Effects of nitrogen vacancy sites of oxynitride support on the catalytic activity for ammonia decomposition

Introduction

Hydrogen has emerged as a crucial alternative energy carrier in the face of progressing global warming caused by the extensive consumption of fossil fuels and subsequent CO₂ emissions¹. However, serious problems exist in the storage of liquid hydrogen, such as boiling of the liquid and the requirement of extremely low liquefaction temperatures (–259 °C)². Ammonia has garnered attention as a hydrogen carrier because of its high hydrogen storage density and ease of liquefaction under mild conditions. To realize the practical utilization of ammonia as a hydrogen carrier, the development of an efficient process for hydrogen production via catalytic ammonia decomposition is highly important. The catalytic activity for ammonia decomposition over transition metal catalysts is governed by the interaction between the transition metal surface and ammonia molecules, which results in a volcano-shaped relationship, and Ru has the optimum activity according to this relationship. Ni-based catalysts have been investigated as alternatives to precious metal Ru catalysts for ammonia decomposition. However, the catalytic activity of Ni catalysts is significantly lower than that of Ru catalysts, even though the adsorption of nitrogen on Ni can be enhanced by the formation of alloys or electron donation from the catalyst support^3,4,5. As a result, high reaction temperatures are required to achieve high ammonia decomposition activity. Therefore, the development of Ni-based catalysts that are highly active at low reaction temperatures is strongly desirable.

Recently, it has been reported that nitride-based materials promote the activity of nonnoble metal catalysts for the decomposition of ammonia^6,7,8,9. For example, composite catalysts of alkali or alkaline earth metal imides/amides and transition metals (TMs) are efficient for ammonia decomposition. In particular, the catalytic performance of TM-Li₂NH composite catalysts is comparable to or greater than that of conventional Ru catalysts such as Ru/carbon nanotubes and Ru/MgO⁹. In these catalysts, Li₂NH acts as an NH₃ transmitting agent, which enhances the bonding between the TM and nitrogen. However, these alkali or alkaline earth metal imides/amides are very sensitive to moisture, which leads to deactivation after exposure to air. We recently developed an air-stable oxynitride support of perovskite BaTiO_3-xN_y with a hexagonal structure (h-BaTiO_3-xN_y) that promotes the activity of nonnoble metal catalysts such as Ni, Fe, and Co for ammonia decomposition¹⁰. The chemical composition of BaTiO_3-xN_y was determined to be BaTiO_2.00N_0.34ϒ_0.66, where ϒ denotes an anion vacancy site. The BaTiO_3-xN_y catalyst thus allows many anion vacancy sites in the lattice and promotes ammonia decomposition via a lattice N- and/or anion vacancy-mediated Mars–van Krevelen-type mechanism. The catalytic performance of Ni/BaTiO_2.00N_0.34ϒ_0.66 is comparable to that of a Ca-imide-supported Ni catalyst and remains unchanged even after exposure to water¹⁰. This oxynitride was demonstrated to be a useful catalyst support for ammonia decomposition¹⁰; however, the role of the oxynitride support in ammonia decomposition has yet to be clarified owing to a lack of studies.

In this study, the catalytic performance of Ni/BaTiO_2.00N_0.34ϒ_0.66 (hereafter denoted as Ni/BaTiO_2.00N_0.34) was compared with that of other oxynitride-supported Ni catalysts to clarify the role of the oxynitride support in the ammonia decomposition reaction. Some perovskite oxynitrides have been well studied as photocatalysts for the water splitting reaction because of their high stability in water and efficient light absorption ability^11,12. BaTaO₂N and LaTiO₂N were selected as oxynitride catalyst supports because they have high nitrogen concentrations and are stable in water. The role of the oxynitride support in the catalytic cycle was investigated experimentally and computationally.

Results and discussion

Figure 1a shows the ammonia decomposition activity of Ni supported on various perovskite oxynitrides as a function of reaction temperature. Ba₅Ta₄O₁₅ and La₂Ti₂O₇ were also tested as references. These oxides were also used as starting materials for the synthesis of BaTaO₂N and LaTiO₂N. The oxynitride-supported Ni catalysts showed greater activity than the oxide-supported Ni catalysts did, and 2.5–18 times greater catalytic performance was obtained for the oxynitride-supported Ni catalysts at 500 °C. The oxynitrides without Ni loading exhibited low ammonia conversion (<15%), even at a reaction temperature of 650 °C (Fig. S1), which is much inferior to the catalytic performance of the Ni-loaded oxynitrides. The Ni/h-BaTiO_2.00N_0.34 catalyst exhibited the highest ammonia decomposition activity among the catalysts tested. The apparent activation energy (E_a) for ammonia decomposition over the oxynitride-based catalysts was in the range of 69–76.2 kJ mol^–1, which is relatively lower than that for the oxide-based catalysts (83.8–100.2 kJ mol^–1). This result suggests that the use of an oxynitride as a support changes the reaction pathway compared to that of oxide-based catalysts. Metal oxide supports with alkaline earth or rare earth elements effectively promote the decomposition of ammonia over supported TM catalysts^{13,14,15,16,17}. However, the promotion effect is much more prominent for oxynitride supports that consist of the same elements as the oxide supports, except for nitrogen. As shown in Fig. S2, the ammonia decomposition activities of the Ni-loaded oxynitrides were repeatedly confirmed, with a standard deviation of 5%, demonstrating the high reproducibility of the oxynitride-based catalysts.

**Fig. 1: Ammonia decomposition over various catalysts.**

The stability of these oxynitride-supported Ni catalysts was evaluated at similar NH₃ conversion levels (Fig. 1b). The catalytic activities of Ni/BaTaO₂N and Ni/LaTiO₂N decreased slightly during the first 10 h and subsequently became constant. In contrast, the Ni/h-BaTiO_2.00N_0.34 catalyst exhibited stable activity during the entire test period. Furthermore, the catalytic activities of Ni/h-BaTiO_2.00N_0.34 in the second run were almost identical to the initial activities in all the temperature ranges (Fig. S3). The crystal structure of the oxynitrides remained almost unchanged after the ammonia decomposition reaction, with only Ni/h-BaTiO_2.00N_0.34 exhibiting slight peak broadening (Fig. S4). However, there is no significant difference in the N₂ desorption behavior between the Ni/h-BaTiO_3-xN_y catalysts before and after ammonia decomposition (Fig. S5), which indicates that the composition of the h-BaTiO_3-xN_y support remains unchanged during the reaction. These results suggest that the crystallinity of h-BaTiO_2.00N_0.34 slightly decreases because of surface reconstruction during the ammonia decomposition reaction through a nitrogen vacancy-mediated reaction mechanism. This surface reconstruction may lead to the aggregation of Ni particles on h-BaTiO_2.00N_0.34, which accounts for the slight decrease in catalytic activity with increasing reaction time.

Structural characterization was conducted to elucidate the difference in the ammonia decomposition activity, as shown in Fig. 1. The Brunauer‒Emmett–Teller (BET) surface areas of the samples used in the experiments are summarized in Table S1. These results indicated little difference between the surface areas of different samples. Structural modifications of the Ni nanoparticles caused by interactions with the support were investigated via X-ray absorption fine structure (XAFS) measurements. Figure 2a shows the Ni K-edge X-ray absorption near edge structure (XANES) spectra of the tested catalysts. Although the Ni in Ni/LaTiO₂N was slightly oxidized, the other catalysts were composed of only metallic Ni particles. The partial oxidation of Ni in the Ni/LaTiO₂N catalyst is attributed to the small Ni particle size (Table S2)¹⁸, i.e., the fraction of the Ni-support interface was larger for Ni/LaTiO₂N than for the other catalysts. Fourier transforms of the extended XAFS (EXAFS) spectra (Fig. 2b) indicated that most samples presented signals associated with Ni–Ni bonds with a similar intensity at 2.2 Å, except for Ni/LaTiO₂N. As summarized in Table S2, the Ni particle size in Ni/h BaTiO_2.00N_0.34 was larger than that in the other catalysts. These results suggest that the variation in the ammonia decomposition activity is not due to structural factors such as the size of the supported Ni particles or the surface area of the support.

**Fig. 2: Local structure of Ni determined by XAFS analyses.**

The electronic state of the Ni particle surfaces was further investigated via X-ray photoelectron spectroscopy (XPS) measurements. As shown in Fig. 3, the Ni 2p_3/2 peak for Ni/h-BaTiO_3-x, Ni/h-BaTiO_2.00N_0.34, and Ni/BaTaO₂N appears at a similar position to that of metallic Ni (852.7 ± 0.4 eV)¹⁹. Only Ni/Ba₅Ta₄O₁₅ showed a Ni peak below 852 eV, which indicates that negatively charged Ni formed on this catalyst. The Ni 3p spectra for Ni/LaTiO₂N and Ni/La₂Ti₂O₇ are shown in Fig. 3b, and the La 3d_3/2 peak overlaps with the Ni 2p peak^20,21. The Ni 3p peak position is almost the same as the Ni metal peak position (67 ± 0.2 eV). If electron donation from the support to Ni is a dominant factor for ammonia decomposition activity, then Ni/Ba₅Ta₄O₁₅ should exhibit the best catalytic performance. However, Ni/Ba₅Ta₄O₁₅ had a much lower activity than the oxynitride-supported Ni catalysts. These results indicate that the high ammonia decomposition activity of the oxynitride-based catalyst is not attributed to an electron-donating effect.

**Fig. 3: Electronic state of Ni determined by XPS analyses.**

The decomposition of ¹⁵NH₃ was conducted to clarify the contribution of the oxynitride support to the reaction mechanism. Both ³⁰N₂ and ²⁹N₂ gases were produced from the Ni/h-BaTiO_2.00N_0.34 catalyst, whereas ²⁹N₂ was not produced over Ni/h-BaTiO_3-x (Fig. 4). This suggests that the lattice N^3– in the oxynitride support is involved in the reaction. Figure 4 summarizes the initial ³⁰N₂ and ²⁹N₂ production during the decomposition of ¹⁵NH₃ over various oxynitride-supported Ni catalysts. Ni/h-BaTiO_2.00N_0.34 resulted in the highest ²⁹N₂ production, although h-BaTiO_2.00N_0.34 had the lowest surface area and lowest nitrogen content among the tested oxynitrides. In addition, ²⁹N₂ production from Ni/h-BaTiO_2.00N_0.34 was initiated at ~350 °C (Fig. S6), which is consistent with the onset temperature for ammonia decomposition over Ni/h-BaTiO_2.00N_0.34 (Fig. 1a). The catalytic performance of the oxynitride-supported Ni catalysts is thus governed by the reactivity of the lattice nitride anions on the support surface rather than the structural and electronic properties of the Ni particles.

**Fig. 4: 15NH3 decomposition over various catalysts.**

Figure 5a shows the N₂ temperature programmed desorption (N₂-TPD) profiles for various oxynitride catalysts. N₂ desorption from lattice nitrogen started at a lower temperature for Ni/h-BaTiO_2.00N_0.34 than for the other oxynitride-based catalysts. Figure 5b shows that the N₂ desorption temperature significantly decreased after Ni loading of the h-BaTiO_2.00N_0.34 support, which indicated that the supported Ni particles facilitated the formation of nitrogen vacancy sites at the Ni-support interface. The N₂ desorption temperatures for LaTiO₂N and BaTaO₂N also decreased after Ni loading (Fig. S7). This tendency is very similar to the results of the ammonia decomposition tests. As shown in Fig. 1 and S1, the oxynitrides without Ni loading exhibited much lower activity than did the Ni-loaded oxynitrides. These results imply that N₂ desorption from the oxynitride support is a kinetically significant step that is facilitated by the supported Ni particles.

**Fig. 5: Desorption behavior of nitrogen on oxynitride catalysts.**

Ammonia-temperature programmed surface reaction (NH₃-TPSR) measurements were conducted to explore the reactivity of the nitrogen vacancy sites on each catalyst. Prior to the measurement, each catalyst was heated under Ar flow at 600 °C to form nitrogen vacancy sites, after which NH₃ was adsorbed at 50 °C. N₂ and H₂ were desorbed from the oxynitride-supported Ni catalysts as the temperature increased (Fig. 6), which indicates that the adsorbed NH₃ molecules decomposed on the catalyst surface at elevated temperatures. Ni/h-BaTiO₃ showed very weak desorption peaks because the Ni surface and oxide support have very weak interactions with NH₃ molecules. On the other hand, the oxynitride-supported Ni catalysts exhibited strong N₂ and H₂ desorption peaks. Given the difference in the Ni particle size (Table S2), the large difference in the desorption peaks infers a difference in the number of nitrogen vacancy sites on the oxynitride supports, i.e., NH₃ molecules are preferentially adsorbed at nitrogen vacancy sites on the oxynitride support. As confirmed by N₂-TPD measurements (Fig. 5), many more nitrogen vacancy sites are formed on Ni/h-BaTiO_2.00N_0.34 than on the other catalysts after heat treatment in Ar at 600 °C. Consequently, Ni/h-BaTiO_2.00N_0.34 showed the most intense desorption peaks. N₂ and H₂ desorption was initiated at lower temperatures for Ni/h-BaTiO_2.00N_0.34 than for the other catalysts, which implies that NH₃ molecules are effectively activated at nitrogen vacancy sites on Ni/h-BaTiO_2.00N_0.34 at lower reaction temperatures. Overall, these results strongly suggest that the nitrogen vacancy sites function as adsorption and active sites for ammonia molecules.

**Fig. 6: Surface reaction of catalysts with ammonia.**

Computational calculations

Density functional theory (DFT) calculations were conducted to simulate NH₃ adsorption on the catalyst surface. Prior to NH₃ adsorption on the catalyst surface, the nitrogen vacancy formation energy (E_f(V_N)) on the oxynitride-based catalysts was calculated. E_f(V_N) for h-BaTiO_2.00N_0.34 could not be calculated because h-BaTiO_2.00N_0.34 originally has many intrinsic anion vacancy sites due to the 2N^3– substitution of 3O^2–, which leaves an oxygen vacancy and becomes unstable in the simulation when further V_N sites are created. As summarized in Table S3, E_f(V_N) decreases from 2.98 to 2.68 eV for BaTaO₂N and from 2.23 to 1.98 eV for LaTiO₂N after Ni loading. This result is consistent with the decrease in the nitrogen desorption temperature after Ni loading (Fig. S8). Therefore, it was experimentally and theoretically demonstrated that the supported Ni particles enhance the formation of nitrogen vacancy sites on the oxynitride surface. Next, the adsorption energy for NH₃ (E_ad(NH₃)) on a catalyst surface with V_N sites was calculated, and the optimum adsorption configuration is shown in Fig. 7 for Ni/h-BaTiO_3-xN_y and in Fig. S8 for Ni/BaTaO₂N and Ni/LaTiO₂N. For the Ni/h-BaTiO_2.00N_0.34 catalyst, h-BaTiO₂N_1/3ϒ_2/3 was used as the model structure for the calculation. E_ad(NH₃) for the Ni surface was estimated to be –1.34 eV. On the other hand, NH₃ molecules are adsorbed at Ti cation sites on the support surface adjacent to V_N sites, with an E_ad(NH₃) of –1.41 eV (Table 1). Similarly, the NH₃ molecules strongly interact with the support surface rather than the Ni surface on the Ni/LaTiO₂N and Ni/BaTaO₂N catalysts (Table S3). These results suggest that the ammonia decomposition reaction over the Ni-loaded oxynitride mainly occurs at V_N sites near the interface between the supported Ni and the support.

**Fig. 7: Simulation on ammonia adsorption on the catalyst surface.**

Table 1 Adsorption energy for ammonia molecules on Ni and support surfaces.

Full size table

The calculated lengths of the N‒H bonds of ammonia adsorbed on the catalyst surfaces are summarized in Table 2 for Ni/h-BaTiO_3-xN_y and in Table S4 for Ni/BaTaO₂N and Ni/LaTiO₂N. The N‒H bonds are more elongated on the surface of the oxynitride support with V_N than on the Ni surface, which indicates that N‒H bond cleavage is facilitated on the support surface. A similar phenomenon is observed for the Ni/CaNH catalyst, i.e., ammonia molecules are dehydrogenated at anion vacancy sites on the CaNH surface rather than on the Ni surface, and N‒H bond cleavage occurs even at 50 °C⁶. Furthermore, the NH₃-TPSR results in Fig. 6 show that the desorption temperature for N₂ is higher than that for H₂, which suggests that the rate-determining step for ammonia decomposition on the oxynitride-supported Ni catalyst is N₂ desorption from the oxynitride support.

Table 2 N‒H bond lengths for adsorbed ammonia molecules on Ni and support surfaces of Ni/h-BaTiO_3-xN_y.

Full size table

Conclusion

Compared with oxide-supported Ni catalysts, perovskite oxynitride-supported Ni catalysts exhibited much higher activity for ammonia decomposition. Among the tested catalysts, Ni/h-BaTiO_2.00N_0.34 showed the highest activity. The difference in activity is not due to differences in the support surface area, Ni particle size, or electronic state of Ni but rather to the ease of nitrogen vacancy formation on the oxynitride support. N₂-TPD revealed that the supported Ni particles promoted nitrogen desorption from the oxynitride support and that Ni/h-BaTiO_2.00N_0.34 had the lowest nitrogen desorption temperature. NH₃-TPSR and DFT calculations indicated that NH₃ molecules are activated at nitrogen vacancy sites rather than at the Ni surface. N–H bond cleavage is facilitated at nitrogen vacancy sites on the oxynitride-supported Ni catalysts, and N₂ desorption from the oxynitride support is the rate-limiting step. These results provide new insights for the design of highly active and stable ammonia decomposition catalysts.

Materials and methods

Synthesis of BaTiO_2.00N_0.34 and h-BaTiO_3-x

BaTiO_2.00N_0.34 was synthesized via a solid-phase reaction according to a previously reported procedure^10,22. h-BaTiO_3-xH_x was synthesized via the solid-state reaction of BaH₂ and TiO₂. BaH₂ was prepared by hydrogenation of Ba metal pieces (99.9%, Aldrich) under H₂ gas flow (20 mL min^–1) at 450 °C for 20 h. TiO₂ nanoparticles (99.5%, Aldrich) were dehydrated at 500 °C for 6 h. BaH₂ and TiO₂ were then mixed in an Ar-filled glovebox using an agate mortar for 30 min. The mixture was placed in a quartz tube and heated under a H₂ flow (10 mL min⁻¹) at 800 °C for 20 h. The resultant powder was collected in an Ar-filled glovebox and washed with water several times to remove the small amount of impurity phase present, followed by drying overnight at 70 °C. BaTiO_2.00N_0.34 was prepared by heating h-BaTiO_3−xH_y at 600 °C for 12 h in a stream of pure N₂ gas (10 mL min⁻¹). h-BaTiO_3−x was also prepared by heating h-BaTiO_3−xH_x at 800 °C for 12 h in an air atmosphere.

Synthesis of LaTiO₂N and La₂Ti₂O₇

Powders of LaTiO₂N were obtained by nitridation of oxide precursors prepared via a polymerizable complex method. For oxide precursor synthesis, Ti(C₄H₉O)₄ (Kanto Chemical, 97.0%), La(NO₃)₃·6H₂O (Wako Pure Chemical, 99.9%), and citric acid (Wako Pure Chemical, 98%) were added to CH₃OH (Kanto Chemical, 99.8%). After chelation, propylene glycol (Kanto Chemical, 99.0%) was added, and the mixture was gelatinized. The organic materials were carbonized at 550 °C to obtain the oxide precursors. Nitridation of 1.0 g of oxide precursors was performed under an NH₃ stream (100 mL min⁻¹) at 950 °C for 15 h in a tubular furnace. La₂Ti₂O₇ was obtained by heat treatment of the oxide precursor at 950 °C for 15 h under an air atmosphere²³.

Synthesis of BaTaO₂N and Ba₅Ta₄O₁₅

Powders of BaTaO₂N were obtained by nitridation of oxide precursors prepared via a polymerizable complex method. For oxide precursor synthesis, Ta(C₂H₅O)₅ (Kojundo, 99.999%), BaCO₃ (Kanto Chemical, 99.9%), citric acid (Wako Pure Chemical, 98%), and propylene glycol (Kanto Chemical, 99.0%) were added to CH₃OH (Kanto Chemical, 99.8%). After chelation and gelatinization, the organic materials were carbonized at 550 °C to obtain the oxide precursors. Nitridation of 1.0 g of oxide precursors was performed under an NH₃ stream (100 mL min⁻¹) at 950 °C for 15 h in a tubular furnace. Ba₅Ta₄O₁₅ was obtained by heat treatment of the oxide precursor at 950 °C for 15 h under an air atmosphere²³.

Ni loading method

Nickel-supported catalysts were prepared by H₂ reduction of a mixture of support materials with Ni(η⁵-C₅H₅)₂ (Tokyo Chemical Industry, 98.0%) at 250 °C for 1.5 h.

Characterization

X-ray diffraction analysis was conducted via a diffractometer (D8 ADVANCE, Bruker) with monochromatic Cu Kα radiation (λ = 0.15418 nm) at 45 kV and 360 mA. The air-sensitive samples were placed in airtight X-ray transmitting capsules in an Ar-filled glovebox and measured in air. The BET specific surface area of the catalysts was estimated from N₂ adsorption isotherms measured at 77 K via an adsorption instrument (BELSORP-mini, MicrotracBEL). N₂-TPD and NH₃-TPSR measurements were conducted via a catalyst analyzer (BELCAT A instrument, BELCAT 2 instrument, BELMass, MicrotracBEL) under an Ar flow (10 mL min^–1) in the temperature range of 25–1000 °C. XAFS spectra were analyzed via ATHENA and ARTEMIS software²⁴ and the FEFF6 code²⁵. The Ni particle size was calculated based on the coordination number calculated from the peak fitting results for Ni‒Ni bonds¹⁸. Ni has an fcc structure at room temperature and the nearest neighbor coordination number is 12. The Ni‒Ni bond coordination number for each catalyst was calculated using a value of 12, which is the coordination number for the Ni foil used as a reference in the measurements. XPS spectra (Kratos Ultra 2, Shimadzu) of Ni 2p and 3p were measured using Al Kα radiation at <10^–6Pa. The La 3d_3/2 peaks for Ni/LaTiO₂N and Ni/La₂Ti₂O₇ were compared with the Ni 3p peak because the peak position is close to that of Ni 2p. The binding energies for all the spectra were calibrated with respect to the C 1s peak of the sample. All samples were sealed in transfer vessels in an Ar-filled glovebox to prevent oxidation by air and then introduced into the chamber.

Ammonia decomposition reactions

NH₃ decomposition was performed in a fixed-bed plug-flow silica glass reactor at ambient pressure. Prior to the ammonia decomposition activity tests, the oxide-supported Ni catalysts and the oxynitride-supported Ni catalysts were heated under a N₂/H₂ mixed gas flow (N₂/H₂ = 1:3, 20 mL min^–1) at 500 °C for 2 h for catalyst pretreatment. The catalytic activity and stability tests were performed using 0.1 g of catalyst at 360–660 °C under pure NH₃ gas flow (25 mL min^–1), which corresponds to a weight hourly space velocity (WHSV) of 15,000 mL_NH3 g_cat.^–1 h^–1. Effluent gases were analyzed with an online gas chromatograph (GC-14A, thermal conductivity detector (TCD), Porapak OS column, Shimadzu; He carrier gas).

¹⁵NH₃ gas decomposition tests were performed at 450 °C for 30 min over Ni-loaded oxynitrides and h-BaTiO_3−x (75 mg of catalyst) in a U-shaped silica glass reactor connected to a closed gas circulation system. ¹⁵NH₃ (98% as ¹⁵NH₃, 3.5 mmol) gas was introduced into the glass system, and the product gases were monitored with a mass spectrometer (M-101QATDM, Canon Anelva Corp).

Computational methods

DFT calculations were performed using the generalized periodic boundary conditions and gradient approximation with the Perdew–Vurke–Ernzerhof (PBE) functional²⁶ implemented in the Vienna Ab initio Simulation Package (VASP)^27,28. The core electrons were handled using the projector augmented wave method⁶, and valence electrons were represented with wave functions based on plane waves. The DFT calculations of the oxynitrides were implemented as follows: h-BaTiO_2.00N_0.34(110) contained 260 atoms with the bottom four layers of the 11 atomic layers fixed, BaTaO₂N(001) contained 368 atoms with the bottom four layers of the 9 atomic layers fixed, and LaTiO₂N(010) contained 306 atoms with the bottom four layers of the 10 atomic layers fixed. Each surface was the most stable among the low-index stoichiometric surfaces calculated via DFT. No spatial symmetry was imposed in the calculations. These surfaces were modeled using the supercell approach with sufficient vacuum regions of more than 20 Å to avoid artificial interactions between slabs that are periodically repeated along the axis normal to the surfaces. The z-axis was defined as the direction perpendicular to the surface plane. The cutoff energy of the plane wave basis set was 400 eV. The Monkhorst–Pack k-point grid for Brillouin zone sampling was 2 × 2 × 1 for each model, where the z-axis was defined as the direction perpendicular to the surface plane. The convergence criteria for energy and force were 1.0 × 10^–5eV and 1.0 × 10^–2eV Å^–1, respectively, for all the models. For further correction, the DFT-D3 method of Grimme et al. with a zero-damping function was used²⁹. Ni₆₀-loaded h-BaTiO_2.00N_0.34(110), Ni₇₀-loaded BaTaO₂N(001) and Ni₇₀-loaded LaTiO₂N(010) were used for ammonia adsorption. The number of Ni atoms used was sufficient to fill one side of the structure. Ni is placed in four atomic layers on the catalyst surface and is fully relaxed. Visualization of the crystal structures was performed using the VESTA package³⁰. The formation energy for N^3– vacancies on the BaTaO₂N surface, Ni₇₀/BaTaO₂N interface, LaTiO₂N surface, and Ni₇₀/LaTiO₂N interface was defined as

$${{rm{E}}}_{{rm{f}}}left({{rm{V}}}_{{rm{N}}}right)=left{{rm{E}}left({{rm{S}}}^{{rm{d}}}right)+0.5times {rm{E}}left({{rm{N}}}_{2}right)right}-{rm{E}}left({{rm{S}}}^{{rm{p}}}right),$$

where E(A) represents the total energy of system A, and S^d and S^p represent oxynitrides with and without surface V_N, respectively.

The adsorption energy for the NH₃ [E_ad(NH₃)] species on the surfaces was defined as

$${{rm{E}}}_{{rm{ad}}}left({{rm{NH}}}_{3}right)={{rm{E}}}_{{rm{tot}}}left({{rm{NH}}}_{3}/{rm{surface}}right)-{{{rm{E}}}_{{rm{tot}}}left({rm{surface}}right)+{{rm{E}}}_{{rm{tot}}}left({{rm{NH}}}_{3}right)},$$

where E_tot(NH₃/surface) is the total energy for the optimized NH₃ molecule adsorption configuration and E_tot(NH₃) is the total energy for NH₃.