In situ interfacial engineering of 1D Bi₂S₃/2D g-C₃N₄ heterostructures for antibiotics degradation in aqueous media via light mediated peroxymonosulfate activation

Abstract

Interfacial engineering between metal sulfides (MS) and graphitic carbon nitride (g-C₃N₄) offers a promising strategy to design semiconductors for the efficient degradation of persistent water pollutants. However, conventional multi-step methods used to prepare MS/g-C₃N₄ heterostructures often result in weak interfacial interactions between the building blocks, thereby leading to inefficient charge separation and sub-optimal catalytic performance. To overcome this limitation, we present here a novel single-step strategy for the in situ preparation of 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructures, producing intimate interactions between the 1D and 2D architectures as evidenced by experimental and theoretical findings. Remarkably, these robust interfacial interactions establish a strong internal electric field (IEF), favoring spatial separation of high charge flux at the 1D/2D interface via an S-scheme mechanism. Importantly, the lowered charge transfer barrier at the interface speeds up the activation kinetics of peroxymonosulfate (PMS) and O₂, to achieve a high tetracycline degradation efficiency of 98.5% with a rate constant of 0.06 min⁻¹. DFT calculation results reveal that the effective coupling between the 1D/2D counterparts induced a charge redistribution and electron density accumulation at the interface, facilitating cleavage of the O–O bond in PMS and O₂. Furthermore, DFT calculations identified a unique PMS adsorption configuration on Bi sites and revealed the competence of S atoms in activating the peroxide bond in PMS. This work offers a cost-effective and environmentally friendly approach for the rational engineering of interfacial interactions in MS/g-C₃N₄ heterostructures, enabling highly efficient applications in energy and environmental remediation.

---

New concepts

This communication introduces a new single-step in situ synthesis approach to forming 1D Bi₂S₃(n)/2D g-C₃N₄ S-scheme van der Waals (vdW) heterostructures. The in situ 1D/2D integration allowed the strong interfacial coupling, thereby enabling us to address the poor interfacial interactions faced by the traditional multi-step methods. Furthermore, in this work we clarified the underlying role of intimate interfacial interactions in constructing a robust internal electric field (IEF) for activating the S-scheme charge separation mechanism and decreasing the interfacial charge transfer barrier. Through DFT calculations we demonstrated that charge accumulation near the 1D Bi₂S₃/2D g-C₃N₄ facilitated PMS adsorption and lowered the energy barrier for O–O bond cleavage in PMS, while electronically modifying Bi and S sites served as pivotal catalytic centers. Beyond its superior catalytic activity, this work presents a scalable green approach to engineer metal sulfide/g-C₃N₄ interfaces, offering a universal design strategy for advanced nanomaterials via atomic-level interfacial control. Integrating material design (1D/2D coupling), mechanistic understanding (IEF-driven S-scheme), and practical application (PMS-based advanced oxidation processes), our findings provide significant advances in water purification technologies.

1. Introduction

Rampantly increasing consumption of stubborn antibiotics as medicinal therapy against life-threatening microbial infections has triggered the emergence and transmission of multidrug resistant (MDR) bacterial strains in water reservoirs,^1–3 thus imposing detrimental effects upon both human health and aquatic life.⁴ Considering the severity of the situation, the World Health Organization (WHO) has declared the sub-lethal accumulation and dissemination of antibiotic resistant organisms as an escalating threat to human life,⁵ forecasting a MDR bacterial associated mortality toll of 10 million by the year 2050.^6,7 There is therefore a dire need for a globally collaborative approach from policymakers to mitigate the transmission of MDR strains and prevent waterborne infections. Up to now, adequate degradation and mineralization of refractory antibiotics by conventional wastewater treatment technologies have remained daunting challenges and demand urgent development of a highly efficient and economically feasible technology capable of effectively eliminating antibiotics to address the ongoing safe drinking water crisis.

Advanced oxidation processes (AOPs),^8,9 specifically persulfate based (PS-AOP) ones, have emerged as a cutting edge next generation water treatment technology to effectively degrade a wide spectrum of persistence micropollutants, including antibiotics.^10,11 Compared to the conventional hydroxyl radicals process (˙OH), sulfate SO₄˙⁻ radicals possess higher redox potentials (2.5–3.1 eV), longer half-lives (30–40 μs) and wider pH (2–8) operative windows.^12,13 PS-AOP could completely disintegrate antibiotics into CO₂ and H₂O rather than simply oxidizing them into non-toxic intermediates.^14,15 In practice, strong oxidants such as peroxymonosulfate (PMS) and peroxydisulfate (PDS) with a half-life span of 3–4 × 10⁻⁵ s are employed to generate multiple reactive oxygen species (ROS) through appropriate activation means such as catalysts, light, heat, radiations, etc.^16,17 In comparison with PDS, PMS based PS-AOP has gained more popularity due to the asymmetric geometry of the PMS molecule (H–O–O–SO₃⁻), which renders it to serve as an electron acceptor as well as donor to activate superoxide bonds (O–O, 1.326 Å) for the non-selective generation of free radicals (SO₄˙⁻, ˙OH, and O₂˙⁻) and non-free radicals (¹O²).¹⁸ In recent years, light mediated heterogeneous PMS activation has been demonstrated as an excellent strategy to dramatically improve antibiotics degradation efficiency.¹⁹ This triggers immense research focus because of the natural abundance of solar light, simple operation protocol and high efficiency. Nevertheless, rational design of a robust catalyst that can efficiently capture visible light to generate charge carriers is of tremendous importance for effective light-induced PMS activation.

Since 2009,²⁰ metal-free conjugated polymeric g-C₃N₄ featuring an appealing electronic structure, suitable band gap and superior physiochemical stability has been the material of choice for a breadth of visible light driven catalysis including remediation of hazardous water pollutants.^21–23 However, pristine g-C₃N₄ exhibits inferior PMS activation ability because of poor redox active sites,²⁴ sluggish charge carrier separation,²⁵ inadequate visible light absorption²⁶ and low electric conductivity.²⁷ Hitherto, elemental doping,²⁸ constructing M–N_x atomic sites^29,30 and heterostructure engineering with other suitable semiconductors has been explored in the pursuit of boosting PMS activation and degradation performance.³¹ Interfacial engineering of g-C₃N₄ with narrow bandgap metal oxides, metal chalcogenides, metal phosphides and metal carbides has been developed rapidly in recent years with the aim of constructing catalysts with good photocatalytic activity through simultaneously improved visible light utilization and amplified charges transportation.^32,33 Amongst the semiconductors, metal sulfides,³⁴ especially Bi₂S₃, an n-type narrow bandgap (1.3–1.7 eV) semiconductor, could be the best selection due to its distinct advantages such as low cost, non-toxicity and strong absorption capability in the visible and near-infrared region.^35,36

Given the distinct advantages of g-C₃N₄ and Bi₂S₃ individually, constructing a Bi₂S₃/g-C₃N₄ heterostructure interface is a plausible strategy to significantly enhance catalytic performance.^37,38 However, the synthesis of Bi₂S₃/g-C₃N₄ heterostructures has up until now relied on tedious multi-step methods, often requiring expensive precursors and post anchoring of Bi₂S₃ nanostructures onto the surface of pre-synthesized g-C₃N₄.^39,40 Additionally, most reported Bi₂S₃/g-C₃N₄ heterojunctions suffer from inhomogeneous distribution of Bi₂S₃ and weak chemical or physical interfacial interactions, resulting in sub-optimal photocatalytic performance. Strong interfacial interactions are critical in regulating the electronic structure and charge transport properties. Such interactions could form an internal electric field (IEF) and enhance the adsorption behavior of PMS and antibiotics, thereby accelerating reaction kinetics of PMS-activated AOPs. Consequently, developing a strategy to engineer a robust Bi₂S₃/g-C₃N₄ heterojunction is highly desired to address the weak interfacial interaction challenges and to optimize the photocatalytic performance.

Herein, we present a facile single-step gas–solid reaction strategy for the in situ synthesis of a series of strongly attached 1D Bi₂S₃ chains on 2D g-C₃N₄ nanosheets with tunable Bi₂S₃ content. Our experimental evidences revealed that the interfacial interactions could be appropriately controlled by regulating the growth density of 1D Bi₂S₃ onto the 2D g-C₃N₄ sheets. The synthesized 1D/2D heterostructures exhibited optimal interfacial coupling, enhanced charge separation efficiency and appropriate number of active sites, thereby resulting in great antibiotic degradation performances in light mediated PMS-AOP processes. Density functional theory (DFT) simulations have unveiled the electronic charge redistribution on the Bi and S atoms at the 1D/2D interface. DFT calculations further demonstrated strong PMS adsorption on the electronically modified Bi and S atoms via distinct configurations, stimulating ROS generation and thereby enhancing antibiotic degradation activity.

2. Experimental section

2.1. Materials

Details of the chemicals and reagents employed in this study are provided in the ESI.†

2.2. Synthesis protocol of 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructures

Synthesis of 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructures with varying 1D Bi₂S₃ content was accomplished via a single step strategy. In detail, bismuth acetate (1 mM, 2 mM and 3 mM) was dissolved in 20 mL DI water under stirring in a 50 mL centrifuge tube to make a yellow solution. Next, 3 g of thiourea and 2 g of urea was added into the above yellow solution and stirred at 40 °C until urea and thiourea completely dissolved. Then, the resulting solution was freeze-dried using liquid nitrogen and kept in a lyophilizer at −50 °C for three days. The obtained yellow solid was crashed into powder form and subjected to pyrolysis in a tube furnace at 560 °C under an inert atmosphere (Ar) at a heating rate of 4 °C minute⁻¹ for 2 hours. After cooling to normal temperature, the samples were collected and designated as 1D Bi₂S₃(n)/2D g-C₃N₄, where n = 0.1, 0.2, and 0.3 represent the feed Bi concentration in mM. A control pristine 2D g-C₃N₄ sample was prepared by mixing 3 g thiourea and 2 g urea under the same conditions used to prepare the heterostructures. The procedure for synthesizing the pristine 1D Bi₂S₃ sample is given in the ESI.†

2.3. Characterizations

Details of the instruments used for the characterization of the prepared samples and procedures for evaluating degradation activity are described in the ESI.†

2.4. Density functional theory (DFT) calculations

Details of DFT calculations are provided in the ESI.†

3. Results and discussion

3.1. Synthesis, morphological and structure analysis and DFT structure optimization

Fig. 1a schematically illustrates the preparation of 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructures via the in situ solid–gas phase reaction. In this process, urea and thiourea were employed as the sources of carbon (C), nitrogen (N), and sulfur (S), while bismuth acetate served as the precursor for the bismuth (Bi) ions. Three representative heterostructures were synthesized and labeled as 1D Bi₂S₃(n)/2D g-C₃N₄, where n represents the feed concentration of Bi in millimoles (0.1, 0.2, and 0.3 mM). The formation mechanism of these heterostructures involves firstly, the polymerization of urea and thiourea into 2D g-C₃N₄ nanosheets along with the release of reactive CS₂ gas;⁴¹ and the simultaneous decomposition of bismuth acetate into Bi₂O₃ under the pyrolysis conditions. Next, the in situ solid–gas reaction between Bi₂O₃ and CS₂ resulted in the growth of 1D Bi₂S₃ nanostructures onto the 2D g-C₃N₄ surfaces to ultimately construct the 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructures.

Fig. 1 (a) Schematic illustration of the synthesis process of 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructures. Detailed analysis of the morphology of the 1D Bi₂S₃(0.2)/2D g-C₃N₄ sample: (b) SEM, (c) TEM, (d) HAADF-STEM, and (e)–(h) respective EDX elemental mapping of C, N, Bi and S, and (i) HR-TEM image showing the lattice spacing. DFT optimized models for: (j) 1D Bi₂S₃, (k) 2D g-C₃N₄, and (l) and (m) 1D Bi₂S₃/2D g-C₃N₄, respectively along the α and β directions.

Field-emission scanning electron microscopy (FE-SEM) images of the as-obtained 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructures (Fig. 1b and Fig. S1, ESI†) clearly revealed the presence of 1D Bi₂S₃ nanostructures uniformly distributed on the surface of 2D g-C₃N₄. As a comparison, the pristine g-C₃N₄ prepared under identical conditions gave a typical 2D sheet-like structure (Fig. S2a, ESI†), while the prepared pristine Bi₂S₃ gave a typical 1D morphology (Fig. S2b, ESI†). Notably, as the Bi³⁺ concentration was increased from 0.1 to 0.3 mM, a corresponding increase in the density of 1D Bi₂S₃ nanostructures could be observed on the 2D g-C₃N₄ surfaces (Fig. S1a–c, ESI†). We illustrate detailed morphological analysis of the 1D Bi₂S₃(0.2)/2D g-C₃N₄ sample in Fig. 1b-i, as this n = 0.2 mM feed ratio resulted in the highest catalytic performance in subsequent study. The SEM (Fig. 1b), TEM (Fig. 1c) and HAADF-STEM (Fig. 1d) images of the 1D Bi₂S₃(0.2)/2D g-C₃N₄ sample revealed good contact between the 1D Bi₂S₃ formed in situ onto the 2D g-C₃N₄ nanosheets. Energy-dispersive X-ray (EDX) mapping revealed decent spatial distribution of the Bi and S composition along C and N elements in the 1D Bi₂S₃(0.2)/2D g-C₃N₄ heterostructure (Fig. 1e–h).

In contrast, no such close contact was observed in the SEM (Fig. S3a, ESI†) and TEM (Fig. S3b, ESI†) images taken of a physically mixed 1D Bi₂S₃ and 2D g-C₃N₄ sample. The HR-TEM micrograph (Fig. 1i) clearly unveils the presence of distinct amorphous g-C₃N₄ and crystalline Bi₂S₃ phases, confirming the formation of an intimate heterostructure interface that is of outmost importance to set up IEF. The lattice fringes of 0.37 nm and 0.34 nm spacing correspond well to the d-spacing of the (130) planes of orthorhombic Bi₂S₃, further substantiating the successful formation of the 1D Bi₂S₃ structure.

Spin-polarized DFT simulations were performed to further investigate the structure synthesized. The heterostructure model was constructed with optimized structures of a 1 × 9 1D Bi₂S₃ molecular chain (Fig. 1j) and a 5 × 5 supercell of 2D g-C₃N₄ (Fig. 1k) with a lattice mismatch of less than 3%. As there exists two binding facets of 1D Bi₂S₃ (α and β, as indicated in Fig. S4a, ESI†), either of these facets could be oriented along the a, b or diagonal direction of the 2D g-C₃N₄ surface (Fig. S4b, ESI†). Among the models simulated, the α/diagonal configuration, designated as 1D Bi₂S₃/2D g-C₃N₄-α (Fig. 1l), was found to give stable van der Waals (vdW) interactions at an average interface space of about 3.38 Å and binding energy of −6.94 eV. In a similar manner, the β/diagonal configuration (1D Bi₂S₃/2D g-C₃N₄-β, Fig. 1m) was obtained with a binding energy of −7.04 eV. These energetically almost degenerate 1D Bi₂S₃/2D g-C₃N₄-α and 1D Bi₂S₃/2D g-C₃N₄-β configurations suggest that both heterostructures are stable and could be formed experimentally.

The crystal structure and phase composition of 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructures were further confirmed by executing powder X-ray diffraction (XRD) analysis (Fig. 2a). The XRD pattern of 2D g-C₃N₄ exhibited two characteristic peaks at 13.1° and 27.6° for the graphitic layers of the (002) and (100) planes respectively. On the other hand, the pristine 1D Bi₂S₃ showed diffraction peaks that could be indexed to the crystal planes of the orthorhombic Bi₂S₃ phase (PDF no. 17-0320). All the 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructures exhibited only diffraction peaks that corresponded to 1D Bi₂S₃ and 2D g-C₃N₄, confirming the successful fabrication with high phase purity. Notably, the diffraction peaks of the 1D Bi₂S₃ phase appeared sharper in the 1D Bi₂S₃(n)/2D g-C₃N₄ sample as compared to the pristine 1D Bi₂S₃ sample, indicating better crystallinity in the heterostructures. Furthermore, the peak intensity of the 1D Bi₂S₃ diffraction increased linearly with higher Bi³⁺ feed concentrations, while the 2D g-C₃N₄ peak intensity decreased steadily, as expected with increasing growth of 1D Bi₂S₃ on the surface of 2D g-C₃N₄.

Fig. 2 Comparison of 1D Bi₂S₃, 2D g-C₃N₄ and 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructures: (a) XRD patterns, (b) and (c) FTIR spectra, XPS profiles of (d) C 1s, (e) N 1s and (f) Bi 4f peaks respectively. DFT simulated work function (WF, φ) for: (g) 2D g-C₃N₄, (h) 1D Bi₂S₃ and (i) 1D Bi₂S₃/2D g-C₃N₄ heterostructure respectively. Bader charge analysis for the optimized (j) 1D Bi₂S₃/2D g-C₃N₄-α, and (k) 1D Bi₂S₃/2D g-C₃N₄-β structures.

The effect of in situ growth of 1D Bi₂S₃ on the chemical structure of 2D g-C₃N₄ was investigated by FTIR spectroscopy (Fig. 2b). In the case of pristine 1D Bi₂S₃, the absorption band observed in the range 510–640 cm⁻¹ was assigned to the stretching mode of the Bi–S bond.⁴² The FTIR peaks of 2D g-C₃N₄ at ∼800 cm⁻¹ and 1200–1700 cm⁻¹ were respectively attributed to the bending vibrations of triazine/heptazine rings and the stretching modes of CN heterocycles within these rings.⁴³ Specifically, the peaks at 1320 cm⁻¹ and 1232 cm⁻¹ correspond to the stretching vibrations of N–(C)₃ and C–NH–C linking units within the tri-s-triazine structure.⁴⁴ Broader peaks in the 2900–3400 cm⁻¹ region could be associated to the stretching vibrations of terminal amine groups (–NH and –NH₂) and hydroxyl (OH) groups of physically adsorbed water molecules.⁴⁵ Notably, all the 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructures retained the basic chemical skeleton of g-C₃N₄, as evidenced by the retention of characteristic absorption bands. However, as can be seen from the magnified FTIR spectra (Fig. 2c), the characteristic stretching vibrations of C=C and C–C heterocycles in the heterostructures exhibited a slight blue shift, suggesting the presence of interfacial interactions and electronic charge redistribution between the 1D Bi₂S₃ and 2D g-C₃N₄ counterparts.⁴⁶

XPS analysis was conducted to ascertain the surface chemical structures and affirm the electronic interactions within the 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructures. The XPS binding energy (B.E.) of all the peaks was calibrated with respect to C 1s peak at 284.8 eV. Apart from the dominant peaks of C 1s and N 1s, XPS survey scans of the 1D Bi₂S₃(n)/1Dg-C₃N₄ heterostructures (Fig. S5, ESI†) displayed minor peaks for Bi 4f (157.4 eV) and S 2s (225.01 eV), thus confirming the successful integration of 1D Bi₂S₃ on 2D g-C₃N₄ as observed by EDX mapping. In agreement with the XRD and SEM results, 1D Bi₂S₃(0.3)/2D g-C₃N₄ exhibited comparatively intense Bi 4f and S 2s peaks as compared to the 1D Bi₂S₃(0.1)/2D g-C₃N₄ and 1D Bi₂S₃(0.2)/2D g-C₃N₄ samples. The C 1s spectrum of 2D g-C₃N₄ (Fig. 2d) could be fitted into two well-resolved Gaussian–Lorentzian components, assignable respectively to the sp² hybridized C–C (284.8 eV) in the graphitic structure and sp² carbon of the C–N=C (288.2 eV) bond in the s-triazine ring.⁴⁷ Interestingly, a blue shift in the C 1s B.E. of the C–N=C component was noted for the heterostructures, suggesting a modification in the electronic density around the carbon in the C–N=C bond.

The N 1s spectrum of 2D g-C₃N₄ was deconvoluted into three component peaks (Fig. 2e). The peaks at 398.9 eV and 400.9 eV were respectively attributed to the sp² hybridized N atom of the C–N=C bond and tertiary nitrogen of N–(C)₃,⁴⁸ while the peak at 404.8 eV could be ascribed to π–π* excitations between the stacking 2D g-C₃N₄ interlayers or amino functional group (–NH₂).^49,50 Interestingly, the N 1s B.E. of the C–N=C component peak red-shifted to varying degrees in the heterostructures, with the apparent shift (Δ) decreasing in the order: 1D Bi₂S₃(0.1)/2D g-C₃N₄ (Δ = 0.32 eV) > 1D Bi₂S₃(0.2)/2D g-C₃N₄ (Δ = 0.2 eV) > 1D Bi₂S₃(0.3)/2D g-C₃N₄ (Δ = 0.1 eV). This N 1s B.E. trend provides exclusive evidence for the existence of robust interfacial electronic interactions between 1D Bi₂S₃ and 2D g-C₃N₄, which triggered the electronic charge modification on the N atom of the C–N=C bond and the nearby C atom.

As shown in Fig. 2f, the Bi 4f XPS profile of 1D Bi₂S₃ exhibited two peaks for the characteristic Bi 4f_7/2 (158.5 eV) and Bi 4f_5/2 (163.8 eV) with spin orbital splitting of 5.3 eV, indicating the existence of Bi³⁺.⁵¹ Notably, Bi 4f B.E. of the heterostructures exhibited a blue shift, suggesting a redistribution of electronic charge on Bi. The Bi 4f peaks shifted in the order of 1D Bi₂S₃(0.1)/2D g-C₃N₄ (Δ = 0.7 eV) > 1D Bi₂S₃(0.2)/2D g-C₃N₄ (Δ = 0.64 eV) > 1D Bi₂S₃(0.3)/2D g-C₃N₄ (Δ = 0.54 eV), implying a weakening of interfacial interactions with increasing population of 1D Bi₂S₃ in the heterostructures. From the XPS data, it could be inferred that the existence of strong interfacial interactions in 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructures set up an IEF that directed the charge flow across the interface, similar to a previous report.⁵²

In order to confirm the charge redistribution at the 1D Bi₂S₃/2D g-C₃N₄ interface, we performed the work function, charge differential and Bader charge analysis (Fig. 2g–k). According to band theory, the electron transfer and band alignment of the semiconductor closely relate to its work function (WF, φ). Therefore, we computed the WF from the equation: φ = E_vac − E_F, where E_F and E_vac are respectively the Fermi energy level and electrostatic vacuum level potentials. The values of WFs (φ) for the 2D g-C₃N₄, 1D Bi₂S₃ and 1D Bi₂S₃/2D g-C₃N₄ structures were respectively simulated to be 5.2 eV, 4.9 eV and 4.6 eV (Fig. 2g–i). On the basis of the mismatch of the Fermi energy levels of 2D g-C₃N₄ and 1D Bi₂S₃, it would be expected that the construction of the 1D Bi₂S₃/2D g-C₃N₄ interface would inevitably push electrons to flow from 2D g-C₃N₄ to 1D Bi₂S₃ until a unified Fermi energy level was achieved. This electronic charge redistribution caused an increased electron density at the interface near the 1D Bi₂S₃ side and electron depletion on the 2D g-C₃N₄ side, which eventually led to the formation of IEF directed from 2D g-C₃N₄ to 1D Bi₂S₃. IEF acted as the energizing force to drive the holes and electrons in opposite directions, thereby prolonging their lifetimes and improving catalytic performance. The charge density differences (CDD) analysis conducted by creating isosurfaces further verified the electronic charge accumulation (orange-purple color regions in Fig. 2j and k) around the Bi and S atoms with unsaturated coordination around the newly formed bonds at the 1D Bi₂S₃/2D g-C₃N₄ interface, while charge depletion (green color regions in Fig. 2j and k) on the 2D g-C₃N₄ surface. Additionally, Bader charge analysis revealed about 0.1 electronic charge (e⁻) transfer across the interface, thus fully corroborating the XPS results.

3.2. Optical and photo-electrical properties

As the light absorption capacity of semiconductors profoundly affects their catalytic efficiency, we characterized the 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructures and control samples (pristine 2D g-C₃N₄ and 1D Bi₂S₃) by ultraviolet-visible diffused reflectance spectroscopy (UV-vis DRS) as depicted in Fig. 3a. The absorption edge of 2D g-C₃N₄ at ∼460 nm blue-shifted by the induction of 1D Bi₂S₃, with the 1D Bi₂S₃(0.3)/2D g-C₃N₄ heterostructure found to display the largest shift to ∼420 nm. Importantly, 1D Bi₂S₃ induction significantly extended the light absorption ability of 2D g-C₃N₄ across the 200–800 nm region as indicated by its dark color (Fig. 3c). To evaluate the bandgap energy (E_g), the Kubelka–Munk function, (αhv)ⁿ = k(hv − E_g), was applied to convert the UV-vis DRS spectral data to Tauc plots as shown in Fig. 3b. Compared to 2D g-C₃N₄ (E_g = 2.65 eV), 1D Bi₂S₃(n)/2D g-C₃N₄ samples exhibited narrowed E_g in the order: 1D Bi₂S₃(0.1)/2D g-C₃N₄ (E_g = 2.4 eV) > 1D Bi₂S₃(0.2)/2D g-C₃N₄ (E_g = 2.22 eV) > 1D Bi₂S₃(0.3)/2D g-C₃N₄ (E_g = 2.14 eV).

Fig. 3 (a) and (b) UV-Visible DRS spectra and corresponding Tauc plots, and (c) digital photographs of 2D g-C₃N₄, 1D Bi₂S₃ and 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructures. (d) Schematics showing the energy band alignment and charge transfer mechanism of the 1D Bi₂S₃/2D g-C₃N₄ heterostructure. (e) Comparison of PL spectra of 2D g-C₃N₄, 1D Bi₂S₃ and the various 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructures. (f) and (g) Photocurrent response and Nyquist plots of the 1D Bi₂S₃, 2D g-C₃N₄ and 1D Bi₂S₃(0.2)/2D g-C₃N₄ heterostructures, respectively.

The bandgap narrowing of 2D g-C₃N₄ upon coupling with 1D Bi₂S₃ was also validated by the DFT calculations (Fig. S6, ESI†). The valence band edge (E_VB) potentials and conduction band edge (E_CB) potentials of 2D g-C₃N₄ and 1D Bi₂S₃ were estimated using the empirical formulas: E_VB = χ + 0.5E_g − E₀ and E_CB = E_VB − E_g, where χ stands for absolute electronegativity, which is 4.73 eV and 5.27 eV for g-C₃N₄ and Bi₂S₃ respectively.^53,54 E₀ is the energy of free electrons on the hydrogen scale which is approximately ∼4.5 eV. Correspondingly, the E_VB and E_CB potentials of 2D g-C₃N₄ were respectively estimated to be 1.53 eV and −1.07 eV, while those for 1D Bi₂S₃ were estimated to be 1.7 eV and 0.38 eV. Notably, for the 1D Bi₂S₃(0.2)/2D g-C₃N₄ heterostructure, E_VB and E_CB edges of 2D g-C₃N₄ were lowered to 1.27 eV and −0.93 eV respectively, due to the formation of the intimate interface. According to the E_VB and E_CB potentials, band alignments of 1D Bi₂S₃(0.2)/2D g-C₃N₄ before contact, after contact and under light may be schematically represented as shown in Fig. 3d. Besides light absorption, the efficient separation of charge carriers (e⁻/h⁺) is also essential to activate surface-adsorbed molecules.

Photoluminescence (PL) spectroscopy offers an indirect means to investigate charge transfer characteristics, as the correlation between intensity and wavelength of the emitted photons produces a steady state PL spectrum that reflects the efficiency of charge separation. A significantly decreased PL spectral intensity of the 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructures as compared to 2D g-C₃N₄ unveiled the suppression of e⁻/h⁺ recombination (Fig. 3e). Besides, the lowest PL intensity of 1D Bi₂S₃(0.2)/2D g-C₃N₄ among all the heterostructures implies the existence of strong interfacial interactions in 1D Bi₂S₃(0.2)/2D g-C₃N₄, constructing an IEF that steered up the directional charge transfer and hindered the recombination rate of e⁻/h⁺ pairs. Excessive surface coverage of 2D g-C₃N₄ by 1D Bi₂S₃ in the 1D Bi₂S₃(0.3)/2D g-C₃N₄ heterostructure has resulted in an increased PL intensity, which could be due to the weakening of interfacial interactions as also indicated by XPS analysis. Moreover, the higher transient photocurrent response of 1D Bi₂S₃(0.2)/2D g-C₃N₄ as compared to 2D g-C₃N₄ and 1D Bi₂S₃ (Fig. 3f) under visible light (λ > 420 nm) on–off cycling further indicated the reduced recombination rate of charge carriers and improved electron transport across the 1D Bi₂S₃(0.2)/2D g-C₃N₄ interface. In addition, electrochemical impedance spectroscopy (EIS) analysis was performed to probe the interface charge transfer resistance. Generally, a smaller arc radius of Nyquist plot obtained from EIS analysis signifies a better charge transfer tendency. The smaller arc radius observed for 1D Bi₂S₃(0.2)/2D g-C₃N₄ as compared to those of 2D g-C₃N₄ and 1D Bi₂S₃ indeed inferred the former has better charge transfer ability (Fig. 3g). On the basis of the above analysis, 1D Bi₂S₃(0.2)/2D g-C₃N₄ with abundant active sites and improved charge separation is expected to profoundly improve catalytic activity.

4. Catalytic evaluation

4.1. Catalytic performance of 1D Bi₂S₃(n)/2D g-C₃N₄ towards antibiotics degradation

Tetracycline (TC), due to its high concentration in underground and drinking water, extended half-life and poor metabolization, was selected as the model antibiotic to evaluate the degradation performance of our prepared 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructures as shown in Fig. 4a. Prior to determining the contribution of visible light and PMS, the adsorption–desorption of TC in the dark was carried out for 30 minutes. Pristine 1D Bi₂S₃ exhibited better TC adsorption ability than 2D g-C₃N₄, which could be due to the interaction between surface functional groups on TC molecules and Bi⁺³ sites in the 1D Bi₂S₃ structure. Although both 2D g-C₃N₄ and 1D Bi₂S₃ are visible light responsive semiconductors, they failed to achieve adequate TC degradation efficiency due to the high e⁻/h⁺ recombination rate. Among the prepared heterostructures, 1D Bi₂S₃(0.2)/2D g-C₃N₄ exhibited the best TC degradation competency of 76% (Fig. S7, ESI†), followed by 1D Bi₂S₃(0.1)/2D g-C₃N₄ (65%) and 1D Bi₂S₃(0.3)/2D g-C₃N₄ (54%). This activity trend signifies that the optimal 1D Bi₃S₃ amount with strong contact with the 2D g-C₃N₄ surface is crucially important to optimize the degradation performance. Next, the PMS activation ability of 1D Bi₂S₃(0.2)/2D g-C₃N₄ was investigated. Under dark conditions, the 1D Bi₂S₃/PMS and 2D g-C₃N₄/PMS systems degraded 37% and 16% of TC respectively, while the 1D Bi₂S₃(0.2)/2D g-C₃N₄/PMS system exhibited significantly improved TC degradation of 61%. This may be attributed to the electronic charge accumulation at the 1D Bi₂S₃(0.2)/2D g-C₃N₄ interface that could be readily transferred to the adsorbed PMS for facile activation.

Fig. 4 (a) TC degradation performance of 1D Bi₂S₃, 2D g-C₃N₄ and 1D Bi₂S₃(0.2)/2D g-C₃N₄ under different conditions (with/without PMS/light). (b) Kinetics curves of the TC degradation of photocatalyst/PMS/light systems. (c) Comparison of TC degradation efficiency of this work with some literature reported Bi and g-C₃N₄ heterostructures. (d) TC degradation efficiency of the 1D Bi₂S₃(0.2)/2D g-C₃N₄/PMS/light system in various water matrices. (e) General applicability of the 1D Bi₂S₃(0.2)/2D g-C₃N₄/PMS/light system towards various antibiotics and dye degradation (inset: rate constant (k) values). (Reaction condition: [TC] = 20 mg L⁻¹, [catalyst] = 30 mg, PMS = 2 mM and pH = 7).

Next, we explored the light and PMS synergistic effect to ameliorate the TC degradation. Visible light alone degraded only 14% of TC, exhibiting weak PMS activation ability. Remarkably, under the simultaneous action of light and PMS, 1D Bi₂S₃(0.2)/2D g-C₃N₄ achieved optimized activity of 98.5% TC degradation, exceeding by 22% and 37% respectively when compared to 1D Bi₂S₃(0.2)/2D g-C₃N₄/light and 1D Bi₂S₃(0.2)/2D g-C₃N₄/PMS systems. Time-dependent UV-visible spectra of TC degradation was monitored for the 1D Bi₂S₃(0.2)/2D g-C₃N₄/light/PMS system (Fig. S8, ESI†) and the activities under different conditions are compared in Fig. 4a. The superiority of the 1D Bi₂S₃(0.2)/2D g-C₃N₄ system for the effective degradation of TC under the cooperative action of light and PMS could be credited to the enhanced light harvesting capacity, electronic charge modulation at the 1D/2D interface, and better charge separation via IEF. These thus collectively enabled 1D Bi₂S₃(0.2)/2D g-C₃N₄ to demonstrate impressive PMS activation under light as compared to its pristine 1D Bi₂S₃ and 2D g-C₃N₄ counterparts. In addition, adsorbed PMS itself may capture the photogenerated electrons to produce radicals or non-radical reactive oxygen species. The kinetic data of TC degradation could be described well by pseudo first-order reaction kinetics (Fig. 4b). The rate constant (k) value of TC degradation for the 1D Bi₂S₃(0.2)/2D g-C₃N₄/light/PMS system (0.06 min⁻¹) was determined to be 7 and 4.6 times higher than the 2D g-C₃N₄/light/PMS (0.01 min⁻¹) and 1D Bi₂S₃/light/PMS (0.015 min⁻¹) systems, respectively. TC degradation comparison in Fig. 4c clearly illustrates the superiority of our 1D Bi₂S₃(0.2)/2D g-C₃N₄ system over most of the other literature reported Bi and g-C₃N₄ heterostructures.

The actual implication of the 1D Bi₂S₃(0.2)/2D g-C₃N₄/light/PMS system in real-world scenarios was further investigated by testing TC degradation in Yangtze river, rain, tap, and lake water matrices (Fig. 4d). Interestingly for tap water, 1D Bi₂S₃(0.2)/2D g-C₃N₄ exhibited comparable TC degradation (92%) with that in pure water. In more complex water matrices such as Yangtze river, lake, and rainwater samples, 1D Bi₂S₃(0.2)/2D g-C₃N₄ still degraded 89%, 88%, and 86% of TC, respectively, proving its enormous potential for PMS-AOP. Additionally, the 1D Bi₂S₃(0.2)/2D g-C₃N₄ heterostructure exhibited excellent to moderate degradation ability towards various other organic molecules (Fig. 4e), including levofloxacin, ofloxacin, sulfamethoxazole, ciprofloxacin, rhodamine B and methylene blue, establishing its higher and universal utility.

4.2. Effects of reaction parameters

The effect of various reaction parameters such as: 1D Bi₂S₃ content, PMS dosage, TC concentration, pH value and ion interference were investigated as presented in Fig. 5. First, the TC degradation activity of 1D Bi₂S₃(n)/2D g-C₃N₄ was optimized by varying the 1D Bi₂S₃ to 2D g-C₃N₄ ratio as illustrated in Fig. 5a. Initially, the increase in Bi concentration from 0.1 mM to 0.2 mM produced a positive impact as indicated by enhanced TC degradation from 60% to 98.5%. However, at 0.3 mM Bi, TC degradation sharply declined to 83%. This was due to the over-abundance of 1D Bi₂S₃, which provided an ample number of reaction sites for species quenching⁵⁵ and thus lowered the overall degradation efficiency. These results signify that an appropriate amount of the 1D/2D intimate contact is crucial for effective TC removal. To prove this argument, we physically mixed 1D Bi₂S₃ and 2D g-C₃N₄ to fabricate a composite catalyst. This composite was found to exhibit only slight shifts in its XPS N 1s and Bi 4f B.E. relative to those of the 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructure (Fig. S9, ESI†), confirming its weaker interfacial interactions. Expectedly, the physically mixed sample degraded merely 79% of TC (Fig. S10, ESI†) under the same conditions, thus reaffirming the argument that strong interfacial interactions are essential to promote electron–hole separation and PMS activation towards efficient TC degradation.

Fig. 5 Effect of various reaction parameters on the TC degradation: (a) 1D Bi₂S₃ concentrations, (b) PMS concentration, (c) influence of pH, (d) TC concentration, (e) effect of common ions/molecules, and (f) recycling experiment.

TC degradation increased linearly with PMS concentration due to a proportionate increase in the number of active species until it reached 98.5% (k = 0.06 min⁻¹) at a concentration of 3 mM L⁻¹, then started to decline (Fig. 5b). The observed activity decline could be caused by the reverse reaction between SO₄˙⁻ and ˙OH radicals and excess HSO₅⁻ (eqn (1)–(4)), leading to the formation of radicals with lower oxidation potential such as HO₂˙ and SO₅˙⁻.⁵⁶ In addition, degradation efficiency could be hampered by the self-quenching of SO₄˙⁻ radicals (eqn (5)) prior to attacking the antibiotics.⁵⁶

HSO₅⁻ + SO₄˙⁻ → SO₄²⁻ + SO₅˙⁻ + H⁺ (1)

HSO₅⁻ + ˙OH → SO₅˙⁻ + H₂O (2)

HSO₅⁻ + ˙OH → HO₂˙ + SO₄²⁻ + H⁺ (3)

HSO₅⁻ + SO₄˙⁻ + H₂O → HO₂˙ + 2SO₄²⁻ + H⁺ (4)

SO₄˙⁻ + SO₄˙⁻ → S₂O₈²⁻ (5)

Next, it is of practical importance to investigate the pH tolerance, for the reason that TC can exist as cationic species (pH < 3.3), zwitterionic species (3.3 < pH < 7.7) or anionic species (pH > 7.7).⁵⁷ Consequently, water pH can alter the adsorption behavior of TC onto the surface of the photocatalyst and also influence the selectivity of ROS.⁵⁸ Under strong acidic conditions (pH = 1), TC degradation was found to drop to 69% (Fig. 5c), since the excessive H⁺ ions not only inhibit the generation of ROS by making HSO₅⁻ more stable, but also effectively react with SO₄˙⁻ or ˙OH,⁵⁹ resulting in a deficiency of ROS. As the initial pH was changed progressively to neutral conditions, TC degradation increased gradually and 1D Bi₂S₃(0.2)/2D g-C₃N₄ achieved 86% and 93% TC degradation, respectively, at pH = 4 and pH = 6. More interestingly, 1D Bi₂S₃(0.2)/2D g-C₃N₄ exhibited an exceptional 99.9% TC degradation in alkaline conditions (pH = 11). The plausible reason for this trend is that the alkaline environment promotes TC deprotonation, rendering TC with increased electron density around the ring system, thus making it more vulnerable to oxidation by ˙OH and especially SO₄˙⁻ radical.^60,61 Furthermore, alkaline conditions accelerate PMS activation to generate more ROS, thus enhancing TC degradation.⁶² Also, at pH = 11, 1D Bi₂S₃(0.2)/2D g-C₃N₄ exhibited negligible TC absorption, leaving more active sites available for PMS adsorption and activation. The influence of initial TC concentrations (10–50 mg L⁻¹) revealed the highest TC degradation efficiency of 99.3% at 10 mg L⁻¹ (Fig. 5d). However, the degradation of TC declined stepwise from 99.3 to 66.7%, when the initial concentration increased from 10–50 mg L⁻¹. The reason is that the PMS concentration is fixed and more ROS are required to react with the increased number of TC molecules and its intermediate fragments.

Natural water matrices contain a mixture of inorganic anions and organic species that can potentially compete with ROS or adhere to the photocatalyst surface, resulting in severely compromised antibiotic degradation activity. It is therefore highly critical to study the inhibitory effect of various ions (Cl⁻, NO₃⁻, HPO₄²⁻, CO₃²⁻ and HCO₃⁻) and organic molecules on the catalytic performance. As can be seen in Fig. 5e, the addition of 2 mM Cl⁻ ions did not result in any noticeable interference. In contrast, the addition of 5 mM and 10 mM Cl⁻ ions significantly enhanced the degradation of TC by 98.5% and 99.3% respectively. This could be due to the formation of chlorine species from the reaction between Cl⁻ ion and PMS, ˙OH, and SO₄˙, according to eqn (6)–(10).⁶³ Although the chlorine radicals are less reactive than ˙OH and SO₄˙ radicals, they possess a very strong affinity to attack the electron-rich regions of TC molecules, i.e., OH and NH₂ groups and conjugated double bonds of the aromatic rings, thus leading to a boosted TC degradation as previously reported.^63,64

Cl⁻ + HSO₅⁻ → SO₄²⁻ + HOCl (6)

2Cl⁻ + HSO₅⁻ + H⁺ → SO₄²⁻ + Cl₂ + H₂O (7)

˙OH + Cl⁻ → ClOH˙⁻ (8)

SO₄˙⁻ + Cl⁻ → SO₄²⁻ + Cl˙ (9)

Cl˙ + Cl⁻ → Cl₂˙⁻ (10)

The existence of CO₃²⁻ and HCO₃⁻ ions is expected to have a negative effect on the AOP process because they can act as scavengers for SO₄˙⁻ and ˙OH radicals.⁶⁵ However, in our experiment, a contradictory effect was observed because the existence of 10 mM CO₃²⁻ and HCO₃⁻ ions significantly accelerated the TC degradation. There are a number of reasons that could be credited to the promotional effect of CO₃²⁻ and HCO₃⁻ ions. First, the added CO₃²⁻ and HCO₃⁻ ions could react with ˙OH, and SO₄˙ radicals to produce CO₃˙⁻ and HCO₃˙ radicals (eqn (11)–(15)) of lower reactivity but higher selectivity towards electron rich groups of TC such as phenol and dimethylammonium.⁶⁶ Second, the nucleophilic character of both CO₃²⁻ and HCO₃⁻ ions promoted the activation of asymmetric PMS molecules to produce more reactive species,⁶⁷ thus resulted in higher degradation efficiency. Third, the presence of CO₃²⁻/HCO₃⁻ ions turned the reaction system alkaline, which promoted TC degradation as explained under the pH effect.⁶⁶

CO₃²⁻ + H₂O → OH⁻ + HCO₃⁻ (11)

˙OH + HCO₃⁻ → CO₃˙⁻ + H₂O (12)

˙OH + CO₃²⁻ → CO₃˙⁻ + OH⁻ (13)

SO₄˙⁻ + HCO₃⁻ → HCO₃˙ + SO₄²⁻ (14)

SO₄˙⁻ + CO₃²⁻ → CO₃˙⁻ + SO₄²⁻ (15)

Low concentrations of HPO₄⁻ did not caused any interference, but high concentrations improved TC degradation slightly because of the capability of HPO₄⁻ to break the O–O bond of PMS to yield SO₄˙⁻ radicals. In contrast to Cl⁻, CO₃²⁻, HCO₃⁻ and HPO₄²⁻ ions, NO₃⁻ moderately inhibited the TC degradation via the scavenging of ˙OH, and SO₄˙⁻ radicals by NO₃⁻ ions to form less reactive nitrate radicals (eqn (16) and (17)).⁶⁸

NO₃⁻ + ˙OH → NO₃˙ + OH⁻ (16)

NO₃⁻ + SO₄˙⁻ → NO₃˙ + SO₄²⁻ (17)

The TC degradation efficiency of 1D Bi₂S₃(0.2)/2D g-C₃N₄ reduced significantly from 98 to 57.7% with the addition of 10 mM of sodium salt of humic acid (HA). A possible reason could be due to the fact that HA not only competes for the active sites on the catalyst surface but also scavenges ROS more effectively because the functional groups on HA are more prone to attack from the SO₄˙⁻ and ˙OH radicals as previously described.^69,70

4.3. Identification of reactive species and reaction mechanism

In order to identify the radical and non-radical reactive species formed during the entire TC degradation process, competitive scavenging experiments were carried out by adding tert-butyl alcohol (TBA), ethanol (EtOH), triethanolamine (TEO), L-histidine (L-His), and 2,2,6,6-tetramethylpiperidine-1-oxyl (TEMPOL) as a specific scavenger for ˙OH, ˙OH and SO₄˙⁻, h⁺, ¹O₂ and O₂˙⁻, respectively (Fig. 6a). Quenching experiments revealed a moderate contribution of ˙OH radicals to the degradation of TC since the addition of TBA retarded the removal efficiency to 71%. The hydroxyl group of EtOH is expected to exhibit a strong ability to trap both ˙OH and SO₄˙⁻ radicals at high rates (k = (1.2–2.8) × 10⁹ M⁻¹ s⁻¹ and k = (1.6–7.7) × 10⁷ M⁻¹ s⁻¹, respectively). However, the addition of EtOH only showed a slightly stronger inhibitory effect compared to TBA in our experiment, indicating the minor role of SO₄˙⁻ radicals and their transfer to the other ROS.

Fig. 6 (a) Radical and non-radical quenching effect on TC degradation for the 1D Bi₂S₃(0.2)/2D g-C₃N₄ photocatalyst. Spin-trapping ESR spectra of 1D Bi₂S₃(0.2)/2D g-C₃N₄ for (b) DMPO–˙OH and DMPO–SO₄˙, (c) DMPO–O₂˙⁻, (d) TEMP–¹O₂, and (e) TEMP–h⁺. (f) Comparison of the Bi 4f XPS profile for fresh and spent 1D Bi₂S₃(0.2)/2D g-C₃N₄ samples. (g) Proposed TC degradation mechanism on the 1D Bi₂S₃(0.2)/2D g-C₃N₄ heterostructure under the synergistic action of light and PMS.

Consistently, an electron paramagnetic resonance (EPR) spectra with a DMPO spin trapping agent revealed the characteristic peaks of DMPO˙–OH and DMPO˙–SO₄ (Fig. 6b). This therefore provided direct evidence for the generation of ˙OH and SO₄˙⁻ radicals. Notably, relatively enhanced DMPO˙–OH signal intensity for the 1D Bi₂S₃(0.2)/2D g-C₃N₄/PMS/light system further confirmed the transfer of SO₄˙⁻ radicals to ˙OH radicals according to eqn (18). The O₂˙⁻ and ¹O₂ quenching effect was more obvious because the addition of 2 mM TEMPOL and 2 mM L-His suppressed the TC degradation to 66% and 57%, respectively. The appearance of intense characteristic peaks of DMPO–O₂˙⁻ (Fig. 6c) and TEMP–¹O₂ (Fig. 6d) in the EPR spectrum of the 1D Bi₂S₃(0.2)/2D g-C₃N₄/PMS/light system proved the formation of O₂˙⁻ and ¹O₂. The increase in TEMP–¹O₂ EPR signals of the 1D Bi₂S₃(0.2)/2D g-C₃N₄/PMS/light system, as compared to the 1D Bi₂S₃(0.2)/2D g-C₃N₄/light system, suggests that the activation of PMS generated additional ¹O₂ species. Hole (h⁺) quenching by a TEAO scavenger restricted the TC degradation to 76%, which implies their direct contribution by reacting with TC molecules or indirect contribution by reacting with HSO₅, and O₂˙⁻ to form ¹O₂ species according to eqn (19) and (20). To our surprise, the intensity of the TEMP–h⁺ signal was significantly reduced when PMS was present (Fig. 6e), indicating part of the h⁺ was consumed by PMS to produce ¹O₂ species, which is consistent with the increased intensity of TEMP–¹O₂ signals. Quenching all the ROS (˙OH, SO₄˙⁻, O₂˙⁻ and ¹O₂) with NaN₃ retarded the TC degradation to 21%. Consequently, radical and non-radical species quenching together with EPR results conclusively proved that 1D Bi₂S₃(0.2)/2D g-C₃N₄ would efficiently trigger the activation of O₂ and PMS to produce ROS; and further evidenced that effective TC degradation was achieved via the synergistic action of various reactive species as shown in Fig. 6g.

SO₄˙⁻ + H₂O → SO₄²⁻ + ˙OH + H⁺ (18)

HSO₅ + h⁺ → 1/2¹O₂ + ˙SO₄ + H⁻ (19)

O₂˙⁻ + h⁺ → ¹O₂ (20)

4.4. DFT investigation on O₂ and PMS activation

To further elucidate the activation mechanism, we simulated the adsorption behavior of O₂ and PMS molecules on 2D g-C₃N₄, 1D Bi₂S₃, and separately on the α and β configurations of the 1D Bi₂S₃/2D g-C₃N₄ heterostructure. As the 1D Bi₂S₃(0.2)/2D g-C₃N₄/light system exhibited significant TC degradation under aerobic conditions, we first investigated the adsorption and activation of O₂. The O₂ molecule showed no adsorption on pristine g-C₃N₄ but strong adsorption on 1D Bi₂S₃ (ΔE_ad(O₂) = −1.94 eV, Fig. S11a, ESI†), which stretched and eventually ruptured the O–O bond. In the 1D Bi₂S₃/2D g-C₃N₄-α and 1D Bi₂S₃/2D g-C₃N₄-β configurations, O₂ adsorbed weakly on the 2D g-C₃N₄ sites, with ΔE_ad(O₂) values of −0.06 eV and −0.07 eV, respectively (Fig. S11b and c, ESI†). For the 1D Bi₂S₃/2D g-C₃N₄-α model, two distinct O₂ adsorption sites were identified: one through the unsaturated S atom (ΔE_ad(O₂) = −0.93 eV, Fig. S11d, ESI†) and the other sandwiched between S and Bi atoms (ΔE_ad(O₂) = −0.42 eV, Fig. S11e, ESI†). In contrast, for the case of the 1D Bi₂S₃/2D g-C₃N₄-β model, O₂ adsorption solely occurred at the unsaturated S sites (ΔE_ad(O₂) = −1.82 eV, Fig. S11f, ESI†). The observed slight increase in ΔE_ad(O₂) at the 2D g-C₃N₄ sites of the heterostructure relative to its pristine 2D structure, and slight decrease at the 1D Bi₂S₃ sites relative to its pristine 1D structure could be attributed to the electronic charge redistribution as evidenced by XPS and CDD analysis. These findings confirm the crucial role of 1D Bi₂S₃/2D g-C₃N₄ heterostructures in modulating electronic properties, facilitating O₂ adsorption and activation process.

Next, we examined the possible adsorption configurations of the PMS molecule on the pristine 2D g-C₃N₄, 1D Bi₂S₃, and the 1D Bi₂S₃/2D g-C₃N₄ heterostructure. Simulations revealed weak interactions with 2D g-C₃N₄ (Fig. S12a, ESI†) but robust adsorption on 1D Bi₂S₃ (ΔE_ad = −3.98 eV, Fig. S12b, ESI†) through both Bi and S atoms. These findings confirm the high capability of 1D Bi₂S₃ towards PMS activation and are in good agreement with the observed TC degradation performance. To understand the PMS (HO–OSO₃⁻) activation mechanism, we simulated the adsorption energy of the peroxide O–O bond in PMS on the 1D Bi₂S₃/2D g-C₃N₄-α model. Previous work has revealed that activation of PMS to generate SO₄˙⁻ and ˙OH radicals is accomplished through the adsorption of the peroxide O–O bond, either via the HO– or the –SO₄ side.⁷¹ Our simulations disclosed that the peroxide O–O bond adsorbed strongly on both Bi and S sites on the 1D Bi₂S₃/2D g-C₃N₄-α heterostructure (Fig. 7). Notably on the Bi site (Fig. 7a), the O–O bond exhibited a unique adsorption mode with both the O atoms interacting effectively with the Bi atom (ΔE_ad(HSO₅⁻) = −1.72 eV, system a(II) in Fig. 7a). During adsorption, the O–O bond was instantaneously stretched from 1.33 Å (gas phase) to 1.49 Å and the system energy decreased to −4.00 eV (system a(III) in Fig. 7a), which signifies barrier-less O–O bond breaking. The CDD analysis, as shown in the inset of Fig. 7a, revealed a significant charge depletion around the Bi atom (indicated by green isosurfaces) and charge accumulation around the O atoms of the adsorbed PMS (indicated by orange-purple isosurfaces). Bader charge analysis endorsed these findings and affirmed the transfer of 1.82e⁻ from the Bi site to the adsorbed PMS molecule. This observation thus emphasizes that the Bi atom played a central role in facilitating the activation of PMS. The charge redistribution around the Bi site was also confirmed by performing XPS analysis of the spent catalyst. The B.E. of the Bi 4f_5/2 and Bi 4f_7/2 XPS peaks shifted towards higher values by 0.8 eV (Fig. 6f), indicating a reduction in electron density around the Bi atom, consistent with the charge transfer mechanism predicted theoretically.

Fig. 7 Schematic illustration of DFT simulated free energy changes against the reaction coordinate for PMS adsorption: (a) on the Bi site, and (b) on the S site, for the simulated 1D Bi₂S₃/2D g-C₃N₄-α model. Insets of (a) and (b) are the CDD analysis.

Similarly, DFT simulations predicted strong adsorption of the peroxide O–O bond in PMS on the S site (Fig. 7b) through the O atom of –SO₄ (ΔE_ad(HSO₅⁻) = −3.18 eV, system b(II) in Fig. 7b). The robust adsorption resulted in the stretching of the O–O bond length to 2.54 Å, causing splitting of PMS into ˙OH and SO₄˙⁻ radicals. These radicals adsorbed on the same S atom, with adsorption energies of ΔE_ad(˙OH) = −0.44 eV and ΔE_ad(SO₄˙⁻) = −2.74 eV respectively (systems b(III) and b(IV) in Fig. 7b). CDD analysis (inset in Fig. 7b) and Bader charge transfer calculations revealed the transfer of 1.86e⁻ charge from the S site to PMS, supporting the cleavage of the O–O bond on the S site. The relatively mild adsorption energy of ˙OH radicals on Bi and S sites suggests that they can easily desorb and participate in the TC degradation process. This agrees with the increased intensity of EPR signals for ˙OH radicals measured in the presence of PMS. On the other hand, the stronger affinity of SO₄˙⁻ towards Bi and S sites indicates that it may react with water molecules (eqn (18)) to generate additional reactive oxygen species (ROS), as certified by the radical trapping experiments. These DFT findings together with experiment results predicted that the formation of the 1D Bi₂S₃/2D g-C₃N₄ heterostructure is beneficial for the efficient PMS activation and subsequent radical generation processes.

In light of radical trapping experiments, EPR analysis and DFT calculations, the plausible mechanism of TC degradation by the 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructure in a light-assisted PMS-AOP can be rationally put forward as illustrated in Fig. 6g. First, the loading of 1D Bi₂S₃ nanostructures on the 2D g-C₃N₄ surface facilitated the light absorption, separation of charge carriers and electronic tuning of surface-active sites to collectively improve the activation of PMS activation and antibiotic degradation process. Second, optimized growth of 1D Bi₂S₃ effectively adjusted the interfacial interactions, creating a strong IEF that enabled rapid electron transfer from the 2D g-C₃N₄ substrate to the 1D Bi₂S₃ active sites, as supported by both XPS and DFT results. Third, the increased electron density on the Bi and S sites enabled strong binding with PMS molecules to form a stable transition state structure. As a result, compared to standalone 2D g-C₃N₄ and 1D Bi₂S₃, the in situ engineered 1D Bi₂S₃(n)/2D g-C₃N₄ heterostructure exhibited robust homolytic cleavage of the peroxide O–O bond to yield ˙OH and SO₄˙⁻ radicals at significantly lowered energy barriers as predicted by DFT calculations. Ultimately, the abundance of ROS produced by the simultaneous activation of O₂ and PMS on the DBi₂S₃(0.2)/2D g-C₃N₄/light system collectively reacted with the pollutants, leading to their degradation and mineralization in aqueous media.

4.5. Stability and recyclability of photocatalysts

The structural stability and recyclability of photocatalysts is crucial for practical catalytic implementation. Thus, the stability of 1D Bi₂S₃(0.2)/2D g-C₃N₄ was investigated by carrying out TC degradation experiments for four consecutive cycles. After each cycle, the spent 1D Bi₂S₃(0.2)/2D g-C₃N₄ catalyst was recovered and washed with deionized water and ethanol to remove the adsorb organic and inorganic species and then dried in an oven at 80 °C. As shown in Fig. 5f, the TC degradation efficiency of recycled 1D Bi₂S₃(0.2)/2D g-C₃N₄ declined approximately to 92%, i.e. ∼6% lower than that of first cycle. The apparent loss in activity could be attributed to the loss of catalyst upon filtration and/or re-oxidation of Bi as evident by the post XPS results. The XRD diffraction patterns of the fresh and recycled samples exhibited the characteristic peaks of 1D Bi₂S₃ (Fig. S13a, ESI†), suggesting promising structural stability. Additionally, SEM and TEM images of the recycled catalysts revealed no visible structural changes (Fig. S13b–d, ESI†), indicating the high structural stability of the prepared 1D Bi₂S₃(0.2)/2D g-C₃N₄ heterostructure.

Conclusion

In summary, we have designed a novel single-step in situ solid–gas reaction strategy to fabricate 1D Bi₂S₃(n)/2D g-C₃N₄ (n = 0.1, 0.2 and 0.3) heterostructures with adjustable densities of 1D Bi₂S₃ architectures. XPS and DFT studies verified the presence of strong interfacial interactions, which efficiently modified the electronic structure, especially for the Bi and S atoms at the interface. Notably, the optimized 1D Bi₂S₃(0.2)/2D g-C₃N₄ heterostructure exhibited exceptional performance, and degraded 98.5% and 99.9% of tetracycline (TC) and rhodamine B (RhB), respectively, exceeding most of the documented g-C₃N₄-based systems. DFT calculations unveiled that electronic modification of Bi and S sites at the 1D Bi₂S₃/2D g-C₃N₄ interface promoted the PMS adsorption and decreased the energy barrier for the cleavage of the peroxide O–O bond, thereby accelerating the ROS generation and pollutant degradation efficiency. This in situ strategy is general and could be used for the rational design of metal sulfide/g-C₃N₄ heterostructures to explore their potential for real-world energy and environmental remediation applications.

Conflicts of interest

There is no conflict of interest to declare.

Data availability

The data supporting this article have been included as part of the ESI.†

Acknowledgements

M. Mateen, N. Guo and W. S. Chin contributed equally to this work. The authors appreciate the financial support from the National University of Singapore (Chongqing) Research Institute and Chongqing Liangjiang New Area Municipal Government.

References

1. L. Zhao, W. Zhou, M. Wen, Q. Wu, W. Li, Y. Fu, Q. Zhu, S. Chen and J. Ran, Energy Environ. Mater., 2023, 6, e12299.
2. S. B. Levy and B. Marshall, Nat. Med., 2004, 10, S122–S129.
3. H. Long, S. F. Miller, C. Strauss, C. Zhao, L. Cheng, Z. Ye, K. Griffin, R. Te, H. Lee, C.-C. Chen and M. Lynch, Proc. Natl. Acad. Sci. U. S. A., 2016, 113, E2498–E2505.
4. X. Zhang, B. Xu, S. Wang, X. Li, C. Wang, Y. Xu, R. Zhou, Y. Yu, H. Zheng, P. Yu and Y. Sun, Appl. Catal., B, 2022, 306, 121119.
5. M. Z. Akbari, Y. Xu, Z. Lu and L. Peng, Environ. Adv., 2021, 5, 100111.
6. E. Gómez, R. Cestaro, L. Philippe and A. Serrà, Appl. Catal., B, 2022, 317, 121703.
7. H. Kim, S. Lee, H. W. Seo, B. Kang, J. Moon, K. G. Lee, D. Yong, H. Kang, J. Jung, E.-K. Lim, J. Jeong, H. G. Park, C.-M. Ryu and T. Kang, ACS Nano, 2020, 14(12), 17241–17253.
8. B. C. Hodges, E. L. Cates and J.-H. Kim, Nat. Nanotechnol., 2018, 13, 642–650.
9. S. Zhang, H. Zheng and P. G. Tratnyek, Nat. Water, 2023, 1, 666–681.
10. R. Guo, B. Xi, C. Guo, W. Liu, N. Lv and J. Xu, Environ. Funct. Mater., 2022, 1, 239–252.
11. T. Yu, H. Chen, T. Hu, J. Feng, W. Xing, L. Tang and W. Tang, Appl. Catal., B, 2024, 342, 123401.
12. Q.-Y. Wu, Z. W. Yang, Z. W. Wang and W. L. Wang, Proc. Natl. Acad. Sci. U. S. A., 2023, 120(16), e2219923120.
13. X. Yang, X. Xie, S. Li, W. Zhang, X. Zhang, H. Chai and Y. Huang, J. Hazard. Mater., 2021, 419, 126360.
14. U. Ushani, X. Lua, J. Wang, Z. Zhang, J. Dai, Y. Tan, S. Wang, W. Li, C. Niu, T. Cai, N. Wang and G. Zhen, Chem. Eng. J., 2020, 402, 126232.
15. J. Lee, U. von Gunten and J.-H. Kim, Environ. Sci. Technol., 2020, 54(6), 3064–3081.
16. N. Zrinyi and A. L.-T. Pham, Water Res., 2017, 120, 43–51.
17. J. Wang and S. Wang, Chem. Eng. J., 2018, 334, 1502–1517.
18. Q. Zhou, C. Song, P. Wang, Z. Zhao, Y. Li and S. Zhan, Proc. Natl. Acad. Sci. U. S. A., 2016, 120(13), e2300085120.
19. R. Su, Y. Zhu, B. Gao and Q. Li, Water Res., 2024, 251, 121119.
20. X. Wang, K. Maeda, A. Thomas, K. Takanabe, G. Xin, J. M. Carlsson, K. Domen and M. Antonietti, Nat. Mater., 2009, 8, 76–80.
21. W.-J. Ong, L.-L. Tan, Y. H. Ng, S.-T. Yong and S.-P. Chai, Chem. Rev., 2016, 116(12), 7159–7329.
22. S. Cao, J. Low, J. Yu and M. Jaroniec, Adv. Mater., 2015, 27(13), 2150–2176.
23. G. F. S. R. Roch, M. A. R. da Silva, A. Rogolino, G. A. A. Diab, L. F. G. Noleto, M. Antonietti and I. F. Teixeira, Chem. Soc. Rev., 2023, 52, 4878–4932.
24. W.-D. Oh, V. W. C. Chang, Z.-T. Hu, R. Goei and T.-T. Lim, Chem. Eng. J., 2017, 323, 260–269.
25. L. Li, H. Zeng, R. Tang, Z. Zhou, S. Xiong, W. Li, Y. Huang and Y. Deng, Appl. Catal., B, 2024, 345, 345123693.
26. H. Shi, Y. He, Y. Li and P. Luo, ACS Catal., 2023, 13(13), 8973–8986.
27. C. Lu, Y. Yang and X. Chen, Nano Lett., 2019, 19(16), 4103–4111.
28. W. Zhang, D. Xu, F. Wang and M. Chen, Nanoscale Adv., 2021, 3, 4370–4387.
29. J. Miao, J. Song, J. Lang, Y. Zhu, J. Dai, Y. Wei, M. Long, Z. Shao, B. Zhou, P. J. J. Alvarez, P. J. J. Alvarez and L. Z. Zhang, Environ. Sci. Technol., 2023, 57(10), 4266–4275.
30. C. Zhai, Y. Chen, X. Huang, A. B. Isaev and M. Zhu, Environ. Funct. Mater., 2022, 1, 219–229.
31. V. Hasija, V.-H. Nguyen, A. Kumar, P. Raizada, V. Krishnan, A. A. P. Khan, P. Singh, E. Lichtfouse, C. Wang and P. T. Huong, J. Hazard. Mater., 2021, 413, 125324.
32. J. Fu, J. Yu, C. Jiang and B. Cheng, Adv. Energy Mater., 2018, 8(3), 1701503.
33. A. Zhu, L. Qiao, P. Tan and J. Pan, Inorg. Chem. Front., 2020, 7, 4754.
34. Y. Ren, D. Zeng and W.-J. Ong, Chin. J. Catal., 2019, 40(3), 289–319.
35. G. Manna, R. Bose and N. Pradhan, Angew. Chem., Int. Ed., 2014, 53, 6743–6746.
36. J. Li, Z. Li, X. Liu, C. Li, Y. Zheng, K. Wai, K. Yeung, Z. Cui, Y. Liang, S. Zhu, W. Hu, Y. Qi, T. Zhang, X. Wang and S. Wu, Nat. Commun., 2022, 12, 1224.
37. J. Wang, S. Lin, N. Tian, T. Ma, Y. Zhang and H. Huang, Adv. Funct. Mater., 2021, 31(9), 2008008.
38. H. Zhang, Z. Wang, J. Zhang and K. Dai, Chin. J. Catal., 2023, 49, 42–67.
39. D. Chen, J. Fang, S. Lu, G. Y. Zhou, W. Feng, F. Yang, Y. Chen and Z. Q. Fang, Appl. Surf. Sci., 2017, 426, 427–436.
40. B. Zhanga, H. Shia, Y. Yana, C. Liua, X. Hub, E. Liua and J. Fan, Colloids Surf., A, 2021, 608, 125598.
41. A. Balakrishnan, J. D. Groeneveld, S. Pokhrel and L. Mädler, Chem. – Eur. J., 2021, 27, 6390–6406.
42. B. Shao, X. Liu, Z. Liu, G. Zeng, Q. Liang, C. Liang, Y. Cheng, W. Zhang, Y. Liu and S. Gong, Chem. Eng. J., 2019, 368, 730–745.
43. J. Fan, H. Qin and S. Jiang, Chem. Eng. J., 2019, 359, 723–732.
44. Y.-P. Zhu, T.-Z. Ren and Z.-Y. Yuan, ACS Appl. Mater. Interfaces, 2015, 7(30), 16850–16856.
45. J. Xu, L. Zhang, R. Shi and Y. Zhu, J. Mater. Chem. A, 2013, 1, 14766–14772.
46. F. Zhang, J. Zhang, J. Li, X. Jin, Y. Li, M. Wu, X. Kang, T. Hu, X. Wang, W. Ren and G. Zhang, J. Mater. Chem. A, 2019, 7, 6939–6945.
47. J. Fu, Q. Xu, J. Low, C. Jiang and J. Yu, Appl. Catal., B, 2019, 243, 556–565.
48. R. Shen, J. Xie, X. Lu, X. Chen and X. Li, ACS Sustainable Chem. Eng., 2018, 6(3), 4026–4036.
49. N. Tian, Y. Zhang, X. Li, K. Xiao, X. Du, F. Dong, G. I. N. Waterhouse, T. Zhang and H. Huang, Nano Energy, 2017, 38, 3772–3781.
50. C. Li, Y. Du, D. Wang, S. Yin, W. Tu, Z. Chen, M. Kraft, G. Chen and R. Xu, Funct. Mater., 2017, 27, 1604328.
51. P. Li, X. Zhang, C. Hou, Y. Chen and T. He, Appl. Catal., B, 2018, 238, 656–663.
52. Q. Hao, C. Xie, Y. Huang, D. Chen, Y. Liu, W. Wei and B.-J. Ni, Chin. J. Catal., 2020, 4, 249–258.
53. J. Zhang, Y. Hu, X. Jiang, S. Chen, S. Meng and X. Fu, J. Hazard. Mater., 2014, 280, 713–722.
54. Y. Yan, Z. Zhou, W. Li, Y. Zhu, Y. Cheng, F. Zhao and J. Zhou, RSC Adv., 2014, 4, 38558–38567.
55. S. M. Abdel-Moniem, M. A. El-Liethy, H. S. Ibrahim and M. E. M. Ali, Ecotoxicol. Environ. Saf., 2021, 226, 112808.
56. Z. Li, M. Wang, C. Jin, J. Kang, J. Liu, H. Yang, Y. Zhang, Q. Pu, Y. Zhao, M. You and Z. Wu, Chem. Eng. J., 2020, 392, 123789.
57. M. Li, X. Liu, Z. Xie, C. Du and Y. Su, Environ. Sci.:Adv., 2024, 3, 290–303.
58. Y.-H. Guan, J. Ma, X.-C. Li, J.-Y. Fang and L.-W. Chen, Environ. Sci. Technol., 2011, 45, 9308–9314.
59. H. Li, H. Wang, Q. Gao, B. Han, K. Xia and C. Zhou, J. Mater. Chem. A, 2020, 8, 20953–20962.
60. Q. Wang, Y. Shi, S. Lv, Y. Liang and P. Xiao, RSC Adv., 2021, 11, 18525–18538.
61. X. Ao, W. Sun, S. Li, C. Yang, C. Li and Z. Lu, Chem. Eng. J., 2018, 361, 1053–1062.
62. Y. Wu, G. Liang, W.-B. Li, X.-F. Zhong, Y.-Y. Zhang, J.-W. Ye, T. Yang, Z.-W. Mo and X.-M. Chen, Chem. Sci., 2024, 15, 9733–9741.
63. L. R. Bennedsen, J. Muff and E. G. Søgaard, Chemosphere, 2012, 86, 1092–1097.
64. P. Wang, Y.-L. He and C.-H. Huang, Water Res., 2011, 45, 1838–1846.
65. R. Zhang, P. Sun, T. H. Boyer, L. Zhao and C.-H. Huang, Environ. Sci. Technol., 2015, 49(5), 3056–3066.
66. Y. Liu, X. He, X. Duan, Y. Fu, D. Fatta-Kassinos and D. D. Dionysiou, Water Res., 2016, 95, 195–204.
67. J. Iqbal, N. S. Shah, J. A. Khan, M. Naushad, G. Boczkaj, F. Jamil, S. Khan, L. Li, B. Murtaza and C. Han, Sep. Purif. Technol., 2024, 347, 127458.
68. J. Wang and S. Wang, Chem. Eng. J., 2021, 411, 128392.
69. B. Wu, Z. Wang, Y. Jia, N. Xu, L. Liao, C. Zhang, Z. Wang, Y. Shan, W. Feng and H. Xue, Environ. Technol. Innov., 2024, 35, 103736.
70. S. Zhang, Z. Shao and D. Wu, Arab. J. Chem., 2024, 17, 105483.
71. J. Lee, U. von Gunten and J.-H. Kim, Environ. Sci. Technol., 2020, 54, 3064–3081.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5nh00265f

---