Spin Polarization Study of Benzene Molecule Adsorbed on Fe(100

Oct 3, 2007 - National Institute for Materials Science, 1-2-1 Sengen, Tsukuba 305-0047, Japan, Hefei National ... The Journal of Physical Chemistry C ...
0 downloads 0 Views 662KB Size
J. Phys. Chem. C 2007, 111, 15289-15298

15289

Spin Polarization Study of Benzene Molecule Adsorbed on Fe(100) Surface with Metastable-Atom Deexcitation Spectroscopy and Density Functional Calculations X. Sun,†,‡ Y. Yamauchi,*,† M. Kurahashi,† T. Suzuki,†,§ Z. P. Wang,† and S. Entani† National Institute for Materials Science, 1-2-1 Sengen, Tsukuba 305-0047, Japan, Hefei National Laboratory for Physical Sciences at Microscale and Department of Physics, UniVersity of Science and Technology of China, Hefei, Anhui 230026, China, and PRESTO, Japan Science and Technology Agency, 4-1-8 Honcho, Kawaguchi 332-0012 Japan ReceiVed: May 26, 2007; In Final Form: August 18, 2007

The spin-resolved electronic states and spin polarization of a benzene molecule adsorbed on an Fe(100) surface are investigated using spin-polarized metastable-atom deexcitation spectroscopy (SPMDS) measurements and density functional calculations. The spin asymmetry observed by the SPMDS is found to be negative at the induced π* states close to the Fermi level and positive in molecular orbitals. The opposite spin asymmetry can be explained well by the calculated spin density distributions and plane-averaged density of states on the vacuum side. It is also found that the energy region of the negative spin asymmetry increases with increasing coverage of the benzene molecule. This phenomenon has been discussed in conjunction with the possible multiple adsorption sites, followed by an argument for the site-dependent spin polarization.

I. Introduction The chemisorption of organic molecules on metal surfaces is of great interest because of the significance of the interaction mechanism at the organic/metal interface and the wide range of applications, from the tuning of a self-assembled monolayer to catalysis and organic-based devices and so forth. Previous studies on the adsorption of organic molecules focus on their interaction with nonmagnetic substrates. Recently, great interest on organic spintronics was aroused after the observation of giant magnetoresistance in organic spin-valves.1 To advance toward electron spin manipulation, understanding the magnetic coupling between the adsorbed molecule and the substrate is of utmost importance. Spin-resolved spectroscopies, such as electronenergy loss spectroscopy,2,3 X-ray photoelectron spectroscopy,3,4 photoemission spectroscopy (PES),5 and spin-polarized metastableatom deexcitation spectroscopy (SPMDS),6 have been used to investigate the electronic structure of adsorbate-covered metallic surfaces and their magnetic properties. As the smallest aromatic molecule, benzene is a good model for the fundamental study of more-complicated aromatic hydrocarbons, such as pentacene, which consists of five benzene rings. The adsorption of benzene has been examined experimentally as well as theoretically on a large number of metal surfaces.7 Parallel adsorption of the benzene ring system to the surfaces is quite common, and the π orbitals are responsible for the bond between the benzene molecule and the substrate.8-10 No spin polarization has been found for the benzene adsorbed on magnetic surface, such as Fe(110) and Co(0001), with spin-resolved PES.11 However, the SPMDS measurements give evidence of the spin polarization for pentacene adsorbed on an Fe(100) surface.12 This is apparently puzzling if we simply consider the fact that pentacene is composed of five benzene rings. PES is known to be a useful technique for the determination and assignment of the valance * Corresponding author. E-mail: [email protected]. † National Institute for Materials Science. ‡ University of Science and Technology of China. § PRESTO, Japan Science and Technology Agency.

level of the adsorbate. Spin-resolved PES can precisely detect the exchange splitting of orbitals being responsible for the interaction between adsorbate and substrate. However, it is difficult to distinguish the adsorbate-induced state around the EF and the spin polarization of molecules from the backgrounds because of obscuring by the strong emission from the substrate d band. Unlike spin-resolved PES, SPMDS provides an extremely sensitive tool to study the spin-resolved electronic structure of the adsorbate on magnetic surfaces because the deexcitation involves electrons distributing outside of the surface.13 An evident case is the adsorption of CO on an Fe(110) surface. The adsorbate-induced 2π* states formed by the backdonation of metal electrons to the CO molecule are detected directly by SPMDS and found to be negatively spin-polarized.14 Another well-studied example is the adsorption of oxygen on Fe films. A change of the sign in the asymmetry of the spin polarization right below EF has been found with oxygen exposure above 3 L.6,15,16,42 In this paper, we report SPMDS measurements for benzeneadsorbed Fe(100) surfaces and spin-polarized density functional theory (DFT) calculations for single-molecule adsorption. Consistent with the results obtained by Getzlaff11 with spinresolved PES, exchange splitting has not been observed for either the σ orbitals or the π orbitals by SPMDS. However, our experiments show that spin polarization occurs in both the induced π* states (antibonding π states with e2u symmetry) and other molecular bonding orbitals of benzene. The SPMDS results are in good agreement with the calculated spin density and plane-averaged density of states (PDOS) on the vacuum side because SPMDS predominantly probes the electronic states extending toward the vacuum side of the topmost surfaces.13 On the basis of DFT calculations, we propose that the possible presence of multiple adsorption sites might be responsible for the increase of the magnitude and the energy region of the negative spin asymmetry (positive spin polarization) at high

10.1021/jp0740797 CCC: $37.00 © 2007 American Chemical Society Published on Web 10/03/2007

15290 J. Phys. Chem. C, Vol. 111, No. 42, 2007

Sun et al.

Figure 1. Spin-summed SPMDS spectra (a) and spin asymmetry (b) measured for various benzene-covered Fe(100) surfaces at 85 K. The highenergy cutoffs of SPMDS spectra are replotted with a magnification factor of 50 as in the scales on the right-hand side.

coverages. Electronic explanations of site-dependent spin polarization are also discussed extensively for single-molecule adsorption. II. Experimental Techniques and Results The experimental setup is described in ref 17 in detail. Here, we will present a short overview of the main points. A metastable atom source of a nozzle-skimmer discharge-type generates a metastable helium atom (He*) beam. Optical pumping with circularly polarized laser radiation of 1083 nm polarizes He* in the triplet state He(23S) into the Ms ) 1(-1) level of three magnetic sublevels under a small defining field (about 100 mG). Spin polarization greater than 95% for He(23S) and a population ratio of 9:1 between triplet and singlet He* are confirmed by the Stern-Gerlach measurement. An Fe(100) thin film is formed by depositing pure iron onto a MgO(100) single-crystal surface using an electron bombardment evaporator under an ultrahigh vacuum condition (about 10-8 Pa) and pulse-magnetized in-plane using a coreless coil. The Fe(100) surface is cleaned using Ar+ ion sputtering, annealed, and then cooled to 85 K by liquid nitrogen prior to exposure to benzene vapor introduced into the sample chamber through a variable-leak valve. A retarding energy analyzer consisting of a pair of grid meshes in front of an electron multiplier is employed to measure the energy-integrated energy distributions of secondary electrons ejected from the sample surface. The SPMDS spectra shown in Figure 1 are obtained by differentiating the energy-integrated count data from the electron multiplier. The input axis of the retarding energy analyzer is oriented at

30° with respect to the surface normal, while the spin-polarized He(23S) beam impinges along the normal. The defining field around the sample is adjusted so that the spin direction of He(23S) in Ms ) 1(-1) is aligned parallel (antiparallel) to the majority spin direction of the sample. The measurements are repeatedly taken with parallel (Ip) and antiparallel (Ia) spin orientations. The spin polarization is characterized by the parameter of asymmetry A, which is defined as A ) (Ip - Ia)/ (Ip + Ia). Note that the positive (negative) spin polarization of surface electrons to the majority spin gives negative (positive) spin asymmetry in the SPMDS. In general, two different deexcitation channels are responsible for He*. If the work function is larger than the ionization potential of He*, then the deexcitation mechanism is described by resonance ionization (RI) with subsequent Auger neutralization. Alternatively, RI is suppressed, and a direct Auger deexcitation (AD) may arise as the dominant mechanism if the work function is smaller than the ionization energy of He(23S) or if the wave function overlap of the He(23S) 2s-electron with empty states at the surface is insufficient because of the presence of an adsorbate layer. In the AD process, the 1s hole of the He(23S) atom is filled by an electron from the surface and the 2s electron is ejected with a kinetic energy of Ekin ) E* - EB - φ (E*, effective excitation energy of He(23S); EB, binding energy of electron; φ, work function). Assuming an equalejection probability for electronic states at the major deexcitation positions,18 the SPMDS spectra in the AD process mostly reflect the local density of the occupied surface electronic states at around the interaction surface for He(23S) atoms reaching the

Spin Polarization Study of Benzene on Fe(100) classical turning point. Figure 1a shows the experimental spinsummed SPMDS spectra for benzene adsorbed on Fe(100) surfaces with various exposures at 85 K. The kinetic energy of 16 eV corresponds to the Fermi level (EF). The clean Fe(100) surface spectrum exhibits a fairly smooth feature, which is due to the RI + AN.19 As the exposure of the benzene molecule increases, the dominant deexcitation process becomes the AD mechanism. The peaks due to the AD involving benzene molecular orbitals appear and become more and more dominant at about 3.1, 4.8, 6.0, 7.8, and 10.7 eV as the multilayer region increases. Similar evolutions of peak appearances are reported for benzene on Pt(111),20 Mo(100),21 and Cu(100)22 surfaces. Their typical onset exposure of the AD peaks is 1.5 L, and saturations are observed at exposures higher than 5 L. We assign the AD peaks according to a report for the benzene/Pt(111) surface.20 Among the AD peaks, the 1e1g(π) peak at 11 eV appears at 2 L and then reaches saturation at 4 L, while most of the σ relevant peaks at lower energies become visible at 3 L and still increase at 4 L. This discrepancy implies that the accessibility of He(23S) to the different wave functions of adsorbed molecules changes with increasing exposure. The benzene molecules in the monolayer are oriented parallel to the surface so that He(23S) can predominantly interact with π electrons, while He(23S) can also touch σ electrons of tilted molecules in the multilayer.21 The tilting of molecules in the multilayer, however, allows various kinds of packing patterns, leading to a complex mixture of molecular clusters with different orientations as well as of flat-lying monolayers and even bare surfaces. Although the full analysis of the molecular clusters is not straightforward, we can extract the contribution of flat-lying monolayers and bare surfaces from the SPMDS spectra. The high-energy cutoffs of SPMDS spectra are replotted with a magnification factor of 50, as shown in the scales on the righthand side of Figure 1a. A free benzene molecule or a benzene multilayer has a band gap at the Fermi level because of the closed-shell nature, while Fe has relatively large state density at and below the Fermi level owing to its d orbital. The magnified plots clearly illustrate the change of the electronic states at around the Fermi level. The SPMDS intensity for a clean surface in the bottom panel sharply diminishes at 14.5 eV, indicating the contribution of the significant density of states at the Fermi cutoff through the RI + AN process. On the contrary, in the top panel, almost no intensity is found in the energy region above 13 eV, designating a complete multilayer coverage with a clear band gap. At small exposures (1-4 L) to benzene vapor, a shoulder appears at 16 eV, which is also observed for benzene on Pt(111)20 and Cu(100).22 This shoulder would be originated from the AD involving the new electronic states induced from the π* states of the benzene molecule by the molecule-substrate interaction in the monolayer regions. Because the induced π* states should also distribute above the Fermi level, the unoccupied portion of the new state can keep the RI + AN channel more or less open besides the AD in the monolayer regions. Therefore, the SPMDS spectrum at 1 L exposure does not show appreciable AD peaks, while the high-energy cutoff extends its slope to 16 eV by the small contribution from the AD of the induced π* states. Interestingly, this high-energy cutoff still exhibits a sharp rise at 14-14.5 eV specific to the AN on the clean surface, which suggests the existence of bare surface regions. The disappearance of the sharp rise means the complete coverage by benzene at 2 L exposure. With increasing the exposure up to 3 or 4 L, the

J. Phys. Chem. C, Vol. 111, No. 42, 2007 15291 molecular cluster regions increase. However, the monolayer regions decrease, and, thus, the shoulder at 16 eV declines. No splitting is observed in the SPMDS peak positions between parallel and antiparallel spin orientations. This phenomenon is consistent with the spin-resolved PES result reported by Getzlaff et al.11 However, spin polarization is found for both the benzene molecular orbitals and the induced π* states by our SPMDS measurements. The corresponding spin asymmetry is shown in Figure 1b. The variation of the spin asymmetry profile is obvious, which has not been detected by spin-resolved PES.11 This may be attributed to the high sensitivity of the SPMDS in detecting electrons outside of the surface with the least influence of the electron emission from the substrate. With increasing exposure to benzene, the positive asymmetry at around EF becomes small and reverses to negative, while it keeps positive at the higher binding energy side of the benzene molecular peaks. Because a similar spin polarization behavior has been reported for the pentacene-adsorbed Fe(100) surface,12 the reversal of the spin polarization of the induced π* states can be a common behavior inherent to aromatic rings. In the present investigation, we will concentrate on the behavior of the induced π* state in monolayer regions in comparison with other molecular bonding orbitals. As the monolayer regions fully develop until 2 L exposure and are reduced at larger exposures, the induced π* states contribute to gradually change the SPMDS asymmetry at around EF from positive to negative. Molecular cluster regions, which may also contribute to the positive SPMDS asymmetries at the AD peaks, will be explored elsewhere, although our preliminary investigation has revealed consistent spin polarization for tilted adsorption geometries. III. Calculation Method and Results First-principles calculations are performed to investigate the spin-resolved electronic states of the benzene molecule adsorbed on the Fe(100) surface within the framework of DFT using a plane-wave basis set, as implemented in the Vienna ab initio simulation package (VASP).23,24 Exchange-correlation interactions are included through the generalized gradient approximation (GGA) in the Perdew-Burker-Ernzerhof form.25 The electron-ion interaction is described by the projector-augmented wave method in its implementation of Kresse and Joubert.26,27 The spin interpolation of Vosko et al.28 is adopted for spinpolarized calculations. The Brillouin-zone integration is sampled via a 4 × 4 × 1 k-point mesh, which is automatically generated using the Mokhorst-Pack method.29 The plane-wave energy cutoff is set to 400 eV for all calculations. The surface is modeled by a seven-layer Fe slab through a (3 × 3) unit cell with a vacuum region of 18 Å. A benzene molecule is adsorbed on one side of the Fe slab. During the structural optimizations, the benzene molecule and the top two substrate layers are allowed to relax freely according to Hellmann-Feynman forces. The calculated results for the adsorption geometry, electronic structure, and spin polarization are presented in the following subsections. A. Adsorption Geometry. The geometric structures of bulk Fe and a free benzene molecule are calculated first. The GGA calculation yields a lattice constant of 2.833 Å and a magnetic moment of 2.20 µB/atom for bulk bcc Fe, which agrees well with the experimental values of 2.866 Å and 2.22 µB/atom,30 respectively. The optimized equilibrium C-C and C-H bond lengths of a free benzene molecule are 1.398 and 1.092 Å, again in good agreement with the respective experimental values of 1.399 and 1.101 Å.31 Four adsorption geometries with the center

15292 J. Phys. Chem. C, Vol. 111, No. 42, 2007

Sun et al.

Figure 2. Sketch of the adsorption geometry of benzene molecules on an Fe(100) surface. Note that carbon atoms can be classified into two types and labeled as C2, which represents two identical carbon atoms, and C4, which represents four identical carbon atoms

TABLE 1: Calculated Optimal Geometries and Energies of C6H6/Fe(100) for Different Adsorption Sitesa position

free

hollow

bridge1

bridge2

atop

dC-C (Å) dC-H (Å) θC-H (deg) θC2-C4(deg) dC-Fe (Å) Ead (eV) Edis(molecule) (eV) Edis(surface) (eV) Einteraction (eV) ∇HOMO-LUMO (eV)

1.398 1.092 0 0

1.441/1.455 1.095/1.115 21.7/17.7 0.546 2.018/2.340 -1.070 0.957 0.198 -2.225 4.486

1.441/1.481 1.107/1.091 33.4/23.5 10.642 2.011/2.228 -0.739 1.433 0.229 -2.401 4.005

1.422/1.444 1.090/1.097 14.0/6.6 0.526 2.177/2.821 -0.311 0.403 0.166 -0.880 4.606

1.432/1.407 1.095/1.104 4.7/7.1 -4.314 2.298/2.217 -0.0433 0.251 0.321 -0.615 4.646

0 0 5.087

a C-C bond lengths dC-C, C-H bond lengths dC-H, tilt angle of hydrogen atoms θC-H, bend angle of C2-C4 atoms θC2-C4, C-Fe bond lengths dC-Fe, adsorption energy Ead, distortion energy of molecule Edis(molecule) and surface Edis(surface), interaction energy Einteraction, and energy gap between the HOMO and LUMO orbitals of distorted isolated benzene molecule ∇HOMO- LUMO.

of the benzene ring located at the hollow, bridge1, bridge2, and atop sites, as shown in Figure 2, have been considered in our calculations. The adsorption energy and the optimized geometrical parameters at each site are shown in Table 1. Because of the symmetry of the system, the six carbon atoms of the adsorbed benzene molecule can be classified into two types. C2 represents two identical carbon atoms, and C4 represents four identical carbon atoms. The benzene molecule is predicted to prefer the hollow site with the adsorption energy of 1.070 eV. Adsorptions at the bridge1, bridge2, and atop sites are less favorable in the adsorption energy by amounts of 0.341, 0.759, and 1.027 eV, respectively. The adsorption of benzene molecules on a surface depends not only on the molecule-substrate interaction but also on the intermolecule interaction. The shifted alignment of two pentacene molecules with a shift of a half aromatic ring is energetically more favorable than the aligned alignment because of intermolecule interaction.41 The energy difference between the shifted and aligned alignments increases abruptly with the decrease of the molecule’s distance. The energetic advantage of the shifted alignment of two benzene molecules (with a shift of the half lattice constant of bulk Fe) is also demonstrated by our calculations. For example, the total energy of the shifted alignment with a distance of the double lattice constant (5.66 Å) is 0.46 eV lower than that of the aligned one. Thus, although the hollow site is energetically preferable, the intermolecular interaction will let the benzene molecule adsorb on not only the hollow site but also the bridge1 and bridge2 sites even at 1 L exposure in our experiment. The possibility of the atop-site adsorption is insignificant because of its very small

adsorption energy. It is necessary to investigate the adsorption properties of the benzene molecule at different sites. After adsorption, the carbon rings are found to be enlarged with the hydrogen atoms being pushed away from the substrate for all four adsorption geometries. Similar distortion has been observed when the benzene molecule adsorbs on other metallic surfaces, such as Ni,32 Pt,8 Co,10 and Au.33 The benzene ring remains almost parallel to the surface at the hollow and bridge2 sites but is bent downward at the bridge1 site and upward at the atop site (characterized by the bend angle of C2-C4 atoms θC2-C4). The total energy of the benzene molecule is increased by the molecular distortion. The molecular distortion energy Edis(molecule) is calculated as the energy difference between the free molecule before adsorption and the distorted molecule after adsorption. Similarly, the distortion energy of surface Edis(surface) is computed as the energy difference between the clean relaxed surface and the distorted surface due to adsorption. The adsorption energy can then be decomposed into three parts: Eads ) Edis(molecule) + Edis(surface) + Einteraction Einteraction is the interaction energy between the molecule and the surface. This decomposition has been demonstrated to be helpful to understand the adsorption energy.34 The distortion energy of the molecule at the bridge1 site is the largest, which is accordant with the longest stretching of C-C bonds and the largest tilting angle of C-H bonds. The stretch of the C-C bonds and the hybridization of the Fe-d states with the π states of the benzene molecule decrease the π-orbital overlap in the molecular ring. Compared with the free benzene molecule, the distorted benzene molecule has higher-lying occupied and lower-lying vacant π orbitals, which can develop stronger interactions with the vacant and

Spin Polarization Study of Benzene on Fe(100)

J. Phys. Chem. C, Vol. 111, No. 42, 2007 15293 TABLE 2: Binding Energies (eV), Referenced to the Fermi Level of the Individual Orbitals for a Free Benzene Molecule and Benzene Chemisorbed on Fe(100) Surfaces

Figure 3. Total DOS of a free C6H6 molecule, the local DOS inside the atomic spheres of the C4 atoms at different adsorption sites, and the local DOS of carbon-bonded Fe atom before (dotted line) and after adsorption (solid line) at hollow site, where EF was set to zero.

filled surface electronic states, respectively.8,34 Moreover, the tilting of the hydrogen atoms allows the carbon ring to get closer to the surface and to form stronger bonds. As indicated in Table 1, the larger distortion of the molecule (with larger distortion energy) results in a smaller gap between the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO), allowing for stronger interaction of the molecule with the surface. Then, a strong interaction does not necessarily mean a favorite adsorption because of the large deformation of the adsorption molecule. In the present case, the largest adsorption energy is found at the hollow-site adsorption while the bridge1-site adsorption has the strongest interaction with the substrate. B. Electronic Structure of the Surface Layer. The total DOS of a free benzene molecule, the local DOS inside the atomic spheres of the C4 atoms at different adsorption sites, as well as the local DOS of the carbon-bonded Fe atom before (dotted line) and after adsorption (solid line) at a hollow site are shown in Figure 3, where EF is scaled to zero. For a free benzene molecule, seven peaks in the energy range from -11 eV to EF correspond to the molecular orbitals of 1e1g, 3e2g, 1a2u, 3e1u, 1b2u, 2b1u, and 3a1g, respectively, while the π* states locate above the EF. The energies calculated for these molecular orbitals agree well with previous calculations.32 When a benzene molecule is adsorbed on the Fe(100) surface, as listed in Table 2, the energy levels of molecular orbitals shift toward higher binding energies than those of the free benzene molecule at all adsorption sites. The orbitals of 1a2u, 3e2g, and 1e1g are obviously broadened because of the overlap of these orbitals with the Fe 3 d bands. The π* states, which are empty in a free benzene

orbital

free

hollow

bridge1

bridge2

atop

1e1g(π) 3e2g(σ) 1a2u(π) 3e1u(σ) 1b2u(σ) 2b1u(σ) 3a1g(σ)

2.54 4.43 5.27 6.39 7.07 7.35 9.04

3.62 4.58 5.14 6.98 7.50 7.98 9.31

3.52 4.89 5.41 7.09 7.54 8.46 9.74

3.52 4.72 5.28 7.33 7.85 8.25 9.77

3.88 5.08 6.08 8.04 8.37 8.89 10.57

molecule, are partially occupied after adsorption. The spin polarization in the local DOS of the C atom is clear, particularly at energies close to the Fermi level. The predicted spin polarization of the benzene molecule originates from the interaction of the molecule with the magnetic Fe surface. In comparison with the corresponding Fe atom at a clean surface, several new peaks appear and locate at -9.4, -8.1, -7.6, -7.1/ 7.0, and -5.2 eV in the local DOS of the carbon-bonded Fe atom (inset in the top panel of Figure 3). They are induced by the 3a1g, 1b1u, 1b2u, 3e1u, and 1a2u orbitals of the adsorbed benzene molecule. These new peaks are positively spinpolarized; namely, the spin-up electron densities are higher than the spin-down electron densities. The local DOS(v) of the carbon-bonded Fe atom below EF shifts toward the Fermi level after the molecule adsorption. The shift can be clearly observed from the strongest peak at 2.43 eV by about 0.27 eV. This results in the reduction of the magnetic moment of the carbon-bonded Fe atom from 2.95 µB to 2.59, 2.19, 2.62, and 2.29 µB for adsorption at the hollow, bridge1, bridge2, and atop sites, respectively. A similar adsorbate-induced magnetization reduction has also been theoretically predicted in the system of benzene/Ni(111)35 and CO/Fe(110).14,36 The interaction between the adsorbate and the substrate causes significant electron donation and backdonation processes, which involve the preceding induced peaks of the carbon-bonded Fe atom and the filling of the empty π* states of the benzene molecule. To visualize the charge flow pertinent to the formation of the chemisorption bond, differential charge densities (∆n) are calculated. ∆n is obtained through the subtraction of the electron densities of the isolated adsorbate n(C6H6) and the clean substrate n(Fesub) from that of the adsorbate/substrate system n(C6H6/Fesub).

∆n ) n(C6H6/Fesub) - n(Fesub) - n(C6H6)

(1)

It is worth noting that n(C6H6) and n(Fesub) are computed with the distorted geometries as in the adsorbate/substrate system. The differential charge densities for benzene adsorbed at the hollow site are shown in Figure 4. The yellow or blue color of the differential charge densities indicates the gain or loss of electrons. The green spheres represent the Fe atomic sites at the first layer of the substrate. Here, the energy range is roughly divided into two regions. The first energy window is set from -20.0 to -2.0 eV, which covers all of the occupied molecular orbitals of free benzene (the Fermi level locates at 0 eV). Another energy window is close to the Fermi level (from -2.0 eV to EF), in which no state is present for the free molecule. At the first energy window, as shown in Figure 4a, the blue color around the π ring of the benzene molecule and the yellow color between the molecule and the substrate indicate the donation of electrons from the adsorbate into the Fe substrate, which induces new peaks in the local DOS of the carbon-bonded Fe atom in the inset of the top panel of Figure 3. At energy close to the Fermi level (Figure 4b), the color

15294 J. Phys. Chem. C, Vol. 111, No. 42, 2007

Figure 4. Differential charge density of C6H6/Fe(100) with a molecule adsorbed at the hollow site. The green balls represent Fe atoms. The yellow or blue color indicates a gain or loss of electrons, respectively. The iso-surface level is 0.01e/Å3. The energy region is (a) -20.0 to -2.0 eV, (b) -2.0 eV to EF.

around the benzene molecule becomes yellow, which is evidence of the electron backdonation from the Fe substrate to the π* states of the benzene molecule. Similar donation and backdonation processes have also been found when the adsorption site changes to a bridge1, bridge2, or atop one. C. Spin Polarization of Adsorbed Benzene. Conventional local DOS is obtained by integrating inside the atomic spheres and cannot exactly reflect the experimental SPMDS spectra, which predominantly probes the electronic states extending toward the vacuum side of the topmost surface. The planeaveraged DOS (PDOS), calculated by averaging the density of states over a plane parallel to the substrate surface, is helpful to discuss the experimental SPMDS results.37 Our experiment was conducted at various exposures of the benzene molecule at a temperature of 85 K. At extremely low coverages, the molecule should be dominantly adsorbed at the energetically favorable hollow site. However, even at an exposure of 1 or 2 L, other sites, such as the bridge1, bridge2, or even atop sites, are also possible for molecular adsorption due to intermolecular interaction, as described in subsection A. Thus, it is necessary to investigate various PDOSs of benzene/Fe(100) at different sites. Figure 5 shows the PDOS for four geometries. The distances shown in the right-hand side of Figure 5 are those from the C4 atoms. It is clear that the molecular orbitals can be

Sun et al. assigned in the PDOS and the peak intensity decays reasonably in an exponential way with increasing the distance from the surface. For all adsorptions at these sites, the integration of the PDOS over the energy range from -10.0 to -2.0 eV within the vacuum region results in weak dominance of the spin-down states, suggesting slightly negative spin polarizations at these molecular orbitals. This indicates that more spin-up electrons are transferred from the benzene ring to Fe atoms during the adsorbatesubstrate interaction process. This can also be seen from the positive spin polarizations at the molecule-induced peaks from local DOS of the carbon-bonded Fe atom, as shown in the inset of the top panel of Figure 3. The PDOS of the induced π* states is much smaller than those of other molecular orbitals when the plane is close to the carbon ring. It becomes comparable at a far distance such as 3 Å. This result indicates that the induced π* states extend toward the vacuum region and can be detected by SPMDS. Most interestingly, the spin polarization of the induced π* states depends on the adsorption site. For the hollow- and bridge1-site adsorption, the spin polarization of the induced π* states is obviously negative at a near distance (as shown in the inset figure in Figure 5 at a distance of 1 Å) but slightly positive at a far distance, such as 3 Å. For the bridge2-site adsorption, the PDOS at a distance of 1 Å shows a slightly negative spin polarization, while large positive spin polarization can be clearly seen at a distance of 3 Å. For the atop-site adsorption, positive spin polarization is found at a distance of not only 3 Å but also 1 Å, although the magnitude is much smaller at a near distance. The spin polarization of electrons can be well characterized by the spin-density distribution. Figure 6 shows the isosurface of the spin densities around the benzene molecule and the first layer of the substrate in the energy region from -20.0 to -2.0 eV for the hollow-site adsorption. The position of the benzene molecule is shifted upward about 2.6 Å to avoid hiding. The yellow or blue indicates the dominance of spin-up or spin-down electrons, namely, positive or negative spin polarization. Negative spin polarization can be clearly observed around the carbon ring. This is caused by the inequivalence between the spin-up and spin-down electron donation processes. As indicated in the inset of the top panel of Figure 3, more spin-up electrons are donated to Fe atoms, resulting in the positive spin polarization of the new peaks in carbon-bonded Fe atoms and negative spin polarization of the adsorbed benzene molecular orbitals. For the different adsorption sites, a small difference can be observed in the binding energy of the new peaks, but the polarity of the spin polarization should be the same. In contrast, there is much more difference in the spin polarization among these adsorption sites in the energy region from -2.0 eV to EF, as shown in Figure 7. The carbon ring is negatively spin-polarized for hollow-site adsorption. However, it is positively spin-polarized around the center of the benzene molecule. These positively polarized electronic states are found to extend toward the vacuum side much farther than the negative polarized states around the carbon ring. This is consistent with Figure 5, which shows negative spin polarization near the carbon ring plane but slightly positive spin polarization at a far distance from the surface. For bridge1-site adsorption, positive spin polarization is found within a belt surrounded by the four C4 atoms. Alternatively, outside this belt, especially near the two C2 atoms, it is negatively spin-polarized. At this adsorption geometry, the C2 atoms locate higher than the C4 atoms because of the strong molecular distortion, resulting in the farther extension of the electronic states of C2 atoms than those of C4

Spin Polarization Study of Benzene on Fe(100)

J. Phys. Chem. C, Vol. 111, No. 42, 2007 15295

Figure 5. PDOS at different adsorption sites. The distances shown are those from C4 atoms.

is observed around the C4 atoms, and positive one, near the C2 atoms. Moreover, the electronic states of the C2 atoms spill out much farther than those of the C4 atoms. This clearly explains the reason for the slightly negative spin polarization near the carbon ring plane and the large positive spin polarization at far distances for the bridge2-site adsorption. For the atop-site adsorption, the whole carbon ring including the center is positively spin-polarized, giving a reasonable explanation for the large positive spin polarization on the vacuum side. IV. Discussion

Figure 6. Spin density of C6H6/Fe(100) in the energy region from -20.0 to -2.0 eV with the molecule adsorbed at the hollow site. The yellow or blue color indicates positive or negative spin polarization, respectively. The z positions of the benzene molecule are shifted upward about 2.6 Å to avoid hiding.

atoms. Thus, from the vacuum side, the positively spin-polarized electronic states from the C4 atoms are partially shielded by the negatively spin-polarized ones from the C2 atoms. This can explain the fact that only a slightly positive spin polarization is obtained at its PDOS, even at far distances. The situation is the reverse for bridge2-site adsorption. Negative spin polarization

A. Comparison of Results. Here, we compare the SPMDS observations and the DFT calculations. As reported above, the AD process becomes dominant at exposures of benzene higher than 2 L. The SPMDS spectra in this process reflect the density of occupied states on the vacuum side of adsorbed molecules. The PDOS in the vacuum region shows a peak at around -10.0 eV, corresponding to the SPMDS peak at the kinetic energy of 3.1 eV. The PDOS peaks in the energy regions of -9.0 to -6.0 eV, -6.0 to -4.0 eV, and -4.0 to -2.0 eV correspond to the SPMDS peaks at about 6.0, 7.8, and 10.7 eV, respectively. Although the calculated energy levels are shallower, that are affected by the approximation of the many-body effects, the energy differences among them are in relatively good agreement with the experiment. The calculated PDOS shows slightly negative spin polarization in the energy region from -20.0 to -2.0 eV. This agrees well with the positive asymmetry of the SPMDS in the same energy region because the 1s hole of He(23S) atom, which is filled by a surface electron, has a spin opposite of the excited He* atom.

15296 J. Phys. Chem. C, Vol. 111, No. 42, 2007

Sun et al.

Figure 7. Spin density of C6H6/Fe(100) in the energy region from -2.0 eV to EF with a molecule adsorbed at a different site: (a) hollow, (b) bridge1, (c) bridge2, (d) atop. The yellow or blue color indicates positive or negative spin polarization, respectively. The z positions of the benzene molecule are shifted upward about 2.6 Å to avoid hiding.

The benzene molecule may adsorb at all possible sites, even at 1 L exposure, except at the atop site because of its very small adsorption energy (0.043 eV). However, because the PDOS shows a very small value of the spin polarization for the bridge1site adsorption, the spin density of the induced π* states should be contributed mainly by the hollow- and bridge2-site adsorption. For the bridge2-site adsorption, the PDOS(v) shows a large peak at -1.0 eV (Figure 5, at distances of 2 Å and 3 Å). Obviously, this leads to a highly positive spin polarization of the peak at -1.0 eV. The PDOS(V) also has a small peak at -0.25 eV, resulting in the decrease of the spin polarization in this energy region. For the hollow-site adsorption, the spin polarization of the induced π* states in PDOS is positive, particularly in the energy region close to the Fermi level. The total electron density of the induced π* states should be mainly positively spin-polarized if the number of benzene molecules at the hollow site is greater than or comparable to that at the bridge2 site. This positive spin polarization at around the Fermi level is consistent with the measured asymmetry of SPMDS close to the Fermi level, which changes gradually from positive (clean Fe surface) to negative after exposure of benzene molecules (Figure 1b). Our preliminary calculations for doublelayer or tilting molecule cases show no induced π* states, as expected from the weak interaction between the outmost atoms of adsorbed benzene and the substrate. Thus, the induced π* states exist only in monolayer adsorption regions, which may survive at higher exposures although the area is not much. The asymmetry of SPMDS shows a negative valley at 13.5 eV when the exposure increases to 4 L. This character corresponds well with the positive spin polarization at around -0.25 eV in the PDOS for the bridge2-site adsorption, indicating the preference of bridge2-site adsorption to hollow-site adsorption. Although the induced π* states are contributed by the

monolayer adsorption, multilayer stacks or tilting molecular clusters should develop with increasing exposure to 2 or 3 L. The second-layer molecule might prefer to adsorb on the firstlayer molecule at the hollow site due to its stability. Then, the ratio of the bridge2-site to the hollow-site adsorptions at residual monolayer would increase with the exposure. The dominant adsorption site in the monolayer region might be the bridge2 site when exposure increases to 4 L, resulting in a negative valley in the measured asymmetry. Further investigation should be conducted to verify the present description because our calculation is connected with single molecule adsorption isolated from surrounding multilayer stacks or tilting molecular clusters, which may influence the induced π* states. When the exposure of benzene reaches 8 L, a complete multilayer is formed and, as a consequence, the induced π* states disappear. However, the molecular orbitals still retain a positive spin polarization to some degree. The origin of the residual spin polarization will be discussed elsewhere concerning multilayer stacks or tilting molecular clusters. B. Site Dependence. Even within the monolayer adsorption, it has been found that the spin polarization of the carbon ring depends so much on the adsorption site. A detailed analysis of this remarkable phenomenon provides insight into the induced spin polarization. The electrons filling the π* states are brought through the electron backdonation from Fe atoms. The adsorption site dependence of the spin polarization of the induced π* states means that the spin-polarized backdonation processes are highly sensitive to the locations of Fe atoms. As indicated in Figure 2, four, two, two, and one Fe atoms at the first layer of the substrate are covered by or bonded to the adsorbed benzene ring at the hollow, bridge1, bridge2, and atop sites, respectively. For the hollow-site adsorption, the interactions or the backdonations between the substrate and the carbon ring come mostly

Spin Polarization Study of Benzene on Fe(100)

J. Phys. Chem. C, Vol. 111, No. 42, 2007 15297

Figure 8. Differential charge density for spin-up (a) and spin-down (b) electrons for the bridge2-site adsorption in the energy region from -2.0 eV to EF. The yellow or blue color indicates a gain or loss of electrons, respectively. The z positions of the benzene molecule are shifted upward about 2.6 Å to avoid hiding.

from the four carbon-bonded Fe atoms because the second nearest distances between the Fe atom and each C atom (d2nd) are evidently much longer than the nearest ones (d1st). For the bridge1-site adsorption, d2nd (3.384 Å) of the C2 atoms is much longer than d1st (2.011 Å), while d2nd (2.417 Å) of the C4 atoms is only slightly longer than d1st (2.228 Å). Thus, the electron backdonations to the C2 atoms are contributed chiefly by the two carbon-bonded Fe atoms (labeled as Fe1st-c), but those to the C4 atoms may come partially from the second-nearest Fe atoms (labeled as Fe2nd), which are not bonded directly to the carbon ring. Conversely, for the bridge2-site adsorption, d2nd (3.013 Å) of the C2 atoms is close to d1st (2.821 Å), while d2nd (3.061 Å) of the C4 atoms is significantly longer than d1st (2.177 Å). The induced π* states of the C2 atoms should be backdonated by both the Fe1st-c and Fe2nd atoms, and those of the C4 atoms, mainly by the Fe1st-c atoms. Similarly, for the atop-site adsorption, the interaction between the Fe2nd atoms and each carbon atom cannot be ignored because of the insignificant difference in d1st and d2nd of both the C2 and C4 atoms. Consequently, the Fe2nd atoms are involved in the interactions or backdonations with the C4 atoms for the bridge1-site adsorption, with the C2 atoms for the bridge2-site adsorption, and with all six carbon atoms for the atop-site adsorption. In reference to the spin density shown in Figure 7, all of the carbon atoms described above exhibit a positive spin polarization, while the others show a negative spin polarization. A possible conjecture here is that the spin polarities of the backdonated electrons from the Fe1st-c and Fe2nd atoms are the opposite of each other. To examine this, we plotted the differential charge density for spin-up and spin-down electrons, in the energy range from -2.0 eV to EF for the bridge2-site adsorption (in Figure 8). The blue (yellow) color in Figure 8a between the benzene molecule and the substrate indicates the loss (gain) of spin-up electrons. As indicated by the arrows, spin-up electrons are partially drawn back by the Fe1st-c atoms, and the residue is pushed toward the C4 atoms, as specified by the yellow color in these regions. Usually, the density of electrons outside the atomic spheres is sparse. Thus, the quantity of the backdonation of the spin-up electrons to the C4 atoms should be quite small. The colors are reversed inside the atomic spheres of the Fe1st-c atoms and at the space between the substrate and the C4 atoms for spin-down electrons (Figure 8b). This means that the spindown electrons at the space between the substrate and the C4 atoms become dense by the contribution from the Fe1st-c atoms. Subsequently, the backdonation of spin-down electrons is

expected to be more than that of spin-up electrons, resulting in a negative spin-polarized contribution from the Fe1st-c atoms to the C4 atoms. A similar backdonation process occurs on all six carbon atoms for the hollow-site adsorption and the C2 atoms for the bridge1-site adsorption, which corresponds to the negative spin polarization. As for the C2 atoms for the bridge2-site adsorption, the blue color is observed not only inside the Fe2nd atomic spheres but also at the space between the atoms of Fe2nd and C2 in Figure 8a, clearly indicating the backdonation of spin-up electrons from the Fe2nd to C2 atoms. The backdonation of spin-down electrons from the Fe2nd atoms seems somewhat less because the blue color (electron loss) is only observed inside the Fe2nd atomic spheres but does not appear at the space between the Fe2nd and the C2 atoms in Figure 8b. This means that these Fe2nd atoms provide positively spin-polarized electrons to the C2 atoms. The spin polarization of the C2 atoms is finally determined by the competition of the electrons from the Fe2nd and Fe1st-c atoms. Many more electrons from the Fe1st-c atoms are expected to be backdonated to the C4 atoms than to the C2 ones because the distance from the Fe1st-c to the C4 atoms (2.177 Å) is much shorter than that to the C2 atoms (2.821 Å) and each carbonbonded Fe1st-c atom has two C4 atoms already. Alternatively, there are two Fe2nd atoms contributing positively spin-polarized electrons to each C2 atom. Thus, the induced π* states of the C2 atoms is believed to be predominantly filled by the electron backdonation from the Fe2nd atoms and be positively spinpolarized. The C4 atoms for the bridge1-site adsorption and all six carbon atoms for the atop-site adsorption can obtain positively spin-polarized electrons in a similar way. For the atop-site adsorption, the backdonation from the covered Fe1st-c atom is distributed to six carbon atoms, resulting in the failure of the competition with that from the Fe2nd atoms and the appearance of positive spin polarization of the total carbon ring. For the bridge1-site adsorption, the positive spin polarization of the C4 atoms is not very large. One reason is that each Fe1st-c atom is bonded only to one C2 atom and its backdonation to the C4 atom cannot be ignored. Another reason is that one C4 atom has only one Fe2nd atom to contribute the positive spin-polarized electrons. Furthermore, considering the higher positions of C2 than C4 atoms, only weak positive spin polarizations in the PDOS at this site seem reasonable. As for the positive spin polarization around the center of the benzene ring adsorbed at the hollow site, it can be attributed to the extended dz2 orbital of the Fe atom (shown in Figure 9), which is located im-

15298 J. Phys. Chem. C, Vol. 111, No. 42, 2007

Sun et al. China (no. 60306006). We thank T. Sasaki and Q. X. Li for advice on total energy. References and Notes

Figure 9. Local DOS of dz2 states of the Fe atom located at the second layer of the substrate below the benzene ring for hollow-site adsorption.

mediately below the benzene ring and at the second layer of the substrate. The site dependence of the spin density around the Fermi level can be explained on the basis of the conjecture that the backdonation from the Fe1st-c and Fe2nd atoms causes opposite spin polarization. Although the leading mechanism is still unclear, the opposite spin polarization for a magnetic atom at different positions is not particularly rare. An example is found in clean ferromagnetic films, such as Fe(100) and Fe(110). The spin polarization of valence electrons is positive inside the magnetic atomic sphere but negative in the interstitial region.38-40 This negative spin polarization might be caused by the faster radial decrease of the wave functions for spin-up valence electrons, which have higher binding energies, than for the spindown valence electrons. Alternatively, for surface magnetic atoms positive spin polarization is found not only inside but also outside the atomic sphere that extends toward the vacuum region.38-40 This is attributed to the half reduction of the nearestneighboring atoms. The uncovered Fe2nd atoms may show a similar tendency to the clean-surface Fe atom, which has positively spin-polarized electrons in the vacuum region. As for the covered Fe1st-c atom, although it cannot be simply regarded as the Fe atom at the second clean surface layer, which is also covered by the first surface layer and has negative spin polarization in the interstitial region. There is similarity in its isolation from the vacuum by the covering layer. It should be noted, however, that the interaction between an Fe atom and an adsorbed benzene molecule is much less than that between Fe-Fe atoms. V. Conclusions The spin-polarized electronic structure of a benzene molecule adsorbed on an Fe(100) surface was investigated by SPMDS measurements and first-principles calculations. The peaks of 3a1g, 1b2u/3e1u, 1a2u/3e2g, and 1e1g and π* were observed on the SPMDS spectra. Positive spin polarization was detected for the induced π* peak, whereas other peaks corresponding to deeper levels indicated negative spin polarization. Theoretical calculations of the PDOS and spin density in the vacuum region provide ample support for the opposite spin polarizations. The site dependence of the spin polarization at the induced π* states is proposed to result from the competition of two backdonation processes between the carbon-bonded Fe atoms and the secondnearest-neighboring Fe atoms. Acknowledgment. This work was partly supported by the JSPS postdoctoral grant of Japan, the Atomic Energy Commission of Japan, and the National Natural Science Foundation of

(1) Xiong, Z. H.; Wu, D.; Vardeny, Z. V.; Shi, J. Nature 2004, 427, 821. (2) Moon, D. W.; Bernasek, S. L.; Dwyer, D. J.; Gland, J. L. J. Am. Chem. Soc. 1985, 107, 4363. (3) Moon, D. W.; Bernasek, S. L.; Lu, J. P.; Gland, J. L.; Dwyer, D. J. Surf. Sci. 1987, 184, 90. (4) Lu, J. P.; Albert, M. R.; Bernasek, S. L. J. Phys. Chem. 1990, 94, 6028. (5) Cameron, S. D.; Dwyer, D. J. Langmuir 1988, 4, 282. (6) Fo¨rster, S.; Baum, G.; Mu¨ller, M.; Steidl, H. Phys. ReV. B 2002, 66, 134427. (7) Steinru¨ck, H. P.; Huber, W.; Pache, T.; Menzel, D. Surf. Sci. 1989, 218, 293. (8) Morin, C.; Simon, D.; Sautet, P. J. Phys. Chem. B 2004, 108, 5653. (9) Chen, W. K.; Cao, M. J.; Liu, S. H.; Xu, Y.; Li, J. Q. Chem. Phys. Lett. 2005, 407, 414. (10) Pussi, K.; Lindroos, M.; Katainen, J.; Habermehl-Cwirzen, K.; Lahtinen, J.; Seitsonen, A. P. Surf. Sci. 2004, 572, 1. (11) Getzlaff, M.; Bansmann, J.; Scho¨nhense, G. Surf. Sci. 1995, 323, 118. (12) Suzuki, T.; Kurahashi, M.; Ju, X.; Yamauchi, Y. Surf. Sci. 2004, 549, 97. (13) Onellion, M.; Hart, M. W.; Dunning, F. B.; Walters, G. K. Phys. ReV. Lett. 1984, 52, 380. (14) Sun, X.; Fo¨rster, S.; Li, Q. X.; Kurahashi, M.; Suzuki, T.; Zhang, J. W.; Yamauchi, Y.; Baum, G.; Steidl, H. Phys. ReV. B 2007, 75, 035419. (15) Getzlaff, M.; Egert, D.; Rappolt, P.; Wilhelm, M.; Steidl, H.; Baum, G.; Raith, W. J. Magn. Magn. Mater. 1995, 727, 140-144. (16) Hammond, M. S.; Dunning, F. B.; Walters, G. K.; Prinz, G. A. Phys. ReV. B 1992, 45, 3674. (17) Yamauchi, Y.; Kurahashi, M.; Kishimoto, N. Meas. Sci. Technol. 1998, 9, 531. (18) Wang, S. W.; Ertl, G. Surf. Sci. 1980, 93, L75. (19) Penn, D. R.; Apell, P. Phys. ReV. B 1990, 41, 3303. (20) Sesselmann, W.; Woratschek, B.; Ertl, G.; Ku¨ppers, J. Surf. Sci. 1984, 148, 17. (21) Gu¨nster, J.; Liu, G.; Kempter, V.; Goodman, D. W. Surf. Sci. 1998, 415, 303. (22) Kurahashi, M.; Yamauchi, Y. Surf. Sci. 2005, 590, 21. (23) Kresse, G.; Furthmu¨ller, J. Phys. ReV. B 1996, 54, 11169. (24) Kresse, G.; Furthmu¨ller, J. Comput. Mater. Sci. 1996, 6, 15. (25) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. ReV. Lett. 1996, 77, 3865. (26) Kresse, G.; Joubert, D. Phys. ReV. B 1999, 59, 1758. (27) Blo¨chl, P. E. Phys. ReV. B 1994, 50, 17953. (28) Vosko, S. H.; Wilk, K.; Nusair, N. Can. Phys. J. 1980, 58, 1200. (29) Monkhorst, H. J.; Pack, J. D. Phys. ReV. B 1976, 13, 5188. (30) Kittel, C. Introduction to Solid State Physics; Wiley: New York, 1996. (31) Structure Data of Free Polyatomic Molecules, Landolt-Bo¨rnstein, New Series, Group II; Spinger: Heidelberg, 1992; Vol. 21. (32) Mittendorfer, F.; Hafner, J. Surf. Sci. 2001, 472, 133. (33) Chen, W. K.; Cao, M. J.; Liu, S. H.; Lu, C. H.; Xu, Y.; Li, J. Q. Chem. Phys. Lett. 2006, 417, 414. (34) Morin, C.; Simon, D.; Sautet, P. J. Phys. Chem. B 2003, 107, 2995. (35) Yamagishi, S.; Jenkins, S. J.; King, D. A. J. Chem. Phys. 2001, 114, 5765. (36) Ge, Q.; Jenkins, S. J.; King, D. A. Chem. Phys. Lett. 2000, 327, 125. (37) Kurahashi, M.; Suzuki, T.; Ju, X.; Yamauchi, Y. Surf. Sci. 2004, 548, 269. (38) Wang, C. S.; Freeman, A. J. Phys. ReV. B 1981, 24, 4364. (39) Ohnishi, S.; Weinert, M.; Freeman, A. J. Phys. ReV. B 1984, 30, 36. (40) Hong, S. C.; Freeman, A. J.; Fu, C. L. Phys. ReV. B 1988, 38, 12 156. (41) Lee, K.; Yu, J. Surf. Sci. 2005, 589, 8. (42) Moroni, R.; Bisio, F.; Canepa, M.; Mattera, L. Appl. Surf. Sci. 2001, 797, 175.