Theoretical Insights into Synergistic Effects at Cu/TiC Interfaces for Promoting CO₂ Activation

Abstract

The adsorption behaviors of CO₂ at the Cu_n/TiC(001) interfaces (n = 1–8) have been investigated using the density functional theory method. Our results reveal that the introduction of copper clusters on a TiC surface can significantly improve the thermodynamic stability of CO₂ chemisorption. However, the most stable adsorption site is sensitive to the size and morphology of Cu_n particles. The interfacial configuration is the most stable structure for copper clusters with small (n ≤ 2) and large (n ≥ 8) sizes, in which both Cu particles and TiC support are involved in CO₂ activation. In such a case, the synergistic behavior is associated with the ligand effect introduced by directly forming adsorption bonds with CO₂. For those Cu_n clusters with a medium size (n = 3–7), the configuration where CO₂ adsorbs solely on the exposed hollow site constructed by Cu atoms at the interface shows the best stability, and the charger transfer becomes the primary origin of the synergistic effect in promoting CO₂ activation. Since the most obvious deformation of CO₂ is observed for the TiC(001)-surface-supported Cu₄ and Cu₇ particles, copper clusters with specific sizes of n = 4 and 7 exhibit the best ability for CO₂ activation. Furthermore, the kinetic barriers for CO₂ dissociation on Cu₄- and Cu₇-supported TiC surfaces are determined. The findings obtained in this work provide useful insights into optimizing the Cu/TiC interface with high catalytic activation of CO₂ by precisely controlling the size and dispersion of copper particles.

1. Introduction

The explosive growth of industry and the massive combustion of fossil fuel have caused the rapid increase in the concentration of CO₂ in the atmosphere, leading to a series of environmental issues. Therefore, it is highly desired to develop novel strategies for CO₂ capture and utilization technologies. Especially, the conversion of CO₂ to alcohols or other valuable hydrocarbon compounds has attracted great attention in recent years,^1,2 which provide an important approach to achieve the recycling of the released CO₂. However, this is a challenging task due to the chemical inertness of CO₂, and catalysts with high activity and selectivity are required to realize CO₂ conversion.

In the past few years, extensive works have been carried out to investigate the performances of various catalysts for CO₂ activation and conversion.³⁻¹⁹ Among them, transition-metal carbides (TMCs) have been regarded as very promising catalysts since they often exhibit catalytic properties similar to precious metals for many reactions.²⁰ At the same time, TMCs also are good candidates to design cost-effective catalysts due to their abundance and relatively low cost with a high melting point and good thermal stability.²¹ It has been found that CO₂ binds strongly on many TMC surfaces, which will facilitate the cleavage of the C–O bond of CO₂, suggesting that TMCs are effective catalysts for CO₂ activation.^{12,14,15,22−24} As a typical example, experiments and density functional theory (DFT) calculations indicate that CO₂ is adsorbed well on different molybdenum carbide surfaces, and the dissociation of CO₂ can occur at low temperatures.^{12,14,22−24} Besides molybdenum carbide, the low-index surfaces of other TMCs, including TiC, ZrC, HfC, NbC, and TaC, can also adsorb and activate the CO₂ molecule with considerable binding strength.^8,12,25 More importantly, TMCs are excellent supports for metal particles,²⁶ and some recent investigations have revealed that the catalytic activity of TMCs can be obviously prompted by depositing metal on the surface.²⁷⁻³⁰ For instance, compared with the pristine Ni(111) and TiC(001) surfaces, the TiC-surface-supported Ni particles exhibit a stronger binding strength for CO₂ adsorption, which will reduce the energy barrier for the scission of the C–O bond.²⁷ Similarly, after deposing Cu particles on the Mo₂C surface, the CO₂ conversion rate is increased by about 25–35% compared to the bare Mo₂C surface, and the yield of methanol on Cu/Mo₂C is also greater than that of Mo₂C(001) and Cu(111).²⁹ DFT calculations confirm that CO₂ molecules strongly anchor on Cu₄ and Au₄ clusters supported on the TiC(001) surface, and the rate of methanol formation correlates well to the CO₂ binding strength with a sequence of Cu₄/TiC(001) > Au₄/TiC(001) > Cu/ZnO(001) > Cu(111).¹⁵ Therefore, the depositions of small Au and Cu clusters on TMC surfaces are alternative catalysts for the hydrogenation of CO₂ to methanol. A typical research carried out by Rodriguez and co-workers¹⁶ indicates that small Au, Cu, and Ni particles supported on TiC(001) have high activity for CO₂ hydrogenation, and using Cu₄/TiC(001) as a prototype, DFT calculations prove that arising from the strong metal–support interaction, the binding strength of CO₂ is obviously higher than on the pristine TiC(001) surface. However, it is worth noting that besides the type of metal particle, the catalytic properties of metal/TMCs show a significant dependence on the size of the metal cluster. The methanol synthesis from CO₂ hydrogenation over the Cu/TiC(001) surface with different coverage of the admetal has been investigated experimentally, and the maximum activity is observed when the coverage of Cu is 0.1 ML.³¹ The scanning tunneling microscopy (STM) images indicate that at this small coverage, the heights and the diameters of most copper particles are about 0.2 and 0.3–0.5 nm, respectively. A similar size effect on CO₂ conversion has also been found for Cu particles deposited on various oxide supports, such as Al₂O₃³² and CeO₂,³³ in which the size-selected Cu₄ cluster exhibits the highest CO₂ hydrogenation activity. The above results illustrate that the activation and conversion of CO₂ on both metal/TMC and metal/oxide catalysts can be tuned by controlling the size of metal clusters.

It is well known that the activation of CO₂ is an essential prerequisite for its utilization, and the generated CO₂^δ− species with a bent configuration is an important precursor to realize the conversion of CO₂ into organic products. Although the CO₂ conversion on Cu/TMC surfaces has been investigated theoretically,²⁹ the existing research are performed by assuming that a small Cu₄ cluster is supported on the TMC surface, and there is still a lack of systematic investigation on the relationship between CO₂ activation and the size of the copper particle. Especially, detailed knowledge about how change in the size of Cu particles affects the nature of the Cu/TMC interface and the resulting influences on the CO₂ activation are still poorly understood. In this paper, we report a detailed investigation on a series of Cu_n clusters with different sizes and morphologies supported on the TiC(001) surface using DFT calculation, in which the number of copper atoms varies from 1 to 8. Our results reveal a strong size dependence of CO₂ activation on the Cu_n/TiC(001) surface, and the preferred configuration of CO₂ adsorption is sensitive to the size and morphology of the Cu_n cluster. Unlike findings in many metal/oxide surfaces that CO₂ usually favors the sites at the interface,⁴ the exclusive chemisorption of CO₂ on the metal particle is predicted to be the most stable structure for certain copper clusters. Especially, Cu_n clusters with special sizes of n = 4 and 7 exhibit a superior ability for CO₂ activation. Furthermore, the mechanism of the synergistic effect between the Cu_n cluster and the TiC support in promoting CO₂ activation is discussed. The results obtained in the present work are of great significance for optimization of Cu_n/TiC via precise controlling of the number of Cu atoms in the particle, which is the key for designing highly efficient catalysts of CO₂ activation and conversion.

2. Results and Discussion

2.1. Configurations of Cu_n Clusters in the Gas Phase

We first study the ground-state configurations of Cu_n clusters in the gas phase according to the results reported in previous theoretical works,³⁴⁻³⁷ and the optimized structures obtained here are displayed in Figure 1.

Figure 1.

Optimized structures of Cu_n clusters in the gas phase. The Cu–Cu bond length (in Å) and the binding per atom (E_b) (in eV) are shown, in which E_b is defined as E_b = [n × E(Cu) – E(Cu_n)]/n, where E(Cu_n) and E(Cu) represent the energies of Cu_n clusters at the ground state and the copper atom itself, respectively.

For the Cu₂ dimer, it has a linear structure with a bond length of 2.219 Å (Figure 1a), and the calculated binding energy per atom (E_b) is 1.13 eV. The copper trimer (Cu₃) exhibits an equilateral triangle configuration (Figure 1b), and the Cu–Cu bond length and E_b are 2.234 Å and 1.22 eV, respectively. For the ground state of the Cu₄ cluster, a planar rhombic configuration is obtained (Figure 1c). In this structure, the side length and the short diagonal of the rhombus are 2.373 and 2.264 Å, respectively. As for Cu₅ and Cu₆ clusters, they still favor planar two-dimensional (2D) configurations, which are isosceles trapezoid (Figure 1d) and equilateral triangle (Figure 1e). The values of E_b for Cu₄, Cu₅, and Cu₆ clusters are 1.60, 1.74, and 1.94 eV, respectively. By contrast, three-dimensional (3D) structures are the global minimums for those clusters larger than Cu₆. For example, the Cu₇ cluster exhibits a pentagonal bipyramid structure (Figure 1f) and a dodecahedral configuration (Figure 1g) is predicted for Cu₈. Meanwhile, the larger values of E_b = 2.06 and 2.17 eV mean that the Cu–Cu bonds of two clusters become stronger.

Overall, since the Cu–Cu bond length and the E_b value are increased with the increase in the number of copper atoms (Figure 1), it can be expected that as the size of the Cu_n cluster grows, the length of the Cu–Cu bond and E_b will gradually approach the corresponding values of the copper bulk, which are 2.570 Å and 3.73 eV, respectively. Moreover, the trends of bond length as well as E_b values are consistent with the results obtained in other theoretical investigations.³⁵⁻³⁷

2.2. Structures of Cu_n Clusters Supported on the TiC(001) Surface

Based on the structures derived from ab initio molecular dynamics (AIMD) simulations, the most stable geometries of the Cu_n cluster deposited on the TiC(001) surface (Figure 2) are determined after extensive calculations. Table 1 lists the average lengths of Cu–Cu and Cu–C bonds, as well as the calculated binding energy (E_b) of each cluster. For the single Cu atom (Figure 2a), it is adsorbed on the top site above the surface carbon atom with a Cu–C bond length of 1.839 Å and E_b = 2.57 eV. For the Cu₂ dimer, it tends to form bonds with two adjacent C atoms on the surface (Figure 2b), and the average length of two Cu–C bonds is 1.899 Å. After deposition, the Cu–Cu bond is stretched from 2.219 Å in the gas phase to 2.383 Å, and the value of E_b is 3.42 eV. The structure of Cu₃ on the TiC surface is shown in Figure 2c, and each Cu atom is bonded to one C atom on the support with a length of 1.987 Å. The average distance of the Cu–Cu bond (2.476 Å) also shows an obvious elongation of the Cu–Cu bond. For the deposition of Cu₄, the cluster exhibits a square configuration by forming four Cu–C bonds at the interface (Figure 2d). The average lengths of Cu–Cu and Cu–C bonds are 2.420 and 1.921 Å, respectively, and the value of E_b is 6.42 eV. The structure of Cu₅ supported on the TiC surface is somewhat complicated, and two configurations with 3D and 2D arrangements of Cu₅ are obtained, which are denoted Cu₅(3D) and Cu₅(2D) in Figure 2e,f, respectively. The Cu₅(3D) shows a configuration of a tetragonal pyramid, and four Cu–C bonds are observed at the interface with an average length of 1.913 Å. While for Cu₅(2D), the cluster adopts a highly dispersed arrangement, and there are five Cu–C bonds at the interface with an average length of 1.950 Å. In terms of thermodynamic stability, E_b values of two configurations are nearly identical, both being 7.24 eV, indicating two depositing configurations may coexist on the TiC(001) surface. In the case of Cu₆, a rectangle structure where each Cu is bound to one surface C atom is observed (Figure 2g), and the average lengths of Cu–C and Cu–Cu bonds are 1.955 and 2.460 Å, respectively. For the depositions of Cu₇ and Cu₈ with larger sizes, the configurations of copper particles are completely different from the structures in the gas phase. The most stable configuration of Cu₇ (Figure 2h) can be constructed from the supported Cu₆ cluster by assuming an additional Cu atom at one of two 4-fold hollow sites (Figure 2g). Therefore, the number of Cu–C bonds of Cu₇/TiC is still six, and the corresponding average length is 1.940 Å. The E_b of Cu₇ is calculated to be 8.62 eV. If two 4-fold hollow sites of Cu₆ deposited on TiC are occupied by Cu atoms, the most stable structure of Cu₈/TiC is produced. As shown in Figure 2i, there are still six Cu–C bonds at the interface with an average length of 1.941 Å. The E_b of the Cu₈ cluster is predicted to be 9.49 eV. It seemed that although the numbers of Cu–C bonds at the interface are the same for the depositions of Cu₆, Cu₇, and Cu₈ clusters, the binding strength of copper particles enhances gradually. The above results indicate that Cu_n clusters are anchored on the TiC(001) surface through the formation of Cu–C bonds. Meanwhile, because the distance between two adjacent C atoms on the TiC(001) surface is about 3.05 Å, which is obviously larger than the length of the Cu–Cu bond of the cluster in the gas phase, the elongation of the Cu–Cu bond can be expected after forming Cu–C bonds at the interface to match the relatively large C–C distance.

Figure 2.

Top and side views of the most stable configurations of Cu_n clusters supported on the TiC(001) surface. The blue, gray, and cyan balls represent Cu, C, and Ti atoms, respectively.

Table 1. Calculated Binding Energy (E_b) (in eV), Average Lengths of Cu–C and Cu–Cu Bonds (in Å), Height (h) (in Å) of Cu_n Particles, the Bader Charge of Cu_n (in e), and the Work Function Change (in eV) of Cu_n/TiC(001) Surfaces.

systems	Eb	dCu–Cu	dCu–C	h	Bader charge of the Cun clustera	work function changeb
Cu1 (Figure 2a)	2.57		1.839	1.84	+0.13	–0.25
Cu2 (Figure 2b)	3.42	2.383	1.899	1.86	+0.17	–0.13
Cu3 (Figure 2c)	5.60	2.476	1.987	1.83	+0.28	–0.13
Cu4 (Figure 2d)	6.42	2.420	1.921	1.85	+0.34	–0.24
Cu5(3D) (Figure 2e)	7.24	2.491	1.913	3.49	+0.33	–0.28
Cu5(2D) (Figure 2f)	7.24	2.434	1.950	1.84	+0.31	–0.16
Cu6 (Figure 2g)	8.04	2.460	1.955	1.85	+0.37	–0.21
Cu7 (Figure 2h)	8.62	2.488	1.940	3.44	+0.47	–0.33
Cu8 (Figure 2i)	9.49	2.517	1.941	3.35	+0.28	–0.13

^aThe positive value denotes the loss of electrons for the copper cluster after deposition.

^bThe negative value means the decrease of the surface work function with respect to the pristine TiC(001) surface.

Since the positive binding energies are obtained for all Cu_n clusters, the deposition of the copper cluster on the TiC(001) substrate is thermodynamically feasible. Due to the strong interactions with the support, the configurations of Cu_n clusters with n ≥ 4 are changed greatly. According to the optimized structure shown in Figure 2, we can see that the 2D configuration (i.e., one single copper layer) is the dominant appearance of those small copper particles with n ≤ 6, and the calculated height with respect to the TiC substrate is about 1.8 Å (Table 1). When the number of Cu exceeds six, the admetal tends to form a 3D particle with a double-layer arrangement, in which the additional copper atoms fill the 4-fold hollow site of the first layer and the height of the particle is about 3.4 Å. Actually, as displayed in Figure S1 in the Supporting Information (SI), even when the number of Cu atoms is increased to 13, such a double-layer structure still exists. Experimentally, STM images of the Cu/TiC surface with 0.1 ML of Cu reveal that a large fraction of Cu particles exhibits a height of about 0.2 nm and has diameters in the range of 0.3–0.5 nm.³¹ Therefore, in comparison with STM data, our results suggest that for such a Cu/TiC surface with high catalytic activity, it mainly contains small Cu_n clusters with n ≤ 6.

To further reveal the interactions between Cu_n particles and the support, the Bader charge of Cu_n and the variation of the charge density introduced by the deposition of the Cu_n particle are determined. As shown in Table 1, due to the formation of the Cu–C adsorption bond at the interface, the electrons are transferred from the cluster to the TiC(001) support, and consequentially, Cu_n clusters carry positive charges. Among them, the most obvious charge transfer (0.47e) is found for the deposition of the Cu₇ cluster. Similar charge transfers are observed when the copper cluster is loaded on the other surfaces including Al₂O₃.³⁸ The above electron transfers can be further verified by the difference electron-density maps displayed in Figure 3. It is clear that the electron density at the interface increases obviously, while the electron density around the copper atom toward the vacuum side is reduced. Furthermore, the loss of electrons of the copper cluster will result in a decrease in the surface work function of TiC(001). The calculated work function change with respect to the pristine TiC(001) surface is listed in Table 1. The decrease in work function shows an order of Cu₇ > Cu₅(3D) > Cu₁ ∼ Cu₄ > Cu₆ > Cu₅(2D) > Cu₂ ∼ Cu₃ ∼ Cu₈. In other words, except for the extreme case of a single Cu atom, the deposition of the Cu₇ cluster causes the most obvious reduction (−0.33 eV) of the surface work function, followed by the cases of Cu₅(3D) (−0.28 eV) and Cu₄ (−0.24 eV) clusters. This also means that the Cu₇/TiC, Cu₅(3D)/TiC, and Cu₄/TiC surfaces have a relatively strong ability to donate electrons, which may be favorable for the activation of the CO₂ molecule.

Figure 3.

Difference electron-density maps (isovalue = 0.002 e/ų) of Cu_n clusters supported on the TiC(001) surface. The yellow and blue colors denote the gain and loss of electrons, respectively.

2.3. Adsorption of CO₂ Molecules on the Cu_n/TiC(001) Surface

It is well known that CO₂ activation usually leads to the generation of CO₂^δ− species with a bent configuration, which is a key step to realize CO₂ conversion. For the pristine TiC(001) surface, the most stable structure for CO₂ adsorption is shown in Figure 4a, and some structural parameters including the lengths of two C–O bonds, the O–C–O bond angle, and the lengths of different adsorption bonds, as well as the adsorption energy (E_ads) are given in Table 2. In this structure, the carbon atom of the CO₂ molecule is bonded to the surface C atom, and meanwhile, two oxygen atoms form adsorption bonds with surface Ti atoms, indicating that both C and Ti atoms on the surface are active sites for CO₂ chemisorption. The optimized lengths of C–C and Ti–O adsorptions bonds are about 1.50 and 2.11 Å, respectively. After adsorption, the O–C–O bond angle (123.3°) is close to the value of the CO₂^– anion (120°), and the elongations of two C–O bonds (0.12 Å) are observed. Additionally, a positive value of E_ads = 0.76 eV is predicted. The above results indicate that the pristine TiC(001) surface has a certain ability to activate the CO₂ molecule. Moreover, our previous studies showed that the formation of the CO₂^δ− anion is energetically unfavorable on low-index copper surfaces.³⁹ For example, the most stable chemisorption structure on the Cu(001) surface is that CO₂ is adsorbed above the 4-fold hollow site by forming two Cu–C and two Cu–O bonds. However, due to the obvious antibonding feature of the Cu–O adsorption bond, a negative E_ads = −0.51 eV is predicted, indicating that CO₂ could not be stabilized on the Cu(001) surface from a thermodynamic point of view.

Figure 4.

Top views of the optimized structures of the adsorption of CO₂ on the Cu_n/TiC(001) surface. The blue, cyan, gray, green, and red balls represent Cu, Ti, C of TiC support, C of CO₂, and O atoms, respectively. For clarity, the symbols O_a and O_b are used to distinguish two oxygen atoms of the CO₂ moiety.

Table 2. Calculated Adsorption Energy (E_ads) (in eV), Lengths of Two C–O Bonds (in Å), the ∠O–C–O Bond Angle (in Degree), and the Lengths of Different Adsorption Bonds for the Adsorption of CO₂ on the Cu_n/TiC(001) Surface^a.

systems	Eads	C–Oa	C–Ob	∠O–C–O	Cu–Oa	Cu–Ob	C–C	Ti–Oa	Ti–Ob	Cu–C
pristine surface	0.76	1.292	1.292	123.3			1.500	2.111	2.107
Cu1-M1	1.47	1.285	1.312	119.0		1.894	1.504	2.099
Cu1-M2	1.23	1.315	1.320	117.1	2.313	2.258	1.461	2.150	2.152
Cu2-M1	1.20	1.285	1.313	118.7		1.910	1.501	2.109
Cu2-M2	1.13	1.315	1.314	117.1	2.342	2.347	1.466	2.130	2.131
Cu3-M1	1.14	1.285	1.321	119.7		2.028	1.488	2.112	2.328
Cu3-M2	1.03	1.311	1.314	122.9	2.113	2.111	1.452	2.389	2.353
Cu3-M3	1.41	1.297	1.297	121.3	1.937	1.937				1.936
Cu4-M1	1.24	1.287	1.315	119.6		1.989	1.491	2.121	2.524
Cu4-M2	1.11	1.318	1.306	118.3	2.373	2.461	1.468	2.151	2.142
Cu4-M3	1.25 (1.11)b	1.287 (1.29)	1.357 (1.36)	116.8	1.957	2.018 (×2)				1.943 (×2)
Cu5(3D)-M1	1.23	1.287	1.310	120.1		1.992	1.494	2.140
Cu5(3D)-M2	1.19	1.319	1.309	118.0	2.385	2.461	1.467	2.164	2.135
Cu5(2D)-M1	1.14	1.269	1.370	118.9		2.085 (×2)	1.467	2.227
Cu5(2D)-M2	1.06	1.316	1.301	118.4	2.351	2.642	1.477	2.132	2.122
Cu5(2D)-M3	1.17	1.278	1.300	123.0	1.990	1.930				1.966
Cu6-M1	1.13	1.288	1.313	119.9		2.058	1.494	2.117	2.367
Cu6-M2	1.10	1.324	1.290	119.6	2.181	2.927	1.477	2.160	2.133
Cu6-M3	1.00	1.274	1.358	117.9	2.031	2.048 (×2)				1.936 (×2)
Cu7-M1	1.25	1.288	1.313	119.4		2.018	1.493	2.121	2.494
Cu7-M2	1.16	1.292	1.325	119.3	2.844	2.193	1.475	2.143	2.171
Cu7-M3	1.30	1.306	1.336	116.0	2.096 (×2)	1.963 (×2)				1.941 (×2)
Cu8-M1	1.02	1.285	1.318	120.9		2.140	1.489	2.132	2.298
Cu8-M2	1.03	1.310	1.300	119.8	2.573	2.700	1.482	2.142	2.136

^aFor the free CO₂ molecule, the optimized length of the C–O bond is 1.176 Å.

^bResults obtained by Rodriguez et al.¹⁶

For TiC(001)-surface-supported copper particles, many possible configurations for the chemisorption of CO₂ have been explored. Our results reveal that for copper clusters with a small size (n ≤ 2) and a large size (n ≥ 8), the interface is the preferred adsorption site, in which both Cu_n clusters and the TiC support directly participate in the binding and stabilizing of CO₂. However, for the remaining Cu_n clusters with a medium size (n = 3–7), the noninterfacial structures that CO₂ adsorb solely on the copper particles are also observed. More specifically, there are two typical configurations for the interfacial adsorption, which correspond to one or two O atoms of CO₂ to form the Cu–O bond with the copper cluster, respectively. For clarity, in the following sections, the suffixes M1 and M2 are used to distinguish these two structures, and the configuration where only the Cu_n cluster is involved in CO₂ adsorption is denoted by the suffix M3. Furthermore, the symbols O_a and O_b are used to distinguish two oxygen atoms of the CO₂ moiety. Here, we mainly focus on the stable configurations of the above three typical adsorption models, as presented in Figure 4, and the corresponding structural parameters are listed in Table 2.

As shown in Figure 4b,c, two typical structures are obtained for the TiC(001) surface with the predeposition of a single Cu atom, in which CO₂ is chemisorbed at the interface. In the M1 structure, the CO₂ molecule is located at the interface through the formation of Cu–O_b, Ti–O_a, and C–C bonds with lengths of 1.894, 2.099, and 1.504 Å, respectively. After adsorption, the configuration of the CO₂ moiety exhibits a bent structure with an O–C–O bond angle of 119°, and C–O_a and C–O_b bonds are stretched to 1.285 and 1.312 Å, respectively. When both oxygen atoms of CO₂ are bonded to the Cu atom, the M2 structure (Figure 4c) is produced, in which five adsorption bonds, including two Cu–O, two Ti–O, and one C–C bond, are formed at the interface. Compared with the M1 model (Table 2), the longer C–O bonds (1.315 and 1.320 Å) and the smaller O–C–O bond angle (117.1°) are predicted for the M2 structure, suggesting a more obvious activation of CO₂. However, the values of E_ads (1.47 vs 1.23 eV) indicate that M1 is more stable than the M2 model.

The configurations of CO₂ adsorption on the Cu₂/TiC(001) surface are presented in Figure 4d,e. Since it is still only one Cu atom involved in the interaction with CO₂, the adsorption structures of Cu₂/TiC are similar to those of Cu₁/TiC. For example, the structural parameters of M1 models of two surfaces are nearly identical (Table 2). According to the values of E_ads (1.20 vs 1.13 eV), the M1 model is still the most stable configuration for CO₂ adsorption. However, the energy difference between M1 and M2 becomes small (<0.1 eV). Moreover, due to the decrease in E_ads, the relatively weak binding strength of CO₂ on Cu₂/TiC can be expected with respect to Cu₁/TiC.

For CO₂ adsorption on the Cu₃/TiC surface, the optimized structures of two interfacial models are given in Figure 4f,g, and the M1 model is still determined to be the most stable structure with E_ads = 1.14 eV. However, since the O_b atom of CO₂ is close to the surface Ti, one additional Ti–O_b bond with a length of 2.328 Å is found (Figure 4f). Thus, CO₂ is anchored at the interface of Cu₃/TiC through forming four adsorption bonds in the M1 model. Accompanied by the formation of the Ti–O_b bond, the elongation of the Cu–O_b bond (about 0.1 Å) is observed. Based on the results shown in Table 2, the configurations of the CO₂ moiety of three M1 models containing single Cu atom, Cu₂, and Cu₃ clusters are similar. Besides M1 and M2 structures, a new typical configuration, namely, the M3 model (Figure 4h), where the CO₂ molecule is exclusively adsorbed on the Cu₃ cluster, is found. In this structure, CO₂ occupies the 3-fold hollow site of the Cu₃ cluster, and three adsorption bonds including one Cu–C and two Cu–O bonds are formed. The large value of E_ads = 1.41 eV indicates that the M3 model is more stable than interfacial structures on the Cu₃/TiC surface.

In the case of the Cu₄/TiC surface, three models of CO₂ adsorption are obtained (Figure 4i–k). Unlike Cu₃/TiC, a different configuration of the M3 model is observed. Now, CO₂ favors the 4-fold hollow site built by four Cu atoms, and five adsorption bonds including two Cu–C, one Cu–O_a, and two Cu–O_b bonds are formed (Figure 4k). The average lengths of Cu–C and Cu–O bonds are 1.943 and 1.998 Å, respectively. The values of E_ads for M1–M3 models are calculated to be 1.24, 1.11, and 1.25 eV, respectively. The small difference in E_ads between M1 and M3 models indicates that these two structures can coexist from a thermodynamic point of view; however, the configurations of the CO₂ moiety in M1 and M3 models are different. Since the O_b atom is adsorbed at the bridge site between two Cu atoms in the M3 structure (Figure 4k), the weakening of the C–O_b bond is more pronounced than in the case of the M1 model. Consequently, the C–O_b bond in M3 (1.357 Å) is somewhat longer than in M1 (1.315 Å). In addition, a relatively small O–C–O bond angle (116.8°) is predicted for the M3 model. The above results suggest that the activation of CO₂ is more obvious for the M3 model on the Cu₄/TiC surface. Furthermore, the configuration of the CO₂ moiety obtained here for the M3 model is nearly identical to the result reported in the previous theoretical work (Table 2),¹⁶ in which only Cu₄/TiC(001) is selected as a prototype for studying CO₂ activation and hydrogenation.

As mentioned above, there are two different arrangements of the Cu₅ cluster on the TiC(001) surface, namely, Cu₅(3D) and Cu₅(2D), as shown in Figure 2e,f, respectively. For the surface-supported three-dimensional Cu₅(3D) cluster, the chemisorption of the CO₂ molecule on the copper cluster is difficult, and only interfacial structures are obtained (Figure 4l,m). Similar to the previous cases, the M1 structure (E_ads = 1.23 eV) is still more stable than M2 (E_ads = 1.19 eV). By contrast, when Cu₅ adopts a dispersed 2D arrangement, in addition to two interfacial configurations (Figure 4n,o), CO₂ molecules can be adsorbed only forming bonds with the copper particle. In the M3 model (Figure 4p), CO₂ shows a configuration similar to the Cu₃/TiC system, in which the 3-fold hollow site is preferred. This structure is more stable than the case where CO₂ occupies the 4-fold hollow site above four Cu atoms. The E_ads values of M1–M3 models are 1.14, 1.06, and 1.17 eV, respectively, namely, a slight advantage in energy for the M3 structure. The above differences indicate that the adsorption behavior of CO₂ is also sensitive to the morphology of the Cu_n cluster supported on the TiC(001) surface.

For Cu₆/TiC, as presented in Figure 4q–s, all three typical structures of CO₂ adsorption are achieved. It is noted that for the interfacial structures, i.e., M1 and M2 models, the low-coordinated Cu atoms at four corners are the prior sites to form the Cu–O adsorption bond. Since the configuration of Cu₆ on the TiC(001) surface can be regarded as composed of two edge-shared Cu₄ clusters (see Figure 2g), the configurations of CO₂ in the M3 structure of both Cu₆ and Cu₄ clusters are quite similar, and CO₂ tends to locate at the 4-fold hollow site above four Cu atoms through forming five bonds (Figure 4k,s). Based on the values of E_ads, the M3 model (E_ads = 1.00 eV) is thermodynamically less stable than M1 (E_ads = 1.13 eV) and M2 (E_ads = 1.10 eV). Therefore, unlike on the Cu₄/TiC surface where M1 and M3 structures are very close in energy, now the CO₂ molecule has a tendency to be adsorbed at the interface.

The most stable configuration of Cu₇ deposited on the TiC(001) surface (Figure 2h) can be constructed by adding one copper atom at a 4-fold hollow site of the Cu₆ cluster, and two kinds of Cu atoms at the corners of the particle are found. Correspondingly, there are more possible sites for the interfacial adsorptions of CO₂. Our calculated results show that for the M1 model the Cu atom at the corner of the pyramid bottom is the preferred adsorption site (Figure 4t), while for the M2 model CO₂ is favorable for the Cu atom at the corner of the square part (Figure 4u). Similar to the above Cu/TiC interfaces, the M1 structure (E_ads = 1.25 eV) is still more stable than the M2 model (E_ads = 1.16 eV). For the M3 model, CO₂ adopts an adsorption configuration observed in Cu₄/TiC and Cu₆/TiC systems, and it occupies the 4-fold hollow site of the Cu₇ cluster (Figure 4v). However, because the Cu atom at the vertex of the tetragonal pyramid is near the hollow site, an additional Cu–O_a bond is formed, resulting in a total of six adsorption bonds between CO₂ and the Cu₇ cluster. Consequently, among all M3 structures studied here, the largest E_ads with a value of 1.30 eV is predicted for the Cu₇/TiC(001) surface. In the M3 model, both C–O bonds of CO₂ are stretched to a length larger than 1.30 Å, and the smallest O–C–O bond angle (116°) is obtained (Table 2). These changes in the configuration mean a significant activation of the CO₂ molecule. Moreover, the M3 model is the most stable structure for CO₂ adsorption on the Cu₇/TiC surface.

As presented in Figure 2i, the Cu₈ cluster supported on the TiC(001) surface has a 3D structure, and the close stacking of eight Cu atoms cannot provide a suitable hollow site for the chemisorption of CO₂. Consequently, only the interfacial structures (Figure 4w,x) are obtained for Cu₈/TiC. Similar to the above cases, CO₂ still favors the Cu atom at the corner of the cluster, and the small difference in E_ads (1.02 vs 1.03 eV) shows that the stability of M1 and M2 structures is similar. Furthermore, using Cu₁₃/TiC(001) as an example for the deposition of copper particles with a large size, the optimized structures of three adsorption models are given in Figure S2 in the SI. It is interesting that although CO₂ can be chemisorbed on the top of the Cu₁₃ cluster, the small E_ads with a value of 0.21 eV indicates that the binding strength between CO₂ and the metal particle is weak. For comparison, the interfacial configuration still has a good stability for CO₂ adsorption (E_ads = 0.94 eV). Thus, for the TiC(001)-surface-supported copper particle with a large size, the interface site is preferred for the adsorption of the CO₂ molecule.

Figure 5a presents a comparison of E_ads of three typical configurations for the adsorption of CO₂ on the TiC(001)-surface-supported copper cluster with different sizes. It is clear that E_ads values of all Cu_n/TiC(001) systems are larger than that of the pristine TiC surface (0.76 eV), revealing the obvious promotion for CO₂ chemisorption on the Cu_n/TiC surface due to the synergistic effect between the metal particle and the support. For two interfacial adsorption configurations, they can be observed on all Cu_n/TiC surfaces, and the M1 model is more stable than M2 in energy. However, with the increase in size of the metal particle, the energy difference between M1 and M2 structures decreases gradually. Therefore, for the deposition of copper clusters with a large size, various interfacial adsorption configurations may coexist. If different kinds of Cu atoms appear at the interface, the copper atom at the cluster corner is the preferential position for interfacial adsorption. It is noted that for the exclusive chemisorption of CO₂ on the copper cluster (i.e., M3 model), it only occurs on TiC(001) surfaces containing Cu_n particles with a moderate size (i.e., n = 3–7). From Figure 5, in most cases of these Cu/TiC interfaces (except for Cu₆-M3), the E_ads of the M3 structure is larger than that of M1 and M2 models, so now M3 corresponds to the most stable pattern for CO₂ adsorption. As for the configurations of the CO₂ moiety in three models, although the elongations of two C–O bonds are observed for all systems with respect to the free CO₂ molecule (see Table 2), the specific changes are dependent on the adsorption structure. For two interfacial adsorption models, if the oxygen atom of CO₂ forms a bond with the Cu atom, the associated C–O bond can be stretched to more than 1.30 Å. Additionally, the O–C–O bond angle of the M2 model is smaller than that of the M1 structure. For the M3 structure, CO₂ can be adsorbed at a 3-fold (such as Cu₃-M3 and Cu₅-M3) or 4-fold (such as the surfaces containing Cu₄ and Cu₇ clusters) hollow site. As a result of more adsorption bonds being formed in the latter case, two C–O bonds of CO₂ are elongated more obviously for the adsorption at the 4-fold hollow site, and at the same time, the decrease in the O–C–O bond angle is more pronounced. For example, in the Cu₇-M3 surface, the lengths of two C–O are larger than 1.30 Å, and the O–C–O bond angle is reduced to 116.0°.

Figure 5.

(a) Variations of adsorption energy (E_ads) of three typical configurations of CO₂ adsorption on different Cu_n/TiC(001) surfaces. (b) Deviation of the average length of two C–O bonds and (c) the O–C–O angle from free CO₂ as a function of the total Bader charge of the CO₂ moiety.

It is widely acknowledged that the degree of CO₂ activation is closely related to the magnitude of charge transfer caused by the chemisorption, and the more electrons CO₂ obtains, the more significantly it is activated. Therefore, the charge transfer between CO₂ and the Cu_n/TiC(001) surface is further analyzed, and the calculated Bader chargers of Cu_n clusters, CO₂ moiety, and C and O atoms are listed in Table 3. First, after CO₂ chemisorption, the copper cluster carries a more positive charge due to the charge transfer. By comparing the charges of the same Cu_n cluster in different adsorption structures, we can see that in two interfacial adsorption models the charges of the copper cluster are similar, but more obvious charge transfer can be observed in the M3 structure (if it exists). Taking Cu₄/TiC as an example, the charges of Cu₄ in M1 (+0.56e) and M2 (+0.65e) are similar; however, they are remarkably smaller than the value of the M3 model (+1.36e). This also means that CO₂ receives more electrons from the surface when it is exclusively adsorbed on the copper cluster. It is worth noting that in M1 and M2 models, the number of electrons obtained by the CO₂ molecule is obviously larger than that lost from the copper cluster, implying that both the Cu_n particle and the TiC support are involved in the charge transfer. This is consistent with the fact that both Cu–O and Ti–O adsorption bonds are formed for the interfacial adsorption, so the ligand effect is responsible for CO₂ activation in the interfacial configurations. By contrast, in the M3 structure, most of the accumulated electrons on CO₂ originate from the Cu_n cluster. Therefore, the synergistic effect between clusters and the support shows a dependence on the adsorption site of the CO₂ molecule. From Table 3, we can see that the CO₂ moiety in Cu₇-M3 carries the most negative charge (−1.13e), followed by Cu₄-M3 (−1.08e) and Cu₆-M3 (−1.06e), indicating that CO₂ adsorbed on these surfaces with the M3 structure can be activated to CO₂^δ− species more easily. This result agrees with the lower surface work function after loading of Cu₄, Cu₆, and Cu₇ clusters (see Table 1). As for C and O atoms belonging to CO₂, the adsorption has a small effect on the charges of two O atoms, while a significant reduction of the positive charge of the C atom is found, suggesting that the carbon atom of CO₂ receives most of the electrons from the Cu/TiC surface. Considering that the number of electrons obtained by the C atom is an important indicator to measure the degree of CO₂ activation, the most obvious activation of CO₂ can be expected in Cu₄-M3, Cu₆-M3, and Cu₇-M3 by comparing the charges of the C atom for all systems, in which the carbon atom carries the smallest positive charge of about +1.1e.

Table 3. Bader Charges (in e) of the Cu_n cluster, the CO₂ Moiety, and C and O Atoms for the Adsorption of CO₂ on Different Cu_n/TiC(001) Surfaces^a.

systems	Cun cluster	CO2 moiety	C of CO2	Oa/Ob
free CO2		0	+2.04	–1.02/–1.02
Cu1-M1	+0.44	–0.78	+1.36	–1.07/–1.07
Cu1-M2	+0.44	–0.97	+1.15	–1.07/–1.06
Cu2-M1	+0.54	–0.78	+1.37	–1.08/–1.07
Cu2-M2	+0.47	–0.96	+1.22	–1.09/–1.09
Cu3-M1	+0.49	–0.83	+1.32	–1.06/–1.09
Cu3-M2	+0.66	–1.07	+1.15	–1.11/–1.12
Cu3-M3	+1.08	–0.90	+1.19	–1.04/–1.05
Cu4-M1	+0.56	–0.84	+1.29	–1.05/–1.08
Cu4-M2	+0.65	–0.95	+1.24	–1.09/–1.10
Cu4-M3	+1.36	–1.08	+1.06	–1.08/–1.06
Cu5(3D)-M1	+0.67	–0.85	+1.32	–1.07/–1.09
Cu5(3D)-M2	+0.71	–0.99	+1.19	–1.10/–1.08
Cu5(2D)-M1	+0.92	–0.92	+1.24	–1.09/–1.07
Cu5(2D)-M2	+0.61	–0.93	+1.21	–1.09/–1.05
Cu5(2D)-M3	+1.12	–0.91	+1.23	–1.08/1.06
Cu6-M1	+0.65	–0.87	+1.32	–1.07/–1.11
Cu6-M2	+0.64	–0.92	+1.22	–1.06/–1.07
Cu6-M3	+1.36	–1.06	+1.11	–1.10/–1.07
Cu7-M1	+0.77	–0.86	+1.31	–1.07/–1.10
Cu7-M2	+0.76	–0.92	+1.27	–1.09/–1.10
Cu7-M3	+1.62	–1.13	+1.11	–1.10/–1.13
Cu8-M1	+0.58	–0.84	+1.35	–1.08/–1.10
Cu8-M2	+0.60	–0.90	+1.25	–1.09/–1.06

^aThe positive and negative values denote the loss and obtaining of electrons, respectively.

To see how the size of the copper cluster affects the activation of the CO₂ molecule on the TiC(001) surface, Figure 5b,c displays the deviations of the average length of two C–O bonds and the O–C–O bond angle from the free CO₂ as a function of charge of the CO₂ moiety for the most stable adsorption structure of all Cu_n/TiC surfaces. It is clear that the charge transfer from the Cu_n/TiC surface to CO₂ approximately follows a linear relationship with the deviations of the C–O bond length and the O–C–O bond angle. Therefore, the more electrons CO₂ receives, the more the corresponding configuration is changed. The most obvious deformation of the CO₂ moiety is observed for Cu₇-M3 and Cu₄-M3 structures, indicating that the TiC(001) surfaces with the depositions of Cu₄ and Cu₇ clusters have the best ability for the activation of CO₂.

2.4. Dissociation of CO₂ on Cu₄/TiC(001) and Cu₇/TiC(001) Surfaces

The reduction of CO₂ to CO is an important process since CO can be used as feedstock in the Fischer-Tropsch synthesis of fuels or as the starting point for the production of hydrocarbon compounds used in the industry. Based on the above results, the TiC(001)-surface-supported Cu₄ and Cu₇ particles are adopted as prototypes to investigate the CO₂ conversion to CO. Using the climbing image-nudged elastic-band (CI-NEB) method,⁴⁰ the minimum energy paths for the cleavage of one C–O bond of CO₂* to generate CO* on the above surfaces are determined, and two typical CO₂ adsorption configurations, M1 and M3 models, are taken into account due to the small energy difference between them. Figure 6 displays the predicted energy profiles and the optimized configurations of the transition state (TS) and the final state contained CO* and O* species.

Figure 6.

Energy profiles for CO₂ dissociation on (a) Cu₄/TiC(001) and (b) Cu₇/TiC(001) surfaces. The zero of energy is set to the total energy of the Cu_n/TiC(001) surface and the CO₂ molecule in the gas phase. In the figures, top views of the optimized structures of the transition state (TS) and the final state are also presented.

For the Cu₄/TiC(001) surface, the calculated activation barriers (E_a) of CO₂ dissociation are 1.09 and 0.70 eV for M1 and M3 models (Figure 6a), respectively. It is noted that E_a of M3 is obviously smaller than that of the M1 structure. When the Cu₇ cluster is deposited on the TiC(001) surface (Figure 6b), although the difference in E_a values of M1 and M3 adsorption structures (0.91 vs 0.98 eV) is small, it is still relatively easier to decompose CO₂ on the copper cluster than at the interface. Therefore, the kinetic studies of CO₂ dissociation also indicate that the CO₂ adsorbed on the copper cluster is more activated compared to the case of interfacial adsorption.

3. Conclusions

In this work, extensive DFT calculations have been performed to study the size effects of subnanometer copper clusters supported on the TiC(001) surface for the adsorption and activation of CO₂. By comparing the adsorption energies of the pristine and Cu_n-supported TiC(001) surfaces, it is clear that no matter whether CO₂ molecules are adsorbed at the Cu/TiC interface or on the copper cluster, the larger values of E_ads are predicted. Thus, introducing Cu_n clusters on the TiC surface can significantly improve the thermodynamic stability of CO₂ chemisorption due to the synergistic effect between the metal particles and the carbide support. Our results show that the specific adsorption site is sensitive to the size and the morphology of the Cu_n cluster deposited on the surface. First of all, the interfacial M1 structure is observed for all Cu_n/TiC(001) surfaces, in which both Cu particles and TiC support directly participate in the binding and activation of CO₂. In this case, the synergistic behavior of two constituents of the system is associated with the ligand effect that is introduced by forming adsorption bonds with CO₂. Second, for those Cu_n clusters with a medium size (n = 3–7) where there is an exposed hollow site built by three or four Cu atoms at the interface, CO₂ can be exclusively adsorbed on the copper particles. In most cases, this structure (i.e., M3 model) exhibits the best stability compared to the interface adsorption configuration. Although similar hollow sites also exist on the bare copper surfaces, CO₂ cannot be stabilized on such surfaces. This is due to the obvious antibonding characteristic of the Cu–O adsorption bond,³⁹ which implies that removing some 3d electrons from the copper atom will enhance the strength of the Cu–O adsorption bond. It is worth noting that for the Cu_n cluster supported on the TiC(001) surface, those Cu atoms at the interface lose electrons due to the formation of Cu–C bonds, and this is beneficial for strengthening the Cu–O adsorption bonds. Consequently, CO₂ molecules have good stability on the copper clusters supported on the TiC(001) surface by adopting an M3 structure. In other words, although the TiC support in the M3 model does not directly participate in the interactions with CO₂, the charge transfer between TiC and Cu_n clusters activates those copper atoms at the interface. Therefore, in the M3 structure, the synergistic effect can be mainly attributed to the charge transfer between the different constituents of the system. In addition, considering that the exposed hollow site is difficult to appear at the Cu/TiC interface for the Cu particle with a large size (e.g., Cu₈), the interfacial configuration is still the main pattern for the chemisorption of CO₂. The Bader charge analyses show that the activation mechanism of CO₂ in M1 and M3 configurations is different, and the CO₂ moiety in the latter case has the most negative charge, indicating that the chemisorption of CO₂ with the M3 structure can be more easily activated to CO₂^δ− species. Among the Cu_n/TiC studied here, the TiC(001)-surface-supported Cu₄ and Cu₇ clusters have a superior ability in CO₂ activation. Thus, controlling the size and the dispersion of the metal cluster with an optimized metal/TMC interface may help in designing excellent catalysts for CO₂ activation and conversion, and it will be interesting to study the mechanisms of subsequent CO₂ hydrogenation on Cu₄- and Cu₇-supported TiC(001) surfaces in a future work.

4. Computational Details

All spin-polarized density functional theory (DFT) calculations were performed using the Vienna ab initio simulation package (VASP).^41,42 The generalized gradient approximation (GGA) of the Perdew–Burke–Ernzerhof (PBE)⁴³ exchange–correlation functional was employed, and the interactions between the electron and ion were described by the projector augmented wave (PAW)⁴⁴ pseudopotentials. The effects of van der Waals interactions were considered using the vdW-DF3 functional,⁴⁵ and the dipole correction in the surface normal direction was applied. The kinetic cutoff energy of the plane-wave expansion was set to 400 eV, and the Brillion zone integration was sampled using a (4 × 4 × 1) Monkhorst–Pack grid. The convergence thresholds of the energy change and the maximum force were set to 10^–5 eV and 0.03 eV/Å, respectively.

A four-layer slab with a (4 × 4) supercell was adopted to simulate the TiC(001) surface, which contains 64 Ti and 64 C atoms. During structural optimizations, the top two layers were fully relaxed while the atoms at the bottom two layers were fixed to their bulk positions, and the vacuum between adjacent slabs was set to 15 Å to avoid interactions between neighboring slabs. Since there are many possible arrangements of copper clusters on the TiC(001) surface, an ab initio molecular dynamics (AIMD) simulation using the Nosé algorithm⁴⁶ was carried out to explore possible configurations for the deposition of Cu_n on the TiC(001) surface by a low cutoff energy of 250 eV with one K point. The simulation length was 10 ps with a time step of 1 fs at a temperature of 500 K. Then, typical configurations were sampled from the results of AIMD simulations, and finally further structural optimizations using the above accurate settings were performed to determine the most stable configuration of Cu_n/TiC(001). Our previous works on W₃O₉/TiO₂(110),⁴⁷ TiO_x/Mo(112),⁴⁸ and Au/TiO_x/Mo(112) surfaces⁴⁹ showed that the above approach could reasonably predict the stable structure of the complicated systems.

The binding energy (E_b) for the copper cluster on the TiC(001) surface was calculated by the following equation

where E(Cu_n/TiC), E(Cu_n), and E(TiC) represent the total energies of the Cu_n/TiC(001) system, the Cu_n cluster in the gas phase, and the pristine TiC(001) surface, respectively. For the chemisorption of the CO₂ molecule on the Cu_n/TiC(001) surface, the adsorption energy (E_ads) was defined as

where E(CO₂/Cu_n/TiC), E(Cu_n/TiC), and E(CO₂) denote the total energies of the adsorbed system, the clean Cu_n/TiC(001) surface, and the CO₂ molecule, respectively.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.1c04040.

• Optimized configurations of the Cu₁₃ cluster supported on the TiC(001) surface; and top views of the structures for the adsorption of CO₂ on the Cu₁₃/TiC(001) surface (PDF)

The authors declare no competing financial interest.