Indium chloride (InCl₃·4H₂O, ≥99.99%), zinc sulfate heptahydrate (ZnSO₄·7H₂O, ≥99.5%), urea (CO(NH₂)₂, ≥99.99%) thioacetamide (TAA, ≥99.0%), 5,5-dimethyl-1-pyrroline-N-oxide (DMPO, ≥98%), 2,2,6,6-Tetramethylpiperidine-1-oxyl (TEMPO, 97%), hexadecyl trimethyl ammonium bromide (CTAB, 99%), ammonium oxalate (OA, ≥99.8%), bisphenol A (BPA, ≥99%), sodium sulfate (Na₂SO₄, ≥99%), benzoquinone (BQ, ≥99%), potassium chloride (KCl, ≥99.5%), isopropyl alcohol (IPA, ≥99.7%), barium sulfate (BaSO₄, ≥99.99%) and methyl viologen dichloride (MVCl₂, ≥99%) were purchased from Sinopharm Chemical Reagent Co., Ltd. and directly used.

To synthesize 2D g-C₃N₄ (CN), place 10 g CO(NH₂)₂ in the crucible, and then calcine 773.15 K for 4 h 2D-3D Zn₅In₂S₈ (ZIS) was synthesized by the modification of previous studies (Shen et al., 2010; Du et al., 2023). Specifically, 1.438 g of ZnSO₄·7H₂O, 0.586 g of InCl₃·4H₂O, 0.650 g of CTAB and a double excess of TAA were dissolved in deionized water (70 mL). The solution is to be transferred into a 100 mL hydrothermal reactor, with the temperature being maintained at 433.15 K for a period of 12 h. Subsequently, the yellowish-white precipitate is to be collected and washed with deionized water and ethanol. Finally, the precipitate is to be dried in a vacuum oven at 333.15 K, resulting in the formation of the 2D-3D ZIS (Supplementary Figure S1).

A total of 30 mg of sample powders was dispersed in 30 mL BPA solution (20 ppm BPA). This solution was stirred in darkness for 30 min to ensure equilibrium between desorption and adsorption. A 300 W Xe lamp, placed in a circular hollow chamber with water flowing through the annular casing, served as the visible light source (λ > 400 nm). At specified time intervals, 3 mL of the solution was extracted for analysis. The residual pollutants in the solution were measured using a UV-Vis spectrometer (Shimadzu UV 3600). The degradation efficiency (DE) was measured by using the formula: DE = (1 –Ct/Co) × 100%, where C_t and C_o are the BPA concentration after illumination time t and initial concentration of reactant, respectively. The rate of photocatalytic degradation (k) was assumed to follow pseudo-first-order kinetics, represented by the equation: C_t = C_oe^-kt.

The morphologies were analyzed using a transmission electron microscope (TEM, FEI Tecnai G2 F20) and a field emission scanning electron microscope (FESEM, Regulus 8200). Elemental mappings were conducted with an energy-dispersive X-ray spectrometer (EDX) attached to the SEM. X-ray diffraction (XRD) patterns were obtained using a Bruker D8 X-ray powder diffractometer with Ni-filtered Cu Kα radiation. X-ray photoelectron spectroscopy (XPS) measurements were performed on a Thermo Scientific ESCA Lab250 spectrometer with an Al Kα source. UV-vis diffuse reflectance spectra (DRS) of the samples were recorded on a UV-vis spectrophotometer (Shimadzu UV-3600), with BaSO₄ used as a reference material. Electron paramagnetic resonance (EPR) was measured using a Bruker A300. Typically, DMPO-methanol, TEMPO-water, and TEMPO-acetonitrile solution were applied to detect superoxide radicals, photoexcited electrons, and photoexcited holes, respectively. The solution containing 2 mg catalyst was loaded into a quartz tube for EPR measurements. The photoelectrochemical (PEC) properties were conducted by using an electrochemical workstation (CHI-660E) and were evaluated through photocurrent responses (i-t), open circuit potential-time (v-t) measurements, Mott-Schottky (M-S) tests, and Nyquist plots and Bode plots of electrochemical impedance spectroscopy (EIS). During PEC tests, A 300 W Xe lamp with a cutoff filter (λ > 400 nm) served as the visible light source. Ag/AgCl with 3 M KCl was used as the reference electrode. A Pt wire was applied as the counter electrode. FTO substrate deposited by the photocatalyst served as the work electrode. Na₂SO₄ aqueous solution (0.2 M) was used as the electrolyte for photocurrent, Mott-Schottky (M-S) plots, and photovoltage tests. KCl (0.1 M) contained K₃ [Fe(CN)₆]/K₄ [Fe(CN)₆] (0.01 M) aqueous solution was used for measuring electrochemical impedance spectroscopy (EIS) tests.

The crystal phase structures of carbon nitride (CN) and zinc indium sulfide (ZIS) were analyzed by using X-ray diffraction (XRD). As presented in Figure 1A, the characteristic peak at 27.3° can be assigned to the (002) plane of graphitic carbon nitride (g-C₃N₄) (Wang et al., 2009b). For ZIS, the diffraction peaks at 27.2°, 28.6°, 47.6°, and 56.3° were indexed to the (102), (104), (112), and (202) lattice planes of Zn₅In₂S₈, respectively (Shen et al., 2010; Du et al., 2023). The XRD spectra of both CN and ZIS are broadened, indicating their nano-structured nature. Based on the Debye-Scherrer formula (D = Kγ/Bcosθ. D, K, B, θ, and γ are the average grain size, Scherrer constant, half-height width of the diffraction peak of the measured sample, Bragg angle, and X-ray wavelength (1.54 Å), respectively.) (Akter et al., 2024), the average grain sizes of CN and ZIS were calculated to be about 10.0 nm and 12.7 nm, respectively. In addition, the dislocation density (δ) could also be obtained by the formula δ = 1/D² (Akter et al., 2024). The dislocation densities of CN and ZIS were calculated to be about 1.0 × 10⁻² and 6.2 × 10⁻³, respectively. These results indicate that CN and ZIS have fewer structural defects, and ZIS exhibits good crystal quality (the characterization results below will also prove it).

The nano-structural characteristics of CN and ZIS were visually observed by using Field Emission Scanning Electron Microscopy (FESEM) and Transmission Electron Microscopy (TEM). As presented in Figure 1B, CN displays a 2D nanosheet-like morphology, consistent with previous reports (Wang W. et al., 2024; Gao et al., 2021). Figure 1C shows the FESEM image of ZIS, exhibiting a 3D spherical structure with a rough surface. Further investigation through TEM images (Figures 1D, E) revealed that the 3D sphere of ZIS is composed of multiple thin nanosheets. The lattice fringes are clearly visible in the high-resolution TEM (HRTEM) image (Figure 1F). The lattice fringe of 0.32 nm (Figure 1G) is attributed to the (102) lattice plane of ZIS (Du et al., 2023). Additionally, the crystallographic characteristics can be verified by using the Fast Fourier Transform (FFT) pattern (Figure 1H). Energy-dispersive X-ray Spectroscopy (EDS) was carried out to investigate the constituent elements and their distribution in ZIS. As depicted in Figures 1I, J, the components of ZIS include Zn, In, and S, while C and O originate from the graphite conductive adhesive. EDS mapping confirms the uniform dispersion of indium (In), zinc (Zn), and sulfur (S) elements within the 2D-3D ZIS structure.

To further examine the composition and elemental valence of ZIS, X-ray Photoelectron Spectroscopy (XPS) was conducted. Figure 2A is the survey spectrum of ZIS, confirming the presence of indium (In), zinc (Zn), and sulfur (S) elements. The C element and C-O species (286.2 eV) are attributed to the graphite conductive adhesive (Li X. et al., 2019). The binding energies of the constituent elements were adjusted based on the C 1s peak at 284.6 eV (Figure 2B). The fine-structured Zn 2p spectrum was deconvoluted into two peaks at approximately 1,044.6 eV and 1,021.5 eV (Figure 2C), which can be assigned to Zn 2p_1/2 and Zn 2p_3/2, respectively (Liu et al., 2021; Du et al., 2023). In addition, the splitting energy between 2p_1/2 and 2p_3/2 is 23.1 eV, indicating a Chemical valence of Zn²⁺ in the ZIS photocatalyst. The fine-structured XPS spectrum for In species shows peaks at binding energies of 452.0 eV and 444.4 eV (Figure 2D), which can be indexed to In 3d_3/2 and In 3d_5/2, respectively (Liu et al., 2021; Du et al., 2023). The splitting energy between 3d_5/2 and 3d_3/2 is 7.6 eV, which corresponds to the In³⁺ state in the ZIS. The binding energy peaks for S 2p are observed at about 161.3 eV and 162.5 eV (Figure 2E), with a splitting energy of 1.2 eV, associated with S^2- 2p_3/2 and S^2- 2p_1/2, respectively (Liu et al., 2021; Du et al., 2023). Thus, the above analysis indicates that a 2D-3D hierarchical structure of ZIS nanosheets and spheres has been successfully synthesized.

CN and ZIS’s optical properties and band energy structures were studied using UV-vis diffuse reflectance spectroscopy (DRS), Tauc plots, and Mott-Schottky (M-S) tests. Light absorption spectra of CN and ZIS are displayed in Figure 3A. Compared to CN, ZIS exhibits increased light absorption intensity. The enhanced light absorption in the range of less than 450 nm falls to its intrinsic absorption, while the boosted light absorption larger than 450 nm may be attributable to the rough surface of the 2D-3D ZIS nanosheet-sphere hierarchical structure. The band gap energy (E_g) can be obtained by the Tauc equation: αhv = A (E - E_g)ⁿ. The product of h and v is energy (E = hv). A is the proportionality constant, h is the Planck constant, α is the absorption coefficient, and v is the light frequency (Zhang et al., 2024). The values of n equal 1/2 or 2 for semiconductors with direct absorption or indirect absorption, respectively. As depicted in Figure 3B, the band gap energies of ZIS and CN are about 2.90 eV and 2.82 eV, respectively, which are close to the reported values (Liu et al., 2015; Kalomiros et al., 1987; Du et al., 2023). On the basis of the above results, it is interesting that, compared to CN, ZIS possesses strong light absorption and wide band gap energy. High light absorption and wide band gap energy would help to improve the generation and separation of photoexcited charge carriers, respectively. To demonstrate the energy band positions [valence band (VB) and conduction band (CB)] of CN and ZIS, Mott-Schottky (M-S) measurements and the formula (E_g = E_VB - E_CB) could be used. Firstly, the flat band (FB) could be obtained by the M-S formula: 1/C² = 2(V–V_FB–kT/e)/(N_Deεε₀A²) (Sharma et al., 2024). C, V, V_FB, T, A, k, e, N_D, ε, and ε₀ are the capacitance, applied potential, FB potential, absolute temperature, area, Boltzmann constant, electron charge, carrier density, dielectric constant and vacuum permittivity, respectively. For n-type semiconductors, the Fermi level (FB) is typically about 0.1 V below the conduction band (CB) (Che et al., 2024). As illustrated in Figures 3C, D, the slopes of the Mott-Schottky (M-S) plots for both CN and ZIS are positive. The results demonstrate that both ZIS and CN are n-type semiconductors. The FB values for CN and ZIS are approximately −1.05 V and −1.04 V (vs. NHE), respectively. Consequently, the CB potentials for CN and ZIS are about −1.15 V and −1.14 V, respectively, while the valence band (VB) potentials are around 1.67 V for CN and 1.76 V for ZIS (as shown in Figure 3E). The CB potential of ZIS is similar to that of CN and is more negative than the potential for producing superoxide radicals from oxygen (E (O₂/•O₂ ⁻) = −0.33 V) (Che et al., 2024). These results indicate that the reduction potential of photogenerated electrons in the CB of ZIS is comparable to that of CN, allowing for the reduction of oxygen molecules to generate superoxide radicals, which are active species involved in photocatalytic degradation. Although the oxidizing potentials of photoexcited holes in both CN and ZIS are insufficient to produce hydroxyl radicals (with an energy of E (•OH/H₂O) = 2.40 V) (Che et al., 2024), the oxidizing potential of the photoexcited holes in the VB of ZIS is stronger than that of CN. Given these results, ZIS has stronger light absorption, larger band gap energy, and a higher VB potential compared to CN, facilitating greater photoexcited charge production, better charge separation, and enhanced degradation of BPA.

The photoelectrical properties of CN and ZIS were examined through a series of photoelectrochemical tests, including photocurrent measurements, electrochemical impedance spectroscopy (EIS), and photovoltage assessments. As illustrated in Figures 4A,B, the photocurrent density of ZIS is greater than that of CN under visible light irradiation. This observation indicates a higher yield and more efficient separation and transport of photoexcited electron-hole pairs. Specifically, the charge separation and transport efficiency (η) can be evaluated by using the empirical formula: η = J_H2O/J_MVCl2 (Wan et al., 2022). J_H2O and J_MVCl2 are photocurrent densities of a photocatalyst without and with the addition of methyl viologen dichloride (MVCl₂), respectively. As expected, ZIS (η(ZIS) = 70.8%) exhibits higher separation-transport efficiency than CN (η(CN) = 52.2%). The charge separation-transport efficiency of ZIS is about 1.356 times of CN. Figure 4C exhibits EIS Nyquist plots of ZIS and CN. The smaller arc radius means the more effective charge separation and the faster charger transportation (Meng et al., 2021). Outwardly, in contrast to CN, ZIS has a smaller arc radius, suggesting the charge separation and transfer of ZIS is superior to CN. Figure 4D is the Bode phase plots at the open-circuit voltage. According to the relationship between the lifetime of electron recombination with a time constant (t_e) and the characteristic maximum frequency (t_e = 1/(2πf_max)) (Zhou et al., 2014), the characteristic maximum frequency peak (f_max) of ZIS (∼118.9 Hz) decrease obviously with respect to CN (∼262.1 Hz), manifesting that ZIS possesses about 2.2 times enhancement of electron lifetime than CN. Consequently, a low recombination rate of photoexcited charge carriers is highly desired to achieve in ZIS. To investigate the origin of the improved photoelectrochemical performance of ZIS, we conducted open circuit photovoltage decay (OCPV) measurements. The OCPV technique is utilized to study the lifetime of photoelectrons and the recombination rate of photoexcited charge carriers (Ning et al., 2018). As shown in Figure 4E, ZIS exhibits a higher photovoltage response compared to CN. This result aligns with the findings from photocurrent response and electrochemical impedance spectroscopy (EIS) tests, further confirming the efficient charge separation and transport in ZIS. To gain more insights, we evaluated the average photoelectron lifetime (τ_a) using a specific equation: τ_a = (k_BT/e)/(dV_OC/dt) (Ning et al., 2018). T, k_B, e, and t are the temperature, Boltzmann constant (1.380610 × 10⁻²³ J K⁻¹), electron charge (1.602 × 10⁻¹⁹ C), and time, respectively. Obviously, ZIS possesses a significantly prolonged lifetime of photoexcited electrons in comparison with CN (Figure 4F), thus contributing to the remarkably improved photoelectrochemical properties. These results imply that the photocatalytic performance of ZIS would be superior to CN.

The performances of CN and ZIS for the photocatalytic degradation of BPA under visible light illumination are illustrated in Figure 5. To study the photocatalytic performance of the catalysts, initial tests for blank photolysis (without catalysts) and dark adsorption were conducted. As shown in Figure 5A, the results of the photolysis demonstrate that the photo-induced self-degradation of BPA is negligible, which aligns with the persistent nature of BPA. On the other hand, the dark adsorption results indicate that ZIS exhibits superior adsorption capabilities compared to CN, likely due to the 2D-3D hierarchical structure of ZIS. Under visible light irradiation, the concentration of BPA decreases rapidly, with a reduction of nearly 92.3% after 120 min of illumination. In comparison, BPA removal efficiencies of CN (Figure 5A), ZnIn₂S₄ (Supplementary Figure S2), and In₂S₃ (Supplementary Figure S2) are only 31.3%, 72.6%, and 40.4%, respectively. In addition, the BPA removal efficiency of ZIS is superior to that of the majority of the reported catalysts (Supplementary Table S1). The remarkable reactivity of ZIS is further supported by a comparison of the apparent reaction rate constants (denoted as k). The linear relationship of ln (C₀/C) versus time (t) suggests that the degradation of BPA follows pseudo-first-order reaction kinetics (Supplementary Figure S3). The k values for CN and ZIS are approximately 0.36 h^-1 and 2.36 h^-1, respectively (as shown in Figure 5B). The k value for ZIS is 6.56 times higher than that for CN, highlighting ZIS’s superiority in photocatalytic degradation of BPA. Moreover, the total organic carbon (TOC) results showed that the mineralization rate of BPA over ZIS can reach about 63.5% under visible light irradiation for 120 min (Supplementary Figure S4). To confirm the stability and reusability of this photocatalyst, recycling experiments were conducted using ZIS for the photocatalytic degradation of BPA. As illustrated in Figure 5C, the photodegradation efficiency for BPA over reused ZIS remains comparable to that over fresh ZIS after three cycles. Additionally, the morphology (Figure 5D), the XRD pattern (Supplementary Figure S5), and the light absorption spectrum (Supplementary Figure S6) of the used ZIS are similar to that of the fresh ZIS, indicating that no significant changes occurred before and after the photocatalytic reaction. Thus, ZIS proves to be a stable and promising photocatalyst for the degradation of BPA.

In the context of photocatalytic degradation of organic pollutants, hydroxyl radicals (•OH), photogenerated electrons (e⁻), superoxide radicals (•O₂⁻), and photoexcited holes (h⁺) are considered active species (Song et al., 2022; Habibi Zare and Mehrabani-Zeinabad, 2023). A scavenger study was conducted to demonstrate the roles of these active species and their effect on ZIS for BPA degradation. Benzoquinone (BQ), ammonium oxalate (AO), isopropanol (IPA), and oxygen (O₂) serve as trapping agents for •O₂⁻, photogenerated h⁺, •OH, and photogenerated e⁻, respectively (Song et al., 2022; Habibi Zare and Mehrabani-Zeinabad, 2023). As shown in Figure 6A, the degradation efficiency of BPA is obviously inhibited by the addition of BQ (decreasing to 46.4%) and AO (decreasing to 70.3%), while IPA has minimal impact on BPA degradation. These results indicate that photogenerated holes and superoxide radicals play more critical roles than hydroxyl radicals in the degradation of BPA over ZIS. The limited effect of IPA is expected since ZIS does not generate hydroxyl radicals. Notably, when oxygen is introduced into the reaction system, almost complete degradation efficiency is achieved (99.9%). Oxygen can trap electrons, which not only creates superoxide radicals but also enhances the separation of photoexcited charges.

To verify the generation of superoxide radicals and charge separation, electron paramagnetic resonance (EPR) technology was employed (Che et al., 2024; Lei et al., 2024), as presented in Figures 6B–D. Figure 6B demonstrates that under visible light irradiation, characteristic sextet peaks of the DMPO-•O_₂⁻ adduct can be observed, demonstrating the generation of superoxide radicals on the surface of ZIS. Additionally, TEMPO shows triplet EPR peaks. The intensity of the TEMPO EPR signal decreases when photoexcited electrons or holes are captured by TEMPO (Lei et al., 2024). Hence, TEMPO is utilized to detect photoexcited electrons and holes, and a reduction in intensity signifies photogenerated charge separation. As presented in Figures 6C, D, the TEMPO EPR intensities for holes and electrons are reduced by about 38.8% and 37.5% under light irradiation for 10 min, respectively. The TEMPO EPR peaks diminish upon detecting photoexcited holes and electrons, indicating that photoexcited charge carriers in ZIS can be effectively separated and transported to the surface for the target redox reactions.

In light of the results and analyses above, a photocatalytic mechanism for ZIS is proposed and illustrated in Figure 6E. The unique 2D-3D hierarchical structure of ZIS, with its rough surface, enhances light absorption and generates electron-hole pairs when exposed to visible light. The separated holes can directly participate in BPA degradation, while the separated electrons are captured by oxygen molecules to produce superoxide radicals, which also play a role in the degradation process of BPA.