Shortage of ammonia (NH₃), which is used as the raw material for chemical fertilizer, chemical pharmaceuticals, and zero-carbon emission fuel, has become an urgent energy problem with the continuous growth of the population. Fixing N₂ in the atmosphere to NH₃ is an important method to realize nitrogen storage and the conversion of Earth’s nitrogen cycle (Service, 2014; Brown et al., 2016). Up to now, the Haber–Bosch method (N₂ + 3H₂ → 2NH₃) has been used to synthesize NH₃ in the industry, which works at high temperatures and pressure with high energy consumption (Van Der Ham et al., 2014). Most raw materials of hydrogen and energy consumption are obtained through the steam reformation of fossil fuels, while a large amount of CO₂ is also produced (Kyriakou et al., 2020). Therefore, the more economical, environmental, and sustainable method of producing NH₃ is in great demand.

Electrochemical reduction is a prospective method for artificial nitrogen fixation (Amar et al., 2011; Costentin et al., 2013; She et al., 2017; Guo et al., 2018). However, the N₂ reduction reaction (NRR) requires breaking the inert molecular structure of N₂, which requires sufficient energy to break the N≡N bond (∼941 kJ mol⁻¹) (Han et al., 2021). Noble metal catalysts (Ru, Au, Ag, and Rh) have shown to have superior NRR activity, but the high cost and rareness limit their commercial applications (Aioub et al., 2017; Bao et al., 2017; Nazemi et al., 2018; Liu et al., 2019; Tao et al., 2019). The search for inexpensive, earth-abundant materials with high activity and excellent stability has recently attracted significant research interest and has become an important pursuit toward ammonia economy. In recent years, transition metal has attracted extensive attention as NRR catalysts because of their high activity, abundant reserves, and low price. It has been reported that molybdenum-based catalysts have been designed and applied to artificial nitrogen fixation (Banerjee et al., 2015; Sun et al., 2017; Wickramasinghe et al., 2017; Yang et al., 2017; Han et al., 2018; Li et al., 2019; Ou et al., 2019). Molybdenum nitride has attracted much attention due to its platinum-like properties. Specifically, Mo₂N is a typical interstitial compound in which the metal atoms express a tight stacking pattern to stabilize the lattice structure, where the top element is mainly a metal atom on the surface. The nearest metal ions with short distances will provide the possibility to form metallic bonds to facilitate electron transport in the connected channels. MoN and Mo₂N nanosheets also have an electrocatalytic activity of N₂ reduction, with the FE of 1.15% and 4.5%, respectively (Ren et al., 2018; Zhang et al., 2018). Although molybdenum-based catalysts have excellent performance, they still suffer from weak FE efficiency. Thus, it is attractive to search for new Mo-based catalysts with enhanced FE performance.

In catalytic and electrocatalytic systems, uniform dispersion of active sites on the surface of porous structures ensures the high activity of porous materials (Zhang et al., 2021; Lu et al., 2022). Single crystals with almost no grain boundary reduce the resistance of charge transmission between grain boundaries. Therefore, porous single crystals with anisotropy and a long-range structural order and structural coherence combined with a large specific surface area provide unsaturated nitrogen coordination on the surface of metal nitride, and its internal structures exhibit regularity and periodicity in three-dimensional space. Due to the porous single-crystal structure, the electrocatalytic properties are also improved (Wu et al., 2019; Zhang et al., 2019; Lin et al., 2020, 2022; Xiao and Xie, 2022). Different from the conventionally used nano-catalyst materials such as nanosheets or nanorods, which are smooth in surface and poor in single crystallinity, porous single crystals have a large amount of nanopores and good crystallinity which increase the active site and improve the stability during the reaction, thus improving the catalytic performance of the catalyst.

In this paper, we prepare the micron-scale porous Mo₂N single crystal (PSC Mo₂N) and polycrystalline Mo₂N (PPC Mo₂N) cubes by the lattice reconstruction strategy and investigate their catalytic performance in the electrocatalytic reduction of N₂ to NH₃. PSC Mo₂N can provide a large surface area for chemical adsorption, and the internal interconnected pores and channels are conducive to the diffusion of ions in the electrolyte. PSC Mo₂N exhibits ultra-high yield of NH₃ and FE, which greatly enhances the catalytic performance of NRR.

Figure 1A shows the XRD patterns of MoO₃, PSC Mo₂N, and PPC Mo₂N. The XRD patterns of MoO₃ and Mo₂N single crystals conform to the JCPDS standard cards 75-0512 and 75-1150, respectively. The diffraction peaks are sharp and narrow, indicating that molybdenum nitride has excellent crystallinity. From a polycrystalline versus a single crystal perspective, molybdenum nitride single crystals tend to be [111] oriented. The inset in Figure 1A is ball-and-stick model of the crystal structure of Mo₂N. Supplementary Figure S1 illustrates the growth mechanism for obtaining micron-scale metal nitride cubes from metal oxides. When the intrinsic oxygen atoms in the MoO₃ lattice are heated in an ammonia atmosphere, the oxygen atoms gradually leave the intrinsic position in the lattice and slowly disappear. The nitrogen atom in ammonia gradually enters the position before the oxygen atom. Due to the different sizes of oxygen and nitrogen atoms, the lattice configuration also changes, and then, the molybdenum nitride skeleton is gradually formed. By changing the ammonia flow rate and pressure, the molybdenum nitride lattice is reconstructed to form a porous Mo₂N cube.

To explore the relationship between the structure and properties of porous single-crystal molybdenum nitride, we perform XPS tests. The test results are shown in Figure 1B; the results specifically show that the characteristic peaks at 229.8 and 232.3 eV are caused by Mo 3d_5/2 and Mo 3d_3/2 (Ozkan et al., 1997; Afanasiev, 2002; Liu et al., 2013). In addition, the combining energies at 228.9 and 235.9 eV are consistent with Mo^δ+ and Mo⁶⁺, respectively, which is ascribed to the surface oxidation of Mo₂N (Yoon et al., 2012). The peak at 397.81 eV is the N 1 s of Mo₂N, while the peak value of 394.77 eV corresponds to the shoulder of Mo3p_3/2 with Mo₂N, indicating the presence of Mo–N bonds (Ozkan et al., 1997; Afanasiev, 2002; Yoon et al., 2012; Liu et al., 2013). To further confirm the presence of chemical elements in Mo₂N, energy dispersive X-ray (EDX) and its mapping are used for further identification (Figure 1C). The results indicate that only two chemical elements (Mo and N) can be detected in the final product Mo₂N, and no other impurity phases are observed. Moreover, the chemical elements Mo and N are uniformly distributed in the sample, which confirms well the complete transformation process of the non-porous MoO₃ single crystal.

We observe the changes in the microscopic morphology of the crystals during the growth and preparation processes by SEM, and Figure 1D shows the MoO₃ cubes of about 3–10 μm obtained after calcination of the precursor MoO₃-PEG. Figures 1E,F display the microcube scanning diagram of PSC Mo₂N and PPC Mo₂N formed by ammoniation of MoO₃, which have the same morphology before and after calcination from the microscopic scale. In addition, Figure 1E reveals the magnified morphology, and it can be seen that the pore channels are uniformly connected roughly around 50 nm, while the polycrystalline surface is in a sintered state with some cracks. Therefore, we speculate that the specific surface area of single crystals should be larger than that of polycrystals. We use a lattice reconstruction strategy to grow porous molybdenum nitride from non-porous molybdenum nitride, whose BET results (Supplementary Figure S2) indicate that the BET-specific surface areas of PSC Mo₂N and PPC Mo₂N are 14.347 and 7.676 m² g⁻¹, respectively. The larger specific surface area has more active sites, which is consistent with our subsequent test results (PSC Mo₂N > PPC Mo₂N).

For the purpose of obtaining further information about the microstructure and crystal orientation of the synthesized material, TEM and HRTEM are used for characterization and analysis. Figure 2A demonstrates that the precursor MoO₃ has good single crystallinity, where SAED (Figure 2B) corresponds to the (002) plane of MoO₃. HRTEM (Figure 2C) indicates a crystal surface spacing of 0.198 nm that corresponds to the (002) face of MoO₃, which is in agreement with both results. Mo₂N retains the morphology and structure of the precursor, irrespective of it being a single crystal (Figures 1E, 2D) or polycrystalline (Figures 1F, 2G). Figure 2F taken with HRTEM from a single Mo₂N cube (Figure 2I) reveals highly resolved lattice stripes between the crystal planes with a crystalline spacing of 0.238 nm associated with the (111) plane of Mo₂N, which is consistent with SAED (Figure 2E). This further establishes that the sample has a defined crystal orientation. However, in Figures 2H,I, the SAED is tending to a ring-like pattern, and there are no distinct and regular lattice stripes, which suggests that the sample is polycrystalline.

NRR is a prototypical reduction process by hydrogenation. As shown in Figure 3A, N₂ bubbles to the cathode surface, and H⁺ can be transferred through the electrolyte of 0.1 M HCl to react with the cathode Mo₂N to form NH₃. At first, we test the electrochemical performance of the sample and the catalytic activity of N₂. The LSV curves for the reduction reaction (NRR) are obtained by saturating with N₂ and Ar in 0.1 M HCl electrolyte (Figure 3B), and it is clear that the current density of N₂ increases slightly compared to Ar in the saturated electrolyte, which indicates the potential NRR activity of PSC Mo₂N. Apart from that, the AC impedance method is used to verify the electron transfer resistance of molybdenum nitride by fitting the EIS data on the sample to an equivalent circuit. The Nyquist curve (Figure 3C) shows that PSC Mo₂N has a smaller resistance than PPC Mo₂N and therefore has a higher current for the same voltage conditions applied. This indicates that the porous single crystal is conducive to carrier migration, which has better catalytic properties. All these results clearly indicate that the porous Mo₂N single crystals are more effective as electrocatalysts than polycrystals.

In the next step, we collect the products at different voltages at the same time and analyze their contents by cation chromatography. The results are shown in Figures 3D,E, which exhibit the NH₄ ⁺ yields and FEs of PSC Mo₂N and PPC Mo₂N at different voltages. Among them, the catalytic effect of the catalysts is best at −0.6 V, when the NH₄ ⁺ production rate of PPC Mo₂N is 77.78 μg h⁻¹ mg⁻¹ and the FE reached 0.6%, while the NH₄ ⁺ production rate of PSC Mo₂N reached 272.56 μg h⁻¹ mg⁻¹ and the FE reached 7.3%. Obviously, PSC Mo₂N implies extremely high catalytic activity compared to PPC Mo₂N. The NH₄ ⁺ yield and FE content increase with the increase in negative voltage; however, the yield of NH₄ ⁺ and FE decreases significantly when it passes −0.6 V vs. RHE. It is possible that the hydrogen evolution reaction (HER) competes fiercely at the Mo₂N electrode to preferentially occupy the active site with a large amount of free H and limit the effective trapping of N₂, hindering the forward progress of the NRR reaction. Compared with other water-based catalysts, PSC Mo₂N has higher ammonia yield and FE. The specific performance of other catalysts is shown in Supplementary Table S1.

The effect of other impurities in the experiments is excluded by conducting control experiments on the precursor MoO₃ and carbon paper on the results. The results demonstrate that the NRR rate of Mo₂N in Figure 3F is much higher than that of the MoO₃ single crystal (26.21 μg h⁻¹ mg⁻¹), which implies that N plays an important role in the NRR and excludes the influence of surface oxidation on this reaction. At the same time, the bare CP shows weak NH₄ ⁺ production, which has been subtracted from the starting data to reduce the experimental error. PSC Mo₂N and PPC Mo₂N have similar morphology and dimensions, but they have different specific surface areas, pore size, and lattice arrangement, leading to different ammonia yields of both. This suggests porous single crystals with large specific surface area and lattice continuity can reduce the energy loss of electron transfer between crystals, which results in higher catalytic activity. Therefore, PSC Mo2N exhibits excellent catalytic performance, which is consistent with the experimental results.

Durability is an important credential for assessing NRR response. Figures 4A,B indicate that the relative current density of PSC Mo₂N and PPC Mo₂N catalysts reduced to ammonia at different potentials respectively is stable, which shows the excellent durability of catalysts. Moreover, as the cathodic potential increases, the current density slightly decreases and the HER process is enhanced, inhibiting the NRR reaction, which is also consistent with the decrease of NH₄ ⁺ production and FE. To further evaluate the stability of this catalytic process, we test the samples for 10 cycles at −0.6 V (Figures 4C,D), and the NH₄ ⁺ production rate and FEs have no significant changes, indicating that they have good stability for NRR. In addition, the effect of a different N₂ flow rate on the product is also tested in this electrocatalytic process. As shown in Figures 4E,F, the approximately constant yields of NH₄ ⁺ and FEs demonstrate that the N₂ diffusion rate has no obvious effect on the reaction, which ruled out the effect of the inlet concentration on this reaction. We perform some controlled experiments to verify the reliability of Mo₂N/CP for the electrochemical reduction of N₂ to NH₃. N₂ is led into the electrochemical reaction cell under open-circuit conditions, and saturated-Ar and saturated-N₂ gas flows are introduced separately for 2 h at −0.6 V vs. RHE, respectively. In summary, the influence of the external environment on the reaction during the experiment is excluded, indicating that the obtained NH₃ is derived from Mo₂N/CP.

We test the long-term electrolysis process, as shown in Figure 5A, and the current density electrochemical NRR process hardly changes significantly within 16 h, indicating the excellent performance of the Mo₂N catalyst. We perform a series of characterizations on the tested PSC Mo₂N such as XRD (Figure 5B), XPS (Figure 5C), SEM (Figures 5D,E), EDX (Supplementary Figure S4), and Mapping (Figure 5F). The test results clearly show that the phase composition, valence state, and elemental distribution of the catalyst itself have not changed significantly compared with those before the reaction, indicating that no other impurity phases are generated in the catalyst during the reaction. Further from the microscopic scale, PSC Mo₂N is able to maintain its initial composition and original cubic morphology after electrolysis, which also proves that the catalytic process is reproducibly cyclable and stable.

The aforementioned results and previous literature reports confirm that the possible NRR mechanism of Mo₂N can be described by the Mars–van Krevelen (MvK) process (Supplementary Figure S5) (Abghoui et al., 2016; Yang et al., 2017; Ren et al., 2018; Jin et al., 2019). First, H⁺ is adsorbed on the N atoms on the surface of Mo₂N. Second, N vacancies appear on the catalyst face due to the departure of NH₃ molecules. Then, N₂ molecules occupy the vacancies, attracting H⁺ to approach. Finally, NH₃ is released from the catalyst surface, and the Mo₂N catalyst returned to its original state. A steady stream of hydrogen ions in the electrolyte promotes the cyclic reaction. In the process of reaction, the porous structure increases the reaction area of the N atom and the active site of the reaction. Single-crystal materials are characterized by lattice continuity, so the migration process of N atoms is not blocked by grain boundaries. Compared with polycrystalline materials, it reduces the reaction kinetic energy of N atom migration, promotes the charge transfer of N atoms at the grain boundary, and accelerates the reaction rate. These results indicate that porous single-crystal molybdenum nitride has fine catalytic activity for the NRR reaction.

In summary, we prepare PSC Mo₂N and PPC Mo₂N. Among them, PSC Mo₂N reduces the charge transfer rate between grain boundaries and improves the rate and stability of the catalytic reaction due to its single-crystal nature. Compared with PPC Mo₂N, PSC Mo₂N exhibits excellent catalytic performance at −0.6 V vs. RHE, resulting in FE and yield of NH₄ ⁺ up to 7.3% and 272.56 μg h⁻¹ mg⁻¹. In addition, the reaction has superior recyclability and electrochemical durability, which maintains approximately constant yields of NH₃ and FE during 10 cycles and hardly weakens after 16 h long-term electrolysis. This work explores a novel avenue for the reasonable design and advancement of the molybdenum-based porous single crystal as the electrocatalysts of the artificial nitrogen fixation reaction.