Abstract

Quality of interfaces is a key factor determining photoexcited charge transfer efficiency, and in turn photocatalytic performance of heterostructure photocatalysts. In this paper, we demonstrated CdS-MoS2/RGO-E (RGO-E: reduced graphene oxide modified by ethylenediamine) nanohybrid synthesized by using a facile one-pot solvethermal method in ethylenediamine, with CdS nanoparticles and MoS2 nanosheets intimately growing on the surface of RGO. This unique high quality heterostructure facilitates charge separation and transportation, and thus effectively suppressing charge recombination. As a result, the CdS-MoS2/RGO-E exhibits a state-of-the-art H2 evolution rate of 36.7 mmol g
1 h
1 and an apparent quantum yield of 30.5% at 420 nm, which is the advanced performance among all the same-type photocatalysts (see Table S1), and far exceeding that of bare CdS by higher than 104 times. This synthesis strategy gives an inspiration for the synthesis of other compound catalysts, and higher performance photocatalyst may be obtained.

© 2018 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.

Introduction

Carbon-neutral energy is the pursuit for today's society owing to severe environmental pollution caused by fossil energy excessive consumption [1e5]. Hydrogen energy is a perfect candidate with highest energy of 142 MJ kg
1 and only clear H2O discharge. Photocatalytic hydrogen evolution from water splitting is an environmental and simple technique to produce hydrogen in mass degree, and has potential to be applied in future. Since the early 1970s, when TiO2 was firstly reported having good photocatalytic activity in water splitting, this research project has attracted widespread interest and achieved great progress [6e8]. But most reported photocatalysts can only harvest UV light which makes up only about 3% of total solar energy. Therefore, much attention should be paid on the photocatalysts driven by visible light which accounts for about 45% of total solar energy.

CdS is widely employed as a visible-light-driven photocatalyst owing to its suitable conduction band edge for H2 generation, and relative narrow band gap. However, ultrafast recombination of electron-hole pairs resists its photocatalytic performance further enhancement. Combing cocatalysts with CdS to construct heterostructures is an effective approach to enhance the photocatalytic activity [9,10]. Noble metal Pt working as cocatalyst exhibits high catalytic performance because it requires low overpotential for Hþ reduction, but its low reserve and cost of up-scaling severely hindered its application. Recently, a great number of earth abundant materials have been developed as alternatives, such as MoS2 [11e16], WS2 [17], Ni2P [18,19], CoP [20], carbon dot [21], graphdiyne [22] and black phosphorus [23,24]. Among them, MoS2 shows excellent performance and has been intensively studied [25e30]. Both computational and experimental results revealed that the edges of MoS2 layers were the active sites for hydrogen evolution, but their basal planes were catalytic inert [31,32]. Therefore, increasing the amount of edge sites and accelerating electron transfer from basal to edge planes will be beneficial for the enhancement of photocatalytic activity. In previous study, we developed solvothermal approach to synthesize MoS2/CdS heterostructure with amorphous-like MoS2 anchored on CdS nanorods, which exposed a great many activity sites for H2 evolution, but the weak conductor of MoS2 curbed electron transfer and in turn affected H2 generation activity [33,34]. It is well known that graphene is an ideal conductor and has been widely applied in photo-catalysis [35e37], electro-catalysis [38e41] and solar cell [42e44]. MoS2-RGO has been proved to be good dual-cocatalysts [12,45e47], but the high quality interface between MoS2 and RGO is still the pursuit. Moreover, the synthesis of CdS/MoS2- RGO ternary catalysts usually needs multistep, which is time consumption and energy intensive.

Usually, graphene oxide (GO) was used as precursor for the synthesis of graphene based photocatalysts [27,35,48e50]. But its electronegativity resists the absorption of precursor MoO4 2
used in this manuscript. It was reported that the graphene modified with ethylenediamine rendered it easily absorbing electronegativity precursors and transition metals owing to the electrostatic attraction and complex effect [51e53]. Inspired by this, herein, we adopted one-pot solvothermal approach in ethylenediamine solution to synthesize CdSMoS2/RGO-E ternary catalyst with CdS nanoparticles and amorphous-like MoS2 nanosheets in-situ growth on the surface of RGO, which guarantees intimately interfacial contact between composites, and offers a great many active sites for H2 generation. This well-defined nano-heterostructure makes for photo-excited charge separation and transportation, and thus significantly retarding charge recombination. As a result, the optimized CdS-MoS2/RGO-E exhibits high performance with H2 evolution rate of 36.7 mmol g
1 h
1 , which is 104 times higher than that of pure CdS, indicating it has potential for applications.

Experimental

Chemicals

Cadmium acetate dihydrate (Cd(CH3COO)2$2H2O, 98%), Sodium molybdate dihydrate (Na2MoO4$2H2O, 99%), Thiourea (CN2H4S, 99%), ethylenediamine and graphite (>99.8%) were purchased from Alfa Aesar chemical co., USA. All agents were used directly without further purification. GO was synthesized by a modified Hummers' method [54,55]. Deionized water with a resistivity of 18.2 MU cm, produced by using a Milli-Q apparatus (Millipore), was used in all the experiments.

Synthesis

Synthesis of CdS-MoS2/RGO-E

The CdS-MoS2/RGO-E composites were synthesized through a facile one-pot solvothermal method. In a typical preparation, 2 mL GO solution (1.0 mg/mL), 25 mL ethylenediamine, a varying amount of Na2MoO4$2H2O solution (0.08 M), 0.2 g Cd(CH3COO)2$2H2O and 0.3 g CN2H4S were mixed together with strong stirring, and followed by sonication for 10 min. Then the mixture was added into 50 mL Teflon-lined stainless steel autoclave, and held at 210 C for 24 h. After naturally cooled to room temperature, yellow-green powders were collected by centrifuging, washing with deionized water and then drying at 80 C for 12 h.

Synthesis of control samples

As a control, CdS, MoS2, CdS-MoS2-E, CdS/RGO-E, and MoS2/ RGO-E were synthesized in parallel following the same procedure except for no addition of (GO þ Na2MoO4$2H2O), (GO þ Cd(CH3COO)2$2H2O), GO, Na2MoO4$2H2O, or Cd(CH3COO)2$2H2O, respectively. CdS-MoS2/RGO-W was also fabricated at the same reaction conditions except for addition of deionized water instead of ethylenediamine. Note that all the loadings in percentage in this manuscript represent the theoretical mass ratios of MoS2 or graphene to CdS given that the reactants were completely converted into the products. The actual loading amounts of MoS2 to CdS in all samples were tested using an inductively coupled plasma-atomic emission spectrometry (ICP-AES, ICPE-9000 Shimadzu), and the actual content of graphene to CdS was calculated by using subtraction method. The results were listed in Table 1.

Evaluation of photocatalytic activities

Photocatalytic H2 evolution was performed in a Pyrex glass cell which had a flat, round upside-window with an irradiation area of 38 cm2 for external light incidence. A 300 W Xenon arc lamp with a 420 nm cut-off filter (CEL-HXF 300, Beijing China Education Au-light Co., Ltd) was used to simulate the visible light source.

 The illumination intensity was adjusted to 100 mW cm
2 . The H2-solar system (Beijing China Education Au-light Co., Ltd) with a gas chromatogram (GC), equipped with a thermal conductivity detector (TCD), TDX-01 column and Ar carrier gas, was used to collect and on-line detect evolved H2. 0.02 g of photocatalyst was suspended in glass cell with 72 mL of deionized water containing 8 mL lactic acid as hole scavenger. The cell was kept at 5 C by using a circulating water system. Before irradiation, the reaction system was pumped to vacuum. The H2 evolution rate was determined by GC. The apparent quantum yield (Ø) was estimated by the following equation: 

where Ø is the apparent quantum yield, ne- is the number of reacted electrons, np is the number of incident photos, nH2 is the number of evolved H2 molecules, q is the total energy of incident photos (J), h is the Planck constant (J s
1 ), n is the frequency of light (Hz), I is the illumination intensity (W m
2 ) determined with a ray virtual radiation actinometer, t is the irradiation time (s), S is the irradiation area (m2 ).

Characterization

Transmission electron microscopy (TEM) images were obtained on a JEM 2100F (JEOL, Japan) operated at 200 kV. X-ray powder diffraction (XRD) was carried out with a Rigaku D/ max-7000 using filtered Cu Ka irradiation. Raman spectrum was recorded on a Thermo Scientific DXR confocal Raman Microscope equipped with a 532-nm laser. X-ray photoelectron spectroscopy (XPS) data were recorded with an ESCALab 220 i-XL electron spectrometer from VG Scientific using 300 W Al Ka radiation, in which the binding energies were referenced to the C1s line at 284.8 eV from adventitious carbon. The UVevisible absorption spectra were recorded with a UVevisible spectrophotometer (UV-2550, Shimadzu, Japan). Fourier transform infrared spectra (FTIR) were obtained on a FTIR spectrometer (Bruker Tensor 27). The ethylenediamine in centrifuged solution was tested by using GC with a FID detector equipped with an Rtx-1701 Sil capillary column (Shimadzu GC-2014C). Transient photocurrent was measured on a CHI 760 E electrochemical system (shanghai, china) using AgAgCl as reference and Pt as counter electrodes. The work electrode was prepared by dispensing sample suspension in ethanol onto ITO/glass of fixed area (1.96  10
5 m2 ). The electrolyte is lactic acid solution (1.33 M) which was filled in a quartz cell with a side window for external light incidence. Light on and off was controlled by a baffle installed on a stainless steel black box. Brunauer-Emmett-Teller (BET) measurements were performed on a Micro-meritics's Tristar 3000. Inductively coupled plasma atomic emission spectrometry (ICP-AES, ICPE-9000 Shimadzu) was used to measure the Mo and Cd contents.

Results and discussion

The synthesis process for CdS-MoS2/RGO-E is schematically illustrated in Fig. 1a. GO, CN2H4S, Na2MoO4 and Cd(CH3COO)2 were added into ethylenediamine solution in one batch, and reacted at 210 C for 24 h to get the end-products (Detailed experiments see experimental section). The morphology of GO and as prepared products were detected by transmission electron microscope (TEM). As shown in Fig. 1b, the raw material GO is cleanly wrinkled nanosheets, providing large surface for catalysts deposition. After reaction, TEM images (Fig. 1c) of the end-products show that nanoparticles with radius ranging from ca. 15e185 nm firmly grew in-situ on graphene nanosheets. Detailed statistics (see Fig. S1) display that nanoparticles with radius less than 60 nm account for about 71.5%. The further enlarged images of nanoparticals as shown in Fig. 1d indicate that the lattice fringes spacing is ca. 0.36 nm, which is well corresponding to the (100) planes of hexagonal CdS [33,34]. Besides, the irregular nanosheets (delineated by blue dashed line in Fig. 1d) were also observed growth on graphene or CdS nonoparticles. High-resolution TEM (HRTEM) images exhibit short-range continuous lattice fringes on these nanosheets. The lattice spacing of 0.27 and 0.61 nm can be indexed as d-spacing of crystallographic (101) and (003) planes of rhombohedral MoS2 [33,34,56]. Generally, the lattice fringes of MoS2 are some amorphous features, indicating that they are partially crystalline MoS2 nanosheets. Those amorphous-like MoS2 can efficiently enhance H2 evolution performance owing to the exposed edges providing a great many active sites for HER, which has been verified in our previous study [33]. Furthermore, the energy dispersive X-ray spectroscopy (EDS) mapping analysis (Fig. 1e) of the obtained products shows elemental S and Cd signal are homogeneously distributed on the surface of nanoparticles, manifesting that the nanoparticles on graphene are of CdS. However, the signal of C distinctively presents two sections of strong and weak, which can be attributed to graphene, and carbon membrane on copper grid for supporting sample, respectively. The relatively weak but homogeneous signal of Mo matches well with the sheetlike structure MoS2 grown on CdS nanoparticles and graphene.

The further chemical state information of those elements was detected by X-ray photoelectron spectroscopy (XPS). As shown in Fig. 2a, the XPS survey spectrum reveals the existence of C, S, Cd and Mo in the ternary composite. The highresolution XPS spectrum in Fig. 2b displays two peaks at 412.0 and 405.1 eV, which are in good consistency with the characteristic binding energies of Cd2þ 3d3/2 and Cd2þ 3d5/2 in CdS, respectively [57]. XPS spectrum in Fig. 2c shows a typical strong doublet at 231.4 and 228.2 eV, which match well with the binding energies of Mo 3d3/2 and Mo 3d5/2, respectively, suggesting the dominant existence of Mo4þ species in product [13]. The peak for S 2p (Fig. 2d) can be well deconvoluted into two separate peaks at around 162.4 and 161.1 eV, which are the typical XPS signals of S 2p1/2 and S 2p3/2 in form of S2
[33]. The XPS signal (Fig. 2e) of C1s can be well fitted into four separate peaks centered at 284.6, 285.7, 286.9 and 288.7 eV, corresponding to (C-C and C-H), (C-OH and C-N), (C-O-C) and (O-C]O and N-C]O), respectively. In comparison with C 1s XPS signal (Fig. S2) of GO, there exists significantly decrease for the signal of oxygen functionalities, but an additional peak at 285.7 (C-N), suggesting that the GO thermally treated in ethylenediamine solution is successfully reduced into RGO and modified by ethylenediamine. Together with TEM

observation, these results corroborate that the as-obtained products are CdS-MoS2/RGO composites.

The phase structure and crystallinity of pure CdS as well as CdS-MoS2/RGO-E nanohybrids with graphene mass ratio fixed at 2 wt% vs. CdS, and MoS2 mass ratio from 4 to 30 wt% vs. (CdS-RGO), were investigated by X-ray powder diffraction (XRD) (Fig. 2f). All the XRD peaks for the pure CdS correspond to the hexagonal CdS (JCPDS Card no. 65e3414). However, no graphene and MoS2 characteristic patterns were detected for the nanohybrid samples compared with the pristine CdS, although they can be clearly observed in TEM images as mentioned above, which can be reasonably attributed to the low loading for the graphene [35], and the low crystalline of MoS2 in keeping well with the TEM characterization results [33,34].

To uncover the role of ethylenediamine for the synthesis of CdS-MoS2/RGO-E, a control experiment was carried out, which kept equal reaction conditions except for adding deionized water instead of ethylenediamine, and the obtained catalyst is designed as CdS-MoS2/RGO-W. Compared with CdS-MoS2/ RGO-E, TEM characterization (Fig. S3) of CdS-MoS2/RGO-W reveals that similarly dimensional CdS nanoparticles grew on graphene. But it is distinctly different for the MoS2 naonosheets which presents flowerlike aggregation. Furthermore, equal mixture of GO, Na2MoO4 and CH4N2S were added into ethylenediamine and deionized water to synthesize MoS2/ RGO composites marked as MoS2/RGO-E and MoS2/RGO-W, respectively. As shown in Fig. 3a, small MoS2 nanosheets with radius ca. 10 nm tightly and uniformly grew on the surface of RGO for MoS2/RGO-E. However, the MoS2 in MoS2/RGO-W (Fig. 3b) exhibits large layered structure random distributing on RGO. Besides, the different image contrast for the MoS2 and RGO in Fig. 3b indicates that most parts of MoS2 stretched outside of RGO, demonstrating the weak contact between MoS2 and RGO, which will result in its low cocatalytic activity. Those observations further indicate ethylenediamine can significantly affect the amorphous of MoS2 and the contact state between MoS2 and RGO. This influence may stem from the modification of RGO by ethylenediamine. To prove this hypothesis, RGO-E and RGO-W were synthesized through solvothermal treatment of pure GO in ethylenediamine and water, respectively. It should be mentioned here, the obtained RGO-E was thoroughly washed to remove the adsorbed ethylenediamine, and the GC analysis of the last centrifuged solution shows that no characteristic peaks of ethylenediamine appeared. The Fourier-transform infrared (FT-IR) spectra of GO, RGO-E are shown in Fig. 4a. In comparison, after solvothermal treatment, there is a dramatic decrease intensity for the peaks of oxygen-containing functional groups at 1727 (COOH stretching vibration peak), 3425, 1399 (eOH deformation vibration peaks), 1224 (epoxy) and 1060 cm
1 (alkoxy) in RGO-E in compared with that of GO, but meanwhile, new peaks at 1564 and 1260 cm
1 appeared in RGO-E, which can be ascribed to the strong in-plane C-N scissoring absorptions and C-O stretching vibrations [58]. Raman spectra (Fig. 4b) of GO and RGO-E show that two peaks appeared at 1343 and 1582 cm
1 corresponding to the D- and G- band of graphene, where the D band can be attributed to the edges, defects, and disordered carbon, while the G band is assigned to the vibration of ordered sp2 C atoms [59e61]. Here, the ID/IG (ratio of peaks intensity) is used to assess the reduction degree of GO. The higher of this value represent the higher reduction degree of GO. Obviously, the ID/IG of RGO-E (1.13) is higher than that of GO (0.85). The above analysis of FT-IR and Raman spectra further conform the successful reduction of GO and modification of RGO [51]. In addition, the efficient modification of RGO was further verified by the dispersity of RGO-E and RGOW in water, as shown in Fig. 3c. The RGO-E suspension is stable but the RGO-W can quickly precipitate out of solution within about 5 min. The good stability for the RGO-E manifests its good dispersion and hydrophilia owing to the modification of eNH2 which can effectively retard aggregation and provide anchored sites for the precursors. Therefore, it can be reasonably deduced that the eNH2 on RGO easily absorb MoO4 2
and provide confined environment for the growth of MoS2, thus rendering small MoS2 nanosheets intimate growth on RGO.

The photocatalytic performances of synthesized samples were assessed under visible light irradiation (l  420 nm). Lactic acid, as a green and efficient sacrificial agent, was employed in photocatalytic HER. Through the oxided reaction of lactic acid to generate pyruvic acid [29,62], the photogenerated holes were successfully captured and consumed, and thus the photocorrosion and charge recombination were

suppressed in some degree. The effect of mass percentage of RGO and MoS2 on photocatalytic activity was first investigated, every sample was repeatedly synthesized and tested for five times. The screening experiment results are shown in Fig. 5a and b, respectively. It demonstrates that the CdSMoS2/RGO-E with RGO and MoS2 mass ratio of 2 and 4 wt% exhibits the best photocatalytic performance with the average H2 evolution rate of 36.7 mmol g
1 h
1 , corresponding to the apparent quantum efficiency of 30.5% at 420 nm, which was calculated using the formula listed in experiment section. Additionally, the quantum efficiency of optimized CdS-MoS2/RGO-E dependence on light wavelength was tested. As shown in Fig. S4, with the increment of monochromatic light wavelength, the quantum efficiency reduced. This tendency is similar with that of light absorption of CdSMoS2/RGO-E with light wavelength less than 520 nm as mentioned later in this manuscript, indicating the photocatalytic activity was significantly affected by light wavelength.

This optimized catalyst exhibits highest performance among the same type catalysts (see Table S1). Considering its good performance, all the following experiments were carried out by using this sample, which was marked by CdS-MoS2/ RGO-E for brevity, hereafter. It should be mentioned that the H2 evolution rate in this manuscript represents the normalized value. The real gram-scale reaction results have been listed in Table S2. It shows that the H2 evolution rate was not proportional enhanced in compared with that of the milligram-scale. This can be reasonably attributed to the agglomeration of nano-catalyst and the light shielding effect [29]. Maybe proportionally enlarged reactor will achieve proportional enhancement of H2 evolution rate. Fig. 5c-d shows

the rate of H2 evolution on CdS-MoS2/RGO-E, CdS-MoS2/RGOW, CdS/MoS2-E, CdS/RGO-E, CdS and MoS2/RGO-E, as well as the long-term stability test for the CdS-MoS2/RGO-E. All the synthesized processes of those catalysts see details in experimental section. The real composition of all the prepared samples was analyzed and listed in Table 1, which is matched well with the theoretical values. No H2 signal was detected for the MoS2/RGO-E, suggesting that it is not active for photocatalytic water splitting. Pristine CdS shows low photocatalytic activity with H2 evolution rate of 0.35 mmol g
1 h
1 owing to the ultrafast charge recombination. But the optimized CdS-MoS2/RGO-E can significantly increase the photocatalytic activity with H2 evolution of 36.7 mmol g
1 h
1 which is 104 times higher than that of bare CdS, indicating the good cocatalytic activity of MoS2/RGO-E. The CdS/MoS2-E and CdS/ RGO-E show inferior activity of 18.5 and 1.7 mmol g
1 h
1 , respectively, in compared with that of CdS-MoS2/RGO-E. Besides, the sum H2 evolution rate of CdS/MoS2-E and CdS/RGOE is much lower than that of CdS-MoS2/RGO-E (20.2 vs. 36.7 mmol g
1 h
1 ), manifesting that RGO and MoS2 have synergistic effect for boosting photocatalytic H2 generation. But this synergistic effect is significantly affected by the contact station between RGO and MoS2, which is deduced from the activity comparison of CdS-MoS2/RGO-E and CdS-MoS2/ RGO-W, the H2 evolution rate of the former is 1.5 times higher than that of the later (36.7 vs. 24.3 mmol g
1 h
1 ). The BET (see Fig. S5.) specific surface areas of CdS-MoS2/RGO-E (66.7 m2 /g), and CdS-MoS2/RGO-W (66.3 m2 /g) are similar, which indicates the effect of specific surface for the photocatalytic activity can be ignored. The CdS-MoS2/RGO-E heterostructure with welldefined interfaces between MoS2 and RGO as mentioned from the HRTEM characterization results, will accelerate photogenerated charge separation and transportation. However, the MoS2 and RGO in CdS-MoS2/RGO-W have a weak contact, which deteriorates charge transportation and separation, and thus results in the lower H2 generation performance in compared with that of CdS-MoS2/RGO-E. The outstanding catalytic activity of CdS-MoS2/RGO-E is further verified by the video in supporting information. It can be observed that no bubbles generate without light irradiation, but bubbles rapidly generate once the xenon lamp is turned on. The H2 generation dependence on time for four cycling runs and 24 h of tests shows that no obvious activity deterioration for CdS-MoS2/RGO-E, indicating its good stability. In addition, the XRD and XPS characterizations (see Figs. S6 and 7) indicate the samples after 24 h reaction still keep its original phase and surface chemical state, further suggesting its good stability.

Additionally, the photocatalytic performance dependence on sacrificial donor concentration was studied over CdS-MoS2/ RGO-E. The results (see Fig. S8) show that the H2 evolution rate enhances along with the concentration increment of lactic acid before volume ratio up to ca. 25%. However further increment concentration of lactic acid will result in the deterioration, which can be ascribed to the high formation of intermediates at high concentration of lactic acid, being similar to the previous report [29].