Spin polarized d surface resonance state of fcc Co/Cu(001)

Spin- and angle-resolved photoemission spectroscopy has been applied to the study of the surface and bulk electronic structures of a face-centered cubic (fcc) Co thin film. We have experimentally resolved a negatively spin-polarized surface resonance state of fcc Co/Cu(001) at 0.4 eV below the Fermi energy. Moreover, we have found that the surface resonance state exhibits a strong spin–orbit interaction through an investigation of magnetic dichroism in the angular distribution spectrum of Co/Cu(001).


Introduction
Spin-polarized electron tunneling through the surface electronic states enables us to map the spin distribution on the surface of ferro-or anti-ferromagnetic materials [1]. It also contributes to the tunneling magnetoresistance (TMR) effect [2,3]. It is widely accepted that the magnetoresistance is mainly determined by the presence of spin polarized interfacial states between the insulating barrier and magnetic metal [4].
Spin-polarized surface electronic states of body-centered cubic (bcc) Cr(001), bcc Fe(001) and hexagonal closed packed Co(0001) have been well established and they are actually used for magnetic imaging in spin-polarized scanning tunneling microscopy (SP-STM) [2,3]. Facecentered cubic (fcc) Co(001) is one of the most widely used ferromagnetic materials for its attractive features of epitaxial film growth and good film stability and quality. However, surface electronic states near the Fermi level (E F ) of fcc Co(001) have not been experimentally identified, although there have been several studies with spin resolved and non-spin resolved photoemission on these films [5,6]. Recently, it has been pointed out theoretically that resonant electron tunneling should occur between fcc Co films separated by vacuum, where the electron wave number is restricted to some hot spots in the surface Brillouin zone (SBZ). Such ballistic tunneling could be caused by the formation of bonding and anti-bonding minority interface states [7]. In a realistic tunneling junction that contains an oxide barrier instead of a vacuum barrier, such surface electronic states at the ferromagnetic layer could be hybridized with the oxygen p state, a situation which might also be responsible for TMR.
Thus, the spin-polarized surface and interface states play quite important roles in atomically resolved magnetic microscopy/spectroscopy as well as in spin polarized electron tunneling through magnetic junctions. In the present paper, we have clarified the spin-polarized electronic states localized at the surface, as well as the spin-dependent bulk band structures, of fcc Co(001) thin film grown on Cu(001) by means of spin-and angle-resolved photoemission (SARPE) spectroscopy utilizing our recently developed spin polarimeter [8].

Experimental
A clean surface of the Cu substrate was obtained by repeated cycles of Ar-ion bombardment (1-2 keV) and annealing at 720 K in ultra high vacuum. A Co thin film was epitaxially grown on the Cu(001) single crystal using a commercial evaporator at a deposition rate of 0.1-0.2 monolayer (ML) min −1 in ultra-high vacuum conditions with the pressure below 3 × 10 −8 Pa. The substrate was kept at room temperature during the film growth. Cleanliness of the substrate and the thin film surfaces were confirmed by Auger electron spectroscopy (AES) and reflection high-energy electron diffraction (RHEED). The film thickness was monitored by the RHEED oscillation during the Co deposition and checked later by AES. SARPE spectroscopy was performed using our homemade system equipped with a combination of a hemispherical electron analyzer and a retarding type Mott spin polarimeter [8]. The photoemission measurements were done at room temperature. We used the unpolarized He resonance line (hν = 21.2 eV) for excitation. The angular acceptance of photoelectrons was set to 2 • at the pass energy of 5 eV, where the estimated total energy resolution was 110 meV. The photon incidence angle was 50 • with respect to the center axis of the electron analyzer, which corresponds to the surface normal for the emission angle θ = 0 • . The sample was magnetized along the in-plane 110 -direction by a coil wound around a µ-metal yoke. The 3 spectra were recorded by repeated scans and the direction of magnetization was switched for every sweep. Measurement time to complete one spin-resolved spectrum was as short as 20 min to avoid possible surface contamination. Although the measurement time was limited, data statistics were sufficient enough to resolve several spectral features, which are benefited from the improved efficiency of the spin detector combined with the high flux of incident photons.

Ab initio calculation
A first principles calculation was performed using density functional theory in the local density approximation (LDA). We used the full-potential linearized augmented plane-wave method. For the calculations, a cutoff of R MT K max = 2.25 × 3.8 was used and the two-dimensional Brillouin zone was sampled with 28 special k points. We calculated a 27-layer fcc Co(001) film with an in-plane lattice constant of 2.56 Å. The interlayer distance, d = 1.70 Å, was deduced from a total-energy minimization with the in-plane fcc Co lattice constant being coincident with the fcc Cu(001) lattice parameter. The position of the top three layers was allowed to relax in the direction of the surface normal. Compared to the ideal interlayer distance, the first interlayer spacing contracted by 4.8%, wheras the distance between the second and the third layer expanded by 2.6%. The contraction of the third layer spacing was small (0.36%). Figure 1(a) represents the SARPE spectra of a 6.5 ML Co thin film on Cu (001) at various emission angles (θ ) from 0 • to 47 • , corresponding to the¯ -X line of the SBZ. Here, the open up-and the filled down-triangles indicate majority and minority spin spectra, respectively. One can see that the spectra are dominated by the minority spin intensity for all the emission angles. In contrast to the clear peak structures in the minority spin channel, the majority spin spectrum shows broad and weak features. In figure 1(b), the majority spin spectra are enlarged. At θ = 0 • , the minority spin spectrum shows a clear double peak structure at about 0.12 (±0.06) and 0.36 (±0.03) eV below E F , whereas the majority spin spectrum only shows a broad peak around 0.8 (±0.1) eV. These structures are denoted as A and B for the minority spin spectrum and C for the majority spin spectrum.

Results and discussion
Except for the structure B, these observed valence band features at θ = 0 • are consistent with previously reported results, where the excitation energy was similar to that used in the present case (hν = 24 eV) [9]. The improved efficiency of the present spin polarimeter combined with a high flux of the incident photons enables us to complete a measurement while the surface remains clean. One can clearly see the spin-dependent dispersion curves of fcc Co(001). With increasing θ , the structure A in the minority spin channel shifts to higher E B and reaches a maximum around 37 • , corresponding to theX point. In contrast, the structure B appears only around θ = 0 • . For the majority-spin states, the peak positions are assigned on the majority-spin spectra by the arrows in figure 1(b). Here, the thin vertical bars indicate the photoelectron signals from the Cu 3d states excited by the HeI β satellite line; the dispersion of these features faithfully follows those appearing in the higher E B region excited by HeI α . The structure C shows an upward energy dispersion up to θ = 37 • and shifts back to higher E B from 42 • . The second majority spin state D comes in at θ = 12 • and shows a similar dispersion to the state C. Next, we compare the experimental band structures of fcc Co to the results of the tightbinding calculation at the measured wave numbers (k = 0.512 figure 2) assuming a free-electron parabola as the final state. Here, φ and V 0 are the work function (4.4 eV) and the inner-potential (15 eV) [10]. The experimental energy dispersion curves are retrieved from the intensity maxima of the observed spectra in both majority and minority spin channels. The energy positions determined in this way could sometimes deviate slightly from the realistic band structure, we believe that such minor errors do not affect the arguments used throughout the rest of this paper. For the tightbinding calculation, the transfer integrals such as (spσ ) and (pdσ ) were determined not only for the first, but also for the second and third nearest-neighbor atoms so as to reproduce the result of the pseudo-potential calculation except for the exchange splitting [11]. The value of 1.5 eV was used for the exchange splitting E ex , which is close to the reported values of 1.4 eV by Schneider et al [12] and 1.55 eV by Clemens et al [5]. Note that the correspondence between the experimental and the calculated bands becomes worse if we use the smaller value of 1.2 eV reported by Mankey et al [13]. The open (full) circles represent the calculated band structures, and are denoted by α and α for the minority spin, and β, γ and δ for the majority spin. The experimental band dispersions are interpolated between the data points along the solid (dashed) line for the band A (C). The experimental curve A can be assigned to the calculated band α or α because their dispersion relations are similar to each other. Due to the limited energy and momentum resolutions, it is difficult to distinguish α or α experimentally. In the majority spin channel, the band C corresponds to β. In contrast, no reasonable agreement is found for the second-lowest band (D). The discrepancy could be partly solved if we suppose that the data points below k ∼ 0.7 Å −1 belong to the calculated branch γ and those above k ∼ 0.8 Å −1 correspond to δ. In the intermediate region of k = 0.4-0.8 Å −1 , we still find some data points that deviate from the calculated bands. This mismatch might be caused by the unresolved intensities contributed from the upper γ and the lower δ bands. Here, it should be noted that the perpendicular wave number component in the present experiment for θ = 0 • corresponds to the middle of the -X line of the fcc BZ. The observed features A and C at θ = 0 • are consistent with the former spin resolved ARPES experiments with tunable photons, which disperse along the -X line (normal to the [001]-direction) [5,9,13]. This further supports that the features A and C come from the bulk.
Thus, the observed features except for the structure B have been assigned to the bulk band. In order to check if the structure B comes from electron confinement inside the Co film, we have measured the spectra of films with various thickness. It is expected that the peak position of the quantum well state (QWS) would shift depending on the film thickness. Figure 3(a) shows the SARPE spectra for 5, 6.5 and 12 ML Co films. We find no energy shift of the peak B for all the Co film thickness. Furthermore, there is no spectral feature whose E B is dependent on the film thickness in this energy range. Therefore, the QWS can be excluded as being responsible for B. The band structure that is relevant to QWS formation is the 1 band along the -Xdirection. The energy gap of the Cu 1 band appears in the region 2-8 eV above E F , whereas no energy gap is formed below E F . Based on spin-resolved inverse photoemission spectroscopy, one can expect QWS formation inside the Co film in the unoccupied state, whereas the QWS is unlikely in the occupied state [14]. This is consistent with the present result.
Next, we show the effect of oxygen adsorption on B in order to see if this feature derives from a surface state or not. Figure 3(b) shows the SARPE spectra at θ = 0 • of a clean Co film and one with 0.1 L of adsorbed oxygen. As can be clearly seen here, the intensity of the structure B is diminished upon oxygen adsorption, whereas the intensity of the structure A remains unchanged. Therefore, the structure B can be assigned to a surface derived electronic state. At the bottom of figure 3(b), the difference (dot) between the minority spin spectra for the clean and oxygen adsorbed surface is shown as the fitted curve (solid line) by a Voigt function. Thus, the extracted difference spectrum shows a quite symmetric feature, whose energy width and position are 300 meV (full width at half maximum (FWHM)) and 450 meV. We note that the energy width of the surface derived state is much narrower than that of bulk state (∼600 meV). Here, one needs to consider the surface Umklapp process, which transfers the bulk state at the next¯ of the 2nd SBZ by the reciprocal lattice vector G 110 . In this case, one might possibly observe a similar surface-sensitive feature to the structure B. However, in that case, the energy width should be similar to, or wider than, the corresponding bulk state, which is opposite to the sharp peak observed in the present experiment. Therefore, the possibility of a bulk state derived from a surface Umklapp process can also be excluded. Figure 4(a) exhibits the minority spin spectra (down-triangles) with a smaller angle step of 4 • from θ = 0 • (¯ ) to 12 • ( 1/3 X ) for a clean 5 ML Co/Cu(001) surface, compared with those for the contaminated surface (thin solid line). Note that the surface derived state shifts toward E F and it seems to cross E F around θ = 12 • as clearly shown in the shaded areas of the spectra. Figure 4(b) shows the E-k relation for this minority surface state, which shows an upward dispersion and crosses E F around 1/3 X . Figure 5 shows theoretical majority and minority spin bands obtained by an ab initio slab calculation for the states with even symmetry with respect to a mirror plane that is parallel to the k-vector and perpendicular to the surface of fcc Co(001). The black squares indicate the states, of which more than 50% are localized at the surface. The gray line shows the projected bulk band. The calculated surface resonance state appears around 0.55 eV below E F at¯ , shows an upward dispersion, and crosses E F around 1/3 X . The other surface states are found near the energy gap at 1 eV around the middle of¯ -X . In contrast, we find no surfacederived states with odd symmetry along¯ -X (not shown). The energy position of the calculated surface state, mainly composed of a d z 2 component, is close to B at¯ and shows an upward dispersion, crossing at 1/3¯ -X . These features are consistent with our experimental result. Such a minority-spin surface resonance state was not seen in the experiments previously reported with hν = 45 eV because of the s-polarized incident light, which could only probe states with odd  symmetry [5]. This result supports our finding of a surface resonance state with even symmetry, which is accessible only with the use of p-polarized light.
Finally, we try to elucidate the influence of spin-orbit coupling on this minority-spin surface resonance state by changing the magnetization direction, as depicted in figure 6. Here, I +M and I −M denote the spin integrated photoemission intensities of the clean Co thin film magnetized along [110] and [110]. The spectrum for the plus magnetization direction contains two peaks (A + , B + ) at E B ∼ 0.1 and 0.5 eV, whereas only a single peak appears at E B ∼ 0.5 eV (B − ) for the minus magnetization direction. The normalized difference spectra defined as (I +M − I −M )/(I +M + I −M ) are shown in figure 6(b). Here, these difference spectra are referred 9 to as MDAD. The MDAD spectrum of the clean surface shows a hump (−4%) at 0.36 eV in the negative MDAD signal and a maximum (+10%) at E F . Upon oxygen adsorption, the hump at 0.36 eV disappears and a simple dispersive structure with a minimum at 0.4 eV and maximum at E F remains. We consider that the observed MDAD for the oxygen-covered surface originates mainly in the bulk, and is caused by the mixing between the two Co 3d bands with 5 and

Conclusion
In summary, we have studied the spin-dependent electronic band structures of a fcc Co thin film on Cu (001) using our new, higher efficiency, spin-and angle-resolved photoemission spectrometer. Thanks to the higher energy resolution and the reduced time needed for the measurements, we have successfully resolved the spin polarized Co 3d minority-spin surface resonance state around the¯ point in the SBZ. Moreover, from the MDAD spectrum, we found that the surface resonance states have a strong spin-orbit interaction. This discovery of a new surface resonance state could play an important role in SP-STM as well as in the fundamental understanding of the magnetic transport phenomena of tunneling junctions.