A density functional study of the adsorption of  CO on Fe(111) surface

Chen Yunhong¹, Yang Jun^1,2, Sun Yuchun¹, Chen Manying¹
(¹Department of Chemistry, Jinan University, Guangzhou 510632 China;  ²State Key Laboratory of Coal Conversion, Institute of Coal Chemistry, Chinese Academy of Sciences, Taiyuan, 030001)

Received on May 26, 2004; Support by the Committee of Science and Technology of China via 863 plan (No.2001AA523010)

Abstract The adsorption of CO on the Fe (111) surface has been investigated by ab initio density functional methods and the generalized gradient approximation (GGA). The surface is modeled using an infinite periodic slab. CO adsorption at five different surface sites have been studied and it shows that the energetically preferred site for CO adsorption corresponds to the shallow hollow(a₂) adsorption followed by bridge site(a₄), bridge-like site(a₅), deep hollow site(a₃), and on-top site(a₁). However, CO is mostly activated in the bridge-like site. The significantly increased theoretical CO bond length of 1.221Å on the bridge-like site (compared to 1.144Å in the gas phase) indicates the activation of CO bond and this site is expected to be the precursor for CO dissociation on the Fe (111) surface.
Keywords Density functional methods; Fe(111) surface; carbon monoxide; adsorption

1. INTRODUCTION
The adsorption of CO on transition metal low index surfaces has been the subject of numerous experimental and theoretical studies in recent years because the CO adsorption and dissociation play a crucial role in many catalytic processes, for instance, in car exhaust or in the Fischer-Tropsch synthesis (FTS). As the iron-CO interaction plays an important role in the FTS, the interaction of CO with clean iron surface has been investigated by various techniques. There are many experimental and theoretical studies on the Fe(001) and Fe(110) surfaces^[1,2]. Both of these two surfaces are close-packed surfaces, whereas, the Fe(111) surface is a very open surface. Difference in the surface structure will affect the CO adsorption on the surface. There are relative few studies on CO adsorption on the Fe(111) surface. Christmann and Seip ^[3] et al have investigated the interaction of CO with an Fe (111) surface with low energy electron diffraction (LEED). They found that the adsorption of CO on Fe (111) below 300 K present three different non-dissociated species as distinguished by their C-O stretch frequencies. According to this result, they thought that CO adsorbed on three different sites on the Fe(111) surface, namely as "on top" , "shallow hollow" and "deep hollow" sites. Whitemam et al ^[4] found that the CO adsorption site occupation on the Fe (111) depends upon coverage and temperature, and CO adsorbs on four sites on the Fe(111) surface. A new additional "bridge" site is found except the three sites mentioned above. Mehndru and Alfred et al ^[5] have used semi-empirical molecular orbital theory to study the binding and orientations of CO on Fe (100), Fe (110) and Fe(111) at low coverage. Their investigation shows that CO binds on the di-s bridging site in the lying-down orientation on Fe (111) which do not support the previous proposed "shallow-hollow" binding site for CO on Fe (111) surface^[3]. It is obviously that the level of the theory is insufficient and it can not be expected to give a quantitative description of the process. Despite considerable valuable structure information obtained from the experimental and theoretical studies described above, a detailed mechanism and quantitative estimation of the energies of various processes of CO interaction with the iron surface are still lacking.
    In this paper, we present a systematic study of CO adsorption on the Fe (111) surface at high coverage using ab initio density functional theory. The aims of the present paper is to build up a detailed picture of the geometry, energetic and electronic structure of the clean and adsorbed surface, and show some light in the CO adsorption mechanism. The magnetism of the system and small energy differences between various spin states make the study quite involved and challenging.

2 COMPUTATIONAL DETAILS
In the present study, all calculations described herein were performed within the framework of density functional theory (DFT) using a basis set consisting of plan waves, as implemented in the CASTEP package program ^[6] available in the Materials studio 2.2 version of the Accelrys Inc. The electron-ion interactions were described by ultrasoft pseudopotentials and electron exchange and correlation energies were calculated with the Perdew, Burke and Ernzerhof formulation (PBE)^[7] of the generalized gradient approximation (GGA). Spin-polarization was included in the calculations for systems with Fe to correctly account for its magnetic properties. Spin-polarization has been shown to have a major effect on the adsorption energies for magnetic systems and may alter the topology of the potential energy surfaces^[8].
    The surface of Fe (111) is modeled by a slab which contained seven layers. The three topmost layers are allowed to relax, and CO is adsorbed only on one side of the slab. The one-sided slab has been used intensively in the literature and has been proven to be accurate. The plane-wave energy cutoff was set to 340 eV throughout all calculations and the Brillouin zone of the surface unit cell was sampled with a 6× 6×1 Monkhorst-Pack mesh. The smearing width is 0.1eV. The coverage of the CO is defined as the following: the coverage=the number of the CO molecule/the number of the surface Fe atoms. CO can adsorb on the three topmost layers. So one CO adsorbed on the p(1×1) unit cell is to represent the coverage of 1/3 ML. The adsorption energy per CO molecule is defined as: E_ads=E_(slab+CO)-E_slab-E_CO. In order to distinguish the iron atoms in different layers, we labeled the iron atoms of the first layer, second layer and third layer as Fe1, Fe2,Fe3, respectively.

3.RESULTS AND DISCUSSIONS
To make sure that the GGA-PBE is suitable for the system of CO adsorption on the Fe(111) surface, we calculated the CO bone length and a -Fe lattice constant and magnetic. The results listed in Table 1 show in agreement with the experiments results.

Table 1: Comparison between PBE and experiments

    This method has also been used to reproduce the results of CO adsorption on the Fe(001)^[1] and Fe(110)^[2], and the results show in agreement.

3.1 Clean surface
On the Fe (111) surface, an iron atom of the first layer is surrounded by three shallow hollows and three deep hollows, shown in Fig. 1. A shallow hollow is formed by a Fe2 atom and three Fe1 atoms (which are the nearest to the Fe2 atom), and the Fe2 atom sits directly below the hollow. A deep hollow is formed by a Fe3 atom and three Fe1 atoms (which are the nearest to the Fe3 atom), and the Fe3 atom sits directly below the hollow. The very open structure of the clean Fe(111) surface is known to exhibit large multilayer relaxation. With d_ik being spacing between the ith and kth atomic layer, and d the bulk spacing. Sokolov et al ^[11] found the following relaxations: Dd₁₂/d=(-16.9± 3.0)%, Dd₂₃/d=(-9.8± 3.0)%, Dd₃₄/d=(4.2± 3.6)%,and Dd₄₅/d=(-2.2± 3.6)%. The relaxations of our seven layer slab are shown in Table 2. Here only the three topmost layers are relaxed and the relaxations are tabulated as relative changes in the nearest neighbor distances d_i(i+3) instead of d_i(i+1).The agreement is reasonable considering the uncertainties in the calculations due to the small number of layers in the slab^[12] and different k-points in the bulk and the Fe (111) surface. The result shows that the first and second layers shifts inward and the third layer shifts outward, and this agrees with the experiments qualitatively. The average magnetic moment in the surface layer is slightly enhanced (m_s=2.68u_B) compared to the bulk (m_s=2.2u_B), and this can be explained by the open structure of the Fe(111) surface and the reduced coordination of the surface atoms.

Fig.1 Top view and side view of Fe(111) surface

Table 2 Relaxation of the clean Fe(111) surface. The experimental data is from Ref [11]

3.2 CO adsorption
As mentioned above, four possible adsorption sites on the Fe (111) surface at this coverage have been experimentally detected. In our study, five adsorption sites are identified, including a new bridge-like site, namely: on-top site (a₁), shallow hollow site (a₂), deep hollow site(a₃), bridge site(a₄), and bridge-like site (a₅) (shown in Fig.2). On a₁ and a₂ sites, the CO molecules are adsorbed perpendicularly to the surface. On the a₃ and a₄ sites, the CO molecules are found to be slightly tilted to the surface. The two tilting angles are about 1oand 12owith respect to the surface normal. The bridge-like site (a₅) is a new adsorption side which has not been mentioned both in experiments and theoretical studies. On this site, the carbon end of the CO is bonded to two Fe1 atoms and one Fe2 atom. The oxygen atom is also bonded to a Fe1 atom and the O-Fe bond length is 2.175Å. The CO tilting angle is 39° with respect to the surface normal on this site. Table 3 summarizes the results for various sites. The results indicate that at this coverage, CO prefers to adsorb on the a₂ site, followed by the a₄ site, a₅ site, a₃ site, and a₁ site according to the adsorption energy of these sites. The adsorption energy difference between the a₂ site and the a₄ site is very small, 0.02eV, and that between the a₂ site and the a₅ site is also very small (0.1 eV). Both the adsorption energy and geometry of a₂, a₄ and a₅ are very similar, these three adsorption sites may be detected as one adsorption site by experiments. This is probably the reason that there are three or four adsorption sites observed by experiments. The next question is to decide which site to be the precursor for CO dissociation on the Fe (111) surface. Two factors are considered: one is the adsorption energy which decides the appearance possibility of the adsorption site; the other is the activation of the C-O bond (C-O bond length). Table 3 shows that the C-O bond length of a₃ and a₅ both are 1.221Å, but the adsorption energy of a₅(2.33eV) is higher than that of a₃(1.79eV). So the CO adsorbed on the bridge-like site is considered to be the precursor for CO dissociation on the Fe (111) surface. On the a₅ site, the oxygen atom bonding to a Fe1 atom also shows the tendency to dissociate.

Fig.2 Side view and top view of a₁, a₂, a₃, a₄, a₅

Table 3 Results for CO adsorption on Fe(111) surface.

Here E_ads is referred to adsorption energy; d_C-O, d_C-Fe are referred to the bond length of C-O bond, and C-Fe bond

3.2.2 DOS analysis
A vast amount of literature has described the nature of the interaction between CO and transition metals. It can be briefly summarized as follows: the occupied 5s orbital of CO can readily overlap with the unoccupied d_z² and 4s orbitals of the transition metal atoms with electron donation: the unoccupied 2p* orbitals of CO are also capable of overlapping with occupied d_xz and d_yz orbitals of the metal atoms with electron back-donation^[13]. The electronic density（DOS）is carried out in the text to analyze the binding mechanism. Fig. 3 shows the total density of states of the free CO and the partial density of states(PDOS) of the C and O. The four peaks below E=0 are 3s 、4s 、1p 、5s (from the left to the right) while the 2p* antibonding orbital is about 7eV above the 5s orbital. These results agree with the gas-phase photoelectron spectrum of CO qualitatively ^[14]. Comparison of the DOS of the free CO molecule with the PDOS of C and O atoms shows that the components of the 5s and 2p* orbitals are mainly the C atom orbitals. That is the reason why in most cases the CO is adsorbed on the metal surface with the C atom. Fig. 4 shows the PDOS of CO adsorbed on the five sites where one can notice that both the 5s and 2p* orbitals are broadened over a range due to hybridization with the iron orbital. The 2p* orbital, empty in the gas phase, is now partially occupied upon adsorption, resulting from interacting with the d-orbitals of the surface. The partial charge transfer leads to the broadening of the 2p* orbital with an edge below the Fermi level and the significant elongation of the C-O bond due to the antibonding nature of the 2p* orbital. For all these adsorption sites, the 5s orbital intensity is decreased dramatically. The PDOS of the CO adsorbed at various sites are very similar.

Fig.3 Total density of states of the free CO and the PDOS of the C and O

Fig.4 The PDOS of CO adsorbed on the five sites

4. SUMMARY
The interaction of CO with the Fe(111) surface have been studied by means of ab initio density functional theory. Five adsorption sites are investigated including a new site (bridge-like site) which is not discovered by the early experiments and theoretical studies. The new site is expected to be the precursor for CO dissociation on the Fe (111) surface. Our study finds that the CO molecule is mostly strongly bonded to the iron surface at the shallow hollow, which is in good agreement with the early experiments. The C-O bond is mostly activated in the bridge-like site. The binding mechanism is also analyzed by using DOS analysis.

REFERENCES
[1] Nayak S K, Nooijen M, Bernasek S L J. Phys. Chem. B, 2001, 105: 164.
[2] Stibor A, Kresse G, Eichler A et al. Surf. Sci, 2002, 507-510: 99.
[3] Seip U, Tsai M C, Christmann K et al. Surf. Sci, 1984, 139: 29.
[4] Whitman L J, Bartosch C E. J. Chem. Phys, 1989, 90: 2050.
[5] Mehandru S P, Anderson A B. Surf. Sci, 1988, 201: 345.
[6] Payne M C, Teter M P, Allan D C et al. Rev. Mod. Phys, 1992, 64 (4): 1045.
[7] Perdew J P, Burke K, Ernzerhof M. Phys. Rev. Lett, 1996, 77: 3865.
[8] Ge Q, Jenkins S S, King D A. Chem. Phys. Lett, 2002, 327: 125.
[9] Herzberg G. Molecular spectra and Molecular structure. I. Spectra of Diatomic Molecules, 1950, p.521
[10] Kittel C. Introduction to Solid State Physics, New York: Wiley, 1996.
[11] Sokolov J, Jona F, Marcus P M. Phys. Rev B, 1986, 33: 1397.
[12] Mortensen J J, Ganduglia-Pirovano M V, Hansen L B. Surf. Sci, 1999, 422: 8.
[13] Hansong C, David B R, Sheldon W D. J. Phys.Chem, B, 2001, 105: 12547.
[14] Plummer E W, Salaneck W R, Miller J S. Phys. Rev. B, 1978, 18: 1673.