Interfacial chemical bond and internal electric field modulated Z-scheme Sv-ZnIn2S4/MoSe2 photocatalyst for efficient hydrogen evolution

Construction of Z-scheme heterostructure is of great significance for realizing efficient photocatalytic water splitting. However, the conscious modulation of Z-scheme charge transfer is still a great challenge. Herein, interfacial Mo-S bond and internal electric field modulated Z-scheme heterostructure composed by sulfur vacancies-rich ZnIn2S4 and MoSe2 was rationally fabricated for efficient photocatalytic hydrogen evolution. Systematic investigations reveal that Mo-S bond and internal electric field induce the Z-scheme charge transfer mechanism as confirmed by the surface photovoltage spectra, DMPO spin-trapping electron paramagnetic resonance spectra and density functional theory calculations. Under the intense synergy among the Mo-S bond, internal electric field and S-vacancies, the optimized photocatalyst exhibits high hydrogen evolution rate of 63.21 mmol∙g−1·h−1 with an apparent quantum yield of 76.48% at 420 nm monochromatic light, which is about 18.8-fold of the pristine ZIS. This work affords a useful inspiration on consciously modulating Z-scheme charge transfer by atomic-level interface control and internal electric field to signally promote the photocatalytic performance.

W ith the rapid development of the industrial society, the energy crisis and environmental pollution issues are getting more and more serious. Therefore, finding an alternative energy source is of great significance for the long-term development of human society. Hydrogen (H 2 ) has long been considered as an excellent candidate to substitute the fossil fuel, due to its advantages of clean, renewable, high energy density, and transportability [1][2][3] . However, at present, the low efficiency, high energy consumption, and environmentally hazardous H 2 production technology seriously restrict the commercial application of hydrogen energy. By comparison, photocatalytic water splitting can tactfully convert the sustainable solar energy to H 2 energy without discharging any pollutant during the whole process, thus has been considered as a sustainable and promising technique 1,4,5 .
In the past few years, metal chalcogenide semiconductor photocatalyst, such as ZnS, CdS, PbS, ZnIn 2 S 4 , have absorbed extensively attention due to the favorable visible-light response ability [6][7][8] . ZnIn 2 S 4 is a typical ternary layered metal chalcogenide semiconductor with adjustable band gap of 2.06~2.85 eV, besides, the conduction band is about −1.21 eV, suggesting the intense reducing capacity of the photogenerated electrons 9 . In addition to the suitable band structure, ZnIn 2 S 4 also possess the prominent photo-stability, environmental and human friendliness in comparison to CdS and PbS 10 . Whereas, the photocatalytic property of the single ZnIn 2 S 4 is unsatisfying because of the serious carrier recombination. For pursuing the higher photocatalytic activity of ZnIn 2 S 4 , researchers have thrown tremendous efforts, including phase and morphology regulating, elements doping, cocatalystloading, defect engineering, and heterojunction constructing [11][12][13][14][15] . Among these strategies, defect engineering and heterojunction constructing are the two-effective means. In photocatalytic field, introducing anion vacancies in semiconductor can not only enhance the light absorption ability of the pristine semiconductor, but also introduce mid gap states in the band gap, which can serve as effective electron "traps" accelerating the separation efficiency of photocarriers 16 . Nevertheless, the excessive defects in photocatalyst can also act as the recombination sites of photocarriers, thus deteriorating the photocatalytic performance 17 . Therefore, regulating the defect in an appropriate concentration would ensure the high activity and stability of photocatalyst 18 . In addition, as known from the reported literatures, the only defect introduction is not enough for realizing efficient photocatalytic property.
Heterojunction constructed by coupling different materials with diverse energy level structure is another effective means to improve photocatalytic performance [19][20][21][22][23] . In recent years, Z-scheme heterostructure, especially the direct Z-scheme heterostructure, has become one of the most effective strategy for obtaining high-efficient photocatalyst 22,23 . For example, Huang et al. reported a HxMoO 3 @ZnIn 2 S 4 direct Z-scheme photocatalyst for efficient hydrogen production. The results demonstrates that the H x MoO 3 @ZnIn 2 S 4 presents a 10.5 times higher H 2 -production activity (5.9 mmol·g −1 ·h −1 ) than pristine ZnIn 2 S 4 24 . To fabricate the Z-scheme heterostructure, the primary premise is the matching band structure, in which the conduction band of one semiconductor should locate as close to the valence band of another semiconductor as possible. It is reported that the conduction band potential of MoSe 2 (about −0.45 eV 25 ) is lower than the conduction band of ZnIn 2 S 4 , but very close to its valence band (0.99 eV 9 ), which suggests that the photogenerated electrons in the conduction band of MoSe 2 are likely to recombine with the photogenerated holes in the valence band of ZnIn 2 S 4 following Z-scheme pathway. However, as known from the current literatures, MoSe 2 can only play the role of cocatalyst in ZnIn 2 S 4 /MoSe 2 instead of realizing Z-scheme charge transfer 26,27 . The question is that there is no direct and intimate interfacial connection between MoSe 2 and ZnIn 2 S 4 . The poor interfacial contact is like erecting a "wall" between the two semiconductors, seriously preventing the trajection of charge flow. Therefore, the formation of intimate interface contact became the hinge to Z-scheme photocatalyst fabrication.
Recently, defect-induced heterostructure construction have opened thought for assembling the heterostructure with specific atomic-level interfacial contact 22 . Its basic principle lies on that the defective sites with abundant coordinative unsaturation atoms and delocalize local electrons can act as the anchoring sites for other semiconductors to form a unique heterostructure contact interface with chemical bond connection 28 . The interfacial chemical bond can act as specific "bridge" accelerating charge transfer between semiconductors. In addition to the intimate interface combination, internal electric field also emerging as a viable strategy to promote Z-scheme charge transfer 29 . Under the effect of internal electric field, the photogenerated electrons in the conduction band of one semiconductor with lower Fermi level could directionally transfer to the valence band of another semiconductor with higher Fermi level, thus realizing the Z-scheme charge transfer 30 . Inspired by the above considerations, an efficient Z-scheme photocatalyst can be obtained through establishing intimate interfacial chemical bond connection between two semiconductors with specific band structure and Fermi level. Up to now, however, the interfacial bonding and internal electric field are always considered separately, the jointly modulation and their synergy effect on photocatalytic performance still remains a challenging task.
Herein, taking S vacancies-rich ZnIn 2 S 4 (S v -ZIS) and MoSe 2 as model material, through a defect-induced heterostructure constructing strategy, an interfacial Mo-S bond and internal electric field modulated Z-scheme S v -ZIS/MoSe 2 photocatalyst was fabricated. The addition of hydrazine monohydrate (N 2 H 4 • H 2 O) provides pivotal prerequisite for the formation of S vacancies and coordinative unsaturation S atoms, where the S vacancies can enhance light absorption and facilitate photocarriers separation, while the abundant coordinative unsaturation S atoms can serve as anchoring sites for Mo atoms, thus contributing the formation of Mo-S bond and the in-situ growth of MoSe 2 on the surface of S v -ZIS (as showing in Fig. 1). During photocatalytic reaction, the internal electric field induced by the different work function between S v -ZIS and MoSe 2 provide intense driving force steering the photogenerated electrons on the conduction band of MoSe 2 transfer to the valence band of S v -ZIS, that's the Z-scheme mechanism. Meanwhile, the interfacial Mo-S bond afford the fast pathways for charge transfer from MoSe 2 to S v -ZIS, thus accelerating the Z-scheme charge transfer process. This work provides a constructive reference for atomic-level interfacial and internal electric field regulating Z-scheme heterostructure for efficient photocatalytic reaction.

Results and discussion
Characterizations of as-prepared photocatalysts. The morphology and microstructure of the as-synthesized ZIS, MoSe 2 and S v -ZIS/MoSe 2 (the optimized sample) were analyzed by the SEM, TEM and HRTEM characterizations. As observed in Fig. 2a, the basic morphology of ZIS is flower-like hierarchical microsphere composed by plenty of intersecting nanoflakes, which benefits to the exposure of active surface. The TEM image in Fig. 2b further reveals the hierarchical microsphere of ZIS assembled by nanoflakes. Furtherly, as shown in the HRTEM image in Fig. 2c, the clear lattice stripes with interplanar spacing (d) of 0.32 nm can be well indexed to the (102) lattice plane of hexagonal ZnIn 2 S 4 (JCPDS:65-2023) 9 . Figures S1-S2 are the elements mapping and  NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24511-z ARTICLE EDS spectrum of ZIS, it can be clearly seen the evenly distributed Zn, In and S elements, and the atomic ratio of Zn/In/S can be calculated to be about 1.00/1.85/4.13 (as listed in Table S1), very close to the stoichiometric ratio in ZnIn 2 S 4 . Figure S3 presents the SEM, TEM and element mapping images of the S v -ZIS. It is found that the S v -ZIS appears the identical morphology and structure with ZIS, suggesting that the N 2 H 4 • H 2 O-assisted hydrothermal treatment cannot destroy the flower-like microsphere structure of ZIS. The atomic ratio of Zn/In/S in S v -ZIS sample is~1.00/1.92/ 3.35 (as displayed in Table S2), the distinctly deficient of S atom compared to that in ZIS confirms the existence of abundant S vacancies in ZIS. Figure 2d is the TEM picture of MoSe 2 , which manifests the nanosheet feature. The HRTEM image ( Fig. 2e and f) present the d-spacing of 0.65 and 0.24 nm, assigning to the (002) and (103) lattice planes of 2H-MoSe 2 (JCPDS: 29-0914), respectively 31 . Figure 2g is the SEM image of S v -ZIS/MoSe 2 , which exhibits almost the same morphology with ZIS, moreover, the ZIS and MoSe 2 in the S v -ZIS/MoSe 2 structure are undistinguishable, indicating that the MoSe 2 was grown on the surface of ZIS intimately to form a 2D/2D contact, and the introduction of MoSe 2 can hardly affect the hierarchical microsphere morphology of ZIS. The TEM image displaying in Fig. 2h and i further reveal the hierarchical flower-like microsphere structure of S v -ZIS/MoSe 2 , which could lead to the enhanced light absorption by the multilevel reflection and scattering of the incident light 32 . Furthermore, the HRTEM picture displaying in Fig. 2j shows the different lattice stripes with d value of 0.32 and 0.24 nm, respectively, which can be indexed to the (102) crystal face of hexagonal ZnIn 2 S 4 (JCPDS:65-2023) and the (103) lattice planes of 2H-MoSe 2 (JCPDS: 29-0914), respectively. The HRTEM results indicate that MoSe 2 are directly grown and attach on the ZIS nanosheets substrate. Figure 2k-p is the EDS spectra and element mapping of S v -ZIS/MoSe 2 , as displayed, the distribution of Zn, In, S elements are dense and uniform, meanwhile, the Mo and Se elements are relatively sparse but still evenly distributed. From the EDS spectrum, the mass ratio of MoSe 2 to ZIS can be calculated to be about 4.8% (as presented in Table S4), which is very close to the ratio of the added raw materials. What's more, the atomic ratio of Zn/In/S in S v -ZIS/MoSe 2 was determined to be 1.00/1.83/3.25, indicating that there is still a mass of S vacancies inside S v -ZIS/MoSe 2 .
The ZIS, S v -ZIS, MoSe 2 and S v -ZIS/MoSe 2 were further characterized by X-ray diffraction (XRD) to determine the phase composition. As displayed in Fig. 2q, the XRD pattern of MoSe 2 matches well with 2H-MoSe 2 (JCPDS:29-0914) 31 9 . It is worth noting that the S v -ZIS sample shows almost the same XRD pattern with ZIS, indicating that the introduction of S vacancies can hardly affect the size and crystal structure of ZIS. Moreover, in the XRD patterns of S v -ZIS/MoSe 2 , in addition to the peaks of hexagonal ZIS, a new peak at about 13.7°can be well assigned to the (002) crystal face of MoSe 2 , reconfirming the successful synthesis of S v -ZIS/MoSe 2 composite.
To further characterize the chemical structures of the assynthesized photocatalyst, the Raman spectra were carried out (shown in Fig. 2r). As observed in the Raman spectra of MoSe 2 , the peaks located at 235.4, 277.4 and 330.8 cm −1 stem from the A 1g , E 2g and B 2g modes of 2H-MoSe 2 , respectively, while the peak at 142.1 cm −1 is associated to the E 1g mode of the in-plane bending of Se atoms in 2H-MoSe 2 32 . For the Raman spectra of S v -ZIS, the peaks located at 244.8 and 348.9 cm −1 can be severally assigned to the F 2g and A 1g modes of ZnIn 2 S 4 . Furtherly, as for the S v -ZIS/MoSe 2 (the red line), in addition to the E 1g mode of 2H-MoSe 2 , and the F 2g and A 1g modes of ZnIn 2 S 4 , a new emerging peak situated at about 404.9 cm −1 can be indexed to the Mo-S bonding state 33 , suggesting that the S v -ZIS and MoSe 2 were combined intimately by Mo-S bond. Additionally, it can be observed that all the peaks in S v -ZIS/ MoSe 2 exhibited evidently blue-shift compared to that in S v -ZIS, further revealing the intense chemical coupling effect between the S v -ZIS and MoSe 2 34 .
To further testify the existence of S vacancies, the electron paramagnetic resonance (EPR) was carried out (Fig. 2s). For the original ZIS sample, the EPR intensity can hardly be observed, in comparison, the S v -ZIS sample shows the sharply increased EPR signal at a g-factor of 2.009, confirming the abundant S-vacancies in S v -ZIS 35,36 . In addition, it is interesting to observe that the EPR intensity of S v -ZIS/MoSe 2 exhibits slightly decreased compared to that of S v -ZIS, which should be contributed to the bonding effect among Mo and unsaturated S in S v -ZIS, decreasing the number of unpaired electrons, but the S vacancies in ZIS have not been sewed up by compositing MoSe 2 37 .
The X-ray photoelectron spectroscopy (XPS) was applied to investigate the surface composition and chemical states of ZIS, S v -ZIS and S v -ZIS/MoSe 2 , and the results are showing in Fig. 3. As can be found from the survey spectrum ( Fig. 3a), the Zn, In and S peaks are coexisting in ZIS and S v -ZIS, in comparison, Mo and Se peaks can also be observed in the S v -ZIS/MoSe 2 , which is agree with the EDS test results. As observed in Fig. 3b, the S 2p 3/2 and 2p 1/2 of the original ZIS located at 161.72 and 162.97 eV, respectively, in accordance with the reported literature 36 . In comparison, the S 2p 3/2 and 2p 1/2 of S v -ZIS presented evident negative-shift of about 0.14 eV and 0.19 eV, respectively, verifying the generation of S vacancies in ZIS. The S-vacancies can serve as strong electron-withdrawing group for facilitating the ZIS electrons transfer to S-vacancies, thus decreasing the equilibrium electron cloud density of S atoms inside ZIS, and further leading to the decreased binding energy 38,39 . Furtherly, it can be noted that the S 2p 3/2 and 2p 1/2 of S v -ZIS/MoSe 2 exhibited a positiveshift of about 0.13 and 0.17 eV compared to that of S v -ZIS, which should be caused by the strong interfacial interaction between MoSe 2 and S v -ZIS 34 . Besides, as shown in Fig. 3c and d, the Zn 2p and In 3d in S v -ZIS also exhibited a slightly negative-shift compared to that in ZIS, which could be explained that the generation of S vacancies leading to the decreased coordination number of Zn and In 37 . After combining with MoSe 2 , the Zn 2p and In 3d peaks re-shift to the high binding energy region, revealing that the bonding effect between Mo atoms in MoSe 2 and unsaturated coordination S in S v -ZIS contributing to the slightly increased electron cloud density around Zn and In. Interestingly, it can also be observed that the binding energy variation of Zn 2p in ZIS, S v -ZIS, and S v -ZIS/MoSe 2 are more notable than that of In 3d, revealing that the Mo were mainly bonded with the S around Zn sites 37 . What's more, according to the XPS peak area, the actual atomic ratio of Zn/In/S in ZIS, S v -ZIS and S v -ZIS/MoSe 2 are 1.00/2.15/3.87, 1.00/2.20/3.29, and 1.00/2.14/3.36, respectively. The lower S atom ratio in S v -ZIS and S v -ZIS/MoSe 2 further confirm the presence of abundant S vacancies. As shown in Fig. 3e, the peaks at 228.05 and 230.5 eV can be attributed to Mo 3d 5/2 and 3d 3/2 of Mo 4+ in MoSe 2 , meanwhile, the peak at 227.1 eV verified the formation of Mo-S bond 40 . Figure 3f is the Mo 3p spectrum, as observed, four distinct XPS peaks can be distinguished, where the peaks at 400. 55  Photocatalytic H 2 evolution activity measurements. The photocatalytic H 2 evolution were evaluated under the visible light (λ > 420 nm) irradiation, the corresponding test results are showing in Fig. 4. As shown in Fig. 4a and b, all the tested samples exhibit H 2 production activity except for MoSe 2 . The pristine ZIS exhibits the poor H 2 production activity of about only 3.36 mmol•g −1 ·h −1 , in comparison, the S v -ZIS presents a slightly improved H 2 evolution rate of 4.77 mmol•g −1 ·h −1 . The improved photocatalytic performance of S v -ZIS should be ascribed to the accelerated photocarriers separation induced by S vacancies as the electrons trap. Furtherly, the introduction of MoSe 2 gave rise to the distinctly improved H 2 evolution activity, and the H 2 evolution rate of S v -ZIS/MoSe 2 increased with the mass ratio of MoSe 2 to ZIS increasing. Until the mass ratio of MoSe 2 to ZIS reaches to 5.0%, the H 2 evolution rate reaches to the highest of 63.21 mmol•g −1 ·h −1 , which is about 18.8 and 13.3 times higher than that of pristine ZIS and S v -ZIS, respectively, and superior to the recently reported ZnIn 2 S 4 -based photocatalytic system (as listed in Table S6). It can also be observed that the S v -ZIS-5.0MoSe 2 (synthesized by mixing S v -ZIS and MoSe 2 by ultrasound) performs obvious inferior H 2 evolution property compared to that of S v -ZIS/5.0MoSe 2 , indicating that the in-situ growth of MoSe 2 on S v -ZIS connecting by Mo-S bond plays critical influence on the photocatalytic performance of the ZIS-MoSe 2 composite, which should be attributed to that the Mo-S bond could facilitate the charge transfer between S v -ZIS and MoSe 2 . Besides, Fig. S5 shows the wavelength dependent hydrogen evolution efficiency of S v -ZIS/MoSe 2 , which was tested following the similar procedure of photocatalytic H 2 evolution, except that the band-pass filter was equipped to obtain monochromatic incident light (λ=380, 420, 500 and 600 nm). The detailed test results and the light power of different monochromatic light are displaying in Table S5. Accordingly, the AQY of photocatalytic H 2 evolution over the S v -ZIS/MoSe 2 photocatalyst can be calculated (the detailed calculation process is shown in the Supporting Information) and the action spectrum was displayed in Fig. 4c. As observed, the action spectrum of S v -ZIS/MoSe 2 matches well with the UV-vis absorption spectra, besides, the AQY values of S v -ZIS/MoSe 2 are about 93.08% (380 nm), 76.48% (420 nm), 29.7% (500 nm) and 0.15% (600 nm), indicating the favorable optical absorption and utilization capacity of S v -ZIS/ MoSe 2 photocatalyst. Fig. S6 is the AQY of ZIS and S v -ZIS, it can be observed that under different monochromatic light wavelength, the AQY of S v -ZIS are larger than that of ZIS, suggesting the more efficient photons to H 2 conversion ability of S v -ZIS, which should be caused by the enhanced light absorption and the promoted photocarriers separation efficiency by introducing abundant S-vacancies in S v -ZIS. In addition to the excellent photocatalytic H 2 evolution efficiency, the recycling stability is also a pivotal factor for the practical application of photocatalyst. As discerned in Fig. 4d, the H 2 evolution amount of the optimized S v -ZIS/MoSe 2 photocatalyst remains about 90.5% after 20 h of 5 cycles of photocatalytic tests, signifying the favorable Photophysical and Electrochemical Properties. Figure 5a is the UV-vis absorption spectra of ZIS, S v -ZIS, MoSe 2 and S v -ZIS/ MoSe 2 . It is apparent that the MoSe 2 shows the intense light absorption in the whole UV-vis light range, which should be caused by its dark black color. Meanwhile, it can be observed that light absorption intensity of S v -ZIS is higher than that of ZIS, indicating that the introduction of S vacancies can influence the band structure of ZIS. Furtherly, after combining with MoSe 2 , the light absorption of S v -ZIS/MoSe 2 increased again compared to S v -ZIS. The improved light absorption is in favor of the generation of photocarriers, and beneficial for the enhancement of photocatalytic performance 9 . Figure 5b is the PL spectroscopy. As displayed, under the 375 nm excitation wavelength, the pristine ZIS displays a prominent emission peak, indicating the intense recombination of photogenerated carriers inside ZIS. In comparison to ZIS, the emission peak intensity of S v -ZIS decreases lightly, which should be contributed to that S vacancies can act as electrons trap for facilitating the photocarriers separation. It is  worth noting that the PL signal of S v -ZIS/MoSe 2 sample is further quenched compared to that of S v -ZIS, revealing the positive effect of MoSe 2 for suppressing the recombination of photocarriers. Figure 5c is the photocurrent response. As observed, all the tested samples exhibit the light-response characteristic under the FX-300 Xe lamp. Obviously, the photocurrent density is in the order of S v -ZIS/MoSe 2 > S v -ZIS > ZIS. The highest photocurrent density of S v -ZIS/MoSe 2 reveals the most accelerated photocarriers separation and migration efficiency. Figure 5d is the electrochemical impedance spectroscopy (EIS). As compared, MoSe 2 express the smallest semicircle, meanwhile, the semicircle of ZIS is the largest. Obviously, the semicircle of S v -ZIS is slightly lower than that of ZIS, and the semicircle of S v -ZIS/MoSe 2 is significantly decreased than that of pristine ZIS and S v -ZIS, manifesting that the introduction of S vacancies and the combination with MoSe 2 can decrease the interfacial charge transfer resistance, which is in favor of photogenerated carriers transfer and separation, and finally facilitate the photocatalytic property.
In order to investigate the effects of the MoSe 2 to ZIS mass ratio on the photocatalytic performance of S v -ZIS/MoSe 2 composites. The light absorption, photocarriers separation and photocurrent density of S v -ZIS/MoSe 2 photocatalysts with different mass ratio of MoSe 2 to ZIS were also characterized by UV-vis absorption, steady-state PL spectroscopy and photocurrent response. As observed in Fig. S7, with increasing the mass ratio of MoSe 2 to ZIS, the light absorption intensity enhance gradually. It is worth mentioning that the S v -ZIS/7.0MoSe 2 sample displays the strongest light absorption ability, but its photocatalytic H 2 production performance is not the best (as known from Fig. 4a), suggesting that the light absorption is not the only decisive factor for the photocatalytic activity. Fig. S8 is the PL spectra, it can be observed that the PL peak of S v -ZIS/ 5.0MoSe 2 is the lowermost, revealing the most effective photocarriers separation when the mass ratio of MoSe 2 to ZIS is 5%, which directly explains why the S v -ZIS/5.0MoSe 2 sample has the best photocatalytic performance. Figure S9 shows the photocurrent response. As displayed, the S v -ZIS/5.0MoSe 2 shows the highest photocurrent density, which is the result of highefficiency separation and transfer of photogenerated electron and hole, further revealing the optimum photocatalytic performance of S v -ZIS/5.0MoSe 2 . As known from the above results, the prominent photocatalytic performance requires the coordination among the efficient light absorption, photocarrier separation and transfer ability.
Mechanism analysis. Furtherly, the bandgap value (E g ) of the tested sample can be obtained from the Kubelka-Munk function vs. the energy of incident light plots 41 . As displayed in Fig. 6a, the E g of ZIS, S v -ZIS and S v -ZIS/MoSe 2 can be estimated to be 2.35, 2.28 and 2.19 eV, respectively. The narrower E g is beneficial for the incident light absorption and photocarriers generation, thereby contributing to the photocatalytic property 42 . The Mott-Schottky (M-S) plot can be obtained by the following formula of C À2 sc ¼ 2 , in which C SC represents space charge capacitance, ɛ represents the dielectric constant, ɛ 0 represents the permittivity of vacuum, e represents the single electron charge, N D represents the charge carrier density, E fb represents the flat band potential, k B represents the Boltzmann constant, and T represents the temperature, E represents the electrode potential 9 . As displayed in Fig. 6b-d, the E fb of ZIS, S v -ZIS and MoSe 2 can be determined to be −0.96, −0.9 and −0.1 V (vs. NHE), respectively, by extending the linear part of M-S plots. Besides, all the tested samples exhibit the positive slope of M-S plots, indicating the n-type semiconductor traits 43 . As known, the conduction band potential (E CB ) of n-type semiconductor is~0 .2 eV negative than the E fb 44 , thus the E CB of ZIS, S v -ZIS and MoSe 2 can be discerned to −1.16, −1.1 and −0.3 V (vs. NHE), respectively. According to the equation of E VB = E CB + E g (E VB is the potential of valence band (VB)), the E VB of the ZIS and S v -ZIS can be estimated to 1.19 and 1.18 V vs. NHE, respectively. According to the reported literature, the E g of MoSe 2 is about 1.89 eV, therefore, the E VB of MoSe 2 can be determined to be 1.59 eV 25 .
The work function (ɸ) is an important nature for reflecting the escaping ability of free electron from Fermi level (E f ) to vacuum level 45 . To investigate the mechanism for the excellent photocatalytic performance of S v -ZIS/MoSe 2 , the ultraviolet photoelectron spectroscopy (UPS) with He I as the excitation source was conducted. As displayed in Fig. 6e, the secondary cutoff binding energy (E cutoff ) of S v -ZIS and MoSe 2 can be respectively determined as 17.65 and 16.87 eV, by extrapolating the linear part to the base line of the UPS spectra. Based on the formula of ɸ=hv-E cutoff , the ɸ of S v -ZIS and MoSe 2 can be calculated as 3.57 and 4.35 eV, respectively. Hence, the E f of S v -ZIS and MoSe 2 can be determined as −0.93 and −0.15 V (vs. NHE), respectively. Based on the above calculation and analysis results, the detailed band structure of S v -ZIS, MoSe 2 and S v -ZIS/MoSe 2 were depicted in Fig. 6f. As observed, the E f of MoSe 2 is below that of S v -ZIS, hence, when S v -ZIS and MoSe 2 contact and form an intimate interface, the free electrons in S v -ZIS with high E f would spontaneously diffuse to MoSe 2 with low E f , until a new equilibrium state E f fabricated. The electron drifting from S v -ZIS to MoSe 2 result in the charge redistribution on the interface between S v -ZIS and MoSe 2 , in which the interface near S v -ZIS side is positively charged, while negatively charged near the MoSe 2 side, as result, an internal electric field from S v -ZIS to MoSe 2 was built 46 .
To further reveal the photocatalytic reaction mechanism of S v -ZnIn 2 S 4 /MoSe 2 heterostructure, the density functional theory (DFT) calculations were conducted out. Figure 7(a) is the optimized structure of S v -ZnIn 2 S 4 /MoSe 2 heterostructure, where the coordinative unsaturation S atoms was simulated by breaking two Zn-S bonds in the surface of ZnIn 2 S 4 . According to Population analysis and Hirshfeld analysis results, the population of Mo 001 -S 018 is 0.34, and the transferred charge between MoSe 2 and S v -ZnIn 2 S 4 is 0.12 | e | . The above results directly demonstrate the intense bonding effect between the Mo atom in MoSe 2 and the coordinative unsaturation S atom in ZnIn 2 S 4 . Figure 7(b) shows the side view of charge density difference of S v -ZnIn 2 S 4 / MoSe 2 , where the red and blue iso-surfaces denote the accumulation and depletion of electron density, respectively. As observed, the electron cloud density presents distinctly localized distribution between the Mo atom in MoSe 2 and the coordinative unsaturation S atoms in S v -ZnIn 2 S 4 , which more intuitively manifests the intense bonding effect between Mo and S. In addition, it can be noted that the surface of MoSe 2 was dominantly covered by red color, while S v -ZnIn 2 S 4 was chiefly filled by blue color, suggesting that the electrons in S v -ZnIn 2 S 4 were transfer to MoSe 2 along the intimate heterointerface, which would subsequently induce the internal electric field in S v -ZnIn 2 S 4 /MoSe 2 heterostructure 47 .
Accordingly, the photocatalytic reaction mechanism of S v -ZIS/ MoSe 2 can be elaborated in Fig. 7c. Under the irradiation of visible light, a mass of photoinduced electrons (e -) with enough energy would transfer from the VB of S v -ZIS and MoSe 2 to the CB of S v -ZIS and MoSe 2 , respectively, while the holes (h + ) be left on the VB of S v -ZIS and MoSe 2 , respectively. It should be mentioned that the abundant S vacancies inside ZIS could introduce new donor level in the band gap of ZIS, which can act as efficient electrons trap to suppress the photogenerated electron-hole pairs recombination 48 . Furtherly, under the driving effect of the internal electric field, the electrons on the CB of MoSe 2 would migrate to the VB of S v -ZIS to recombine with the holes. The Mo-S bond acting as atomic-level interfacial "bridge" can promote the photoexcited carriers migration between S v -ZIS and MoSe 2 , thus significantly accelerating the Z-scheme charge transfer. To validate the Z-scheme charge transfer mechanism, the SPV and EPR measurements were carried out. Figure 7d is the SPV spectra of S v -ZIS, MoSe 2 and S v -ZIS/MoSe 2 samples. It is noted that the pristine MoSe 2 presents no SPV signals in the whole wavelength, suggesting the poor photocarriers separation efficiency inside the MoSe 2 , that's why MoSe 2 performed very poor hydrogen evolution. In comparison, a significant positive photovoltage response can be observed in the SPV spectra of S v -ZIS, suggesting that the holes migrate to the surface of S v -ZIS, which is the typical trait of n-type semiconductor 49 . Meanwhile, the SPV response of S v -ZIS/MoSe 2 is significantly lower than that of S v -ZIS, which means that fewer photogenerated holes migrate to the surface of S v -ZIS/MoSe 2 . This phenomenon should be contributed to that the photogenerated electrons on the CB of MoSe 2 transfer to the VB of S v -ZIS and recombine with the photogenerated holes, that's the Z-scheme mechanism 50 . EPR spin-trapping experiment with DMPO as spin-trapping reagent was further proceeded to support the Z-scheme charge transfer mechanism in S v -ZIS/MoSe 2 . As displayed in Fig. 7e, 52 . Therefore, the •O 2should be the reaction product between the photoinduced electrons on the CB of S v -ZIS and O 2 (the CB potential of S v -ZIS is about −1.10 eV, lager than the redox potential of O 2 / • O 2 -), indicating that a mass of photogenerated electrons were accumulated on the CB of S v -ZIS under irradiation of visible light, which should be contributed by the recombination between the electron on the CB of MoSe 2 and the hole on the VB of S v -ZIS, thus verifying the direct Z-scheme charge migration mechanism. Above SPV and EPR spin-trapping technique provides the direct proof for the direct Z-scheme charge transfer mechanism inside the S v -ZIS/MoSe 2 photocatalyst.
In summary, we have successfully demonstrated an interfacial Mo-S bond and internal electric field modulated Z-scheme S v -ZnIn 2 S 4 /MoSe 2 photocatalyst through a defect-induced heterostructure constructing strategy for boosting the photocatalytic H 2 evolution performance. The internal electric field provide the necessary driving force steering the photogenerated electrons on the conduction band of MoSe 2 transfer to the valence band of S v -ZnIn 2 S 4 following the Z-scheme mechanism, while the interfacial Mo-S bond creates direct charge transfer channels between S v -ZnIn 2 S 4 and MoSe 2 , further accelerates the Z-scheme charge transfer process. What's more, the abundant S-vacancies also contribute to the enhanced light absorption and accelerated photocarriers separation. The above factors together lead to the efficient photocatalytic performance of the S v -ZnIn 2 S 4 /MoSe 2 . Specifically, the optimized photocatalyst exhibits a high AQY of 76.48% at 420 nm, and an ultrahigh H 2 evolution rate of 63.21 mmol·g −1 • h −1 under visible light (λ > 420 nm), which is about 18.8 times higher than that of pristine ZnIn 2 S 4 . Besides, the S v -ZnIn 2 S 4 /MoSe 2 also shows favorable recycling stability by remaining above 90% rate retention after 20 h of 5 continuous photocatalytic tests. This work not only provides an efficient direct Z-scheme ZnIn 2 S 4 -based heterostructure photocatalyst, but also affords a beneficial prototype for designing other Z-scheme photocatalyst for efficient green energy conversion.
Synthesis of ZnIn 2 S 4 and S v -ZnIn 2 S 4 . In a representative experiment, InCl 3 (1 mmol), Zn(CH 3 COO) 2 ·2H 2 O (0.5 mmol), and TAA (4 mmol) were orderly dissolved into 50 mL deionized water, and then stirred at room temperature for 30 min. Thereafter, the clear solution was poured into 100 mL stainless steel autoclave, and maintained at 180°C oven for 18 h. After cooling naturally to indoor temperature, the sediment was separated by centrifugation, followed by washing with deionized water and ethanol, and drying at 60°C for 10 h. The obtained yellow powder ZnIn 2 S 4 were labeled as ZIS. S v -ZnIn 2 S 4 was prepared via a N 2 H 4 ·H 2 O-assisted hydrothermal method. Typically, 100 mg the as-synthesized ZIS was dispersed into 20 mL deionized water for 1 h, then, 5 mL N 2 H 4 ·H 2 O was added into the mixing solution and stirred for another 30 min. After that, the mixture was transfer to 50 mL stainless steel autoclave, and maintained at 240°C oven for 5 h. Finally, the precipitate was separated by centrifugation, and washing with deionized water for several times, then drying at 60°C for 10 h. The obtained light-yellow powder was labeled as S v -ZIS.
Synthesis of S v -ZnIn 2 S 4 /MoSe 2 heterostructure. The S v -ZnIn 2 S 4 /MoSe 2 heterostructure were synthesized by the similar process with S v -ZnIn 2 S 4 , except that Na 2 MoO 4 ·2H 2 O and Se powders were added into the mixture. The S v -ZnIn 2 S 4 / MoSe 2 with different mass ratio of MoSe 2 to ZnIn 2 S 4 (0.5%, 1.0%, 3.0%, 5.0%, and 7.0%) were synthesized by adjusting the addition of Na 2 MoO 4 ·2H 2 O and Se, and the synthesized samples were labeled as S v -ZIS/0.5MoSe 2 , S v -ZIS/1.0MoSe 2 , S v -ZIS/ 3.0MoSe 2 , S v -ZIS/5.0MoSe 2 , S v -ZIS/7.0MoSe 2 , respectively. For comparison, the pure MoSe 2 was prepared following the above steps without adding ZIS. Besides, the S v -ZIS-5.0MoSe 2 mixture was also fabricated by ultrasonic mixing the S v -ZIS with MoSe 2 for 1 h.  X-ray spectrometer (EDS), and JEM-2100 plus transmission electron microscope (TEM). The crystalline and phase information were characterized by Bruker D8 Advance X-ray diffraction (XRD). The chemical states were investigated by Thermo ESCALAB 250 XI X X-ray photoelectron spectroscopy (XPS, monochromatic Al Kα radiation), and the XPS data was calibrated by C 1 s spectrum (binding energy is 284.8 eV). The light absorption property was researched by the PerkinElmer Lambda 750 S UV-vis spectrophotometer using barium sulfate as standard reference. The recombination of photogenerated carriers was tested by F-4600 spectrofluorometer (375 nm excitation wavelength). The secondary cutoff binding energy was measured by AXIS SUPRA X-ray photoelectron spectroscopy with He I as the excitation source. The surface photovoltage (SPV) measurement were carried out on the system consisting a 500 W Xe lamp source equipped with a monochromator, a lock-in amplifier with a light chopper, a photovoltaic cell, and a computer. The Raman spectra were conducted on LabRAM HR Evolution Raman spectrometer with 325 nm excitation wavelength to analysis the composition. The electron paramagnetic resonance (EPR) measurement was conducted on JEOL JES-FA200 EPR spectrometer with a 9.054 GHz magnetic field. The 5,5-dimethylpyrroline N-oxide (DMPO) was adopted as spin-trapping reagent and the •O 2and •OH were tested in methanol and aqueous solution, respectively.
Photocatalytic water splitting for hydrogen evolution. The hydrogen production experiments were proceeded on Labsolar-6A (Beijing Perfectlight). Typically, photocatalyst (50 mg) was ultrasonically suspended into 100 mL solution involving 0.1 M ascorbic acid sacrificial agent. Prior to exerting light, the reaction system was degassed for 1 h to thoroughly exclude the air and the dissolved oxygen in reaction system. Then the reaction was proceeded under PLS-SEX300D 300 W Xenon lamp (Beijing Perfectlight) with a 420 nm cut-off filter. The light intensity was determined by PLMW2000 photoradiometer (Beijing Perfectlight) to be about 254 mW/cm 2 . The generated hydrogen was analyzed by GC 7900 gas chromatograph (Techcomp, 5 Å molecular sieve stainless steel packed column, Ar as carrier gas and TCD detector).
Photoelectrochemical and electrochemical measurements. All the electrochemical and photoelectrochemical measurements were conducted by a threeelectrode system on CHI-660E electrochemical workstation. In the typical threeelectrode system, the working electrode was a piece of nickel foam coating with the as-prepared photocatalyst, the reference electrode was Hg/HgO, while the counter electrode was Pt wire. The electrolyte was 0.5 M Na 2 SO 4 aqueous solution. The electrochemical impedance spectroscopy (EIS) was conducted under open-circuit potential with 0.01 to 1×10 5 Hz frequency range and 0.005 V AC amplitude. The photocurrent response was tested under FX-300 Xe lamp. Mott-Schottky (M-S) plots were collected from −1 to −0.2 V under 10 kHz frequency and 0.01 V amplitude.
The working electrode was fabricated as follows: a certain amount of photocatalyst, carbon black and polyvinylidene fluoride were weighted according to the mass ratio of 8:1:1, and then dispersed into N-methyl-2-pyrrolidone to gain a homogeneous paste. The paste was daubed on a piece of pre-cleaned 1×1 cm 2 FTO collector, and then dried in 60°C vacuum for 1 h.

Data availability
The experimental data that support the findings of this study are available from the corresponding author upon reasonable request. Source data are provided with this paper.