Coupled molybdenum carbide and reduced graphene oxide electrocatalysts for efficient hydrogen evolution

Electrochemical water splitting is one of the most economical and sustainable methods for large-scale hydrogen production. However, the development of low-cost and earth-abundant non-noble-metal catalysts for the hydrogen evolution reaction remains a challenge. Here we report a two-dimensional coupled hybrid of molybdenum carbide and reduced graphene oxide with a ternary polyoxometalate-polypyrrole/reduced graphene oxide nanocomposite as a precursor. The hybrid exhibits outstanding electrocatalytic activity for the hydrogen evolution reaction and excellent stability in acidic media, which is, to the best of our knowledge, the best among these reported non-noble-metal catalysts. Theoretical calculations on the basis of density functional theory reveal that the active sites for hydrogen evolution stem from the pyridinic nitrogens, as well as the carbon atoms, in the graphene. In a proof-of-concept trial, an electrocatalyst for hydrogen evolution is fabricated, which may open new avenues for the design of nanomaterials utilizing POMs/conducting polymer/reduced-graphene oxide nanocomposites.

T o address the energy crisis and ameliorate environmental contamination, researchers have devoted considerable attention to hydrogen as promising alternative to fossil fuels. Electrochemical water splitting to produce hydrogen, or the hydrogen evolution reaction (HER), is the most economical and sustainable method for large-scale hydrogen production. Achieving this goal requires inexpensive electrocatalysts with high efficiency for the HER 1,2 . Although the best electrocatalysts are Pt or Pt-based materials, their high cost and low abundance substantially hamper their large-scale utilization [3][4][5] . Thus, the development of low-cost and earth-abundant non-noble-metal catalysts to replace Pt is an important and urgently needed for practical applications.
Because of their Pt-like catalytic behaviours 6 , Mo-based compounds, such as Mo 2 C 7-10 , MoN [11][12][13] , MoS 2 (refs [14][15][16][17], and others [18][19][20] have attracted substantial interest as a new class of electrocatalysts. To further enhance the HER activity, Mo-based compounds have been anchored onto conductive supports, such as carbon nanosheets (NSs) [21][22][23] and carbon nanotubes (CNTs) 11,24,25 , which not only prevent Mo-based compounds from aggregating but also increase the dispersion of active sites. Among these conductive supports, reduced graphene oxide (RGO), particularly nitrogen (N)-doped RGO, has garnered much attention because of its excellent electron transport properties and chemical stability 26,27 . Therefore, RGO-supported Mo-based compounds appear to be highly active and stable electrocatalysts 11,25,[28][29][30] . However, carbonization at high-reaction temperature during synthesis procedures leads to the sintering and aggregation of Mo-based-compound nanoparticles (NPs), thus reducing their number of exposed active sites and their specific surface area 8,19 . In addition, due to its strong p-stacking and hydrophobic interactions, RGO NSs usually aggregate, which hinders their practical application 31,32 . Preventing the RGO from re-stacking and the Mo-based compound NPs from aggregating during the synthesis of a porous uniform thin layer RGOsupported Mo-based electrocatalysts is critical to enhancing their catalytic performance.
We developed a new approach to integrate polyoxometalates (POMs) and pyrrole (Py) on graphene substrates via a ''one-pot'' method to obtain ternary POMs-polypyrrole/RGO (POMs-PPy/ RGO) nanohybrid sheets with a uniform distribution. As an important family of transition-metal oxide clusters with excellent redox features 33,34 , POMs provided an essential oxidizing medium for the oxidative polymerization of Py 35 , and the POMs finally were converted into ''heteropoly blue'' 36 . Heteropoly blue can be used as a highly localized reducing agent and can further react with graphene oxide (GO) to restore the original POMs. With the polymerization of the Py monomers, POMs were dispersed into the PPy framework. Meanwhile, RGO was homogeneously dispersed and segregated by both the POMs and PPy during the synthesis of POMs-PPy/RGO. Thus, RGO-supported Mo-based catalysts prepared with POMs-PPy/RGO as a precursor may efficiently hinder the Mo sources and graphene from aggregating during the process of forming the RGO-supported NPs. To the best of our knowledge, reports on POMs, PPy and RGO ternary hybrids by a green and one-pot redox relay reaction are rare. More importantly, the coupled hybrid with both Mo 2 C and RGO has not been previously prepared with a ternary hybrid as the precursor.
In this work, we carefully design and fabricate a two-dimensional (2D) coupled hybrid consisting of Mo 2 C encapsulated by N, phosphorus (P)-codoped carbon shells and N, P-codoped RGO (denoted as Mo 2 C@NPC/NPRGO) using a PMo 12 (H 3 PMo 12 O 40 )-PPy/RGO nanocomposite as the precursor. Notably, the entire polymerization and the reductive reactions are triggered by PMo 12 without any additional oxidants or reductants, leading to a synthetic process that is green, efficient and economical. PPy was used as both the carbon and nitrogen sources as well as the reducing agent for GO. Three main advantages of this method are attributed to the Mo 2 C@NPC/ NPRGO hybrid: (1) due to the unique structure of PMo 12 -PPy/ RGO, the Mo 2 C NPs are nanosized and uniformly embedded in the carbon matrix without aggregation; (2) the Mo 2 C NPs are coated with carbon shells, which effectively prevent Mo 2 C NPs from aggregating or oxidizing and impart them with fast electron transfer ability; and (3) owing to the heteroatom dopants (N, P), a large number of active sites are exposed. Overall, taking advantage of the synergistic catalytic effects, the Mo 2 C@NPC/ NPRGO catalyst exhibits excellent electrocatalytic activity for the HER, with a low onset overpotential of 0 mV (vs reversible hydrogen electrode (RHE)), a small Tafel slope of 33.6 mV dec À 1 , and excellent stability in acidic media. Its HER catalytic activity, which is comparable to that of commercial Pt-C catalyst, even superior to those of the best reported non-noble-metal catalysts. In addition, we further investigate the nature of catalytically active sites for the HER using density functional theory (DFT). This approach provides a perspective for designing 2D nanohybrids with transition-metal carbides and RGO as HER catalysts.

Results
Catalyst synthesis and characterization. Mo 2 C@NPC/NPRGO was synthesized as follows: (1) the PMo 12 -PPy/RGO nanocomposite was synthesized via a green one-pot redox relay reaction. The nanocomposite was then carbonized under a flow of ultrapure N 2 at 900°C for 2 h at a heating rate of 5°C min À 1 . Finally, the obtained samples were acid etched in 0.5 M H 2 SO 4 for 24 h with continuous agitation at 80°C to remove unstable and inactive species. The etched samples were thoroughly washed with deionized water until the pH of the wash water was neutral (Fig. 1). Figure 2a shows a scanning electron microscopy (SEM) image of PMo 12 -PPy/RGO. The rough surfaces and wrinkled edges on the sheet-like structures were due to the intercalation and polymerization of Py. A transmission electron microscopy (TEM) image of PMo 12 -PPy/RGO revealed that a large amount of PPy/PMo 12 NPs were homogeneously coated onto the RGO NSs and that voids were present (Fig. 2b). As evident in Fig. 2c and d, the morphologies of Mo 2 C@NPC/NPRGO were similar to that of PMo 12 -PPy/RGO after carbonization. The nanosized Mo 2 C NPs with diameters of B2-5 nm were uniformly decorated on the RGO sheets at a high density, which was attributed to the distinct porous structure of PMo 12 -PPy/RGO. The high-resolution TEM (HRTEM) image exhibited clear lattice fringes with an interplanar distance of 0.238 nm, corresponding to the (111) planes of Mo 2 C (Fig. 2e) 37 . Notably, the Mo 2 C NPs were embedded in the carbon shells, which can efficiently prevent the aggregation and/or excessive growth of Mo 2 C NPs 22 . Figure 2f shows the scanning TEM (STEM) and corresponding energy dispersive X-ray spectroscopy (EDX) elemental mapping images, which confirmed that C, Mo, N and P were distributed on the Mo 2 C@NPC/NPRGO surface, consistent with the EDX spectrum ( Supplementary Fig. 1). These results confirm the successful synthesis of the Mo 2 C@NPC/NPRGO nanocomposite. For comparison, the nanohybrid of Mo 2 C encapsulated by N, P-codoped carbon (defined as Mo 2 C@NPC) was also synthesized through a similar preparation procedure without GO. Supplementary Fig. 2a shows aggregation of PPy/PMo 12 NPs.
Supplementary Fig. 2b and c reveals that Mo 2 C NPs tended to agglomerate during the heat treatment to form large NPs, which decreased the exposed active surface. Supplementary Fig. 2d demonstrates the STEM and EDX elemental mapping images of Mo 2 C@NPC. These data verified that the Mo 2 C@NPC material contained C, N, P and Mo elements, consistent with the EDX results ( Supplementary Fig. 1b). Hence, these results sufficiently confirm that the presence of GO plays an important role in the generation of highly dispersed and nanosized Mo 2 C NPs. Supplementary Fig. 3 shows the powder X-ray diffraction patterns of Mo 2 C@NPC and Mo 2 C@NPC/NPRGO. The broad peak at B25°was ascribed to carbon 38,39 . The other peaks located at 37.9, 43.7, 61.6 and 75.6°were indexed to the (111), (200), (220) and (311) planes of Mo 2 C (JCPDS, No. 15-0457), respectively; these peaks were broad and exhibited low intensity because of the smaller crystallites of Mo 2 C or Mo 2 C coated with amorphous carbon 21,40,41 . Beside, the degrees of graphitization of the two catalysts were analyzed by Raman spectra (Supplementary Fig. 4). As is well-known, the ratio between the D (1,350 cm À 1 ) and G band (1,580 cm À 1 ) intensities (I D /I G ) is an important criterion to judge the degree of the graphitization 9,28 . Compared to Mo 2 C@NPC, the I D /I G of Mo 2 C@NPC/NPRGO is higher, implying that more defects formed on the RGO sheets, thus favoring the accessibility of more active sites and enhancing the electrocatalytic performance. The Brunauer-Emmett-Teller (BET) surface areas of Mo 2 C@NPC and Mo 2 C@NPC/NPRGO calculated by the N 2 sorption isotherms were 55 and 190 m 2 g À 1 , respectively (Supplementary Fig. 5a). Mo 2 C@NPC showed a microporous structure, with pore sizes mainly in the range from 1 to 2 nm ( Supplementary Fig. 5b), whereas the corresponding pore size distribution of Mo 2 C@NPC/NPRGO was mainly concentrated in the range from 1 to 10 nm, which was characteristic of a microporous and mesoporous structure ( Supplementary Fig. 5c). Overall, the large surface area and enriched porous structures efficiently facilitate electrolyte penetration and charge transfer 9 .
X-ray photoelectron spectroscopy (XPS) analyses of Mo 2 C@NPC/NPRGO catalysts were carried out to elucidate their valence states and compositions. As observed, the XPS spectrum of Mo 2 C@NPC/NPRGO ( Supplementary Fig. 6) indicated the presence of C, N, O, P and Mo in the catalyst. The deconvoluted C1s spectrum is shown in Fig. 3a, and the main peak at 284.6 eV implies that the graphite carbon is the majority species 22 . The deconvolution of N1s energy level signals for Mo 2 C@NPC/ NPRGO revealed the peaks at 398.6 and 401.3 eV, which were assigned to pyridinic and graphitic N (Fig. 3b), respectively 21,27 . From Fig. 3c, it can be seen that the P2p peaks at about 133.5, and 134.8 eV were attributed to P-C and P-O bonding, respectively 18,28 . Besides, the high-resolution Mo 3d XPS revealed that the peak at 228.8 eV was attributable to Mo 2 þ , stemming from Mo 2 C. In parallel, as a consequence of surface oxidation, the peaks at 232.05 and 235.2 were attributable to MoO 3 and that at 232.7 eV was assignable to MoO 2 (refs 8,21); both of these species are inactive toward the HER (Fig. 3d). For comparison, Mo 2 C@NPC is shown in Supplementary Fig. 7. All of these data were similar to those for Mo 2 C@NPC/NPRGO. The corresponding atomic percentages of the different catalysts measured by XPS are listed in Supplementary Table 1.
Electrocatalytic HER performance. A three-electrode system was adopted to evaluate the electrocatalytic activities of Mo 2 C@NPC/ NPRGO toward the HER in 0.5 M H 2 SO 4 at 100 mV s À 1 . For comparison, Mo 2 C@NPC and commercial Pt-C (20 wt% Pt on carbon black from Johnson Matthey) were also assessed. The corresponding polarization curves without IR compensation are shown in Fig. 4a. All potentials in this work are reported vs RHE. As expected, the commercial Pt-C displayed the highest ARTICLE electrocatalytic activity, with an onset overpotential of nearly zero 30 . The Mo 2 C@NPC catalyst exhibited far inferior HER activity. Impressively, Mo 2 C@NPC/NPRGO exhibited the lowest onset overpotential of 0 mV, approaching the performance of commercial Pt-C. Moreover, it was clearly observed that the cathodic current rose sharply with more negative potentials. Generally, the potential value for a current density of 10 mA cm À 2 is an important reference because solar-light- coupled HER apparatuses usually operate at 10-20 mA cm À 2 under standard conditions (1 sun, AM 1.5) 4 . To achieve this current density, Mo 2 C@NPC requires an overpotential of 260 mV. Strikingly, Mo 2 C@NPC/NPRGO required only B34 mV to achieve a 10 mA cm À 2 current density, even superior to commercial Pt-C (40 mV) (  Table 2).
To elucidate the HER mechanism, Tafel Plots were fitted to Tafel equation (that is, Z ¼ b log (j) þ a, where b is the Tafel slope, and j is the current density), as shown Fig. 4b. The Tafel slope of commercial Pt-C was B30 mV dec À 1 , which was in agreement with the reported value, thus supporting the validity of our electrochemical measurements 30 . The Tafel slope of Mo 2 C@NPC/ NPRGO was 33.6 mV dec À 1 , which indicated higher performance than that of Mo 2 C@NPC (126.4 mV dec À 1 ). Meanwhile, the Tafel slope of Mo 2 C@NPC/NPRGO suggested that hydrogen evolution on the Mo 2 C@NPC/NPRGO electrode probably proceeds via a Volmer-Tafel mechanism, where the recombination is the ratelimiting step 17 . The exchange current density (j 0 ) was extrapolated from the Tafel plots. Notably, Mo 2 C@NPC/NPRGO displayed the largest j 0 of 1.9 Â 10 À 3 A cm À 2 , which was nearly three times larger than the j 0 of Pt-C (0.39 Â 10 À 3 A cm À 2 ) ( Table 1) and was substantially greater than those of other recently reported nonnoble-metal catalysts (Supplementary Table 2). This performance of Mo 2 C@NPC/NPRGO demonstrates favourable HER kinetics at the Mo 2 C@NPC/NPRGO/electrolyte interface.
The electrochemical double-layer capacitance (EDLC, C dll ) was measured to investigate the electrochemically active surface area. Cyclic voltammetry (CV) was performed in the region from 0.27 to 0.37 V at rates varying from 20 to 200 mV s À 1 (Fig. 4c and Supplementary Fig. 8). The C dll of Mo 2 C@NPC/NPRGO (17.9 mF cm À 2 ) was B195 times larger than that of Mo 2 C@NPC (0.092 mF cm À 2 ). Thus, the large j 0 value of Mo 2 C@NPC/ NPRGO may benefit from both its large BET surface area and its large EDLC.
To gain further insight into the electrocatalytic activity of Mo 2 C@NPC/NPRGO for the HER, we performed electrochemical impedance spectroscopy (EIS). The Nyquist plots of the EIS responses are shown in Supplementary Fig. 9. Compared with the Nyquist plot of Mo 2 C@NPC, that of Mo 2 C@NPC/ NPRGO showed a much smaller semicircle, suggesting that Mo 2 C@NPC/NPRGO has lower impedance. This result proves that the catalyst affords markedly faster HER kinetics due to the presence of the RGO support.
Long-term stability is also critical for HER catalysts. To probe the durability of the Mo 2 C@NPC/NPRGO catalyst, continuous CV was performed between À 0.2 and 0.2 V at a 100 mV s À 1 scan rate in 0.5 M H 2 SO 4 solution (Fig. 4d). As observed, the polarization curve for Mo 2 C@NPC/NPRGO remained almost the same after 1,000 cycles. In addition, the durability of Mo 2 C@NPC/NPRGO was also examined by electrolysis at a static overpotential of 48 mV. The inset of Fig. 4d shows that the current density experienced a negligible loss at B20 mA cm À 2 for 10 h. For comparison, the durability of the Mo 2 C@NPC catalyst was examined by the same methods ( Supplementary  Fig. 10). This is reconfirming that Mo 2 C@NPC and Mo 2 C@NPC/ NPRGO are stable electrocatalysts in acidic solutions.
In control experiments, we investigated the effect of the PMo 12 content on electrocatalytic performance. Two other catalysts with different PMo 12 contents (1.1 and 3.3 g) were synthesized (denoted as Mo 2 C@NPC/NPRGO-1.1 and Mo 2 C@NPC/ NPRGO-3.3). The morphology, structure and composition of these two catalysts were studied by SEM, TEM, HRTEM, STEM, EDX, elemental mapping, powder X-ray diffraction patterns and XPS in detail ( Supplementary Figs 11-16). The HER activities of Mo 2 C@NPC/NPRGO-1.1 and -3.3 were also evaluated using the same measurements. As seen from Fig. 5a,b, Mo 2 C@NPC/ NPRGO showed the lowest onset overpotential and the smallest Tafel slope among the three samples. We speculate that these results are likely related to the amount and distribution of active sites. Because of the lower amount of Mo 2 C NPs in Mo 2 C@NPC/ NPRGO-1.1, the corresponding electrocatalytic activity was poorer than that of Mo 2 C@NPC/NPRGO. In contrast, a larger number of Mo 2 C NPs in Mo 2 C@NPC/NPRGO-3.3 aggregated together, which is also unfavourable for the HER. These results demonstrate that the amount of PMo 12 substantially influences the HER performance.
We subsequently studied the influence of carbonization temperature under the given conditions. Supplementary  Figs 17-22 show the morphology, structure and composition of the two samples carbonized at 700 and 1,100°C (defined as PMo 12 -PPy/RGO-700 and Mo 2 C@NPC/NPRGO-1100), respectively. The onset overpotentials of PMo 12 -PPy/RGO-700 and Mo 2 C@NPC/NPRGO-1100 were 20 and 27 mV, respectively, and the Tafel slopes were 48.4 and 70.1 mV dec À 1 , respectively ( Fig. 5c and d). Among these catalysts, the Mo 2 C@NPC/NPRGO catalyst exhibited the optimal HER activity, possibly because active sites of Mo 2 C were not produced when PMo 12 -PPy/RGO is carbonized at 700°C; the high-carbonization temperature led to substantial sintering and aggregation of Mo 2 C NPs, which further reduced the density of highly active sites. Meanwhile, the N content decreased with increasing carbonization temperature (Supplementary Table 1). All of these results were consistent with the SEM, TEM, XRD, thermogravimetric analysis and XPS results ( Supplementary Figs 17-22). Therefore, in this work, the selection of the correct PMo 12 content and carbonization temperature was critical to forming high-HER active sites.
Theoretical investigation. The aforementioned experimental results demonstrated that the Mo 2 C@NPC/NPRGO composite exhibits excellent electrocatalytic activity toward the HER because of the synergistic effects of Mo 2 C and NPC/NPRGO. To elucidate the mechanism underlying the superior HER activity of the Mo 2 C@NPC/NPRGO composite, we performed a series of DFT calculations ( Supplementary Fig. 23 and Supplementary Table 3). Theoretically, the HER pathway can be depicted as a three-state  E0), which can provide a fast proton/electron-transfer step as well as a fast hydrogen release process 42 . Because only trace amounts of P were present in the Mo 2 C@NPC/NPRGO hybrid compared to the N content, we investigated only the effect of N doping (graphitic N and pyridinic N) on the catalytic effect of the hybrids. Figure 6 shows the calculated free energy diagram for the HER in various studied systems. According to our computational results, pristine graphene had an endothermic DG H* of 1.82 eV, implying an energetically unfavourable interaction with hydrogen. Therefore, the HER can barely proceed on pristine graphene because of the slow proton/ electron transfer. On the other hand, the (001) surface of Mo 2 C had a strong interaction with H, as indicated by the exothermic DG H* of À 0.82 eV, which would subsequently lead to poor HER performance because of the foreseeable difficulty of hydrogen release. Moreover, N-doped graphene exhibited low catalytic activity toward the HER. Specifically, the DG H* values for graphitic-N-and pyridinic-N-doped graphene were 0.89 and À 2.04 eV, respectively.
However, the catalytic activity of graphene and N-doped graphene were substantially improved when they were anchored to the surface of Mo 2 C. For example, the DG H* values for Mo 2 C@C and Mo 2 C@C-graphitic N were 0.41 and 0.69 eV, respectively, which were much lower than those of suspended graphene (1.82 eV) and N-doped graphene (0.89 eV). The DG H* of Mo 2 C@C (0.41 eV) indicated that the graphene C atoms in the hybrid also play an important role in the HER activity. In particular, due to the synergistic effect between Mo 2 C and C-pyridinic N, Mo 2 C@C-pyridinic N had a favourable DG H* ( À 0.22 eV) for the adsorption and desorption of hydrogen. Therefore, the active sites for the HER should be composed mainly of pyridinic N atoms and C atoms of graphene rather than graphitic N atoms. We note here that, according to the results of XPS analysis, the major type of N in Mo 2 C@NPC/NPRGO was pyridinic N, which means that Mo 2 C@NPC/NPRGO would manifest a high density of active sites and would consequently present a high-current density at a low overpotential for the HER. Overall, the experimental and theoretical results verified that as-synthesized Mo 2 C@NPC/NPRGO is an unexpected and highly efficient HER electrocatalyst.

Discussion
In view of the aforementioned considerations, the amazing HER activities of the Mo 2 C@NPC/NPRGO are postulated to originate from the following reasons: (1) the small size of Mo 2 C NPs favors the exposure of an abundance of available active sites, which may enhance the catalytic activity for the HER 7,21,28 ; (2) the introduction of heteroatoms (N, P) into the carbon structure results in charge density distribution and asymmetry spin, thus enhancing the interaction with H þ (refs 18,27). Especially, pyridinic N is favourable for highly efficient catalytic performance 43,44 ; (3) as an advanced support, RGO can increase the dispersion of PMo 12 to further obtain highly dispersed Mo 2 C during the carbonization process. Meanwhile, the outstanding electrical conductivity of RGO facilitates charge transfer in the catalyst 11,25 ; (4) the robust conjugation between Mo 2 C and NPC/ NPRGO provides a resistance-less path favourable for fast electron transfer. The carbon shells may hamper the aggregation of Mo 2 C NPs 21 and promote electron penetration from Mo 2 C to RGO 22 . Furthermore, the geometric confinement of Mo 2 C inside the carbon shells can also enhance the catalytic activity for the HER 40 and (5) the unique structure of Mo 2 C@NPC/NPRGO is favourable for the fast mass transport of reactants and facilitates electron transfer 26,39 . Because of the synergistic catalytic effects of the aforementioned factors, the Mo 2 C@NPC/NPRGO catalyst exhibits potent HER activity.
In summary, we designed and developed a novel architecture that is composed of Mo 2 C NPs, NPC and NPRGO by simply carbonizing a ternary PMo 12 -PPy/RGO nanocomposite. The effect of the PMo 12 content and carbonization temperature on the HER activity was investigated in detail. The RGO-supported Mo-based catalysts prepared with PMo 12 -PPy/RGO as the precursor may efficiently hinder Mo sources and graphene from aggregating during the formation of RGO-supported Mo 2 C NPs. The obtained Mo 2 C@NPC/NPRGO nanocomposite exhibits the best HER performance and high stability as an electrocatalyst in an acidic electrolyte reported to date. Theoretical studies demonstrated that the synergistic effect between Mo 2 C and C-pyridinic N contributes to the excellent HER activity of the Mo 2 C@NPC/NPRGO nanocomposite, in accordance with the experimental results. This proof-of-concept study not only offers novel hydrogen-evolving electrocatalysts with excellent activity but also opens new avenues for the development of other 2D coupled nanohybrids with transition-metal carbides and RGO using POMs/conducting polymer/RGO as a precursor. These catalysts can also be explored as highly efficient electrocatalysts for oxygen reduction reaction (ORR), HER and lithium batteries.

Methods
Synthesis of PMo 12 -PPy/RGO and Mo 2 C@NPC/NPRGO hybrids. In a typical synthesis, GO NSs were pre-synthesized by chemical oxidation exfoliation of natural graphite flakes using a modified Hummers method 45 . The obtained GO NSs were dispersed in de-ionized water by ultrasonication to form a suspension with the concentration of 1 mg ml À 1 . Around 12.5 ml of such GO suspension and 150 ml of 2 mM PMo 12 solution were added into a clean three-necked flask, respectively, and mixed uniformly under a strong ultrasonication bath. Subsequently, Py monomer solution by dispersing 230 ml of Py in 15 ml de-ionized water, was slowly dropped into the above mixed PMo 12 /GO suspension. With the addition of Py monomer solution, the reaction system gradually turned from yellow-brown to deep blue and a black precipitate began to generate after about 5 min. Finally, the reactor was transferred to an oil bath and allowed to react for 30 h at 50°C under vigorously magnetic stirring. After separated by centrifugation and washed with deionized water and anhydrous ethanol for several times, the black PMo 12 -PPy/RGO ternary nanohybrids were obtained, which were dried in vacuum at 50°C. In control experiments, PMo 12 -PPy/RGO (1.1) and PMo 12 -PPy/ RGO (3.3) were synthesized by identical condition except that the amount of PMo 12 is 1.1 and 3.3 g, respectively.
To prepare the Mo 2 C@NPC/NPRGO nanocomposite, 2 g of PMo 12 -PPy/RGO was carbonized in a flow of ultrapure N 2 at 900°C for 2 h with the heating rate of 5°C min À 1 . The obtained samples were acid etched in H 2 SO 4 (0.5 M) for 24 h with continuous agitation at 80°C to remove unstable and inactive species. The etched samples were then thoroughly washed with de-ionized water until reaching a neutral pH, and defined as Mo 2 C@NPC/NPRGO, Mo 2 C@NPC/NPRGO-1.1 and -3.3, respectively.
Synthesis of PMo 12 -PPy and Mo 2 C@NPC composites. The synthetic procedure is very similar to PMo 12 -PPy/RGO without GO. Likewise, the preparation of Mo 2 C@NPC composite is identical with that of Mo 2 C@NPC/NPRGO.
Characterization. The TEM and HRTEM images were recorded on JEOL-2100F apparatus at an accelerating voltage of 200 kV. Surface morphologies of the carbon materials were examined by a SEM (JSM-7600F) at an acceleration voltage of 10 kV. The EDX was taken on JSM-5160LV-Vantage-typed energy spectrometer. The XRD patterns were recorded on a D/max 2500VL/PC diffractometer (Japan) equipped with graphite monochromatized Cu Ka radiation (l ¼ 1.54060 Å).
Corresponding work voltage and current is 40 kV and 100 mA, respectively. XPS was recorded by a scanning X-ray microprobe (PHI 5000 Verasa, ULAC-PHI, Inc.) using Al Ka radiation and the C1s peak at 284.8 eV as internal standard. The Raman spectra of dried samples were obtained on Lab-RAM HR800 with excitation by an argon ion laser (514.5 nm). The nitrogen adsorption-desorption experiments were operated at 77 K on a Micromeritics ASAP 2050 system. BET surface areas were determined over a relative pressure range of 0.05-0.3, during which the BET plot is linear. The pore size distributions were measured by using the nonlocalized density functional theory method. Before the measurement, the samples were degassed at 150°C for 10 h.
Electrochemical measurements. All electrochemical experiments were conducted on a CHI 760D electrochemical station (Shanghai Chenhua Co., China) in a standard three electrode cell in 0.5 M H 2 SO 4 at room temperature. A glassy carbon electrode (3 mm in diameter), an Ag/AgCl with saturated KCl, and a Pt wire were used as the working electrode, reference and counter electrode, respectively. A total of 4 mg of the catalysts were dispersed in 2 ml of 9:1 v/v water/Nafion by sonication to form a homogeneous ink. Typically, 5 ml well-dispersed catalysts were covered on the glassy carbon electrode and then dried in an ambient environment for measurements. The electrocatalyst was prepared with a catalyst loading of 0.14 mg cm À 2 . Commercial 20% Pt-C catalyst was also used as a reference sample. Linear sweep voltammetry was tested with a scan rate of 5 mV s À 1 . EIS measurements were carried out from 1,000 kHz to 100 mHz with an amplitude of 10 mV at the open-circuit voltage. The electrochemical stability of the catalyst was conducted by cycling the potential between À 0.3 and 0.3 V vs RHE at a scan rate of 100 mV s À 1 . The Chronoamperometry were tested at an overpotential of À 0.12 V vs RHE after equilibrium. To estimate the electrochemical active surface areas of the catalysts, CV was tested by measuring EDLC under the potential window of 0.19-0.39 vs RHE with various scan rate (20, 40, 60, 80, 100, 120, 140, 160, 180 and 200 mV s À 1 ). A flow of N 2 was maintained over the electrolyte during the experiment to eliminate dissolved oxygen. The potential vs RHE was converted to RHE via the Nernst equation: Computational details. DFT calculations were performed using the plane-wave technique implemented in the Vienna ab initio Simulation package 46 . The ion-electron interaction was treated within the projector-augmented plane wave pseudopotentials 47,48 . The generalized gradient approximation expressed by Perdew À Burke À Ernzerhof functional 49 and a plane-wave cutoff energy of 360 eV were used in all computations. The electronic structure calculations were employed with a Fermi-level smearing of 0.1 eV for all surface calculations and 0.01 eV for all gas-phase species. The Brillouin zone was sampled with 3 Â 3 Â 1 k-points. The convergence of energy and forces were set to 1 Â 10 À 5 eV and 0.02 eV Å À 1 , respectively. A vacuum region of around 12 Å was set along the z direction to avoid the interaction between periodic images. More computational details are provided in Supplementary Note 1.