Molecular oxygen confinement ladder enables stable electro-Fenton treatment of nanofiltration concentrate Authors: Yufei Cheng, Mengzhuang Yang, Yuchen Hu, Peng Liu, Weichen Zeng, Xiaohang Han, Tianyi Wang, Langming Bai, Jinlong Wang, Han Zhang, Xiaoguang Duan, Heng Liang Published: 2026 ABSTRACT Clean water scarcity intensifies the need for advanced treatment of emerging contaminants in nanofiltration (NF) concentrates. Conventional electro-Fenton (EF) systems face unstable oxygen supply and electrode wetting in bubbleless aeration, limiting selective reactive oxygen species generation in high-salinity matrices. Here we show a molecular oxygen confinement ladder strategy by developing a conductive aeration membrane cathode in which boron nitride spontaneously forms wedge-shaped structures inside carbon nanotubes. This architecture achieves kinetic antiwetting and physical oxygen confinement at the nanoscale, accelerating Fe(II)/Fe(III) cycling and directing the pathway to dominant singlet oxygen (1O2) generation. Experiments and simulations demonstrate fivefold faster degradation kinetics, complete sulfamethoxazole removal within 15 min, and sustained high efficiency over 48 h of continuous operation, including a switch to real NF concentrate. The structure-oriented gas confinement approach thus offers a robust platform for selective and stable EF treatment of complex industrial wastewater. Introduction Nanofiltration (NF) technology has demonstrated high efficiency in the advanced purification of drinking water [1]. However, the resulting NF concentrates enrich various emerging contaminants (ECs), including antibiotics, pesticides, and endocrine disruptors, which may increase ecological risks upon discharge [[2], [3], [4]]. Electro-Fenton (EF) oxidation is a promising approach for degrading recalcitrant organic pollutants due to its high efficiency and environmental friendliness [5,6]. Nevertheless, conventional EF systems largely rely on bubble aeration for oxygen supply, which is limited by low gas–liquid mass-transfer efficiency and the poor solubility of O2 in water [7]. Moreover, aeration can trigger a stripping effect of volatile organic compounds and greenhouse gases, leading to secondary atmospheric pollution. To address these issues, bubbleless aeration has been proposed as an effective method for supplying oxygen [8,9]. This technology employs gas-permeable dense membranes (e.g., silicone rubber) or microporous membranes (e.g., hydrophobic polymeric membranes) to maintain gas pressure below the bubble point. O2 from air or pure oxygen is continuously supplied into the membrane lumen, while the liquid phase flows externally, driving O2 diffusion directly into the liquid phase via a concentration gradient. While bubbleless aeration achieves near-100% oxygen utilization and eliminates stripping risks, hydrophobic membranes face severe challenges due to pore wetting during long-term electrocatalytic operation [10]. The instability of the tri-phase interface under an electric field often leads to oxygen supply failure, restricting practical applications. In membrane science, wedge-shaped structures can enhance antiwetting performance. This specific geometry kinetically enhances surface hydrophobicity, firmly “locking” gas within the microscopic architecture to maintain a stable gas–liquid interface. However, existing research has primarily focused on constructing such structures on macroscopic membrane surfaces; how to implement and maintain effective antiwetting mechanisms within nanoscale catalytic-confined spaces remains unclear [11]. Carbon nanotubes (CNTs), classic confinement carriers, are frequently used to enhance reaction kinetics [[12], [13], [14], [15], [16], [17]]. Boron nitride (BN) lattices tend to spontaneously form wedge-shaped topologies when grown within confined spaces. Few-layer BN has been reported to activate hydrogen peroxide (H2O2) for pollutant degradation [18], with nitrogen atoms acting as Lewis base centers that interact with H2O2 to promote reactive oxygen species (ROS) generation via electron donation [19,20]. BN modification can also enhance electron-transfer efficiency and accelerate the Fe(II)/Fe(III) cycle [1,21,22]. However, these works mainly focus on regulating the “hydrophilic phase.” Whether the wedge-shaped structures formed by B–N within confined spaces can serve as anchoring units for a hydrophobic gas phase by leveraging kinetic antiwetting to achieve efficient oxygen enrichment and stable transport remains to be systematically investigated [23]. Here, we developed a “molecular oxygen confinement ladder” by inducing boron–nitrogen (B–N) wedge-shaped structures to grow within the interior of CNTs, yielding a BN-in-CNT conductive aeration membrane cathode. We hypothesized that the kinetic antiwetting behaviour of these internal wedge structures resists electrolyte intrusion and physically confines molecular oxygen, sustaining a locally oxygen-rich microenvironment that promotes selective singlet oxygen (1O2) generation and continuous Fe(II)/Fe(III) cycling [24]. To test this, we compared BN-in-CNT with BN-out-CNT under tunable oxygen-diffusion conditions and interpreted the electrochemical and degradation behaviour with density functional theory (DFT) and molecular dynamics (MD) simulations of oxygen transport and retention [25,26]. The confined architecture enabled sustained removal of ECs from real NF concentrate, driven predominantly by 1O2 with continuous iron cycling. This establishes a structure-oriented route to stable EF treatment of complex, high-salinity effluents. Materials and methods Chemicals and catalyst preparation Details on chemicals and materials used are presented in Supplementary Text S1. All reagents were analytical grade and used without further purification. The NF concentrate used in this study was prepared from Songhua River water using ultrafiltration and NF equipment, and its water-quality characteristics were analyzed (Supplementary Table S1). All aqueous solutions were prepared using Milli-Q ultrapure water. Building on previous studies of nanoconfinement in CNTs [23,25], we modified the reported method and fabricated BN-in-CNT and BN-out-CNT (Supplementary Fig. S1). The BN-in-CNT catalyst and its conductive aeration membrane cathode were synthesized by introducing a boron precursor into CNTs through ultrasonic-assisted capillary infiltration, followed by high-temperature calcination under nitrogen to form and confine BN inside the CNTs and minimize outer-surface nucleation through washing. The resulting BN-in-CNT was immobilized onto a polytetrafluoroethylene (PTFE) and polypropylene (PP) membrane using vacuum filtration, and BN-out-CNT was prepared for comparison. Details of the synthesis process are provided in Supplementary Text S2. Electrochemical experiments Electrochemical experiments were conducted in an acrylic reactor using a BN-in-CNT, BN-out-CNT, and CNT conductive aeration membrane as the cathode and a Pt sheet as the anode, with O2 continuously supplied through the membrane (Fig. 1a). The system contained 0.2 mmol L−1 FeSO4·7H2O and 0.05 mol L−1 Na2SO4 as the supporting electrolyte, and its pH was adjusted to 3.0 using 0.1 mol L−1 H2SO4. Detailed electrochemical measurements are provided in Supplementary Text S3, the characterization methods in Supplementary Text S4, and the analytical methods in Supplementary Text S5, Table S2, and Fig. S2.Fig. 1Configuration of the membrane oxygen-supply electro-Fenton system and characterization of boron nitride (BN)–carbon nanotube (CNT) materials. a, Schematic diagram of the membrane oxygen supply electro-Fenton (EF) system. b, Transmission electron microscopy (TEM) image showing the distribution of BN nanoparticles confined within the CNT channels (BN-in-CNT). White dashed circles indicate the BN nanoparticles. c, High-resolution transmission electron microscopy (HRTEM) image of BN-in-CNT. The inset shows lattice fringes assigned to the BN (003) and (101) planes. d, HRTEM image of boron nitride nanoparticles located outside the carbon nanotubes (BN-out-CNT). The inset shows lattice fringes assigned to the BN (003) plane. e, Particle-size distributions of BN nanoparticles in BN-in-CNT and BN-out-CNT. Bars show the particle-size distribution histograms; solid curves are the corresponding Gaussian fits. f–g, B 1s X-ray photoelectron spectroscopy (XPS) spectra of BN-in-CNT (f) and BN-out-CNT (g). Percentages denote the relative peak-area fractions of the B–N and B–O components from B 1s XPS deconvolution. h, X-ray diffraction (XRD) patterns of BN-in-CNT, BN-out-CNT, and CNTs. The yellow shaded regions highlight the weak diffraction peaks at 28.0° and 40.5°, assigned to the (003) and (101) planes of BN, respectively. i, N2 adsorption–desorption isotherms of BN-in-CNT and BN-out-CNT with the Brunauer–Emmett–Teller (BET) specific surface area (SBET) indicated. Configuration of the membrane oxygen-supply electro-Fenton system and characterization of boron nitride (BN)–carbon nanotube (CNT) materials. a, Schematic diagram of the membrane oxygen supply electro-Fenton (EF) system. b, Transmission electron microscopy (TEM) image showing the distribution of BN nanoparticles confined within the CNT channels (BN-in-CNT). White dashed circles indicate the BN nanoparticles. c, High-resolution transmission electron microscopy (HRTEM) image of BN-in-CNT. The inset shows lattice fringes assigned to the BN (003) and (101) planes. d, HRTEM image of boron nitride nanoparticles located outside the carbon nanotubes (BN-out-CNT). The inset shows lattice fringes assigned to the BN (003) plane. e, Particle-size distributions of BN nanoparticles in BN-in-CNT and BN-out-CNT. Bars show the particle-size distribution histograms; solid curves are the corresponding Gaussian fits. f–g, B 1s X-ray photoelectron spectroscopy (XPS) spectra of BN-in-CNT (f) and BN-out-CNT (g). Percentages denote the relative peak-area fractions of the B–N and B–O components from B 1s XPS deconvolution. h, X-ray diffraction (XRD) patterns of BN-in-CNT, BN-out-CNT, and CNTs. The yellow shaded regions highlight the weak diffraction peaks at 28.0° and 40.5°, assigned to the (003) and (101) planes of BN, respectively. i, N2 adsorption–desorption isotherms of BN-in-CNT and BN-out-CNT with the Brunauer–Emmett–Teller (BET) specific surface area (SBET) indicated. Theoretical calculation method This study systematically investigated the interaction mechanisms between BN-in-CNT and BN-out-CNT and pollutants, as well as the spatial distribution and dynamic behavior of ROS, using DFT, finite element, and MD simulations. DFT calculations were performed with CP2K, Gaussian, and Multiwfn to evaluate adsorption and free energies [27,28]; COMSOL simulated the distribution of •OH and 1O2 in microchannels; and GROMACS was employed to study nanoscale dynamics (Supplementary Texts S6–S8). Results and discussion Structural and morphological characterization of BN-in-CNT Transmission electron microscopy (TEM) images demonstrated that the confined growth of adjacent BN nanoparticles was not a simple aggregation. Instead, a wedge-shaped structure spontaneously formed by leveraging the curvature of the CNT walls (white dashed circles, Fig. 1b; Supplementary Fig. S3–S4). This continuous geometric configuration, formed by rigid BN particles and the CNT inner walls, created a microscopic ladder of molecular-oxygen confinement. This architecture provides stable physical gas traps for oxygen molecules and imparts kinetic antiwetting capabilities to the channels due to its specific capillary geometry, thereby resisting electrolyte intrusion into the gas-phase pathways. In BN-in-CNT, BN nanoparticles were orderly distributed along the internal channels of the CNTs (Fig. 1c), whereas in BN-out-CNT, they were randomly dispersed on the external surfaces (Fig. 1d). Lattice fringes corresponding to the crystal plane of BN nanoparticles are shown (insets, Fig. 1c and d). The particle size distribution showed average diameters of 18.1 ± 1.3 nm for BN-in-CNT and 9.9 ± 1.2 nm for BN-out-CNT (Fig. 1e). The relatively larger size of BN nanoparticles within the CNT channels can be attributed to the spatial confinement effect of the inner walls—which reduces the number of random nucleation sites and restricts the outward diffusion of precursors, thereby promoting oriented growth and particle fusion [29,30]. In addition, structural defects were observed on both the inner and outer walls of the CNTs in the BN-in-CNT system (Supplementary Fig. S5). These defects may have originated from internal stress and lattice mismatch introduced during the BN encapsulation process or from damage to the CNT outer walls caused by external thermal treatment and reaction processes [31,32]. We characterized the chemical composition of the catalysts by X-ray photoelectron spectroscopy (Fig. 1f and g). The results indicated that the proportion of B–N bonds in BN-in-CNT reached 80.3%, higher than the 24.2% observed in BN-out-CNT. Analysis of the C 1s spectra (Supplementary Fig. S6) also revealed structural modulations, including variations in π–π∗ transition features, functional group distributions, and sp2 carbon structures. The X-ray diffraction patterns show characteristic diffraction peaks of CNTs at (002) and (100) positions for both BN-in-CNT and BN-out-CNT, indicating the maintenance of a typical hexagonal graphite structure (Fig. 1h). Weak diffraction peaks at 28.0° and 40.5° correspond to the (003) and (101) planes of BN, consistent with the X-ray photoelectron spectroscopy findings. The more pronounced BN diffraction peaks in BN-in-CNT align with the larger average particle size and the confinement-induced growth mechanism within the CNT channels. We employed Raman spectroscopy to examine the interactions between the encapsulated BN nanoparticles and the CNTs [[33], [34], [35]]. The intensity ratio of the D to G bands (ID/IG) characterizes the degree of graphitic disorder caused by structural defects and heteroatom doping [36]. The higher ID/IG value for BN-in-CNT suggests that BN encapsulation reduces graphitization and increases defect density, thereby enhancing catalytic activity (Supplementary Fig. S7). The characteristic E2g peak of BN [37], which arises from the in-plane optical phonon vibration of boron and nitrogen atoms within the basal plane, is overlapped by the stronger G-band signal in both BN-out-CNT and BN-in-CNT (Supplementary Fig. S8). BN-in-CNT and BN-out-CNT exhibited distinct N2 adsorption--desorption behaviors and pore size distributions (Fig. 1i; Supplementary Fig. S9). Based on the Barrett–Joyner–Halenda method, the calculated pore volumes (Supplementary Table S3) were 82.7 cm3 g−1 for BN-out-CNT and 94.9 cm3 g−1 for BN-in-CNT. The sharp rise in the isotherm of BN-in-CNT at relative pressure (P/P0) > 0.8 confirms the formation of effective wedge-shaped internal cavities between the core BN particles and the CNT shells [38]. This unblocked internal void demonstrates that the molecular oxygen confinement ladder provides sufficient physical space to accommodate and enrich oxygen, thereby serving as a critical microenvironment for efficient gas-confined reactions. Consequently, the synergy of CNT condensation and spatial confinement leads to distinct differences between internal and external doping, resulting in varied B-doping levels and compositions in BN-in-CNT compared to the external-doping system [36]. Electrochemical performances of the cathode materials Time-resolved in situ Raman monitoring spectroscopy was used to confirm the capture and activation mechanism of oxygen by the molecular oxygen confinement ladder at the molecular level. As the electrocatalytic reaction progressed, the BN-in-CNT system exhibited a gradual emergence and intensification of a characteristic peak in the 1500–1580 cm−1 range (Fig. 2a). This peak corresponds to the O–O stretching vibration mode of the key reaction intermediate, ∗OOH. The continuous accumulation of signal intensity over time directly reflects the prolonged residence of oxygen species within the confined space. Rather than escaping rapidly, the diffusing oxygen molecules are progressively captured and locked by the internal wedge-shaped structures, thereby continuously enriching the reaction microzone with high-concentration active intermediates. In contrast, no ∗OOH characteristic signals were detected in the BN-out-CNT system (Fig. 2b), which lacks the internal confinement architecture. This distinct temporal evolution provides kinetic evidence that the wedge-shaped structure in BN-in-CNT serves as an effective gas trap, confining oxygen molecules within the reaction microenvironment and enabling efficient catalysis.Fig. 2Electrochemical properties and oxygen-reduction selectivity of boron nitride (BN)–carbon nanotube (CNT) materials. a–b, Raman spectra of the BN-in-CNT (a) and BN-out-CNT (b) recorded at 0, 1, 2, 3, 5 min. Spectra are arranged from bottom to top in order of increasing reaction time and are vertically offset for clarity. Colored curves correspond to successive reaction times; the dashed box marks the 1500–1580 cm−1 region of the ∗OOH intermediate. c, Cyclic voltammograms of BN-in-CNT, BN-out-CNT, and CNTs in O2-saturated 0.1 mol L−1 Na2SO4 solution at a scan rate of 10 mV s−1. The red dashed line marks the cathodic O2-reduction peak of BN-in-CNT at −0.6 V versus the reversible hydrogen electrode (RHE). d, Electrochemical impedance spectroscopy Nyquist plots; the inset shows the equivalent circuit used for fitting, where Rs is the solution resistance, Rct the charge-transfer resistance, and CPE the constant-phase element. e, Linear sweep voltammograms (LSV) of the BN-in-CNT, BN-out-CNT, and CNTs. f, Tafel plots derived from the LSV in panel e, with the fitted Tafel slopes indicated. g, Faradaic efficiency for hydrogen peroxide (H2O2) production calculated from the ring and disk currents. h, Electron-transfer number derived from rotating ring-disk electrode (RRDE) measurements. Error bars represent standard deviations from n = 3 independent replicates. Electrochemical properties and oxygen-reduction selectivity of boron nitride (BN)–carbon nanotube (CNT) materials. a–b, Raman spectra of the BN-in-CNT (a) and BN-out-CNT (b) recorded at 0, 1, 2, 3, 5 min. Spectra are arranged from bottom to top in order of increasing reaction time and are vertically offset for clarity. Colored curves correspond to successive reaction times; the dashed box marks the 1500–1580 cm−1 region of the ∗OOH intermediate. c, Cyclic voltammograms of BN-in-CNT, BN-out-CNT, and CNTs in O2-saturated 0.1 mol L−1 Na2SO4 solution at a scan rate of 10 mV s−1. The red dashed line marks the cathodic O2-reduction peak of BN-in-CNT at −0.6 V versus the reversible hydrogen electrode (RHE). d, Electrochemical impedance spectroscopy Nyquist plots; the inset shows the equivalent circuit used for fitting, where Rs is the solution resistance, Rct the charge-transfer resistance, and CPE the constant-phase element. e, Linear sweep voltammograms (LSV) of the BN-in-CNT, BN-out-CNT, and CNTs. f, Tafel plots derived from the LSV in panel e, with the fitted Tafel slopes indicated. g, Faradaic efficiency for hydrogen peroxide (H2O2) production calculated from the ring and disk currents. h, Electron-transfer number derived from rotating ring-disk electrode (RRDE) measurements. Error bars represent standard deviations from n = 3 independent replicates. This structure-induced oxygen enrichment effect is directly mapped onto the macroscopic electrochemical performance. Cyclic voltammetry showed that BN-in-CNT exhibited the highest electrochemical response (Fig. 2c). The catalytic activity for the oxygen reduction reaction (ORR) was enhanced when BN was encapsulated within the CNTs compared to BN-out-CNT. In the O2-saturated electrolyte, a reduction peak appeared at −0.6 V versus the reversible hydrogen electrode (RHE) during the cathodic scan, whereas this peak completely disappeared under an N2 atmosphere (Supplementary Fig. S10) [39]. Conversely, no significant reduction peaks were observed in the cyclic voltammetry curves of pristine CNTs, indicating that the CNTs themselves contributed minimally to the ORR. These disparities demonstrate that BN-in-CNT's electrochemical activity stems from the high-concentration local oxygen-rich microenvironment established by the molecular oxygen confinement ladder, which facilitates the ORR. Furthermore, the Nyquist arc radius of BN-in-CNT, fitted using an equivalent circuit model, was smaller than those of pristine CNTs and BN-out-CNT (Fig. 2d). This phenomenon is attributable not only to the inherent conductivity of the material but also to the kinetic antiwetting properties imparted by the wedge-shaped structure. Leveraging its specific geometry, this structure effectively prevents excessive electrolyte infiltration into the gas-phase channels during long-term operation. This maintains the long-term stability of the gas–liquid–solid tri-phase interface, thereby reducing interfacial transport resistance and accelerating electron-transfer kinetics [40]. The ORR activity of different cathode materials was further evaluated using linear sweep voltammetry in an O2-saturated electrolyte at a scan rate of 10 mV s−1 (Supplementary Texts S9–S10). The BN-in-CNT conductive aeration membrane cathode exhibited the highest ORR performance among the tested materials, characterized by the most positive onset potential and the highest cathodic current density (Fig. 2e). We conducted Tafel analysis to further compare the ORR kinetic characteristics. The Tafel slope of the BN-in-CNT cathode was 61.1 mV dec−1 (Fig. 2f), lower than those of the control materials, indicating faster reaction kinetics. These results collectively demonstrate the distinct advantages of the BN-in-CNT cathode in both ORR activity and reaction kinetics (Supplementary Table S4). We conducted rotating ring-disk electrode (RRDE) tests to reveal the ORR pathways and H2O2 generation capacity of the different catalysts. BN-in-CNT exhibited substantially higher H2O2 selectivity (Fig. 2g and Supplementary Fig. S11)—with a Faradaic efficiency (FE) of approximately 91.7%, markedly exceeding those of BN-out-CNT (approximately 60.9%) and pristine CNTs (approximately 25.0%). Meanwhile, the electron transfer number for BN-in-CNT remained stable at approximately 2.2 across the tested potential range (Fig. 2h), indicating that it primarily follows a two-electron ORR pathway for H2O2 generation. In contrast, the electron transfer number for BN-out-CNT and CNTs was approximately 2.8 and 3.5, respectively, suggesting a partial shift toward the four-electron reduction pathway. Fenton-like catalytic performance We used the oxygen-encapsulation cathode systems, constructed from PTFE, CNT, BN-out-CNT, and BN-in-CNT conductive aeration membranes, to systematically evaluate the Fenton-like catalytic performance of different cathode materials. In the BN-in-CNT system, the removal efficiency of sulfamethoxazole (SMX) reached 90.3% within 9 min and achieved complete degradation (100%) at 15 min, outperforming the other three systems (PTFE (43.6%), CNT (45.9%) and BN-out-CNT (86.9%), respectively; Fig. 3a). Concurrently, the apparent reaction rate constant of the BN-in-CNT system was 5.5 times higher than that of the CNT system (Supplementary Text S11). This performance enhancement validates the role of the wedge-shaped structures as gas traps that create a high-concentration oxygen microenvironment within the nanoconfined space. This allows reactants to be physically confined near the active sites, thereby accelerating the reaction process.Fig. 3Reactive oxygen species (ROS) and iron cycling during sulfamethoxazole degradation. a, Sulfamethoxazole (SMX) degradation in various reaction systems. b, Cyclic stability performance test of BN-in-CNT (top) and BN-out-CNT (bottom) over nine consecutive cycles, continuing from cycle 1 in panel a. Colors and symbols distinguish cycles 2–10 from left to right; each cycle lasted 15 min and the time axis is reset at the start of every cycle. Panels a and b together comprise 10 consecutive cycles and 150 min of cumulative operation. c, Boron leaching from BN-in-CNT and BN-out-CNT during repeated SMX-degradation cycles. d, Effects of reactive-oxygen-species quenchers on SMX degradation in the BN-in-CNT and BN-out-CNT systems. Quenchers: furfuryl alcohol (FFA), a scavenger of singlet oxygen (1O2) and hydroxyl radicals (•OH); tert-butanol (TBA), an •OH scavenger; L-histidine (L-His), a 1O2 scavenger; and catalase (CAT), a hydrogen peroxide (H2O2) scavenger. Bars show SMX degradation with each quencher; horizontal lines indicate the quencher-free controls. e, Electron paramagnetic resonance (EPR) spectra for •OH detection using 5,5-dimethyl-1-pyrroline N-oxide (DMPO) as the trapping agent. f, EPR analysis for 1O2 detection with 2,2,6,6-tetramethyl-4-piperidinol (TEMP) as the trapping agent. Stars denote the characteristic peaks of the DMPO-•OH adduct in panel e and the TEMP-1O2 adduct in panel f. g, SMX degradation under N2 purge or without Fe2+. h, Temporal changes in adsorbed iron and dissolved Fe2+ and Fe3+ during iron redox cycling in the CNT, BN-out-CNT and BN-in-CNT systems. i–j, Time-dependent SMX degradation in the BN-in-CNT, BN-out-CNT, and broken-CNT systems in the absence or presence of L-His (i) or TBA (j). Connected symbols indicate degradation without quencher, and bars indicate degradation in the presence of the corresponding quencher. Broken CNTs denote CNTs opened at both ends by nitric acid reflux, used as a control that lacks the continuous confinement channels of intact CNTs. Error bars represent standard deviations from n = 3 independent replicates. Reactive oxygen species (ROS) and iron cycling during sulfamethoxazole degradation. a, Sulfamethoxazole (SMX) degradation in various reaction systems. b, Cyclic stability performance test of BN-in-CNT (top) and BN-out-CNT (bottom) over nine consecutive cycles, continuing from cycle 1 in panel a. Colors and symbols distinguish cycles 2–10 from left to right; each cycle lasted 15 min and the time axis is reset at the start of every cycle. Panels a and b together comprise 10 consecutive cycles and 150 min of cumulative operation. c, Boron leaching from BN-in-CNT and BN-out-CNT during repeated SMX-degradation cycles. d, Effects of reactive-oxygen-species quenchers on SMX degradation in the BN-in-CNT and BN-out-CNT systems. Quenchers: furfuryl alcohol (FFA), a scavenger of singlet oxygen (1O2) and hydroxyl radicals (•OH); tert-butanol (TBA), an •OH scavenger; L-histidine (L-His), a 1O2 scavenger; and catalase (CAT), a hydrogen peroxide (H2O2) scavenger. Bars show SMX degradation with each quencher; horizontal lines indicate the quencher-free controls. e, Electron paramagnetic resonance (EPR) spectra for •OH detection using 5,5-dimethyl-1-pyrroline N-oxide (DMPO) as the trapping agent. f, EPR analysis for 1O2 detection with 2,2,6,6-tetramethyl-4-piperidinol (TEMP) as the trapping agent. Stars denote the characteristic peaks of the DMPO-•OH adduct in panel e and the TEMP-1O2 adduct in panel f. g, SMX degradation under N2 purge or without Fe2+. h, Temporal changes in adsorbed iron and dissolved Fe2+ and Fe3+ during iron redox cycling in the CNT, BN-out-CNT and BN-in-CNT systems. i–j, Time-dependent SMX degradation in the BN-in-CNT, BN-out-CNT, and broken-CNT systems in the absence or presence of L-His (i) or TBA (j). Connected symbols indicate degradation without quencher, and bars indicate degradation in the presence of the corresponding quencher. Broken CNTs denote CNTs opened at both ends by nitric acid reflux, used as a control that lacks the continuous confinement channels of intact CNTs. Error bars represent standard deviations from n = 3 independent replicates. The BN-in-CNT conductive aeration membrane system demonstrated versatility under various operating conditions. The oxidation kinetics were effectively enhanced with increasing oxygen supply rates and current densities; however, excessive BN loading or catalyst dosage reduced SMX degradation efficiency (Supplementary Fig. S12). The BN-in-CNT system exhibited only a slight inhibitory effect in the presence of various competitive anions—indicating that the nanoconfined electrode possesses anti-interference capabilities, supporting the selective removal of ECs in real water matrices. Cyclic stability tests provided evidence for the kinetic antiwetting effect of the wedge-shaped structures. After 10 repetitive cycles, the SMX degradation efficiency in the BN-out-CNT system dropped to 40.7% (Fig. 3b). This rapid attenuation is attributed to the lack of an internal physical hydrophobic barrier. As the reaction progressed, the electrolyte gradually invaded the unprotected gas transport channels via electrowetting, leading to electrode flooding and cutting off the oxygen supply. In contrast, the BN-in-CNT system maintained a degradation efficiency of nearly 100%. This demonstrates that the wedge-shaped structures spontaneously formed by BN within the CNT channels establish a kinetic antiwetting barrier, preventing electrolyte infiltration into the internal gas-phase pathways and ensuring the long-term stability of the gas–liquid–solid tri-phase interface. After 150 min of operation, the concentrations of leached boron in the BN-in-CNT and BN-out-CNT systems were 0.006 and 0.02 mg L−1, respectively (Fig. 3c), both of which are well below the World Health Organization drinking water standard (0.5 mg L−1) and the European Union directive limit (1 mg L−1) [41]. Identification of ROS in Fenton-like reactions To elucidate the ROS involved in SMX degradation, we conducted comparative experiments with various radical scavengers (Fig. 3d and Supplementary Fig. S13–S14). In both BN-in-CNT and BN-out-CNT systems, the removal efficiency of SMX decreased to 30–50% upon the addition of tert-butanol (•OH scavenger) and furfuryl alcohol (FFA), a 1O2 scavenger. This initially suggested the simultaneous involvement of •OH and 1O2. However, as the reaction rate constant of FFA with •OH is comparable to that with 1O2 (1.2 × 108 L mol−1 s−1), the inhibitory effect of FFA alone can only suggest the potential participation of 1O2 rather than confirming its specific contribution, while •OH also plays a key role in ROS generation or pollutant oxidation. To further verify the role of 1O2, we employed L-histidine (L-His), a more specific scavenger for 1O2. Unlike FFA, L-His selectively quenches 1O2 via its imidazole group, thus exhibiting stronger inhibition when 1O2 is the dominant ROS. Upon adding L-His, the SMX degradation efficiencies in the BN-in-CNT and BN-out-CNT systems dropped sharply to 6.9% and 9.5%, respectively. This result confirms the dominance of the nonradical species 1O2 and indicates that the wedge-shaped structure within BN-in-CNT modulates the ROS generation pathway, favoring nonradical oxidation. The addition of catalase (an H2O2 scavenger) reduced SMX removal by 34.8% and 27.3% in the BN-in-CNT and BN-out-CNT systems, respectively, indicating that H2O2 also synergistically participates in the degradation process (Supplementary Fig. S15). To further confirm the involvement of 1O2 and •OH, electron paramagnetic resonance (EPR) spectroscopy was performed with 5,5-dimethyl-1-pyrroline-N-oxide (DMPO) and 2,2,6,6-tetramethyl-4-piperidinol (TEMP) as trapping agents for •OH and 1O2, respectively (Fig. 3e and f). No significant signals were observed with the CNT cathode, indicating minimal ROS generation in the absence of BN. However, replacing the CNT cathode with BN-out-CNT or BN-in-CNT enhanced the 1O2 signals, with corresponding concentrations of 3.3 × 10−5 and 4.0 × 10−5 mol L−1, respectively. The concentration advantage of BN-in-CNT is attributed to the kinetic antiwetting effect of the wedge-shaped structure, which maintains a high-concentration gas-phase microenvironment. This allows oxygen molecules to continuously enter and be converted into high-concentration 1O2 without the rapid loss of intermediates or solvent quenching typical of the open environment in BN-out-CNT. In contrast, •OH signals were relatively weak in both BN-modified systems. FFA probe degradation experiments (Supplementary Fig. S16) further corroborated this conclusion after excluding •OH interference. These results do not imply that 1O2 exists only in the BN-in-CNT system but rather indicate that nanoconfinement primarily enhances the flux of ROS conversion from •OH and superoxide radical (O2•−) precursors to 1O2. In the BN-out-CNT system, the absence of wedge-shaped physical restrictions within the pores allows the precursors to readily diffuse into the bulk solution, leading to reduced conversion flux and a lower absolute 1O2 concentration. Conversely, in the BN-in-CNT system, the efficient mass transfer and antiwetting properties of the nanocanyons promote the effective conversion of intermediates (such as •OH) into 1O2, thereby making 1O2 the primary ROS driving SMX degradation. Although 1O2 is the dominant ROS responsible for SMX degradation, the role of •OH in the system should not be overlooked. The •OH concentration in the BN-in-CNT system was slightly higher than that in the BN-out-CNT system (1.3 × 10−5 versus 8.2 × 10−6 mol L−1). Through analysis of intermediate products, we investigated the degradation pathways of SMX and validated the reaction mechanism (Supplementary Text S12, Supplementary Table S5 and Fig. S17–S24). Combined with the competitive quenching kinetics and liquid chromatography–mass spectrometry (LC-MS) analysis (see Supplementary Table S6–S7 and Fig. S25–S28 for details), the steady-state •OH signal does not independently dominate the elimination of the contaminant. Beyond degradation efficiency, the total organic carbon removal after 120 min indicated that BN-in-CNT outperformed BN-out-CNT (40.1% versus 20.7%; Supplementary Fig. S29). This further demonstrates that the stable gas–liquid–solid tri-phase interface maintained by the antiwetting wedge-shaped configuration not only achieves efficient contaminant degradation but also constructs a continuous high-energy oxidation environment through the selective attacks dominated by 1O2 and deep reactions assisted by •OH. This high-energy oxidation environment promotes the deep mineralization of the products rather than simple physical adsorption, which contributed only 6.3%, establishing a solid foundation for the deep treatment of ECs in complex water matrices. As a nonradical electrophilic oxidant, 1O2 selectively targets electron-rich organic moieties while exhibiting negligible reactivity toward common inorganic anions (e.g., Cl− and HCO3−) that readily scavenge •OH to form weaker secondary radicals, fundamentally underpinning the anti-interference capability of this system [42,43]. The theoretical simulation results presented below, along with the consistent SMX removal efficiency observed in diverse ionic backgrounds and real NF concentrate, further validate this selective advantage. Fe cycling efficiency and its mechanism In the absence of Fe(II), the degradation efficiency of SMX in the BN-in-CNT and BN-out-CNT systems decreased to 31.3% and 40.6%, respectively (Fig. 3g). Consequently, the iron cycling process in these systems was further investigated to elucidate the contribution of internal BN within CNTs to the Fe(II)/Fe(III) redox cycle and its accelerating mechanism (Fig. 3h). The iron cycling efficiency of the BN-in-CNT system primarily stems from the electronic modulation of the B–N bonds. Unlike the chemically inert B–O bonds [44], the nitrogen atoms in the confined B–N bonds act as Lewis base electron donors [18,45]. This enables the Fe(II) concentration in the BN-in-CNT system to remain at 0.9 mg L−1 after 60 min of operation, whereas Fe(II) ions were almost undetectable in the BN-out-CNT system. Quantitative gradient experiments revealed that the iron cycling efficiency and SMX degradation efficiency exhibit a consistent trend with varying B–N contents, confirming that an optimal B–N site density effectively regulates the Fe(III)/Fe(II) redox kinetics (Supplementary Fig. S30 and Table S8–S10). However, chemical effects alone are insufficient to explain the performance disparity, which also reflects the physical microenvironment created by the molecular-oxygen confinement ladder. The wedge-shaped structure leverages geometric confinement to force Fe(III) into intimate contact with the enriched O2 within the nanoslits, thereby increasing the effective collision frequency and enhancing the efficiency of ferrous cycling. The confined electric field further amplifies the electron distribution differences induced by the B–N bonds [46], effectively lowering the energy barrier for electron transfer from the CNTs to Fe(III) and providing sufficient precursors for continuous 1O2 generation [47]. Moreover, the wedge-shaped structure enhances cycling stability through physical anchoring. After cleaning the used BN-in-CNT and BN-out-CNT materials with ethanol, the eluate from BN-out-CNT was darker in color (Supplementary Fig. S31), indicating the presence of more loosely bound or aggregated Fe species on its surface. These results suggest that the B–N sites within BN-in-CNT effectively inhibit Fe(III) aggregation and precipitation through dual geometric and polarization effects, thereby improving the stability of the iron redox cycle. This conclusion is further supported by TEM images of the catalysts after use (Supplementary Fig. S32). Evidence of oxygen confinement based on ROS quenching and oxygen diffusion control In conventional EF systems, continuous O2 supply enables electrogenerated H2O2 to react with dissolved Fe(II), producing •OH in the bulk solution [48,49]. Consequently, this study further investigated the generation of 1O2. The catalytic performance under membrane aeration with O2 was higher than that under continuous N2 purging (Fig. 3g), confirming that O2 is the critical precursor for 1O2 generation. In the N2-purging experiment, however, the degradation efficiency of the BN-out-CNT system was higher than that of BN-in-CNT. This phenomenon indirectly corroborates the notion that the molecular oxygen confinement ladder exhibits distinct directional transport characteristics via physical encapsulation. Under N2 purging, only trace O2 generated at the anode exists in the system, diffusing toward the cathode driven by a concentration gradient. BN exposed on the external surface can more easily capture this free oxygen from the bulk solution. Conversely, the internal wedge-shaped structures of BN-in-CNT feature kinetic antiwetting properties and a specific geometry. These antiwetting properties and the specific wedge geometry characteristics create substantial steric hindrance, effectively preventing oxygen from diffusing from the bulk solution into the channels. This further validates the effectiveness of the confined structure in inhibiting the outward escape of internal oxygen under normal membrane aeration conditions. This conclusion is supported by consistent results obtained when the aeration site was moved from the cathode membrane side to a position near the anode (Supplementary Fig. S33). To distinguish the contributions of physical structure and chemical catalysis, we introduced a broken CNT conductive aeration membrane cathode as a control (Fig. 3i). Quenching experiments indicated that 1O2 remained the dominant ROS in this system, suggesting that the confinement effect of the CNT channels may alter the reaction energy barriers and pathways for 1O2, rather than relying solely on the catalyst itself. When the •OH-mediated 1O2 generation pathway was blocked by tert-butanol, the performance of the broken CNT system surpassed that of BN-out-CNT (Fig. 3j), indicating that the contribution of oxygen encapsulation outweighs the catalytic effect of BN external to the CNTs. The BN-in-CNT system, integrating the dual advantages of physical confinement reinforced by wedge structures and chemical catalysis, exhibited the highest catalytic performance. The variations in dissolved oxygen (DO) concentration during the reaction (Supplementary Fig. S34) further support this view. Furthermore, pulsed O2-supply experiments showed that, after the oxygen supply was cut off at 15 min, the DO and H2O2 concentrations in the BN-out-CNT system rapidly plateaued (Supplementary Fig. S35). In contrast, both parameters continued to rise in the BN-in-CNT system, indicating that the enriched O2 within the wedge-shaped cavity can undergo a sustained slow release even after the gas supply is halted. Concurrently, the O2 flow-rate dependence experiments demonstrated that the degradation efficiency of the BN-in-CNT system varied by only about 20% between flow rates of 4.0 and 0.5 mL min−1, a variance smaller than that observed in the BN-out-CNT system (Supplementary Fig. S36). This underscores a substantially reduced dependence of its catalytic activity on external O2 input. Furthermore, the low-pressure desorption residue of BN-in-CNT reached as high as 40.9%, which was approximately 1.7 times that of BN-out-CNT (Supplementary Fig. S37). Moreover, BN-in-CNT exhibited an open H2-type hysteresis loop, consistent with a pore structure comprising narrow necks and wide cavities. Together with the pulse-experiment results, this observation supports sustained release of confined O2 in terms of both the release extent and the proposed mechanism. At the 3 min mark, the catalytic performance of the BN-in-CNT system was higher than that of BN-out-CNT because of the instantaneous response mechanism of the wedge-shaped structure (Fig. 3j). As O2 traverses the membrane interface, it is immediately trapped by the internal wedge-shaped ladder and converted in situ into 1O2, bypassing diffusion and re-adsorption into the bulk solution. In contrast, the BN-out-CNT system relies on the gradual accumulation of DO to establish a concentration gradient before exerting its catalytic effect. Although the broken CNT system possesses physical channels, it lacks antiwetting guiding structures. Consequently, O2 primarily accumulates at the interface between the membrane and electrode and cannot effectively overcome the diffusion resistance to penetrate deeper channels. Fe(II) concentration also remained sufficient within 10 min (Fig. 3h), ruling out Fe(II) deficiency as a limiting factor. The interference of surface chemical changes introduced during the broken CNT preparation was excluded in evaluating the respective contributions of physical channels and BN chemical catalysis (Supplementary Fig. S38–S39, Table S11, and related discussion). These results confirm that oxygen diffusion and capture efficiency are the factors for early-stage performance disparities. By 15 min, DO concentration rose, and the differences in degradation efficiency among the systems gradually narrowed. Oxygen activation and thermodynamic advantages of the wedge-shaped confined interface DFT calculations revealed the essence of the molecular oxygen confinement ladder from the perspectives of energy and electronic structure. The results indicated that the adsorption energies (Eads) of O2 in CNT, BN-in-CNT, and BN-out-CNT systems were −0.3, −1.8, and −0.9 eV, respectively (Fig. 4a). The wedge-shaped confined structure enhances the thermodynamic stability and residence probability of O2 within the channels, creating a stable oxygen encapsulation state. This physical capture directly induces chemical activation, as evidenced by the increase in the O–O bond length of O2 adsorbed at the wedge sites (Supplementary Table S12). This indicates that in the nanoconfined environment, O2 not only achieves effective physical retention but also undergoes electronic modulation and chemical activation by the BN nanoparticles, thereby reaching a metastable state with a higher propensity to react. These findings align with the rapid kinetic characteristics observed during the initial 3 min of the reaction. Furthermore, the electrostatic potential distributions showed that the wedge-shaped interface, formed by confining BN nanoparticles within the CNT inner walls, established a distinct electron density gradient (Fig. 4b). Combined with the electron-donating capability of the N sites, this effectively suppresses side reactions and imparts high oxidative selectivity toward target pollutants [25,50,51].Fig. 4Computational analysis of oxygen confinement and reactive oxygen species (ROS) formation in boron nitride (BN)–carbon nanotube (CNT) composites. a, Adsorption energies of O2 on BN-in-CNT, BN-out-CNT, and CNT sites. b, Electrostatic potential (ESP) mapped on the electron density isosurfaces for CNTs, BN-in-CNT, and BN-out-CNT. c–d, Reaction free-energy profiles for the confined BN-in-CNT system (c) and unconfined BN-out-CNT system (d). e, Molecular dynamics simulation snapshots showing O2 diffusion into the CNT channels. f–g, Finite-element simulations of the spatial concentration distributions of singlet oxygen (1O2; f) and •OH (g) in a 20 nm-diameter BN-in-CNT at the indicated times. Computational analysis of oxygen confinement and reactive oxygen species (ROS) formation in boron nitride (BN)–carbon nanotube (CNT) composites. a, Adsorption energies of O2 on BN-in-CNT, BN-out-CNT, and CNT sites. b, Electrostatic potential (ESP) mapped on the electron density isosurfaces for CNTs, BN-in-CNT, and BN-out-CNT. c–d, Reaction free-energy profiles for the confined BN-in-CNT system (c) and unconfined BN-out-CNT system (d). e, Molecular dynamics simulation snapshots showing O2 diffusion into the CNT channels. f–g, Finite-element simulations of the spatial concentration distributions of singlet oxygen (1O2; f) and •OH (g) in a 20 nm-diameter BN-in-CNT at the indicated times. We analyzed the reaction coordinates to further understand how the molecular oxygen confinement ladder establishes the preferential pathway for 1O2 from a thermodynamic standpoint. In the BN-in-CNT system, 1O2 generation involves O2•− formation and O2•− and HO2• recombination, which is thermodynamically favorable. Benefiting from the electronic modulation within the wedge-shaped confinement, the energy barrier for O2•− and HO2• recombination to generate 1O2 was as low as −2.6 eV, lower than the values for the •OH-mediated pathway (−1.9 eV; Fig. 4c) and BN-out-CNT system (−0.3 and −1.5 eV; Fig. 4d). This theoretical thermodynamic preference was concurrently corroborated by the pronounced EPR signals of superoxide radicals (Supplementary Fig. S40). Concurrently, the electron-deficient concave surface of the CNT facilitated the decomposition of H2O2 into •OH through π–π interactions with the BN nanoparticles (Supplementary Fig. S41). This suggests that the wedge-shaped structure not only spatially aggregates reactants but also energetically filters them, thereby establishing the dominant pathway for 1O2 generation. From a kinetic perspective, MD simulations demonstrated that a large amount of O2 can spontaneously enter the CNT lumen within 1 ns, providing theoretical support for the successful entry and in situ confined catalysis of O2 within BN-in-CNT (Fig. 4e). Based on these analyses, we constructed a composite model to simulate the reaction process within the confined microchannels and calculated the concentration gradients of •OH and 1O2 inside the BN-in-CNT nanotubes (Fig. 4f and g). The 20-nm diameter of the CNT restricted the diffusion of •OH and 1O2, leading to elevated steady-state concentrations within the tube. Within 1 × 10−6 s, the 1O2 concentration can reach 4.8 × 10−5 mol L−1. This high concentration is maintained because the wedge-shaped structure physically prevents dilution of the microreactor by the bulk solution, thereby reducing radical loss [52]. In contrast, when the CNT diameter increases to 50 nm, the radical concentration decreases (Supplementary Fig. S42), although 1O2 still maintains a higher concentration than •OH. This phenomenon is likely related to the extremely short half-life of •OH compared to 1O2 (1 × 10−3 μs versus 4 μs) [53]. Radicals rapidly diffuse from the catalyst surface and decay in the bulk solution, reducing their likelihood of collision with pollutants. Therefore, without effective confinement, maintaining high radical concentrations for optimal catalytic activity is challenging. In summary, by allowing O2 entry and locking it within antiwetting wedge structures, the system achieves efficient transformation and in situ utilization from nonselective radicals to highly selective 1O2. Practical potential in NF concentrate To evaluate the system in complex ionic environments, the removal efficiency of SMX was examined in the presence of Na2SO4, NaCl, Na3PO4, and NaHCO3 electrolytes. SMX maintained high removal efficiencies within 15 min, reaching 100% in Na2SO4 and exceeding 80% in other systems, demonstrating the environmental adaptability of the nanoconfined electrode in multi-ion coexisting systems (Fig. 5a). DFT calculations further revealed the molecular selectivity mechanism of the system. The degradation efficiencies of five pollutants indicated a negative correlation with their highest occupied molecular orbital–lowest unoccupied molecular orbital energy gaps (Fig. 5b and c). According to frontier molecular orbital theory, a narrower energy gap corresponds to higher chemical softness. In the oxygen encapsulation system, the electrophilic 1O2, induced by the wedge-shaped structure, preferentially attacked electron-rich molecules with easily polarizable electron clouds. This explains why the system precisely identifies and degrades target micropollutants in the presence of complex background organic matter.Fig. 5Pollutant degradation and performance in nanofiltration (NF) concentrate using boron nitride (BN)–carbon nanotube (CNT) materials. a, Effects of background anions on sulfamethoxazole (SMX) degradation by BN-in-CNT. b, Degradation kinetics of SMX, 17α-ethinylestradiol (EE2), carbamazepine (CBZ), benzoic acid (BA), and atrazine (ATZ). c, Highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) energy levels and corresponding energy gaps of the five pollutants. The calculated frontier molecular orbitals are shown, with the LUMO in green and the HOMO in blue; for each orbital, the dark and light shades denote the positive and negative phases of the wavefunction, respectively. d, SMX degradation in nanofiltration (NF) concentrate by BN-in-CNT and BN-out-CNT systems with N2 purging or in the absence of Fe2+. e, Effects of different quenchers on SMX degradation by BN-in-CNT in NF concentrate. f, Temporal changes in adsorbed iron and dissolved Fe2+ and Fe3+ during iron redox cycling in the CNT, BN-out-CNT and BN-in-CNT systems in NF concentrate. g–i, Excitation–emission matrix (EEM) fluorescence spectra of the initial NF concentrate (g), and NF concentrate after treatment by BN-in-CNT (h) or BN-out-CNT (i). j, Long-term SMX degradation by the BN-in-CNT system during continuous operation. The feed was switched from Milli-Q water to NF concentrate after 24 h, as indicated by the vertical dashed line. The O2 and liquid flow rates were 2.5 and 3.3 mL min−1, respectively. Error bars represent standard deviations from n = 3 independent replicates. Pollutant degradation and performance in nanofiltration (NF) concentrate using boron nitride (BN)–carbon nanotube (CNT) materials. a, Effects of background anions on sulfamethoxazole (SMX) degradation by BN-in-CNT. b, Degradation kinetics of SMX, 17α-ethinylestradiol (EE2), carbamazepine (CBZ), benzoic acid (BA), and atrazine (ATZ). c, Highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) energy levels and corresponding energy gaps of the five pollutants. The calculated frontier molecular orbitals are shown, with the LUMO in green and the HOMO in blue; for each orbital, the dark and light shades denote the positive and negative phases of the wavefunction, respectively. d, SMX degradation in nanofiltration (NF) concentrate by BN-in-CNT and BN-out-CNT systems with N2 purging or in the absence of Fe2+. e, Effects of different quenchers on SMX degradation by BN-in-CNT in NF concentrate. f, Temporal changes in adsorbed iron and dissolved Fe2+ and Fe3+ during iron redox cycling in the CNT, BN-out-CNT and BN-in-CNT systems in NF concentrate. g–i, Excitation–emission matrix (EEM) fluorescence spectra of the initial NF concentrate (g), and NF concentrate after treatment by BN-in-CNT (h) or BN-out-CNT (i). j, Long-term SMX degradation by the BN-in-CNT system during continuous operation. The feed was switched from Milli-Q water to NF concentrate after 24 h, as indicated by the vertical dashed line. The O2 and liquid flow rates were 2.5 and 3.3 mL min−1, respectively. Error bars represent standard deviations from n = 3 independent replicates. To verify the system's applicability in real water matrices, degradation experiments were conducted using actual NF concentrate, yielding sustained high removal rates for both pollutants and total organic carbon (Fig. 5d and Supplementary Fig. S43). Furthermore, the reaction mechanism was elucidated in depth through radical quenching, N2 purging, and monitoring of iron cycling (Fig. 5e and f). The results indicated that 1O2 was the primary oxidative species, while •OH, O2•−, and H2O2 served as precursors for 1O2 generation, with the reaction pathway dominated by nonradical oxidation. The narrow wedge-shaped inlets of BN-in-CNT acted as molecular sieves, effectively intercepting most macromolecular organic matter and colloidal components (Fig. 5g–i and Supplementary Fig. S44). This prevents clogging of active sites or quenching of internally generated 1O2 [54]. The competitive adsorption of dissolved organic matter (DOM) is therefore largely confined to the outer CNT surface and does not penetrate the hydrophobic wedge-shaped nanochannels, as evidenced by the progressive reduction in humic- and fulvic-like fluorescence peaks observed in the three-dimensional excitation–emission matrix (EEM) spectra (Fig. 5g–i and Supplementary Fig. S44). Simultaneously, as a salt-tolerant oxidant, 1O2 remains inert toward Cl−, inherently avoiding the formation of toxic halogenated by-products such as ClO3−. Therefore, the steric exclusion by wedge-shaped nanochannels and the electrophilic oxidation of 1O2 jointly govern the system's selective oxidation behavior. Beyond laboratory-scale studies, this research highlights the application potential of the BN-CNT conductive aeration membrane system in advanced water treatment. We conducted a 48-h continuous flow experiment (Fig. 5j and Supplementary Fig. S45). After 24 h of operation, the influent matrix was switched from ultrapure water to complex NF concentrate, and the SMX degradation efficiency remained stable at approximately 100%. This confirms the system's interfacial stability during long-term dynamic operation. Scanning electron microscopy (SEM) images after long-term operation (Supplementary Fig. S46) revealed that the BN-in-CNT membrane surface maintained an intact, interconnected porous network, with no evidence of polymer skeleton collapse due to electrolyte infiltration. X-ray diffraction patterns (Supplementary Fig. S47) also confirmed the integrity of the catalyst's crystal structure, as the characteristic diffraction peaks of BN and CNTs showed no significant changes in position or intensity. These results demonstrate that the self-assembled wedge-shaped structure imparts kinetic antiwetting capability and achieves physical anchoring at the microscopic scale, enabling it to withstand prolonged flushing and oxidative corrosion from high-salinity fluids. Consequently, combined with its low raw-material cost of approximately 1.4 USD m−2 and highly integrated membrane module configuration [55], this technology enables a convenient plug-and-play operation mode. The preassembled membrane module requires no additional civil engineering or specialized installation, enabling direct integration into existing treatment trains as a functional unit [56,57]. While pilot-scale validation and comprehensive energy auditing, including the one-time synthesis costs associated with high-temperature calcination, remain necessary steps toward full engineering deployment, the dual advantages of technical performance and economic feasibility support application of this molecular oxygen confinement ladder system in NF concentrate treatment and the in situ upgrading of wastewater treatment plants. Conclusion By integrating experimental characterization with simulation results, this study confirms that O2 molecules can enter and be confined within the wedge-shaped structures of BN-in-CNT, validating the molecular oxygen confinement ladder strategy. The results further reveal a fundamental transition in the catalytic pathway of the Fe(II)/Fe(III) EF process, triggered by the O2 nanoconfinement effect. While previous research has predominantly focused on using nanoconfinement to enhance adsorption capacity and enrich pollutant molecules, the in situ activation mechanism of gas-phase molecules within antiwetting microchannels has often been overlooked. The wedge-shaped confined structure developed in this study leverages the kinetic antiwetting effect to effectively maintain a high-concentration local O2 microenvironment. This further alleviates mass-transfer limitations and enables efficient in situ generation of ROS within CNTs. This gas-confinement strategy may be applicable to other nanoconfined catalytic systems. Benefiting from the shielding effect of the physical confinement structure, the reaction pathway in which 1O2 dominates, alongside •OH participation, demonstrates anti-interference capabilities and structural stability under high-salinity conditions. These findings support treatment of complex, high-concentration wastewater environments. Nevertheless, future scale-up will require addressing key engineering challenges, including cost-effective fabrication of conductive aeration membranes, reactor configuration optimization, and long-term operational stability across variable influent conditions. Overall, the membrane-aerated EF system enables stable pollutant removal and provides a strategy for protecting drinking water sources from ECs. CRediT authorship contribution statement Yufei Cheng: Writing – original draft, Conceptualization, Methodology, Visualization. Mengzhuang Yang: Investigation. Yuchen Hu: Investigation. Peng Liu: Software, Formal analysis. Weichen Zeng: Investigation. Xiaohang Han: Investigation. Tianyi Wang: Investigation. Langming Bai: Investigation. Jinlong Wang: Investigation. Han Zhang: Writing – review & editing, Funding acquisition, Supervision. Xiaoguang Duan: Writing – review & editing. Heng Liang: Writing – review & editing, Funding acquisition, Supervision. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.