Environmental and energy issues are the two major factors hindering the sustainable development of human society. Photocatalysis is considered one of the effective routes to solve the above two issues ^[1-7]. Since the pioneering work of Fujishima and Honda in 1972 ^[8] on photoelectrochemical water splitting over a TiO₂ electrode, photocatalytic H₂ generation has been investigated as one of the most promising routes to convert solar energy into available chemical energy ^[9-15]. Until now, many semiconductors, such as TiO₂ ^[16-18], g-C₃N₄^[19-22], ZnO ^{[23, 24]}, and CdS ^{[25, 26]}, have been explored as photocatalysts for hydrogen generation by water splitting. However, single semiconductors have some disadvantages such as high recombination probability of photocarriers and limited redox potential ^[27-29].

Heterostructures (HSs) consisting of two or more semiconductors have the following advantages. First, the bandgaps of HSs can be easily adjusted by introducing guest semiconductors with suitable bandgap positions. This helps in expanding the light absorption region of the photocatalyst and improving the utilization efficiency of solar energy ^{[30, 31]}. Second, HSs provide abundant active interfaces for the photocatalytic process; moreover, the heterojunctions in HSs promote the separation and transfer of photogenerated electrons and holes ^[32-34].

Binary semiconductor HSs with three different configurations: type-Ⅰ ^[35], type-Ⅱ ^{[36, 37]}, and Z-scheme ^[38] have been reported. In recent years, ternary HSs consisting of three semiconductors have attracted considerable attention. For example, Xie et al. ^[39] prepared new ternary ZnTiO₃/Zn₂Ti₃O₈/ZnO HSs with high efficiency for photocatalytic hydrogen evolution and polluted water treatment. Jang et al. ^[40] fabricated ternary CdS/ZnIn₂S₄/TiO₂ HSs with excellent photoelectrochemical performance. In our previous study, we successfully synthesized a series of ternary HSs: MoO₃/MoS₂/g-C₃N₄^[41], CdS/WS₂/g-C₃N₄ ^[42], In₂S₃/CdIn₂S₄/In₂O₃ ^[43], and CdS/ZnS/PdS ^[44]. These ternary HSs exhibited enhanced photocatalytic H₂ evolution performances compared to the corresponding binary HSs and single semiconductors. Therefore, ternary HS-based materials are considered promising photocatalysts.

In our previous work ^[32], we fabricated new ternary CdS/ZnS/ZnO (abbreviated as CZZ) HSs by two steps. Firstly, ZnO nanosheets were sulfurized in situ to form ZnS/ZnO binary HSs. Secondly, CdS quantum dots (QDs) were deposited in situ on the ZnS/ZnO HSs to obtain CZZ ternary HSs. The CZZ ternary HSs exhibited a hydrogen evolution rate of 27.25 mmol h^-1g^-1 in the presence of a Pt co-catalyst. In order to further improve the photocatalytic performance of the CZZ ternary HSs, in the current work, we constructed new CZZ ternary HSs by the same procedure used in our previous work; however, the two-dimensional (2D) ZnO nanosheets were replaced with 1D ZnO nanofibers, which were prepared by electrospinning. The resultant CZZ nanofibers exhibited a much higher hydrogen evolution rate of 51.45 mmol h^-1g^-1 without any co-catalyst. The reason for the enhancement in photocatalytic H₂ evolution was analyzed in detail. Furthermore, these photocatalysts are economical because they are co-catalyst-free (noble metals, such as Pt and Au, are the most widely used co-catalysts), thus making our study more significant.

ZnO nanofibers were synthesized by electrospinning. Firstly, 0.226 g of zinc acetate [Zn(CH₃COO)₂] was dissolved in 12 mL of N, N-dimethylformamide (DMF), and then, 4 g of polyvinylpyrrolidone (PVP) powder (molecular weight = 900000) was added and stirred at room temperature for 6 h to obtain the Zn(CH₃COO)₂-PVP precursor for electrospinning. Subsequently, the precursor was placed in an injector for electrospinning. The electrospinning device and procedure used were the same as those used in our previous study ^[45]. In this work, the supply voltage was 15 kV, distance between the steel needle and the aluminum foil was 10 cm, speed of the syringe pump was 0.4 mL/h, and speed of the cylinder was 200 rpm. The electrospun PVP/Zn(CH₃COO)₂ composite nanofibers were calcined at 550 ℃ in air for 3 h at a heating rate of 1 ℃/min to obtain ZnO nanofibers.

ZnS/ZnO binary HSs were prepared by in situ sulfidation. Typically, 25 mg of the prepared ZnO nanofibers were ultrasonically dispersed in aqueous thiourea solutions with three different concentrations obtained by dissolving 30, 35, and 40 mg of thiourea in 5 mL of water, respectively. The resultant suspensions were then heated at 150 ℃ for 45 min in a Teflon autoclave (20 mL) to achieve in situ sulfidation of ZnO nanofibers, yielding ZnS/ZnO binary HSs. The obtained ZnS/ZnO binary HSs were labeled as ZZ-x (x stands for the weight of thiourea added).

CdS/ZnS/ZnO ternary HSs were synthesized by in situ deposition of CdS QDs on ZnS/ZnO binary HSs (ZZ-35) by the successive ionic layer adsorption and reaction method ^[46]. The amount of CdS QDs deposited was controlled by adjusting the number of deposition cycles. One deposition cycle involved the following steps. (1) The prepared ZnS/ZnO binary HSs were immersed in 0.1 mol/L Cd(CH₃COO)₂ solution for 5 min, and then, washed with deionized water to remove the excess Cd²⁺ ions. (2) The resulting sample was dried in air, immersed in 0.1 mol/L Na₂S solution for 5 min, and finally, washed with deionized water. Accordingly, a series of CdS/ZnS/ZnO ternary HSs were obtained. The obtained CdS/ZnS/ZnO ternary HSs were denoted as CZZ-y (y stands for the number of deposition cycles). For comparison, a pristine CdS sample was synthesized according to the procedure described in our previous report ^[47]. The synthesis of CdS/ZnS/ZnO ternary HSs is illustrated in Scheme 1.

X-ray diffraction (XRD) was performed with an X-ray diffraction system using the Cu K_α radiation. Transmission electron microscopy (TEM) and high-resolution TEM (HRTEM) were performed with a JEOL JEM-2100F microscope at an operating voltage of 200 kV. The morphology of the prepared catalysts was characterized by scanning electron microscopy (SEM, SU6600 Hitachi, Japan). The elemental compositions of the catalysts were determined with an energy-dispersive X-ray spectroscopy (EDX) system integrated into an SEM instrument (Phenom Prox). Ultraviolet-visible (UV-vis) absorption spectra were recorded with a V-670 spectrophotometer. X-ray photoelectron spectroscopy (XPS) was performed with a Thermo Fisher Scientific ESCALAB spectrometer using the aluminum K_α radiation. The photoluminescence (PL) spectra were recorded on a fluorescence spectrometer (Edinburgh Instruments FLS 980) using an excitation wavelength of 350 nm. Photoelectrochemical analysis (photocurrent measurements and electrochemical impedance spectroscopy (EIS)) was carried out by the same method described in our previous report ^[38].

A top-irradiation quartz cell connected to a closed glass gas circulation system (CEL-SPH2N, Beijing) was used for photocatalytic hydrogen production. Firstly, the sacrificial agents Na₂SO₃ (0.35 mol/L) and Na₂S·9H₂O (0.25 mol/L) were dissolved in 25 mL of water. Secondly, 10 mg of the prepared photocatalyst was added under stirring (note: no co-catalyst was added), and then, the resultant suspension was completely evacuated. Subsequently, the suspension was irradiated with a xenon arc lamp (225 W, 320–780 nm). The temperature of the whole reaction was maintained at about 6 ℃, and the H₂ produced was detected with an on-line thermal conductivity detector (TCD) gas chromatograph (GC-9720).

The XRD patterns of the prepared samples are displayed in Fig. 1. The patterns showed three strong peaks at 31.77°, 34.42°, and 36.26° which were assigned to the (100), (002), and (101) crystal planes of hexagonal wurtzite ZnO (JCPDS No. 65-3411), respectively. Moreover, ZZ-30, ZZ-35, and ZZ-40 showed three new XRD peaks at 29.06°, 48.37°, and 57.42°, which were indexed to the (111), (220), and (311) planes of ZnS (JCPDS No.79-0043), respectively, indicating the formation of ZnS on the surface of ZnO via in situ sulfidation. Furthermore, as the addition amount of thiourea increased, the ZnO peak became weaker, while the ZnS peak became stronger, demonstrating that ZnO was gradually consumed and converted to ZnS. Besides the characteristic ZnO and ZnS diffraction peaks, weak wide diffraction peaks in the range of 24°-30° were observed in the XRD patterns of the CZZ-y samples, which were attributed to the deposited CdS QDs (JCPDS 41-1049) ^[48].

Figure 2 shows the typical SEM images of the as-prepared nanofibers. As observed in Fig. 2A and 2B, the PVP/Zn(CH₃COO)₂ composite fibers with lengths of several micrometers and diameters of 250–300 nm exhibited smooth surfaces. The ZnO nanofibers (Fig. 2C and 2D) obtained after the heat treatment at 550 ℃ in air for 3 h exhibited lower lengths and diameters: a dozen or even dozens of micrometers and 100–150 nm, respectively, due to the decomposition of PVP. The SEM images of ZZ-35 are shown in Fig. 2E and 2F. Although the 1D nanofiber morphology was maintained during in situ sulfidation, the nanofibers were broken (several micrometers in length), as shown in Fig. 2E. However, after the deposition of CdS QDs, no obvious change in the morphology of the nanofibers was observed (Fig. S1). Notably, the pristine CdS sample exhibited nanoparticle morphology (Fig. S2).

Figure 3 shows the TEM and HRTEM images of CZZ-4. CZZ-4 exhibited the original 1D nanofiber morphology; the diameters of the nanofibers were about 200 nm (Fig. 3A). The HRTEM image of the sample (Fig. 3B) exhibited lattice fringes separated by distances of 0.31, 0.35, and 0.28 nm (for clarity, see the magnified images in Fig. 3C), which were assigned to the (111) planes of ZnS, (100) planes of CdS, and (100) planes of ZnO, respectively, demonstrating the successful synthesis of ternary HSs composed of ZnO, ZnS, and CdS.

Figure 4 shows the EDX mapping images of CZZ-4. As observed, Zn and Cd were uniformly distributed at the nanoscale. The density of the yellow dots (i.e., Cd concentration; see Fig. 4B) was significantly lower than that of the red dots (i.e., Zn concentration; see Fig. 4C), demonstrating that CdS QDs in small amounts were uniformly distributed on the ZnO/ZnS matrix. The amount of CdS QDs in CZZ-4 was determined with an EDX system integrated into an SEM instrument (Fig. S3). As shown in Table S1, the amount of CdS QDs in CZZ-4 was about 2.34 at.%.

The chemical compositions and surface chemical states of CZZ-4 were further studied by XPS. The Cd 3d XPS profile (Fig. 5A) exhibited two strong peaks at binding energies of 404.9 and 411.7 eV, which were assigned to the Cd 3d_5/2 and Cd 3d_3/2 core levels, respectively ^[49-51]. The O 1s spectrum was deconvoluted into two peaks at 530.8 and 532.1 eV (Fig. 5B), corresponding to the lattice oxygen species (lower binding energy) and chemisorbed oxygen species (higher binding energy), respectively ^{[52, 53]}. The S 2p XPS profile (Fig. 5C) exhibited two peaks centered at 161.4 and 162.6 eV, corresponding to the S 2p_1/2 and S 2p_3/2 electronic states, respectively ^[54-56]. The high-resolution Zn 2p XPS profile exhibited two peaks centered at 1021.9 and 1044.9 eV (Fig. 5D), which were assigned to Zn 2p_3/2 and Zn 2p_1/2, respectively, suggesting that Zn existed in the oxidation state of Zn^{2+ [31, 57]}.

Fig. 6A shows the UV-vis absorption spectra of all samples. The ZnO sample exhibited strong absorption at wavelengths lower than 400 nm because of its wide bandgap (3.05 eV) ^[58]. The optical absorption band edge of ZZ-x was blue shifted compared to that of ZnO and the extent of blue shift gradually increased with increasing vulcanization degree, which can be attributed to the transformation of more ZnO into ZnS and the larger bandgap of ZnS (3.68 eV) compared to that of ZnO (3.2 eV). In contrast, the optical absorption band edge of CZZ-y was red shifted and the extent of red shift gradually increased with the increasing number of CdS QD deposition cycles, which can be attributed to the deposition of more CdS QDs on ZZ-35 and the smaller bandgap of CdS (2.4 eV) compared to those of ZnO and ZnS ^[59]. Fig. 6B shows the Tauc plots of all samples, from which their optical band gap (E_g) values were determined. E_g was calculated using the equation α(hν) = A(hν - E_g)^1/2, where α, ν, E_g, and A are the absorption coefficient, light frequency, band gap energy, and a constant, respectively ^{[60, 61]}. The estimated E_g values are presented in Fig. 6B.

Photocatalytic analysis of the samples was carried out in an aqueous solution without additional co-catalysts, such as Pt and Au, under simulated solar irradiation (320–780 nm). As shown in Fig. 7A, the pristine ZnO sample exhibited the lowest hydrogen evolution rate (0.55 mmol h^–1 g^–1). The samples obtained after sulfidation: ZZ-30, ZZ-35, and ZZ-40 exhibited enhanced photocatalytic performances with H₂ evolution rates of 17.86, 22.57, and 17.83 mmol h^–1 g^–1, respectively, due to the presence of ZnS/ZnO binary HSs, which facilitated the separation and transfer of photogenerated charge carriers. Moreover, the samples obtained after the deposition of CdS QDs on ZZ-35: CZZ-2, CZZ-4, and CZZ-6 exhibited further enhanced photocatalytic H₂ evolution rates: 42.00, 51.45, and 40.08 mmol h^–1 g^–1, respectively, owing to the presence of CdS/ZnS/ZnO ternary HSs. Among them, CZZ-4 exhibited the highest H₂ evolution rate of 51.45 mmol h^–1 g^–1, which was 2.28 times higher than that of ZZ-35 (22.57 mmol h^–1 g^–1), indicating that the photocatalytic H₂ production performance of ternary HSs (CdS/ZnS/ZnO) was better than that of binary HSs (ZnS/ZnO). Furthermore, the photocatalytic H₂ evolution rate measured over the CZZ-4 sample under visible light irradiation (420–780 nm; a 420 nm filter was used to filter out the UV light) was 30.56 mmol h^-1 g^-1 (Fig. S4). In addition, the H₂ evolution rate for CZZ-4 (1D nanofibers) in the presence of a Pt co-catalyst was 118.62 mmol h^-1 g^-1 and that for 2D CZZ nanosheets without any co-catalyst was 9.98 mmol h^–1 g^–1 (Fig. S5). Notably, under the same conditions, P25 exhibited a photocatalytic H₂ evolution rate of 9.8 μmol g^-1 h^-1, which was significantly lower than that exhibited by CZZ-4.

Moreover, the durability and stability of the optimized CZZ-4 sample were determined by six consecutive processes under the same conditions (the photocatalyst was recovered by centrifugation and redispersed in deionized water with a sacrificial agent after each run). As shown in Fig. 7B, the H₂ production remained stable for six cycles (total time: 24 h), without significant decline, indicating the good durability and stability of CZZ-4.

The apparent quantum efficiency (AQE) is regarded as an important parameter in the determination of the activity of a photocatalyst; it reflects the conversion efficiency of solar energy into chemical energy during the photocatalytic process. To measure the AQE for CZZ-4, different bandpass filters (at 420, 450, and 500 ± 10 nm) were used. The method used for the calculation of AQE is described in the supporting information. As displayed in Fig. 8, CZZ-4 exhibited the highest AQE (26.88%) at 420 nm, without any co-catalyst. The AQE values at 450 and 500 nm were 14.26% and 11.42%, respectively.

To investigate the photoinduced electron-hole transport mechanism, PL, transient photocurrent response, and EIS measurements were performed. Fig. 9A presents the PL spectra of ZnO, ZZ-35, and CZZ-4 at an excitation wavelength of 270 nm; the PL spectra of other samples are shown in Fig. S6. ZnO exhibited the highest PL intensity, which was attributed to the fast recombination of photogenerated electrons and holes. ZZ-35 exhibited a lower PL intensity, indicating that the recombination of photoinduced charge carriers was reduced, which was attributed to enhanced separation efficiency of electron-hole pairs in the binary HSs. The PL intensity of CZZ-4 was lower than those of the other two, indicating that the separation efficiency of photogenerated electrons and holes of the ternary HSs was significantly improved ^[62]. A comparison of Figs. 7A and 9A indicates that the PL intensity is inversely proportional to the hydrogen evolution rate. The higher the PL intensity, the lower is the hydrogen evolution rate, demonstrating that the separation efficiency of photogenerated electron-hole pairs is very critical to the improvement of the hydrogen production efficiency.

The transient photocurrent density versus time curves for the ZnO, CdS, ZZ-35, and CZZ-4 samples under simulated solar light (320–780 nm) irradiation are presented in Fig. 9B. The photocurrent responses for the other samples are shown in Fig. S7. As shown in Fig. 9B, the samples exhibited steady and reproducible photocurrent responses during repeated on/off illumination cycles. Compared with ZnO, CdS, and ZZ-35, CZZ-4 exhibited a significantly improved photocurrent response, demonstrating the most efficient separation and transfer of photogenerated charge carriers. In addition, the EIS Nyquist plots confirmed the efficient separation and transfer of charge carriers in CZZ-4 (Fig. 9C), which exhibited the smallest radius, indicating that the transfer resistance of charge carriers in the ternary HSs was significantly reduced. The EIS Nyquist plots for other samples are shown in Fig. S3.

On the basis of the above experimental results, we proposed a mechanism for photocatalytic H₂ evolution over the prepared CZZ ternary HSs. As shown in Scheme 2, the excitation of all three semiconductors (ZnO, ZnS, CdS) under simulated sunlight (320–780 nm) irradiation causes the generation of electrons and holes (the energy levels of ZnO, ZnS, and CdS, as well as the ·OH/H₂O and H⁺/H₂ redox potentials are displayed in Scheme S1). The conduction band (CB) position of ZnS is more negative than those of ZnO and CdS; therefore, the photogenerated electrons in the CB of ZnS will be transferred to the CBs of ZnO and CdS, which will then reduce H⁺ to H₂. Meanwhile, the photogenerated holes produced in the valence band (VB) of ZnO will be transferred to the VB of ZnS, which will then react with the sacrificial agents. Moreover, the interface between ZnS and CdS induce the transfer of photogenerated holes in the VB of CdS to zinc vacancies (V_Zn) and interstitial sulfur (I_S) sites of ZnS ^{[44, 63]}.

Compared to the CZZ ternary HSs with nanosheet morphology prepared in our previous study ^[32], the CZZ ternary HSs with nanofiber morphology prepared in the current work exhibited a significantly higher photocatalytic H₂ evolution performance. This could be attributed to the following factors. A 2D nanosheet can be regarded as a parallel stack of several 1D nanofibers. Therefore, the geometric size of a nanofiber is considerably smaller than that of a nanosheet. Accordingly, the transfer distance of photogenerated charge carriers from the inside to the surface will be significantly lower, resulting in effective inhibition of recombination of photogenerated electrons and holes and improved photocatalytic hydrogen production efficiency.

In summary, new ternary CdS/ZnS/ZnO HS nanofibers were successfully prepared by a series of processes: electrospinning ZnO nanofibers, sulfurizing ZnO in situ to form ZnS/ZnO HSs, and depositing CdS QDs in situ on ZnS/ZnO HSs. The resultant CdS/ZnS/ZnO nanofibers exhibited a hydrogen evolution rate of 51.45 mmol h^-1 g^-1 without any co-catalyst, which was significantly higher than that of CdS/ZnS/ZnO nanosheets in the presence of a Pt co-catalyst (27.25 mmol h^-1 g^-1) prepared by the same sulfurization and deposition procedures. The enhanced photocatalytic H₂ evolution could be attributed to the considerably smaller sizes of 1D nanofibers compared to those of 2D nanosheets, resulting in effectively suppressed recombination of photogenerated charge carriers and promotion of photocatalytic hydrogen production performance. This work indicates that the photocatalytic performance could be improved by reducing the geometric size of the photocatalyst.

TEM and SEM were carried out at Jia-Lab for Interface and Atomic Structure of Xi'an Jiaotong University. EDX mapping and XPS were carried out at Analysis and Test Center of Xi'an Jiaotong University. We thank Hu Nan, Yanzhu Dai, Jiao Li, and Jiamei Liu for their help in using TEM, SEM, EDS mapping and XPS analyses, respectively.