Introduction

The growing presence of emerging organic micropollutants, including pharmaceuticals, personal care products, antibiotics, and endocrine disruptors, in water systems has intensified the global water crisis, posing serious risks to aquatic ecosystems and human health1. Photocatalytic water treatment technology, harnessing abundant and renewable solar energy to degrade micropollutants, stands out as one of the most promising approaches2,3,4. A major obstacle in advancing photocatalysis for practical applications lies in the unsatisfactory removal efficiency of trace micropollutants (ng/L to μg/L) in real water systems. This challenge arises because the oxidative species, such as hydroxyl radicals (HO•), are primarily consumed by water matrices, including natural organic matter (NOM) and anions (e.g., Cl−, HCO3−, and NO3−), which exist at concentrations several orders of magnitude higher than the target micropollutants5. For instance, 5 mg/L of NOM, commonly found in natural waters, can quench approximately 90% of the generated HO•6. Moreover, anions can react with HO• to generate secondary radicals, such as chlorine, bromine, and carbonate radicals. These secondary radicals have a lower oxidation capacity than HO• and may interact with NOM, potentially forming toxic disinfection byproducts (DBPs)7. However, limited attention has been paid to strategies for selectively degrading trace micropollutants in complex real-world water conditions.

Nanoconfined catalysis exhibits exceptional potential for selectively degrading micropollutants by minimizing the interference from water matrices. The nanoconfined structure leverages physical size-exclusion effects to block larger NOM from reaching the catalytic sites, allowing smaller micropollutants to preferentially access and undergo degradation8. For example, graphene oxide (GO)-wrapped defective TiO2 creates nanoconfined catalytic sites on the TiO2 surface within GO, enabling the selective degradation of micropollutants by excluding NOM through the nanoscale pores of GO9. Additionally, the high local concentrations of reactants and unique interfacial effects within the nanoconfined environment can promote the interactions between oxidative species and target micropollutants10,11,12,13. Thus, nanoconfined catalysis provides an innovative strategy for the selective and efficient removal of micropollutants in complex aquatic environments.

In this context, core-shell structured composite photocatalysts offer stable nanoconfined catalytic sites, as the shell can be precisely engineered to prevent NOM from reaching the reaction sites while allowing micropollutants to access these sites for rapid degradation. Among potential shell materials, graphitic carbon nitride (C3N4) and covalent carbon nitride (C2N) have attracted attention due to their chemical stability, favorable bandgaps, and promising photocatalytic activity14,15,16,17. C3N4 possesses a nitrogen-rich polymeric framework with limited π-conjugation18, whereas C2N features a highly ordered 2D architecture19. These structural and electronic differences can give rise to distinct heterojunction types (e.g., type II or Z-scheme) and interfacial charge transfer behaviors when integrated with semiconductor cores such as TiO2. Notably, a systematic comparison of the photocatalytic performance between TiO2/C3N4 and TiO2/C2N core-shell heterostructures has not yet been reported.

Herein, we carefully designed two core-shell structured composite photocatalysts, TiO2/C3N4 and TiO2/C2N, using melamine and 1,4,5,8,9,11-hexaazatriphenylenehexacarbonitrile (HAT-CN) as carbon nitride precursors, respectively, followed by calcination in argon and natural cooling (Fig. 1a). Characterizations and density functional theoretical (DFT) calculations reveal that TiO2/C3N4 follows a Z-scheme heterostructure, while TiO2/C2N adopts a type II heterostructure. Compared with commonly used oxidants, including H2O2, peroxydisulfate (PDS), and peroxymonosulfate (PMS), free chlorine (FC) is a more cost-effective option and can produce 2–5 times more HO• than systems without FC, outperforming H2O2, PDS, and PMS under comparable conditions20,21,22,23,24. Therefore, FC was selected as the oxidant to develop UV/FC/TiO2/C3N4 and UV/FC/TiO2/C2N systems. The primary reactive species in both systems are identified as HO•. However, the Z-scheme heterostructure of TiO2/C3N4 results in the formation of HO• on the C3N4 surface, where it is quickly consumed by water matrices. In contrast, the type II heterostructure of TiO2/C2N confines HO• near the TiO2 surface, achieving efficient and stable degradation of trace micropollutants when exposed to NOM, anions, and real water samples. This work pioneers a TiO2/C2N core-shell heterostructure that enables the selective degradation of target micropollutants under complex real water conditions. This design establishes a paradigm for developing core-shell catalysts with robust resilience to water matrices, advancing practical water treatment technologies.

Fig. 1: Synthetic illustration and characterization of core-shell structured catalysts.
Fig. 1: Synthetic illustration and characterization of core-shell structured catalysts.
Full size image

a Schematic illustration of the synthesis for core-shell structured catalysts. b, d TEM images of TiO2/C3N4 and TiO2/C2N, respectively. c, e TEM-EDS mapping images of TiO2/C3N4 and TiO2/C2N, respectively. f Raman spectra (inset: magnification of local areas) of TiO2, TiO2/C3N4, and TiO2/C2N. g Solid-state 13C nuclear magnetic resonance (NMR) spectra of TiO2/C3N4 and TiO2/C2N. h O 1s XPS spectra of TiO2, TiO2/C3N4, and TiO2/C2N. i Pore size distributions of TiO2/C3N4 and TiO2/C2N.

Results

Characterization of the catalysts

Thermogravimetric analysis (TGA) of C3N4 and C2N shows that both materials were nearly fully decomposed by 800 °C (Supplementary Fig. 1a). At 800 °C, the mass losses of TiO2/C3N4 and TiO2/C2N were 1.77% and 1.82% higher than that of TiO2 (Supplementary Fig. 1b), corresponding to the mass percentages of C3N4 and C2N in the composites. Transmission electron microscopy (TEM) images of TiO2/C3N4 and TiO2/C2N reveal that the TiO2 particles, with an average size of approximately 100 nm, are encapsulated by thin carbon nitride shells (Supplementary Fig. 2a−d). Enlarged TEM images provide further confirmation of the core-shell structure, clearly showing the TiO2 core surrounded by carbon nitride shells (Fig. 1b, d). HRTEM images display a TiO2 core with a crystal plane spacing of 0.35 nm, corresponding to the TiO2 (101) facet (Supplementary Fig. 3a, b). Energy-dispersive X-ray spectroscopy (EDS) mapping of TiO2/C3N4 (Fig. 1c) and TiO2/C2N (Fig. 1e) indicates a uniform distribution of Ti and O in the core and C and N in the shell.

X-ray diffraction (XRD) patterns illustrate that all samples closely matched the anatase TiO2 (PDF#01-0562) (Supplementary Fig. 4). Raman spectra of TiO2/C3N4 and TiO2/C2N reveal the coexistence of TiO2 and carbon nitride, with the TiO2 peak shifting from 142 to 145 cm−1, suggesting the strong interfacial contact between TiO2 and carbon nitride (Fig. 1f). The intensity ratios of the D band and the G band (ID/IG) in TiO2/C3N4 and TiO2/C2N are 0.98 and 0.90, respectively, indicating that the C2N structure is more ordered and exhibits fewer defects compared to C3N425. As shown in Fig. 1g, solid-state 13C nuclear magnetic resonance (NMR) spectrum of TiO2/C3N4 exhibits two distinct peaks at 165 and 156 ppm, corresponding to C2N-NHx (a) and C3N (b) within the heptazine units of C3N4, respectively26. In the solid-state 13C NMR spectrum of TiO2/C2N, the peaks at 153 and 148 ppm are attributed to the carbon atoms at the phenyl edges linked to C = N (a) and the carbon atoms of triphenylene cores (b), respectively, which confirms the presence of pyrazine core structure in C2N25. Fourier transform infrared (FTIR) spectra of TiO2/C3N4 and TiO2/C2N show Ti-O-Ti signals in the 500–700 cm−1 range (Supplementary Fig. 5). For TiO2/C3N4, the characteristic peaks at 1200–1700 and 810 cm−1 are assigned to the aromatic C3N4 heterocyclic units and the heptazine rings, respectively. In contrast, the peaks at 1755 and 1280 cm−1 in TiO2/C2N correspond to C = C/C = N and C-C/C-N bonds in C2N, respectively27.

The C 1s (Supplementary Fig. 6a) and N 1s (Supplementary Fig. 6b) X-ray photoelectron spectroscopy (XPS) spectra reveal the CN heterocyclic frameworks in C3N4, with the signals for N = C-N, N-(C)2, and N-(C)3 located at 288.3, 398.5, and 399.0 eV, respectively26. The characteristic peaks of C-C, C = N, and C-N in C2N were observed at 284.5, 398.5, and 400.9 eV, respectively27. The peaks in the O 1s XPS spectra at 530.5, 532, and 533 eV correspond to lattice oxygen (OL), vacancy oxygen (OV), and chemisorbed oxygen species (OC), respectively (Fig. 1h)28. The OV content in TiO2/C2N was determined to be 39.7%, higher than that in TiO2/C3N4 (26.6%) and TiO2 (12.5%, obtained by calcination without precursors), which can be attributed to the greater reduction of TiO2 by the reducing gases (e.g., carbon and nitrogen monoxides) generated from HAT-CN during the calcination process. Moreover, solid-state electron paramagnetic resonance (EPR) spectra confirm that the OV content (g = 2.003) follows the order TiO2/C2N > TiO2/C3N4 > TiO2 (Supplementary Fig. 7). Nitrogen adsorption-desorption isotherms of TiO2/C3N4 and TiO2/C2N exhibit a type IV pattern, suggesting the presence of mesopores (Supplementary Fig. 8)29,30. Pore size distribution curves reveal similar average mesopore sizes of TiO2/C3N4 (4.0 nm) and TiO2/C2N (3.7 nm) (Fig. 1i). These mesopores are primarily attributed to the interfacial voids between the TiO2 core and the carbon nitride shells, as well as to structural imperfections or partial stacking irregularities generated during the polymerization process.

Identification of TiO2/C3N4 and TiO2/C2N heterostructures

UV−vis diffuse reflectance spectroscopy (UV−vis DRS) was conducted to determine the optical properties of TiO2, C3N4, and C2N, revealing their absorption edges at approximately 390, 477, and 603 nm, respectively (Fig. 2a). The corresponding Tauc plots indicate bandgap of 3.18 eV for TiO2, 2.60 eV for C3N4, and 2.06 eV for C2N (Fig. 2a inset). Furthermore, the XPS valence band (VB) spectra show that the VB positions (EVB, XPS) of TiO2, C3N4, and C2N are 2.37, 1.12, and 1.19 eV, respectively (Fig. 2b). Using the equation EVB, RHE = ϕ + EVB, XPS−4.44, where ϕ is the work function of the XPS instrument (4.8 eV) and EVB, RHE represents the VB position relative to the reversible hydrogen electrode (RHE)31, the calculated EVB, RHE values are 2.73 V for TiO2, 1.48 V for C3N4, and 1.55 V for C2N. Consequently, the conduction band (CB) positions versus RHE (ECB, RHE) for TiO2, C3N4, and C2N are determined to be −0.45, −1.12, and −0.51 V, respectively. Figure 2c illustrates the band edge potential diagram for TiO2, C3N4, and C2N. The ECB, RHE of the materials are higher than the reduction potential of O2, suggesting that their photogenerated electrons can reduce O2 to O2•− (−0.33 V vs. RHE). The EVB, RHE of TiO2 is lower than the oxidation level of H2O, enabling its photogenerated holes to oxidize H2O to HO• (2.70 V vs RHE). In contrast, the EVB, RHE of C3N4 and C2N are insufficient to produce HO• from H2O.

Fig. 2: Heterostructural characteristics of TiO2/C3N4 and TiO2/C2N.
Fig. 2: Heterostructural characteristics of TiO2/C3N4 and TiO2/C2N.
Full size image

a UV−vis diffuse reflectance spectroscopy (UV−vis DRS) of TiO2, C3N4, and C2N (inset: the Tauc plots of TiO2, C3N4, and C2N). b XPS valence band (VB) spectra of TiO2, C3N4, and C2N. c Band edge potentials for TiO2, C3N4, and C2N. d, e Electron paramagnetic resonance (EPR) signals for DMPO–O2•− and DMPO–HO•, respectively. f Operando Ti 2p XPS spectra of TiO2/C3N4 and TiO2/C2N. g Interfacial charge density distribution at the TiO2/C3N4 and TiO2/C2N junctions. Ti atoms are depicted as blue spheres, and O atoms are depicted as red spheres. h Schematic diagram of charge transfer in TiO2/C3N4 and TiO2/C2N under photoexcitation.

As shown in Fig. 2d, e, 5,5-dimethyl-1-pyrroline N-oxide (DMPO) was used as a trapping agent to detect O2•− and HO• species. EPR signals corresponding to O2•− were observed for TiO2, C3N4, C2N, TiO2/C3N4, and TiO2/C2N (Fig. 2d), as confirmed by comparison with the simulated DMPO–O2•− spectrum (Supplementary Fig. 9a, b). However, HO• signals were detected only for TiO2 and TiO2/C3N4 (Fig. 2e). This indicates that at the TiO2/C3N4 interface, holes from TiO2 directly oxidize H2O to HO•. At the TiO2/C2N interface, the holes transfer to the VB of C2N, where they cannot produce HO• from H2O. Operando XPS analysis of TiO2/C3N4 shows that, upon light irradiation, the Ti 2p peak shifted to a higher binding energy (Fig. 2f), suggesting the migration of photogenerated electrons from TiO2 to C3N4. Meanwhile, TiO2/C2N exhibited the opposite trend, which reveals that more photogenerated electrons remain in TiO2.

Heterointerface models of TiO2/C3N4 (Supplementary Fig. 10) and TiO2/C2N (Supplementary Fig. 11) were constructed, and their interface charge density difference was analyzed based on DFT calculations. As shown in Fig. 2g, charge accumulation in TiO2/C3N4 occurs on C3N4, indicating the presence of an internal electric field that drives electron flow from TiO2 to C3N4. In TiO2/C2N, charge predominantly accumulates on TiO2 rather than C2N, suggesting that electrons tend to transfer from C2N to TiO2. Consequently, the EPR spectroscopy, operando XPS analysis, and DFT calculations collectively confirm that TiO2/C3N4 follows a Z-scheme heterostructure, while TiO2/C2N adopts a type II heterostructure. Figure 2h presents a schematic diagram of charge transfer under photoexcitation: the internal electric field in TiO2/C3N4 drives electron flow from the CB of TiO2 to the VB of C3N4, resulting in electron accumulation on C3N4 and hole accumulation on TiO2; the internal electric field in TiO2/C2N directs electron transfer from C2N to TiO2, with holes moving from TiO2 to C2N.

Catalytic activity and mechanism

Photoluminescence (PL) and time resolved photoluminescence (TRPL) spectroscopy were utilized to examine the charge carrier transfer dynamics. The samples displayed a prominent PL peak at approximately 480 nm, attributed to the characteristic emission of TiO2 (Fig. 3a). The PL intensity of TiO2/C3N4 and TiO2/C2N was significantly lower than that of TiO2, indicating that the heterostrucures effectively suppressed the recombination of electrons and holes. Moreover, the average charge carrier lifetimes (τave) for TiO2, TiO2/C3N4, and TiO2/C2N were determined to be 6.28, 1.61, and 0.97 ns, respectively (Fig. 3b and Supplementary Table 1). Compared to TiO2, the remarkably shorter τave values of TiO2/C3N4 and TiO2/C2N suggest superior charge migration in both heterostrucures, which is consistent with the PL spectra.

Fig. 3: Catalytic performance and mechanisms of TiO2, TiO2/C3N4, and TiO2/C2N.
Fig. 3: Catalytic performance and mechanisms of TiO2, TiO2/C3N4, and TiO2/C2N.
Full size image

a–c Steady-state photoluminescence (PL), time resolved photoluminescence (TRPL) spectra, and transient photocurrent curves of TiO2, TiO2/C3N4, and TiO2/C2N. d Degradation rate constants of carbamazepine (CBZ) by photocatalytic and photocatalytic free chlorine (FC) activation systems. Conditions: [CBZ] = 5 µM, [FC] = 2 mg/L, [catalysts] = 10 mg/L, pH 7. e Comparison of state-of-the-art photocatalytic advanced oxidation systems for CBZ degradation20,21,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52. In each data point (a, b), a is the ratio of k′CBZ to catalyst dosage, and b is the ratio of k′CBZ to oxidant dosage. f Effects of different scavengers on CBZ degradation. g Concentrations of O2•− in different photocatalytic systems. h Gibbs free energy profiles for HO• generation in different photocatalytic FC activation systems. i Specific CBZ degradation rate constants contributed by different reactive species. The error bars in the figure (d, f, g, i) represent the standard deviation, calculated from at least two independent experiments.

Furthermore, photoelectrochemical characterization was performed to evaluate the charge carrier transfer capability. As shown in Fig. 3c, the current densities of TiO2/C3N4 (4.7 μA/cm2) and TiO2/C2N (5.2 μA/cm2) were obviously higher than that of TiO2 (1.0 μA/cm2). Additionally, electrochemical impedance spectroscopy (EIS) Nyquist plots reveal that the arc radii of TiO2/C3N4 and TiO2/C2N were significantly smaller than that of TiO2 (Supplementary Fig. 12). These results confirm the promoted interfacial charge transfer facilitated by the heterostructures.

As shown in Fig. 3d, carbamazepine (CBZ) was chosen as the target micropollutant for degradation using various photocatalytic and photocatalytic FC activation processes (Supplementary Fig. 13). The first-order degradation rate constants of CBZ (k′CBZ) by UV/TiO2, UV/TiO2/C3N4, and UV/TiO2/C2N were 0.08, 0.19, and 0.19 min−1, respectively. Notably, the UV/FC/TiO2/C3N4 and UV/FC/TiO2/C2N systems achieved k′CBZ of 0.52 and 0.62 min−1, respectively, which are much higher than that by UV/FC/TiO2 (0.30 min−1). Moreover, the k′CBZ by UV/FC/TiO2, UV/FC/TiO2/C3N4, and UV/FC/TiO2/C2N were 3.8, 2.7, and 3.2 times higher, respectively, compared to their counterparts without FC. The accelerated FC decay in UV/FC/TiO2/C3N4 and UV/FC/TiO2/C2N, compared to UV/FC/TiO2, suggests more efficient FC utilization due to the improved separation of h+-e− pairs in the heterostructures (Supplementary Fig. 14). Notably, the ratios of k′CBZ to catalyst dosage (k′/catalyst) and oxidant dosage (k′/oxidant) in UV/FC/TiO2/C2N reached 62 and 310 L min−1 g−1, respectively, the highest among all surveyed photocatalytic advanced oxidation systems (Fig. 3e and Supplementary Table 2). This demonstrates that the system achieves the most efficient CBZ degradation with minimal catalyst and oxidant usage, highlighting its potential for practical water treatment.

As shown in Fig. 3f, reactive species scavenging experiments reveal that the addition of Na2C2O4 (a hole scavenger)21 had minimal impact on CBZ degradation by the UV/FC/TiO2/C3N4 and UV/FC/TiO2/C2N systems, indicating that photogenerated holes play a minor role in CBZ degradation. However, CBZ degradation was severely inhibited by adding AgNO3 and tert-Butanol (TBA), which act as scavengers for electrons and HO•, respectively7,32,33. Moreover, purging nitrogen into the solution to lower the dissolved oxygen (DO) concentration (<0.5 mg/L) obviously suppressed CBZ degradation. These results suggest that electrons, HO•, and O2•− are likely to play important roles in the CBZ degradation process. DFT calculations show that the adsorption energies of TiO2/C3N4 and TiO2/C2N for DO are 4.88 and −3.45 eV respectively, indicating that DO is more strongly adsorbed on TiO2/C2N than TiO2/C3N4. Based on nitroblue tetrazolium (NBT) probe experiments, the O2•− yields in UV/TiO2/C2N reached 4.8 μM within 5 min, surpassing the 3.9 and 1.0 μM observed in UV/TiO2/C3N4 and UV/TiO2, respectively (Fig. 3g), likely due to the more favorable DO adsorption on TiO2/C2N. Furthermore, the O2•− signals in UV/TiO2/C3N4 and UV/TiO2/C2N disappeared immediately upon adding FC (Supplementary Fig. 15a), while the HO• signals increased significantly (Supplementary Fig. 15b). Accordingly, the reaction mechanism proceeds as follows: (i) photogenerated electrons reduce DO to form O2•−; (ii) O2•− reacts with FC to produce HO•; and (iii) HO• as the primary reactive species effectively degrades CBZ, in agreement with the literature20,21,22. The Gibbs free energy was calculated to evaluate the thermodynamic feasibility of HO• generation in the UV/FC/TiO2, UV/FC/TiO2/C3N4, and UV/FC/TiO2/C2N systems. Initially, HOCl is adsorbed onto the active sites, and subsequently reacts with O2•− to generate HO•. As shown in Fig. 3h, the first stage is thermodynamically favorable in all three systems, whereas the second stage requires overcoming an energy barrier. Among these systems, UV/FC/TiO2/C2N requires the smallest energy barrier, while UV/FC/TiO2 encounters the largest. Therefore, the thermodynamic ease of HO• generation follows the order: UV/FC/TiO2/C2N > UV/FC/TiO2/C3N4 > UV/FC/TiO2.

To specifically identify the contributions of HO• and other reactive species to CBZ degradation, probes including nitrobenzene (NB), benzoic acid (BA), caffeine (CAF), and 1,4-dimethoxybenzene (DMOB) were utilized to measure their steady-state concentrations (Supplementary Figs. 16−19). As shown in Fig. 3i, HO• generated in the UV/FC/TiO2/C3N4 and UV/FC/TiO2/C2N systems contributed to k′CBZ values of 0.45 and 0.52 min−1, respectively, which are 2.0 and 2.4 times higher than that in the UV/FC/TiO2 system (0.22 min−1). We classify radicals adsorbed on the catalysts as surface radicals and those released into the bulk solution as free radicals. Interestingly, surface HO• on TiO2/C2N accounted for 82.7% of the total HO• contribution to k′CBZ, while surface HO• on TiO2/C3N4 and TiO2 contributed only 45.8% and 39.9%, respectively. Likewise, surface ClO• on TiO2/C2N accounted for 77.6% of the overall ClO• contribution to k′CBZ, whereas for TiO2/C3N4 and TiO2, the surface and free ClO• contributions were nearly equivalent. These results suggest that in the UV/FC/TiO2/C2N system, HO• and ClO• are predominantly confined near the catalyst surface, creating a localized catalytic environment. In contrast, in the UV/FC/TiO2/C3N4 and UV/FC/TiO2 systems, less than half of the HO• and ClO• remain surface-bound, with the majority diffusing into the bulk solution, where they are more susceptible to being consumed by background water matrices.

Resistance to water matrices

The resistance of the core-shell structured TiO2/C3N4 and TiO2/C2N to varying concentrations of NOM in photocatalytic FC activation processes was evaluated, using TiO2 as a control. As shown in Fig. 4a, in the presence of 1, 5, and 10 mg/L NOM, the k′CBZ by UV/FC/TiO2 obviously decreased from 0.30 min–1 to 0.16, 0.05, and 0.03 min–1, respectively (degradation kinetics shown in Supplementary Fig. 20). Similarly, the k′CBZ by UV/FC/TiO2/C3N4 decreased from 0.52 min–1 to 0.40, 0.21, and 0.12 min–1, respectively. The significant decline in k′CBZ can be attributed to the extensive consumption of HO• by NOM. Notably, NOM at concentrations of 1, 5, and 10 mg/L had a negligible impact on the k′CBZ by UV/FC/TiO2/C2N, demonstrating the strong resistance of TiO2/C2N to NOM.

Fig. 4: Nanoconfinement-enabled resistance to water matrices for selective micropollutant degradation.
Fig. 4: Nanoconfinement-enabled resistance to water matrices for selective micropollutant degradation.
Full size image

a Resistance to NOM, Cl−, HCO3−, and NO3− at varying concentrations. b Resistance to pH and real water samples, and the reusability of TiO2/C2N in tap water. Conditions: [CBZ] = 5 µM, [FC] = 2 mg/L, [catalysts] = 10 mg/L, pH 7. c Removal efficiencies of eight coexisting micropollutants in pure water and tap water. Conditions: [micropollutants] = 1 µg/L, [FC] = 0.5 mg/L, [catalysts] = 2 mg/L, pH 7. d Photocatalytic FC activation mechanisms of the UV/FC/TiO2/C3N4 and UV/FC/TiO2/C2N systems. The error bars in this figure (a–c) represent the standard deviation, calculated from at least two independent experiments.

The hydrated CBZ molecule, with a size of approximately 1.5–1.7 nm9,34, is smaller than the pore sizes of C3N4 (4.0 nm) and C2N (3.7 nm), while the overall size of HOCl is even smaller, at less than 1 nm35. In contrast, the average particle size of NOM, approximately 325 nm as measured by dynamic light scattering (Supplementary Fig. 21), far exceeds the pore sizes of C3N4 and C2N. Therefore, the C3N4 and C2N shells can effectively prevent NOM from reaching the TiO2 core while allowing CBZ and FC to diffuse through. Notably, in the Z-scheme heterostructure of TiO2/C3N4, photogenerated electrons remain on the C3N4 shell, where they sequentially react with DO and FC to generate HO•. However, the HO• formed on the C3N4 surface can be rapidly consumed by nearby NOM. In the type II heterostructure of TiO2/C2N, photogenerated electrons accumulate on the TiO2 core, creating a nanoconfined reaction space within the C2N shell. This confinement keeps HO• localized near the TiO2 surface, thereby reducing its interaction with NOM. Consequently, CBZ diffuses through the C2N shell and be effectively degraded within this nanoconfined space.

Excitation-emission matrix (EEM) measurements of NOM before and after 20 min of reaction show a significant decrease in fluorescence intensity in UV/FC/TiO2 and UV/FC/TiO2/C3N4, whereas it remained almost unchanged in UV/FC/TiO2/C2N (Supplementary Fig. 22). These findings confirm that NOM was degraded in UV/FC/TiO2 and UV/FC/TiO2/C3N4, but was effectively excluded from the nanoconfined reaction sites in UV/FC/TiO2/C2N.

Common anions in water including Cl–, HCO3–, and NO3– significantly inhibited the degradation of CBZ by the UV/FC/TiO2 and UV/FC/TiO2/C3N4 systems (Fig. 4a, degradation kinetics shown in Supplementary Fig. 23). Specifically, the k′CBZ of UV/FC/TiO2 decreased by 45.7%, 71.7%, and 57.2% in the presence of 10 mM Cl–, HCO3–, and NO3–, respectively, while that of UV/FC/TiO2/C3N4 dropped by 21.2%, 46.6%, and 25.8%, respectively. Interestingly, the anions had negligible effects on the k′CBZ by the UV/FC/TiO2/C2N system, indicating the strong resistance of TiO2/C2N to anions. The Zeta potential of TiO2/C2N at pH 7 was measured to be −28.4 mV, more negative than that of TiO2/C3N4 (−13.5 mV) and TiO2 (−5.3 mV) (Supplementary Fig. 24). The highly negatively charged surface of the C2N shell effectively hindered the entry of most anions into the nanoconfined reaction space near the TiO2 surface through electrostatic repulsion, thereby reducing the consumption of HO• by the anions. However, the HO• formed in UV/FC/TiO2 and UV/FC/TiO2/C3N4 were exposed to Cl– and HCO3–, and were likely converted into less reactive Cl• and CO3•−, respectively.

Figure 4b shows that the k′CBZ by the three systems increased slightly when the pH dropped from 7 to 6. However, when the pH was raised from 7 to 9, the k′CBZ by UV/FC/TiO2 and UV/FC/TiO2/C3N4 decreased by 69.4% and 69.7%, respectively (degradation kinetics shown in Supplementary Fig. 25). In contrast, the k′CBZ by UV/FC/TiO2/C2N at pH 9 remained almost unchanged compared to its value at pH 7. Notably, FC (pKa = 7.5) predominantly exists as HOCl at pH values below 7, whereas it transitions to OCl– at pH values above 7. OCl–, as a strong scavenger of HO•, can convert it into less reactive ClO•36, which likely accounts for the sharp decline in k′CBZ by UV/FC/TiO2 and UV/FC/TiO2/C3N4 at higher pH. In UV/FC/TiO2/C2N, HOCl can easily diffuse through the pores of the C2N shell to access the nanoconfined catalytic sites, while OCl– faced difficulty penetrating the negatively charged C2N surface due to electrostatic repulsion. Consequently, the stable k′CBZ at increasing pH by UV/FC/TiO2/C2N is attributed to the negatively charged C2N shell, which serves as a barrier to OCl–.

The effects of water matrices from real water samples (i.e., tap water, river water, and wastewater treatment plant (WWTP) effluent) on k′CBZ are presented in Fig. 4b (degradation kinetics shown in Supplementary Fig. 26), with the corresponding water parameters listed in Supplementary Table 3. The k′CBZ by UV/FC/TiO2 significantly decreased by 50.7%, 62.2%, and 75.3% in river water, tap water, and WWTP effluent, respectively, whereas for UV/FC/TiO2/C3N4, the reductions were 21.6%, 40.8%, and 59.2%, respectively. Notably, UV/FC/TiO2/C2N exhibited exceptional stability in degrading CBZ across all real water samples. This enhanced stability can be ascribed to the size exclusion and electrostatic forces of the C2N shell, which effectively protect the nanoconfined reaction sites from interference by various water matrices. Additionally, the UV/FC/TiO2/C2N system maintained its degradation efficiency across ten consecutive cycles in tap water (Fig. 4b), highlighting its stability and suitability for practical water treatment scenarios. This can be attributed to (i) the highly nitrogen-doped C2N with a well-ordered porous network19, which enhances its intrinsic chemical stability, and (ii) the nanoconfinement effect, which increases the local concentration of micropollutants10,11,12,13, favoring interactions between HO• and the micropollutants rather than with C2N.

A mixture of eight coexisting micropollutants including CBZ, diclofenac, naproxen, tetracycline, gemfibrozil, propranolol, ciprofloxacin, and ibuprofen, each at a concentration of 1 µg/L, was prepared to simulate their degradation by the photocatalytic FC activation systems under environmentally relevant conditions, as these micropollutants are commonly detected in wastewater and surface water. The degradtion efficiencies of the eight micropollutants by UV/FC/TiO2 and UV/FC/TiO2/C3N4 were notably lower in tap water (Fig. 4c) and WWTP effluent (Supplementary Fig. 27) compared to pure water, while that by UV/FC/TiO2/C2N remained nearly consistent. This result demonstrates that the UV/FC/TiO2/C2N system is highly effective for practical water treatment, successfully mitigating coexisting micropollutants at environmentally relevant concentrations.

Figure 4d illustrates the mechanism underlying the nanoconfinement-enhanced selective micropollutant degradation. In the UV/FC/TiO2/C3N4 system, photogenerated electrons accumulate on the C3N4 surface, driven by the Z-scheme heterostructure. These electrons then reduce DO to form O2•−, which subsequently reacts with FC to produce HO•. However, the HO• formed on the C3N4 surface is susceptible to rapid consumption by nearby NOM and anions. In contrast, the UV/FC/TiO2/C2N system displays distinct behavior owing to its type II core-shell heterostructure. Here, photogenerated electrons accumulate on the TiO2 core, creating a nanoconfined reaction space within the C2N shell. Micropollutants, DO, and HOCl, all smaller than the pore size of the C2N shell, can diffuse into this confined space. However, larger NOM molecules and negatively charged anions are effectively excluded by the C2N shell through size exclusion and electrostatic repulsion, respectively. As a result, the type II core-shell heterostructure of TiO2/C2N enables nanoconfined catalysis, localizing the degradation of micropollutants by HO• within the C2N shell and minimizing undesired interactions with the surrounding water matrices. The degradation products of CBZ by UV/FC/TiO2/C2N were identified using QTOF-MS analysis (Supplementary Table 4). The proposed degradation pathways for CBZ, primarily induced by HO•, include hydroxylation, deamination, and decarboxylation (Supplementary Fig. 28).

Impact of C2N thickness on nanoconfined catalysis

Given the critical role of the C2N shell in constructing the core-shell nanoconfined catalytic system, fine-tuning its properties is essential for maximizing catalytic efficiency and achieving selective degradation of target micropollutants. The thickness of the C2N shell influences several factors, including the OV content in TiO2, light transmittance and O2 diffusion through the C2N layer, and the surface electronegativity of C2N (Fig. 5a).

Fig. 5: Characterization and catalytic activity of TiO2/C2N with varying C2N thicknesses.
Fig. 5: Characterization and catalytic activity of TiO2/C2N with varying C2N thicknesses.
Full size image

a Overview of the effects of C2N thickness on the nanoconfined core-shell TiO2/C2N photocatalysis. b TEM images of TiO2/C2N-1, TiO2/C2N-2, and TiO2/C2N-3 with varying C2N thicknesses (scale bar: 10 nm). c–h Vacancy oxygen (OV) contents, solid-state EPR spectra, UV−vis DRS, oxygen temperature-programmed desorption (O2-TPD) profiles, O2•− EPR spectra, and Zeta potentials of TiO2, TiO2/C2N-1, TiO2/C2N-2, and TiO2/C2N-3. i Impacts of C2N thickness on the resistance to water matrices. Conditions: [CBZ] = 5 µM, [FC] = 2 mg/L, [catalysts] = 10 mg/L, pH 7. The error bars in this figure (i) represent the standard deviation, calculated from at least two independent experiments.

We prepared core-shell TiO2/C2N composites with varying C2N thicknesses by adjusting the amount of the C2N precursor. The resulting thicknesses of TiO2/C2N-1, TiO2/C2N-2 (referred to as TiO2/C2N above), and TiO2/C2N-3 were approximately 1–2, 5–6, and 10–11 nm, respectively (Fig. 5b). O 1s XPS spectra show that the OV content in TiO2/C2N increased with the thickness of the C2N shell (Fig. 5c and Supplementary Fig. 29). Moreover, solid-state EPR spectra confirm that the OV content was highest in TiO2/C2N-3, followed by TiO2/C2N-2, and lowest in TiO2/C2N-1 (Fig. 5d). However, TiO2/C2N-3 exhibited weaker UV light absorption compared to TiO2/C2N-1 and TiO2/C2N-2 (Fig. 5e). This is because the thicker C2N shell in TiO2/C2N-3 hindered light penetration and reduced the light absorption by the TiO2 core. According to oxygen temperature-programmed desorption (O2-TPD) analysis (Fig. 5f), the O2 desorption temperature of TiO2/C2N-2 was 195.0 °C, higher than that of TiO2/C2N-1 (180.8 °C) and TiO2/C2N-3 (189.6 °C). This suggests that TiO2/C2N-2 exhibited the strongest O2 adsorption among the samples, indicating that a moderate C2N thickness facilitates the mass transfer of O2 from solution to TiO2. NBT probe experiments show that the O2•− yields in UV/TiO2/C2N-2 were 1.5 and 1.4 times higher than those in UV/TiO2/C2N-1 and UV/TiO2/C2N-3, respectively (Supplementary Fig. 30), consistent with the qualitatively observed EPR intensities of O2•− (Fig. 5g). Therefore, TiO2/C2N-2, with its moderate C2N thickness, is more conducive to the adsorption and activation of O2. Moreover, the more negative Zeta potentials of TiO2/C2N-2 and TiO2/C2N-3 compared to TiO2/C2N-1 imply better repulsion of anions (Fig. 5h). These characterizations demonstrate that TiO2/C2N-2 achieves an optimal balance of OV concentration, light absorption, O2 adsorption, and surface electronegativity, all of which likely contribute collectively to its photocatalytic performance.

The PL intensity of TiO2/C2N-2 was remarkably lower compared to TiO2/C2N-1 and TiO2/C2N-3 (Supplementary Fig. 31a). Additionally, TRPL analysis reveals that TiO2/C2N-2 exhibited the shortest τave (Supplementary Fig. 31b), highlighting its superior charge separation and migration. Moreover, TiO2/C2N-2 achieved the highest k′CBZ values compared to TiO2/C2N-1 and TiO2/C2N-3 in both photocatalytic and photocatalytic FC activation systems (Supplementary Fig. 32). The UV/FC/TiO2/C2N-2 system generated the highest concentration of HO• through the reaction between O2•− and FC, as confirmed by the EPR signals for HO• (Supplementary Fig. 33). Interestingly, surface HO• on TiO2/C2N-1, TiO2/C2N-2, and TiO2/C2N-3 accounted for 62.4%, 82.7%, and 84.8% of the total HO• contribution to k′CBZ, respectively (Supplementary Fig. 34). This suggests that in UV/FC/TiO2/C2N-2 and UV/FC/TiO2/C2N-3, HO• is primarily confined within the nanoconfined reaction sites, whereas in UV/FC/TiO2/C2N-1, a larger amount of HO• is released into the bulk solution due to its thinner C2N shell.

As shown in Fig. 5i, both UV/FC/TiO2/C2N-2 and UV/FC/TiO2/C2N-3 exhibited high selectivity in CBZ degradation even in the presence of 5 mg/L NOM and 10 mM common anions (Cl–, HCO3–, and NO3–). Also, both systems maintained stable CBZ degradation under varying pH conditions and in tap water, river water, and WWTP effluent. In contrast, UV/FC/TiO2/C2N-1 showed an obvious reduction in CBZ degradation efficiency due to the interference from the water matrices. These findings underscore that a well-tuned C2N shell thickness can effectively balance key properties, such as light absorption, O2 diffusion, and the confinement of HO• within the nanostructure, thereby improving both efficiency and stability of photocatalytic processes. A thinner C2N shell enhances light absorption by the TiO2 core and promotes O2 diffusion into the nanoconfined space, boosting photocatalytic performance. However, thicker C2N shells provide better confinement of HO•, facilitating the selective degradation of micropollutants even in complex environmental matrices. Therefore, precise control over the C2N shell thickness is essential not only for optimizing photocatalytic efficiency but also for ensuring the long-term stability and adaptability of the photocatalyst in real-world applications.

Evaluation of DBP formation

In water treatment applications, the reaction between organic compounds (e.g., NOM) and FC or reactive chlorine species (RCS, e.g., Cl•, Cl2•−, and ClO•) can generate undesirable DBPs, which is a pressing issue. To evaluate this, the formation of DBPs and total organic chlorine (TOCl) was monitored in the presence of 2 mg/L NOM during the time needed to achieve 90% degradation of CBZ (Supplementary Fig. 35a). The UV/FC system produced DBPs at a concentration of 61.3 μg/L, with trichloroacetic acid (TCAA) and dichloroacetic acid (DCAA) determined as the dominant byproducts (Fig. 6a). Interestingly, the addition of TiO2, TiO2/C3N4, and TiO2/C2N-2 to the UV/FC system reduced DBP formation to 22.8, 14.1, and 0.9 μg/L, respectively. One key factor contributing to this reduction is the significantly lower FC exposure observed upon the addition of the photocatalysts (Supplementary Fig. 35b). Notably, the enhanced DBP control achieved with TiO2/C2N-2 compared to TiO2 and TiO2/C3N4 highlights the role of nanoconfinement in limiting NOM access to the nanoconfined space. Similarly, the TOCl formation in the UV/FC system, initially measured at 709.2 μg Cl/L, was decreased to 325.0, 181.7, and 19.3 μg Cl/L with the addition of TiO2, TiO2/C3N4, and TiO2/C2N-2, respectively (Fig. 6b).

Fig. 6: Formation of disinfection byproducts (DBPs) and total organic chlorine (TOCl) during the photocatalytic FC activation systems.
Fig. 6: Formation of disinfection byproducts (DBPs) and total organic chlorine (TOCl) during the photocatalytic FC activation systems.
Full size image

a, b Formation of DBPs and TOCl from 90% degradation of CBZ, respectively. c, d Formation of DBPs and TOCl within 30-min reactions, respectively. The error bars in this figure represent the standard deviation, calculated from at least two independent experiments.

Additionally, the generation of DBPs and TOCl was evaluated in the presence of 2 mg/L NOM over 30 min. In this case, FC exposure across different systems was relatively similar (Supplementary Fig. 36). The DBP formation in the UV/FC system reached 20.4 μg/L; however, the addition of TiO2 and TiO2/C3N4 increased DBP formation to 34.5 and 29.0 μg/L, respectively (Fig. 6c). The primary byproducts identified included TCAA, DCAA, 1,1,1-trichloropropanone (1,1,1-TCP), and trichloromethane (TCM). Notably, the addition of TiO2/C2N-2 resulted in a significant reduction of DBP formation, bringing it down to 1.0 μg/L. Moreover, the TOCl formation in UV/FC/TiO2/C2N-2 (32.6 μg Cl/L) was remarkably lower than that in UV/FC (271.1 μg Cl/L), UV/FC/TiO2 (413.0 μg Cl/L), and UV/FC/TiO2/C3N4 (391.6 μg Cl/L) (Fig. 6d). These results confirm that the nanoconfined effect effectively mitigates DBP formation through the exclusion of NOM from the reactive sites.

Economic cost-benefit evaluation

To assess the cost-effectiveness of the UV/FC/TiO2/C2N system under practical conditions, a comprehensive analysis of the 100-day operational costs was conducted and compared with those of the conventional UV/FC system (Supplementary Tables 5 and 6). In industrial practice, sorbents and heterogeneous catalysts are generally replaced or regenerated every 3–6 months, depending on influent water quality and operating conditions. Accordingly, our cost–benefit model incorporates different scenarios by assuming a material lifetime of 100 days, during which the catalyst is considered to maintain stable performance without replacement. As shown in Supplementary Fig. 37, over a 100-day operation period, the FC consumption in UV/FC/TiO2/C2N was estimated at 48 t, representing an 86.7% reduction compared to 360 t in UV/FC. This substantial decrease is attributed to the efficient photocatalytic activation of FC. Furthermore, the total treatment cost of UV/FC/TiO2/C2N was evaluated to be 0.038 $/m3, which is 78.9% lower than that of UV/FC (0.18 $/m3). This notable cost reduction is primarily owing to the decreased FC consumption. Therefore, the UV/FC/TiO2/C2N system provides significant economic advantages over the conventional UV/FC process, making it a more sustainable and cost-effective solution for water purification.

Discussion

In conclusion, we developed type II TiO2/C2N core-shell heterostructures to create nanoconfined catalytic sites, enabling the selective degradation of target micropollutants by effectively excluding NOM and anions via the nanopores of the C2N layer. The UV/FC/TiO2/C2N system achieved the highest k′/catalyst and k′/oxidant values for CBZ degradation among all surveyed photocatalytic advanced oxidation systems. More importantly, it consistently maintained nearly 100% micropollutant degradation efficiency across a wide range of NOM and anion concentrations, diverse pH conditions, and real water samples, whereas unconfined counterparts exhibited a 40%–80% performance decline. The precise tuning of the C2N shell thickness plays a pivotal role in optimizing light absorption, O2 diffusion, and HO• confinement, collectively enhancing the photocatalytic efficiency and selectivity. Furthermore, the formation of DBPs was successfully suppressed by restricting NOM access to the nanoconfined reaction sites.

Real-world water sources, such as drinking water, groundwater, domestic sewage, and high-salinity industrial wastewater, often contain varying concentrations of NOM and anions across a wide range of pH conditions. Our approach provides a robust and straightforward solution by engineering core-shell heterostructures to achieve efficient and selective micropollutant degradation in real-world environments. This strategy can inspire the development of next-generation core-shell catalysts for sustainable water treatment applications.

Methods

Chemicals

All chemicals used in this study are commercially available and are described in Supplementary Method 1.

Synthesis of TiO2/C3N4 and TiO2/C2N

Typically, 5 mg of melamine (precursor for C3N4) or HAT-CN (precursor for C2N) was dispersed in 10 mL of ethanol via ultrasonication. Next, 250 mg of TiO2 powder was added to the solution, and the mixture was ultrasonically mixed for 1 h. The solution was then dried in an oven at 80 °C for 8 h to obtain the precursor mixture TiO2/melamine or TiO2/HAT-CN. The mixture was heated in a tube furnace under a flow of argon gas at 100 mL min−1. The temperature was first increased from room temperature to 80 °C at a rate of 2 °C min−1, held at 80 °C for 60 min, then increased from 80 °C to 550 °C at 4 °C min−1, and maintained at 550 °C for 120 min to complete the polymerization. After calcination, the samples were allowed to cool naturally to room temperature and were subsequently collected. Additionally, TiO2 without carbon nitride shell was synthesized using the same procedure but without the carbon nitride precursor, and was used as a control.

To synthesize TiO2/C2N with varying C2N thicknesses, 250 mg of TiO2 powder was dispersed in 10 mL of ethanol using ultrasonication. Subsequently, 2.5 mg, 5 mg, and 10 mg of HAT-CN were added to the solution, respectively, and each mixture was ultrasonically mixed for 1 h. The resulting solutions were dried to obtain TiO2/HAT-CN, which were subsequently heated in a tube furnace under an argon flow (100 mL min−1) and naturally cooled to room temperature, yielding the final products TiO2/C2N-1, TiO2/C2N-2, and TiO2/C2N-3, respectively.

Catalyst characterization

The core-shell structures and elemental mapping of Ti, O, C, and N in the catalyst were analyzed using a JEM-2100F field-emission high-resolution TEM (JEOL, U.S.A.). The crystalline structures were characterized via XRD on a D8 Advance diffractometer with Cu Kα radiation (λ = 1.5406 Å) at 40 kV (Bruker, Germany). Surface properties were examined using a LabRAM Aramis Raman spectrometer (Horiba Scientific, Japan). The chemical structures of C3N4 and C2N were analyzed using solid-state 13C NMR spectroscopy (Agilent, U.S.A.). Element composition and oxygen vacancies were determined through XPS with an ESCALAB 250 spectrometer (Thermo Fisher, U.S.A.). Solid-state EPR signals and DMPO spin-trapped EPR signals were measured on an EMXplus EPR spectrometer (Bruker, Germany). Pore size distribution and specific surface area were evaluated via N2 adsorption–desorption tests at 77 K using Brunauer-Emmett-Teller (BET) analysis with a Belsorp-max instrument (MicrotracBEL, Japan). Light absorption properties were recorded using UV-vis DRS on a UV-2600 spectrometer (Shimadzu, Japan). Zeta potentials of the catalysts and the diameters of NOM were determined with an Omni multi-angle particle size and high-sensitivity Zeta potential analyzer (Brookhaven, U.S.A.), with the pH adjusted using H2SO4 and NaOH. Chemical adsorption of O2 was measured through O2-TPD using a Micromeritics AutoChem II apparatus, equipped with a computer-controlled CryoCooler and a thermal conductivity detector (Micromeritics, U.S.A.). Steady-state PL spectra and TRPL decay curves were obtained with an FLS 1000 photoluminescence spectrometer using a 365 nm excitation laser (Edinburgh, U.K.).

Photocatalytic experiments

Photocatalytic experiments were performed in a 100 mL glass reactor with continuous magnetic stirring. A light-emitting diode (LED) with an emission wavelength of 365 nm was used as the UV light source to minimize the activation of chlorine by UVB (280–320 nm) and UVC (200–280 nm) wavelengths. A water circulation system was applied to keep the temperature at 25 ± 0.2 °C. The degradation experiments were conducted in a 100 mL solution containing 5 μM micropollutants and 2 mM phosphate buffer to maintain a pH of 7. The solution was sequentially spiked with 10 mg/L catalysts and 2 mg/L FC (as Cl2), and simultaneously exposed to UV irradiation. Degradation samples were collected at predetermined time intervals and promptly quenched with excessive ascorbic acid, then filtered through a 0.22 μm membrane. The concentration of each scavenger was set to 5 mM. A low DO concentration (<0.5 mg/L) was achieved by purging nitrogen gas into the solution. The used catalysts were collected by centrifugation and washed three times with deionized water and ethanol, respectively, to remove surface organics before the next degradation cycle in tap water. The formation of DBPs and TOCl was monitored in a 100 mL solution containing 10 mg/L catalysts, 2 mg/L FC, 2 mg/L NOM, 5 μM CBZ, and 2 mM phosphate buffer (pH 7). After the reaction, three 20 mL samples were collected and promptly quenched, then filtered through a 0.22 μm filter membrane. The first and second portions were used to determine volatile DBPs and haloacetic acids, respectively, while the third portion was utilized to detect the formation of TOCl.

Analytical methods

The concentration of FC was determined using the titration method with N,N-diethyl-p-phenylenediamine sulfate and ferrous ammonium sulfate hexahydrate. The details of photoelectrochemical measurements are provided in Supplementary Method 2. NBT was employed as a probe to determine the concentration of O2•−, as it can be selectively reduced by O2•− to form formazan, which absorbs strongly at 560 nm2,37. NB, BA, CAF, and DMOB were employed as probe compounds to calculate the steady-state concentrations of surface and free radicals, along with their corresponding contributions to CBZ degradation (Supplementary Method 3). Concentrations of CBZ and the probe compounds were determined using a 1260 high performance liquid chromatography (HPLC, Agilent, U.S.A.). Degradation products of CBZ were determined using a Synapt G2-Si ion mobility ultrahigh-performance liquid chromatography system (Ion-Mobility-UPLC-QTOF-MS, Waters, U.S.A.), with detailed information in Supplementary Method 4. Concentrations of DBPs were determined using a 7890 A gas chromatograph (GC, Agilent, U.S.A.). TOCl was determined using an adsorbable organic halogen analyzer (Xplorer, Germany). The details of DFT calculation are provided in Supplementary Method 5.