As a clean and renewable resource, hydrogen is believed to be one of the most promising alternative energy carriers [1] [2] . Hydrogen can be generated by electrolysis or photoelectrolysis of water via the hydrogen evolution reaction (HER). Effective electrocatalysts for HER are essential for efficient hydrogen generation from electrolysis and photoelectrolysis. Though platinum-group metals have shown excellent catalytic activity in HER, their widespread commercial application is inhibited by high cost and low abundance. The exploitation of effective and low-cost HER catalysts is therefore highly desirable.

Recently, the development of cost-effective HER catalysts has gained extensive attention. Great efforts have has been devoted to explore HER catalysts among transition metal chalcogenides (e.g., molybdenum sulfide [3] - [7] , tungsten sulfide [8] [9] , cobalt dichalcogenide [10] [11] , iron dichalcogenide [10] [11] , and nickel dichalcogenide [10] [11] ), carbides (e.g., molybdenum carbide [12] - [14] , and tungsten carbide [15] [16] ), as well as nitrides and carbonitrides (e.g., molybdenum nitride [13] , cobalt-molybdenum nitride [17] , tungsten carbonitride [18] ). On the other hand, nickel phosphide has been predicted to be an excellent catalyst in HER in 2005 [19] , whereas the catalytic activity of nickel phosphide (Ni₂P nanoparticles) was experimentally confirmed in 2013 [20] . The successful demonstration of catalytic activity of Ni₂P inspires research concerning the application of metal phosphide in hydrogen generation (e.g., Ni₁₂P₅ [21] , and CoP [22] [23] ).

Herein the catalytic activity of MoP flakes in HER is demonstrated. We found that MoP flakes show efficient catalytic activity in HER in both acidic and basic solution, and the corresponding overpotential required for 20 mA∙cm⁻² current density is 155 mV in acidic solution and 184 mV in basic solution, respectively. The performance is favorably comparable to most values of reported nonprecious metal catalysts. Potentiostatic electrolysis and accelerated degradation experiments demonstrate the long-term stability of MoP flakes in hydrogen generation in acidic and basic solution. The influence of synthesis temperature on catalytic activity of MoP flakes was revealed. Tafel slope suggests that the HER occuring on the surface of MoP flakes proceeds along Volmer-Heyrovsky mechanism. Mo and P in MoP flakes were found to have slight charge, and the catalytic activity of MoP might be correlated with these features.

MoP was synthesized by the reduction of hexaammonium heptamolybdate tetrahydrate ((NH₄)₆Mo₇O₂₄∙4H₂O) and ammonium dihydrogen phosphate (NH₄H₂PO₄) at 800˚C. The XRD pattern of product Figure 2(a) exhibits

distinct peaks, and these peaks can be assigned to those of hexagonal phase MoP (JCPDS No. 65-6487, a = 3.223 Å, c = 3.191 Å). These distinct peaks suggest the good crystallinity of MoP. The composition information of product was accessed by EDX analysis carried out in SEM. The average atomic ratio of Mo to P recorded from five different sites is 0.976 ± 0.053 (Figure 3 and Table 1), in accordance with that of stoichiometrical MoP (1:1).

Typical morphology of MoP was investigated by SEM and TEM. Figure 2(b) shows the low magnification SEM image of MoP, revealing a flake like morphology. The lateral size of MoP flakes ranges from less than 1μm to larger than 5 μm, and the thickness of MoP fakes is ca. 200 nm (Figure 2(c)). It is further revealed that the MoP flakes are porous and constructed by MoP nanoparticles with diameter ca. 100 nm (Figure 4(d)). The high-resolution TEM (HRTEM) image suggests that these small MoP nanoparticles are single crystalline (Figure 4(e)). The good crystallinity of the MoP nanoparticles can be suggested by the well defined lattice fringes. The distances between fringes are measured to be 2.8 and 3.2 Å, respectively, corresponding well to those between the (100) and (00-1) planes of hexagonal phase MoP. The pattern resulted from fast Fourier transform (FFT) of the lattice fringes is shown in the inset of Figure 2(e), matching well that of the (010) zone axis diffraction pattern of hexagonal phase MoP.

The MoP flakes exhibit effective catalytic activity in HER. Linear sweep voltammetry (LSV) experiments were carried out to obtain polarization curves of MoP flakes. Figure 4 shows the optimal performance of MoP flake loaded on glassy carbon electrode (GCE) with and without addition carbon black (Vulcan XC-72R). For comparison, the polarization curves of bare GCE and GCE loaded with carbon black (C on GCE) are also shown. It is found that in acidic solution (H₂SO₄, 0.5 M) overpotential required for current density of 10 mA∙cm⁻² (η₁₀)

for MoP on GCE is ca. 423 mV, while η₁₀ is markedly reduced to 141 mV and η₂₀ (overpotential required for current density of 20 mA∙cm⁻²) is as small as 155 mV for the hybrid materials of MoP flakes and carbon black (MoP/C on GCE). The current density of carbon black on GCE (C on GCE) is smaller than that of MoP on GCE, and bare GCE shows neglectable current density in the potential range of −0.5 - 0 V vs RHE, suggesting that the current in MoP/C on GCE sample can be associated with MoP flakes. The small current density in MoP on GCE sample might be correlated with slow electron transport between less-conducting MoP flakes. The introduction of carbon black enables rapid electron transport from electrode to MoP flakes at catalyst/electrolyte interface, and results in markedly enhanced hydrogen generation performance.

The η₁₀ and η₂₀ are usually adopted as key parameters for the comparison of the catalytic activity of different HER catalysts, because the typical current density of photoelectrochemical water splitting cell ranges in 10 - 20 mA∙cm⁻² under the solar photon flux of 1 Sun AM 1.5 illumination [24] . Typical η₁₀ and η₂₀ of nonprecious HER catalysts are listed in Table 2. It is shown that the optimal η₁₀ and η₂₀ of MoP/C on GCE are favorably comparable to most values of reported nonprecious metal HER catalyst.

The HER performance depends heavily on the loading amount of MoP flake and carbon black. A series of

^ah₁₀: overpotential required for 10 mA∙cm⁻² current density; ^bh₂₀: overpotential required for 20 mA∙cm⁻² current density.

LSV experiments were carried out to determine the optimal performance of MoP flake. These results are shown in Figure 5. The η₁₀ of MoP on GCE sample firstly increases with the loading amount of MoP, with the optimal loading amount of 0.855 mg∙cm⁻², and further increase the loading amount of MoP results in the decreasing of η₁₀. The relationships between η₂₀ and the loading amount of MoP in MoP/C on GCE samples with different weight ratios of MoP to C are summarized (Figure 5(b)), and the corresponding polarization curves are shown in panels c-e of Figure 5. It is shown that the smallest η₂₀ can be found in MoP/C on GCE with the weight

ratio of MoP to C being 2:1 and the loading amount of MoP being 1.425 mg∙cm⁻².

The fabrication temperature influences the catalytic activity of MoP flake in HER. MoP flakes were fabricated at different temperatures, and the products are denoted as MoPxxx, where xxx is the fabrication temperature. The plots of h₂₀ versus the loading amounts of MoP750 and MoP850 are shown in Figure 6(a), and the polarization curves of MoP750, and MoP850 can be found in Figure 6(b) and Figure 6(c), respectively. The optimal h₂₀ is 210 mV for M750 (the loading amount of MoP: 1.425 mg∙cm⁻²) and 190 mV for M850 (the loading amount of MoP: 1.425 mg∙cm⁻²), respectively, both larger than that of MoP800 (Figure 4). The XRD experiments reveal that the structure of M750 and M850 is the same as that of MoP800 (Figure 7(a)), and no impurity phase can be detected from the corresponding XRD patterns. The peaks are more distinct for sample obtained at higher temperature, suggesting that higher temperature would be beneficial for the crystallization of MoP flake. The specific surface areas of MoP750, MoP800, and MoP850 were measured by Brunauer, Emmett and Teller (BET) method Figure 7(b), being 4.798, 4.835, and 3.767 m²∙g⁻¹, respectively. The results of XRD and BET experiments suggest that the crystallinity and surface area of MoP flake work cooperatively, resulting in an optimal catalytic activity of sample obtained at 800˚C.

The faradaic efficiency of MoP flakes in HER was evaluated by the comparison of measured volume and theoretical volume of generated hydrogen. The generated hydrogen was measured by water displacement method, and the theoretical volume was computed by assuming that all elecrons passing through the circuit are consumed by the reduction reaction of H⁺ (2H⁺ + 2e⁻ → H₂). Figure 8 shows the plots of measured volume and theoretical volume of hydrogen generated during a potentiostatic electrolysis. The measured volume of hydrogen matches well the theoretical one within experimental error, suggesting that the faradaic efficiency of MoP flakes is nearly 100% in hydrogen generation.

Long-term stability during hydrogen generation is also important for the practical application of HER catalyst. MoP flake was found to work stably in acidic solution during hydrogen generation, as indicated by potentiostatic electrolysis and accelerated degradation experiments (Figure 9). A potentiostatic electrolysis experiment shows that only slightly decrease of current density occurs in the first 5000 s, and thereafter the current density is nearly unchanged. An accelerated degradation experiment was carried out by repeated cyclic voltammetry (CV) sweep. Negligible difference can be found between the polarization curves of the initial and 2000th scan. The potentiostatic electrolysis and accelerated degradation experiments demonstrate the long-term hydrogen generation capability of MoP flakes.

To obtain insight into the HER process of MoP flakes, EIS experiments were carried out in acidic solution. The spectra corresponding to different applied potentials are shown in a Nyquist plot in Figure 10(a). All spectra contain two semicircles, and they can be well fitted with a two-time-constant equivalent circuit (Figure 11). Charge transfer resistance (R_ct) at solid/liquid interface was derived from semicircles at low frequencies range. R_ct is associated with HER kinetics of a catalyst, and a smaller R_ct suggests faster kinetics. Figure 10(a) shows that R_ct decreases with increasing applied potential, in accordance with larger current density at larger applied potential. The applied potentials are plotted versus inverse R_ct on a logarithmic scale (Figure 10(b)), and the Tafel slope was determined by the slope in the plot, being 71.77 mV∙dec⁻¹.

The HER process can be revealed by Tafel slope. In general, a classic two-electron-reaction model suggests that the HER process can proceed in two steps: a discharge step (Volmer reaction: H₃O⁺ + e⁻ → H_ads + H₂O) followed by a desorption step (Heyrovsky reaction: H_ads + H₃O⁺ + e⁻ → H₂ + H₂O), or a discharge step followed by a recombination step (Tafel reaction: H_ads + H_ads → H₂), where H_ads represents a H atom absorbed at the active site of the catalyst. The rate-limiting step can be identified by Tafel analysis of the catalyst, and a Tafel slope of 116, 38, or 29 mV∙dec⁻¹ assigns the rate-determining step in the HER process to Volmer, Heyrovsky, or Tafel reaction. The Tafel slope of 71.77 mV∙dec⁻¹ lies between 38 and 116 mV∙dec⁻¹, suggesting that a Volmer- Heyrovsky mechanism might be responsible for the HER process [13] , and that the rates of the discharge step and the desorption step might be comparable during the HER process [25] .

The chemical states of Mo and P in MoP flake were investigated by XPS experiments to find possible origin of the HER catalytic activity of MoP flake. The corresponding results are shown in Figure 12. The XPS spectrum of the Mo 3d window can be fitted to two doublets with peaking energy at 228.3 eV and 232.5 eV, respectively Figure 12(a). The peak at 228.3 eV can be associated with Mo in MoP [26] [27] , and this binding energy is very close to that of elemental Mo (227.9 eV) [28] , implying that these Mo species have very a small positive charge. On the other hand, the peak at 232.5 eV suggests the presence of Mo⁶⁺ species in product [26] [27] . Two doublets (129.5 and 133.8 eV) can be found in the P 2p window Figure 12(b). The doublet at 129.5 eV can be attributed to P bonding to Mo [26] [27] , and these P species have a slightly negative charge, because their binding energy (129.5 eV) is only 0.5 eV smaller than that of elemental P (130.0 eV) [29] . The doublet at 133.8 eV

can be related to P in species [26] [27] . The Mo⁶⁺ and possibly come from un-reacted precursor or oxidation product of MoP due to air exposure.

In Mo₂C and Mo₂N, Mo was found to have very small positive charge because of charge transfer from Mo to C or N [13] [14] . The d band electron structure becomes similar to that of Pt as a result of the charge transfer, and the HER catalytic activity of Mo₂C and Mo₂N have been associated with the modified d band structure of Mo [14] . The XPS spectrum of the Mo 3d window from MoP flakes reveals that Mo species in MoP flakes also have a very small positive charge. In addition, active sites in hydrogenase [19] , its analogues (Ni(PS3^*)(CO)]⁻ and [Ni(PNP)₂]²⁺) [19] and Ni₂P [19] contain hydride acceptors (Ni), which have a small positive charge. The charged nature of Mo in MoP resembles that of Mo in Mo₂C and Mo₂N, as well as the hydride acceptors in hydrogenase, its analogues, and Ni₂P, implying that the weakly charged Mo species in MoP might contribute to the catalytic activity of MoP. On the other hand, the proton acceptors are another kind of active sites in hydrogenase,

its analogues, and Ni₂P. The proton acceptors are nonmetal sites having a small negative charge to trap protons (e.g., O of Glu23 in hydrogenase, −0.44 e; S in Ni(PS3^*)(CO)]⁻, −0.4 e; N in [Ni(PNP)₂]²⁺, −0.34 e; P in Ni₂P, −0.07 e) [19] . The P species in MoP flakes are also slightly negatively charged. This charged nature is analogous to that of the proton acceptors in hydrogenase, its analogues, and Ni₂P. It is reasonable to postulate that the slightly charged P species in MoP work similarly to the proton acceptors in hydrogenase, its analogues, and Ni₂P, also contributing positively to the catalytic activity of MoP flakes. Therefore, the catalytic activity of MoP flakes is likely to be correlated with the charged natures of Mo and P.

Finally, the HER performance of MoP flakes in basic solution (KOH, 1M) was evaluated. The results are summarized in Figure 13. The corresponding η₁₀ and η₂₀ is 166 and 184 mV, respectively (Figure 13(a)). Current density recorded in a potentiostatic electrolysis experiment shows slight decrease in 10000 s, and the increase of η₂₀ after 2000 scan in repeated CV scans is as small as 14 mV (inset of Figure 13(b)). In addition, the faradaic yield is also nearly 100% within the experimental error (Figure 13(c)). These experiments demonstrate that MoP flake can also work efficiently and stably in basic solution for hydrogen generation.

In summary, MoP flake was found to show efficient and stable catalytic activity in HER in acidic and basic solution. The optimal η₂₀ of MoP flakes mixed with carbon black is as small as 155 mV in acidic solution and 184 mV in basic solution. The fabrication temperature of MoP flakes was found to influence the catalytic activity of MoP flakes. The samples fabricated at 800˚C showed superior performance to those at 750˚C or 850˚C. The difference of catalytic activities might be associated with the crystallinity and specific surface area of products obtained at different temperatures. Charge transport resistances at the interface of catalyst/electrolyte suggest a Tafel slope of 71.77 mV∙dec⁻¹, implying that a Volmer-Heyrovsky mechanism might be responsible for the HER process on the surface of MoP flakes. Potentiostatic electrolysis and accelerated degradation experiments show that MoP flakes can work stably in long-term hydrogen generation in both acidic and basic solution. In

addition, the faradaic yield of MoP flakes in acidic and basic solution is nearly 100%. XPS experiments reveal that Mo and P in MoP flakes have a slight charge, and the catalytic activity of MoP flakes might be associated with the charged natures of Mo and P.

This research was financially supported by the National Natural Science Foundation of China (61006049, 50925207), the Ministry of Science and Technology of China (2011DFG52970), the Ministry of Education of China (IRT1064), Jiangsu Innovation Research Team, the Education Department of Jiangsu (10KJB430004), Jiangsu Province (2011-XCL-019, and 2013-479), and Jiangsu University (09JDG043 and Outstanding Youth Project).