CoO-Mo2N hollow heterostructure for high-efficiency electrocatalytic hydrogen evolution reaction

Driving the electrocatalytic hydrogen evolution reaction (HER) with solar-energy cells is considered a green and sustainable way to produce H2. Herein, CoO-Mo2N hollow heterojunctions were designed for effective HER based on the combined virtues of the hollow structure and heterojunctions. The hollow CoMoO4-Co(OH)2 precursor was first synthesized via the reaction of Co2+ from ZIF-67 with MoO42− and OH− in a Na2MoO4 solution. A series of experiments indicate the formation of the hollow Co-Mo-O precursor followed a mechanism analogous to the nanoscale “Kirkendall Effect”. After heating in NH3, the CoO-Mo2N hollow heterostructure was obtained. The Mo species in the precursor played an important role in maintaining the morphology under nitridation treatment. The hollow structure is favorable for contact and diffusion of electrolyte with (in) catalysts, while the CoO in CoO-Mo2N is favorable for the dissociation of water. Both promote the HER. Under optimized conditions, the hollow catalyst exhibited good HER performance with an overpotential of 65 mV at 10 mA cm−2 in 1 M KOH. The performance is better than that of many nonprecious metal-based catalysts. An electrolyzer composed of CoO-Mo2N heterojunctions as the cathode and NiFe-LDH as the anode can be driven by a solar cell to achieve effective overall water splitting. The adjudication of MOFs makes the route promising for the design of robust catalysts for advanced application. A fabrication technique that turns metal compounds into tiny hollow crystals may enable economical energy generation from water. Hydrogen fuel production from the electrochemical splitting of water molecules is often restricted by the need for expensive platinum catalysts. Chungui Tian, Honggang Fu and colleagues from Heilongjiang University in Harbin, China, now report that mixtures of transition metal compounds can split water using small amounts of electricity harvested from solar cells. Using a metal–organic nanocrystal as a template, the team adjusted heating conditions until the template’s inner core dissolved, leaving behind hollow shells containing cobalt–molybdenum oxides. The resulting CoO-Mo2N hollow heterostructure from the nitridaton of oxide precursor provides electronic properties favorable for hydrogen evolution reactions, while ensuring easy transport of reactants and products from catalytic sites. The Mo2N-CoO hollow heterostructures are designed synthesis by the nitridation of a hollow Co-Mo-O precursor from controllable reaction of ZIF-67 (Co source and template) with Na2MoO4 (Mo source and OH- source). The catalyst exhibits good HER performance with an overpotential of 65 mV at 10 mA cm−2 benefited from the combined virtues of hollow structure and heterojunction. The adjudication of MOFs makes current route promising toward the design of the transition metal-based catalyst for catalytic application.


Introduction
Electrocatalytic water splitting provides a clean method for producing hydrogen energy 1 . It is especially promising for converting low-voltage electricity from solar-energy cells (or wind) into easily used H 2 energy, which is very important to promote the utilization of sustainable energy. Due to the "uphill" reaction process, the hydrogen evolution reaction (HER) should be driven by an additional potential (overpotential), which results in high energy consumption and high equipment requirements 2 . Pt can reduce the overpotential effectively but suffers from high cost and scarcity 3 . Consequently, the development of nonprecious metal-based catalysts has been a topic in recent years. A variety of catalysts (sulfides, carbides and phosphides) have been successfully developed [4][5][6][7][8] . By tuning the components and microstructure, the kinetics for the HER can be largely improved, thus decreasing the overpotential 9 . It should be noted that effective catalysts should have good electrical conductivity, easy contact with the electrolyte and abundant active sites for the adsorption and activation of reactants 10 . A careful regulation of structure should be performed to produce effective HER catalysts 11 .
The transition metal nitrides (Mo 2 N, WN, etc.) are an important class of materials in catalytic fields 12,13 . Their good conductivity and Pt-like characteristics are favorable for use as HER catalysts 14 . The theoretical calculation indicated that hydrogen can easily adsorb on nitrides, thereby facilitating the HER 15 . The Gibbs free energy (ΔG) for adsorption/adsorption of reaction species on nitrides can be reduced in comparison with that for corresponding precursor metals 16 . Although, single metal nitrides (WN or Mo 2 N) have shown low activity for the HER 17 . Fortunately, it has been shown that the performance of nitrides can be improved by doping, tuning the structure or constructing heterojunctions 18 . The activity of WN can be improved by a P-modification method. By constructing a defective 2D sheet, the initial overpotential of molybdenum nitride can be reduced to <100 mV 19 . Furthermore, the heterostructure of Mo 2 N with Mo 2 C can provide an active interface to promote catalytic performance. It was also shown that Mo 2 C/Mo 2 N synthesized by a urea-based route has an overpotential of 176 mV at 10 mA cm −2 (ref. 20 ).
The above studies show the importance of adjusting the composition and structure of catalysts for effective HER performance 21 . The hollow structure has the advantages of easy contact with the electrolyte and the transmission and diffusion of reagents [22][23][24] . The masstransfer process that can affect diffusion impedance is an important factor for electrocatalysis 25 . In addition, heterojunctions containing late transition metal oxides or hydroxides are promising for water splitting because of their ability to promote the dissociation of water in alkaline systems 26 . Therefore, it is expected that suitable catalysts for the HER can be obtained by combining the advantages of hollow structures and heterostructures composed of Mo 2 N and oxides. Nitrides are usually synthesized by heat treatment of the corresponding oxides at high temperature 27 . To obtain hollow nitrides, it is necessary to construct a suitable precursor that cannot collapse during the heating process. Metal-organic frameworks (MOFs) are highly porous materials composed of metal ions and organic ligands 28,29 . The incorporation of metal ions makes MOFs promising precursors for functional nanomaterials 30 . Recently, we developed a "trapping" route with Mo X Co X C@C bimetallic carbides based on size matching of H 3 PMo 12 O 40 (PMo 12 cluster) (~1 nm) and cages (~1.16 nm) of ZIF-67 31 . In comparison with the PMo 12 cluster, MoO 4 2− can more easily combine with Co 2+ to form CoMoO 4 . Therefore, it is possible to produce a Co-Mo-based precursor by a replacement reaction of molybdate with ZIF-67. It is shown that a hollow CoMoO 4 -Co(OH) 2 precursor can be formed by the reaction of ZIF-67 with Na 2 MoO 4 32 . The formation of Co (OH) 2 can be ascribed to the reaction of OH − in Na 2 MoO 4 with Co 2+ at an elevated temperature. After elaborately tuning parameters during the synthesis of Co-Mo precursor and nitridation process, we have shown that CoO-Mo 2 N hollow heterojunctions can be obtained. Benefiting from the combined advantages of a hollow structure and heterojunctions, CoO-Mo 2 N showed good HER activity in 1 M KOH with an overpotential of 65 mV at 10 mA cm −2 . Although there are some reports on Mo 2 N-based catalysts, this report about Mo 2 N-based hollow heterojunctions is rare. This robust strategy offers an opportunity for designing highly efficient catalysts.

Synthesis of ZIF-67
The ZIF-67 polyhedra were synthesized according to a previous report with some modifications 31 . Typically, 12 mmol of Co(NO 3 ) 2 ·6H 2 O and 48 mmol of 2-methylimidazole were dissolved in two beakers containing 300 mL of methanol. Then, the Co(NO 3 ) 2 ·6H 2 O solution was slowly dropped into a 2-methylimidazole solution under continuous magnetic stirring. After the addition was completed, the mixture was stirred for an additional 6 h at room temperature. The purple precipitate was collected via filtration and washed with methanol repeatedly. The solids were dried at 60°C in a vacuum oven for 12 h to obtain the final ZIF-67 for subsequent use.

Synthesis of hollow CoMoO 4 -Co(OH) 2 precursor
The synthesis of CoMoO 4 -Co(OH) 2 is based on the modified reaction of ZIF-67 with Na 2 MoO 4 32 . Typically, ZIF-67 (0.3 mmol, 0.104 g) was dispersed in 40 mL of ethanol followed by sonication for 10 min to obtain a uniform dispersion. Na 2 MoO 4 ·2H 2 O (0.4 mmol, 0.10 g) was dissolved in 20 mL deionized water. The Na 2 MoO 4 ·2H 2 O solution was mixed with the ZIF-67 dispersion and then refluxed at 92°C for 2 h. After being centrifuged and washed with ethanol, the hollow CoMoO 4 -Co(OH) 2 precursor (Co-Mo-O-2h) was obtained and dried overnight at 60°C in a vacuum oven. Time-dependent experiments were performed (10 min, 1 h, 2 h and 3 h) to tune the structure of the precursor and study the formation process of the hollow structure.

Synthesis of the CoO-Mo 2 N hollow heterostructure
To synthesize the CoO-Mo 2 N hollow heterostructure, a porcelain boat with a Co-Mo-O-2h precursor was loaded in a quartz tube that was heated from room temperature to 600°C at a rate of 5°C min −1 and held at 600°C for 2 h under an NH 3 atmosphere (50 sccm) in a tube furnace. After naturally cooling to room temperature, the CoO-Mo 2 N hollow heterostructure (Co-Mo-N-2-600-2h) was obtained. For comparison, the heat treatment was performed at 500°C (Co-Mo-N-2-500-2h) and 700°C (Co-Mo-N-2-700-2h) for 2 h to study the effect of the calcination temperature. For comparison, the corresponding Co-based and Mo-based catalysts were also prepared. Typically, a porcelain boat loaded with Na 2 MoO 4 was put into a tube furnace and then heated to 600°C at a rate of 2°C min −1 and maintained at this temperature for 2 h under flowing NH 3 gas (50 sccm) (Mo-N-600-2h). For the Cobased catalyst (Co-N-600-2h), Na 2 MoO 4 was replaced by ZIF-67 while remaining consistent other conditions with the Mo-based catalyst. The sample and corresponding synthetic parameters are shown in Table S1.

Characterization
Scanning electron microscopy (SEM) was carried out on a Hitachi S-4800 instrument at an accelerating voltage of 5 kV. Transmission electron microscopy (TEM) and highresolution TEM (HRTEM) were performed on a JEM-F200 electron microscope (JEOL, Japan) with an acceleration voltage of 200 kV. X-ray diffraction (XRD) patterns were recorded using a Rigaku D/max-2600/PC X-ray diffractometer (XRD, Cu-Kα radiation). X-ray photon-electron spectroscopy (XPS) analysis was performed using a VG ESCALABMK II spectrophotometer with Mg-Kα radiation (1253.6 eV). The Brunauer-Emmett-Teller (BET) test was performed at 77 K with a Micromeritics Tristar II surface area and a porosimetry analyzer. The surface area was determined from desorption isotherms by the Barrett-Joyner-Halenda (BJH) and BET methods.

Electrochemical measurements
The electrochemical measurements were performed on a CHI 760E electrochemical workstation (CH Instruments Inc., China) in a 1 M KOH electrolyte with a typical three electrode configuration. A Hg/HgO electrode filled with 1 M KOH and a graphite rod were used as the reference electrode and counter electrode, respectively. The working electrode was prepared by mixing 5 mg of catalyst and 1 mg of carbon black dispersed in 3 μL of water/ethanol (v/v = 1:1). Then, 30 μL of Nafion was added to the mixture. The mixture was treated to obtain a dispersed slurry. The slurry was pasted on a piece of NF current collector (1.0 × 1.0 cm 2 ) and dried in a vacuum oven. Before coating, the NF was washed with acetone, HCl aqueous solution (2 M), deionized water and ethanol in sequence (immersion time: 15 min per solvent) 31 . Linear sweep voltammetry (LSV) was performed at a rate of 5 mV s −1 . The Tafel data were fitted according to the Tafel equation η = b log(j) + α (j: current density and b: Tafel slope). The scan rates of the cyclic voltammetry (CV) tests were varied from 20 to 200 mV s −1 . Chronoamperometry (CA) tests were carried out at −65 mV. The stability of the catalysts was evaluated by performing 1000 CV cycles at a scan rate of 100 mV s −1 . All potential mentioned is referred to the reversible hydrogen electrode (RHE), for which the Nernst equation was used: Hg/HgO . The faradaic efficiency (FE) for the HER was calculated by comparing the experimentally determined and theoretically calculated amounts of H 2 . In the test, potentiostatic electrolysis was performed at 100 mA cm −2 for 60 min in an airtight H-type electrolytic cell. The amount of gas generated was analyzed by gas chromatography (GC). The theoretical H 2 amount was calculated as follows: where n(H 2 ) is the number of moles of hydrogen produced, Q is the charge passed through the electrodes, F is the Faradaic constant (96,485 C mol −1 ), and n is the number of electrons transferred during water splitting (for HER, n is 2).

Results and discussion
Structural characterization Figure 1 illustrates the process for the synthesis of CoO-Mo 2 N hollow heterojunctions. The ZIF-67 rhombohedral dodecahedron was used as the Co source and template based on its regular morphology with plentiful Co. ZIF-67 was mixed with Na 2 MoO 4 solution and treated under  (Fig. 2a) 32 . The XRD results indicate the formation of CoMoO 4 -Co (OH) 2 by the reaction of ZIF-67 and Na 2 MoO 4 . The transformation was further verified by SEM and TEM analyses. As revealed in Fig. 2b, ZIF-67 shows rhombic dodecahedra with a uniform size of~500 nm and a smooth surface. The corresponding TEM image (Fig. 2c) reveals the solid nature and smooth surface of ZIF-67, which is consistent with previous reports [31][32][33] . After reacting with Na 2 MoO 4 for 2 h, uniform rhombic dodecahedra with a slightly rough surface are formed (Fig. 2d). Some broken polyhedral observed in the SEM image demonstrate the formation of hollow structures. In particular, the TEM images show the formation of a uniform hollow structure, which is obviously different from the solid structure of the ZIF-67 (Fig. 2e). The transparent characterization indicates the thin characteristics of the shell. The thickness of the shell is~10 nm, as estimated from the magnified TEM image (Fig. 2f). The scanning transmission electron microscopy (STEM) image and corresponding energy dispersive X-ray (EDX) mapping show homogeneous distributions of Co, Mo and O throughout the hollow structure ( Fig. 2g-g3). All characterization results demonstrate the transformation of ZIF-67 into hollow CoMoO 4 -Co(OH) 2 .
The conversion of ZIF-67 into CoMoO 4 -Co(OH) 2 was also verified by XPS. The XPS wide spectrum in Fig. S2 shows the presence of Co, Mo, O, N and C in the precursor. Further information can be gained from the highresolution XPS spectra. In Fig. S3a, the peaks located at 781.7 eV and 797.5 eV can be assigned to Co 2p 3/2 and Co 2p 1/2 of Co(OH) 2 , which are accompanied by shakeup satellite peaks (shortened as sat.) at 802.3 eV and 786.7 eV 34 . The binding energies (BEs) at approximately 780.1 eV and 795.9 eV are indexed to Co 3+ (ref. 35 ). The formation of Co 3+ should be from slight oxidation of Co 2+ in Co(OH) 2 . The Mo 3d spectrum can be divided into four peaks (Fig. S3b). The peaks at BEs of 231.8 eV and 234.9 eV are indexed to Mo 3d 5/2 and Mo 3d 3/2 , respectively, of Mo 6+ in CoMoO 4 (ref. 36 ). The BEs at approximately 230.9 eV and 233.7 eV are indexed to Mo 4+ , which is from the slight reduction of Mo 6+ (ref. 36 ). The high-resolution XPS spectrum in the O 1s region (Fig.  S3c) shows three peaks at 532.3, 531.3 and 530.5 eV, corresponding to the metal-O, hydroxy oxygen (O H ) and adsorbed water (H 2 O), respectively. The high-resolution N 1s peak (Fig. S3d) shows three deconvoluted peaks at 400.6, 399.4 and 379.5 eV, which can be assigned to graphitic N, pyrrolic N and pyridinic N, respectively. The presence of N can be ascribed to the slight residue of the imidazole ligand after the conversion reaction.
Time-dependent experiments were performed (10 min, 1 h, 2 h and 3 h) to tune the structure of the precursor and to study the formation process of the hollow structure. To examine the initial process, the reaction of ZIF-67 with Na 2 MoO 4 was performed for 10 min. The yolk-shell structure was composed of a thin outer shell and an inner solid particle (Fig. S4a). The inner solid particles should be unreacted ZIF-67. By prolonging the reaction time to 1 h, we can see the formation of a hollow structure (Fig.  S4b). The structure is similar to that from the reaction for 2 h but with a relatively smooth surface. The hollow structure remains upon increasing the reaction time to 2 h (as shown in Fig. 2) and even 3 h but with a slightly increased surface roughness (Fig. S4c). Previous reports have shown that the hollow structure can be formed through hard-templated, soft-templated and selftemplated routes 22 . The use of presynthesized solid templates containing at least one of the elements of the final nanoshell usually results in a mechanism analogous to the nanoscale Kirkendall effect 32,37 . The process first includes the reaction of the elements (A) in the template with the external reagents (B) to produce a layer of shell materials (AB). The formed shell will hinder further direct reaction of A and B. Then, A and B react by diffusion through the shell. The difference in the diffusion rate can result in the formation of voids (hollow structure). The timedependent experiment indicates that the formation of hollow precursors should be ascribed to a nanoscale "Kirkendall effect" 38,39 . The MoO 4 2− from Na 2 MoO 4 can react with Co 2+ to give CoMoO 4 . Heating can promote the hydrolysis of Na 2 MoO 4 (a weak acid-strong alkali salt, pH of~9 at room temperature) to provide more OH − , thus forming Co(OH) 2 . As the reaction continues, the outward diffusion of Co 2+ from ZIF-67 leads to the formation of voids in the CoMoO 4 -Co(OH) 2 shell along with the consumption of ZIF-67 cores. With completion of the dissolution of the ZIF-67 cores, a well-defined hollow CoMoO 4 -Co(OH) 2 structure finally formed. The use of Na 2 MoO 4 is very important because it not only provides MoO 4 2− for the formation of CoMoO 4 but also provides an alkaline environment for the formation of Co(OH) 2 . The direct reflux of ZIF-67 without Na 2 MoO 4 only results in a slight morphology change of ZIF-67. Some thin sheets (Co(OH) 2 ) can be observed on the surface of ZIF-67 (Fig.  S4d). The low number of sheets is due to the minimal OH − in water relative to that in the Na 2 MoO 4 solution. Nevertheless, the phenomena show the importance of OH − replacement for the formation of the thin hollow structure.
The simultaneous incorporation of Co and Mo species and a hollow structure makes CoMoO 4 -Co(OH) 2 suitable as a precursor for hollow Co-Mo-based catalysts. Based on the promising characteristics of nitrides, we treated the precursor under NH 3 atmosphere at 600°C for 2 h. After the treatment, the XRD pattern shows obvious differences from those of the original precursor and ZIF-67. Typically, the XRD pattern (Fig. 3a) shows the peaks located at 37.7°and 43.2°, which are attributed to the (112) and (200) planes, respectively, of Mo 2 N (JCPDS no. 25-1368), while the peaks located at 36.5°, 42.4°and 61.5°can be attributed to the (111), (200) and (220) planes, respectively, of CoO (JCPDS no. 43-1004). The SEM image (Fig. 3b) shows the well-retained polyhedron morphology after heating. Some broken polyhedra can be observed, indicating the formation of the hollow structure. In comparison with CoMoO 4 -Co(OH) 2 , the surfaces are further roughened and showed many small particles. The TEM images (Fig. 3c-f) solidly indicate the formation of hollow structures composed of small particles. The HRTEM (Fig. 3f) image shows two sets of lattice fringes of 0.20 nm and 0.25 nm in two adjacent particles, corresponding to the (200) plane of Mo 2 N and the (111) plane of CoO, respectively. To provide additional information about the hollow structure, EDS elemental mapping was performed. Mo, N, Co and O were uniformly distributed through the hollow polyhedron (Fig. 3g-g4), demonstrating the formation of hollow CoO-Mo 2 N heterojunctions. The morphological change can be attributed to the phase change from CoMoO 4 -Co(OH) 2 to CoO-Mo 2 N heterojunctions. The well-retained hollow structure after heating should be relative to the presence of Mo. The heat treatment of ZIF-67 results in severe damage to the morphology (the formation of irregular particles) (Fig.  S5a). Furthermore, heating of the precursor prepared by the reaction of ZIF-67 and Na 2 MoO 4 for 10 min also gives particles with irregular morphology (Fig. S5b), and the yolk-shell structure in the precursor (Fig. S4a) cannot be observed. In contrast, heating the precursor at reflux for 1 h (Co-Mo-O-1 h) and 3 h (Co-Mo-O-3 h) can result in the formation of a well-retained hollow structure (Fig.  S5c, d). Previous reports have shown that heat treatment of Mo-based MOFs (NENU-5) under NH 3 atmosphere can produce a catalyst with a morphology similar to that of the original NENU-5 (ref. 40 ). In contrast, the treatment of Co-based ZIF (as shown here) can only form irregular particles. The combination of our results and previous reports show the important role of Mo species in maintaining the morphology under nitridation treatment 41 .
To further provide information about the composition of CoO-Mo 2 N heterojunctions, an XPS test was conducted. The wide spectrum shows the presence of Co, Mo, O and N in the sample (Fig. S6). Figure 4a shows the high-resolution spectrum of Co. The peaks at approximately 781.2 eV and 796.7 eV correspond to Co 2p 3/2 and Co 2p 1/2 , respectively, of Co 2+ in CoO 42 , which are accompanied by a couple of satellite peaks located at 802.6 eV and 786 eV. The BEs at~779.5 eV can be attributed to Co in Co-N 43 . The peaks of Co-N are less intense than those of Co-O, indicating the predominant phase is CoO. The Mo 3d spectrum can be divided into six peaks, as shown in Fig. 4b. The peaks at BEs of 230.2 eV and 233.9 eV are assigned to Mo 3d 5/2 and Mo 3d 3/2 , respectively, of Mo 4+ in MoO 2 . The peaks at BEs of 232.0 eV and 235.2 eV are attributed to Mo 3d 5/2 and Mo 3d 3/2 , respectively, of Mo 6+44 . The peaks located at 229.1 eV and 232.8 eV are assigned as Mo-N bonds in Mo 2 N 18 . The presence of oxides is related to surface oxidation of nitrides, which can be observed in many Mobased carbides and nitrides 45 . Additionally, the O 1s spectrum (Fig. 4c) can be subdivided into two peaks at  16 . The high-resolution N 1s XPS (Fig. 4d) shows five peaks located at 402.1, 400.1, 398.9, 397.5 and 395.6 eV. The peaks located at 402.1, 398.9 and 379.5 eV can be assigned to graphitic N, pyrrolic N and pyridinic N, respectively. The peak at 395.6 eV is the typical BE of N to Mo, which verifies the formation of a nitride 46 48 ). Ni-doped Co-Co 2 N is synthesized by first annealing MOFs to form oxide precursors (NiCo 2 O 4 ) followed by ammonization 49 . In the present work, the CoMoO 4 -Co(OH) 2 hollow polyhedra are the precursors for the synthesis of CoO-Mo 2 N. We speculate that the formation of CoO, but not Co 4 N, is related to the difference in the precursor (i.e. Co(OH) 2 vs Co 3 O 4 ). The heating temperature in the ammonia atmosphere has some effect on the final structure. The sample from calcination at lower temperature (500°C) exhibits a hollow structure and shows a small particle size (Fig. S7a). At the higher temperature (700°C), the polyhedral morphology also remains. However, the particles in the hollow structure have a larger size than those from heating at low temperature. Many broken and fragmented polyhedra can be observed (Fig. S7b). The difference in structure should affect the application performance.  superior activity for the HER 53 . The combined advantages of hollow structures and heterostructures make CoO-Mo 2 N a promising catalyst for the HER. To implement this proof-concept, we evaluated the HER activity of the hollow CoO-Mo 2 N structure. The catalyst was pasted on a piece of a NF current collector and then tested for HER activity in a 1 M KOH solution. The LSV test for Co-Mo-N-2-600-2h was performed first (Fig. S8a). To minimize the capacitive current, a slow scan rate of 5 mV s −1 was applied to obtain all LSV curves. For comparison, Co-Mo-N-2-500-2h and Co-Mo-N-2-700-2h were also tested. The current densities in the whole potential range follow the order of CoMo-N-600-2h > CoMo-N-700-2h > CoMo-N-500-2h. The CoMo-N-600-2h catalyst delivers higher HER activity in 1 M KOH and only requires overpotentials of 65 mV and 230 mV to reach current densities of 10 mA cm −2 (η 10 ) and 100 mA cm −2 (η 100 ), respectively. The values for Co-Mo-N-2-500-2h and Co-Mo-N-2-700-2h are approximately 126 mV (η 10 ) and 321 mV (η 100 ) and 89 mV (η 10 ) and 302 mV (η 100 ), respectively (Fig. S8b). Co-Mo-N-600-2h shows the closest value to Pt/C (Fig. 5a). Especially, at a higher current density (>100 mA cm −2 ), the performance of Co-Mo-N-2-600-2h can even exceed that of the Pt/C catalyst (Fig. 5a, b). Notably, all Co-Mo-N samples show superior activity to Mo-N-600-2h and Co-N-600-2h (Fig. 5a, XRD patterns indicate the presence of Co 5.47 N for Co-N-600-2h, and MoN and Mo 2 N for Mo-N-600-2h, Fig. S9). Co-N-600-2h and Mo-N-2-600-2h require overpotentials of 207 mV and 240 mV to reach η 10 and 378 mV and 410 mV to reach η 100 , respectively (Fig. 5b). In addition, the NF shows an η 10 of 250 mV and η 100 of 485 mV (Fig. S10). The values are much higher than those of NF supported by CoO-Mo 2 N heterojunctions (CoO-Mo 2 N/NF). This indicates that good activity of CoO-Mo 2 N/NF is mainly from the CoO-Mo 2 N heterojunctions. The better activity of Co-Mo-N-2-600-2h than that of Co-Mo-N-2-500-2h should be related to the improved crystallinity of the nitrides in the early sample. The formation of nitrides (Mo 2 N) usually requires a high temperature of~600°C 54 . Co-Mo-N-2-500-2h from heat treatment at 500°C can result in incomplete conversion of Mo in the Co-Mo-O precursor into the corresponding nitrides, thus leading to low activity for the HER. On the other hand, a higher temperature during the heat treatment can result in the growth of particles, which is also adverse for enhancing the activity. The better activity of Co-Mo-N-2-700-2h than Co-Mo-N-2-500-2h indicates the promotion role of nitrides on the HER.

Electrochemical performance
The catalyst was obtained by heating the Co-Mo-O precursor. Therefore, the structure of the precursor can affect the final structure of the Co-Mo-N catalyst. Figure  S11a, b shows the electrochemical performance of the catalysts derived from the precursor by reflux for 1 h (Co-Mo-N-1-600-2h) and 3 h (Co-Mo-N-3-600-2h). The LSV curves for Co-Mo-N-1-600-2h and Co-Mo-N-3-600-2h show that the η 10 is 82 and 92 mV, respectively. The value is higher than that for Co-Mo-N-2-600-2h. The low activity of Co-Mo-N-1-600-2h should be related to the low amount of Mo in the catalyst and the dense surface (Fig. S5). A long reaction time of 3 h can result in breakage of the shell (Fig. S5), which damages the final hollow catalysts and thus lowers the activity.
To gain further insights into the HER catalytic kinetics, the Tafel plots were created. Co-Mo-N-2-600-2h shows a small Tafel slope of 72 mV dec −1 . The values for Co-Mo-N-2-500-2h and Co-Mo-N-2-700-2h are approximately 82 and 79 mV dec −1 , respectively. The lower Tafel slope of Co-Mo-N-2-600-2h indicates its high intrinsic HER catalytic kinetics. The value is also lower than that of Co-Mo-N-1-600-2h (90 mV dec −1 ) and Co-Mo-N-3-600-2h (120 mV dec −1 ) (Fig. S11c). In particular, Co-Mo-N-2-600-2h showed a lower Tafel slope than Co-N-600-2h (98 mV dec −1 ) and Mo-N-600-2h (101 mV dec −1 ) (Fig.  5c). The improved performance of Co-Mo-N-2-600-2h compared to that of Co-N-600-2h and Mo-N-600-2h is related to its hollow structure, which can produce a large accessible surface area, as shown by the N 2 adsorption-desorption isotherm tests (Fig. S13). The specific surface area (S BET ) of the CoO-Mo 2 N hollow heterojunction (Co-Mo-N-2-600-2h) is~56.23 m 2 g −1 . However, the S BET of Co-N-600-2h and Mo-N-600-2h is 4.04 and 26.31 m 2 g −1 , respectively. The large S BET of the CoO-Mo 2 N hollow heterojunction can promote fast mass transport, which is considered beneficial for electrocatalysis. The performance is better than that of the control samples (Table S2) and many reported Mo-based catalysts (Table S3). The good performance of the CoO-Mo 2 N hollow heterojunction is related to the presence of CoO promoting the dissociation of water in alkaline systems. To support this point, we pasted the catalysts on carbon cloth and then tested the HER activity in 0.5 M H 2 SO 4 . The η 10 for Co-N-600-2h, CoO-Mo 2 N and Mo-N-600-2h is 153, 208, and 271 mV, and η 100 is 530, 493, and 688 mV, respectively. CoO-Mo 2 N did not show a larger advantage over the corresponding Co-N-600-2h and Mo-N-600-2h in acidic media. The results indicate the promoting role of CoO for the HER in alkaline media (Fig. S12). Electrochemical impedance spectroscopy (EIS) tests (Fig. 5d) were conducted to explore the HER kinetics. The charge transfer resistance (R ct ) of the CoO-Mo 2 N hollow heterojunction was measured to be 21 Ω, much lower than that of Co-N-600-2h (46 Ω) and Mo-N-600-2h (122 Ω). The results suggest the fast chargetransfer ability and mass-transfer ability of hollow heterojunctions. There is a typical interaction between different compositions of heterojunction materials, which is largely helpful in promoting catalytic ability. Qiao et al.
have shown that the enhanced electrocatalytic performance of 110 mV at 10 mA cm −2 for the HER benefited from the interaction of different components in heterojunctions 11 . We have also found the enhancing influence of constructing heterojunctions (Mo 2 N-Mo 2 C, Ni-Vbased interstitial compound) on the catalytic performance 16 . To provide insight into the interaction between Mo 2 N and CoO, we synthesized molybdenum nitride and CoO (see Fig. S14   electrons is helpful in promoting the catalytic performance for the HER. The low Tafel slope implies fast reaction kinetics on Co-Mo-N-2-600-2h due to more active sites, which can be supported by the analysis of the electrochemical active surface area (ECSA) reflected by double-layer capacitance (C dl ). The C dl is proportional to the active surface area (ECSA), which was evaluated based on the CV curves at different scan rates (Figs. 5e, S15). As shown in Fig. 5e, the C dl value of Co-Mo-N-2-600-2h is~26.5 mF cm −2 . This value is the largest among all the catalysts, implying the existence of more active sites in the hollow heterostructure. Moreover, the Faradic efficiency of Co-Mo-N-2-600-2h was obtained by GC measurement of evolved H 2 . The catalyst gives ∼100% Faradic efficiency for the HER (Fig. S16). In an accelerated degradation test, there was no obvious loss in the current density after 1000 cycles (Fig. S17). There was no obvious change in morphology after the stability test, as shown by the SEM and TEM images (Fig. S18a, b). In addition, XRD shows similar peaks before and after the stability test (Fig. S18c).
All results indicate the good stability of the CoO-Mo 2 N heterojunctions for the HER. As illustrated in Fig. 5f, there was no significant increase in the overpotential during continuous electrolysis for 48 h, verifying its good stability. The good HER activity and stability of the CoO-Mo 2 N heterostructures indicate their potential for water hydrolysis. A series of tests showed the good activity of CoMo-N-600-2h with low η 10 (65 mV) and η 100 (230 mV), low Tafel slope (72 mV dec −1 ), and large C dl value (26.5 mF cm −2 ). The good activity is related to the special structure of the CoO-Mo 2 N heterojunctions. (1) The hollow structure is favorable for contact of the electrolyte with the catalyst and the transmission/diffusion of reagents, thus decreasing the diffusion impedance; (2) the interaction (the transfer of electrons) between Mo 2 N and CoO is favorable for forming an active interface for the adsorption and activation of reactants; and (3) the presence of CoO is helpful for promoting the dissociation of water in alkaline systems. All of them are conductive to promoting the HER.
Inspired by the outstanding HER performance of the CoN-Mo 2 N heterojunction, we assembled an overall water splitting device using Co-Mo-N-2-600-2h as the cathode and NiFe-LDH as the anode in 1 M KOH (Fig. 6a). NiFe-LDH was synthesized based on a previous report 55 . CoO-Mo 2 N (−)//NiFe-LDH (+) required a Fig. 6 The performance of the catalysts for overall water splitting. a LSV polarization curve in 1 M KOH for overall water splitting; b the stability test performed at a constant voltage of 1.57 V; c photo of a water-splitting device driven by a solar cell (ca. 1.57 V); d photo of the evolution of H 2 and O 2 . Co-Mo-N-2-600-2h was used as the cathode, and NiFe-LDH was used as the anode. voltage of 1.57 V to drive the current density of 10 mA cm −2 . More importantly, the electrolyzer showed good stability at a constant voltage of 1.57 V. Only a slight current change was observed after 48 h of operation (Fig. 6b). Additionally, the electrolyzer can be driven by a solar cell (1.57 V) to realize substantial water splitting (Fig. 6c, d). This shows the promising potential of the catalyst to actually store indirect energy.

Conclusions
In summary, we have demonstrated a robust route toward CoO-Mo 2 N hollow heterostructures by nitridation of a hollow Co-Mo-O precursor from the controllable reaction of ZIF-67 (Co source and template) with Na 2 MoO 4 (Mo source and OH − source). A series of experiments showed the tunability of the Co-Mo-O structure, and the formation followed a mechanism analogous to that of the nanoscale "Kirkendall effect". The hollow CoO-Mo 2 N structure can be obtained by elaborately tuning the synthetic parameters during the nitridation process. The presence of Mo is helpful in preserving the original morphology of the precursor after nitridation. The hollow CoO-Mo 2 N exhibited good HER performance with an overpotential of 65 mV at 10 mA cm −2 in 1 M KOH and benefited from the combined advantages of the hollow structure and heterojunction. The performance is better than that of many nonprecious metal catalysts. The adjudication of MOFs makes the current route promising for the design of transition metal-based catalysts for catalytic applications.