Overcoming catalyst deactivation with in-situ electrochemical regeneration in packed-bed reactors enabling long-lived wastewater purification

Abstract

Heterogeneous advanced oxidation processes (HG-AOPs) can efficiently remove persistent pollutants but are constrained by catalyst deactivation from polymeric deposits and by energy- and chemical-intensive regeneration. We report a conductive packed-bed reactor that couples continuous treatment with in-situ electrochemical regeneration, maintaining high pollutant removal across diverse contaminants and complex matrices. Upon loss of activity due to catalyst deactivation, applying a mild current with low-cost H₂O₂ restores performance without reactor downtime. Mechanistic analyses show that electrochemical modulation lowers surface lipophobicity and weakens the interaction between polymeric products and the catalyst, enabling their desorption. Meanwhile, electrogenerated ^•OH promotes the coupling of residual phenolic polymers into insoluble aggregates and mineralizes quinoid compounds, recovering interfacial mass and electron transfer. Industrial coking wastewater can be continuously treated using a series-parallel reactor system, which alternates between oxidation and regeneration phases. This approach maintains over 80% total organic carbon removal efficiency for more than 300 hours of operation. Compared with conventional advanced oxidation processes, this strategy reduces operating costs by ~68%, minimizes external chemical inputs, and avoids high-temperature regeneration. The reactor architecture and regeneration logic are scalable, compatible with distributed deployments, and readily retrofittable to existing units, providing a sustainable pathway to extend catalyst lifetimes, lower resource intensity, and advance next-generation oxidation technologies.

A conductive packed-bed reactor integrates continuous catalytic oxidation with in-situ electrochemical regeneration. The system uses electrically switchable wettability to reduce foulant adhesion and enhance the conversion and separation of •OH-mediated polymeric products.

Introduction

The global water crisis is worsening under mounting pressures from population growth, accelerated industrialization, and intensifying climate change, demonstrating the urgent need for innovative and sustainable water purification technologies^1,2. Emerging recalcitrant contaminants (ECs) persist through conventional treatment and threaten ecosystems and human health, drawing attention to advanced oxidation processes (AOPs) as a central abatement strategy^1,3–9. Within AOPs, heterogeneous systems (HG-AOPs) are attractive alternatives to homogeneous Fenton because they broaden the pH window, avoid iron-sludge generation, and circumvent slow Fe(III) reduction^7,10–15. Developing stable, efficient, and heterogeneous catalysts is therefore pivotal to realizing HG-AOPs for large-scale remediation. Together, these considerations focus the agenda on durability under realistic operating conditions as the gatekeeper to impact.

However, the large-scale application of HG-AOPs is hindered by catalyst deactivation^16–18. Although the initial catalytic activity is often high, performance typically declines sharply over repeated cycles^19–22. Prior work shows that interfacial polymerization of target pollutants increases the product degree of polymerization, strengthening van der Waals and hydrophobic interactions with catalyst surfaces^17,23–25. The resulting adherent, high-molecular-weight deposits form a fouling layer that (i) blocks active sites, (ii) hinders mass transfer of pollutants and oxidants, and (iii) disrupts interfacial electron transfer required for redox activation²³. Carbon-based catalysts (e.g., carbon nanotubes, graphene) and several transition-metal oxides are particularly vulnerable to this self-inhibition, necessitating frequent replacement and offline servicing that raise costs and compromise reliability. Consequently, the design problem shifts from maximizing initial activity to ensuring regenerability, restoring function efficiently and repeatedly without replacement or downtime.

Addressing deactivation thus requires effective regeneration strategies^18,23. Conventional methods (annealing, organic solvent washing) can partially restore activity^26,27, but are inherently ex situ and often destructive²³. Reactors must be halted, spent media removed, and harsh treatments applied, leading to downtime, high energy demand (especially for calcination), solvent footprints, and incomplete recovery or structural damage^23,28,29. These limitations highlight the need for in situ, continuous, efficient, and environmentally benign regeneration. Electrochemical regeneration is compelling because it enables tunable control via electrical stimuli, mild operating conditions, low energy use, and seamless integration for online operation^30–34. Yet the approach remains underexplored: a key gap is the absence of reactor designs that couple continuous catalytic oxidation with localized, scalable electrochemical regeneration in practical flow systems^16,35. Filling this gap calls for a conductive, modular architecture that embeds regeneration as a low-energy control within continuous operation.

Here, we present a conductive packed-bed reactor that integrates continuous catalytic oxidation with in situ electrochemical regeneration to overcome polymerization-induced deactivation. The reactor employs a simple N-doped carbon nanotube (NCNT) as both a structured catalyst support and functional electrode. Under mild current/voltage with controlled H₂O₂ dosing, deactivated catalysts are easily regenerated on demand within the bed. Regeneration proceeds through three synergistic processes: (i) electrochemical modulation of NCNT wettability reduces surface lipophobicity and weakens polymer-surface interactions, facilitating desorption; (ii) electro-activation of H₂O₂ accelerates mineralization of polymeric products (e.g., quinones); and (iii) localized ^•OH generation within the fouling layer further oxidizes desorbed phenolic polymers into insoluble ultra-high-molecular-weight aggregates, reopening mass and electron transfer pathways. This integrated design achieves in situ regeneration without complex operations, deep oxidation plus physical removal of fouling polymers, and uninterrupted continuous-flow operation via flexible series or parallel configurations. Our results demonstrate that an electrochemically regenerable conductive packed-bed system effectively addresses HG-AOP deactivation, establishing a sustainable pathway for next-generation advanced water treatment.

Results

Design and performance of a catalytic packed-bed reactor

Most reported heterogeneous catalysts have been evaluated in batch Fenton-like reactors (Fig. 1a), where mechanical stirring ensures adequate mass transfer^27,36,37. After the reaction, the deactivated catalysts were separated from the effluent by gravity sedimentation³⁸. These spent catalysts were often either discarded as hazardous waste or regenerated via chemical cleaning or thermal annealing. Such regeneration processes usually required transferring the catalyst to dedicated regeneration equipment (e.g., a muffle furnace)²⁷ and then returning it into the reactor, making conventional thermal/solvent routes time-consuming and operationally cumbersome while increasing labor and capital costs^20,23,39. To overcome these drawbacks, this study developed an integrated packed-tower reactor that coupled continuous catalytic oxidation with in situ electrochemical regeneration (Fig. 1b). Organic wastewater was premixed with oxidants and pumped upward through the packed column to purify the influent. Nitrogen-doped carbon nanotubes (NCNT) were selected as the packing because of their proven persulfate-activation activity, high electrical conductivity, and mechanical robustness³². These properties allowed NCNT to serve simultaneously as the catalytic scaffold and as an electrode, providing a solid foundation for investigating reaction and regeneration mechanisms. Material characterizations are provided in Supplementary Text 4.

Fig. 1. Design and performance of a catalytic packed-bed reactor.

a Schematic diagram of a conventional heterogeneous Fenton-like reaction system. b Schematic diagram of a catalytic packed-bed reactor. From left to right, the packed-bed reactor model illustrates the general mode of the in‑situ catalytic reaction; the purple‑to‑green gradient arrows indicate the progressive deactivation of the catalyst layer from bottom to top, and the schematic diagram on the far right shows the deposition of oxidation products of various molecular weights onto the catalyst surface. c Phenol removal efficiency in the packed-bed reactor at varying oxidant concentrations. d Phenol removal efficiency in the packed-bed reactor under different hydraulic retention time (HRT). e Removal efficiency of selected phenolic and sulfonamide micropollutants by the conductive packed tower within a 5-h reaction window (BPA bisphenol A, 2-CP 2-chlorophenol, 2,4-DMP 2,4-dimethylphenol, PCMX p-Chloro-m-xylenol, SMX sulfamethoxazole, SPY sulfapyridine, SAT sulfathiazole, SMM sulfamonomethoxine). f Phenol removal efficiency of the packed-bed reactor over two consecutive operational cycles. Data for (c, d–f) are presented as mean values ± SD (n = 3). Unless otherwise specified, the reaction conditions in the conductive packed tower are as follows: [Oxidant]₀ = 500 μM; [OMP]₀ = 100 μM, [HRT] = 1.44 min, T = 298.15 K. Source data for this figure are provided as a Source data file.

The key operational parameters of the catalytic packed-bed tower were first optimized for wastewater treatment. Pollutant adsorption (Fig. 1c) declined to ~20% after 200 min in the absence of oxidant peroxymonosulfate (PMS), as the adsorption sites became progressively saturated. With 300 μM oxidant, the removal efficiency reached an initial efficiency of 96% via catalytic oxidation but decreased to ~73% after 200 min due to active-site occupation and oxidant insufficiency. Increasing the oxidant dose to 500 μM enhanced the apparent reaction kinetic constant (k_obs) (Supplementary Fig. 4) and improved pollutant removal efficiency within the packed bed (Fig. 1c). Nevertheless, the enhancement of pollutant removal efficacy by oxidant concentration had an upper limit. Further increase in oxidant dose beyond a certain threshold did not lead to additional improvement in the apparent reaction rate (Supplementary Fig. 5). Hydraulic retention time (HRT) also governed performance. By adjusting the influent flow rate (Fig. 1d), the HRT could be altered to modulate the extent of reaction. Reducing the influent flow rate from 1.0 to 0.5 mL min⁻¹ (HRT extended from 1.01 to 2.01 min) increased the pollutant removal efficiency of continuous 5 h operation from 30% to 100%. Therefore, tuning the oxidant dosage and HRT could significantly enhance the apparent removal efficiency of organic pollutants, enabling the efficient removal of phenolic and sulfonamide contaminants (Fig. 1e).

Different from the typical steep saturation curves observed in previously studied electrofiltration membrane systems⁴⁰, an axial concentration gradient developed along the catalyst bed, with pollutant concentration gradually decreasing from the inlet to the outlet (Fig. 1b). The lower section of the bed was exposed to the highest pollutant concentrations, leading to the fastest accumulation of polymeric products (as explained in the subsequent reaction mechanism). Consequently, this region became fouled first, and catalyst deactivation progressively advanced from the bottom (inlet) toward the top (outlet) over time, until the entire bed required regeneration. During extended continuous operation, the removal efficiency in a subsequent cycle fell to ~21.5% (Fig. 1f), reflecting catalyst deactivation and decreased oxidant utilization over prolonged operation (Supplementary Fig. 6) and indicating stability as a key challenge.

Deactivation pathways of packing materials

The catalytic reaction and deactivation mechanisms were first probed for the subsequently developing targeted regenerative technologies. Neither ethanol nor tert-butanol inhibited the reaction kinetics (Supplementary Fig. 7), consistent with previous reports that NCNTs predominantly activated oxidants via an electron transfer pathway (ETP)^41,42. To verify ETP, a divided electrochemical oxidation system (Supplementary Fig. 8) was constructed. The oxidant and pollutant were placed in the cathode and anode chambers, respectively, with an NCNT-modified electrode and a salt bridge ensuring transfer of only electrons and an ionic conductor. Introducing oxidant to the cathode chamber induced a sharp positive shift in cathodic potential (Fig. 2a), suggesting oxidant was adsorbed on the cathode surface. Subsequent addition of pollutant into the anode chamber caused a pronounced potential decrease and the emergence of an oxidative current. This phenomenon arose from electron loss by the pollutant in the anode chamber, with the electrons transferred through the external circuit to the cathode, where they drove reductive decomposition of the adsorbed oxidant²⁵. These results established electron transfer as the dominant pathway in pollutant removal. To elucidate the removal mechanism, we conducted an electron stoichiometry analysis by measuring total organic carbon (TOC) removal at varying oxidant concentrations. After correcting for adsorption, the theoretical minimum oxidant demand for complete mineralization (q_theo) of phenol was calculated and compared with the actual consumption (q_real) (Fig. 2b and Supplementary Table 3). Notably, q_real was substantially lower than q_theo, indicating that pollutant removal proceeds mainly via polymerization rather than full mineralization^21,35,43,44.

Fig. 2. Deactivation pathways of packing materials.

a Open-circuit potential (OCP) curves in a divided cell system. Insets show atomic models with added oxidizers and pollutants. b Theoretical versus actual peroxymonosulfate (PMS) consumption for complete phenol removal. c Differential pulse voltammetry (DPV) of the phenol (PE) solution and NCNT_used. d Total-ion-current (TIC) chromatograms of reaction solution and surface-extracted product (FTMS: Fourier transform mass spectrometry). e Possible reaction pathways of PE (Phenol undergoes dehydrogenation at multiple sites to form radical intermediates, which then polymerize through C-O, ortho C-O, and para C-C coupling, eventually forming polymers of varying chain lengths). f Reaction energy barriers for the proton-coupled electron transfer (PCET) driven oxidative coupling of phenol. Insets show atomic models of the reactants at each step. g Van der Waals (VDM) interaction between the polymer and the carbon surface. The green areas represent VDM forces, and the blue areas represent chemical bonding forces. The rectangles at the bottom represent the catalyst surface, with monomers and polymers of phenol on top. As the degree of polymerization of phenol increases, its VDM forces with the catalyst surface clearly increase. h Relationship between the degree of polymerization and the adsorption energy (Fit: y = −0.1258x − 0.4592, n = 6, R² = 0.8732, p-value = 0.004, Shaded area: 95% confidence interval). i Relationship between the degree of polymerization and XlogP3 (Fit: y = 1.5757x + 0.0767, n = 6, R² = 0.9986, p-value < 0.0001, Shaded area: 95% confidence interval). Source data for this figure are provided as a Source data file.

To elucidate pollutant transformation pathways, differential pulse voltammetry (DPV) was first applied to direct oxidation of phenol in solution (Supplementary Fig. 9). The first scan displayed a distinct oxidation peak (Peak 1, Fig. 2c), corresponding to hydrogen abstraction from phenol and subsequent radical formation (Supplementary Fig. 10)⁴⁵. In the second scan, two additional peaks (Peaks 2 and 3, Fig. 2c) emerged, attributable to oxidation of hydroxylated phenols (hydroquinone and catechol) into quinones (Supplementary Fig. 10)⁴⁵. Similar oxidation peaks were also detected on the used packing material (NCNT_used) modified electrode (Supplementary Fig. 11 and Fig. 2c), confirming that oxidation reactions occurred at the catalyst surface. Shifts in peak positions further suggested structural rearrangement or polymerization of deposited products²³. To clarify the product structures and deactivation mechanism, we compared extracts from the deactivated packing with residual products in solution. Total ion current (TIC) chromatograms (Fig. 2d and Supplementary Fig. 12) revealed markedly stronger signals from the packing material extracts across both 50–750 and 750–2000 m/z ranges. The strong response in the 750–2000 m/z range obviously demonstrated extensive polymerization of PE on the NCNT surface, yielding high-molecular-weight products. Further analysis and structural matching (Supplementary Table 4 and Supplementary Fig. 13) suggested that phenol underwent dehydrogenation at multiple sites, generating radical intermediates that polymerize via para C-O, ortho C–O, and para C–C coupling (Fig. 2e), ultimately forming phenolic polymers of varying chain lengths⁴⁶. Moreover, given that the oxidation potential of the NCNT–PMS^* complex exceeded that of phenolic compound oxidation (as indicated by the peaks 2 and 3 in Fig. 2c), some polymers could undergo further oxidation to quinone structures (Supplementary Table 5 and Supplementary Figs. 14–15). These complex polymeric products strongly adhered to the catalyst surface, forming a fouling layer that deactivated the catalyst.

The formation of organic radicals was a key step in generating polymeric products^46,47. Results from the divided electrochemical oxidation system also suggested that pollutant removal proceeded predominantly via a proton-coupled electron transfer (PCET) pathway. Phenol molecules usually formed organic radicals through hydrogen atom abstraction (HAT) or PCET (Supplementary Fig. 16). The HAT pathway required direct collisions between reactive species and phenol, which was precluded in the divided setup where only electron and ion transport were permitted (Supplementary Fig. 17). However, the aqueous environment provided abundant protons, enabling phenol to lose an electron and exchange protons with the water environment, generating radicals via PCET. Density functional theory (DFT) calculations supported a PCET-driven coupling mechanism (Fig. 2f) (Atomic coordinates for the structural model are available in Supplementary Data 1). The results indicated that phenol and PMS could co-adsorb on the carbon surface, where phenol underwent PCET to lose a proton and an electron, yielding an organic radical. PMS accepted the electron and proton, decomposing reductively to H₂O and SO₄²⁻. A second phenol followed the same sequence, and the resulting radicals coupled to form a dimer. The thermodynamic rate-determining step was the PCET-mediated radical formation, and higher-order polymers arose from successive radical coupling.

As the degree of polymerization (DP) increased, both van der Waals (vdW) volume and hydrophobicity rose^17,23, thus strengthening their interactions with the solid surface. The density overlap region indicator (DORI) was employed to visualize the VDW interaction between the polymers and the carbon surface (Atomic coordinates for the structural model are available in Supplementary Data 1). For phenol monomers, VDW interactions were colored in green regions (Fig. 2g). As polymerization progressed, the enlarged van der Waals volume increased the interaction regions with the carbon surface (Supplementary Fig. 18 and Fig. 2g, hexamer), resulting in observably higher adsorption energies. DP and adsorption energy revealed a strong positive linear correlation (R² = 0.8732, Fig. 2h). Hydrophobicity similarly increased linearly, further reinforcing adhesion (Fig. 2i). Collectively, amplified van der Waals and hydrophobic interactions drove the accumulation of highly polymerized products on the catalyst packing, impeding mass transfer and interfacial electron transfer, and ultimately causing catalytic deactivation.

In-situ electrochemical regeneration in a packed-bed tower

To mitigate performance loss from catalyst deactivation, we integrated electrochemical regeneration into the catalytic packed-bed, enabling in situ regeneration and uninterrupted operation. A schematic of the regeneration device is shown in Fig. 3a. The catalytic packing was functionalized as the cathode, with a counter electrode installed at the reactor top and voltage supplied externally. During regeneration, an appropriate amount of electrolyte and H₂O₂ solution was added through a unified interface (the electrolyte should submerge the packing materials without continuous pumping). After regeneration, flow switching restored the treatment mode. This integrated design coupled regeneration and catalysis seamlessly, enhancing modularity and operational convenience.

Fig. 3. In-situ electrochemical regeneration in packed-bed tower.

a Schematic of a sustainable catalytic packed-bed reactor enabling in-situ electrochemical regeneration. b Regeneration efficiency of the catalyst under different applied currents. c Simulated electric field distribution within the catalyst bed of the regeneration unit. See the Methods section for specific reactor dimensions. d Regeneration efficiency of the catalyst at varying hydrogen peroxide concentrations. e Regeneration efficiency of the catalyst as a function of regeneration time. f Theoretical total organic compound (TOC) corresponding to the removed phenol versus the residual TOC measured in the electrolyte. g Cyclic experiments for phenol removal. h Differential pulse voltammetry (DPV) of NCNT_x. i Electrochemical impedance spectroscopy (EIS) of NCNT_x. Insets show a zoomed-in view and the equivalent circuit diagram employed for the impedance fitting. j Open-circuit potential (OCP) curves of NCNT_x. Data for (b, d-g) are presented as mean values ± SD (n = 3). Unless otherwise specified, the reaction conditions in the conductive packed tower are as follows: [Oxidant]₀ = 500 μM; [OMP]₀ = 100 μM, [HRT] = 1.44 min, T = 298.15 K. Source data for this figure are provided as a Source data file.

The regeneration efficiency was systematically assessed as a function of applied current (Fig. 3b and Supplementary Fig. 19). At an applied current of 0.3 mA, the regenerated packing achieved partial recovery, with a catalyst regeneration efficiency (R_5h) of 54.72%. Increasing the applied current to 0.5 and 1.0 mA further enhanced R_5h to 67.62% and 72.42%, respectively. However, a further increase in the applied current to 1.5 mA led to a significant reduction in R_5h to 21.49%, which could be attributed to the quenching of ^•OH by an excessive electron supply (as discussed in detail in the mechanism section). These results highlight the necessity of moderate current conditions. Since the conductive packing collectively functions as the cathode, the electric field distribution within the regeneration device was simulated (cross-sectional and top views shown in Fig. 3c). The working voltage between the anode and cathode was approximately 8.5 V under constant current of 1 mA (Supplementary Fig. 20), but the actual potential of the functionalized cathode was maintained at about 1 V due to the voltage loss induced by resistance of the electrodes and packing materials, enabling H₂O₂ activation under mild potential conditions. Further increasing the current raised the cathode surface potential beyond the suitable window for H₂O₂ activation.

Varying the H₂O₂ concentration from 0 to 30 mM progressively improved regeneration (with R_5h increasing from 51.32 to 72.42%), whereas an excessive dose (50 mM) reduced the efficiency, likely due to the self-quenching of ^•OH by H₂O₂ (Fig. 3d and Supplementary Fig. 21)^48,49. Extending the regeneration time also enhanced regeneration performance (Fig. 3e and Supplementary Fig. 22). Under the conditions of a constant current of 1 mA and 30 mM H₂O₂, R_5h was only 29.23% after 15 min of regeneration time. Prolonging the regeneration to 30 min increased R_5h to 72.42%, and extending it to 45 min achieved an R_5h of 100%. A further extension to 60 min maintained R_5h at 100%. Thus, the regeneration time was critical and should be optimized to balance efficacy and cost.

Simultaneously, the TOC residual in the electrolyte solution was measured for all regeneration experiments (Supplementary Table 6 and Fig. 3f). The regeneration efficiency showed a positive correlation with the residual TOC in the solution: better regeneration results corresponded to higher residual TOC. The highest TOC value was observed under 1 mA current, 30 mM H₂O₂, and 60 min of regeneration. Further extending the regeneration time reduced the TOC level, while R_5h remained steady at 100% (Supplementary Figs. 23–25). This decrease can likely be attributed to the further mineralization of the desorbed and transformed polymeric products remaining in the regeneration solution. In all cases, the measured TOC was far below the theoretical TOC equivalent of phenol removed by the packed tower (Fig. 3f), indicating partial mineralization as well as partial desorption and solubilization of polymeric species. After optimizing the operational parameters, continuous cycling regeneration experiments were performed under the selected optimal conditions (Fig. 3g and Supplementary Fig. 26). After five in-situ removal-regeneration cycles, the catalytic packed column still maintained a high phenol removal efficiency (R_5h > 95%) and the corresponding TOC removal (>94%), demonstrating good operational stability and regeneration reliability of the coupled system. Multiple characterization techniques were further employed to evaluate the effectiveness of electrochemical regeneration in restoring the catalyst surface sites, mass transfer, and electron transfer properties. DPV tests showed that the oxidative product signals on the surface of the regenerated packing (NCNT_regen) were almost diminished and closely resembled the signal characteristics of the pristine packing (NCNT_pris), indicating effective desorption of deposited polymeric products (Fig. 3h). Electrochemical impedance spectra (EIS) showed that the deactivated catalyst (NCNT_used) exhibited an enlarged Nyquist arc radius with increased impedance values, suggesting an increase in interfacial charge-transfer resistance. After regeneration, the arc radius decreased markedly to a level comparable to that of NCNT_pris, accompanied by a reduction in impedance, confirming that electron transfer capacity was effectively restored by the regeneration process (Fig. 3i and Supplementary Fig. 27). Chronopotentiometry (CP) further evidenced the cleaning of the reaction interface by regeneration (Fig. 3j). For the deactivated catalyst, PMS addition led to a delayed potential stabilization and a reduced steady-state potential of 0.78 V (vs. Ag/AgCl). After regeneration, both parameters reverted to their initial values, indicating re-exposure of active sites and markedly improved interfacial reaction kinetics.

Additionally, Fourier-transform infrared (FT-IR), X-ray photoelectron spectroscopy (XPS), and scanning electron microscope (SEM) collectively confirmed the removal of oxygenated surface species from the catalyst surface (Supplementary Figs. 28–30). To further visualize the regeneration efficacy at the nanoscale, transmission electron microscope (TEM) analysis was conducted on the fresh, spent, and regenerated NCNTs (Supplementary Fig. 31). The results revealed that the spent catalyst exhibited a distinct overlayer with an enhanced oxygen signal, corresponding to the deposition of polymeric products. In contrast, the regenerated catalyst displayed a clean surface comparable to that of the fresh catalyst. Brunauer-Emmett-Teller surface area analysis (BET) (Supplementary Fig. 32) showed that the regenerated catalyst possessed a higher specific surface area and pore volume than the spent catalyst, further confirming that the catalyst surface was effectively cleaned and the porous structure was restored. In-situ FTIR spectroscopy was conducted throughout the electrochemical regeneration process to directly observe the dynamic evolution of polymeric species on the catalyst surface during regeneration (Supplementary Fig. 33). The spectral evolution reveals a progressive attenuation of the characteristic bands associated with surface-adsorbed organic polymers, indicating their gradual removal during regeneration. These observations provided direct evidence for the dynamic cleaning process occurring on the deactivated catalyst surface.

Interfacial processes in electrochemical regeneration

Although electrochemical regeneration achieved excellent performance, its underlying mechanism required further clarification. We thus examined both interfacial interactions and interfacial reactions. Polymer adhesion to the carbon surface was governed primarily by van der Waals (vdW) forces and hydrophobic interactions, while electrochemical bias could tune surface wettability and thereby modulate organic-surface interactions^50,51. We first measured the contact angle between the material and toluene under various conditions. The pristine NCNT exhibited a contact angle of 96.5° with toluene (Fig. 4a). Polymer deposition on the carbon surface reduced the contact angle (Fig. 4b) because adhered polymeric products were lipophilic. However, applying a negative potential to NCNT markedly increased its contact angle with toluene (Fig. 4c), suggesting that negative polarization weakened the interaction between polymeric products and the catalyst surface, thereby promoting their desorption. This finding verified that electrochemical treatment could effectively tune the surface properties and organic interactions of NCNT.

Fig. 4. Interfacial processes in electrochemical regeneration.

a Lipophilic contact angle of pristine NCNT_pristine. b Lipophilic contact angle of pristine NCNT_used. c Lipophilic contact angle of NCNT after application of a cathodic potential. d Probe experiments of nitrobenzene (NB) and benzoic acid (BA) ([NB]₀ = [BA]₀ = 200 mg L⁻¹, [Na₂SO₄] = 100 mM, [Current] = 1 mA, [H₂O₂] = 30 mM, [Time] = 60 min). e Electron spin resonance (ESR) spectra recorded during electrochemical regeneration. f Reaction energy barrier for H₂O₂ activation to ^•OH. Insets show atomic models of the reactants at each step. g Total-ion-current (TIC) chromatograms of surface-extracted products and residual species in the spent regenerant (FTMS: Fourier transform mass spectrometry). h Size exclusion chromatography of surface-extracted products and residual species in the regenerated liquid. i Schematic diagram illustrating the interplay among hydrophobicity, solubility, and surface adsorption capacity. j Schematic illustration of the deactivation-regeneration cycle of the catalytic packing. Source data for this figure are provided as a Source data file.

Applying voltage alone (without H₂O₂) partially promoted polymer desorption and dissolution (Fig. 3d), resulting in limited recovery of catalytic activity. The addition of H₂O₂ markedly enhanced regeneration, underscoring the critical role of electrochemical activation of H₂O₂. Although NCNT exhibited negligible intrinsic activity toward H₂O₂ activation and pollutant removal (Supplementary Fig. 34), the system efficiently removed an ^•OH radical probe after applying voltage (Fig. 4d), demonstrating ^•OH generation. Quenching experiments on the regeneration solution with ethanol and tert-butanol exhibited obvious inhibition on activity recovery (Supplementary Fig. 35). Electron spin resonance (ESR) spectroscopy further showed a time-dependent increase in the DMPO-^•OH signal (Fig. 4e), supporting continuous ^•OH production in electrochemical regeneration. DFT calculations (Fig. 4f and Supplementary Fig. 36) revealed that the energy barrier for H₂O₂ activation by pristine NCNT under open-circuit conditions was associated with the single-electron transfer step (2.36 eV) for ^•OH formation (Atomic coordinates for the structural model are available in Supplementary Data 1). When an external potential of approximately 1 V was applied, the energy barrier decreased to 1.35 eV, demonstrating that the electric field markedly promoted H₂O₂ activation. Thus, the applied potential effectively lowered the reaction energy barrier and accelerated ^•OH-mediated polymer transformation and removal.

Further investigation was conducted to elucidate the fate and transformation pathways of the desorbed polymers. According to prior TOC analysis (Fig. 3f), some of the polymers entered the regeneration electrolyte after desorption. However, the TOC in the regeneration electrolyte was lower than the theoretically calculated value of removed phenol, suggesting that only a fraction remained in the liquid phase. Meanwhile, DPV results showed a substantial reduction in polymer-related signals on the regenerated catalyst surface. These findings together indicated that another portion of the products had been mineralized. The TIC of residual organic compounds in the regeneration electrolyte revealed shorter retention times as compared with extracts from NCNT_used (Fig. 4g and Supplementary Fig. 37), reflecting weakened interactions between these substances and the chromatographic stationary phase. This suggested that the regenerated polymers exhibited a reduced tendency to adhere to solid surfaces. Furthermore, the signal intensity of organic compounds in the regeneration solution decreased markedly, consistent with the TOC trend. To more accurately characterize changes in the molecular weight and quantity of the polymers, size exclusion chromatography was employed to compare extracts from NCNT_used (SE-OM) and residual organic compounds in the regeneration electrolyte (SR-OM) (Fig. 4h and Supplementary Fig. 38). The results showed that SR-OM eluted earlier than SE-OM but with reduced peak intensity, suggesting a higher molecular weight. This implied that some desorbed polymers underwent further polymerization in the liquid phase, forming higher-molecular-weight products. This suggests that some desorbed polymers undergo further coupling in solution to form higher-molecular-weight species that remain suspended without re-adsorption. Previous results (Fig. 2g–i) and literature suggested that increasing polymer polymerization typically enhances hydrophobicity and van der Waals volume, promoting surface adsorption^17,23. However, this interpretation overlooked the fact that excessive hydrophobicity leads to a sharp decline in solubility and dissolution kinetics (Fig. 4i). As solubility approaches zero, the dissolution–adsorption dynamics become limited by the slow dissolution of suspended polymer species, rendering the process kinetically hindered within the experimental timeframe. This leads to an apparent non-adsorption behavior. In sum, the electrochemical regeneration process not only altered the surface interaction properties of the polymers but also retards adsorption-dissolution kinetics by increasing their molecular weight and reducing solubility. In short, electrochemical regeneration both weakens polymer-surface interactions (via wettability tuning) and retards re-adsorption (by increasing molecular weight and lowering solubility), enabling effective separation of polymers from the packing and restoring catalytic activity within the regeneration window.

Based on electron paramagnetic resonance (EPR) spectroscopy, regeneration experiments, and probe tests, ^•OH were identified as key species responsible for the chain growth and mineralization of desorbed polymeric products. While applied voltage alone modulated surface wettability and induced partial polymer desorption, the introduction of H₂O₂ enabled electrochemical generation of ^•OH, which subsequently participated in two primary pathways: i) mediating ring-opening reactions that mineralized quinone compounds⁵², and ii) abstracting hydrogen atoms from phenolic species to form organic radicals, leading to coupling and the formation of higher-molecular-weight products²⁹. Consequently, only a fraction of the TOC equivalent of removed phenolics was detected in the regeneration solution.

The deactivation and regeneration mechanisms of the catalytic packing could be summarized as follows (Fig. 4j): During the catalytic operation stage, organic radicals were continuously generated via a PCET process and continually coupled to form polymers. As the DP increased, the enhanced hydrophobicity and van der Waals interactions between the polymeric products and the catalyst surface promoted polymer deposition, which blocked active sites, hindered mass and electron transfer, and ultimately led to catalyst deactivation. During the in situ electrochemical regeneration process (Fig. 4j), the applied voltage modulated the surface wettability of the material, enhancing its lipophobic nature, thereby weakening the interaction between organic substances and the catalyst surface. The desorbed polymers underwent either mineralization or further coupling into higher-molecular-weight products via the ^•OH-mediated pathway. Simultaneously, active sites of the catalyst were re-exposed, and interfacial transport was restored, recovering the packed-bed catalytic performance.

Application prospects of an integrated electrochemically regenerative reactor

The proposed in situ electrochemical regeneration could be directly integrated into existing packed-bed reactors, demonstrating excellent transferability and adaptability. Within the wastewater treatment process, this technology could be applied either for direct treatment of recalcitrant organic wastewater or as an advanced purification step following biological treatment units (Fig. 5a). The electrocatalytic packed-bed towers might be flexibly arranged in series or parallel to accommodate varying water quality demands and continuous operation requirements. Precise control of wastewater and regeneration solution flows could be achieved by adjusting the power supply and flow valve settings. Using coking wastewater as an example, we configured two serially connected catalytic packed-tower groups (Fig. 5b). Operating three catalytic packed towers in series enhanced the treatment capacity for high-concentration wastewater, while a parallel configuration of two series groups enabled simultaneous processing and regeneration operations. As illustrated in Fig. 5b and Supplementary Figs. 49–51, when Group 1 was in treatment mode, Group 2 underwent regeneration. When the treatment efficiency of Group 1 declined, the system switched to Group 2 for treatment, and Group 1 entered the regeneration phase. This alternating scheme secured uninterrupted operation. Flow and current switching for the in situ regeneration were implemented through automated controls, indicating strong engineering feasibility.

Fig. 5. Continuous-flow treatment efficiency and application prospects of an integrated electrochemically regenerative reactor.

a Layout drawing of the proposed system integrated into a municipal wastewater treatment plant. b Schematic of the continuous-flow system. c Performance of total organic compound (TOC) removal from coking wastewater in a continuous-flow system. The inset shows the visible changes in color and turbidity before and after treatment of the coking wastewater. Data are presented as mean values ± SD (n = 3). d Relative environmental impact of environmentally relevant descriptors for treating 1 m³ of 0.1 mM phenol wastewater with the catalysis-regeneration continuous-flow system across different catalysis-regeneration cycles (O-C: only catalysis, C-R(x): catalysis-regeneration cycles(x times)). Eighteen environmental impact factors: 1. Global warming; 2. Stratospheric ozone depletion; 3. Ionizing radiation; 4. Ozone formation, Human health; 5. Fine particulate matter formation; 6. Ozone formation, Terrestrial ecosystems; 7. Terrestrial acidification; 8. Freshwater eutrophication; 9. Marine eutrophication; 10. Terrestrial ecotoxicity; 11. Freshwater ecotoxicity; 12. Marine ecotoxicity; 13. Human carcinogenic toxicity; 14. Human non-carcinogenic toxicity; 15. Land use; 16. Mineral resource scarcity; 17. Fossil resource scarcity; 18. Water consumption. For the calculation methods and related data of Life Cycle Assessment (LCA), see Supplementary Text 6 and Supplementary Tables 8 and 9. e Cost comparison for treating 1 m³ of 0.1 mM phenol wastewater with and without catalyst regeneration. f Techno-economic analysis (TEA) of in-situ electrochemical catalyst regeneration versus single-use operation. For related data and sources, see Supplementary Tables 10 and 11. Boxplots of mung bean g root length and h shoot length distribution in different systems. The center point indicates the mean; box limits denote the upper (Q3) and lower (Q1) quartiles; whiskers extend to 1.5× the interquartile range (IQR); all data points are shown (n = 10); the curve is a normal distribution curve. Source data for this figure are provided as a Source data file.

In the practical advanced treatment of coking wastewater (water quality parameters are provided in Supplementary Table 12), the coupled system demonstrated stable continuous operation (Fig. 5c). During the first 50 h of operation using Group A, the removal efficiency of total organic micropollutants remained above 85%, declining by only about 8% by the end of this period. The system was then switched to Group B for continuous treatment, while Group A underwent in situ electrochemical regeneration. Under a fixed 50 h switching cycle, the system uninterruptedly operated and removed >80% TOC for 300 h. The treated effluent exhibited noticeably reduced color and turbidity, confirming the feasibility and stability of this process for real wastewater.

Life-cycle assessment (LCA) and techno-economic analysis (TEA) compared an electrochemical-regeneration packed tower to a non-regenerating packed tower for phenolic wastewater. The normalized environmental hotspot profile across the loop (Supplementary Fig. 54) identified catalyst (NCNT) consumption as the dominant contributor to adverse impacts, primarily because of large amounts of chemical feedstocks and energy in material synthesis (Supplementary Table 8). Subsequent comparisons of life-cycle impact under varying regeneration cycles further underscored the importance of catalyst sparing (Fig. 5d). However, the marginal benefit of impact reduction diminished beyond five cycles. Accordingly, five regeneration cycles were set as the operational optimum for balancing environmental gain with practical performance and cost. The catalysis-regeneration loop outperformed the catalysis-only alternative, lowering the unit treatment cost by 68.2% (Fig. 5e). This advantage was derived from the fact that catalyst preparation dominated the total cost of the catalysis scenario. However, when the catalyst could be utilized over an extended period, the expenditure associated with an in-situ electrochemical regeneration catalysis system was close to the costs of only chemicals and electricity (Fig. 5f, Supplementary Tables 10 and S11). By maximizing catalyst utilization via a low-cost and integrated electrochemical regeneration protocol, the integrated system substantially reduced both critical material demand and overall environmental impact.

Phytotoxicity and animal toxicity assays were performed to evaluate the ecological compatibility of the treatment system. As shown in Fig. 5g, h, 0.2 mM phenol observably inhibited the elongation of both radicles and plumules in Vigna radiata. In contrast, seedlings exposed to the treated effluent, the corresponding oxidant solution, or leachates from fresh and spent catalysts exhibited root and shoot lengths statistically indistinguishable from those of the blank control, indicating complete detoxification. Residual PMS and any catalyst leaching into the environment induced negligible phytotoxic effects, an advantage attributed to the high physicochemical stability of NCNT and the non-leachable nature of surface-bound polymeric products. Zebrafish (Danio rerio) embryogenesis tests further confirmed the detoxification efficacy. Whereas phenolic wastewater induced severe mortality within 24 h (Supplementary Fig. 55), embryos cultured in the treated solution developed normally without observable malformations, serving as a control for the teratogenic group. The consistent results from plant and animal bioassays highlighted the system’s effectiveness for sustainable remediation of refractory wastewater.

Overall, the integrated catalysis-electrochemical regeneration packed-bed reactor combines economic advantage with environmental safety and exhibits clear potential for full-scale deployment.

Discussion

This study presented a packed-bed reactor that seamlessly coupled continuous heterogeneous Fenton-like oxidation with in situ electrochemical regeneration, effectively addressing catalyst deactivation in conventional processes. The regeneration relied on surface-wettability modulation and oxidative polymer transformation to reopen active sites and reestablish interfacial electron-transfer pathways. The reactor demonstrated excellent stability across multiple reuse cycles and during treatment of real industrial effluents, achieving >80% TOC removal from coking wastewater over 300 h of continuous operation. LCA and economic analysis showed substantial reductions in catalyst consumption, operating costs, and environmental burden. After five regeneration cycles, the total cost declined by 68.2% and greenhouse-gas emissions decreased by ~80%. Owing to its compact, modular architecture, the system integrated readily with existing infrastructure, either as a standalone unit for refractory organics or as an advanced polishing step following biological treatment.

Beyond providing a practical route to mitigate polymer-induced deactivation, the electrochemical regeneration strategy advances the real-world applicability of HG-AOPs. Future efforts should target low-cost, high-strength, and high-flux monolithic conductive packing electrodes suitable for scale-up, and embed in situ electrochemical sensing with machine-learning-based control. With millisecond-level signal acquisition and multi-parameter data curation, the system can enable real-time process monitoring and adaptive regulation, laying the groundwork for intelligent water-treatment operations. The integrated approach is extensible to other catalytic platforms, suggesting a potential pathway toward next-generation, high-efficiency, sustainable, and self-cleaning wastewater technologies. In sum, this work provides a practical solution to deactivation challenges in heterogeneous catalysis and suggests a promising development path toward intelligent and modular water treatment systems.

Methods

In situ catalytic-electrochemical regeneration packed-bed reactor setting

A small packed column with an inner diameter of 8 mm and a height of 100 mm was filled with 100 mg of catalyst. The column mainly consists of three parts: a 20 mm high catalytic packing layer, an anode, and a cathode. A 10 mm high nano-sponge block as a support layer was filled at both ends of the catalytic packing layer. Pure titanium wire was chosen as the anode and cathode, with the cathode titanium wire penetrating the catalytic packing layer to polarize the catalyst cathodically during electrochemical regeneration. The electrode interval between the anode and cathode was kept at 10 mm. During catalytic treatment, the mixed solution containing the pollutant and the oxidant (peroxymonosulfate, PMS) was pumped from the bottom of the packed column by a peristaltic pump, flowed upward through the catalyst, and was discharged from the top outlet. During in situ electrochemical regeneration, 100 mM Na₂SO₄ and an appropriate amount of concentrated H₂O₂ solution were fed into the packed column as electrolytes, and electrochemical regeneration was carried out under constant current (all potentials are not iR-corrected). An electrochemical workstation of model CHI760F from Chenhua was used for the power supply. Samples were collected from the reactor outlet at different time intervals for analysis.

Preparation of the catalyst

The synthesis method of NCNT is as follows: 0.5 g of carbon material and 0.5 g of urea were thoroughly ground to mix uniformly. Then, the mixture was placed in a tube furnace and pyrolyzed at 800 °C for 2 h under a nitrogen atmosphere with a heating rate of 10 °C min⁻¹. After pyrolysis, it was cooled to room temperature. The materials obtained from multiple pyrolysis processes were mixed and stored for later use.

Divided electrochemical oxidation system test

In a divided electrochemical oxidation system, 200 µL of catalyst ink (40 mg catalyst powder + 60 µL Nafion solution + 4 mL DMF, sonicated for 1 h) was drop-cast onto a 25 × 25 × 1 mm conductive glass plate (6.4 mg cm⁻²) that simultaneously served as the electron conduit and adsorption substrate for both the pollutant and the oxidant. The two electrodes were connected with a copper wire to enable inter-compartment electron transfer. A salt bridge for ionic communication between the compartments was prepared by dissolving 1.5 g agar and 20 g KCl in 100 mL of water at 90 °C; the hot mixture was poured into a U-shaped glass tube and allowed to solidify at room temperature. The current flowing between the anode and cathode was monitored with a micro-amperemeter. In-situ cathode potentials were recorded by open-circuit potentiometry on an electrochemical workstation (CHI760E, CH Instruments).

Electrochemical analysis tests

Differential pulse voltammetry (DPV), open-circuit potential (OCP), and electrochemical impedance spectroscopy (EIS) were performed in a conventional three-electrode cell comprising an Ag/AgCl (4 M KCl) reference electrode, a platinum-foil counter electrode, and a catalyst-modified glassy-carbon working electrode (d = 3 mm, GaossUnion). The supporting electrolyte for all measurements was 0.1 M phosphate buffer (pH 7.0).

The working electrode was prepared as follows: 5 mg of catalyst powder was dispersed in 1 mL N, N-dimethylformamide (DMF) containing 40 µL Nafion solution (5 wt%, Sigma-Aldrich) and sonicated for 1 h to obtain a homogeneous catalyst ink. Subsequently, 3 µL of the ink was drop-cast onto the glassy-carbon surface (0.5 mg cm⁻²) and dried at ambient temperature.

Experimental procedure for kinetics

Removal kinetics experiments under varying oxidant concentrations were conducted in a glass beaker (100 mL) equipped with magnetic stirring at 500 rpm. The reaction mixture contained 100 µM of the target pollutant and 0.1 g L⁻¹ catalyst in 100 mL of aqueous solution (pH 6.7, 298.15 K). All subsequent experiments were performed in a packed-bed reactor. Quantification of each organic pollutant was carried out by high-performance liquid chromatography (1260 Infinity, Agilent Inc., USA) with a ZORBAX SB C-18 (5 μm, 4.6 × 250 mm) column maintained at 35 ± 1 °C and a UV detector. The mobile-phase composition for each analyte is listed in Supplementary Table 1.

To evaluate PMS adsorption on the NCNT catalyst, 1.0 mL of the sample was mixed with 0.5 mL potassium hydrogen phthalate (0.1 M) and 0.5 mL KCl (0.4 M), allowed to develop for 30 min, and the absorbance was measured at 350 nm with a UV-Vis spectrophotometer.

The oxidant adsorption capacity (Q_e) was obtained via a pseudo-first-order model:

$$Q_t=Q_e\times \left(1-e^{-k_{\mathrm{ads}}t}\right)$$ 1

Where Q_t (mmol g⁻¹) is the adsorption concentration of oxidant at time t, Q_e (mmol g⁻¹) is the adsorption capacity at equilibrium, k_ads is the pseudo-first-order rate constant (min⁻¹), and t is the adsorption time (min).

Calculation method for catalyst regeneration efficiency

In this study, the catalyst regeneration efficiency (R_5h) is defined as the ratio of the catalytic activity of the regenerated catalyst to that of the fresh catalyst under identical reaction conditions.

The regeneration efficiency is calculated as follows:

$$R_{5\mathrm{h}}=\frac{R_{5\mathrm{h},\mathrm{regen}}}{R_{5\mathrm{h},\mathrm{fresh}}}\times 100\%$$ 2

Where R_5h,fresh is the pollutant removal efficiency (%) of the fresh catalyst during the first 5 h of continuous operation in the packed-bed reactor, R_5h,regen is the pollutant removal efficiency (%) of the regenerated catalyst during a subsequent 5-hour operational cycle.

Details for the size exclusion chromatography test

To accurately characterize changes in both the relative molecular mass and the abundance of organic species, size exclusion chromatography (Agilent 1260 Infinity II, USA) was employed to analyze the products extracted from the surface of the deactivated catalyst and the residual organics remaining in the spent regenerant electrolyte.

• (i).Surface extract of the deactivated catalyst: The catalyst was removed from the packed-bed tower after complete deactivation and immersed in 5 mL of an alkaline methanolic solution (methanol: 0.9 mol L⁻¹ NaOH = 1:1 of v/v). The suspension was sonicated for 30 min to desorb surface-bound organics, then filtered through a 10 μm PTFE syringe filter to remove solid particles; the filtrate was directly injected for SEC analysis.
• (ii).Residual organics in the regenerant electrolyte: Upon completion of the regeneration cycle, the electrolyte was withdrawn from the packed-bed tower and passed through a 10 μm aqueous-phase membrane filter; the resulting filtrate served as the analytical sample.

Details of theoretical calculations

The highest occupied molecular orbital energies (E_HOMO, eV) of the pollutants were investigated using quantum chemical density functional theory (DFT). Geometric configurations were optimized with the B3LYP-D3 (BJ) method under the B3LYP-D3/SVP basis set. The energies of the HOMO and LUMO (E_HOMO and E_LUMO, eV) were calculated using the B3LYP-D3/TZVP basis set. The calculated frontier orbital energies for the various pollutants are presented in Supplementary Table 7 and Supplementary Figs. 39–48.

For the calculation of periodic systems, the calculation was performed with the Perdew–Burke–Ernzerhof (PBE) functional within the generalized-gradient approximation (GGA). The catalyst slab was sampled with a 1 × 1 × 1 Γ-centered k-mesh and expanded in a plane-wave basis truncated at 500 eV. Valence occupancies were smeared by a 0.2 eV Gaussian width; electronic self-consistency was declared when successive total energies differed by <10⁻⁵eV. Ionic relaxations were considered converged once the force on every atom fell below 0.02 eV Å⁻¹. A 20 Å vacuum spacer was inserted perpendicular to the surface to eliminate spurious periodic interactions, and semi-empirical D3 dispersion corrections (Grimme) were included to capture long-range van der Waals forces. The density overlap region indicator (DORI) was calculated following established methodologies, where the dimensionless parameter θ is mathematically expressed as:

$$\theta \left(\mathrm{r}\right)=\frac{{\left(\nabla k^2\left(\mathrm{r}\right)\right)}^2}{{\left(k^2\left(\mathrm{r}\right)\right)}^3}=\frac{{\left(\nabla {\left(\frac{\nabla \rho \left(\mathrm{r}\right)}{\rho \left(\mathrm{r}\right)}\right)}^2\right)}^2}{{\left(\frac{\nabla \rho \left(\mathrm{r}\right)}{\rho \left(\mathrm{r}\right)}\right)}^6}$$ 3

Where k(r) is the local momentum vector, defined as:

$$k\left(\mathrm{r}\right)=\frac{\nabla \rho \left(\mathrm{r}\right)}{\rho \left(\mathrm{r}\right)}$$ 4

Where $\rho \left(\mathrm{r}\right)$ is the electron density in real space, $\nabla \rho \left(\mathrm{r}\right)$ is the electron density gradient in real space. The DORI could be defined as:

$$\mathrm{DORI}=\frac{\theta }{1+\theta }$$ 5

To account for the effect of an applied electrode potential, the Gibbs free-energy change of any electrochemical step is corrected as:

$$\mathrm{\Delta }G\left(\mathrm{U}\right)={\mathrm{\Delta }G}_{\mathrm{DFT}}-neU$$ 6

Where ΔG_DFT is the zero-potential free energy obtained from standard DFT, n is the number of electrons transferred, e is the elementary charge, and U is the potential reported versus the standard hydrogen electrode (SHE).

For periodic systems under applied electrode potentials, structural optimizations of H₂O₂ activation and radical generation processes were performed using the constant-potential method within the implicit solvation model VASPsol, as implemented in the CP‑VASP code^53,54. Solvation effects were accounted for by VASPsol, which simulates the dielectric environment of water (dielectric constant ε = 78.4), with an appropriate Debye length set to describe electrolyte screening effects. To determine reaction pathways and energy barriers, the Climbing Image Nudged Elastic Band (CI‑NEB) method was employed to search for transition states. The force convergence criteria were consistent with those used for structural relaxations, and the obtained transition states were verified by frequency analysis to ensure the presence of a single imaginary frequency.

Reporting summary

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

Author contributions

J. J. Huang, P. J. Duan, and F. Chen conceived and designed the experiments. J. J. Huang carried out the experiments. P. J. Duan performed the theoretical calculations. J. J. Huang, P. J. Duan, C. W. Bai, Z. Q. Zhang, J. Wang, X. J. Chen, J. S. Guo, Y.J. Zhang, W.J. Wang, and F. Chen assisted in analyzing various characterizations. J. J. Huang and P. J. Duan wrote the paper. Y.J. Zhang, W.J. Wang, P. J. Duan and F. Chen further revised the paper. All authors contributed to the discussion of the manuscript.

Peer review

Peer review information

Nature Communications thanks Chad Vecitis and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

F.C., P.J.D., and W.J.W. disclose support for the research of this work from the National Natural Science Foundation of China (52270149, 42507077, 52300204), New Chongqing Youth Innovative Talent Program (CSTB2025YITP-QCRCX0056), the Open Fund of the State Key Laboratory of Advanced Environmental Technology (SKLAET202521), the Fundamental Research Funds for the Central Universities grant (2024IAIS-QN013), the China Postdoctoral Science Foundation (2024M763859), and the National Funded Postdoctoral Researcher Program (GZC20233318).

Data availability

The data supporting the findings of the study are included in the main text and supplementary information files. Raw data can be obtained from the corresponding authors upon request. Source data are provided in this paper. 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/s41467-026-76638-6.