Reducible Inverse CeO_x-Based Catalyst as a Potential Candidate for Electroreduction

Abstract

The inverse metal oxide/metal catalyst is very suitable for electrochemical reaction due to unique catalytic properties of metal oxide with small size and good conductivity of metal. To clarify the potential applications of inverse catalyst in electrochemistry, especially for reducible oxides, an inverse CeO_x/Ag(111) model electrocatalyst was constructed and investigated by Density Functional Theory (DFT) for CO₂ electrochemical reduction. It is found that Ag atoms acting as an electron donor, can partially reduce Ce⁴⁺ to Ce³⁺ in the supported CeO_x cluster leading to the formation of interfacial Ce³⁺ active sites, which could promote the adsorption and reduction of CO₂. As expected, all elementary reaction involved in the CO₂ electrochemical reduction are more facile on CeO_x/Ag(111) than pure Ag catalyst. Besides, the generation of CH₃OH and CH₄ is favored on CeO_x/Ag(111), whereas the formation of CO, CH₂O and H₂ is obviously suppressed. More importantly, the weak interaction between H₂O and CeO_x cluster is beneficial for the desorption of OH intermediate, which makes the regeneration of the catalyst become easier and result in a great recyclability. All those results demonstrate that CeO_x/Ag(111) is a potential excellent electrochemical catalyst.

1. Introduction

Oxide supported metal nanoparticles catalysts are widely used in heterogeneous catalysis and play an important role in industrial production. In the study of the adsorption of H₂ and CO on the oxides supported noble metal catalysts, the different adsorption capacities of H₂ and CO on pre-reduced catalysts at different temperatures was observed, which was attributed to strong metal-support interactions (SMSI) [1]. Long-term studies have shown that this kind of interaction exists only in reductive metal oxides carrier (CeO₂ [2], FeO_x [3]), which is not only act as supports to prevent the sintering of metal nanoparticles but also have a great influence on the catalytic activity of the catalysts. Recently, it is found that the inverse catalysts also exhibit excellent catalytic performance towards some reactions. In all cases, inverse catalysts exhibit strong interaction between the metal and oxide (SMSI) through sharing part of oxygens [4,5]. Density Functional Theory (DFT) research reveals that all Ce atoms are Ce³⁺ in CeO_x/Metal (Cu and Ag) catalyst due to the covalent bond between CeO_x cluster and metal carrier, leading to the formation of mixed oxides in the interface of the catalyst [6]. Thus, it can be speculated that the unique interface in inverse catalysts will play a significant role in the reaction.

At present, studies have shown that the inverse catalysts can catalyze many important reactions such as dissociation of O₂ [5], water-gas shift reaction [7] and CO₂ reduction to produce methanol [8]. The hydrogenation of CO₂ on CeO_x/Cu catalyst indicate that CO₂ molecule can be activated when it adsorbed on the interface of the catalyst [8]. This inspires that the Ce-related inverse catalysts may be a potential candidate for electrochemical CO₂ reduction reaction (CO₂RR) to produce useful products such as CH₄, CH₃OH, HCOOH or CH₂O. Recently, the DFT investigation about CO₂RR on (TiO₂)₃/Ag(110) have also been carried out. It is demonstrated that the introduction of metal oxides to a metal electrode provides the interfacial active sites for adsorption and activation of CO₂ and facilitates further reduction process. The metal carrier can guarantee good electrical conductivity, meanwhile the oxide cluster especially those reductive metal oxides can provide an electron rich environment, which can promote the adsorption and activation of reactant. However, it is also pointed out that the strong binding between O-containing species and Ti makes the OH* removal remain challenging [9]. Thus it is very important to design a novel catalyst for facile removal the OH*.

Based on the above analysis, the inverse CeO_x-based catalysts were constructed. For CeO_x/M (M = Cu, Ag, Au) catalyst, different amounts of electrons could be transferred from metal to oxide clusters, in the order of Cu > Ag > Au [6]. Herein, the CeO_x/Ag(111) was employed, since it is relatively stable and its 4f states of Ce³⁺ has high activity [6]. Besides, detailed reaction mechanism of CO₂ electrochemical reduction to CO, HCOOH, CH₄ and CH₃OH has been systematically studied. The presence of Ce³⁺ generated by the SMSI effect between CeO_x and Ag reduces the free energy for the reduction of CO₂. Especially, it is beneficial to generate CH₃OH and CH₄ in thermodynamic. Different from other catalysts, the removal of OH is also facile, due to the weak interaction between H₂O and CeO_x cluster. Those results provide valuable reference to design more efficient catalysts for electrochemical reduction of CO₂.

2. Results and Discussion

2.1. Electronic Characteristics for the CeO_x/Ag(111) Catalyst

A Bader analysis of the charge was performed firstly for CeO_x/Ag(111) catalyst, in order to reveal the electronic structure and interactions between ceria and Ag at the interface of this catalyst. As reported, the Bader charge of Ce in the particle is +2.00 e [5]. While the calculated Bader charge for Ce in the CeO₂/Ag(111) are ~+2.1 e and ~+2.2 e, respectively (

Table 1. Bader charge of the Ce and Ag atoms under the CeO_x cluster of CeO_x/Ag(111).

Bader Charge
Ce	Ag Atoms Under the CeOx Cluster
2.18	0.07	−0.04	0.00
2.12	0.14	0.14	0.10
2.20	0.05	0.11	−0.02
2.11	0.08	0.10	−0.01
2.11	0.14	0.11	0.00
2.21	0.00	−0.01	−0.01
	0.00	0.14	0.01
	0.13	0.12	−0.01
	0.12	0.05	−0.03
	0.11	−0.02	−0.02
	0.10	0.11
	0.11	0.16

). Thus it can be inferred that the formal oxidation states of Ce atoms in the ceria cluster are +4 and +3, respectively [5]. Spin Density Difference (SDD) versus Z direction of the CeO_x/Ag(111) catalyst is plotted in Fig. 1a, the isosurface of SDD is plotted in Figure 1b. From Figure 1a,b, obvious electron accumulation is founded on Ce atoms at the edge of a cluster. Therefore it can be concluded that the electrons are mainly located on the edge of the CeO_x cluster. In other words, single electrons are concentrated in certain areas of ceria cluster, the computational results show that only parts of Ce atoms which located on the edges of ceria cluster have been reduced (Figure 1b). All those Ce atoms can be considered as Ce³⁺. In addition, the Electron Density Difference (EDD) versus Z direction of the CeO_x/Ag(111) catalyst is plotted in Figure 2a, the isosurface of EDD is plotted in Figure 2b. From Figure 2, it was found that EDD indeed increased on CeO_x cluster and decreased on surface Ag atoms, which indicates that there is a charge transfer from the Ag slab to the ceria cluster. And the electrons are mainly coming from the surface Ag layer, which located under the CeO_x cluster, resulting in an oxidation of the Ag atoms and a reduction of the CeO_x cluster.

2.2. CO₂ Adsorption and Activation

The adsorption and activation of CO₂ is considered as the first step for CO₂ reduction. In the initial CO₂ adsorption configurations, no matter where CO₂ was placed (on the interface of CeO_x/Ag(111), Ag and CeO_x, linear or bent structure). Only two stable structure of CO₂ adsorption (denoted as CO₂*, * = CeO_x/Ag(111)) were reached in the final (Figure 3 and Figure 4c). CO₂ is linearly adsorbed on the edge (Ce³⁺) or corner (Ce⁴⁺) of the CeO_x cluster using its O atom. The adsorption energy of CO₂ for those two structures is 0.42 eV and 0.40 eV, respectively. The shortest distance between O atom of CO₂ and Ce atom of CeO_x cluster is 3.03 Å and 3.04 Å. Gradient isosurfaces of reduced density gradient (RDG) versus sign(λ₂)/ρ (Figure 3a) [10] plot shows that there are Van Der Waals interactions between O-Ce and C-Ag for CO₂/*(Ce³⁺) simultaneously. Whereas, in CO₂/*(Ce⁴⁺), only O-Ce interaction has been found (Figure 3c). Spikes near zero in the corresponding scatter plot (Figure 3b,d) also reflect there exist weak interactions in CO₂/*(Ce³⁺) than CO₂/*(Ce⁴⁺). The synergism by Ag and CeO_x in CO₂/*(Ce³⁺) configuration may be beneficial to the subsequent hydrogenating of CO₂. Therefore CO₂/*(Ce³⁺) is chosen as the initial structure in this work.

2.3. Formation of Formic Acid and Methanol

The CO₂ adsorption configuration shown in Figure 3a is set to be the starting point for further reduction. The first hydrogenation of CO₂* may occur at either the O or C atom, forming carboxyl (HOCO*) or formate (HCO₂*) species, respectively. Herein, the routes through both HOCO* and HCO₂* intermediates were investigated. The Gibbs free energy diagrams for all possible products involved in CO₂ reduction process on CeO_x/Ag(111) were also examined.

As mentioned above, hydrogenation of CO₂ could occur either on oxygen or carbon atom of CO₂. HOCO* or HCO₂* intermediates are eventually formed at the interface of CeO_x and Ag. The HCO₂* (Figure 4d) intermediate is formed with a free energy release of −1.14 eV (Figure 4a). While the HCO₂* intermediate can change to bidentate form using its O atoms with CeO_x cluster, which is endothermic with free energy barrier of 0.73 eV. When the hydrogenation of CO₂ occurs on the O atom (Figure 4b), forming HOCO* (Figure 4f) with a free energy barrier of 0.67 eV. This intermediate is bind to the catalyst using its C (C-Ag = 2.25 Å) and O (O-Ce = 2.56 Å) atom simultaneously. The Ag atom which bonds to C atom obviously migrates up from Ag surface. Obviously, the HCO₂* formation is thermodynamically preferable over HOCO*.

When HCO₂* or HOCO* intermediate is formed on the catalyst. Formic acid as an important product for CO₂ reduction can be generated firstly by hydrogenation of HCO₂* or HOCO* intermediate. For the hydrogenation of HOCO*, this is thermodynamically preferable. The formation of cis-HOHCO* and trans-HOHCO* (Figure 4h,i) is exothermic by 0.72 and 0.82 eV, respectively. For the hydrogenation of bidentate HCO₂*, subsequent reaction process can be achieved through two different ways. (a) The second hydrogen is bonded with carbon of HCO₂*, a H₂CO₂* intermediate is formed. The free energy change for this process is 0.42 eV endothermic. (b) The hydrogenation occurs on oxygen atom of HCO₂*, HOHCO* intermediate can be generated. There are also two possibilities here, (1) formation of cis-HOHCO* (0.47 eV, endothermic), (2) formation of trans-HOHCO* (0.37 eV, endothermic). The free energy change for the rate-limiting step of formic acid formation from HOCO* and HCO₂* intermediate are 0.67 and 0.73 eV, respectively. Obviously, HOCO* reaction path is slightly more facile for the formation of formic acid.

As shown in Figure 4, trans-HOHCO* can further hydrogenate by (H⁺ + e⁻) to generate HOH₂CO* intermediate. CH₃OH and O atom co-adsorbed on the catalyst can be generated through the further hydrogenation of HOH₂CO. Next, the formed CH₃OH can be desorbed from the surface of the catalyst, leaving only OH*. Thermodynamically, the formation of CH₃OH through HOH₂CO* is uphill in free energy by only 0.05 eV. In general, once HOHCO formed, it is very easy to convert into CH₃OH, although more hydrogenation steps are needed for this process.

2.4. Formation of Methane

With the aim of comparison, both HCO₂* and HOCH* reaction path have been plotted in Figure 5. When the hydrogenation of HOH₂CO* firstly occurs on the carbon atom, methoxy (CH₃O*) and hydroxyl (OH*) intermediates are formed and co-adsorbed on the catalyst. This process is endothermic with free energy change of 0.71 eV. Further hydrogenation for the formation of H₂O and CH₄ are all uphill with free energy barrier of 0.54 eV and 0.20 eV, respectively. Obviously, the total free energy change from CO₂* to CH₃OH and CH₄ are almost the same. Thus it is speculated that both CH₃OH and CH₄ are easy to form on the interface of this catalyst, as both of process are free energy downhill from the CO₂*. However, it is noteworthy that the formation of CH₄ needs more steps than that of CH₃OH. Thus, CH₃OH formation is probably facile than the formation of CH₄. The most endothermic steps for the formation of HOCO* or trans-HOHCO* are observed, which will be the potential-limiting steps for the formation of CH₃OH and CH₄ on the CeO_x/Ag(111) catalyst.

2.5. Formation of Carbon Monoxide

Although CO₂ only weakly binds to CeO_x/Ag(111), less energy is required for the formation of HOCO* from CO₂* than that on pure Ag [9], on which more than 1.2 eV is needed for the hydrogenating of CO₂ to form HOCO*. The formation of CO* (Figure 6b) from HOCO* is endothermic on CeO_x/Ag(111) with a free energy change of 0.69 eV. However, further hydrogenating of HOCO* is exothermic with a free energy change of −0.82 eV. Clearly, further hydrogenating of HOCO* to form HOHCO* is more facile on CeO_x/Ag(111) than the formation of CO (0.18 eV, CO*). Simultaneously, CO adsorption is quite stable on CeO_x/Ag(111). As CO desorption is endothermic with a free energy change of 0.83 eV. Consequently, CO* may be further reduced on CeO_x/Ag(111) to other products rather than desorbed from the catalyst. The reduction pathway and the free energy profile of the corresponding intermediates from CO* have been mapped out (Figure 6). As shown in Figure 6, apart from the formation of HCO* (0.25 eV), all other processes after HCO* formation are downhill in free energy.

As shown in Figure 6b, CO* is bonded to Ag only through C atom. Bond length of C-O and C-Ag are 1.15 and 2.15 Å, respectively. This indicates that CO is chemically adsorbed on the Ag surface and has been activated. With the formation of CHO*, O atom is weakly bonded to the Ce atom of the CeO_x cluster, distance of C-Ag and O-Ce are 2.20 and 2.60 Å. There is an obvious deformation of Ag surface. The Ag atom, which interacts with C atom, is moved upward from the Ag surface. Gradient isosurfaces of RDG and sign(λ₂)/ρ [10] plot also shows clearly a strong attractive interaction between carbon and Ag (Figure S1). With further hydrogenating of CHO*, the formed CH₂O* bonds to CeO_x/Ag(111) only through O-Ce interaction (C-Ag 4.28 Å, O-Ce 2.71 Å). The C-Ag in CHO* is completely broken upon formation of CH₂O*. The same situation occurs in the formation of CH₃O*, in which methoxyl bonds to the CeO_x/Ag(111) only through O-Ce (C-Ag 6.14 Å, O-Ce 2.13 Å). The high oxygen affinity of CeO_x/Ag(111) keeps the O atom of CH₃O* bound to Ce strongly. And this, in turn, will cause the breaking of the C-O bond of CH₃O*. Eventually, CH₄ can be formed in the subsequent hydrogenation step.

2.6. Formation of Acetaldehyde

The formation of formaldehyde can be achieved through (a) hydrogenation of CO* and HCO*, which is mentioned above in 2.5, or (b) dehydrogenation of methoxy in the (CH₃O_OH)* intermediate. As plotted in Figure 7, Breaking C-O bond of HOH₂CO* intermediate can form CH₂O and OH co-adsorption intermediate (Figure 7c). With the formation H₂O, CH₂O* is ultimately formed. Free energy change for this process is 0.63 eV, which is slightly lower than the formation of HOCO*, thus the rate-limiting step for this reaction path is also the HOCO* generation. Besides, the dehydrogenation of CH₃O* intermediate can also lead to the formation of CH₂O* (0.88 eV). Obviously, this reaction path is relatively difficult to achieve. Compared with the formation of CH₃OH and CH₄, more energy is needed for the generation of CH₂O. Thus CH₂O formation is unfavorable on this catalyst.

To sustain the catalytic cycle, no matter in what kind of reaction path, the O* species has to be cleared through consecutive hydrogenating to OH* and further to H₂O. The hydrogenating of OH* to H₂O is endothermic, with a free energy barrier of 0.70 eV, which indicate that OH is the most stable intermediate species at U = 0 V. This makes OH removal become the free energy limiting step. However, the free energy change is still lower than the desorption energy (1 eV) needed on TiO₂/Ag system [9] or pure metals (such as Pt (0.75 eV)) [11]. The free energy change needed for the formation and removal of H₂O is only slightly higher than the energy required for the generation of HOCO* intermediate (0.67 eV) on CeO_x/Ag(111) catalyst. The weak interaction between H₂O and CeO_x cluster is beneficial for the formation of H₂O (the distance of O-Ce has enlarged from 2.15 Å to 2.72 Å, for OH* and H₂O* intermediate), which makes the regeneration of the catalyst become easier, resulting in a great recyclability.

Finally, the hydrogen evolution reaction (HER), which is the main competitive reaction of CO₂RR, has been calculated. The computational result shows that the free energy change for the formation of H₂ is 1.22 eV, which is far higher than the rate-limiting step of CO₂RR on CeO_x/Ag(111) (Figure 7e). Based on all those advantages mentioned above, it is reasonable to believe that the CeO_x/Ag(111)is a potential excellent catalyst for CO₂ electroreduction.

The above analysis shows that CH₃OH and CH₄ are the main products on the surface of CeO_x/Ag(111) catalyst. In order to change into exothermic and spontaneous process for all the reaction steps, the free energy diagram with an external potential of −0.67 V (RHE) was depicted and shown in Figure 8. The potential is selected based on the potential-limiting step, that is the formation of HOCO*. Under this applied potential (U = −0.67 V vs RHE), the free energy change of all steps involving the proton-electron pairs will be corrected according to Equation (2). Under this potential, all steps following HOCO* formation become downhill on the free energy diagram.

When hydrogenation of CO₂ occurs on the oxygen atom, HOCO* is the key intermediate and plays a central role in the entire mechanism. Further hydrogenating HOCO* leads to the formation of HOHCO* (trans-formic acid) intermediate. When hydrogenation of CO₂ occurs on the carbon atom, trans-HOHCO becomes the critical intermediate. A HOH₂CO* intermediate can be formed through the hydrogenation of trans-HOHCO. Breaking different C-O bond of HOH₂CO* will lead to the formation of CH₃OH, CH₂O or an adsorbed methoxy (CH₃O*) species on the Ce site and an OH*. Subsequently, CH₃O* can be hydrogenated to produce CH₄ or CH₂O. And the catalyst will recover its original state with the desorption of H₂O by the hydrogenation of OH*. Based on the free energy profile for the HOCO* path shown in Figure 8, CH₃OH desorption from the catalyst release more energy (−0.69 eV) than the formation of CH₃O* (−0.13 eV), whereas, the CH₃O* intermediate is more stable (0.19 eV lower) than CH₃OH + OH*. This result indicates that CH₃OH formation is less favorite than CH₄ thermodynamically. However, CH₃OH formation requires fewer proton reduction and hydrogenation steps, therefore, both CH₃OH and CH₄ are likely to be generated.

3. Computational Details

Spin-polarized calculations are conducted with the Quickstep [12] module of the CP2K program package [12,13], which based on the Gaussian and plane waves formalism [14]. The exchange-correlation energy was described by the generalized-gradient approximation (GGA) with spin-polarized revised PBE functional (revPBE). In order to investigate the influence of dispersion, the structures and energies for all the stationary points included in our work are evaluated at the DFT-D3 level with the approximation suggested by Grimme [15] added to the revPBE calculated energy. Fronzi et. al have carried out studies on water adsorption on the stoichiometric and reduced CeO₂(111) surface [16]. They have pointed out that the DFT+U method has some inherent defects, such as no unique way to decide on what value to use for U; weather DFT+U method indeed provides a systematic improvement of the energetics of a system compared to plain DFT and so forth. Moreover, theoretical studies for the CO oxidation on Au_xCe_1−xO₂ catalyst also pointed out there is no hard evidence that GGA gives unusually large errors when used to calculate the energy differences [17]. In short, it is impossible to find a universal method which gives a satisfactory description of both the electronic structure and the energetics. Because this article focuses on the energy changes of the CO₂RR reaction process, Therefore DFT+U method is not considered in this paper. The wavefunctions were expanded in an optimized double-ζ Gaussian basis sets [16,18] with a cutoff energy of 500 Rydberg [14]. Core electrons have been modelled by scalar relativistic norm-conserving pseudo potentials [19] with 12, 6, 1, 4 and 11 valence electrons for Ce, O, H, C and Ag, respectively. Brillouin zone integration is performed with a reciprocal space mesh consisting of only the gamma point. Thermochemistry calculation is implemented by TAMkin [20], a post-processing toolkit for normal mode analysis. All other electronic property analyses deal with Multiwfn software, a program for realizing electronic wavefunction analysis [21].

The following equation was used to calculate the adsorption energy of an adsorbate

ΔE_ads = E_{(adsorbate/(CeOx/Ag(111))} − E_(adsorbate) − E_{(CeOx/Ag(111))}

(1)

where E_{(adsorbate/(CeOx/Ag(111))} is the total energy of an adsorbate bound to CeO_x/Ag(111), E_(adsorbate) is the total energy of the isolated adsorbate and E_{(CeOx/Ag(111))} is the total energy of bare CeO_x/Ag(111). A negative value of ΔE_ads corresponds to an exothermic adsorption process.

Spin Density Difference (SDD) and Electron Density Difference (EDD) for CeO_x/Ag(111) is computed by the following equation

SDD(EDD) = SD(ED)_{(CeOx/Ag(111))} − SD(ED)_(CeOx) − SD(ED)_(Ag(111))

where SD(ED)_{(CeOx/Ag(111))} is the Spin Density (Electron Density) of the catalyst, SD(ED)_(CeOx) and SD(ED)_(Ag(111)) are Spin Density (Electron Density) of CeO_x and Ag(111) respectively, the cartesian coordinates used for the computation of CeO_x and Ag(111) are consistent with that in the CeO_x/Ag(111).

The Gibbs free energy diagrams for CO₂ reduction along different pathways were calculated with reference to the computational hydrogen electrode (CHE) proposed by Nørskov et al. [22]. The free energy of each species is obtained using TAMKin (Table S1) [20]. The free energy change of each elementary step at external potential U will be shifted by eU as

ΔG(U) = ΔG(U = 0) + eU

(2)

where e is the electronic charge. According to Nørskov et al., the thermodynamic activation barrier equals the largest of the free energy differences [22].

A Ce₆O₁₃ cluster, which is the most stable structure found by other theoretical work [6], is selected and supported on an Ag(111) metal surface described by means of 7 × 7, five atomic layers thick of Ag slab (245 Ag atoms) and a vacuum region of 20 Å between repeated slabs. In our calculations, the atoms in the bottom two layers were fixed at their bulk position and those in the top four layers together with the supported CeO_x cluster and the adsorbates were allowed to relax.

4. Conclusions

The CeO_x/Ag(111) as a model of the metal supported metal oxide catalyst is constructed for electroreduction of CO₂. It is found that the transfer of electrons from top layer Ag atoms to CeO_x cluster leads to the reduction of Ce⁴⁺ to Ce³⁺ at the interface. Those interfacial Ce active sites could promote the adsorption and further reduction of CO₂. The dominating pathway on the CeO_x/Ag(111) catalyst to produce CH₃OH and CH₄ is via the HOH₂CO* intermediate, a key intermediate along both HOCO* or HCO₂* pathway. Once the HOH₂CO* is generated, the formation and desorption of CH₃OH and CH₄ is nearly thermal-neutrality. Nevertheless, the formation of acetaldehyde and CO is unfavorable. It is worth noting that the moderate binding of O-containing species makes OH* removal facile, resulting in a great recyclability of catalyst. The main competitive reaction (HER) has also been tested. Higher energy is needed for the formation of H₂ than that for the rate-limiting step of CO₂RR. Those results provide valuable reference to design more efficient catalysts for electrochemical reduction of CO₂.