In situ iodine generation enables solution-phase polymerization of organic pollutants for continuous resource recovery from water

Abstract

Oxidation-driven polymerization of organic pollutants offers a route to convert contaminants into value-added products and enable continuous resource recovery from water, but conventional heterogeneous processes are plagued by polymeric overlayers that foul catalyst surfaces and hinder long-term operation. Here, we realize solution-phase polymerization of 4-chlorophenol (4-CP) by relocating key bond-forming steps from catalyst interfaces into the bulk aqueous phase. In a facet-engineered bismuth oxyiodide/hydrogen peroxide (BiOI/H₂O₂) system, BiOI directs H₂O₂ activation toward in situ generation of freely diffusing, moderately oxidizing iodine (I₂) as the dominant reactive species. Accordingly, I₂ drives 4-CP polymerization predominantly in solution rather than on the solid surface, delivering a record 73% retention of polymeric products in the aqueous phase. In situ scanning electrochemical cell microscopy (SECCM) directly visualizes the spatial generation of I₂ and its subsequent reaction with 4-CP. Complementary spectroscopy and density functional theory reveal that the (110) facet of BiOI stabilizes H₂O₂ adsorption via a hydrogen-bond bridging configuration that promote O–O bond cleavage and I₂ formation, whereas other favors radical pathway. This “facet–species–phase” strategy, in which catalyst facets govern the dominant reactive species and thus direct solution-phase polymerization, enables fouling-resistant advanced oxidation that couples water purification with continuous resource recovery.

Catalyst fouling hinders oxidation-driven polymerization of organic pollutants. This study shows a facet-programmed BiOI/H₂O₂ system generating I₂ in situ to drive solution-phase polymerization, eliminating fouling while enabling resource recovery.

Introduction

Water pollution is a global challenge, with over 4.4 billion people lacking access to safely managed drinking water services. Contamination with hazardous organic pollutants such as industrial chemicals and pharmaceuticals leads to waterborne diseases, causing approximately 1.4 million deaths annually¹. Advanced oxidation processes (AOPs), particularly Fenton and Fenton-like reactions, have shown promise in addressing these issues by generating reactive species. Hydrogen peroxide (H₂O₂) is commonly used as a green oxidant in these reactions, decomposing into water and oxygen, thus enabling sustainable water treatment². However, the practical application of H₂O₂-based Fenton-like systems is constrained by narrow pH windows, high carbon dioxide (CO₂) emissions, and the costly continuous H₂O₂ supply³.

Recent advances in Fenton-like systems have redirected research focus from complete mineralization of pollutants to their oxidative polymerization^4–6. Polymerization drastically lowers oxidant consumption, minimizes secondary pollution, and enables recovery of polymeric products for value-added reuse, aligning with sustainability and circular economy objectives^7,8. However, unlocking the full potential of polymerization-based AOPs requires overcoming three critical challenges: (1) ensuring high selectivity for polymerization over mineralization, (2) mitigating the deposition of polymeric products on catalyst surfaces, which block active sites and lead to catalyst deactivation, and (3) enabling the efficient separation and recovery of polymeric products for value-added utilization.

Addressing these challenges motivates directing selective polymerization into the aqueous phase, where polymer formation proceeds away from the catalyst surface, thereby minimizing catalyst fouling while facilitating downstream product separation and recovery. The feasibility of such solution-phase polymerization depends critically on both the oxidative capacity of reactive species and their spatial localization—whether freely diffusing in solution or binding to the catalyst surface. Free reactive species are more effective in driving aqueous-phase polymerization. For example, in a thermally activated persulfate system, free radicals, particularly hydroxyl radicals (^•OH), have been shown to induce polymerization of organic pollutants with 24.8–91.2% carbon recovery⁹. However, the strong oxidation power of ^•OH (1.9–2.8 V) also triggers competing mineralization reactions that compromise polymerization selectivity¹⁰. To overcome this limitation, nonradical activation of H₂O₂ has been developed to generate moderately oxidative species, such as surface-bound H₂O₂ complexes (H₂O₂*, 0.6–1.2 V)^11,12 and high-valent metal species (e.g., Fe(IV) = O¹³ and Mn(IV), 1.0–2.0 V)^14,15, which favor polymerization. Unfortunately, these species primarily initiate polymerization at the solid-liquid interface, and the resulting products deposit on the catalyst surface via π–π interactions or coordination adsorption, thereby leading to catalyst deactivation and impeding polymer recovery^16,17. Therefore, achieving efficient solution-phase polymerization calls for the development of new reactive species that selectively drive polymerization without excessive oxidation or reliance on the catalyst surface.

Compounding these challenges, although persulfate-driven polymerization has been extensively studied, polymerization in H₂O₂ systems remains strikingly underexplored. This disparity is particularly notable given the intrinsic advantages of H₂O₂ as an environmentally benign oxidant, including the absence of secondary anionic byproducts (e.g., sulfates), and its amenability to in situ generation, which facilitates long-term, sustainable wastewater treatment. The research gap stems from the high O–O bond dissociation energy that kinetically disfavors polymerization activity¹⁸, together with its inherent propensity for homolytic cleavage that generates ^•OH and compromises polymerization selectivity¹⁹. Collectively, these limitations underscore the need for efficient H₂O₂-driven polymerization, expanding the role of this environmentally benign oxidant from pollutant destruction to resource-oriented valorization. Of equal importance is shifting H₂O₂-driven polymerization into the solution phase, avoiding catalyst fouling and enabling continuous and sustainable water treatment.

Bismuth oxyhalides (BiOX, X = Cl, Br, I) provide a compelling platform for exploring reactive species that drive solution-phase polymerization in H₂O₂-based systems, owing to their low cost, low toxicity, limited π–π interaction with polymeric products, and ability to mediate multi-pathway H₂O₂ activation^20–23. In this work, we present a facet-programmed bismuth oxyiodide (BiOI) to decouple polymerization from the catalyst surface into the aqueous phase within an efficient heterogeneous H₂O₂-based AOPs system. This distinctive reactivity arises from the (110) facet of BiOI, which selectively activates H₂O₂ to in situ generate molecular iodine (I₂), a moderately oxidizing (0.62 V) and freely diffusing species that can rapidly react with phenolic compounds (k = 10⁵–10⁸M^–1 s^–1) in the solution phase to produce polymeric products^24–26. By redirecting polymerization into the solution phase, we promote selective 4-chlorophenol (4-CP) coupling while avoiding product fouling.

Using this facet-programmed approach, polymeric products are effectively retained in the solution phase, coupled with high polymerization selectivity. In situ scanning electrochemical cell microscopy (SECCM) and density functional theory (DFT) calculations confirm that I₂ acts as the dominant reactive species, formed through the facet-controlled O–O cleavage of H₂O₂. Additionally, integration of KI-based iodine replenishment closes the iodine loop, enabling continuous resource recovery and sustained fouling-resistant performance. This work establishes a general design principle of “facet–species–phase”, in which catalyst facets govern the dominant reactive species and thus decouple polymerization from the catalyst surface, offering a scalable and durable solution for continuous water treatment with recoverable resources.

Results

Characterization of BiOI with different facet orientations

X-ray diffraction (XRD) patterns of the as-synthesized BiOI catalysts exhibited diffraction peaks that matched well with the standard tetragonal BiOI structure (JCPDS No. 73-2062, Fig. 1a), confirming the successful synthesis of BiOI. The preferential orientation of (hkl) facets in different catalysts was quantified via texture coefficients (TC_hkl), wherein higher TC_hkl values denote a more dominant orientation of the respective (hkl) facet²⁷. As summarized in Supplementary Table 1, the three prepared BiOI catalysts exhibited the highest TC₁₁₀, TC₁₀₂, and TC₀₀₁ values, respectively, indicating preferential exposure of (110), (102), and (001) facets. Accordingly, they were denoted as BiOI-110, BiOI-102, and BiOI-001.

Fig. 1. Facet characterization of BiOI.

a XRD patterns of BiOI with different facet orientations. b, d, f TEM (insets are SAED images) and c, e, g HRTEM images of BiOI-110 (b, c), BiOI-102 (d, e), and BiOI-001 (f, g), respectively. The corresponding raw fast Fourier-transform (FFT) images for (c, e, and g) are provided in Supplementary Figs. 3–5. TEM and HRTEM images represent one of three independent experiments showing consistent results. h–j Side-view atomic structure models of BiOI-110 (h), BiOI-102 (i), and BiOI-001 (j).

Morphological and structural characterizations were performed using transmission electron microscopy (TEM) and scanning electron microscopy (SEM). TEM and SEM images (Fig. 1b, Supplementary Figs. 1, 2) revealed that BiOI-110 featured a stacked nanosheet structure with an approximate width of 100 nm and a thickness of 16 nm. High-resolution TEM (HRTEM) further resolved lattice fringes with a spacing of 0.28 nm, which corresponded to the (110) and (1–10) planes of BiOI (Fig. 1c, Supplementary Fig. 3). The selected-area electron diffraction (SAED) pattern of BiOI-110 exhibited a 90° angle, consistent with the theoretical angle between the (002) and (1–10) planes. The set of diffraction spots could be indexed as the (110) zone axis of tetragonal BiOI, confirming dominant (110) facet exposure²⁸. Likewise, BiOI-102 and BiOI-001 preferentially exposed the (102) and (001) facets, respectively, as evidenced by HRTEM and SAED analyses (Fig. 1d–g, Supplementary Figs. 1, 4, 5). Both BiOI-102 and BiOI-001 samples displayed stacked nanosheet structures with larger lateral dimensions than BiOI-110, resulting in lower Brunauer-Emmett-Teller (BET) surface areas (Supplementary Fig. 6).

X-ray photoelectron spectroscopy (XPS) analysis revealed negligible facet-dependent variations in the valence states of Bi and I in BiOI (Supplementary Fig. 7). Raman spectra detected characteristic peaks at around 87 and 150 cm^–1, assigned to the A_1g and E_g stretching modes of Bi–I vibration, respectively. Notably, BiOI-110 exhibited a subtle redshift in both peaks compared to BiOI-102 and BiOI-001, suggesting elongation of Bi–I bonds arising from a facet-specific structural distortion (Supplementary Fig. 8), which may enhance its catalytic activity²⁹. Collectively, distinct facet orientations of BiOI resulted in different surface structures, as schematically depicted in Fig. 1h–j.

Fenton-like performance for model pollutant removal

The Fenton-like performance of as-prepared catalysts was assessed using 4-CP as a model pollutant. Systematic parameter optimization based on degradation efficiency identified the optimal reaction conditions as 2 mM H₂O₂ and a catalyst dosage of 0.3 g L⁻¹ (Supplementary Figs. 9, 10). Control experiments confirmed negligible 4-CP removal via either catalyst adsorption or direct H₂O₂ oxidation, verifying that catalytic oxidation dominated 4-CP elimination (Supplementary Fig. 11). As shown in Fig. 2a, BiOI-110-activated H₂O₂ achieved nearly complete 4-CP degradation within 30 min, with a first-order rate constant (k_obs) of 0.26 min⁻¹. In contrast, both BiOCl-110 and BiOBr-110 exhibited negligible ability to activate H₂O₂, with less than 4% 4-CP removal (characterized in Supplementary Fig. 12 and Supplementary Table 1). This dramatic activity gap underscored the essential role of iodide sites in H₂O₂ activation. Crucially, substituting BiOI-110 with iodide ions (KI) resulted in minimal 4-CP degradation, demonstrating that lattice iodide, rather than dissolved iodide ions, governs H₂O₂ activation.

Fig. 2. Catalytic performance of BiOI in Fenton-like reactions.

a, b 4-CP degradation via H₂O₂ activation on (110) facet-exposed bismuth oxyhalides (a), and BiOI with different facet orientations (inset shows corresponding pseudo-first-order rate constants of 4-CP removal) (b). c Correlation between TC_hkl and k_obs. Reaction conditions in (a–c): [4-CP]₀ = 50 μM, [catalyst] = 0.3 g L⁻¹, [H₂O₂] = 2 mM, initial pH = 7.0. d Comparison of 4-CP removal performance (by normalized rate constant k_n) across different Fenton-like systems. e 4-CP removal in the BiOI-110/H₂O₂ system with the co-existing water matrix components. Reaction conditions: [4-CP]₀ = 50 μM, [catalyst] = 0.3 g L⁻¹, [H₂O₂] = 2 mM, [NO₃⁻] = [Cl⁻] = [SO₄²⁻] = 2 mM, [HA] = 5 mg L⁻¹, initial pH = 7.0. f Correlation between IP and k_obs. Reaction conditions: [pollutant]₀ = 50 μM, [catalyst] = 0.3 g L⁻¹, [H₂O₂] = 2 mM, initial pH = 7.0. The error bars in the figures represent the standard deviations (mean ± s.d., n = 3) from triplicate tests.

To further verify the contribution of lattice iodide, BiOI with distinct facet orientations was comparatively assessed for H₂O₂ activation. Despite sharing similar nanosheet morphologies and valence states, their catalytic activities diverged markedly (Fig. 2b). BiOI-001, which lacks exposed I sites, showed poor H₂O₂ activation performance (k_obs = 0.03 min⁻¹). Notably, BiOI-102, though possessing surface-exposed I sites, exhibited marginal activity, with k_obs (0.05 min⁻¹) only one-fifth that of BiOI-110. These results indicated that H₂O₂ activation by BiOI relied not only on the presence of lattice I sites but also on their spatial arrangement across different crystal facets.

A quantitative correlation analysis between facet orientations (quantified by TC_hkl) and catalytic activity (k_obs) provided direct evidence for facet-dependent catalysis. While neither TC₁₀₂ nor TC₀₀₁ correlated with k_obs, TC₁₁₀ exhibited a strong positive correlation with activity (R² = 0.91, Fig. 2c and Supplementary Fig. 13), identifying (110) as the active facet for H₂O₂ activation. Consistently, BiOI-110 exhibited the highest H₂O₂ decomposition efficiency, confirming the predominance of the (110) facet in reactive species generation (Supplementary Fig. 14). Benchmarking against state-of-the-art catalysts, the BiOI-110-driven Fenton-like system achieved a normalized 4-CP degradation rate constant (k_n) of 21.7 × 10⁻³ min⁻¹ g⁻¹ L, outperforming previously reported metal oxide-catalyzed H₂O₂ systems, thus highlighting effective H₂O₂ activation capability of BiOI-110 (Fig. 2d and Supplementary Table 2).

The BiOI-110/H₂O₂ system maintained efficient 4-CP removal from pH 4 to 9 (Supplementary Fig. 15). The pH adaptability implied that 4-CP removal was independent of pH-sensitive reactive species (e.g., ^•OH), and demonstrated the stability of BiOI-110 under varying pH conditions. This system also sustained robust 4-CP removal efficiency in the presence of common water matrix components, including typical anions and dissolved organic matter (humic acid (HA) as a model), highlighting its practical applicability (Fig. 2e and Supplementary Fig. 16). A structure-activity relationship analysis for phenolic pollutants revealed a strong linear correlation (R² = 0.98) between degradation kinetics and pollutant ionization potential (IP, Fig. 2f and Supplementary Fig. 17). As IP reflects the electron-donating ability, this correlation indicated that pollutant oxidation in the BiOI-110/H₂O₂ system was primarily governed by the electron-donating ability of target pollutants, which was likely attributed to the electrophilic nature of the generated reactive species³⁰.

Distribution of polymeric products from 4-CP oxidation

4-CP oxidation products in the BiOI-110/H₂O₂ system were analyzed using chemical oxygen demand (COD) and ultra-performance liquid chromatography-mass spectrometry (UPLC-MS). Suspension COD (including both solution and catalyst) remained nearly constant throughout the reaction, indicating minimal mineralization of 4-CP to CO₂ (Fig. 3a). UPLC-MS revealed no detectable intermediates related to mineralization pathways, further suggesting that 4-CP removal proceeded mainly through polymerization (Supplementary Figs. 18, 19).

Fig. 3. The distribution of polymeric products in the BiOI-110/H₂O₂ system.

a The removal of chemical oxygen demand (COD) and 4-CP in the BiOI-110/H₂O₂ system. Reaction conditions: [4-CP]₀ = 0.5 mM, [catalyst] = 1.0 g L⁻¹, [H₂O₂] = 20 mM, initial pH = 7.0. b Polymer partition coefficient in solution phase (θ_sol) at different initial 4-CP concentrations in the BiOI-110/H₂O₂ system. Reaction conditions: [catalyst] = 1.0 g L⁻¹, [H₂O₂] = 20 mM, initial pH = 7.0. c Comparison of θ_sol across different Fenton-like systems (data were extracted from published references; detailed information was provided in Supplementary Table 3). The error bars in the figures represent the standard deviations (mean ± s.d., n = 3) from triplicate tests.

Reactive intermediates governing 4-CP polymerization were then investigated by employing ferulic acid and acetonitrile as probes. Ferulic acid can revert phenoxyl radicals (PhO^•) to their phenolic precursor⁵, whereas acetonitrile scavenges electrophilic phenoxonium ions (PhO⁺) through nucleophilic attack⁴. Compared with acetonitrile, the addition of ferulic acid completely suppressed 4-CP removal, demonstrating PhO^• as the key intermediate driving 4-CP polymerization (Supplementary Fig. 20). Based on these results, a plausible oxidation pathway for 4-CP in the BiOI-110/H₂O₂ system was proposed in Supplementary Fig. 21. 4-CP underwent selective one-electron oxidation to form PhO^• as the key reactive intermediates. Subsequent radical coupling of PhO^• yielded polymeric products, including polyphenyl ethers (via C–O coupling), polyphenylenes (via C–C coupling), and corresponding dechlorinated derivatives.

In polymerization-dominated oxidation systems, the distribution of polymeric products exerts little influence on suspension COD but strongly affects solution COD. Solution-retained polymers maintain aqueous COD despite significant pollutant removal. Conversely, catalyst-deposited polymers reduce aqueous COD in proportion to degradation efficiency. Therefore, polymer distribution can be quantified from the divergence between solution COD and pollutant removal efficiency, from which we proposed an indicator, θ_sol, representing polymer partition coefficient in the solution phase. As defined in Eq. 1, θ_sol was calculated based on the catalyst-deposited polymer fraction (θ_cat), while θ_cat was reflected by the ratio of solution COD (or TOC) removal efficiency (η_cat, representing the transfer of COD/TOC to catalyst surface) to pollutant removal efficiency (η_p).

$${\theta }_{\mathrm{sol}}=1-{\theta }_{\mathrm{cat}}=1-\frac{{\eta }_{\mathrm{cat}}}{{\eta }_{\mathrm{p}}}$$ 1

To ensure measurable COD, we increased the initial 4-CP concentration (0.5 mM) for θ_sol determination. Under the applied conditions, 4-CP removal efficiency was 43%, while the solution COD decreased by only 11%, resulting in a θ_sol of 73% (Fig. 3a). This indicated that 73% of 4-CP-derived polymeric products were distributed in the aqueous phase in the BiOI-110/H₂O₂ system. Solution-phase-dominated polymerization was observed across varied initial 4-CP concentrations, with θ_sol values ranging from 67% to 97% (Fig. 3b and Supplementary Fig. 22). The increase in θ_sol at higher initial 4-CP concentrations was attributed to the elevated pollutant-to-catalyst ratio that reduced the fraction of polymeric products deposited on the BiOI-110 surface. By comparison, reported θ_sol for other polymerization-dominated Fenton-like systems were typically below 20% (Fig. 3c and Supplementary Table 3), suggesting that polymeric products in most systems predominantly deposited on catalyst surfaces through π–π interaction or coordination adsorption^4,16,31. The distinctive aqueous-phase polymerization in the BiOI-110/H₂O₂ system mitigated catalyst deactivation and facilitated polymeric product recovery.

Thermogravimetric analysis (TGA) of BiOI-110 recovered from COD testing further validated minimal catalyst fouling. The mass loss difference between pristine and used catalysts corresponded to polymer mass deposited on the catalyst surface. As shown in Supplementary Fig. 23, only 0.9 mg of surface-accumulated polymer was detected, compared to 2.8 mg of removed 4-CP, indicating that most (~68%) polymeric products remained in solution.

Complementing the above results, UPLC-MS analyses were employed to examine 4-CP oxidation products both in the aqueous phase and on the catalyst surface (extracted with ethanol). As summarized in Supplementary Table 4, only trace dimers (<0.2% of the aqueous-phase concentration, m/z = 252.98) were detected in the ethanol-extracted sample, while trimers (m/z = 378.97 and 345.01) and tetramers (m/z = 504.96) were exclusively observed in the bulk solution. Since ethanol effectively dissolves most surface-accumulated polymers⁴, the combined UPLC-MS, COD, and TGA results consistently indicated that up to 73% of 4-CP polymerization products were distributed in the solution phase in the BiOI-110/H₂O₂ system. This result was markedly different from most reported Fenton-like systems, where polymeric products predominantly accumulated on the catalyst surface, and the mechanism underlying this distinction will be investigated in subsequent sections^31,32.

Investigation of reactive species in the BiOI-driven Fenton-like system

To elucidate the mechanism of solution-phase polymerization, reactive species in the BiOI-110/H₂O₂ system were identified through quenching experiments and electron paramagnetic resonance (EPR) analyses. Figure 4a and Supplementary Fig. 24 revealed marginal suppression of 4-CP removal by methanol (MeOH, free ^•OH scavenger), tert-butanol (TBA, free ^•OH scavenger), NaF (surface-bound ^•OH scavenger), p-benzoquinone (p-BQ, O₂^•− scavenger), and furfuryl alcohol (FFA, ¹O₂ scavenger). Kinetic analysis in the presence of scavengers quantified that ^•OH, O₂^•−, and ¹O₂ contributed only 0.3%, 13.3%, and 2.9% to the overall 4-CP removal, respectively (inset of Fig. 4a). Complementary EPR analyses, using 5,5-dimethyl-1-pyrroline-N-oxide (DMPO, for ^•OH and O₂^•−) and 2,2,6,6-tetramethylpiperidine (TEMP, for ¹O₂) as trapping agents, detected no observable signals for DMPO-^•OH, DMPO-O₂^•− and TEMP-¹O₂ adducts (Supplementary Fig. 25). Collectively, these results implied a limited role of ^•OH, O₂^•−, and ¹O₂ in the oxidative removal of 4-CP in the BiOI-110/H₂O₂ system.

Fig. 4. Identification of reactive species.

a Effect of different scavengers on 4-CP degradation in the BiOI-110/H₂O₂ system (inset shows the corresponding contribution of different reactive species). Reaction conditions: [4-CP]₀ = 50 μM, [catalyst] = 0.3 g L⁻¹, [H₂O₂] = 2 mM, [MeOH] = [TBA] = 100 mM, [FFA] = [p-BQ] = [NaF] = 5 mM, initial pH = 7.0. b In situ Raman spectra of the BiOI-110/H₂O₂ system. c Semi-in-situ I 3d XPS spectra of BiOI-110 during H₂O₂ activation. d Variation of I₂ concentration in different systems (the dash-dotted line illustrated I₂ concentration evolution in the BiOI-110/H₂O₂ system following 4-CP addition (+4-CP) at the indicated time point). Reaction conditions: [4-CP]₀ = 50 μM, [catalyst] = 0.3 g L⁻¹, [H₂O₂] = 2 mM, initial pH = 7.0. e Schematic of local electrochemical measurement for H₂O₂ activation by BiOI-110 using scanning electrochemical cell microscopy (SECCM). f, g SEM image of BiOI-110 (f) and their colocalized high-resolution SECCM current map at 0.57 V vs. Ag/AgCl (g). h Representative local cyclic voltammetric curves in different systems (i-v indicated catalyst-containing regions). Each experiment was repeated three times independently with similar results. i Effect of different scavengers on 4-CP degradation in the BiOI-102/H₂O₂ system (inset shows the corresponding contribution of different reactive species). Reaction conditions: [4-CP]₀ = 50 μM, [catalyst] = 0.3 g L⁻¹, [H₂O₂] = 2 mM, [MeOH] = [TBA] = 100 mM, [FFA] = [p-BQ] = [NaF] = 5 mM, initial pH = 7.0. j EPR spectra of the BiOI-102/H₂O₂ system using DMPO and TEMP as trapping agents. k, l Illustration of H₂O₂ activation pathways on BiOI-110 (k) and BiOI-102 (l). The error bars in the figures represent the standard deviations (mean ± s.d., n = 3) from triplicate tests.

In situ Raman tests were then conducted to probe the formation of H₂O₂*. As shown in Fig. 4b, a characteristic O–O stretching vibration of H₂O₂ appeared at approximately 874 cm^–1. Following the introduction of BiOI-110, a new peak emerged at 850 cm^–1, ascribed to H₂O₂* with an elongated O–O bond^33,34. Critically, the H₂O₂*-associated signal gradually diminished upon the addition of 4-CP, indicating H₂O₂* consumption during 4-CP removal.

However, ambiguity remained regarding whether H₂O₂* directly oxidized 4-CP or acted indirectly as an intermediate for generating other reactive species. To clarify this, semi-in-situ XPS analysis was conducted to investigate the involvement of Bi and I sites in H₂O₂ activation. Negligible shifts were observed in the Bi 4f orbitals, indicating a limited role of Bi sites in the Fenton-like process (Supplementary Fig. 26). In contrast, pristine BiOI-110 featured I 3d peaks at 619.2 and 630.7 eV, characteristic of I⁻¹¹¹. Upon H₂O₂ addition, new peaks emerged at 620.1 and 631.6 eV, suggesting the generation of I₂ during H₂O₂ activation^35,36. These I₂-related peaks diminished after introducing 4-CP, accompanied by the emergence of weak C–I bond peaks (Fig. 4c). It suggested that I₂ participated in 4-CP degradation, with a small fraction of polymeric products accumulated on the catalyst surface despite the majority dispersing in the solution phase.

UV-vis spectroscopy further confirmed I₂ generation during H₂O₂ activation via a characteristic absorption peak at 460 nm^37,38. In the absence of 4-CP, the I₂ concentration plateaued at approximately 0.08 mM after 2 min—a concentration well below the solubility limit of I₂ (1.3 mM), ensuring the generated I₂ diffused freely into the aqueous phase (Supplementary Figs. 27, 28 and Fig. 4d). Introducing 4-CP at this time point induced a rapid decline in I₂ concentration to nearly 0 mM, indicating its instantaneous consumption during 4-CP oxidation. When H₂O₂ and 4-CP were co-present initially, I₂ was barely detectable, further confirming that in situ-generated I₂ was instantaneously consumed by 4-CP. The oxidative capability of I₂ toward 4-CP was further validated in an I₂-only system, achieving approximately 80% 4-CP removal within 2 min (Supplementary Fig. 29). The rapid oxidation of phenolic compounds by I₂ was consistent with a previous report²⁵. Although I₂ can spontaneously hydrolyze to HOI, the low reaction rate constant between HOI and 4-CP (5 M^–1 s^–1) excluded HOI as a major contributor to 4-CP degradation³⁹.

To directly visualize the spatial generation of I₂ and its subsequent reaction with 4-CP, we performed in situ scanning electrochemical cell microscopy (SECCM) integrated with local cyclic voltammetry (CV), as schematically illustrated in Fig. 4e. BiOI-110 was sparsely deposited onto a glassy carbon substrate. A pipette with an orifice diameter of 600 nm was filled with 20 mM KCl and 50 mM H₂O₂. The pipette was moved toward the substrate surface while a potential was applied between the substrate electrode and the quasi-reference counter electrode (QRCE). Upon contact between the pipette droplet and the substrate, CV curves were collected. The pipette was then retracted 3 μm and hopped to a new position, and repeating this process over different locations enabled CV mapping.

From the CV mapping data, the current response at 0.57 V (vs. Ag/AgCl) across various positions was extracted and displayed as a heat map. Each pixel in the heat map corresponded to a single CV measurement. The heat map exhibited some hot spots, and colocalization analysis confirmed these hot spots overlapped with BiOI-110 catalyst locations (Supplementary Fig. 30 and Fig. 4f, g). This spatial correlation clearly demonstrates that H₂O₂ activation occurred exclusively on the BiOI-110 surface. The CV curve obtained at position i showed two oxidation peaks (Fig. 4h). The approximately 0.5 V gap between the two cathodic peaks corresponded to the I⁰/I⁺ redox couple, verifying that I₂ was produced during the activation of H₂O₂ by BiOI-110⁴⁰. The I⁺ peak was derived from the further electro-oxidation of I₂, with the stabilization of I⁺ by Cl⁻ in electrolyte⁴¹. Similarly, SECCM tests were performed in the 4-CP/H₂O₂ and BiOI-110/H₂O₂/4-CP systems (Supplementary Fig. 31 and Fig. 4h). 4-CP exhibited a weak oxidation peak at around 0.7 V (vs. Ag/AgCl). When 4-CP was added to the BiOI-110/H₂O₂ system, both the oxidation peaks of I₂ and 4-CP decreased, suggesting that the in situ-generated I₂ was quickly consumed during the oxidation of 4-CP. Notably, the I⁺ peak disappeared in the presence of 4-CP as I₂ consumption inhibited its further oxidation to I⁺⁴². Taken together, these SECCM findings visualized that BiOI-110 triggered the activation of H₂O₂ to produce I₂, which served as the key reactive species responsible for 4-CP degradation.

Open-circuit potential measurements provided additional evidence for the transient formation and rapid diffusion of I₂ (Supplementary Fig. 32). The potential of the BiOI-110-coated glassy carbon electrode increased immediately upon H₂O₂ addition, consistent with the rapid generation of oxidative I₂ species. However, subsequent addition of 4-CP induced only a minor potential drop, which was distinct from the sharp decrease typically observed in systems dominated by surface-confined reactive species (e.g., H₂O₂*⁴³ and high-valent metals⁶). These observations further implied that the generated I₂ diffused freely into the aqueous phase for 4-CP oxidation.

Consequently, I₂ was the primary oxidative species for the solution-phase 4-CP polymerization in the BiOI-110/H₂O₂ system, and H₂O₂* served as an intermediate for I₂ formation instead of a reactive species (Fig. 4k). Notably, the potential toxicity of the generated I₂ was negligible, as excess I₂ could be decomposed by H₂O₂ to generate O₂ (Supplementary Figs. 33–35).

For comparison, reactive species in the BiOI-102/H₂O₂ system were also investigated. 4-CP oxidation was almost completely suppressed by p-BQ, with the contribution of O₂^•− estimated to be 86.0% (Fig. 4i and Supplementary Fig. 36). EPR analyses detected strong DMPO-O₂^•− signals in the presence of BiOI-102 and H₂O₂, which decreased markedly following 4-CP addition (Fig. 4j), confirming O₂^•− as the main reactive species in the BiOI-102/H₂O₂ system (Fig. 4l). In addition, negligible I₂ generation was observed in the BiOI-102/H₂O₂ system (Fig. 4d). These findings established that reactive species generation in BiOI-driven Fenton-like systems was facet-dependent. The (110) facet selectively generated I₂ via H₂O₂ activation, while the (102) facet favored O₂^•− formation.

Mechanism of facet-programmed in situ iodine generation

Given that H₂O₂ activation initiates with its adsorption onto catalyst surfaces, we investigated H₂O₂ adsorption behaviors on BiOI-110 and BiOI-102 to elucidate the facet-dependent divergence in activation pathways. Under high ionic strength conditions (100 mM NaClO₄), 4-CP degradation by BiOI-110/H₂O₂ was markedly suppressed (k_obs decreased from 0.26 min⁻¹ to 0.05 min⁻¹), whereas the BiOI-102/H₂O₂ system maintained nearly the same activity (Supplementary Fig. 37). This contrast implied that H₂O₂ underwent physisorption on BiOI-110, whereas chemisorption dominated on BiOI-102⁴⁴.

In situ Fourier-transform infrared (FTIR) spectra provided further evidence of distinct adsorption configurations. On BiOI-110, the characteristic O–O stretching vibration of H₂O₂ at 877 cm⁻¹ was preserved with no additional peaks detected (Fig. 5a). It indicated physisorption (e.g., hydrogen-bonded adsorption) between H₂O₂ and BiOI-110, consistent with previous reports^11,45,46. In contrast, H₂O₂ adsorption on BiOI-102 showed a distinct peak at 847 cm⁻¹, attributed to the formation of Bi-OOH species (Fig. 5b)^47,48. These results demonstrated that BiOI-110 and BiOI-102 facilitated fundamentally different H₂O₂ adsorption configurations, which potentially led to distinct H₂O₂ activation pathways.

Fig. 5. Selective I₂ generation through facet-programmed BiOI.

a, b In situ FTIR spectra of H₂O₂ adsorption on BiOI-110 (a) and BiOI-102 (b) surfaces. c Optimized H₂O₂ adsorption configurations on BiOI-110 and BiOI-102. Red labels indicate the bond lengths (H-bonds and the Bi–O bond), and yellow dashed lines mark the corresponding interactions. d, e The dissociation energy (d) and diagram (e) of H₂O₂ bond cleavage on the BiOI-110 surface. Results shown in (c–e) were obtained from DFT calculations.

To mechanistically interpret the above experimental observations, DFT calculations were conducted to model and optimize H₂O₂ adsorption configurations on BiOI-110 and BiOI-102 (Fig. 5c and Supplementary Figs. 38, 39). On BiOI-110, H₂O₂ preferentially adsorbed via hydrogen-bond bridging between two H atoms of H₂O₂ and two interlayer I atoms of BiOI-110, with an adsorption energy (E_ads) of −0.97 eV. In contrast, the most favorable adsorption configuration on BiOI-102 featured coordination between the O atom of H₂O₂ and a surface Bi atom (–1.30 eV). The calculated results aligned well with the experimental observations, confirming the facet-dependent nature of H₂O₂ adsorption on BiOI surfaces.

Bond dissociation energy calculations further corroborated that H₂O₂ adsorption configurations governed activation pathways. As demonstrated in Fig. 5d, e, H₂O₂ adsorbed on the BiOI-110 surface exhibited thermodynamically unfavorable O–H bond cleavage, with a dissociation energy of 2.04 eV, whereas O–O bond cleavage was exothermic (–2.66 eV). It indicated that H₂O₂ on the BiOI-110 surface preferentially underwent O–O bond cleavage, facilitating I₂ generation, rather than O–H bond cleavage that would produce O₂^•−. In contrast, H₂O₂ adsorbed on BiOI-102 underwent spontaneous O–H bond cleavage (–0.51 eV), thereby enabling H₂O₂ activation toward O₂^•− production (Supplementary Fig. 40) ⁴⁹.

Taken together, BiOI-110 adsorbed H₂O₂ via a hydrogen-bond bridge configuration, promoting O–O bond cleavage and selective I₂ formation. The generated trace I₂ diffused into the solution phase to trigger 4-CP oxidation, yielding PhO^• as the key reactive intermediates. PhO^• then underwent C–C and C–O coupling for the generation of recoverable polymers (Fig. 6). In contrast, H₂O₂ adsorbed on BiOI-102 was predominantly through Bi–O coordination, which enabled O–H bond cleavage to generate O₂^•− (Supplementary Fig. 41). Thus, selective I₂ generation, critical for solution-phase polymerization, was achieved through a facet-programmed BiOI catalyst.

Fig. 6. Proposed mechanism for I₂ generation and solution-phase 4-CP polymerization in the BiOI-110/H₂O₂ system.

a H₂O₂ activation for I₂ generation via a hydrogen-bond bridging configuration that promotes O–O bond cleavage. b 4-CP polymerization in the solution phase mediated by diffused I₂.

Continuous pollutant removal and resource recovery via sustained I₂ generation

The practical applicability of the I₂-dominated Fenton-like system was evaluated through long-term pollutant removal tests. Although BiOI-110 retained its crystal structure, morphology, and band structure after reaction (Supplementary Figs. 42–44), 4-CP removal efficiency decreased dramatically over consecutive cycles. The efficiency fell to 62% in the second cycle and further declined to only 16% by the fifth cycle (Fig. 7a). XPS analysis confirmed that this performance deterioration originated from iodide consumption within BiOI-110, with iodine content decreasing from 6.0 wt% to 1.6 wt% (Supplementary Table 5).

Fig. 7. Application of the BiOI-110/H₂O₂ system in water treatment.

a Cyclic 4-CP removal and KI regeneration of the BiOI-110/H₂O₂ system. Reaction conditions: [4-CP]₀ = 50 μM, [catalyst] = 0.3 g L⁻¹, [H₂O₂] = 2 mM, [KI] = 5 g L⁻¹, initial pH = 7.0. b RhB removal in the continuous-flow reactor with and without KI-assisted catalyst regeneration. c Schematic of the continuous-flow reactor for 4-CP removal and polymerization product collection. d 4-CP removal performance of the continuous-flow reactor with KI-assisted catalyst regeneration. Reaction conditions in (b–d): [pollutant]₀ = 10 μM, [catalyst] ≈90 mg, [H₂O₂] = 2 mM, [KI] = 5 g L⁻¹, initial pH = 7.0. e Photograph of the self-Fenton system in ambient conditions. f H₂O₂ generation, RhB removal, and corresponding experimental conditions for the self-Fenton system. Reaction conditions: [RhB]₀ = 10 μM, [catalyst] ≈90 mg, [KI] = 5 g L⁻¹. The error bars in the figures represent the standard deviations (mean ± s.d., n = 3) from triplicate tests.

Inspired by the layered structure of BiOI, wherein [Bi₂O₂]²⁺ slabs are sandwiched between two iodide ions (I⁻) layers via nonbonding van der Waals interactions along the (001) direction^50,51, we hypothesized that exogenous I⁻ could refill iodine vacancies in used BiOI-110 through van der Waals interactions. If this hypothesis held, BiOI-110 could function as an iodine shuttle to sustain in situ I₂ generation from I⁻ and H₂O₂. To verify this, KI was employed as an iodide source to replenish lattice I sites in BiOI-110. After immersing the used BiOI-110 in KI solution for 60 min, the surface I atomic ratio recovered from 1.6% to 7.3%, accompanied by an increase in the proportion of lattice I⁻ from 91% to 93% (Supplementary Table 5 and Supplementary Fig. 45). As a result, the BiOI-110 catalyst subjected to five reaction cycles regained I₂ generation capability, with 4-CP removal efficiency increasing from 16% to 82% (Fig. 7a). It suggested that KI treatment closed the iodine loop, enabling I₂ generation from I⁻ and H₂O₂ through BiOI-110-driven Fenton-like reactions.

The iodine-loop-driven in situ I₂ generation was further corroborated using other bismuth oxyhalides, BiOCl-110 and BiOBr-110. As Supplementary Fig. 46, although BiOCl-110 and BiOBr-110 alone displayed negligible H₂O₂ activation performance, both systems exhibited a two-stage 4-CP degradation profile in the presence of I^‒. The initial lag phase corresponded to the formation of halide vacancies via redox leaching during H₂O₂ activation, followed by efficient 4-CP removal (55% and 48%, respectively) as I⁻ gradually occupied the vacancies to form active lattice I sites. In contrast, KI-pretreated spent BiOCl-110 and BiOBr-110, which contained halide vacancies, showed no lag phase during H₂O₂ activation, as lattice I sites were already established during KI treatment. These results highlighted the generality of BiOX catalysts as iodine shuttles for mediating sustaining in situ I₂ generation from I^‒ and H₂O₂.

The above findings motivated the construction of a continuous-flow reactor for sustainable pollutant removal based on in situ I₂ generation (Supplementary Fig. 47). To minimize the environmental risk associated with iodide release, KI was periodically supplied to replenish lattice I in BiOI-110, and the added KI was recycled throughout the experiment. BiOI-110 was immobilized on melamine sponges, preserving its catalytic activity (Supplementary Fig. 48). Reactor performance was evaluated by Rhodamine B (RhB) degradation at a hydraulic retention time of 20 min. As shown in Fig. 7b, without KI supplementation as an iodine source, initial near-complete RhB removal declined sharply to only 30% by the 8th cycle and stabilized thereafter. The decreased performance was attributed to lattice I depletion, which mirrored observations from batch experiments. In contrast, periodic KI supplementation maintained RhB removal above 95% over 15 consecutive cycles. This stability verified that the BiOI-110-driven Fenton-like system sustained in situ I₂ generation from H₂O₂ and KI, addressing the long-term stability limitation of traditional Fenton-like systems.

We then evaluated the capability of the in situ I₂ generation reactor for continuous solution-phase 4-CP polymerization and resource recovery. An adsorption column packed with XAD-2 resin was connected to the effluent for the polymeric products collection (Fig. 7c, Supplementary Fig. 49). With I⁻ replenishment, the reactor maintained stable 4-CP removal over 15 consecutive cycles, whereas 4-CP degradation rapidly declined in its absence (Fig. 7d, Supplementary Fig. 50). 4-CP polymerization products were successfully collected in the adsorption column, which exhibited profiles consistent with those obtained from batch experiments (Supplementary Fig. 51). Notably, the as-obtained oligomeric polyphenyl ethers and polyphenylenes can serve as comonomer modifiers for polyphenylene oxide (PPO) resin, one of the five major engineering plastics, to effectively enhance its processability and mechanical toughness⁵². These results illustrated that sustained in situ I₂ generation enabled continuous solution-phase polymerization of phenolic pollutants, facilitating product recovery while preventing catalyst deactivation.

Extending this concept, we developed a self-Fenton system by integrating a solar-driven H₂O₂ generation unit upstream of the BiOI-110 reactor (Fig. 7e, Supplementary Fig. 52). This configuration eliminated the need for external H₂O₂ dosing, reducing operation costs. In the H₂O₂ generation unit, pure water was exposed to natural sunlight under ambient conditions, with H₂O₂ produced in situ via a triphenyltriazine-alkynyl-anthraquinone (TPT-alkynyl-AQ) photocatalyst fabricated as an electrospun nanofiber membrane, as we previously reported⁵³. During a 2-h accumulation phase, approximately 0.6 mM H₂O₂ was produced. The effluent containing in situ-generated H₂O₂ and RhB was then directed into the BiOI-110-packed reactor for pollutant degradation. RhB was continuously and efficiently removed throughout daylight hours, with removal efficiency remaining nearly 100%. Remarkably, H₂O₂ concentration peaked at 1.2 mM by 16:00 and plateaued as sunlight intensity declined. Nevertheless, complete pollutant removal persisted even after sunset, owing to the residual H₂O₂ accumulated during the daytime (Fig. 7f). These findings demonstrated a fully self-sustained system that achieved long-term stable water treatment under ambient conditions. Operating solely on natural sunlight, water, KI, and a small amount of BiOI-110 catalyst, the system produced recoverable polymeric products while minimizing resource inputs. Such a configuration offers a cost-effective and environmentally benign strategy for scalable water purification with inherent resource recovery.

Discussion

This study introduces a facet-programmed approach that decouples polymerization from the catalyst surface and relocates it into the solution phase. Such spatial separation addresses the longstanding challenge of product-overlayer fouling in oxidation-driven water treatment. By programming the exposed facets of BiOI, we selectively activate H₂O₂ to generate molecular iodine, a freely diffusing and moderately oxidizing species that drives efficient solution-phase polymerization. This pathway differs from traditional surface-bound or radical-based oxidation processes that often result in catalyst deactivation and over-oxidation.

Our approach leads to 73% retention of polymer products in the aqueous phase, demonstrating high selectivity toward PhO^• generation for C–C/C–O coupling, along with negligible over-oxidation side reactions. SECCM, spectroscopic analyses, and DFT calculations confirm that I₂, generated through facet-controlled H₂O₂ adsorption configuration that facilitates O–O bond cleavage (ΔE = −2.66 eV) is the key reactive species. Additionally, the integration of KI-based iodine replenishment enables continuous resource recovery, closing the iodine loop and ensuring fouling-resistant performance over extended cycles.

The findings establish a “facet–species–phase” design rule, which decouples polymerization from the catalyst surface, offering a scalable, durable, and sustainable solution for continuous, recoverable water treatment. This work demonstrates an advanced oxidation processes that not only minimize catalyst fouling but also enhance the sustainability of water treatment systems by enabling continuous recovery of polymer products.

Methods

Chemicals and materials

Bismuth nitrate pentahydrate (Bi(NO₃)₃ ∙ 5H₂O), potassium iodide (KI), potassium chloride (KCl), potassium bromide (KBr), hydrogen peroxide (H₂O₂, 30 wt%), ethylene glycol, iodine (I₂), 4-chlorophenol (4-CP), furfuryl alcohol (FFA), methanol (MeOH), p-benzoquinone (p-BQ), tert-butanol (TBA), absolute ethanol, ferulic acid, acetonitrile, N,N-diethyl-p-phenylenediamine (DPD), peroxidase (POD), and sodium humate (HA) were purchased from Sinopharm Chemical Reagent Co. Ltd. 5,5-Dimethyl-1-pyrroline-N-oxide (DMPO, 98%) and 2,2,6,6-tetramethyl-4-piperidinyl (TEMP, 99%) were procured from Sigma-Aldrich Co. LLC. COD testing reagents (LH-YE-50 and LH-YD-50) were obtained from Lianhua Technology Co. Ltd. All reagents were of analytical grade used without further purification.

Catalyst preparation

BiOI-110 was synthesized by a facile solvothermal method using ethylene glycol as the solvent¹¹. Briefly, 20 mmol Bi(NO₃)₃ ∙ 5H₂O and 20 mmol KI were dissolved in 30 mL ethylene glycol with continuous stirring for 1 h. The resulting solution was then transferred into a 50 mL Teflon-lined stainless-steel autoclave and heated at 160 °C for 2 h. After cooling to room temperature, the product was filtered, thoroughly washed with ethanol and ultrapure water, and dried at 60 °C. BiOCl-110 and BiOBr-110 were prepared using the same procedure as BiOI-110, except that KCl and KBr were used in place of KI, respectively. BiOI with iodine vacancies (BiOI_1-x) was synthesized by calcining BiOI-110 in air at 350 °C for 2 h⁵⁴.

BiOI-102 was synthesized via a similar solvothermal method, using a 1:1 (v/v) ethanol/water mixture as the solvent and heated at 180 °C for 24 h.

BiOI-001 was prepared via a precipitation method. Specifically, 8 mmol Bi(NO₃)₃ ∙ 5H₂O and 8 mmol KI were dissolved in 70 mL ultrapure water. After aging at 80 °C for 2 h in a water bath, the resulting precipitates were collected by filtration, thoroughly washed, and dried.

Characterization

The crystalline phase of the catalyst was analyzed using XRD (D8 Advance Da Vinci, Bruker) with Cu Kα radiation (λ = 0.154 nm). XPS analysis was performed on an ESCALAB QXi unit (Thermo Scientific). SEM (JSM-7800F, JEOL) and TEM (JEM-2100F, JEOL) images were recorded to observe the morphologies of BiOI. TEM and HRTEM images were acquired in bright-field mode at an accelerating voltage of 200 kV. The BET specific surface area was measured using a gas sorption analyzer (ASAP 2460, Micromeritics). In situ Raman spectra were acquired at 532 nm using a Horiba LabRam HR Evolution Raman microscope. FTIR spectra were collected using a Bruker INVENIO spectrometer equipped with a diamond ATR accessory. The presence of reactive oxygen species was examined by a Bruker micro electron paramagnetic resonance spectrometer.

Catalytic activity evaluation

The catalytic activity of catalysts for H₂O₂ activation was assessed by 4-CP oxidation. In a typical experiment, 0.3 g L⁻¹ BiOI was dispersed into 50 mL of 4-CP solution (50 μM) and stirred until adsorption equilibrium was reached. Then the reaction was initiated by adding H₂O₂ (2 mM) at an initial pH of 7.0. At predetermined intervals, 1 mL samples were withdrawn for analysis. Each test, performed in triplicate, was presented as mean values with error bars depicting ±1 standard deviation.

Analytical methods

The concentration of organic pollutants was quantified by HPLC (Fuli LC5090) equipped with a NanoChrom ChromCore C18 column (5 μm, 4.6 × 250 mm) and a UV-vis detector. The mobile phase consisted of a binary mixture of 0.1 wt% aqueous phosphoric acid solution and methanol (Supplementary Table 6). The intermediates of 4-CP degradation were identified using an Acquity UPLC (Waters) coupled with a timsTOF mass spectrometer (Bruker). Residual H₂O₂ concentration was measured by a modified DPD-POD method⁵⁵. The contributions of different reactive species were calculated from rate constants following scavenger addition (Supplementary Text 1). COD of both the suspension (containing solution and catalyst) and the bulk solution was determined with COD reagents using a multi-parameter water quality analyzer (Lianhua 5B-6C(V10)). I₂ concentration was quantified based on its characteristic absorption at 460 nm using a UV-vis-NIR spectrophotometer (Hitachi UH4150), with the standard curve provided in Supplementary Fig. 27. The texture coefficients were calculated based on the Harris formula as illustrated in Supplementary Text 2^27,56.

Open-circuit potential measurement was performed in a three-electrode cell connected to an electrochemical workstation (CHI 660). A carbon rod, a catalyst-modified glassy carbon electrode, and an Ag/AgCl electrode were used as the counter, working, and reference electrodes, respectively. A KCl solution (20 mM) was used as the electrolyte. To prepare catalyst ink, a mixture of the catalyst (5 mg), isopropanol (570 µL), ultrapure water (400 µL), and Nafion solution (30 µL) was sonicated for 1 h to achieve uniform dispersion. Subsequently, the glassy carbon electrode was coated with catalyst ink (20 µL) and dried.

Scanning electrochemical cell microscopy

The SECCM experiments were conducted on a commercial scanning electrochemical probe system (HEKA Electronik GmbH, Supplementary Text 3). For the constant-contact CV mapping in SECCM, a single-barrel pipette containing 50 mM H₂O₂ and 20 mM KCl (supplemented with 10 mM 4-CP if needed) was approached to the glassy carbon substrate, forming meniscus contact at a series of predefined locations. Upon each landing, a local CV was performed within the confined area defined by the meniscus cell. The resulting current signals were digitally post-filtered using a 60 Hz notch filter in MATLAB to remove the line noise⁵⁷.

Continuous-flow reactor for in situ I₂ generation and water treatment

Melamine sponges (1.5 × 2 × 1.5 cm³) were immersed in 200 mL of aqueous BiOI-110 suspension (5 g L⁻¹). The mixture was sonicated for 20 min and then magnetically stirred for 30 min. The sponges were dried at 160 °C for 6 h to obtain the catalyst-loaded material. To minimize catalyst loss during continuous-flow operation, the obtained sponges were washed three times with ultrapure water and then dried at 80 °C. The final BiOI-110 loading on each sponge was approximately 18 mg.

Continuous-flow experiments involved pollutant oxidation and catalyst regeneration. In each cycle, a solution containing pollutant and H₂O₂ was passed through BiOI-110-loaded melamine sponges in the column reactor, with a hydraulic retention time of approximately 20 min. After a 30-min reaction, a 1 mL sample was withdrawn and filtered for analysis. Subsequently, the sponges were rinsed with a KI solution (5 g L⁻¹) at a flow rate of 2 mL min⁻¹ for 30 min, followed by a 5-min flush with ultrapure water before initiating the next cycle.

Theoretical calculations

All spin-polarized periodic calculations were carried out using the Vienna ab initio Simulation Package (VASP). Details of theoretical calculations were provided in Supplementary Text 4.

Author contributions

Y.W., Y.-X.Y., and Z.G. co-proposed the idea and designed the experiments. Y.W., Y.L., M.Li, and B.H. performed the related experiments. Y.W. and Y.L. analyzed the data. C.D. and M.Long conducted the theoretical calculations. Y.W. wrote the initial draft. Y.-X.Y., G.O., and Z.G. further modified the paper.

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

Z.G. discloses support for the research of this work from the National Natural Science Foundation of China (U24A20514). Y.W. discloses support for the research of this work from the National Natural Science Foundation of China (22506161), China Postdoctoral Science Foundation (2025M771287), and China Postdoctoral Fellowship Program (GZB20250327). Y.-X.Y. discloses support for the research of this work from the National Natural Science Foundation of China (22422611, 22206209) and Southern Marine Science and Engineering Guangdong Laboratory (Zhuhai) (SML2024SP003).

Data availability

The data supporting the findings of the study are included in the paper and its Supplementary Information. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-74369-2.