Density functional study of CO chemisorption on Ni(111) surface

Catalytic reactions are ubiquitous in chemical industry and play most often a crucial role in the production of numerous chemicals. Thus, improving the activity and selectivity of a catalyst is an important task for which a detailed understanding of the catalytic mechanisms at the atomistic level is essential. To gain deeper insights into these mechanisms, atomistic simulations are a highly valuable tool which can help guiding synthesis.
 
In many industrial relevant processes, transition metals like Nickel (Ni) are involved as catalysts. In particular for hydrogenation reactions, which are often found in petrochemical, pharmaceutical, and food industries, Ni has attained special importance because of its ability to adsorb hydrogen [1,2]. Likewise, the adsorption of carbon monoxide (CO) on Ni surfaces is another well-known and industrial relevant example. This reaction plays, for instance, a key role in the methanation reaction, where CO and hydrogen are converted into methane and steam and the Fischer-Tropsch synthesis for producing synthetic oil and fuel from a mixture of CO and hydrogen [3]. Experimental and theoretical studies indicate that the dissociation of adsorbed CO on the Ni surface is the rate-limiting step under realistic conditions used in production [4,5]. Under certain reaction conditions, the CO dissociation is, however, also believed to trigger the deposition of carbon on the Ni surface which entails the deactivation of the catalyst [6].
 
The interest in studying the CO chemisorption on Ni surfaces is not only motivated by its specific relevance in the above mentioned processes, but also by the fact that it can be regarded as a prototype reaction of heterogeneous catalysis in general. In this application note, we present a computational study of CO adsorption and dissociation on the Ni(111) surface.
 
Chemical reactions usually involve the breaking and formation of bonds, which is best described by ab initio methods. In particular for investigating solid state reactions, density functional theory has been proven to be a versatile approach. Plane wave-based DFT methods are a very commonly used approach for studying periodic systems like surfaces. It has been shown that these methods allow for quantitative predictions when compared to experiment [5,7,8]. The calculations which are presented in the following have been performed using the ABINIT module [9] within Scienomics MAPS platform [10].
 
First, the adsorption process is considered. On the Ni(111) surface, there exist four possible high symmetry adsorption sites which are shown in Figure 1.
 

c7-f1-ni111-surface
Top view on a Ni(111) surface. The surface unit cell is indicated with dashed lines. T denotes the site for adsorption on top of a surface atom, B is the bridge site, F and H indicate the fcc (ABCABC...) and hcp (ABABAB...) three-fold hollow site, respectively.

For calculating the adsorption energy, slab models for each adsorption site have been generated based on a 2x2 unit cell. The slab models consist of 3 Ni bulk layers and 5 layers of vacuum. One CO molecule is placed in each surface cell with the carbon being oriented towards the surface. This corresponds to a coverage of 1/4 ML. The DFT generalized gradient approximation (GGA) together with the PBE functional and Fritz-Haber-Institute pseudopotentials have been used for the calculations. During the optimization, the bottom two layers were kept fixed and spin-polarization effects have been taken into account for Ni. The calculations were performed with an energy cutoff of 1000 eV and a 1x1x1 Monkhorst-Pack grid for the k-point generation.
 
The analysis of the absolute energies of the optimized slab models of the four adsorption sites reveals that the fcc and hcp three-fold hollow sites are lower in energy compared to the top and bridge adsorption site. The relative energy differences are listed in Table 1 (second column).
 

screen-shot-2016-12-03-at-17-02-44
Table 1: Energies of the four adsorptions sites given relative to the hcp three-fold hollow site energy (Ex – Ehcp, x = top, bridge, and fcc) and CO adsorption energies (Eads = ECO / Ni slab - ENi slab - ECO, see text for further details) in kJ/mol.

The optimized structure of the hfcc three-fold hollow site which is the energetically most favored adsorption site is illustrated in Figure 2(a). In Figure 2(b), the corresponding optimization history is depicted showing that convergence is smoothly reached after about 65 optimization steps.
 

c7-f2-fcc-optimization
Figure 2: (a) Optimized slab model of the fcc three-fold hollow site. (b) Optimization history.

The adsorption energy is given by Eads = ECO / Ni slab - ENi slab - ECO, where ECO / Ni slab is the energy of the slab model of the adsorbed CO on the Ni(111) surface, ENi slab the energy of the bare surface, and ECO the energy of the isolated CO molecule in the gas phase which was calculated based on a primitive 10x10x10 Å unit cell. The adsorption energies of the four sites are listed in Table 1 (third column). The largest adsorption energies are found for the hcp and fcc hollow sites, respectively, followed by the bridge and top adsorption site. Thus, the same trend is observed as found for the relative energies of the four adsorption sites. This is fully in line with experimental data [11]. The calculated adsorption energies are slightly higher than experimentally observed [12] which is consistent with previous DFT studies which found an overbinding of about 0.5 eV when using the PBE functional within the GGA scheme [13]. However, it could be shown in Ref. [7] that the general trend is correctly reproduced when comparing adsorption energies at different coverages. Since the fcc and hcp hollow sites are significantly lower in energy, the analysis of the co-adsorption of C and O is restricted to these two models in the following.
 
For the co-adsorption of the dissociated CO molecule, six configurations are possible, which are: C(hcp) – O(hcp), C(fcc) – O(fcc), C(hcp) – O(fcc-n), C(hcp) – O(fcc-o), C(fcc) – O(hcp-n), C(fcc) – O(hcp-o), where n indicates that the O atom is sitting in the next adjacent hollow, and o that the O atom is sitting in the hollow lying opposite to the C adsorption hollow. Since the co-adsorption on adjacent sites is most unlikely due to repulsive effects, only the four remaining geometries have been explored.

The relative energies of the four co-adsorption sites are listed in Table 2.
 

screen-shot-2016-12-03-at-17-13-35
Table 2: Energies of the four co-adsorptions sites given relative to the C(hcp)-O(fcc-o) site energy in kJ/mol.

The configurations in which C and O are located in opposite lying holes are energetically preferred over the configurations with C and O sitting both either in hcp or fcc hollows. The optimized structure of the energetically most favored configuration, i.e. C(hcp) – O(fcc-o), is shown in Figure 3 together with the optimization history.
 

c7-f3-co-dissociated-optimization
Figure 3: (a) Optimized slab model of the C(hcp) – O(fcc-o) co-adsorption site. (b) Corresponding optimization history.

Based on the optimized structures of the hcp three-fold hollow site and the C(hcp) – O(fcc-o) co-adsorption site, the reaction pathway of the CO dissociation on the Ni(111) has been evaluated. For this purpose, the nudge elastic band scheme has been used as implemented in the Transition State Locator module of Scienomics MAPS platform. The nudge elastic band scheme does not only localize the transition state (TS), but determines the full reaction path between the reactant (here: CO adsorbed at the hcp three-fold hollow site) and the product state (dissociated CO in C(hcp) – O(fcc-o) configuration). The result of the transition state search is illustrated in Figure 4 showing the energetical change during the reaction.
 

c7-f4-reaction-pathway
Figure 4: Reaction pathway calculated using the nudge elastic band methods.

The energy difference between the CO adsorbed Ni species and the transition state corresponds to the activation energy Ea, while the reaction energy DH is given by the energy difference between product and reactant state. In the present case, the activation energy is 375.7 kJ/mol and the reaction energy is 254.1 kJ/mol.
 

In summary, the study demonstrates that DFT can be used to predict the preferred adsorption site and energy of small molecules on metal surfaces in quantitative and qualitative agreement with experimental measurements. Moreover, it is possible to explore the reaction pathway and thus estimate the activation and reaction energy using first principles methods. Ab initio studies are thus a valuable tool to gain detailed insights into reaction mechanisms in order to guide and improve synthesis.

 

References:

  1. P. Fouilloux, Appl. Catal. 8 (1983) 1-42.
  2. M. Raney, (1927), US Patent 1628190, issued 10. Mai 1927.
  3. J.R.H. Rossin Catalysis: Volume 7: A Review of Chemical Literature, ed. G. C. Bond, Graham A. Webb and Royal Society of Chemistry (1985) 1-45.
  4. J. Sehested et al., J. Phys. Chem. B, 109, (2005) 2432-2438 .
  5. M.P. Andersson et al., Journal of Catalysis, 255 (2008) 6-19 .
  6. C.H. Bartholomew, Applied Catalysis A, 212 (2001) 17-60
  7. A. Eichler, Surface Science, 526 (2003) 332-340 .
  8. V. Shah et al., Surface Science, 537 (2003) 217-227 .
  9. X. Gonze et al., Computer Phys. Commun., 180 (2009) 2582-2615.
  10. MAPS, Version 3.3.2, Scienomics, Paris, France.
  11. L. Becker, Phys. Rev. B, 47 (1993) 9710-9714.
  12. J. Stuckless, J. Chem. Phys., 99 (1993) 2202-2212.
  13. B. Hammer, Phys. Rev. B, 59 (1999) 7413 -7421.
  14. Y. Morikawa et al., Surface Science, 386 (1997) 67-72 .