Chen Zihui, Zhao Chuanlin, Liu Jinxun* , Li Weixue,2*

1. Department of Chemical Physics, School of Chemistry and Materials Science, University of Science and Technology of China, Hefei 230026, China;2. Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China

Abstract: Hydrogen activation plays a pivotal role in hydrogenation reactions over transition metal oxide catalysts. Clarifying hydrogen activation over ceria oxide (CeO2) is an important issue in the acetylene hydrogenation reaction. Employing density functional theory (DFT) calculations, we studied hydrogen activation over stoichiometric and defective CeO2(111), (110), and (100) surfaces. Hydrogen dissociates on the stoichiometric CeO2 surfaces only forming hydroxyl groups. The presence of oxygen vacancies can promote the H2 activation over the defective CeO2 surfaces. Both H+ and H- species can be found on the defective CeO2(111) and (100) surfaces, whereas only H+ species can be observed on the defective CeO2(110) surface. The structure sensitivity of the H2 activation over the stoichiometric and defective CeO2 surfaces is correlated with H+ and H- adsorption energies determined by the ability of the surface oxygen vacancy formation and charge distributions of Ce and O ions. Our work provides more insight into H2 activation on CeO2-based catalysts which will guide better catalyst design for hydrogenation reactions.

Keywords: CeO2; hydrogen activation; surface sensitivity; density functional theory

1 Introduction

Partial hydrogenation of alkynes to olefins is one of the most important industrial reactions and is widely used to purify olefin streams usually containing acetylenic[1]. One of the most commonly used catalysts in partial hydrogenation of alkynes to olefins is Pd which, however, shows a high tendency in excessive hydrogenation in the formation of alkanes and polymerization of alkynes. To tackle this issue, Pd metal alloying with other metals was synthesized to suppress the over-hydrogenation of alkynes[2-4]. For example, alloying Pd with a less active metal can weaken acetylene adsorption and destroy the formation of the β-hydride phase thus improving the selectivity of the partial hydrogenation reaction of alkynes toward olefins[3]. However, the high cost of Pd limits its scale-up application in the industry. To maximize the atom-utilization efficiency, a variety of single-atom catalysts have been synthesized to catalyze alkynes hydrogenation reaction[5]. Many works reported that Pt, Pd, Rh, and other transition metal single atoms deposited on graphene and black phosphorus show high catalytic activity in alkyne hydrogenation[6-8]. However, the single-atom catalysts are often difficult to prepare and ready to aggregate to form large nanoparticles in practice[5]. Therefore, it is highly desirable to design catalysts with high stability, activity, and selectivity for alkyne hydrogenation to replace the usage of expensive noble metal catalysts.

As a popular catalytic material, ceria oxide (CeO2) is widely used as the support and catalyst[9-12]and applied to solid oxide fuel cells[13,14]and oxygen sensors[15,16], which is attributed to its low price[17], excellent acid-base properties, and redox properties[18]. Consequently, CeO2has gained much current interest in catalysis originating from its superior activity and selectivity for many chemical reactions including alkyne semi-hydrogenation reactions[19-23]. Pérez-Ramírez and coworkers[19]found the conversion of propyne and the selectivity of olefins can be achieved as high as 91% and 96% for propyne hydrogenation reaction over bulk CeO2catalyst at the reaction condition ofT= 523 K andP= 1 bar with H2/C2H2ratio of 30:1, respectively. Generally, CeO2displays better catalytic performance in acetylene hydrogenation as compared with conventional noble Pd-based catalysts[2,24]. Carrasco and coworkers[20]revealed that the hydrogen dissociation has the highest activation barrier and can be considered as the rate-determining step for acetylene hydrogenation over the CeO2(111) surface. Consistent with this work, many groups[25,26]also demonstrated the difficulty of hydrogen activation in the hydrogenation of alkynes over the CeO2(111) surface or bulks. Since H2activation plays an important role in the acetylene and/or alkynes hydrogenation reaction, the exploration of H2activation over the CeO2surfaces is pivotal to design better CeO2-based catalysts for acetylene and/or alkynes hydrogenation.

Different from metals, there are generally two different routes for the H2dissociation over metal oxide surfaces, namely, homolytic and heterolytic dissociation of H2. The homolytic H2dissociation often generates two OH groups over metal oxides that are difficult to be reduced, such as MgO[27,28], Al2O3and SiO2[29]. Whereas heterolytic H2dissociation generates an H cation bound to O anion and an Hanion bound to the metal cation. Many experimental and theoretical works indicate that the heterolytic H2dissociation is more likely to occur over reducible oxides[30-32], such as CeO2[33]and TiO2[34]. Density functional theory (DFT) calculations reported that the homolytic H2dissociation is dominant on CeO2(111)[20,30]with an activation barrier larger than 1 eV. Therefore, the H2dissociation is difficult on the CeO2(111) surface at a relatively low temperature corroborated by the experimental measurements[19,20]. A similar high activation barrier of the H2dissociation was also found for the reverse methanol-reforming reaction over different kinds of CeO2surfaces[22].

Although many investigations are done,there are still open on whether and why the oxygen vacancy can promote hydrogenation activation over CeO2catalyst and how the different stoichiometric and defective surfaces of CeO2impact H2activation. In the present work, we studied H2activation over the stoichiometric and defective (111), (110), and (100) facets of CeO2. The structure sensitivity of the H2activation and oxygen vacancy effect on H2activation are revealed on CeO2. The most probable H species on the stoichiometric and defective CeO2surfaces are identified. Our work provides more insight into the H2activation over the CeO2based catalyst which guides a better CeO2-based catalyst design for acetylene hydrogenation reactions.

2 Computational methods

Spin-polarized periodic density functional theory (DFT) calculations were performed by using the Vienna ab initio simulation package (VASP)[41,42]. The exchange-correlation potential was treated by the generalized gradient approximation (GGA) in the Perdew-Burke-Ernzerh (PBE) functional form[43]. The projected-augmented wave (PAW) pseudopotentials were utilized to describe the core electrons[44], and the Kohn-Sham valence states were expanded in a plane-wave basis set with the kinetic energy of 400 eV. The Brillouin zone integration was sampled with 12×12×12 Monkhorst-Pack mesh k-points for the bulk CeO2calculations. The equilibrium lattice constant for bulk CeO2was optimized to be 5.45 Å, in good agreement with the experimental measurement of 5.41 Å[45]. The DFT +Umethodology was used to treat the on-site Coulomb and the exchange interaction of the strongly localized Ce 4f electrons with an effectiveUeff= 5 eV. Here, we adoptedUvalues of 2 eV and 5 eV, which represent a lowUvalue and a highUvalue respectively, to calculate hydrogen dissociation energies on defective CeO2(111) surfaces (Table S1). The difference in H2dissociative adsorption energies by using a low and a highUvalues are 0.29 eV and 1.33 eV forming H+/H-and H+/H+, respectively. DifferentUvalues will result in significantly different H2dissociative adsorption energies. Hydrogen dissociation on the CeO2-x(111) surface is more difficult by using a lowUvalue which conflicts with the experiment results[1,36,37]that oxygen vacancy can promote hydrogen activation. However, a highUvalue of 5 eV is considered to provide localization of the electrons left upon oxygen removal from CeO2[46]. Therefore, we used theUvalue of 5 eV for the surface adsorption and reaction calculations in the present work.

The O-terminated CeO2(111) surface with a (2×2) unit cell, a Tasker Type 2 surface[47], was modeled by a four-layers slab with the bottom two layers fixed at their bulk positions (Figure 1(a)). The CeO2(110) with a (2×2) unit cell, a Tasker Type 1 surface[47], was modeled by a periodic five-layers slab with the bottom two layers fixed at their bulk positions (Figure 1(b)). Whereas the O-terminated CeO2(100) with (3x3) periodicity, a Tasker Type 3 surface[47], was modeled by a periodic seven-layer slab with the bottom two layers fixed at their bulk positions (Figure 1(c)). To eliminate the dipole moment perpendicular to the CeO2(100) surface[23], half of the O atoms from the top layer are removed, which has been commonly applied in many previous computational studies[48,49].

Figure 1. The top view of CeO2(111) (a), CeO2(110) (b) and CeO2(100) (c) surfaces. Red and beige spheres are the O and Ce atoms, respectively. Small spheres stand for the subsurface O and Ce atoms. The dashed black circles represent the oxygen vacancies. This notation is used throughout this paper. The indicated numbers are non-equivalent oxygen atoms.

Figure 2. Top and side views of optimized configurations for heterolytic H2 dissociative adsorption over stoichiometric CeO2(111) (a), CeO2(110) (b), and CeO2(100) (c) surfaces. The H-Ce bond distances are indicated in Å.

A 3×3×1 Monkhorst-Pack mesh k-points were used for calculations of H2activations over three different CeO2surfaces. All slabs were separated by a 12 Å vacuum. All structures were relaxed until forces on each ion were less than 0.02 eV/Å, and the convergence criterion for energy was 10-4eV. Transition structures (TS) for the considered reaction paths were located by using the climbing-image nudged elastic band (CI-NEB) algorithm. The adsorption energy was calculated asEads=Etotal-Eslab-Egas, whereEtotal,Eslab, andEgasrefer to the energy of the slab with adsorbate, the energy of clean CeO2surfaces, and the energy of a gas-phase molecule in a neutral state, respectively. the H2activation barrier is calculated as the energy difference between the transition state and the adsorption state of H2over CeO2surfaces.

3 Results

3.1 Structure sensitivity of H2 activation over stoichiometric CeO2 surfaces

We studied H2activation over the three low-index facets of CeO2(Figure 1), namely (111), (110), and (100) surfaces, which are prototypical examples of three types of ionic crystal facets. The O and Ce atoms are distributed alternately over the O-terminated CeO2(111) surface, where the surface O atom is coordinated with three Ce atoms and each Ce atom binds three O atoms. Similarly, O and Ce atoms are distributed alternately on the O-terminated CeO2(100) surface, whereas surface O atoms are coordinated with three Ce atoms but each Ce atom binds two surfaces and four lattice oxygen atoms. The O and Ce atoms are distributed in the same layer on the CeO2(110) surface, in which the surface O atom binds two Ce atom and Ce atoms are coordinated with four surface and two lattice oxygen atoms. The different surface structures of CeO2(111), (110), and (100) surfaces will exhibit the distinct catalytic performance of H2activation. H2first adsorbs on the CeO2surface and then heterolytic dissociate in the formation of an H+bound to the O anionand an H-bound to the Ce cation(Figure 2). When H atom adsorbs at the O anion site, more electrons can transfer from H to O forming H cation due to the higher electronegativity of O as compared with H. The Bader charges of H atoms at the Ce cation and O anion site are -0.17, -0.15, -0.66 and +0.17, -0.09, +0.21 over CeO2(111), (110), and (100) surfaces, respectively (Table S2). Later, the dissociated H atom migrates from the Ce cation to the O anion site. The calculated potential energy surface diagrams and corresponding configures for H2activation are shown in Figure 3.

Figure 3. The potential energy surface diagram (a) and corresponding transition state configurations (b) for heterolytic H2 dissociation over the stoichiometric CeO2(111), (110), and (100) surfaces. The bond distance between the two dissociated H atoms is indicated in Å.

There is a weak interaction between the H2and stoichiometric CeO2surfaces and H2physically adsorbs on the stoichiometric CeO2surfaces with the adsorption energies higher than -0.10 eV. Two different H2dissociation mechanisms, namely the heterolytic and homolytic dissociation of H2, are considered in the present work. The bond distance of two surface neighboring oxygen atoms in CeO2surfaces is larger than 3.85 Å, which is longer than that of the Ce-O bond, the heterolytic H2dissociation has a high priority to occur. the limiting distance for hemolytic H2dissociation to occur is determined by the distance between neighboring surface O atoms. Therefore, we studied the heterolytic dissociation of H2at the Ce-O pair (Figure 2), and then the adsorbed H+at the Ce cation site might migrate to the O anion thus forming the homolytic H2dissociative adsorption modes on CeO2surfaces.

The heterolytic dissociation of H2is sensitive to the surface structure of CeO2from both thermodynamic and kinetic aspects. Specifically, the heterolytic dissociation of H2to generate an H+and an H-is endothermic by 0.80 eV with an activation barrier of 1.32 eV over the CeO2(111) surface. The high energy cost and activation barrier require a relatively high reaction temperature for the heterolytic dissociation of H2on CeO2(111). As shown in Table S3 and Figure S1, we can see H binding strength increases with the Ce-H bond length increasing from 2.2 to 2.5 Å where a chemical bond can be formed between Ce and H. However, H binding energy becomes +2.2 eV when the bond distance between H and Ce is larger than 2.6 Å indicating the physisorption of the H atom at the Ce cation site. As a result, the formed H anion at the Ce cation is ready to migrate to the neighboring O atom with a low activation barrier of 0.14 eV on the CeO2(111) surface. Therefore, only H+can be found once H2dissociates on the CeO2(111) surface.

Heterolytic dissociation of H2is more feasible on CeO2(110) as compared with that on CeO2(111) surface (Figure 3). The calculated reaction energy and activation barrier for the heterolytic dissociation of H2is moderate over the CeO2(110) surface with the values of 0.37 eV and 0.59 eV, respectively. The activation barriers for the recombination of H-and H+in the formation of H2and the migration of H-from the Ce cation to the O anion are comparable over CeO2(110) surface with the values of 0.22 and 0.15 eV, respectively. Therefore, one can observe abundant H+on the CeO2(110) surface with rare H-anions adsorption at the Ce cation site. Especially on the CeO2(110) surface, some O anions can coordinate with two Ce3+. The Bader charge of this kind of the O anion is high that fewer electrons can be exchanged with H. That’s why the charge of the H ions binding with the O anion is zero or even negative (Table S2).

Figure 4. Potential energy surface diagram (a) and corresponding transition state configurations (b) for the heterolytic H2 dissociation over the defective CeO2(111), (110), and (100) surfaces. The bond distances between the two dissociated H atoms are indicated in Å. Dashed red circles representing oxygen vacancies.

The heterolytic dissociation of H2is most feasible on the CeO2(100) surface with the lowest activation barrier of 0.43 eV among the three considered CeO2surfaces due to the highly exothermic nature of the H2dissociation over the CeO2(100) surface (ΔH=-0.22 eV). After, H anion adsorbed at the Ce cation site can diffuse to the O anion site exothermically with a low activation barrier of 0.32 eV. Therefore, only H+can be found on the CeO2(100) surface which has similar catalytic behavior but more active than the CeO2(111) surface at relatively high temperatures. Our DFT calculations clearly revealed the difficulty of H2activation on the CeO2(111) surface but much feasible on the open CeO2(110) and (100) surfaces. Our result is in line with other theoretical and experimental works that the H2dissociation is difficult on the stoichiometric CeO2(111) surface[20,22,40,50]. A large number of stable hydroxyl groups can be generated on the three different stoichiometric CeO2surfaces for the H2activation. Therefore, a high reaction temperature is required for the hydrogenation of alkynes by the usage of H cation as the hydrogen resources. The hydrogen-to-acetylene ratio is usually larger than 20 that a large number of H2molecules can be activated providing enough H+species used for acetylene hydrogenation reactions[19,20].

3.2 Structure sensitivity of H2 activation over defective CeO2 surfaces

Oxygen vacancies often present on the CeO2surface inevitably under the acetylene hydrogenation reaction conditions. We studied the heterolytic H2dissociation mechanism over defective CeO2(111), (110), and (100) surfaces to reveal the oxygen vacancy effect on H2dissociation activity and corresponding surface H+/H-species distributions. The oxygen vacancy concentrations are different over (111), (110), and (100) surfaces with the values of 25% and 12.5%, 11.1%, respectively. The corresponding concentrations of Ce3+over the three surfaces are around 12.5% to 8.3%, which are similar to the reduced CeO2materials prepared in experiments[36,51-53]with the Ce3+concentration of 10% to 15%. The calculated optimal potential energy diagram for heterolytic H2dissociation and corresponding transition state configurations over the three defective CeO2surfaces are shown in Figure 4. The configurations of heterolytic dissociative of H2and corresponding structural information are shown in Figure S2 and Table S4, respectively. Similar to stoichiometric CeO2surfaces, H2still adsorbs weakly on the three considered defective CeO2surfaces with the adsorption energies of ～ -0.10 eV. However, the heterolytic H2dissociation is accelerated by the introduction of oxygen vacancies in CeO2surfaces from both thermodynamic and kinetic aspects (Figure 4). From the thermodynamic data in Table S4, we can see that on the defective CeO2-x(111), (110), and (100) surfaces, H-prefers to adsorb at the oxygen vacancy site, whereas H+prefer to bound with the surface oxygen atoms rather than subsurface oxygen atoms (Figure S2). For the defective CeO2-x(111) surface, the heterolytic H2dissociative adsorption energy is -0.92 eV, which is much lower than that on the stoichiometric CeO2(111) surface by 1.73 eV. As compared with stoichiometric CeO2(110) and (100) surfaces, the defective surfaces have lower heterolytic H2dissociative adsorption energies of -0.67 eV and -0.68 eV, which are lower than those on the stoichiometric surfaces by 0.98 eV and 0.34 eV, respectively. Heterolytic H2dissociation is structure sensitive and the presence of oxygen vacancy can enhance the heterolytic H2dissociation thermodynamically.

Figure 5. The linear scaling relationship between H+(a)/H-(b) binding energy and oxygen vacancy formation energy on the three reduced CeO2-x surfaces.

The activation barrier of heterolytic H2dissociation forming H+and H-species is reduced by the introduction of oxygen vacancy in CeO2(111), (110), and (100) surfaces by 0.79, 0.23, and 0.30 eV, respectively. Therefore, the heterolytic H2dissociation can be accelerated on the defective CeO2surfaces as compared with the stoichiometric ones. However, H-adsorption at the Ce cation site migrates to the O anion nearby has higher or comparable reaction barriers on the defective CeO2(111), (110), and (100) surfaces as compared with those on the perfective ones by 1.49, -0.01, and 0.83 eV, respectively. This can be attributed to the less exothermic nature of the hydrogenation migration with the introduction of oxygen vacancy in stoichiometric CeO2surfaces.

Different from the observations of merely H+on the three stoichiometric CeO2surfaces, H+and H-can be found at a low temperature on the partially reduced CeO2-x(111) and CeO2-x(100) surfaces due to the higher activation barriers for H-at the Ce cation site migration to O anion site forming H+species of 1.63 and 1.15 eV, respectively. The heterolytic H2dissociation and H-migration from Ce to O forming H+have similar activation barriers of 0.53 and 0.51 eV over CeO2-x(110) surface, respectively. However, the potential energy surface goes down for H2dissociation forming two H+species with low activation barriers that only H+species can be found on the defective CeO2(110) surface. The different catalytic behaviors of the H2dissociation between (110) and (111)/(100) surfaces can be attributed to the different bond strength of H+/ H-on defective CeO2surfaces.

As stated above,the heterolytic H2dissociation is sensitive to the surface structures of CeO2and oxygen vacancy plays an important role in the heterolytic dissociation of H2and migration of H atom from the Ce cation site to the O anion site. This can be originated from different oxygen vacancy formation energies over different stoichiometric CeO2surfaces. The defective CeO2surfaces have higher activities for the heterolytic H2dissociation than H species migration from the Ce cation to the O anion except for the perfect and defective CeO2(110) surface sharing almost the same activities on the H migration. As a result, the surface stably adsorbed H species vary greatly over different stoichiometric and defective CeO2surfaces that only H+can be found on perfect CeO2surfaces and defective CeO2(110) surface but both H+and H-can be observed on the defective (111) and (100) surfaces. The different forms of H species might display distinct catalytic behaviors for acetylene hydrogenation reactions.

To further reveal the reason for the surface sensitivity of the H2dissociation, we calculated the binding energies of H+and H-on stoichiometric and defective CeO2surfaces. In Figure S3, the sum of the isolated adsorption energies of H+and H-is close to the coadsorption energy of the H+& H-configuration. Therefore, there is no obvious interaction between the two H+and H-ions due to the long distance between them for the coadsorption of H+& H-species. By carefully evaluating the calculated H+& H-binding energies and oxygen vacancy formation energies, we find a linear scaling relationships between H+/H-binding energies and oxygen vacancy formation energies (Figure 5). Generally, the binding strength of H+bounded to the O anion decreases by increasing the oxygen vacancy formation energy. The calculated oxygen vacancy formation energies on (111), (110), and (100) are 2.26, 1.67, and 1.17 eV, and the corresponding Bader charge state of O over the three surfaces are -1.09, -0.98, and -0.92, respectively. The stronger interaction between Ce and O will result in a larger Bader charge state of O cation weakening H+adsorption. The same observations can be found on the defective CeO2surfaces(Table S5). The surface sensitivity of homolytic adsorption is strongly correlated with the binding energy of H+, which is determined by the ability of oxygen vacancy formation dependent on the CeO2surfaces. We further calculated hydrogen adsorption over the reduced CeO2-xsurfaces with two oxygen vacancies with the concentration of Ce3+of 25% to 16.6%. The relationship established between the H adsorption energy and the oxygen vacancy formation energy is still valid and universal even considering high Ce3+concentrations in the slab models (Figure S4).

H-adsorption strength has an opposite trend as a function of the oxygen vacancy formation energies (Figure 5(b)). Generally, the harder the surface is to be reduced, the stronger bond can be formed between the Ce cation and the H anion. Compared with perfect surfaces, the Bader charge states of Ce cations over the defective CeO2(111), (110), and (100) surfaces are also reduced to +1.01, +1.91, +2.15, respectively. The weaker charge state of Ce adsorbs H-stronger at the Ce cation site due to the formation of the stronger covalent bond between H and Ce. This finding can be observed over the defective CeO2surfaces with two oxygen vacancies. As shown in Figure S5, it is found that there is no obvious relationship between the hydrogen activation barrier and oxygen vacancy formation energy. Due to the oxygen vacancies on the defective CeO2surface, the energy barrier for hydrogen dissociation is significantly reduced, and the surface is also more difficult to be further reduced. If the data is not classified, it does not comply with the law that the higher the oxygen vacancy formation energy, the higher the dissociation energy barrier. If discussed separately, it can be seen that the activation barriers for hydrogen activation are closely related to oxygen vacancy formation energy, that is, the higher oxygen vacancy formation energy, the higher hydrogen activation barrier on the stoichiometric and defective CeO2surfaces.

4 Conclusions

In the present work, we studied H2activation over the stoichiometric and defective CeO2(111), (110), and (100) surfaces. The heterolytic H2dissociation pathway is dominant on stoichiometric and defective CeO2surfaces. We identified that the heterolytic H2dissociation is difficult on the stoichiometric CeO2(111) surface corroborated by previous experimental measurements. CeO2(100) and (110) surfaces are more active than (111) surfaces for the heterolytic H2dissociation. Only the H+adsorption at the O anions can be found on all three stoichiometric CeO2surfaces. The presence of oxygen vacancies can promote the heterolytic H2dissociation over defective CeO2surfaces. Both H+and H-can be found on the defective CeO2(111) and (100) surfaces, whereas only H+species can be found on the CeO2(110) surface. The oxygen vacancy formation energy can be considered as a key descriptor for the activity of the heterolytic H2dissociation and the distribution of H+/H-species. Our work provides more insight into H2activation on CeO2-based catalysts, which is pivotal for better catalyst design.

Supplementary data

Supplementary data are available at J.Univ.Sci.Tech.China online.

Acknowledgments

This work was supported by the Key Technologies R&D Program of China (2017YFB0602205, 2018YFA0208603), the National Natural Science Foundation of China (91945302), the Chinese Academy of Sciences Key Project (QYZDJ-SSW-SLH054), the start-up funds of University of Science and Technology of China (KY2060000171), USTC Research Funds of the Double First-Class Initiative(YD2060002012) and high-performance computational resources provided by University of Science and Technology of China (http://scc.ustc.edu.cn).

Conflict of interest

The authors declare no conflict of interest.