Bi-axial grown amorphous MoSx bridged with oxygen on r-GO as a superior stable and efficient nonprecious catalyst for hydrogen evolution

Amorphous molybdenum sulfide (MoSx) is covalently anchored to reduced graphene oxide (r-GO) via a simple one-pot reaction, thereby inducing the reduction of GO and simultaneous doping of heteroatoms on the GO. The oxygen atoms form a bridged between MoSx and GO and play a crucial role in the fine dispersion of the MoSx particles, control of planar MoSx growth, and increase of exposed active sulfur sites. This bridging leads to highly efficient (−157 mV overpotential and 41 mV/decade Tafel slope) and stable (95% versus initial activity after 1000 cycles) electrocatalyst for hydrogen evolution.

structure of the edges to improve the activity. However, unsupported nanoscale MoS 2 with a large number of edge sites is thermodynamically unstable, leading to aggregation or transformation of the nanoparticles 10 . In addition, S-terminated edges are easily oxidized in acidic media 11 . These intrinsic properties induce the deactivation and instability of the materials when used for HER. There are two strategies to overcome these challenges: (1) controlling the morphology and (2) designing hybrid structure. Specific morphology control has been achieved using hard or soft templates such as MoO 3 /MoS 2 (core/shell) nanowire, highly ordered double-gyroid MoS 2 , vertically aligned MoS 2 , and MoS 2 flowers [12][13][14][15] . Although these methods could prevent the degradation of MoS 2 electrochemical activity, the synthetic processes are unsuitable for the industrial scale because of their complexity and expense. Meanwhile, the hybrid structure consists of carbon supported MoS 2 , which has exhibited a strong interaction between the TMD and the carbon support, minimizing the thermodynamically unstable properties and improving the morphological and electrochemical stability. However, much remains to be studied regarding the origins of the interaction and the properties of the carbon supported amorphous molybdenum sulfide. Herein, we report a one-pot synthetic strategy to produce the highly-stable and efficient MoS x /r-GO catalyst via oxygen bridging between amorphous MoS x and r-GO. These features are induced by the functional coupling of oxygen bridges between molybdenum sulfide and graphene oxide as shown in Fig. 1a.

Results
The molybdenum sulfide catalysts were easily synthesized by the wet-chemical reaction of (NH 4 ) 2 MoS 4 and HCl in an aqueous dispersion of graphene oxide (GO) at room-temperature. The precursor was reduced to molybdenum sulfide (MoS x , sulfur content (x) changed from 1 to 3) on the graphene supports. To investigate the effects of the amount of deposited MoS x particles on electrochemical hydrogen production, we synthesized the MoS x /r-GO catalysts with the amount of MoS x precursor varying from 0.1 to 0.7 g. (Hereafter, a catalyst prepared using y g of MoS x precursor and a fixed amount of GO is denoted as y MoS x /r-GO sample.) For comparison, unsupported MoS 3 particles were also prepared in the absence of GO using the same process.
High-resolution transmission electron microscopy (HR-TEM) images show that both the size and the amount of MoS x particles on thin graphene flake depend strongly on the weight of precursor used as shown in Fig. 1b-e. For 0.5 MoS x /r-GO, the particles are uniformly deposited on the GO surface with full coverage, whereas catalysts synthesized with either less or more than 0.5 MoS x /r-GO exhibited insufficient or aggregated particle features, respectively. However, the particle size of MoS x /r-GO is relatively smaller than for unsupported MoS 3 (see Supplementary Figure S1). Elemental mapping was conducted using energy dispersive spectroscopy (EDS) to confirm the origin of the particles deposited on the r-GO sheets. For all MoS x /r-GO composite samples, the positions of Mo atoms are highly correlated with the positions of S atoms, and MoS x particles are considered to have been successfully synthesized on the r-GO sheets (see Supplementary Figure S2).
We further investigated the morphological features of as-synthesized MoS x /r-GO composites using atomic force microscopy (AFM) as shown in Fig. 2. Each GO sheet has a thickness of ~1.1 nm consistent with double-or triple-layered GO. Similar to the TEM results for the MoS x /r-GO composites, the width and height of MoS x on the r-GO composites depend on the amount of MoS x precursor. The average width of the MoS x particles on r-GO were increased from 50.2 nm to 58.8, 66.8, and 87.6 nm for 0.1, 0.3, 0.5, and 0.7 MoS x /r-GO, respectively. Average thickness also gradually increased from 2.9 nm to 4.3, 5.6, and 7.7 nm for 0.1, 0.3, 0.5, and 0.7 MoS x /r-GO respectively. The unsupported MoS 3 particles synthesized using the same method show a larger average particle width of 117.8 nm and height of 22.3 nm (see Supplementary Figure S3). Interestingly, all the MoS x /r-GO composites show a very high aspect ratio-above 10, in contrast to the unsupported MoS 3 . This result indicates that MoS x particles were biaxially grown on the r-GO surface in a planar or coin shape. Thus, we believed that interaction between the precursor and the GO might affect the growth and morphology of the particles, which are closely related to the catalytically active edge sites 16 . The structural composition and interaction between the MoS x and r-GO were investigated using X-ray photoelectron spectroscopy (XPS) as shown in Fig. 3. The XPS spectra of all composite samples exhibited predominant C1s, Mo3d, and S2p peaks. The peak intensities of oxygen functional groups on the GO, such as epoxy, carbonyl, and carboxyl groups, which are observed at binding energies (BE) of 286.7, 288.4, and 289.5 eV, respectively, gradually decreased with increasing precursor concentrations (Fig. 3a). Thus, it is believed that the insulating GO substrate can be spontaneously reduced to the conductive r-GO by the hydrazine or ammonium chloride species generated during the growth reaction of MoS x particles 17 . Similar trends were observed in the X-ray diffraction (XRD) patterns of the GO and MoS x /r-GO composites as shown in Figure S4. The sharp (001) peak of GO at 10.8 degrees, which represents the wider interlayer distance between the graphitic layers compared to graphite, was shifted into the (002) plane at 22.0 degrees for the composites.
Remarkably, the stoichiometric S/Mo ratios of the composite gradually increased from 1.5 to 2.3, 2.6, and 3.3 for 0.1, 0.3, 0.5, and 0.7 MoS x /r-GO, respectively, while the ratio of unsupported MoS 3 particles was approximately ~3.0, indicating the MoS 3 structure. The Mo 3d spectrum with Mo 3d 3/2 and Mo 3d 5/2 doublets indicates that the Mo metal in all composite samples had the 4+ oxidation state. In particular, Mo3d 3/2 (233.2 eV) and Mo3d 5/2 (230.0 eV) in the composite samples were observed at higher BE than in unsupported MoS 3 (231.5 eV and 228.4 eV for Mo3d 3/2 and Mo3d 5/2 , respectively). These values can be attributed to the presence of Mo 5+ and indicate that each of the Mo atoms in the composites was randomly bonded with 2 ~ 3S atoms as indicated in the stoichiometric S/Mo ratio. Wang et al. reported that MoS x was a fundamentally and thermodynamically amorphous structure with many active edge sites, in contrast to crystalline MoS 2 , when the stoichiometric ratio of S atoms to Mo atoms is above 2 18 . There are broad diffraction peaks in all our MoS x and MoS 3 particles in XRD patterns of Figure S4. Thus, the resulting MoS x particles have an amorphous structure irrespective of the GO, which is expected to expose more active edge sites of MoS x .
The S 2p spectrum consists of two doublets. One doublet with higher BE (S2p 3/2 = 163.2 eV and S2p 1/2 = 164.6 eV) is attributed to the existence of both bridging S 2 2− and/or apical S 2− ligands. The other doublet with relatively lower BE (S2p 3/2 = 162.0 eV and S2p 1/2 = 163.2 eV) stems from the existence of the terminal S 2 2− and/or S 2− 19 . Considering the previous reports that the HER activity of MoS x is highly correlated to the amount of terminated S-edge sites, it can be expected that the abundance of catalytic edge sites estimated from the deconvolution of S peaks (area ratio of edged versus bridged S = ~5/4) has a beneficial effect on the hydrogen evolution efficiency.
Importantly  20 . Therefore, we believe that most of the epoxide among the oxygen functional groups plays a crucial role in the anchoring or bridging between MoS x and GO. From the XPS analysis, we can conclude that the novel oxygen-bridged structure could induce the modulation of particle growth, Mo/S stoichiometry, and an amorphous configuration with more exposed active sites, which are expected to improve catalytic activity for hydrogen evolution. On the other hand, the HER activity of the 0.7 MoS x /r-GO catalyst was comparatively decreased, probably due to decreased electrical conductivity and reduced catalytically active sites 21 . The conductivity of 0.5 MoS x /r-GO (5.7 × 10 −2 S/cm) measured by the 4-point probe resistivity measurement was three and two orders higher in magnitude than the conductivity of 0.1 and 0.3 MoS x /r-GO, respectively. The increased conductivity would be originated from the reduction of GO. At the same time, 0.7 MoS x /r-GO had a 44% lower value of 3.2 × 10 −2 S/cm, with respect to the relatively low amount of r-GO even though the high degree of reduction. Therefore, we concluded that the current density for the HER is closely related to the conductivity as shown in Fig. 4b; and thus high electrical conductivity would mainly affect the improvement of electrochemical HER activity.
To confirm the quantitative catalytic activity and rate determining step (RDS), we fitted a Tafel plot based on the HER polarization curves as shown in Fig. 4c. The calculated Tafel slopes were 54, 53, 42, 41, and 112 mV/ decade for 0.1, 0.3, 0.5, 0.7 MoS x /r-GO, and unsupported MoS x particles, respectively. The possible HER process in acidic electrolyte generally consists of three steps; Volmer (H + + e − → H ads , < 120 mV/decade), Heyrovsky (H ads + h + + e − → H 2 , < 40 mV/decade), and Tafel (H ads + H ads → H 2 , < 30 mV/decade) 22 . Considering the Tafel slopes of the catalysts, both the unsupported MoS x and the MoS x /r-GO in this study might favor an electrochemical desorption mechanism, in which electrochemical desorption is the RDS, although the inherent mechanism of Mo sulfide based catalysts has been inconclusive to date 16 . However, the resulting Tafel slope of 0.5 MoS x /r-GO is the smallest among the catalysts. Previous studies have reported that the major factors affecting the HER activity are the surface energy for hydrogen desorption and the rate of electron transfer 23 . It is well-known that MoS x itself is a semiconducting material, while the surface energy of MoS x is theoretically limited to desorbing the hydrogen 23 . Thus, it can be concluded that obvious differences in the HER activity of the catalysts attributed to the electron transfer are evident in the electrochemical impedance spectra at 0.2 V (vs RHE). The MoS x /r-GO catalysts, especially 0.5(~14.6 Ω ), show far lower charge-transfer impedance than unsupported MoS x (~432.6 Ω ), leading to higher HER activity (Fig. 4d). In addition, calculated active site and turnover frequency (TOF) of 0.  Fig. 4e [12][13][14]16,19,[25][26][27][28] . The achieved performance for hydrogen production is significantly useful compared to the materials for solar hydrogen production 29,30 .
The catalytic stability of 0.5 MoS x /r-GO over 3,000 cycles was measured by cyclic voltammetry with a potential range from − 0.3 to 0.2 V as shown in Fig. 5a. After 3,000 cycles, there is no significant change in HER performance except for a slight potential shift. Kibsgaard reported that the slight potential shift caused by not the decline of electrocatalytic activity but rigorous H 2 bubble formation in structure of electrodes, which ultimately results in fewer active sites for HER 19 . In contrast, unsupported MoS x showed a considerable decrease in current density from 14.5 to 5.6 mA/cm 2 after 1,000 cycles. It is believed that the excellent durability of MoS x /r-GO originated from the functional coupling of oxygen bridges between MoS x and r-GO, leading to thermodynamic stability of the MoS x particles. The XPS analysis was conducted to investigate the structural changes before and after the durability test. The atomic ratio of S to Mo after the durability test was converted from 2.6 to 2.0 based on the XPS spectrum as shown in Figure S6. Further, the BE of the deconvoluted S 2p peaks was also shifted to lower positions, as in MoS 2 . Previous studies reported that MoS 3 is electrochemically reduced to MoS 2 as the active species for HER 31 . However, covalent S-O and Mo-O bonds are retained after 1,000 cycles, indicating that MoS x particles could be anchored on the r-GO. Therefore, we believe that the oxygen bridges might improve the stability of HER compared to MoS 3 on multi walled carbon nanotubes with no functional coupling between the MoS 3 and the support (88% after 500 cycles vs initial activity) 32 .
The functional coupling between MoS x and r-GO was also significantly effective in preventing oxidation from affecting catalytic stability. The electrochemical oxidation test was conducted in 0.5 M H 2 SO 4 electrolyte at positive potential. Unsupported MoS 3 initially shows two dominant oxidation peaks at approximately 0.50 and 0.95 V as depicted in Fig. 5b. Thermodynamically unstable sulfur atoms located at edge sites are oxidized first at 0.5 V, and the rest of sulfur atoms in the basal plane are then oxidized later at nearly 0.95 V 33 . However, the oxidation potential of 0.5 MoS x /r-GO is positively shifted to 0.65 V (black arrow in Fig. 5c), indicating high oxidation resistance that is closely related to the stability. In addition, after electrochemical oxidation at 0.65 V, 0.5 MoS x /r-GO exhibits a negligible potential shift, whereas the current density of unsupported MoS x decreases significantly as shown in Fig. 5d. Therefore, the novel functional coupling of oxygen could induce anchoring and oxidation-resistance effects through the strong interaction between MoS x and r-GO, leading to the realization of Mo sulfide based catalysts with tremendous activity and durability.

Discussion
In summary, we synthesized MoS x anchored r-GO composite catalysts by a simple one-pot solution process at room temperature. MoS x particles were covalently bonded to r-GO through oxygen functional groups, and GO was simultaneously reduced to conductive r-GO. The oxygen atoms bridged between MoS x and GO play substantial roles in the fine dispersion of MoS x particles, control of planar MoS x growth, and increase of exposed active sulfur sites, leading to highly efficient and stable electrocatalysts for hydrogen evolution. Therefore, biaxially grown MoS x anchored with r-GO could act as promising nonprecious electrocatalysts for the future hydrogen-based energy world.

Methods
The preparation of GO. The GO was prepared via a modified Hummers method as described in a previous report 16 . First, graphite was dispersed in sulfuric acid (133 mg/ml) by sonication and stirring. Then, KMnO 4 was slowly added to suspension at low temperature, which was kept at 45 °C for 6 h. Then, 100 mL of distilled water and 20 mL of H 2 O 2 were added to remove any residual oxidizing agent. The brownish mixture was washed by centrifugation. The resulting gel-like GO was freeze-dried at − 45 °C for 24 h and used for the preparation of the MoS x /r-GO composite materials.
The preparation of MoS x /r-GO composite materials. First, GO was dispersed in deionized water at a concentration of 3 mg/ml with a brief bath-sonication. Then, a specific amount of ammonium thiomolybdate (0.1, 0.3, 0.5, or 0.7 g) as a MoS x precursor was separately added in 100 ml of GO dispersion with constant stirring at room temperature. Hydrochloric acid (5 ml) was slowly added to the homogeneous mixture. After gas evolution was completed, the product was centrifuged at 7000 rpm for 10 min, followed by washed using ethanol and water to remove acidic residues. Finally, the resulting gel-like MoS x /r-GO was freeze-dried at − 45 °C for 24 h and used as the hydrogen evolution catalyst.
Sample characterization. The crystal structure was investigated using XRD equipment (Smartlab 3, Rigaku) with a scan rate of 2 degree/min from 5 to 70 degrees. The morphologies of the prepared materials were analyzed using atomic force microscopy (AFM, Veeco, Digital Instruments Nanoscope IIIA). A sample for AFM measurement was prepared by spin-coating the catalyst dispersed in DMF at a concentration of ~1 mg/mL onto a Si wafer. The surface morphology and atomic contents of Mo, S, and C in the catalysts were analyzed using a field emission transmission electron microscope (FETEM, JEOL, JEM-2200FS) and X-ray photoelectron spectroscopy (XPS, Thermo Fisher, Multilab 2000), respectively. The TEM specimens were prepared by mixing the products in ethanol using an ultrasonic bath for 5 min, and then a drop of the suspension was placed on a copper grid. The XPS data were recorded using Al Kα radiation (hν = 1000 eV). The electrical conductivity was investigated using a four-point probe instrument (FPP-RS8, Dasol Eng.) and the film thickness of each catalyst was analyzed using a surface profiler (Alphastep IQ, KLA Tencor).
Electrochemical analysis. First, 15 mg of each MoS x /r-GO composite powder was dispersed in a mixture of 1000 μ l of DMF and 100 μ L of Nafion with a brief sonication. Then, 8 μ L of the prepared sample was deposited on glassy carbon electrode stand tried at at 50 °C. Linear sweep voltammetry using a potentiostat with a scan rate of 5 mVs −1 was conducted in 0.5 M H 2 SO 4 electrolyte using an Ag/AgCl electrode as the reference electrode and a platinum wire as the counter electrode.
Calculation of electrochemical active sites and TOF. The oxidation peak at lower potential indicated the oxidation potential of edge area of MoS x to MoO 2 as shown in Fig. 5c. Thus, total current of edge oxidation peak was used to calculate the electrochemical active sites. The following equations were used to calculate the active sites and TOF. = TOF (total hydrogen turnover)/(electrochemical active sites) (3) we assumed that the average number of electrons for each Mo oxidation is approximately 8.9 electrons 32 .