS activation rather than from PMS (Fig. 4b). The transferred electron charge (∆q) for MgCo/CN@OH_surf-PMS (0.51 e) was higher than those for Mg/CN@H₂O₂ (0.27 e) and Co/CN@PMS (0.28 e) but not higher than that of MgCo/CN@H₂O₂-PMS (0.62 e). The d-band center of MgCo/CN DSACs@OH_surf (−0.93 eV) was much farther from the Fermi energy level compared to those of MgCo/CN DSACs@H₂O₂ (−0.12 eV) and MgCo/CN DSACs (−0.067 eV) (Supplementary Fig. 34). However, the longer O−O bond in PMS was observed after adsorption on MgCo/CN DSACs@OH_surf (1.87 Å) than MgCo/CN DSACs@H₂O₂ (1.50 Å) and MgCo/CN DSACs (1.23 Å) was observed. These results meant that the superior PMS activation by MgCo/CN DSACs@OH_surf was induced not only by the high electron density of Co. This was completely different from the traditional Fenton-like processes⁵⁶.

Fig. 4. Dynamic electronic structure evolution analysis during Fenton-like reaction.

Energy profiles of a sulfate radical (•SO₄⁻) and b surface hydroxyl radical (OH_surf) formation for MgCo/CN DSACs and Co/CN SACs by peroxymonosulfate (PMS) activation. Green and pink represent peroxymonosulfate adsorption and activation process, respectively. c Differential charge density and spin electron density models of different models. d Partial density of states (PDOS) of Co 3 d in different catalyst models and Op orbital of peroxymonosulfate (PMS). Relationship among e adsorption energy (E_ad) and energy for d-band center (E_d) and f transferred electron amount between the catalyst and PMS (∆q) and O−O bond length (I_O−O) with number of surface hydroxyl radical (OH_surf). g Energy variation of d_xy and d_x²-_y² as the increase in number of surface hydroxyl radical (OH_surf). h Orbital structure between peroxymonosulfate (PMS) and different models. Alternating green and pink are mainly used to distinguish the catalysts and their orbital interactions with H₂O₂, PMS or OH_surf. i The dynamic-catalytic process by dynamic cobalt d-orbital reconstruction for breaking activity-stability trade-off in water decontamination. Source data are provided as a Source Data file.

The orbital hybridization between the Co of different catalysts and Op of PMS and ^*SO₄ was illustrated in Fig. 4d and Supplementary Fig. 35. The overlap between Co d_xy and d_x²_-y² in MgCo/CN DSACs@OH_surf -PMS with Op of PMS and ^*SO₄ was larger than that of Co 3 d orbital with PMS and ^*SO₄. This demonstrated that OH_surf contributed to PMS activation and ^*SO₄ desorption. To further understand the interaction of OH_surf and Co site, the relationship among E_ads, E_d, ∆q and I_O−O with the amount of OH_surf was correlated (Fig. 4e, f). It was found that an optimal amount of OH_surf contributed to the adsorption of PMS, the extension of the O−O bond of PMS, the reduction of d-band center and increase of transferred electrons between the catalyst and PMS. Figure 4g, h, Supplementary Fig. 36 and Supplementary Table 9 showed that OH_surf induced the reconstruction of Co 3 d orbital, leading to the shift of the energy level for d_xy, d_x²_-y², d_xz, d_z² and d_yz orbitals to lower energy levels in MgCo/CN DSACs@OH_surf-PMS. The energies of d_xy and d_x²_-y² orbitals increased from −0.92 to −0.71 eV and from −0.93 to −0.70 eV with the growing number of OH_surf groups. The dynamic energy variations of d_xy and d_x²_-y² were observed. However, these fluctuating energies remained between the Op orbital energies of PMS and SO₄, enabling continuous electron transfer between the Co d_xy or d_x²_-y² orbitals and Op orbital of PMS, as well as appropriate desorption of ^*SO₄. Supplementary Fig. 37 exhibited that the Co d-band center obviously shifted as the amount of OH_surf varied. Supplementary Figs. 38 and 39 showed the strong correlations among the d-band center with PMS adsorption energy (R² = 0.9992), electron transfer charge (R² = 0.9326), and O–O bond length (R² = 0.9948), but a weak correlation with O–H bond length (R² = 0.5262). These results indicated that OH_surf-modulated d-band center facilitated PMS adsorption and activation into •SO₄⁻. Moreover, the valence electron densities of Co sites were 5.09 e, 5.53 e, 6.05 e, and 5.65 e as the amount of OH_surf increased (Supplementary Fig. 40), indicating that appropriate OH_surf adsorption on the Mg site modulated the electron density of the Co site via long-range interaction.

To track electronic change following the modulation of the d-band center, we performed electron paramagnetic resonance (EPR) spectroscopy, electrochemical impedance spectroscopy (EIS), ECSA measurement and Co K-edge XANES spectra by the 4MgCo/CN DSACs after reaction, respectively. The EPR spectra (Supplementary Fig. 41) showed that the EPR signal at g = 2.30 for 4MgCo/CN DSACs intensified with increasing reaction time, indicating a gradual increase in unpaired electrons. The EIS results (Supplementary Fig. 42) exhibited that the radius decreased markedly as the reaction proceeded, suggesting a remarkable enhancement in electron transfer performance. Furthermore, the increased C_dI value in the ECSA of 4MgCo/CN DSACs over reaction time (Supplementary Fig. 43, Supplementary Table 10) demonstrated a rise in active site density, which facilitated the electron participation in the reaction. As shown in Supplementary Fig. 44, the Co K-edge shifted negatively after the reaction compared to that before the reaction, indicating a reduction in the oxidation state of Co in the 4MgCo/CN DSACs after reaction. The slight elongation of Co−N bond in the FT k³-weighted Co K-edge EXAFS spectra of 4MgCo/CN DSACs was mainly attributed to the fact that Co with reduced oxidation state after reaction decreased the nuclear charge attraction to the lone‑pair electrons of N atom (Supplementary Fig. 45). Based on the LCF results, the proportion of Co²⁺ species in the 4MgCo/CN DSACs increased from 32.4% before reaction to 64.8% after reaction (Supplementary Figs. 3 and 46, Supplementary Tables 1 and 11), while the average Co oxidation state decreased from 1.923 to 1.836 (Fig. 1f and Supplementary Fig. 47).

The degradation efficiencies of CBZ by 4MgCo/CN DSACs after the addition of NaF and CAT were reduced by 5.35% and 3.8%, respectively, over 30 cycles. The Co dissolution concentration in the 4MgCo/CN DSACs/PMS/vis system with NaF and CAT exceeded 0.85 mg/L and 0.83 mg/L, which was much higher than that in the system without scavengers (0.43 mg/L) (Supplementary Fig. 48 and 49). These results indicated that OH_surf from H₂O₂ contributed to enhancing stability in the 4MgCo/CN DSACs/PMS/vis system. The morphology, physical phase, functional group, coordination number and energy band structure of 4MgCo/CN DSACs remained unchanged after reaction compared with those before reaction (Supplementary Figs. 50–55). In summary, the dynamic domino catalysis process in overcoming the activity-stability trade-off during water purification was as follows, which was schematically represented in Fig. 4i. Under visible light irradiation, photogenerated electrons drove a two-electron oxygen reduction reaction at the Mg site, enabling the in-situ generation of H₂O₂. The as-formed H₂O₂ was subsequently converted to OH_surf via the Mg site. Critically, these OH_surf species acted as electronic mediators and leveraged the π-conjugated delocalized network of carbon nitride to modulate the d-orbital configuration of non-adjacent Co atoms through a long-range electronic interaction. This interaction specifically downshifted the energy levels of the d_xy, d_x²_-y², d_xz, d_z² and d_yz orbitals^57–59, as illustrated in the corresponding orbital diagram of Fig. 4h. On the one hand, this downshifted d-band center promoted the reduction of Co³⁺ to Co²⁺, thereby accelerating •SO₄⁻ generation during PMS activation for enhanced activity. On the other hand, it reduced the adsorption energy of *SO₄, optimizing its desorption process and thereby suppressing the over-oxidation of the Co site, which effectively prevented Co atom detachment to maintain long-term stability.

Economic and environmental sustainability of dual single atom catalytic system

The CBZ degradation rate constant (k_obs) in the 4MgCo/CN DSACs/PMS/vis system in deionized water (DW) (0.187 min⁻¹) was almost identical to that in tap water (TW) (0.192 min⁻¹), which was slightly higher than that in secondary effluent (SE) (0.174 min⁻¹) (Fig. 5a, Supplementary Table 12). The k_obs was hardly inhibited in the presence of different inorganic salt ions, including Cl⁻, NO₃⁻, CO₃²⁻, HPO₄⁻, Mg²⁺ and Ca²⁺, illustrating that the OH_surf-induced dynamic-catalytic system had an excellent environmental interference resistance performance (Fig. 5b). To evaluate the practical application potential of the 4MgCo/CN DSACs/PMS/vis system, simulated CBZ wastewater and secondary effluent from the Jilin Petrochemical Company wastewater treatment plant were removed in continuous flow reactors (Fig. 5c, d). 85.9% of CBZ was removed from simulated CBZ wastewater and 95.9% of COD, 78.0% of TN, 64.9% of NH₃-N and 69.5% of TP were achieved by the 4MgCo/CN DSACs/PMS/vis system. The in-situ OH_surf-induced dynamic-catalytic system has the potential to be applied in the decontamination sterilization unit (Fig. 5e). The CBZ could be rapidly transformed into small-molecular products in the 4MgCo/CN DSACs/PMS/vis system, which was different from that in the 4Co/CN SACs/PMS/vis system (Supplementary Figs. 56–59). Compared to the toxicity of pristine CBZ, the toxicity of degradation products in the 4MgCo/CN DSACs/PMS/vis system was reduced more obviously than that in the 4Co/CN DSACs/PMS/vis system (Fig. 5f, Supplementary Tables 13 and 14) based on the QSAR-based toxicity prediction using T.E.S.T. software⁶⁰. This was supported by larger mean length of bean sprouts cultured with aqueous solutions of CBZ and its degradation products in the 4MgCo/CN DSACs/PMS/vis system than that in the 4Co/CN SACs/PMS/vis system (Fig. 5g and Supplementary Fig. 60).

Fig. 5. Economic and environmental sustainability evaluation.

Carbamazepine degradation kinetics by the 4MgCo/CN DSACs/peroxymonosulfate (PMS)/visible light (vis) system a under different water conditions (DW: deionized water, TW: tap water, SE: secondary effluent) and b in presence of different inorganic salt ions. c Carbamazepine (CBZ), chemical oxygen demand (COD), total nitrogen (TN), ammonia nitrogen (NH₃-N) and total phosphorus (TP) removal of secondary effluent by the 4MgCo/CN DSACs/peroxymonosulfate (PMS)/visible light (vis) and 4Co/CN DSACs/peroxymonosulfate (PMS)/visible light (vis) systems, respectively. Data for a–c are presented as mean values ± SD from three independent experiments (n = 3). d Schematic illustration for the possible industrial wastewater treatment with dynamic-catalytic system in peroxymonosulfate (PMS)/visible light (vis) system and e potential application of dynamic-catalytic system in wastewater treatment technology. Toxicity evaluation of degradation products by the 4MgCo/CN DSACs/peroxymonosulfate (PMS)/visible light (vis) and 4Co/CN SACs/peroxymonosulfate (PMS)/visible light (vis) systems. The notation P1–P8 and P1–P12 refer to different degradation products of carbamazepine (CBZ) from 1 to 8 and to 12 under different systems, where each number corresponds to a distinct product structure observed during the degradation process (f) and length of bean sprouts (g). Data were obtained from 4 independent bean sprouts per group (n = 4). h Comparison of electrical energy per order (EE/O) between the dynamic-catalytic system and similar reported systems. i Relative environmental impact of environmentally relevant descriptors for treating 1 kg carbamazepine in dynamic catalysis with the MgCo/CN DSACs/peroxymonosulfate (PMS)/visible light (vis), Co/CN SACs/peroxymonosulfate (PMS)/visible light (vis) and Co₃O₄/peroxymonosulfate (PMS)/visible light (vis) systems. For the calculation methods and related data of Life Cycle Assessment (LCA), see the main text and Supplementary Tables 16–18. Source data are provided as a Source Data file.

A comprehensive cost comparison of the 4MgCo/CN DSACs/PMS/vis system was performed with other typical reaction systems previously reported (Fig. 5h, Supplementary Table 15). The treatment cost of the 4MgCo/CN DSACs/PMS/vis system was only 0.132 kWh L⁻¹, which was much lower than that of the other photo-assisted systems. These comparative data highlighted the practical feasibility of the dynamic-catalytic system, especially in large-scale applications. The life cycle assessment (LCA) of the 4MgCo/CN DSACs/PMS/vis system was performed using the ReCiPe methodology⁶¹. Figure 5i exhibited that the environmental impacts for most descriptors in the preparation of MgCo/CN DSACs were much lower than those of Co/CN SACs and Co₃O₄, respectively. Given that the equipment could be operated at full capacity and resources could be maximized when industrial scale production was achieved in the future, the environmental impact of MgCo/CN DSACs would be substantially minimized during the preparation stage (Supplementary Table 16). The MgCo/CN DSACs/PMS/vis system also had a lower environmental impact for most descriptors than the other two systems in treating CBZ wastewater due to the higher intrinsic activity of MgCo/CN DSACs with lower catalyst and PMS requirements. Reducing raw material consumption also meant reducing transportation costs from the production site to the wastewater treatment plant at the practical application level, which further reduced the environmental impact (Supplementary Table 17). These detailed economic and environmental sustainability assessments verified the excellent wastewater treatment potential of the developed dynamic-catalytic system and provided strategic insights for future industrial-scale applications. The green synthesis strategy of hydrogen peroxide from water and dissolved oxygen in water by first using inexhaustible sunlight, and then using the resulting hydrogen peroxide to gradually generate OH_surf for dynamic orbital reconstruction, could break the trade-off between activity and stability of traditional Fenton-like technology for water decontamination, reduce the long-term energy consumption of high-powered equipment and minimize environmental risks. This was in line with the Global Circular Economy Initiative.

Discussion

This study overcomes the long-standing trade-off between catalytic activity and stability in advanced oxidation technology by dynamically reconstructing the d-orbital configuration of active sites in-situ. This reconstruction was facilitated by surface hydroxyl radicals generated from in-situ generated H₂O₂ using green O₂ and H₂O via a domino-catalytic strategy. Using magnesium-cobalt dual single-atom catalysts as an example, it demonstrated that cobalt d-orbital configuration was dynamically changed with the in-situ generation of surface hydroxyl radical. Different from the conventional approach that depended on rapid photoelectron injection to reduce M³⁺ to M²⁺ at the catalyst surface under light irradiation, the domino-catalytic system leveraged cobalt d-orbital reconfiguration to enable the gradual generation of surface hydroxyl radical for dual single atomic electron internal circulation. This mechanism involved the dynamic electron transfer between the d_xy and d_x²_-y² orbitals of cobalt and Op orbital of PMS, resulting in efficient and stable PMS activation. Consequently, this system achieved high removal efficiencies for chemical oxygen demand (75.9%), total phosphorus (69.5%), ammonia nitrogen (64.9%) and total nitrogen (78.0%) from petrochemical secondary effluent over 48 days. This strategy proved to be generally applicable to other AOP systems with different dual single-atom catalysts, effectively resolving the longstanding challenge of balancing catalytic activity and stability and providing a perspective for the sustainable and cost-effective development of advanced oxidation technologies in water decontamination.

Methods

Chemicals and materials

Cobalt chloride hexahydrate (CoCl₂·6H₂O, ≥98%), magnesium chloride hexahydrate (MgCl₂·6H₂O, ≥99%), urea (≥98%), n-octanol (≥99%), ethanol (≥99.8%), SMX (≥98%), CBZ ( ≥ 98%), TCH (≥98%), 4-CP (≥98%), BPA (≥99%), sodium chloride (NaCl, ≥99%), sodium bicarbonate (NaHCO₃, ≥99%), sodium nitrate (NaNO₃, ≥99%), sodium phosphate (Na₃PO₄, ≥ 98%), IPA (≥99.9%), PMSO (≥98%), β-caro (≥96%), NaF (≥99%), CAT (≥95%) and Na₂C₂O₄ (≥99%) were purchased from Aladdin.

Synthesis of catalysts

The synthesis method of carbon nitride was as follows: 20 g urea was transferred to the muffle furnace with the 5 °C/min heating rate to 550 °C for 4 h. The resulting sample was named as CN.

The synthesis method of cobalt atomic catalyst was as follows: a certain amount of CoCl₂·6H₂O and 100 mg of CN was added into 60 mL of n-octanol solution, and stirred for 1 h to obtain turbid solution. The mass ratio of CoCl₂·6H₂O to CN was 4 wt%. Then, the turbid solution was added into a teflon-lined autoclave and heated at 200 °C for 24 h. The sample was washed with ultrapure water and ethanol for several times, followed by drying in a vacuum condition at 60 °C for 10 h. The resulting samples were named as cobalt atomic catalyst (4Co/CN SACs).

The synthesis method of magnesium atomic catalyst was as follows: A certain amount of MgCl₂·6H₂O and 100 mg of CN was added into 60 mL of n-octanol solution, and stirred for 1 h to obtain turbid solution.The mass ratio of MgCl₂·6H₂O to CN was 4 wt%. Then, the turbid solution was added into a tetrachlorofluorine reactor and heated at 200 °C for 24 h. The sample was washed with ultrapure water and ethanol for several times, followed by drying in a vacuum condition at 60 °C for 10 h. The resulting samples were named as magnesium atomic catalyst (4Mg/CN SACs).

The synthesis of magnesium cobalt dual single atom catalysts was as follows: Certain amounts of CoCl₂·6H₂O, MgCl₂·6H₂O and 100 mg of CN were added into 60 mL n-octanol solution. The mass ratios of CoCl₂·6H₂O or MgCl₂·6H₂O to CN were 2, 4 and 6 wt%, respectively. The mixture was stirred for 1 h to obtain a turbid solution. Then, the turbid solution was added into a Teflon-lined autoclave and heated at 200°C for 24 h. The sample was washed with ultrapure water and ethanol several times, followed by drying under the vacuum at 60°C for 10 h. The resulting samples were named as xMgCo/CN (x = 2, 4, and 6).

Statistical analysis protocols for dual single atoms

Histogram of the observed atomic distance of 4MgCo/CN DSACs obtained by the kernel method of frequency statistic and Gauss fitting of dual single atom distances. In an attempt to increase the reliability, we collected the interatomic distance from more than 200 atoms in STEM images with both high and low magnifications, all of which were below 11.91 Å, a threshold distance for non-interacting dual single atoms. To ensure valid statistics, we restricted our atomic distance evaluation to the thin regions, obtained by ion beam thinning, where atoms could be easily counted.

Characterizations

The high-resolution spherical aberration-corrected HAADF-STEM images were recorded by JEOL JEM-ARM200F TEM/STEM with a spherical aberration corrector working at 300 kV. XRD pattern was recorded using a Rigaku D/Max-2550 diffractometer via Cu Kα radiation at 50 kV and 200 mA in the 2θ range between 10° and 70° at a scanning rate of 5°/min. The FT-IR spectra were obtained by a VERTEX 70 spectrophotometer (Bruker, Germany). The XAFS spectroscopy measurements were performed at the Co and Mg K-edges. Co and Mg K-edges analyses were performed with Si (111) crystal monochromators at the BL11B beamlines at the Shanghai Synchrotron Radiation Facility (SSRF), China. The XPS measurements were performed by a Thermo VG Scientific ESCALAB 250 spectrometer using monochromatized Al Kα excitation. The EPR measurements were performed on a Bruker 300 system. The ESR spectra of reactive oxygen species trapped in water and methanol by DMPO or TEMP were analyzed by a Brooke A300 system. The in-situ Raman spectra were recorded with a Renishaw confocal micro-Raman spectroscopy system with a 532 nm excitation laser. 2 mg catalysts were added into 20 mL PMS solution (0.1 mM) to react for different time. Then, the mixture was scanned from 850 to 1100 cm⁻¹ at a resolution of 0.5 cm⁻¹. The in-situ EPR spectra were detected a Bruker 300 system. The EPR signal of catalyst was measured by adding 2 mg catalyst into 20 mL PMS solution (0.1 M) for different time, respectively.

Decontamination performance

0.02 g catalyst was first added into 2.0 L simulated wastewater, consisted of 1.64 mM NaCl, 11.5 mM NaHCO₃, 0.7 mM NaNO₃, 0.016 mM Na₃PO₄·12H₂O and 10.0 mg/L a single contaminant with CBZ, BPA, SMX, ATZ or 4-CP and stirred in the dark for 30 min. Then, the reaction was activated by the visible light irradiation and the addition of 0.1 mM PMS. The light source was a 150 W Xenon lamp (CEL-HXF 300, Beijing Education Au-light Co., Ltd.) with a filter to cut off the ultraviolet light (λ < 420 nm). The light intensity was 122.4 mW/cm² (illumination area = 3.24 cm²). The 2.0 mL suspension was taken out and filtered through a 0.22 μm membrane at intervals of 4 min, and quenched with sodium thiosulfate. The concentration of contaminant was detected by the high performance liquid chromatography (HPLC, Supplementary Table 18). The concentration of H₂O₂ was measured by an iodometric method. All experiments were independently repeated three times to ensure the reliability of the experimental results.

Electrochemical tests

The photoelectrochemical performance was measured on a CHI 660E electrochemical workstation. It was operated in a three-electrode system, Pt wire as the counter electrode, Ag/AgCl electrode as the reference electrode and 0.1 M Na₂SO₄ solution (100 mL) as electrolyte solution.

Quantification of PMS

The 100.0 μL sample was put into 10 mL aqueous solution consisting of KHCO₃ (0.05 g) and KI (0.4 g). Then, the solution was shaken by hand and then detected by a UV-vis spectrometer at 352 nm after reacting for 15 min.

DFT computation

All spin-polarized density functional theory (DFT) calculations were performed using the Vienna Ab initio Simulation Package (VASP) version 5.4.4, employing the projector augmented wave (PAW) method. The exchange-correlation potential was described by the Perdew–Burke–Ernzerhof (PBE) functional within the generalized gradient approximation (GGA). A plane-wave cutoff energy of 450 eV was adopted, and the Brillouin zone was sampled using a 2 × 2 × 1 k-point grid. Structural relaxations were carried out until the energy convergence tolerance reached 1.0 × 10⁻⁵eV/atom and the residual forces on all atoms fell below 3.0 × 10⁻²eV/Å. A vacuum layer of 15 Å was introduced along the surface normal to prevent spurious interactions between periodic images. The DFT-D3 empirical correction was incorporated to account for van der Waals interactions. The computational parameters and k-point mesh settings employed for the structural optimization and electronic properties of all systems were exhibited in Supplementary Data 1 and Supplementary Data 2, respectively. Atomic coordinates of the optimized computational models were exhibited in Supplementary Data 3. The adsorption energies (E_ad) of PMS and H₂O₂ were calculated using the following equations:

$$E_{\mathrm{ad}}=E\left(\mathrm{slab}+\mathrm{PMS}\right)-E\left(\mathrm{slab}\right)-E\left(\mathrm{PMS}\right)$$ 1

$$E_{\mathrm{ad}}=E\left(\mathrm{slab}+{\mathrm{H}}_2{\mathrm{O}}_2\right)-E\left(\mathrm{slab}\right)-\mathrm{E}\left({\mathrm{H}}_2{\mathrm{O}}_2\right)$$ 2

where E(slab + PMS) was the total energy of the system after PMS adsorption on substrate surface, E(slab) was the energy of the substrate surface, E(PMS) was the energy of the PMS radical. E(slab + H₂O₂) was the total energy of the system after H₂O₂ adsorption on substrate surface, E(H₂O₂) was the energy of the H₂O₂ molecule.

Life cycle assessment method

Life cycle assessment (LCA) was a method for assessing the environmental impacts of all inputs and outputs throughout the life cycle of a product or service system (cited from ISO 14040, 14044). In this study, Simapro 9.5 was employed for modeling with data based on the Ecoinvent 3.9 database, and environmental impact quantification was performed using ReCiPe 2016 Midpoint (H) V1.08/World (2010) H. Taking the removal of 1 g carbamazepine from wastewater as the functional unit, the life-cycle environmental impacts of 4Co/CN SACs, Co₃O₄, and 4MgCo/CN DSACs catalysts were compared in wastewater treatment applications. The system boundaries encompassed both catalyst preparation and its subsequent application (Supplementary Fig. 61).

Catalyst production covered the production-related inputs, including chemical reagents, solvents, energy consumption, and auxiliary materials. Catalytic application included the operational inputs associated with PMS activation and carbamazepine degradation. Life cycle inventory (LCI) data for the processes by 4MgCo/CN, 4Co/CN and Co₃O₄ were obtained from experimental data, including catalyst synthesis, reagent dosages and reaction conditions. A series of midpoint indicators were selected to characterize the potential environmental burdens, including global warming, stratospheric ozone depletion, ionizing radiation, ozone formation for human health, fine particulate matter formation, ozone formation for terrestrial ecosystems, terrestrial acidification, freshwater eutrophication, marine eutrophication, terrestrial ecotoxicity, freshwater ecotoxicity, marine ecotoxicity, human carcinogenic toxicity, human non-carcinogenic toxicity, land use, mineral resource scarcity, fossil resource scarcity, and water consumption. In the LCA calculation, intermediate product flows (e.g., electricity, nitrogen) were commonly recorded as inventory inputs or outputs within unit processes. The software tracked the complete upstream supply chain of each intermediate product and decomposed all linked processes into elementary flows (e.g., CO₂, CH₄). Eighteen midpoint impact categories in ReCiPe 2016 were quantified using a standard characterization procedure. For each category, the total impact score was obtained by multiplying the inventory flow of each elementary substance (emission or resource extraction) by its specific characterization factor (CF) and summing over all contributing substances. The general formula was as follows:

$$I_{\mathrm{m}}=\sum \left({\mathrm{EF}}_{\mathrm{i}}\times {\mathrm{CF}}_{\mathrm{i},\mathrm{m}}\right)$$ 3

where Iₘ was the characterization result for midpoint category *m*, EFᵢ was the amount of elementary flow *i* from the life cycle inventory, and CFᵢ,ₘ was the characterization factor that expresses the contribution of flow *i* to midpoint category *m* (in equivalent units). Normalization was performed independently for each impact category. For each indicator, the maximum environment impact value among the three gtoups was set as 100, and the remaining values were proportionally scaled accordingly. The detailed description and calculation formulas for LCA were exhibited in Supplementary Tables 16–18. The resulting 18 environmental impacts were compared among the scenarios mediated by 4MgCo/CN, 4Co/CN and Co₃O₄ to evaluate the environmental feasibility of the system by 4MgCo/CN and to identify the major contributors to its environmental burden.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

J.J. Jiang, C. He, and S.S. Dong conceived and designed the experiments. J.J. Jiang, C. He, T.Z. Sun, B.W. Zhao, R.X. Wang, and Y.N. Zhang carried out the experiments. S.D. Liu, Z.H. Zhao, and X.S. Ye performed the theoretical calculations. J.J. Jiang, C. He, S.D. Liu, T.Z. Sun, B.W. Zhao, R.X. Wang, Y.N. Zhang, Z.H. Zhao, X.S. Ye, H. Yu, Y.F. Fu and Y.N. Liu assisted in analyzing various characterizations. J.J. Jiang and C. He wrote the paper. D. Wu, M.X. Huo, D.D. Zhou, and S.S. Dong further revised the paper. All authors contributed to the discussion of the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous, reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was financially supported by the National Natural Science Foundation of China (52570079, 52300088, W2521110, 52670095, 2023YFC3207704), Science and Technology Development Program of Jilin Province, China (20260602011RC), Fundamental Research Funds for the Central Universities (45124031D066).

Data availability

The data that support the findings of this study are openly available in Figshare at https://doi.org/10.6084/m9.figshare.33315537⁶². Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-78263-9.