Simple hydrothermal synthesis of g-C3N4/Ni9S8 composites for efficient photocatalytic H2 evolution

The prompt recombination between photogenerated electrons and holes is the common problem for improving the hydrogen evolution performance of a photocatalyst, which could be solved greatly by composite co-catalysis. Herein, a simple hydrothermal reaction was utilized to prepare g-C3N4/Ni9S8 composite photocatalysts. Through electroless nickel plating, Ni9S8 nanostructure was homogeneously grown onto the g-C3N4 surface by using sodium hypophosphite as reducing agent. With the optimum loading amount of Ni9S8, the acquired composite, compared with the raw g-C3N4, presented a significant increase in hydrogen evolution rate under visible light irradiation, which was measured as 355.7 μmol g−1 h−1 at 7 °C, being 21.2 times that of raw g-C3N4. The mechanism for the hydrogen evolution reaction over the present g-C3N4/Ni9S8 composite photocatalysts was discussed in detail.


Introduction
Nowadays, human beings are suffering from energy shortage, environmental pollution and extreme climate, because the increasing large-scale use of fossil energy will cause the depletion of easily available energy sources, giving out a pessimistic perspective for the reserve of fossil energy, and release a lot of waste gases and greenhouse gases, resulting in great impact on the earth's ecology [1]. Therefore, the development of new energy is very urgent for mankind. Hydrogen energy, as a pearl of new energy, has attracted much attention because of its high calorific value, no carbon emission and facile recyclability [2]. And since 1972 when Fujishima and Honda discovered that TiO 2 could decompose water into hydrogen and oxygen under ultraviolet light [3], it seems for human beings to find a shortcut to solve the energy problem once and for all. Thus now photocatalysis is such an exciting technology that attracts numerous researchers to study.
It is soon recognized by scientists that photocatalysis is based on the photoelectric conversion of semiconductors, and ultimately the conversion of light into chemical energy [4]. Therefore, the exploration of semiconductor photocatalysts has been rapidly launched. Today, many photocatalytic semiconductors such as ZnO, Ga 2 O 3 , SrTiO 3 and ZnS [5][6][7][8][9] have been found to be similar with TiO 2 in photocatalytic decomposition of water. However, these semiconductor photocatalysts can only respond to ultraviolet light because of their wide bandgap. To effectively make use of sunlight (containing only 5% ultraviolet light), which is the inexhaustible energy source for human beings, researchers have to search for narrow bandgap semiconductors for photocatalytic hydrogen evolution, and have developed some, such as CdS, MoS, Cu 2 O and many other [10][11][12]. But their poor photochemical stability, heavy metal pollution on water and/or less earth reserves limit their large-scale application [13].
In 2009, Wang et al. first found that g-C 3 N 4 has the ability of photocatalytic decomposition of water under visible light [14]. Such ability of this kind of organic semiconductors has soon aroused the great interest of scientists. In literature, a large number of studies have shown that g-C 3 N 4 owns a bandgap of roughly 2.78 eV, which can respond to light sources with wavelength below 460 nm [15]. At the same time, it has the advantages of cadenced carbon-tonitrogen framework (thus providing more active reaction sites for electron donor/acceptors), strong photochemical stability, low-cost, easy availability, non-toxicity and pollution free for environment [16][17][18]. In spite of these virtues, however, there are still some disadvantages on such organic semiconductors for photocatalytical hydrogen evolution, such as high carrier recombination rate due to a large number of intrinsic defects like N and C vacancies, poor electrical conductivity, limited ability to use light source with wavelength only below 460 nm, as well as difficultly dispersible layered structure, which seriously degrades the photocatalytic activity [13].
To reduce the recombination of photogenerated carriers (holes and electrons) in photocatalysts, cocatalysis is an excellent solution [19]. For example, noble metals like Au, Pt, Pd, Rh and so on have excellent catalytic effect due to their unique surface plasmon effect, which can also serve as good co-catalysts [20]. Regrettably, noble metals cannot widely be used in photocatalytic industry due to their high cost and scarce reserves. Therefore, it is very important to find low-cost co-catalysts for photocatalytic hydrogen evolution. In addition, by combining different semiconductor photocatalysts together to form a built-in electric field, photogenerated electrons are easily transferred from the optical semiconductor to the co-catalytic semiconductor. At this time, because of the Schottky barrier, the photogenerated electrons are difficult to return to the original semiconductor catalyst, thus the separation of photogenerated carriers is realized [15]. Meanwhile, for the co-catalytic semiconductor, low overpotential and good conductivity are very helpful for hydrogen evolution [21]. Therefore, it is a feasible idea to find a semiconductor with low overpotential and good conductivity to replace noble metals.
In literature, NiS has been regarded as an excellent co-catalyst in photocatalytic hydrogen evolution, and hopefully it can replace noble metals. This is due to the following advantages of NiS: (1) high-conductivity due to low bandgap, (2) low surface work function that can effectively reduce the reaction activation energy or over potential, (3) high power conversion efficiency, and (4) easy preparation, lowcost as well as environmental friendliness [10]. Therefore, a lot of researches have been executed on the co-catalysis of NiS in photocatalysis, such as CdS/NiS, TiO 2 /NiS, g-C 3 N 4 /NiS and MoO 3 /NiS [22][23][24][25]. However, since nickel sulfides can exist in different component including NiS, NiS 2 , Ni 3 S 4 , Ni 4 S 3 , Ni 9 S 8 and so forth [26], it is difficult for researchers to study them fully. Among them, Ni 9 S 8 is less involved because it is not easily synthesized. As far as we know, no one has reported any composite of Ni 9 S 8 and g-C 3 N 4 . Moreover, Ni 9 S 8 has a near zero bandgap compared with NiS, which implies that Ni 9 S 8 has better conductivity and is more conducive to the transmission of photogenerated electrons, and thus the composite of Ni 9 S 8 can effectively improve the photocatalytic activity of MoS 2 [27]. In addition, Yang et al. reported that Ni 9 S 8 has a low overpotential and can easily expose surface active sites [28]. Therefore, it is much desirable to prepare the composite of g-C 3 N 4 and Ni 9 S 8 on photocatalytic hydrogen evolution.
For the synthesis of the composites of g-C 3 N 4 with nickel sulfides, several methods including hydrothermal synthesis, calcination, precipitation and photodeposition have been proposed [29][30][31][32]. Among them, the calcination methods are energyconsuming and often produce off various polluting gases, and conventional precipitation and photodeposition are difficult to synthesize Ni 9 S 8 . Hydrothermal synthesis has attracted much attention because of its simple operation, safety, high efficiency, energy saving and environmental friendliness. However, the conventional hydrothermal methods to prepare the composites of g-C 3 N 4 and nickel sulfides generally involve in complicated synthesis steps and produce a mixture of several nickel sulfides (NiS and Ni x S y ).
Therefore, in the present work, we adopt a facile, green, one-step hydrothermal route, which is based on the principle of electroless nickel plating, to grow Ni 9 S 8 nanostructures onto the pre-prepared g-C 3 N 4 nanosheet, obtaining the g-C 3 N 4 /Ni 9 S 8 composite [22,33]. In the synthesis of g-C 3 N 4 /Ni 9 S 8 composite, in order to form Ni 9 S 8 , NaH 2 PO 2 was used, which is commonly used in electroless nickel plating, to reduce Ni 2? to metal Ni, and then to conduct the redox reaction with S 0 and S 2to obtain the composites of g-C 3 N 4 with pure Ni 9 S 8 . Compared with the traditional hydrothermal methods, the reduction and sulfuration of Ni 2? were concentrated in onestep hydrothermal synthesis, much reducing the processes. Moreover, the as-acquired g-C 3 N 4 /Ni 9 S 8 composite has a uniform dispersion of Ni 9 S 8 nanostructures on the g-C 3 N 4 nanosheet. Because it possesses high-conductivity, low surface work function and high electron mobility, the as-acquired composite photocatalyst has excellent photocatalytic performance for hydrogen evolution, presenting a H 2 evolution rate as high as 355.7 mmol g -1 h -1 at 7°C. The successful preparation of the present composite will also provide a new perspective for developing other high-performance heterostructure photocatalysts for hydrogen evolution.

Synthesis of g-C 3 N 4 particles
Bulk g-C 3 N 4 samples were first synthesized by directly heating melamine [34]. Typically, in an open box furnace, a half-covered crucible with 5 g of melamine powder is heated from room temperature to 550°C at a speed of 5°C min -1 , soaking at this temperature for 4 h. Afterwards, the furnace was cooled down to ambient temperature, and then the resultant bulk sample was ground, finally obtaining yellow g-C 3 N 4 powders.

Synthesis of g-C 3 N 4 /Ni 9 S 8 nanocomposites
A facile one-step hydrothermal approach was applied in this work to synthesize the proposed g-C 3 N 4 /Ni 9 S 8 nanocomposites. Typically, in a beaker with 50 mL deionized water, 100 mg (1.111 mmol) of the as-obtained g-C 3 N 4 powders was homogeneously dispersed through vigorous stirring for 30 min under ultrasonicating. Then, in order to obtain g-C 3 N 4 / Ni 9 S 8 composites with different loading amounts of Ni 9 S 8 , a series of designed feed amount of nickel acetate (0.05, 0.1, 0.2, 0.3, 0.4, 0.5, 0.7 and 0.9 mmol, respectively) were added into the dispersion solution.
And thiourea with an amount of three times that of nickel acetate was mixed in the dispersion solution together. After further strong stirring for 30 min, 300 mg sodium hypophosphite was mixed in the solution system. After homogeneous mixing, the solution was transferred into an autoclave (100 mL) and then kept in an oven at 140°C for 8 h. By simply shutting the electricity of the furnace, it was cooled down to ambient temperature. And then the autoclave was taken out from the oven. The resultant precipitates were collected after centrifuging, washed twice with 30 mL deionized water as well as absolute ethanol, respectively, and dried out at 60°C for 8 h. Finally, the proposed, dark blue powders could be acquired.

Synthesis of g-C 3 N 4 /Pt nanocomposite
The g-C 3 N 4 /Pt nanocomposite was prepared by a modified photodeposition method [35]. Typically, 50 mg g-C 3 N 4 was firstly dispersed into 100 mL of 10 vol.% TEOA aqueous solution. After 30 min of ultrasonic dispersion, the solution was transferred to a photoreaction cell. And then 1 wt% chloroplatinic acid (in aqueous solution) was added, compared with the amount of g-C 3 N 4 . After that, the air in the reaction system was pumped off, and then a light irradiation was carried out with a 300 W Xe lamp for 1 h to ensure that Pt could be fully deposited on the surface of g-C 3 N 4 . Finally, after pumping out the hydrogen produced during the deposition of Pt, the test on photocatalytic hydrogen evolution over the as-prepared g-C 3 N 4 /Pt nanocomposite was directly carried out.

Materials characterization
In this work, a Zeiss GEMINISEM 500 field emission scanning electron microscope (FE-SEM, Germany) and JEOL JEM-2100F transmission electron microscope (TEM, Japan) were applied to examine the morphology and microstructure of the samples. Then a Thermo ESCALAB MKII X-ray photoelectron spectroscope (XPS, Thermo VG Scientific Ltd., UK) was used to explore their elemental composition and chemical state. With the C1s line (284.8 eV) as reference, the recorded results were corrected. In order to identify their phase compositions, X-ray diffraction (XRD) analysis was executed (GI-XRD, Japan; Cu Ka radiation, k = 1.5418 Å ). For the analysis, the corresponding scanning rate was set as 4°min -1 with 1°of X-ray incidence angle in continuous scanning mode.
The nitrogen adsorption-desorption isotherms were recorded on a Quantachrome Autosorb-iQ Adsorption-Desorption Instrument (America). The surface area and corresponding pore size distribution were obtained by the Brunauer-Emmette-Teller (BET) and Barrett-Joyner-Halenda (BJH) method, respectively. In order to judge the charge transfer efficiency, photoluminescence spectra were recorded on a FLS980 fluorometer (PL, Edinburgh Instrument, England) at an excitation wavelength of 320 nm. Finally, with a Varian Cary 5000 UV-Vis spectrometer (Agilent, America), the UV-visible absorption spectrum of the samples was collected.

Evaluation of photocatalytic hydrogen evolution
To acquire the accurate data of hydrogen evolution, in each test 50 mg catalyst was first diffused into 100 ml TEOA aqueous solution (10 vol.%). Then, the prepared reaction mixture was settled in a Perfectlight LabSolar PhotocatalyticHydrogen Evolution System (Beijing, China). The system possesses a Xe lamp (300 W) with an UV cut-off filter (k C 420 nm) for light source. Afterwards, the system was sealed. Before photocatalytic reaction, the system was first evacuated to a vacuum of -0.1 MPa, and throughout the photocatalytic reaction, the cooling system based on circulating water should work continuously to keep the reaction temperature at 7°C. For the reaction, the applied light source was placed 20 cm far from the reaction vessel, and the area for effective irradiation was measured as 12.57 cm 2 . During the reaction, the generated gas was characterized by a gas chromatography (GC-7900, Xuansheng Scientific Instrument Co. Ltd, Shanghai, China) on-line with nitrogen as carrier gas. Additionally, each of the cyclic tests for photocatalytic hydrogen evolution was executed for 3 h and the gas products were taken for characterization every half an hour. After one round of tests, the reactor will be evacuated for 30 min, and then the subsequent round of tests is repeated without changing the reaction liquid. Totally 6 rounds of tests were performed to evaluate the cyclic performance.

Photoelectrochemical characterization
For each photoelectrochemical test, a working electrode was first fabricated, for which 50 lL ink prepared from 5 mg sample, 500 lL absolute ethanol together with 20 lL Nafion solution (5 wt%) was coating onto a FTO glass substrate, finally forming an electrode with a size of 1 cm 2 . The electrolyte is 0.2 M Na 2 SO 4 aqueous solution.
The measurement on transient photocurrents was carried out by a standard three-electrode system on an electrochemical workstation (CHI 660E, Chenhua Instrument, Shanghai, China). For the measurement, a piece of Pt foil was used as counter electrode, Ag/ AgCl (saturated KCl) was used as reference electrode, and the visible light (k C 420 nm) was provided by a Xe lamp (300 W) using an UV cut-off filter. Prior to the tests, the system should be degassed by high-purity N 2 gas for about 30 min.
The electrochemical impedance spectra (EIS) of the samples were recorded with the same parameters as those in measuring transient photocurrents. During the EIS tests, the frequency falls in the range of 0.01-100,000 Hz, and the applied AC amplitude is 5 mV (vs. Ag/AgCl).

Photocatalytic performance for H 2 evolution
To reveal the photocatalytic activity of the proposed g-C 3 N 4 /Ni 9 S 8 composite, all the samples were tested for hydrogen evolution under the irradiation of visible light at 7°C. Firstly, a series of samples prepared with different feed molar ratios of Ni 2? /g-C 3 N 4 were examined to determine the optimal sample. Figure 1a compares their hydrogen evolution rates (HERs). As can be seen in this figure, with the increase of feeding amount of nickel source, the HER value of the acquired samples can be improved. When the Ni 2? / g-C 3 N 4 feed molar ratio was 0.18, their HER value reached the highest, which was 355.7 lmol g -1 h -1 . With more nickel source added, the performance of the obtained photocatalysts declined again, possibly because the excessive addition of nickel source would lead to a so large amount of nickel sulfides coating on g-C 3 N 4 (see Table 1), which would hinder the absorption of light by g-C 3 N 4 , thus reducing the hydrogen evolution efficiency of the catalysts. So, the optimal Ni 2? /g-C 3 N 4 feed molar ratio for the present composite samples was determined as 0.18. In Fig. 1b, the HER values of the raw g-C 3 N 4 , the pure Ni 9 S 8 , the raw g-C 3 N 4 which loaded with Pt (1 wt%) and the optimal g-C 3 N 4 /Ni 9 S 8 composite are compared. Among them, the sample of g-C 3 N 4 with Pt Fig. 1 Photocatalytic performance on hydrogen evolution: a of the composites prepared with different feed molar ratios of Ni 2? /g-C 3 N 4 , b comparison on raw g-C 3 N 4 , pure Ni 9 S 8 , 1 wt% Pt coated g-C 3 N 4 by light deposition and the optimal g-C 3 N 4 /Ni 9 S 8 sample, and c the cycling tests on the optimal g-C 3 N 4 /Ni 9 S 8 composite.
The photocatalytic reactions were carried out with 20 mg catalyst, 100 mL aqueous solution containing 10 vol.% TEOA and a 300 W Xe lamp with UV cut-off filter (k C 420 nm) was directly tested after the deposition of Pt onto g-C 3 N 4 without any purification (see Sect. 2.4). As is seen in the figure, the sample of raw g-C 3 N 4 only has a very small HER and the pure Ni 9 S 8 has no detectable photocatalytic hydrogen evolution ability. However, the HER value of the optimal composite sample can reach 355.7 lmol g -1 h -1 . This remarkable value is 21.2 times that of the raw g-C 3 N 4 , and only 22.1% less than that of the sample of g-C 3 N 4 loaded with 1 wt% Pt. Therefore, it can be inferred that the compositing between g-C 3 N 4 and Ni 9 S 8 will present a synergistic effect, which can effectively transfer the photogenerated electrons on g-C 3 N 4 to Ni 9 S 8 through the interface charge transfer effect. And then the electrons on Ni 9 S 8 can also combine with H ? to generate hydrogen, which greatly improves the photocatalytic efficiency of g-C 3 N 4 . Additionally, in order to determine the stability of the sample, the cycling tests were carried on the optimal g-C 3 N 4 /Ni 9 S 8 composite. For the cycling tests, the total time was 18 h, which was performed in six rounds and each round was carried out for 3 h in the light reaction cell. For accurate measurement, at the end of each round of test, the hydrogen in the system was evacuated by pumping off and then a new round of test was executed. It can be seen from Fig. 1c that the HER value of the optimal sample has no obvious change after 18 h of cyclic test. Therefore, it can be concluded that the optimal composite has good cycling stability.

Compositional and structural properties
To elucidate the photocatalytic performance of the acquired samples, their composition and structure were investigated. Firstly, SEM imaging was carried out. Figure 2a shows typical SEM micrograph of the optimal g-C 3 N 4 /Ni 9 S 8 composite. As is seen, this composite sample has a similar morphology with the raw g-C 3 N 4 sample (see Fig. 2b). Therefore, it can be deduced that the two-dimensional structure of raw g-C 3 N 4 is not destroyed during the processing. In order to further examine the microstructure of the composite, TEM imaging was performed on the optimal sample. From the low-resolution image (Fig. 2c), it can be observed that the optimal sample has a stacked lamellar structure, and some black blocks are located on the lamellar structure, which might be Ni 9 S 8 nanostructures loading on g-C 3 N 4 . To further confirm the loading of Ni 9 S 8 on g-C 3 N 4 , typical high-resolution image is presented in Fig. 2d. From this picture, it can be seen that the clear lattice fringes (0.28 nm), corresponding to the (311) crystalline plane of Ni 9 S 8 , and the amorphous area (g-C 3 N 4 ) are connected closely with each other, most possibly forming heterostructures between g-C 3 N 4 and Ni 9 S 8 . In addition, to explore the distribution of nickel sulfides in the sample, EDS mapping scanning was executed in the same area for the low-resolution TEM image. And the results are displayed in Fig. 2e-h. It is easily seen that C, N, Ni as well as S atoms are distributed over the sampling area. Therefore, it can be also concluded that the Ni 9 S 8 nanostructures in the obtained composite are uniformly grown onto the g-C 3 N 4 surface, rather than a simple mixture of both components, indirectly confirming the heterostructures between g-C 3 N 4 and Ni 9 S 8 . The uniform composition and structure are one of the reasons why the proposed g-C 3 N 4 /Ni 9 S 8 composite has high photocatalytic activity.
To determine the phase composition of the specimens, XRD analysis was carried out. Figure 3 compares the recorded XRD patterns of the optimal composite, the raw g-C 3 N 4 and pure nickel sulfides samples, in which the raw g-C 3 N 4 specimen was acquired by the calcination of melamine, while the nickel sulfides sample was obtained under the similar conditions as done for the optimal composite but without the addition of g-C 3 N 4 . As is seen, the main XRD peaks of the pure nickel sulfides sample can be indexed to orthorhombic Ni 9 S 8 phase (JCPDS No. , while a small amount of hexagonal NiS (JCPDS No. 12-0041) can be also identified, indicating that it is a mixture of Ni 9 S 8 as the main body with a small amount of hexagonal NiS. The XRD peaks of the raw g-C 3 N 4 sample match perfectly with those of its predecessors, confirming its successful synthesis  13.55 in this work [13]. And in the optimal composite, only the diffraction peaks of g-C 3 N 4 and Ni 9 S 8 can be identified, which proves that it is a relatively pure composite of g-C 3 N 4 and Ni 9 S 8 . The elemental composition as well as chemical state of the optimal g-C 3 N 4 /Ni 9 S 8 composite was further detected by XPS analysis. The obtained results are shown in Fig. 4 and Table 1. XPS survey spectroscopy reveals that the g-C 3 N 4 /Ni 9 S 8 samples are composed of C, N, Ni and S, indicating that they are a composite of these elements as expected (see Fig. 4a). As for the peak of O element, it might be Fig. 2 Microstructure and mainly elemental distribution. SEM images of the optimal sample (a) and g-C 3 N 4 (b). TEM images of the optimal sample: c low-resolution, d high-resolution. And TEM-EDS mapping images of e C, f N, g Ni, and h S caused by the oxidation of g-C 3 N 4 and/or the coupling of hydroxyl groups on the samples during hydrothermal synthesis [24]. Moreover, with increasing feed molar ratio of Ni 2? /g-C 3 N 4 , the loading amount of Ni 9 S 8 on g-C 3 N 4 increased ( Table 1). The high-resolution C 1 s spectrum as exhibited in Fig. 4b presents two distinct peaks at 284.75 and 287.91 eV. Literature survey reveals that the peak at 284.75 eV could be assigned to free carbon, while the one at 287.91 eV should be indexed to the binding energy of C atoms in g-C 3 N 4 [36]. The N 1 s spectrum (see Fig. 4c) can be fitted into four peaks at 398.36, 399.03, 400.69 and 404.25 eV, respectively. Among them, the three peaks at 398.36, 399.03 and 400.69 eV could be ascribed to the sp 2 hybrid nitrogen in C=N-C group, sp 3 hybrid nitrogen in N-(C) 3 structure and nitrogen in amino moiety in g-C 3 N 4 , respectively, while the one at 404.25 eV is owing to the p-excitation of N in the g-C 3 N 4 structures [37]. And the binding energy of C 1s and N 1s in the g-C 3 N 4 /Ni 9 S 8 composite has no obvious change compared with that of the pure g-C 3 N 4 . So it can be deduced that the valence state of g-C 3 N 4 is not changed after their combination. The Ni 2p 3/2 spectrum (Fig. 4d) exhibits three main peaks at 853.15, 855.83 and 861.29 eV, in which the peaks at 853. 15 and 855.83 eV can be attributed to Ni 2? in Ni 9 S 8 , and that at 861.29 eV is assigned to Ni 2? in the hydroxyl compound of nickel [27]. However, as is mentioned above, the phase of hydroxyl compound of nickel could not be identified from the XRD patterns, possibly because its content in the samples was too small. In a word, the main reason for the formation of this binding energy is the hydroxyl hanging bond on the surface of Ni 9 S 8 during hydrothermal reaction [10]. Correspondingly, the S 2p spectrum presented in Fig. 4e displays three main peaks at 161.92, 163.08 and 168.63 eV. The peaks at 161.92 and 163.08 eV can be attributed to the S 2and S 2 2in Ni 9 S 8 lattice, while that at 168.63 eV may be due to the attachment of some sulphate produced by thiourea hydrolysis on the surface of the sample [27]. At the same time, compared with pure Ni 9 S 8 , the binding energies of Ni 2p and S 2p in the g-C 3 N 4 /Ni 9 S 8 composite are slightly shifted to high energy, indicating that the electrons in g-C 3 N 4 are transferred to Ni 9 S 8 , which also confirms the combination between g-C 3 N 4 and Ni 9 S 8 . To obtain the specific surface area (S BET ), pore size distribution and pore volume of the optimal g-C 3 N 4 / Ni 9 S 8 composite, low-temperature nitrogen adsorption tests were carried out. Figure 5a presents the nitrogen adsorption-desorption isotherms in comparison with that of pure g-C 3 N 4 . Their corresponding pore size distribution curves were obtained by the multipoint BET and BJH methods, and the results are shown in Fig. 5b. The calculated S BET , pore volume and pore size of the raw g-C 3 N 4 and g-C 3 N 4 / Ni 9 S 8 composite are listed in Table 2. From Table 2, it can be seen that the specific surface area of the optimal g-C 3 N 4 /Ni 9 S 8 composite became smaller after the compositing of g-C 3 N 4 with Ni 9 S 8 . And from Fig. 5b, it can be observed that after compositing, the pore volume of the optimal g-C 3 N 4 /Ni 9 S 8 composite at the pore size of about 40 nm significantly decreased, but that at the pore size of 4 nm changed very little. These results indicate that, the growth of Ni 9 S 8 happens in the large pores of g-C 3 N 4 , and the deposition of Ni 9 S 8 will have a certain blockage effect on the pores of g-C 3 N 4 . Therefore, it can be concluded that the improved catalytic performance of the optimal g-C 3 N 4 /Ni 9 S 8 composite is not attributed to the increased number of active sites originating from the enhancement in specific surface area of the  Fig. 4 XPS results of the optimal g-C 3 N 4 /Ni 9 S 8 composite. a Survey spectrum. High resolution spectra of b C 1s, c N 1s, d Ni 2p and e S 2p. For comparison, the corresponding data of pure g-C 3 N 4 or Ni 9 S 8 samples are also displayed in (b-e) composite, but to the prompt transfer of electrons instead caused by the heterostructured combination between g-C 3 N 4 and Ni 9 S 8 . Figure 6a compares the UV-visible absorption spectrum of the optimal g-C 3 N 4 /Ni 9 S 8 composite with those of the raw g-C 3 N 4 and Ni 9 S 8 nanostructure. It can be seen that the starting point of the absorption edge for the raw g-C 3 N 4 is about 440 nm, which is well corresponding to the value reported in literature [38]. However, after g-C 3 N 4 was combined with Ni 9 S 8 , the absorption ability for visible light of the obtained composite was greatly enhanced, while the absorption capacity for ultraviolet and near ultraviolet spectrum below 450 nm was also significantly increased. This result reveals that after the Ni 9 S 8 co-catalyst was loaded onto the g-C 3 N 4 nanosheets, the optical absorption region of the composite sample could be effectively broadened. The enhanced ability in optical absorption could be owing to the existence of low bandgap black Ni 9 S 8 in the g-C 3 N 4 /Ni 9 S 8 composite. Such Ni 9 S 8 nanostructures have strong optical absorption to the light with a wavelength from 300 to 800 nm. Furthermore, the bandgaps of the optimal g-C 3 N 4 /Ni 9 S 8 composite and the raw g-C 3 N 4 sample were estimated from their corresponding plots on (ahm) 2 versus E g (see Fig. 6b). As is seen, the E g values of the optimal g-C 3 N 4 /Ni 9 S 8 composite and the raw g-C 3 N 4 are 2.85 and 2.88 eV, respectively. In order to further explore the band gap structure of the present photocatalyst materials, the density of states (DOS) of the valence band of the optimal g-C 3 N 4 /Ni 9 S 8 composite and raw g-C 3 N 4 were measured by valence band XPS (Fig. 6c). Both of them displayed typical valence band DOS characteristics of g-C 3 N 4 with the edge of the maximum energy at about 0.07 and 0.34 eV, respectively. According to the band gap obtained from the UV spectrum, it can be calculated that the conduction band of the optimal g-C 3 N 4 /Ni 9 S 8 composite and the raw g-C 3 N 4 would occur at about -2.87 and -2.51 eV, respectively. Therefore, it can be inferred that thermodynamically, the g-C 3 N 4 /Ni 9 S 8 composite has a better hydrogen evolution ability due to its more negative conduction band. The interfacial charge transfer efficiency of the optimal g-C 3 N 4 / Ni 9 S 8 and raw g-C 3 N 4 can be evaluated from the stable photoluminescence (see Fig. 6d). The raw g-C 3 N 4 has a strong photoluminescence peak at 470 nm. However, compared with the raw g-C 3 N 4 , the photoluminescence peak intensity of the optimal g-C 3 N 4 /Ni 9 S 8 composite becomes significantly weaker. These results indicate that the present g-C 3 N 4 /Ni 9 S 8 has a higher charge transfer efficiency, which can be ascribed to the fast transport of  photoelectrons from g-C 3 N 4 to Ni 9 S 8 , thus finally improving the separation efficiency of photogenerated carriers. In addition, the photoluminescence peak of g-C 3 N 4 /Ni 9 S 8 shifts from 470 to 462 nm in comparison with that of the raw g-C 3 N 4 , which also confirms the loading of Ni 9 S 8 on g-C 3 N 4 in the composite samples.
In summary, the above-mentioned characterizations reveal that a g-C 3 N 4 /Ni 9 S 8 composite with uniform composition and structure has been successfully synthesized in this work.
Finally, the formation mechanism of the present g-C 3 N 4 /Ni 9 S 8 composite as follows was proposed based on the above experimental results (see Fig. 7). In the designed first step, the raw g-C 3 N 4 powder is dispersed in deionized water by ultrasonic stirring to form uniformly dispersed g-C 3 N 4 nanosheets. After nickel acetate and thiourea joined into the reaction system, because of the negative charges on the g-C 3 N 4 nanosheets, Ni 2? could be adsorption on the surface of g-C 3 N 4 , while thiourea could form a complex with Ni 2? to slow down the release of Ni 2? . In the second step, during the hydrothermal reaction, sodium hypophosphite was first decomposed into PH 3 , which could be ionized into Hwith strong reducibility. Then the Hcan reduce Ni 2? attached to g-C 3 N 4 surface into metal Ni. In the third step, the stable thiourea was thermally decomposed to produce a large amount of S 2-, which would react with the metal Ni on the g-C 3 N 4 surface to obtain the proposed composite in the state full of unsaturated Fig. 6 Optical properties. a UV-Vis absorption spectra of the optimal g-C 3 N 4 /Ni 9 S 8 composite, raw g-C 3 N 4 and Ni 9 S 8 , and b the corresponding (ahm) 2 versus E g plots. c Valence-band XPS spectra and d stable photoluminescence spectra of the optimal g-C 3 N 4 /Ni 9 S 8 composite and raw g-C 3 N 4 sulfur. Finally, the g-C 3 N 4 /Ni 9 S 8 composite in the state of unsaturated sulfur was obtained.

Photocatalytic mechanism
To clarify the photocatalytic mechanism for H 2 evolution of the as-acquired g-C 3 N 4 /Ni 9 S 8 composite, its photoelectrochemical features were further explored. For comparison, the Ni 9 S 8 nanostructure and raw g-C 3 N 4 were also investigated. Firstly, the transient photocurrents of the raw g-C 3 N 4 , pure Ni 9 S 8 and the optimal g-C 3 N 4 /Ni 9 S 8 composite are compared in Fig. 8a. It can be seen from this figure that the raw g-C 3 N 4 sample has only 0.3 lA cm -2 of transient photocurrent, which is far less than that of the optimal composite. This is because the photogenerated electrons and holes are easily recombined with each other in the raw g-C 3 N 4 sample due to its low-conductivity. In addition, on the pure Ni 9 S 8 nanostructure the detected photocurrent is not distinct, indicating that although it has strong absorption to the light at the wavelength of 300-800 nm, it cannot contribute to the transient photocurrent of the present composite independently. These facts prove that in the present g-C 3 N 4 /Ni 9 S 8 composite, the Ni 9 S 8 nanostructure is not a photocatalyst but a co-catalyst for g-C 3 N 4 . The Ni 9 S 8 co-catalyst can boost the segregation of photogenerated electrons and holes, thus increasing the transfer efficiency of photogenerated carriers.
Furthermore, the recorded EIS Nyquist curves (see Fig. 8b) reveal that among the three investigated samples, the raw g-C 3 N 4 sample possesses the highest intrinsic impedance, indicating that there is a Fig. 7 Synthesis mechanism of the present g-C 3 N 4 /Ni 9 S 8 composite Fig. 8 Photoelectrochemical data of the optimal g-C 3 N 4 /Ni 9 S 8 composite, raw g-C 3 N 4 and pure Ni 9 S 8 nanostructure: a transient photocurrent responses, and b EIS Nyquist plots higher charge transfer rate in the composite g-C 3 N 4 / Ni 9 S 8 samples after the loading of Ni 9 S 8 onto g-C 3 N 4 . These phenomena may be due to the homogeneous combination of the relatively low-conductivity g-C 3 N 4 nanosheets with the high-conductivity Ni 9 S 8 nanostructures.
According to the above-mentioned experimental results, the possible photocatalytic mechanism over the present g-C 3 N 4 /Ni 9 S 8 composite on hydrogen evolution was proposed (see Fig. 9). Based on the calculated positions of conduction band and valence band of g-C 3 N 4 /Ni 9 S 8 composite, under the irradiation of visible light, the photogenerated electron (e -) will leap into the conduction band of g-C 3 N 4 , which will leave hole (H ? ) in its valence band. A part of photogenerated electrons could move to the g-C 3 N 4 surface and directly reduce H ? ions in solution to produce H 2 . More importantly, due to the close contact in the heterostructure between the main body g-C 3 N 4 and the high-conductivity Ni 9 S 8 nanostructures in the present g-C 3 N 4 /Ni 9 S 8 composite, a builtin electric field is formed after the combination between them. As a result, the photogenerated electrons can easily emigrate to the Ni 9 S 8 nanostructures, thus effectively suppressing the recombination between the photogenerated electrons and holes. Because of the low impedance and high electrocatalytic activity of Ni 9 S 8 , the photogenerated electrons transferred to Ni 9 S 8 can rapidly move to the surface of Ni 9 S 8 , which can also serve as the active site for electron reduction reaction, effectively reducing H ? to H 2 in the solution. In addition, triethanolamine can absorb the holes transferred from the valence band to the surface of g-C 3 N 4 , thus completing a whole set of redox reactions.

Conclusions
An interesting g-C 3 N 4 /Ni 9 S 8 composite was prepared using a novel hydrothermal method, in which nickel acetate and thiourea were used as the Ni and S sources, respectively, while NaH 2 PO 2 was used as the reducing agent, finally synthesizing the g-C 3 N 4 / Ni 9 S 8 composite through a simple one-step process. This method has the potential for large-scale production of g-C 3 N 4 /Ni 9 S 8 composite with the advantages of simple, safety, environmental friendliness and easy controllability. The optimal g-C 3 N 4 /Ni 9 S 8 nanocomposite has high photocatalytic activity under visible light irradiation. The hydrogen evolution rate of the optimal g-C 3 N 4 /Ni 9 S 8 composite prepared with the Ni 2? /g-C 3 N 4 feed molar ratio of 0.18 could reach 355.7 lmol g -1 h -1 at 7°C, which is 21.2 times higher than that of the pure g-C 3 N 4 sample, and only 22.1% less than that of the g-C 3 N 4 with Pt (1 wt%). The greatly enhanced performance on photocatalytic hydrogen evolution over the as-acquired g-C 3 N 4 / Ni 9 S 8 composite could be attributed to the prompt transfer of photogenerated electrons from the lowconductivity g-C 3 N 4 surface to the high-conductivity Ni 9 S 8 cocatalyst, which effectively promotes the segregation of photogenerated electrons and holes, and sponsors the electrocatalytic activity of Ni 9 S 8 via effectively reducing H ? -H 2 as well. The present study will provide a new approach for the construction of photocatalysts by co-catalysis for photocatalytic hydrogen evolution.