Theoretical study of surface dependence of NH₃ adsorption and decomposition on spinel-type MgAl₂O₄

Abstract

Ammonia decomposition is a critical process in the production of renewable hydrogen energy. Although many studies have concentrated on ammonia decomposition, little is known about the effect of the spinel support on the activation of NH₃. The adsorption and dissociation of NH₃ on low-index (100), (110), and (111) MgAl₂O₄ surfaces were investigated using density functional theory. The interaction between NH₃ and MgAl₂O₄ was structure dependent, with the surface geometry and electronic structures determining the active sites and adsorption stability. The NH₃ was likely to point toward a surface protruding metal site with the formation of a hydrogen bond. For dissociation, the adsorption energy of the (111) surface was much more favorable than the energies of the (100) and (110) surfaces. The surface Al_3c-sp state at the Fermi level and the formation of the H–O_3c covalent bond of this surface were the main reasons for the higher adsorption energy in NH₃ dissociation. In particular, the hybridization between the Al_3c-sp state and N-p state on the (111) surface was larger than those of the other two surfaces. In view of the thermodynamics and dynamics, the (111) surface was more favorable for NH₃ adsorption and reaction. For the defective surfaces, the existence of an oxygen vacancy lowered the adsorption ability on metal sites. This was due to the destroyed surface symmetry structure and the reduced charge of the metal site. In addition, only on the (111) surface could the oxygen vacancy act as an active site for adsorption. The energy barrier of NH₃ dissociation on the Vo_3c (111) surface was the lowest, which indicated that the NH₃ reaction on this defective surface was dynamically the most favorable. These findings have an important implication for the decomposition of NH₃ on MgAl₂O₄ surfaces and could provide theoretical guidance for other catalytic reactions.

---

1. Introduction

Ammonia has been used as the hydrogen source and storage material in fuel cells.^1–6 The catalytic decomposition of ammonia could produce a clean hydrogen fuel, which could be an effective way to solve the current energy problems, although ammonia (NH₃) is a toxic gas that could cause environmental pollution. Therefore, the catalytic dissociation of ammonia has important significance for solving energy and environmental problems. Composite catalysts have often been used for heterogeneous catalytic reactions. These materials have included transition metals (active components) and supports. Generally, the support of the catalyst has been a base material providing dispersion, support, and stability to the metal active phase. It has also been reported that the support played an essential role in the activation between the catalyst and adsorbate. For example, an alumina support could improve the reaction activity of a metal catalyst in the catalytic process.^7,8 However, some reports have indicated that the support could participate in the reaction without a metal phase: the alumina could directly activate the CO, CH₄,^9,10 and ethanol¹¹ molecules. These reports indicated that the catalysis was complex. They also suggested that it was necessary to systematically investigate the interaction in the heterogeneous catalytic reaction in order to understand the catalysis mechanism, optimize the catalytic system, and design a highly efficient catalyst.

The spinel MgAl₂O₄ material, which shows much more stability and mechanical strength than some other oxides, was an effective support for an ammonia reaction.^12,13 Szmigiel et al. investigated ammonia decomposition over Ba- or Cs-doped Ru catalysts deposited on the MgAl₂O₄ support.¹³ A multi-composite oxide possesses a complex surface structure in comparison with a single oxide, which would allow better properties. Thus, the microscopic surface structure of MgAl₂O₄ was controllable for the ammonia reaction. It is well known that the surface atom arrangements and active sites are different for different surfaces, which would likely result in different adsorption behaviors.^14–18 In addition, the establishment of oxygen vacancies would change the atomic and electronic surface structure, resulting in different adsorption behaviors.¹⁹ The surface chemical properties of an oxygen defective surface would be different from those of a perfect surface.^20–22 Moreover, oxygen vacancies have been found to be important reactive sites on a surface.^19,23 Besenbacher et al. demonstrated that the oxygen vacancies of a TiO₂ surface were active sites for water dissociation. Thus, it was necessary to pay attention to the ammonia reaction properties on oxygen defective MgAl₂O₄ surface.

In this work, we concentrated on the ammonia adsorption and decomposition on perfect and defective MgAl₂O₄ (100), (110), and (111) surfaces. The surface structure–activity relationship was complicated, and it was necessary to perform a detailed investigation. Three low-index stoichiometric MgAl₂O₄ surfaces were selected to explore its structural sensitivity and surface electronic structure. This paper is organized as follows. Section 2 describes the computational methods used. Section 3 presents and discusses the results of a theoretical analysis of NH₃ adsorption on perfect and defective MgAl₂O₄ surfaces. Section 4 summarizes the main conclusions.

2. Computational methods and surface models

The calculation of the NH₃ reaction with MgAl₂O₄ was performed with density functional theory (DFT)²⁴ calculations using CASTEP.²⁵ The core–electron interaction was described using ultrasoft pseudopotentials.²⁶ For Mg, the 2p and 3s states (8 electrons) were treated as valence states. For Al, the 3s and 3p states (3 electrons) were treated as valence states, and for O, the 3s and 3p states (6 electrons) were treated as valence states. After the NH₃ adsorption, for N, the 2s and 2p states (5 electrons) were treated as valence states, whereas for H, the 1s state (1 electron) was treated as a valence state. The generalized gradient approximation (GGA) proposed by Perdew and Wang in 1991 (PW91)^27,28 was employed for the exchange and correlation functional. The self-consistent convergence accuracy was set at 2.0 × 10⁻⁶ eV per atom, and the convergence criterion for the maximal force between atoms was 0.05 eV Å⁻¹. The maximum displacement was 0.002 Å, and the stress was 0.1 GPa. The wave functions were expanded in a plane wave basis set, while the specified cutoff energy was set at 340 eV, and this cutoff energy was used throughout the calculations. Monkhorst–Pack²⁹ grids of 3 × 3 × 3 κ-points were used for the bulk unit cell, and grids of 2 × 2 × 1 κ-points were used for the (100), (110), and (111) surfaces.

The bulk crystal structure of the spinel MgAl₂O₄ was calculated, along with the results of the lattice parameters (a = b = c = 8.20552 Å, α = β = γ = 90°), which remained consistent with the experimental parameters (a = b = c = 8.08060 Å, α = β = γ = 90°).³⁰ The main exposed low-index surfaces of MgAl₂O₄ included the (100), (110), and (111) surfaces. A careful convergence test had to be conducted to choose the proper slab terminations and slab thickness. After this calculation, we found that the Mg–O–Al terminated surfaces had lower surface energies than the Mg–O or Al–O terminated surface (Table 1). Thus, the Mg–O–Al terminated surfaces were considered to be the adsorption surfaces. In order to choose the proper number of slab layers, detailed convergence tests were conducted. The following surfaces were used for the test slab thickness (Table S1, ESI†): (100) for 5, 7, and 9 slab layers; (110) for 5, 6, and 7 slab layers; and (111) for 5, 7, and 9 slab layers. The three surfaces converged to 0.02 eV Å⁻², 0.01 eV Å⁻², and 0.01 eV Å⁻² respectively, which were within 0.065 eV Å⁻² of each other.³¹ Therefore, to keep the computational cost low, we only chose seven layers for the (100) and (111) surfaces and six layers for the (110) surface as the calculation models to study ammonia adsorption in the present work. Thus, the MgAl₂O₄ (100) surface was modeled using a periodic 2 × 2 surface unit cell, with 16 units possessing 112 atoms. The (110) surface was modeled using a periodic 2 × 1 surface unit cell, with 12 units possessing 84 atoms, and the (111) surface was modeled using a periodic 2 × 2 surface unit cell, with 16 units possessing 112 atoms. To separate the layer and its images in the direction perpendicular to the magnesium aluminate plane, we chose 20 Å as the vacuum region. Before the geometry optimization of the three surfaces, the top three layers were selected to relax, while the remaining layers were constrained to model the bulk effects. Moreover, because the three surfaces are polar, particular treatments^32–35 are employed to eliminate the polarity of the surface in Fig. S4 (ESI).† The results of NH₃ adsorption and reaction on MgAl₂O₄ indicated that the polarity played little effect on the selectivity of surface active sites and the NH₃ dissociation process (Fig. S5–S7, ESI†). A Mulliken population analysis^36,37 was performed to determine the charge transfers and population to gain an understanding of the nature of bonding and the interaction between the NH₃ and MgAl₂O₄ surfaces.

Table 1 Surface energies of (100), (110), and (111) surfaces with different terminations

Surface plane	Termination	Surface energy (J m^−2)
(100)	Al–Mg–O	1.37
	Al–O	2.98
(110)	Al–Mg–O	1.81
	Al–O	3.26
(111)	Al–Mg–O	2.00
	Mg–O	4.58

---

The adsorption energy E_ads was used to evaluate the interaction of the NH₃ with the magnesium aluminate spinel, and it was calculated as follows:

E_ads = [E_slab + E_NH3] − E_slab/NH3

where E_slab and E_NH3 are the total energies of the pristine MgAl₂O₄ and free NH₃ molecule, respectively, and E_slab/NH3 is the total energy of the slab with the adsorbed or dissociative-adsorbed NH₃ molecule. On the basis of the above definitions, negative values of adsorption energy corresponded to an endothermic process, whereas positive values indicated that the adsorption was thermodynamically favorable. In order to accurately determine the activation barriers of the reaction, we chose the complete linear and quadratic synchronous transition (LST/QST) approach to search for the transition states (TS) of the reactions.³⁸ The complete LST/QST approach combines the LST algorithms for constrained minimizations with the QST algorithm. An LST/optimization calculation is first performed. Then, the QST maximization is performed to obtain the TS approximation. After this, another constrained minimization is performed, and the cycle is repeated until a stationary point is found, or the number of allowed steps is exhausted. In this system, a value of 0.25 eV was employed for the RMS convergence, while the maximum number of QST steps was five.

3. Results and discussion

3.1 NH₃ adsorption and dissociation on perfect MgAl₂O₄ surfaces

3.1.1 Structure of perfect MgAl₂O₄ surfaces. We first investigated the optimized perfect structures of MgAl₂O₄ (100), (110), and (111) surfaces, as seen in Fig. 1. It was easy to find that the three surfaces all consisted of unsaturated metal (Mg and Al) and oxygen atoms, but differed in their degree of unsaturation and surface atom arrangement. The unsaturated atoms located at the outermost surfaces were separately labeled as 2-fold-coordinated Mg (Mg_2c), 3-fold-coordinated Mg (Mg_3c), 3-fold-coordinated Al (Al_3c), 4-fold-coordinated Al (Al_4c), 5-fold-coordinated Al (Al_5c), 3-fold-coordinated O (O_3c), and 4-fold-coordinated O (O_4c). For the (100) surface, Mg_2c protruded from the plane and Al_5c was surrounded by four oxygen atoms, which acted as metal adsorption sites. O_3c was a favorable site for bonding the H atom, which acted as an O adsorption site. For the (110) surface, the metal adsorption sites included Mg_3c and Al_4c. The difference compared with the (100) surface was that the two types of metal atoms were both located under the surface. The O_3c site slightly protruded from the surface to capture the H atom. The (111) surface possessed a different structure compared with the other two surfaces. The Al_3c atoms protruded from the surface, while the O_3c atom was isolated in the center. In Table 2 are listed the vertical displacements of the surface atoms relative to their bulk positions that have been calculated from the equilibrium slab geometry. For the MgAl₂O₄ (100) surface, the Mg atom considerably relax outward by 0.023 Å, the Al atom move inward by 0.062 Å, the O_3c and O_4c atoms move outward by 0.057 Å and 0.277 Å. For the (110) and (111) surfaces, the metal and O_3c atoms move inward by 0.001–0.463 Å, the O_4c atoms move outward by 0.121–0.171 Å. This indicates that the MgAl₂O₄ surface remains stable in the full-geometry optimization, the presence of unsaturated atoms do not distort the surface upon relaxation.

Fig. 1 Front and top views of optimized surface structures: (a) (100) surface, (b) (110) surface, and (c) (111) surface. The surface sites and Mulliken charges carried by these are labeled in the top view of each surface. Color coding: red, O atoms; purple, Al atoms; green, Mg atoms.

Table 2 Comparison of the vertical relaxation of the top surface atoms for the bare (100), (110) and (111) surfaces of MgAl₂O₄

Surface	z-shifts/Å
	Mg	Al	O3c	O4c
(100)	0.023	−0.062	0.057	0.277
(110)	−0.043	−0.463	−0.029	0.171
(111)	−0.088	−0.109	−0.001	0.121

---

To further examine the effect of the surface electronic structure on the NH₃ adsorption reaction, we calculated the Mulliken charge (as seen in the top view of Fig. 1). Although the absolute value of this charge was meaningless, the difference between the charge values before and after NH₃ adsorption showed us the charge transfer, which determined the adsorption behavior. Our calculation demonstrated that the surface metal atoms carried a positive charge, while the surface oxygen atoms carried a negative charge. The different charge values of the surface metal and oxygen atoms determined the bonding properties between the NH₃ and surface atoms.

The surface geometric and electronic structures would coordinately influence the adsorption and reaction, which would lead to a complex surface adsorption. Therefore, to correlate the surface atomic and electronic structures with the catalytic activity, the interactions between the NH₃ molecule and surfaces of the MgAl₂O₄ catalyst were carefully examined and are thoroughly discussed in the following sections.

3.1.2 Interaction of NH₃ with perfect surfaces. The NH₃ molecule could interact with the MgAl₂O₄ (100), (110), and (111) perfect surfaces in several different ways by utilizing H atoms and protruding N atoms. This section will discuss the molecular and dissociated adsorption conditions of the three surfaces. Five models were constructed to determine the adsorption energy of the MgAl₂O₄ system (Fig. 2). Fig. 2a is a model that the NH₃ molecule adsorbs via one H atom to the surface. Fig. 2b has a bridging configuration with two H atoms of NH₃ binding with two O atoms. Fig. 2c shows a single N-adsorbed model adsorbed via a surface metal atom. Fig. 2d shows a model of the participation of the N–M bond and hydrogen bond, which generates a bidentate bonding species. In Fig. 2e, the N atom pointes toward the metal atom, while bridged hydrogen bonds are formed. Based on the five models, all of the possible adsorption models of the NH₃ molecule with various orientations and different adsorption sites were calculated and are presented in Fig. S1–S3 (ESI).†

Fig. 2 Possible models of NH₃ adsorbed on MgAl₂O₄ surfaces. Color coding: gray, M atoms (M = metal); red, O atoms; blue, N atoms; white, H atoms.

(100) surface. Several types of the most favorable configurations for the NH₃ adsorption and reaction on the (100) surface are presented in Fig. 3. When we calculated the adsorption model that the NH₃ molecule adsorbs via one H atom to the surface, the N atom would shift to form the N–Mg_2c bond [Fig. S1, ESI†]. This proved that the metal atom was the main active site for adsorption. The metal active sites included the Al_5c and Mg_2c atoms. The strong steric hindrance prohibited the adsorption on the Al_5c site, for which the adsorption energy was only 0.61 eV (Fig. 3a). It was easy to find that the adsorption on the Mg_2c site was more favorable than that of the Al_5c site. This was because the Mg_2c atom protruded from the surface without hindrance. As seen in Fig. 3b and c, the adsorption energy increases by 0.09 eV with the participation of the H–O_3c bond on the Mg_2c site. This suggests that the hydrogen bond had a synergic effect on the adsorption. The dissociated structure is presented in Fig. 3d. When the surface Mg_2c pointed to the N atom, and O_3c bonded with the dissociated H atom to form the H–O_3c bond, this bond, with a length of 0.99 Å, was a covalent bond, which would have promoted the dissociation. However, the adsorption energy was only 1.03 eV, which was much lower than the favorable molecular adsorption energy (1.64 eV). Moreover, the adsorption energy of the NH₃ dissociated adsorption was much lower than that (1.55 eV) of the molecular adsorption without the hydrogen bond. Thus, the difference could not be explained using only the geometric structure. Therefore, the Mulliken charges presented in Table 3 were employed to analyze the surface electronic structure and explain the distinct adsorption energies. There was little difference in the charge values before and after the adsorption on the Al site, which resulted in the lowest adsorption energy on the Al_5c site. The change in the charge on the Mg_2c site was complex. A large number of charges increased (0.35|e| and 0.36|e|) on the Mg_2c site for molecular adsorption, and the charge increased by 0.24|e| on the Mg_2c site for dissociation. This indicated that the strong interaction between the NH₃ and Mg_2c during the NH₃ molecular adsorption and reaction. However, the change in the charge on NH₃ was very different. For adsorption without the hydrogen bond, the NH₃ charge changed by 0.05|e|. The charge decreased by 0.11|e| for the molecular adsorption with the hydrogen bond. The NH₃ could obtain an electron from the surface oxygen through the hydrogen bond. For the dissociation, the NH₃ charge decreased by 0.36|e| (Table 3d), which was a larger decrease than that of the molecular adsorption. This was because the NH₂ group obtained an electron from the dissociated H atom. Meanwhile, this H atom obtained an extra electron from the surface O_3c site. The effect of the left electron state on the N atom would play another role in the NH₃ dissociation.

Fig. 3 Front and top views of the most favorable adsorption configurations and energies of NH₃ on a perfect (100) surface: (a) NH₃ pointing towards the Al_5c atom, (b) NH₃ perpendicular to the surface Mg_2c atom, (c) two H atoms pointing towards O_3c atoms on the Mg_2c site, and (d) dissociation on the Mg_2c site.

Table 3 Mulliken charges for clean and adsorbed (100) surface, where clean corresponds to perfect surface configuration without NH₃ adsorption and (a–d) correspond to configurations in Fig. 3

Mulliken charges, |e|		N	H1	H2	H3	NH3	Mg	Al	O3c1	O3c2
(100)	Clean	−1.26	0.42	0.42	0.42	0	1.27	1.50	−1.16	−1.16
	(a)	−1.10	0.36	0.37	0.37	0	1.27	1.51	−1.17	−1.17
	(b)	−1.23	0.40	0.39	0.39	−0.05	1.62	1.49	−1.16	−1.16
	(c)	−1.18	0.34	0.34	0.39	−0.11	1.63	1.48	−1.16	−1.16
	(d)	−1.38	0.39	0.32	0.31	−0.36	1.51	1.53	−1.01	−1.17

---

To further analyze the surface electronic properties, the local density of states (LDOS) was calculated in this work (Fig. 4). A comparison of Fig. 4a and b shows an obvious peak (Mg_2c-s state) at the Fermi level in Fig. 4b, with the special Mg_2c-s state disappearing after the NH₃ adsorption and reaction. The Mg_2c-s state was called the surface state. Normally, the surface state originated from the dangling bonds at the surface, which could affect the electron density distribution of the surface. This suggested that the existence of the surface state had an effect on the chemical reaction.^39,40 The energy obtained from the disappearing surface Mg_2c-s state drove the NH₃ adsorption, which indicated that the surface state contributed to the favorable adsorption on the Mg_2c site. The LDOS of the surface O_3c atoms was further analyzed. For the molecular adsorption, the O_3c-p state showed little change. For the dissociated adsorption, the bonding interaction between the dissociated H atom and the surface O_3c atom was very strong, which resulted in an obvious downshift in the O_3c-p state. Thus, the system energy decreased. It was easy to find that the H–O_3c bond was a covalent bond, which was an important force for the dissociation. In Fig. 4c, the Mg_2c-s orbital overlaps with the N-sp orbital at the valence band, which verifies the bonding between the NH₃ and surface Mg_2c atom. In addition, the N-sp peaks was different between the molecular adsorption and dissociated adsorption. For the dissociation, a new N-sp state at the −1.05 eV position appeared near the Fermi level, which originated from the left electron state of the dissociated H atom. This corresponded to the 0.12|e| reduction in the charge value of the N atom (Table 3d). The N-sp state was located between −9.50 eV and −6.05 eV for the molecular adsorption, whereas the new state obviously drove the N-sp state shift toward a higher energy (between −6.43 eV and −3.74 eV) for the dissociated adsorption. The up-shift of the N-sp state of dissociation gave the system greater disability compared with the system of molecular adsorption. The new N-sp state caused a large electron gathering at the N atom, which resulted into the disability of the NH₃ dissociation on the Mg_2c site. Therefore, from the viewpoint of thermodynamics, the molecular adsorption on the Mg_2c site of the (100) surface was more favorable than that of the dissociation.

Fig. 4 LDOS of (100) surface: (a) clean surface and molecular adsorption on Al_5c site, (b) clean surface and adsorption on Mg_2c site, and (c) comparison between molecular and dissociated adsorptions on Mg_2c site. The Fermi level is shown by the vertical dotted line.

(110) surface. The surface structures and corresponding adsorption energies of NH₃ on the perfect (110) surface are given in Fig. 5. The (110) surface has the similarity of the active centers and its distributions with (100) surface, thus the adsorption models of (110) surface are similar to that of (100) surface. Mg_3c and Al_4c were still the active sites for NH₃ adsorption. The adsorption on the Mg_3c site was more favorable than that on the Al_4c site, which suffered from strong steric hindrance. However, the adsorption energies on the metal sites for the (110) and (100) surfaces were rather distinct. The adsorption energy on the Mg_3c site decreased to 1.01 eV, because the Mg_3c atom was located under the (110) surface, which was a disadvantage for adsorption compared with the protruding Mg_2c atom of the (100) surface. For the adsorption on the Al_4c site, the adsorption energy increased to 0.88 eV, which could be explained by the synergetic effect of two Al_4c atoms. In order to explore the influence of the hydrogen bond, we established a special model, as shown in Fig. 5c, which possessed two hydrogen bonds without the participation of the Mg_3c site. The adsorption energy of this structure was relative low, which proved the minor effect of the hydrogen bond on adsorption. The dissociation occurred at the Al_4c site of the (110) surface, which was different from that of the (100) surface. The surface geometry structure twisted after the NH₃ dissociation on the Al_4c site, which resulted in the instability of the system. When we calculated the NH₃ dissociation on the Mg_3c site, we found that the H atom would reunite with the NH₂ group to form an NH₃ molecule because the surface O_3c atom remained close to the Mg_3c atom, which suggested that dissociation on the Mg_3c site of the (110) surface was impossible.

Fig. 5 Front and top views of most favorable adsorption configurations and energies of NH₃ on perfect (110) surface: (a) NH₃ pointed toward Mg_3c atom, (b) NH₃ located between two Al_4c atoms, (c) two H atoms pointed toward O_3c atoms, and (d) dissociation on Al_4c site.

The Mulliken charges were calculated to analyze the electronic structure presented in Table 4. A much greater variation in the charge values before and after the molecular adsorption on the Mg_3c site (Table 4a) is seen, which proves that the Mg_3c atom is the main active site. The charge did not change on the Al_4c site for the molecular adsorption. For the dissociation, the charge decreased from 1.52|e| to 1.45|e| on the Al_4c site, which was different from the other metal sites (the charge increased after adsorption). The electron transferred to the Al_4c atom, which lowered the interaction between the NH₃ and Al_4c site. The variation in the charge on the NH₃ molecule of the (110) surface was related to the hydrogen bond and dissociated state, which were consistent with those of the (100) surface.

Table 4 Mulliken charges for clean and adsorbed (110) surface, where clean corresponds to perfect surface configuration without NH₃ adsorption and (a–d) correspond to configurations in Fig. 5

Mulliken charges, |e|		N	H1	H2	H3	NH3	Mg	Al	O1	O2
(110)	Clean	−1.26	0.42	0.42	0.42	0	1.47	1.52	−1.18	−1.18
	(a)	−1.21	0.36	0.36	0.39	−0.10	1.72	1.52	−1.17	−1.18
	(b)	−1.11	0.40	0.41	0.36	0.06	1.44	1.52	−1.16	−1.17
	(c)	−1.13	0.33	0.34	0.41	−0.05	1.49	1.52	−1.17	−1.18
	(d)	−1.25	0.41	0.35	0.35	−0.14	1.47	1.45	−1.06	−1.17

---

The LDOS was employed to further analyze the surface electronic structure, as seen in Fig. 6. The surface Mg_3c-s state was easy to find, while no obvious Al_4c-sp state was found at the Fermi level. Therefore, the surface Mg_3c-s state contributed to the favorable adsorption stability on the Mg_3c site. As shown in Fig. 6b, the Al_4c-sp did not shift during molecular adsorption, whereas the Al_4c-sp state at the valence band shifted to a higher energy band. This hindered the dissociated adsorption, which corresponded to a reduction in the charge value on the Al_4c atom. For the dissociation on the Al_4c site, the dissociated H combined with the surface O_3c to form the H–O_3c covalent bond. This investigation showed that the molecular adsorption on the Al_4c site was more stable than the dissociation on the Al_4c site. In summary, we concluded that the molecular adsorption on the Mg_3c site was the most favorable thermodynamically on the (110) surface.

Fig. 6 LDOS of (110) surface: (a) clean surface and molecular adsorption on Mg_3c site and (b) clean surface and adsorption on Al_4c site.

(111) surface. Fig. 7 presents the favorable adsorption configurations and energies. The NH₃ molecule was more inclined to be adsorbed at a position that combined the active metal sites (Al_3c and Mg_3c sites) and generated hydrogen bonds. Compared with the adsorption energies, the most favorable adsorption site was the Al_3c site, which was different from those for the former two surfaces. This was because the Al atom protruded from the surface. The Mg_3c was located under the Al_3c atoms, which suffered from steric hindrance and adsorption competition from the nearby Al_3c. As seen in Table 5, the changes in the Mulliken charges on the Al_3c and Mg_3c atoms could also illustrate the adsorption difference: the charge value increased by 0.61|e| on the Al_3c site, while it only increased by 0.14|e| on the Mg_3c site. We calculated the dissociation on the Mg_3c site and found that the NH₂ would shift to the Al_3c atom. This meant that dissociated adsorption was inclined to occur on the Al_3c site. Fig. 5c presents the dissociation configuration on the Al_3c site. Its adsorption energy was 3.07 eV, which was much higher than that of the molecular adsorption (2.02 eV). The difference in the charge transfer values on the Al_3c site for the molecular and dissociated adsorptions was quite small, and could not be used to explain this large adsorption energy difference. Therefore, the LDOS was employed to explain the different energies, as seen in Fig. 8.

Fig. 7 Front and top views of most favorable adsorption configurations and energies of NH₃ on perfect (111) surface: (a) H₁ pointed toward O_3c atom on Mg_3c site, (b) H₁ pointed toward O_3c atom on Al_3c site, and (c) dissociation on Al_3c site.

Table 5 Mulliken charges for clean and adsorbed (111) surface, where clean corresponds to perfect surface configuration without NH₃ adsorption and (a–d) correspond to configurations in Fig. 7

Mulliken charges, |e|		N	H1	H2	H3	NH3	Mg	Al	O3c
(111)	Clean	−1.26	0.42	0.42	0.42	0	1.65	1.09	−1.08
	(a)	−1.17	0.33	0.34	0.34	−0.16	1.79	1.03	−1.08
	(b)	−1.14	0.38	0.40	0.40	0.04	1.59	1.70	−1.07
	(c)	−1.35	0.41	0.39	0.37	−0.18	1.62	1.68	−0.96

---

Fig. 8 LDOS of (111) surface: (a) clean surface and molecular adsorption on Mg_3c site, (b) clean surface and adsorption on Al_3c site, and (c) comparison between molecular and dissociated adsorptions on Al_3c site.

As seen in Fig. 8a and b, the surface Al_3c-sp state at the Fermi level promoted adsorption, which determined the favorable Al_3c-sp active site; there was no obvious surface state on the Mg_3c site. As could be observed upon NH₃ dissociation, the downshift of the O_3c-sp state indicated that the H–O_3c covalent bond formed. For the adsorption on the Al_3c site, the H–O_3c covalent bond of the dissociation was much stronger than the hydrogen bond of the molecular adsorption. The LDOS in Fig. 8c was employed to analyze the different adsorption energies of the molecular and dissociated adsorptions. For the dissociation, it was easy to find a new N-sp state near the Fermi level located at the −2.24 eV position, which came from the left electron state of the dissociated H atom. This was the same as that of the (100) surface. The N-sp state between −8.31 eV and −6.61 eV just shifted slightly to a higher energy between −7.79 eV and −5.47 eV, which was different from the higher up-shift of the N-sp state on the (100) surface. In addition, the intensity of the overlap between the Al_3c-sp and N-sp state on the (111) surface (Fig. 8c) was larger than the intensity of the overlap between the Mg_2c-s and N-sp state on the (100) surface (Fig. 4c). In particular, the N-p peak near the Fermi level originated from the orbital splitting of dissociated NH₂ exactly overlapped with the new Al_3c-sp peak. Therefore, the adsorption energy of the NH₃ dissociated on the Al_3c site was the most thermodynamically favorable.

In order to perform a kinetic investigation of the surface reaction, the transition states for dissociative adsorption were calculated and are listed in Table 6. The activation barriers of the three surfaces were 2.63, 2.93, and 1.88 eV, respectively, with the barrier energy of the (111) surface being the lowest. Therefore, from a dynamic view, the (111) surface was the best surface for ammonia dissociation.

Table 6 Energy barriers for dissociative adsorption on different active sites of perfect and defective surfaces

(100) TS-dis (eV)	Perfect-Mg2c	Vo3c-Mg2c	Vo4c-Mg1c	—
	2.63	2.82	4.64	—
(110) TS-dis (eV)	Perfect-Al4c	Vo3c-Al4c	—	—
	2.93	2.49	—	—
(111) TS-dis (eV)	Perfect-Al3c	Vo3c-Al3c	Vo3c	Vo4c
	1.88	1.90	1.28	2.22

---

3.2 NH₃ adsorption and dissociation on defective MgAl₂O₄ surfaces

3.2.1 Structure of defective MgAl₂O₄ surfaces. This section discusses the NH₃ molecule adsorption on defective MgAl₂O₄ surfaces with oxygen vacancies. Five types of oxygen vacancy defects are presented in Fig. S8 (ESI):† O_3c and O_4c vacancies on the (100) surface, O_3c vacancy on the (110) surface, and O_3c and O_4c vacancies on the (111) surface. In order to calculate the surface formation energies of an oxygen vacancy E_Vo, we used the following equation:
where E_perf is the total energy of the perfect clean surface, E_def is the total energy of the defective clean surface, and E_O2 is the total energy of the free O₂ molecule. Fig. 9 presents top views of the oxygen vacancy defect structures and corresponding vacancy formation energies, which are labeled as (100)-Vo_3c, (100)-Vo_4c, (110)-Vo_3c, (111)-Vo_3c, and (111)-Vo_4c. The (111) surface possessed a lower formation energy than the other two surfaces, which made it easier to form the oxygen vacancy compared to the other two surfaces. The establishment of the oxygen vacancy changed the surface geometry structure, as seen in Fig. 9, and the surface symmetry structure was destroyed. In addition, the surface Mulliken charges were also changed after establishing the oxygen vacancy. The value of the charge on the metal site decreased, and the reduction of the charge on the Vo_4c surface was higher than that of the Vo_3c surface, which would result in the different adsorption energies. However, the relative values of the Mulliken charges among the three defective surfaces were consistent with those of perfect surfaces. Therefore, the adsorption models for the defective surfaces were similar to the models for perfect surfaces.

Fig. 9 Top views of optimized structures and corresponding surface energies of defective MgAl₂O₄ (100) surface (a and b), (110) surface (c), and (111) surface (d and e). The surface vacancy formation energies and Mulliken charges are labeled in the figure.

3.2.2 Interaction of NH₃ with defective surfaces.
Adsorption on metal site. The optimized configurations of the three surfaces are presented in Fig. S9–S11 (ESI).† It is easy to find the similarity between the defective and perfect surfaces. Surface Mg and Al atoms were still the most important active sites for the adsorption and reaction, and the NH₃ molecule was inclined to be adsorbed on these metal sites. The influence factor for the adsorption did not change: the N–M (metal) bond and H–O covalent band were important forces for adsorption. The existence of the hydrogen bond would have synergistically promoted the adsorption, whereas the steric hindrance prohibited the adsorption. However, the adsorption energies on the defective surfaces were lower than those on the perfect surfaces. This was because the symmetry of the surface configurations was destroyed, and the charge value of the metal site decreased. With the existence of the oxygen defect, the reduction of the charge value indicated that the electrons partly transferred to the metal site, which changed the surface state on the metal site. As seen in Fig. 10, the surface Mg_2c-s state decreased on the Vo_3c (100) surface, and the surface Mg_1c-s state shifted toward a lower energy on the Vo_4c (100) surface. The (110) surface Mg_2c-s state shifted toward a lower energy on the Vo_3c surface. The (111) surface Al_3c-sp state changed very little on the Vo_3c site, and the Al_2c-sp state was split on the Vo_4c surface. These findings suggest that the existence of the oxygen vacancy on the surface weakened the effect of the surface state, except for the Vo_3c site of the (111) surface. The adsorption energy of the NH₃ dissociated on the Al_3c site of the (111) Vo_3c surface was 2.27 eV, which was the highest among the defective surfaces.

Fig. 10 LDOS for active sites on perfect and defective (100), (110), and (111) surfaces without adsorption.

Adsorption on vacancy site. In addition to the effect of the metal adsorption site, the oxygen vacancy could also act as an adsorption site. Thus, we further calculated the configurations in relation to the adsorption and dissociation on the Vo_3c and Vo_4c sites, as shown in Fig. 11. For the (100) and (110) surfaces, a strong steric hindrance could significantly weaken the stability of the adsorption, and the adsorption energies were very low on these two surfaces. This indicated that the vacancy sites for NH₃ adsorption on the (100) and (110) surfaces were not the active sites. The adsorption energies on the defective sites of the (111) surface were relatively higher, because the oxygen vacancy acted as an active site for adsorption and dissociation. For the NH₃ molecular adsorption on the Vo_3c site (Fig. 11d), the adsorption energy was only 0.61 eV, which suggested that the NH₃ molecule still suffered from some steric hindrance. For the dissociation on the Vo_3c site (Fig. 11e), the adsorption energy was 2.55 eV. The NH₂ was located at the center with a slight steric hindrance, and the dissociated H combined with the Al_3c atom, which resulted in higher adsorption energy. It was impossible for NH₃ molecular adsorption to occur on the Vo_4c site because of the strong steric hindrance and competition of the neighboring Al site. For the NH₃ dissociation on the Vo_4c site, the adsorption energy was 1.52 eV, which was lower than the dissociation on the Vo_3c site. The H–O_3c covalent bond played an important role for adsorption, whereas the NH₂ group located between the Al and Mg atoms suffered from resistance because of steric hindrance. A comparison of the adsorptions and reactions for all of these showed that the dissociation on the Vo_3c site of the (111) surface was the most favorable adsorption on the oxygen vacancy sites.

Fig. 11 Front and top views of the main adsorption and dissociation configurations and energies of NH₃ on Vo site: (a) molecular adsorption on Vo_3c site of (100) surface, (b) dissociated adsorption on Vo_3c site of (100) surface, (c) dissociated adsorption on Vo_3c site of (110) surface, (d) molecular adsorption on Vo_3c site of (111) surface, (e) dissociated adsorption on Vo_3c site of (111) surface, and (f) dissociated adsorption on Vo_4c site of (111) surface.

After comparing the adsorption energies of the defective surfaces, we came to the conclusion that the adsorption and dissociation on the metal or vacancy sites of the (111) surface were much more favorable than those on the other two surfaces. Table 6 presents the barrier energies of the NH₃ reaction on the Vo_3c and Vo_4c surfaces. The existence of an oxygen vacancy could lower the potential barrier, especially for the reaction on the (111) surface. The barrier energy on the Vo_3c site of the (111) surface was only 1.28 eV, which was much lower than the barrier energy (1.88 eV) on the perfect surface. These results suggested that the establishment of an O_3c vacancy on the (111) surface was dynamically favorable for NH₃ reaction.

4. Conclusions

To fully understand the mechanism of NH₃ adsorption on MgAl₂O₄ surfaces, the density functional theory was employed to investigate the NH₃ adsorption and reaction on perfect and defective (100), (110) and (111) MgAl₂O₄ surfaces. The results showed that the adsorption of NH₃ on MgAl₂O₄ surfaces was a structure-sensitive reaction. Surface metal atoms acted as the main sites for the adsorption and reaction, while the hydrogen bond had a synergistic effect on the adsorption stability. The surface electronic structure was another important factor for adsorption. The surface state at the Fermi level was an important force for the adsorption, which determined the activated adsorption metal site. In particular, the overlap between the Al_3c-sp orbital and N-sp state could significantly promote adsorption on the (111) surface. The coordinated action of the surface geometry and electronic structure stabilized the NH₃ adsorption on the (111) surface. Our work demonstrated that the (111) was the most favorable surface for the NH₃ adsorption and reaction in view of the thermodynamics and dynamics.

The generation of an oxygen vacancy resulted in a reduction in the surface metal charge, which lowered the adsorption stability on the metal site. In addition, the oxygen vacancy acted as an active site for adsorption on the (111) surface. In particular, the NH₃ dissociation on Vo_3c site of (111) surface had the lowest reaction barrier. This suggested that the establishment of an oxygen vacancy on the (111) surface was dynamically favorable for the reaction.

In summary, this work suggested that the selection of the exposed surface and oxygen vacancy of MgAl₂O₄ was very important for NH₃ adsorption and dissociation. Information about the adsorbate geometries, reaction activities of various surface sites, and specific electronic structure of the surface metal atoms could provide theoretical guidance for the future design of MgAl₂O₄ catalysts, as well as for an atomistic-level understanding of other structure-dependent reactions.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (Grant no. 21173131), and the Taishan Scholar Project of Shandong Province (China).

References

1. A. K. Avci, D. L. Trimm and Z. I. Onsan, Quantitative Investigation of Catalytic Natural Gas Conversion for Hydrogen Fuel Cell Applications, Chem. Eng. J., 2002, 90, 77–87.
2. F. Rosowski, A. Hornung, O. Hinrichsen, D. Herein, M. Muhler and G. Ertl, Ruthenium Catalysts for Ammonia Synthesis at High Pressures: Preparation, Characterization, and Power-law Kinetics, Appl. Catal., A, 1997, 151, 443–460.
3. J. C. Ganley, F. S. Thomas, E. G. Seebauer and R. I. Masel, A Priori Catalytic Activity Correlations: The Difficult Case of Hydrogen Production from Ammonia, Catal. Lett., 2004, 96, 117–122.
4. T. V. Choudhary, C. Sivadinarayana and D. W. Goodman, Catalytic Ammonia Decomposition: CO_x-Free Hydrogen Production for Fuel Cell Applications, Catal. Lett., 2001, 72, 197–201.
5. R. Metkemeijer and P. Achard, Comparison of Ammonia and Methanol Applied Indirectly in A Hydrogen Fuel Cell, Int. J. Hydrogen Energy, 1994, 19, 535–542.
6. A. Klerke, C. H. Christensen, J. K. Nørskov and T. Vegge, Ammonia for Hydrogen Storage: Challenges and Opportunities, J. Mater. Chem., 2008, 18, 2304–2310.
7. L. G. V. Briquet, C. R. A. Catlow and S. A. French, Structure and Reactivity of Aluminum Oxide Supported Nickel Clusters, J. Phys. Chem. C, 2010, 114, 22155–22158.
8. J. D. Li, E. Croiset and L. Ricardez-Sandoval, Effect of Metal-Support Interface During CH₄ and H₂ Dissociation on Ni/γ-Al₂O₃: A Density Functional Theory Study, J. Phys. Chem. C, 2013, 117, 16907–16920.
9. R. Wischert, P. Laurent, C. Copéret, F. Delbecq and P. Sautet, γ-Alumina: The Essential and Unexpected Role of Water for the Structure, Stability, and Reactivity of “Defect” Sites, J. Am. Chem. Soc., 2012, 134, 14430–14449.
10. G. R. Jenness, M. A. Christiansen, S. Caratzoulas, D. G. Vlachos and R. J. Gorte, Site-Dependent Lewis Acidity of γ-Al₂O₃ and Its Impact on Ethanol Dehydration and Etherification, J. Phys. Chem. C, 2014, 118, 12899–12907.
11. D. Cao, Y. Li, J. Wang and H. Jiao, Mechanism of γ-Al₂O₃ Support in CO₂ Reforming of CH₄ A Density Functional Theory Study, J. Phys. Chem. C, 2011, 115, 225–233.
12. H. Tan, K. Li, S. Sioud, D. Cha, M. H. Amad, M. N. Hedhili and Z. A. Al-Talla, Synthesis of Ru Nanoparticles Confined in Magnesium Oxide-modified Mesoporous Alumina and Their Enhanced Catalytic Performance during Ammonia Decomposition, Catal. Commun., 2012, 26, 248–252.
13. D. Szmigiel, W. Raróg-Pilecka, E. Miskiewicz, Z. Kaszkur and Z. Kowalczyk, Ammonia Decomposition over the Ruthenium Catalysts Deposited on Magnesium–aluminum Spinel, Appl. Catal., A, 2004, 264, 59–63.
14. J. M. H. Lo, T. Ziegler and P. D. Clark, SO₂ Adsorption and Transformations on γ-Al₂O₃ Surfaces: A Density Functional Theory Study, J. Phys. Chem. C, 2010, 114, 10444–10454.
15. J. Handzlik and P. Sautet, Structure of Dimeric Molybdenum(VI) Oxide Species on γ-Alumina: A Periodic Density Functional Theory Study, J. Phys. Chem. C, 2010, 114, 19406–19414.
16. D. H. Mei, V. A. Glezakou, V. Lebarbier, L. Kovarik, H. Y. Wan, K. O. Albrecht, M. Gerber, R. Rousseau and R. A. Dagle, Highly Active and Stable MgAl₂O₄-supported Rh and Ir Catalysts for Methane Steam Reforming: A Combined Experimental and Theoretical Study, J. Catal., 2014, 316, 11–23.
17. F. Frechard, R. A. van Santen, A. Siokou, J. W. Niemantsverdriet and J. Hafner, Adsorption of Ammonia on the Rhodium (111), (100), and Stepped (100) Surfaces: An Ab Initio and Experimental Study, J. Chem. Phys., 1999, 111, 8124–8130.
18. S. B. Tang and Z. X. Cao, Adsorption and Dissociation of Ammonia on Graphene Oxides: A First-principle Study, J. Phys. Chem. C, 2012, 16, 8778–8791.
19. S. Wendt, R. Schaub, J. Matthiesen, E. K. Vestergaard, E. Wahlstrom, M. D. Rasmussen, P. Thostrup, L. M. Molina, E. Laegsgaard, I. Stensgaard, B. Hammer and F. Besenbacher, Oxygen vacancies on TiO₂ (110) and their interaction with H₂O and O₂: A combined high-resolution STM and DFT study, Surf. Sci., 2005, 598, 226–245.
20. C. T. Campbell and C. H. F. Peden, Oxygen vacancies and catalysis on ceria surfaces, Science, 2005, 309, 713–714.
21. M. Fronzi, S. Piccinin, B. Delley, E. Traversa and C. Stampfl, Water adsorption on the stoichiometric and reduced CeO₂ (111) surface: a first-principles investigation, Phys. Chem. Chem. Phys., 2009, 11, 9188–9199.
22. C. H. Sun, T. Liao, G. Q. Lu and S. C. Smith, The Role of Atomic Vacancy on Water Dissociation over Titanium Dioxide Nanosheet: A Density Functional Theory Study, J. Phys. Chem. C, 2012, 116, 2477–2482.
23. R. Schaub, P. Thostrup, N. Lopez, E. Laegsgaard, I. Stensgaard, J. K. Norskov and F. Besenbacher, Oxygen Vacancies as Active Sites for Water Dissociation on Rutile TiO₂ (110), Phys. Rev. Lett., 2001, 87, 266104–266111.
24. M. C. Payne, M. P. Teter, D. C. Allan, T. A. Arias and J. D. Joannopoulos, Iterative Minimization Techniques for Ab Initio Total-energy Calculations: Molecular Dynamics and Conjugate Gradients, Rev. Mod. Phys., 1992, 64, 1045–1097.
25. M. Segall, P. Lindan, M. Probert, M. Pickard, C. Pickard, P. Hasnip, S. Clark and M. Payne, First-principles Simulation: Ideas, Illustrations and the CASTEP Code, J. Phys.: Condens. Matter, 2002, 14, 2717–2743.
26. S. J. Clark, M. D. Segall, C. J. Pickard, P. J. Hasnip, M. J. Probert, K. Refson and M. C. Payne, First Principles Methods Using CASTEP, Z. Kristallogr., 2005, 220, 567–570.
27. J. P. Perdew, J. A. Chevary, S. H. Vosko, A. K. Jackson, R. M. Pederson, D. J. Singh and C. Fiolhais, Atoms, Molecules, Solids, and Surfaces: Applications of the Generalized Gradient Approximation for Exchange and Correlation, Phys. Rev. B: Condens. Matter Mater. Phys., 1992, 46, 6671–6687.
28. J. P. Perdew and Y. Wang, Accurate and Simple Analytic Representation of the Electron-gas Correlation Energy, Phys. Rev. B: Condens. Matter Mater. Phys., 1992, 45, 13244–13249.
29. H. J. Monkhorst and J. D. Pack, Special Points for Brillouin-zone Integrations, Phys. Rev. B: Condens. Matter Mater. Phys., 1976, 13, 5188–5192.
30. T. Yamanaka, Y. Takéuchi and M. Tokonami, Anharmonic Thermal Vibrations of Atoms in MgAl₂O₄ Spinel at Temperatures up to 1933 K, Acta Crystallogr., Sect. B: Struct. Sci., 1984, 40, 96–102.
31. C. G. Zhou, J. Y. Li, S. Chen, J. P. Wu, K. R. Heier and H. S. Cheng, First-Principles Study on Water and Oxygen Adsorption on Surfaces of Indium Oxide and Indium Tin Oxide Nanoparticles, J. Phys. Chem. C, 2008, 112, 14015–14020.
32. P. Li, X. Zhao, C. J. Jia, H. G. Sun, L. M. Sun, X. F. Cheng, L. Liu and W. L. Fan, ZnWO₄/BiOI heterostructures with highly efficient visible light photocatalytic activity: the case of interface lattice and energy level match, J. Mater. Chem. A, 2013, 1, 3421–3429.
33. M. K. Rasmussen, A. S. Foster, B. Hinnemann, F. F. Canova, S. Helveg, K. Meinander, N. M. Martin, J. Knudsen, A. Vlad, E. Lundgren, A. Stierle, F. Besenbacher and J. V. Lauritsen, Stable Cation Inversion at the MgAl₂O₄ (100) Surface, Phys. Rev. Lett., 2011, 107, 036102.
34. Q. L. Tang and Q. H. Luo, Adsorption of CO₂ at ZnO: A Surface Structure Effect from DFT+U Calculations, J. Phys. Chem. C, 2013, 117, 22954–22966.
35. M. M. Rodriguez, X. H. Peng, L. J. Liu, Y. Li and J. M. Andino, A Density Functional Theory and Experimental Study of CO₂ Interaction with Brookite TiO₂, J. Phys. Chem. C, 2012, 116, 19755–19764.
36. D. Sanchez-Portal, E. Artacho and J. M. Soler, Projection of Plane-wave Calculations into Atomic Orbitals, Solid State Commun., 1995, 95, 685–690.
37. X. Gonze, First-principles Responses of Solids to Atomic Displacement and Homogeneous Electric Fields: Implementation of A Conjugate-gradient Algorithm, Phys. Rev. B: Condens. Matter Mater. Phys., 1997, 55, 10337–10354.
38. T. A. Halgren and W. N. Lipscomb, The Synchronous-transit Method for Determining Reaction Pathways and Locating Molecular Transition States, Chem. Phys. Lett., 1977, 49, 225–232.
39. V. P. Indrakanti, J. D. Kubicki and H. H. Schobert, Photoinduced activation of CO₂ on Ti-based heterogeneous catalysts: Current state, chemical physics-based insights and outlook, Energy Environ. Sci., 2009, 2, 745–758.
40. K. Hadidi, R. Hancke, T. Norby, A. E. Gunnces and O. M. Lovvik, Atomistic Study of LaNbO₄; Surface Properties and Hydrogen Adsorption, Int. J. Hydrogen Energy, 2012, 37, 6674–6685.

---

Footnote

† Electronic supplementary information (ESI) available: The convergence surface energies for the different layers of the three surfaces; all the possible configurations of NH₃ molecule adsorbing on different sites; the treatment of the surface polarity; the models of the oxygen vacancy; the configurations of NH₃ adsorbing on defective surfaces. See DOI: 10.1039/c5ra07818k

---