Semiconductor photocatalysis has been considered a promising solution for energy generation and environmental remediation in the past decades [1-7]. The semiconductor-based photocatalysis consists of photon absorption, migration of the photogenerated charge carriers to the surface of photocatalysts, and redox reactions in case of suitable valence and conduction band positions [3-12]. Thus, the semiconductors with wide absorption range of light and suitable alignment of energy levels are the prerequisites for highly efficient photocatalytic reactions. Though the band gap of a photocatalyst can be narrowed by downshifting the level of conduction band (CB) or upshifting the level of valence band (VB), this often weakens the redox driving force of charge carries photogenerated in it [13-15]. With this regard, an implicit contradiction between the light absorption and adequate redox capability cannot be avoid, which can absolutely have an impact on the activity of photocatalysts and, thereby, is a big challenge in the field of semiconductor photocatalysis.

ZnS, an important II-VI group transition metal sulfide, is known for its high theoretical efficiency of charge carrier generation. Moreover, owing to its relatively high position of CB, the photogenerated electrons in ZnS possess strong reduction capability for water splitting into H₂ even in the absence of noble metal co-catalysts like Pt [16]. Being a good photocatalyst, its weakness is that the pristine ZnS only responds to UV light due to its wide band gap, which impedes its further enhancement in the catalytic efficiency [17]. For design and fabrication of ZnS-based photocatalysts with high efficiency, accordingly, it is highly desired to extend the light response range. So far various approaches have been used to improve the light response of ZnS-based photocatalysts, such as heterojunction [18, 19], phase junction [16, 20], doping [21-23] and solid solution [24-34]. Among these, band engineering based on constructing ZnS-based solid solutions has attracted considerable attention because it can balance the driving force (i.e., redox potential) and light absorption via continuously tuning the band structure of a semiconductor. Moreover, precise control of the energy band structure is essential to understand the influence of structure on the photocatalytic activity at a molecular level. Therefore, precise regulation of band structure in ZnS-based solid solutions is of great significance.

It is noted that Zn_1–xCd_xS solid solutions have been widely studied during the past decades [24-28]. Kudo et al. [29-32] also developed some metal-sulfide solid-solution photocatalysts by combining ZnS with a narrow band gap semiconductor (MIn)_xZn_2(1–x)S₂ (M = Cu, Ag), which exhibit enhanced efficiency. Most work focus on the modulation of ZnS by incorporation of the cation ions hitherto, while introduction of the anion ions and its application in photocatalysis has rarely been reported. The substitution of selenide ion (Se^2‒) for sulfide ion (S^2‒) in anion-substituted solid solutions (like La₅Ti₂Cu(S_1–xSe_x)₅O₇, CdS_xSe_1–x and CuSb(S_1–xSe_x)₂) can change the band structure of a chalcogenide semiconductor. The photocatalytic activity can thus be tailored via the modulation of its composition [35]. The light response range can also be manipulated readily with increasing Se incorporation in Ag₂ZnSn(S_1–xSe_x)₄, meanwhile the transfer of charge carriers can be improved as well [36].

Therefore, it is expected that the photocatalytic efficiency of ZnS may be enhanced by substituting S^2‒ ions by Se^2‒ via the formation of ZnS_1–xSe_x solid solutions. In principle, a solid solution structure can be obtained readily if the two components are confined to the same symmetry and space group with similar lattice constant and physicochemical characteristic. Structural homogeneity and chemical compatibility of the two components are the prerequisites for the formation of a solid solution [37, 38]. It is known that ZnS and ZnSe are such a good example, as S (ionic radius r_S = 1.04 Å, Pauling electronegativity χ_S = 2.58) well matches Se (r_Se = 1.17 Å, χ_Se = 2.55) [39]. Furthermore, they both are inclined to exhibit a similar morphology when using ethylenediamine as the solvent in the synthesis, which plays a key role in achieving ZnS_1–xSe_x solid solutions with well-defined morphology [40, 41].

Herein ZnS_1–xSe_x nanobelt solid solutions were prepared by thermal treatment of ZnS_1–xSe_x(en)_0.5 precursors synthesized by a solvothermal method. The band gap of ZnS_1–xSe_x can be continuously tuned from 3.69 to 2.68 eV via increasing Se/S ratio. Composition-dependent band structure was also acquired both experimentally and theoretically. The obtained ZnS_1–xSe_x solid solutions were employed for CO₂ photoreduction. The ZnS-richest sample (ZnS_0.75Se_0.25) exhibited the highest activity among all the samples owing to the optimized balance between light absorption and reduction capability.

All the chemicals were analytical reagent and were used without further purification. Zinc powder and ethylenediamine (> 99%) were bought from Shanghai Aladdin Biological Technology Co., LTD. Sulfur powder and absolute ethanol (≥ 99.7%) were purchased from Beijing Chemical Works. Selenium powder was got from Shanghai Meixing Chemical Co., LTD. High purity water with 18.2 MΩ·cm resistivity was obtained from a Milli-Q Plus system (Millipore, France).

Based on the protocol reported previously [42-45], the ZnS_1–xSe_x(en)_0.5 (x = 0.00, 0.25, 0.50, 0.75 and 1.00) precursors were prepared via a solvothermal route using ethylenediamine as the solvent and structure-directing agent. Using ZnS_0.5Se_0.5(en)_0.5 as an example, typically, 2.25 mmol of Zn powder, 1.125 mmol of S powder and 1.125 mmol of Se powder were dissolved into 90 mL of ethylenediamine under constant stirring for 30 min. Subsequently, the mixture was transferred to an autoclave, followed by being put into a furnace and maintained at 180 ℃ for 24 h. After the autoclave was naturally cooled down to room temperature in air, the resultant white precipitate was separated by centrifugation, and was washed with ethanol and high purity water for several times to remove the residual organic solvent. Finally, the obtained samples were dried at 65 ℃ overnight in a vacuum oven. The ZnS_1–xSe_x nanobelts were prepared by annealing ZnS_1–xSe_x(en)_0.5 precursor at 400 ℃ for 2 h in Ar.

Crystal structure of the obtained samples was determined by X-ray diffraction (XRD) using Bruker D8 focus diffractometer with Cu-K_α radiation, from 20° to 80° at a scan rate of 0.1 step/s. Thermogravimetric (TG) analysis was carried out with a flow rate of 200 mL/min and a heating rate of 10 ℃/min in Ar by Diamond TG/DTA thermal analyzer. Morphology and chemical composition of the samples were analyzed with Hitachi S4800 field emission scanning electron microscope (FE-SEM) and Tecnai G2 F20 U-TWIN transmission electron microscope (TEM). Valence band position of the samples was acquired by X-ray photoelectron spectroscopy (XPS) using ESCALAB 250Xi X-ray photoelectron spectrometer. Fourier transform infrared (FT-IR) absorption spectra were recorded on a Perkin Elmer spectrometer from 4500 to 500 cm^-1. UV-visible light absorption spectra were collected by Lambda 750 UV/Vis/NIR spectrophotometer (Perkin-Elmer, USA). Photoluminescence (PL) spectra were acquired on a NanoLOG-TCSPC spectrophotometer (HORIBA JOBIN YVON, USA) with an excitation wavelength of 350 nm at room temperature. Time-resolved PL spectra were recorded on a spectrometer (FLS920, Edinburgh Instruments) with 375 nm excitation. BET specific surface area and CO₂ adsorption isotherms were obtained by a surface area and porosity analyzer (Micromeritics, Tristar II 3020).

Photocatalytic reduction of CO₂ was carried out in a quartz reactor illuminated by a 300 W Xe lamp (PLS-SXE300) with UV-vis light under constant stirring using Labsolar-IIIAG photoreaction system (Beijing Perfectlight Technology Co., Ltd.). Pure CO₂ (≥ 99.999%) was first bubbled into 100 mL 0.25-M Na₂SO₃ solution for at least 30 min. Then, 35 mg of ZnS_1–xSe_x powder was added in the above solution under sonication for 30 min. Finally, pure CO₂ was introduced into the obtained suspension for another 30 min in the dark to acquire CO₂ adsorption-desorption equilibrium before its reduction. The products were collected and analyzed online every hour by gas chromatograph (Agilent 7890A) equipped with flame ionization detector (FID) and thermal conductivity detector (TCD).

The full-potential augmented plane wave (FP-LAPW) method implemented in the Wien2k code based on density functional theory (DFT) [46] was used to study the band structure of ZnS_1–xSe_x solid solutions (x = 0.00, 0.25, 0.50, 0.75 and 1.00). The exchange correlation potential was calculated by Engel-Vosko GGA (EV-GGA) [47] and modified Becke-Johnson approximation (mBJ) [48]. A 2×2×1 supercell, consisted of 16 atoms (Zn₈S_8–xSe_x, x = 0, 2, 4, 6 and 8), was used for all the given concentration of x. The small size of such a supercell allowed us to sample every possible atomic configuration at each Se concentration in the entire composition region [49]. A mesh of 1000K points was used for a simple cell calculation, while a mesh of 300K points was used for a super cell calculation in full Brillouin zone. Suitable muffin-tin radii (RMT) for Zn, S and Se were selected to stop the overlapping of spheres as well as the leakage of charge out of the MT-sphere. The calculation was started with pure ZnS, and then gradually increasing Se content via substituting S by Se atoms.

All the as-synthesized ZnS_1–xSe_x(en)_0.5 precursors with different Se/S ratios possess similar XRD patterns to that of ZnS(en)_0.5 (Fig. S1); whereas continuous shift to lower 2θ values is observed with increasing Se amount compared with ZnS(en)_0.5, implying the successful incorporation of Se. Nearly all the ethylenediamine can be removed from the precursors after thermal treatment at 400 ℃ based on the TG analysis (Fig. S2), which is further supported by FT-IR results (Fig. S3). The peaks for ZnS_1–xSe_x(en)_0.5 precursor appearing at 3240 and 3120 cm^‒1 are attributed to the stretching vibration of −NH₂, and the peak at 640 cm^‒1 is assigned to the wagging vibration of −NH₂ [50]. The peaks at 2940 and 2870 cm^‒1 are ascribed to −CH₂ stretching vibration. The peaks at 1600 and 1355 cm^‒1 are scissoring vibration of −NH₂ and −CH₂, respectively. The peaks at 1080 and 1030 cm^‒1 correspond to stretching vibration of C-C and C-N, respectively [50]. Evidently, all these peaks disappear after annealing, indicating complete removal of ethylenediamine in the obtained samples.

Typical XRD patterns of the obtained ZnS_1–xSe_x samples are shown in Fig. 1, which demonstrates that all the ZnS_1−xSe_x samples with various Se/S ratios have similar diffraction patterns to those of hexagonal Wurtzite phase ZnS (JCPDS: 36-1450) and ZnSe (JCPDS: 15-0105). Albeit introduction of ZnSe into ZnS host lattice occurs, no splitting in the diffraction peak due to phase separation of ZnS or ZnSe can be observed. Moreover, the XRD peaks continuously shift to lower diffraction angles as the Se content increases, suggesting that the lattice constant of ZnS_1−xSe_x samples becomes larger and larger when the large size Se atoms successively replace S atoms, implying the obtained ZnS_1−xSe_x is a typical solid solution.

The lattice parameters of ZnS_1–xSe_x samples can be determined in light of the XRD data. It is found that the variation of lattice parameters with chemical composition of ZnS_1–xSe_x deviates slightly from linear dependence based on Vegard's law (Fig. S4), indicating again that the obtained ZnS_1–xSe_x solid solutions with various compositions are successfully prepared via changing the mole ratio of S to Se in the precursor materials (i.e., S and Se powder).

The obtained ZnS_1–xSe_x(en)_0.5 precursors exhibit belt-like morphology in terms of the SEM images (Fig. S5), which can even be kept upon annealing (Fig. 2). That is to say, the obtained ZnS_1–xSe_x samples also have belt-like morphology. This can be seen clearly from the following TEM analysis (Fig. 3 and Figs. S6 and S7). Furthermore, chemical composition of ZnS_1–xSe_x can be determined by using energy-dispersive spectroscopy (EDX) attached to SEM and TEM facilities as well as by XPS, which demonstrates that the ZnS_1–xSe_x nanobelts consist of Zn, S and Se elements and the S/Se ratio of the obtained ZnS_1–xSe_x nanobelts is close to that of the corresponding ratio in the precursors (Tables S1–S3). Moreover, EDX-mapping can be utilized to get element spatial distribution. Fig. 4 shows the individual elemental mapping of Zn, S and Se in ZnS_1–xSe_x solid solutions (x = 0.25, 0.50 and 0.75). For comparison, the elemental mapping of pure ZnS and ZnSe are presented in Figs. S8 and S9. Clearly, each element in the obtained samples is homogeneously distributed within the nanobelts. In addition, the interface of composition or phase separation among the three elements of Zn, S and Se cannot be observed. Again, such homogeneity is the solid evidence of the formation of solid solutions, which can be further supported by the line-scan composition profiles of ZnS_0.75Se_0.25 taken along the radial and axial directions (Fig. 3(a) and (b)).

SAED patterns of ZnS_0.75Se_0.25 (Fig. 3(c)) and other solid solutions (x = 0.00, 0.50, 0.75 and 1.00) (Figs. S6–S9) indicate that the obtained nanobelts are single-crystalline with a hexagonal crystal structure, as no diffraction spot splitting (typical for crystalline heterostructure due to the separated ZnSe and ZnS phases) can be seen. The respective inter-planar distance of (002) and (100) obtained from the HR-TEM images for all the samples is about 0.31 and 0.33 nm for ZnS and 0.32 and 0.34 nm for ZnSe, respectively (Figs. S8 and S9). Again, this can confirm the formation of solid solutions.

The chemical states of different elements in the as-prepared ZnS_1–xSe_x solutions are analyzed by XPS. The two peaks at about 1044.5 and 1021.5 eV in the XPS spectra for Zn are attributed to Zn 2p_1/2 and Zn 2p_3/2, respectively (Fig. 5(a)). Fig. 5(b) displays the XPS spectra of Se3d, with a binding energy of 54.1 eV. The respective binding energy of S2p_1/2 and S2p_3/2 for ZnS is 162.5 and 161.3 eV (Fig. 5(c)). Similarly, the two peaks at 165.9 and 160.2 eV for Se 3p signal of ZnSe correspond to Se 3p_1/2 and Se 3p_3/2, respectively (Fig. 5(d)). For the solid solutions (x = 0.25, 0.50 and 0.75), similar to pure ZnS and ZnSe, the respective binding energies of S 2p_1/2 and S 2p_3/2 locate at about 162.5 and 161.3 eV, and these of Se 3p_1/2 and Se 3p_3/2 are about 165.9 and 160.2 eV, respectively (Fig. 5(e)–5(g)). All these are reasonable since the binding energy for all the samples just changes slightly upon doping Se into ZnS lattice homogeneously (or vice versa) because the electronegativity of S (2.58) is similar to that of Se (2.55).

The band gap and optical properties of ZnS_1–xSe_x samples can be tuned by controling their chemical compositions. UV/vis absorption spectra of ZnS_1–xSe_x with different S/Se ratios and the corresponding Tauc plots are shown in Fig. 6. The optical absorption edge shows a systematic red-shift as the Se content increases from 0 to 1, and the band gap of ZnS_1–xSe_x can thus be tuned from 3.69 to 2.68 eV via tailoring Se/S ratio, implying that the incorporated Se in the ZnS narrows down the band gap and, accordingly, extends the absorption of ZnS into visible-light range.

The VB of as-synthesized ZnS_1–xSe_x samples with different composions (x = 0.00, 0.25, 0.50, 0.75 and 1.00) can be determined by XPS VB spectra (Fig. 7(a)), which is about 1.25, 1.16, 1.07, 0.98 and 0.92 V (vs NHE), respectively. Combining with the band gap obtained from UV-vis absorption spectra (Fig. 6(b)), the respective CB is thus determined to be -2.44, -2.16, -1.92, -1.81 and -1.76 V (vs NHE). Accordingly, the alignment of energy level of ZnS_1–xSe_x solid solutions including the CB and VB position (vs NHE) is drawn as shown in Fig. 7(b). The band structure of ZnS_1–xSe_x solid solutions is theoretically calculated too. A 2×2×1 supercell model is used for all given concentration of solid solutions (Fig. S10). The ZnS_0.5Se_0.5 solid solution is modeled in the crystal structure of space group Pmc21, while ZnS_0.75Se_0.25 and ZnS_0.25Se_0.75 solid solutions are crystallized in hexagonal structure of space group P63m. Clearly, the VB maximum and CB minimum for all the ZnS_1–xSe_x samples lie at the same symmetry point Г (Fig. 8), indicating that all these solid solutions are direct band gap materials. Additionally, the simulated band gap of ZnS_1–xSe_x solid solutions also decreases with the increase of Se, consistent with the experimental results.

Fig. 9 shows the results of photocatalytic reduction, the control experiment without CO₂ but just Ar carrier is first carried out. No products like CH₄ and CO can be detected, confirming the absence of carbon contamination in our system. Thus, CO₂ is the only carbon source for any potential reduction products. It is found that CO is the major product from CO₂ photoreduction over all the ZnS_1–xSe_x solid solutions, meanwhile a pretty large amount of H₂ is evolved from the competitive reaction of water splitting. For ZnSe sample, 2.96 µmol/g yield of CO and 0.34 mmol/g of H₂ can be obtained after 4-h irradiation; while it is 32.85 µmol/g for CO and 1.19 mmol/g for H₂ over the ZnS catalyst. After the introduction of little amount of Se into ZnS (such as x = 0.25), the photocatalytic activity can be enhanced obviously. The highest activity with 35.55 µmol/g of CO and 1.63 mmol/g of H₂ are observed over ZnS_0.75Se_0.25 sample; while the photocatalytic activity decreases when the amout of Se in ZnS_1–xSe_x is further increased (such as x = 0.50 and 0.75).

It is noted that the photocatalytic activity of a photocatalyst is highly dependent on the light harvesting, behavior of charge carriers and CO₂ adsorption. The respective BET specific surface area of ZnS_1–xSe_x solid solution (x = 0.00, 0.25, 0.50, 0.75 and 1.00) can be determined to be 108.76, 109.88, 107.80, 94.56 and 76.47 m²/g in terms of the typical N₂ adsorption-desorption isotherms of the ZnS_1–xSe_x samples (Fig. 10(a)). Compared with the pristine ZnS, the BET value of ZnS_1–xSe_x solid solution changes marginally when x = 0.25 and 0.50, and decreases slightly when x = 0.75. As for CO₂ adsorption, all the ZnS_1–xSe_x solid solutions exhibit almost the same CO₂ adsorption capacity but lower than that of ZnS (Fig. 10(b)); while the pristine ZnSe shows the lowest CO₂ adsorption capacity among all the obtained samples. This is further supported by temperature programmed desorption (TPD) results of CO₂ (Fig. S11). Therefore, the influence of CO₂ adsorption is not the major reason for the enhanced photocatalytic activity over ZnS_0.75Se_0.25 samples.

In addition, the photocatalytic activity of a photocatalyst can be affected greatly by the separation of charge carriers. This usually can be probed by the time-resolved photoluminescence decay (TRPL) measurements, for which the PL lifetime is determined from the decay curves. As shown in Fig. 11a, a PL peak at 440 nm is observed for all the samples when excited with the wavelength of 350 nm. Thus, the TRPL measurements are carried out according to this PL peak to acquire the information about the separation of charge carriers (Fig. 11(b)). The obtained PL decay curves are fitted mathematically in terms of the tri-exponential decay kinetics, corresponding to three processes: short lifetime component (τ₁) owing to non-radiative process, medium long one (τ₂) due to radiative process related to the recombination of photogenerated electrons and holes, and long lifetime one (τ₃) because of energy transfer process. The fitting results are shown in Table S4. The τ₂ for all the ZnS_1–xSe_x solid solutions follows the sequence of ZnS_0.5Se_0.5 > ZnSe > ZnS_0.25Se_0.75 > ZnS_0.75Se_0.25 > ZnS. That is to say, when x = 0.25, 0.50, 0.75 and 1.00, the τ₂ for the ZnS_1–xSe_x solid solutions is longer than the pristine ZnS.

Thus, the observed photocatalytic activity can be explained based on the aforementioned results. When little amount of Se (x = 0.25) is introduced into ZnS, the VB maximum shifts upward slightly due to the contribution from Se 4p states (besides S 3p and Zn 3d), and CB minimum shifts downward slightly owing to the contribution from Se 4p and Se 4s states (besides Zn 4s and S 3p), leading to a decrease in the band gap and, thus, enhancement in the light harvesting. Meanwhile, the separation of charge carriers is improved since the τ₂ of ZnS_0.75Se_0.25 is longer than ZnS; whereas the CO₂ adsorption capacity decreases slightly. Here we believe that the improved charge separation and light harvesting plays the primary role, resulting in enhanced photocatalytic activity; while the changes in the CO₂ adsorption and alignment of energy levels play a secondary role that may be ignored here.

When the Se amount in ZnS_1–xSe_x further increases (i.e., x = 0.50 and 0.75), the separation of charge carriers is further increased slightly due to a longer τ₂ and light harvesting is also further improved owing to the decreased band gap. Meanwhile, the change in the CO₂ adsorption capacity can hardly be observed. However, the reduction ability of photogenerated electrons decreases greatly because the CB minimum continuously shifts downward as the contribution from Se4p and Se4s states becomes larger and larger. Here it is suggested that the latter (change in the CB position) plays primary role in the photocatalytic activity, while the other factors play a secondary role.

As for the pure ZnSe (i.e., x = 1.0), it has the lowest reduction ability for the photogenerated electrons and the least positive potential for holes. Moreover, it also exhibits the lowest CO₂ adsorption capacity and BET value. This may explain why it shows the lowest photocatalytic activity, although it seems to have the highest light harvesting. In addition, the value of τ₂ obtained here may not be accurate, as the 440 nm used to determine the τ₂ may not the right one since the shape of its PL spectrum is different from the others (Fig. 11(a)).

In addition to the photocatalytic activity, the stability is another important issue for the evaluation of a photocatalyst. Here the stability of obtained ZnS_1–xSe_x solid solutions is studied by XRD, XPS and SEM before and after CO₂ photoreduction. The XRD patterns of ZnS_1–xSe_x solid solutions before and after photocatalysis are similar to each other, and the original wurtzite structure is always kept very well (Fig. 12(a)). Similarly, almost no change occurs in the XPS spectra of S 2p, Se 3p and Se 3d for the ZnS_1−xSe_x samples before and after photoreduction (Fig. 12(b)–(e)). Moreover, the SEM images of the samples also indicate that there is almost no change in the morphology upon photoreduction (Fig. S12). Hence, the obtained ZnS_1−xSe_x solid-solution photocatalysts exhibit relatively good stability in the photocatalytic process.

In summary, the ternary ZnS_1–xSe_x nanobelt solid solutions with different compositions have been successfully prepared by thermal treatment of the ZnS_1–xSe_x(en)_0.5 precursors. Composition-dependent band structure is obtained based on the experimental results and theoretical calculations. The band gap can be continuously modulated from 3.69 to 2.68 eV, dependent on the Se/S ratio. The ZnS-richest sample (ZnS_0.75Se_0.25) exhibits the highest photocatalytic activity for solar-fuel production among all the obtained ZnS_1–xSe_x solid solutions mainly because of the optimized balance between the effective light harvesting and adequate alignment of conduction band minimum. We envision that this work may afford a better understanding about the design and preparation of solid-solution photocatalysts with high efficiency, especially the balance of conduction/valence band position (reduction/oxidation power of photogenerated charge carriers) and band gap (light harvesting).

Supporting Information associated with this article can be found in the online version, including XRD patterns, TG curves and SEM images of the as-prepared ZnS_1–xSe_x(en)_0.5 precursors; FT-IR spectra of ZnS_1–xSe_x(en)_0.5 and ZnS_1–xSe_x samples; lattice constants "a" and "c" of ZnS_1–xSe_x solid solutions against its Se amount; TEM, SAED pattern and HRTEM images of ZnS_0.5Se_0.5 and ZnS_0.25Se_0.75 samples; TEM image, EDX mapping and SAED patterns of ZnS and ZnSe samples; supercell models of the ZnS_1–xSe_x solid solutions; TPD curves of CO₂ over the ZnS_1–xSe_x solid solutions; SEM images of ZnS, ZnSe and ZnS_0.75Se_0.25 after photoreduction; element composition of ZnS_1–xSe_x solid solutions derived from SEM-EDX, TEM-EDX and XPS analysis; PL lifetime of the ZnS_1–xSe_x solid solutions.