√ 21 × √ 21 phase formed by Na adsorption on Si(111) √ 3 × √ 3 -Ag and its electronic structure

Through electron diﬀraction observations, we found a √ 21 × √ 21 phase formation by Na adsorption of 0 . 1 ∼ 0 . 2 ML on a Si(111) √ 3 × √ 3-Ag surface below 250K. Its electronic structure was investigated by angle-resolved photoemission spectroscopy using synchrotron radiation. The phase was found to be metallic. At least six diﬀerent states were identiﬁed as surface states within the bulk band gap, whose dispersions were determined along a primary symmetry direction. The observed two-dimensional band structure had a very close resemblance with those of √ 21 × √ 21 phases induced by noble-metal adsorption on the same substrate. [DOI: 10.1380/ejssnt.2005.107]


I. INTRODUCTION
Adsorption of atoms or molecules on metal-atom induced Si reconstructed surfaces has recently attracted much attention among many researchers since it provides new nano-scale materials exhibiting interesting physical properties. [1][2][3] The systems are essentially regarded as playgrounds for low-dimensional physics because such surface modifications typically induce changes of surface states, which are electronically isolated from the bulk states.
In the present research, we have studied Na adsorption on the √ 3 × √ 3-Ag surface in order to acquire a general picture on electronic properties of the alkali-metal induced √ 21 × √ 21 phases. We have chosen Na, because it is the simplest alkali metal that does not interdiffuse into the Si substrate. [11] Through detailed electron diffraction observations, together with X-ray photoelectron spectroscopy * This paper was presented at The Sixth Japan-Russia Seminar on Semiconductor Surfaces (JRSSS-6), Toyama, Japan, 10

II. EXPERIMENTS
The ARPES experiments were done on a beam line BL-18A (Institute of Solid State Physics, University of Tokyo) in KEK-Photon Factory, Japan. [12,13] The measurements using linearly-polarized synchrotron radiation were performed with a commercial angle-resolved photoelectron spectrometer (VG ADES 500) with help of lowenergy electron diffraction (LEED) equipped at the beamline. The base pressure during the experiments was better than 1 × 10 −10 mbar.
First, a clean Si(111)7 × 7 surface was prepared by a cycle of in situ resistive heat treatments. The sample was p-type crystal (1 ∼ 10 Ω cm) with dimensions of 10 × 3 × 0.5 mm 3 . A Si(111) √ 3× √ 3-Ag surface was made by monolayer Ag deposition at the substrate temperature of ∼ 520 • C. A monolayer, 1 ML, corresponds to 7.83×10 14 atoms/cm 2 , the number density of Si atoms in the topmost layer of the (111) face. Deposition of Ag was done using a graphite effusion cell or alumina basket. The quality and cleanliness of the √ 3× √ 3-Ag substrate was ascertained by a sharp √ 3× √ 3-LEED pattern, Si 2p core-level photoemission spectra, and strong surface state signals in the valence band spectra. [14] Sodium adatoms were evaporated by using commercial SAES-getter sources which were thoroughly outgassed to minimize any impurity effects.
The ARPES spectra were measured at various emission angles (θ e 's ) along [110] direction with incident angles of photon (θ i 's ) at 45 • and 0 • . A photon energy (hν) of 21.2 eV was used with the overall angular resolution of ±0.5 • and the energy resolution of 0.1 eV. The E F  was determined from a Ta sample holder plate in good electrical contact with the Si sample. The ARPES spectra shown in this paper were taken at 130 K. prepared by additional Ag deposition at 130 K, and it is a summation of electron diffraction over two such domains. We found that the same Fig. 1(c). The Na coverage was calibrated through comparisons of the normalized Na 2s photoemission intensity, the intensity ratio between two core-level emission lines Na 2s/ Si 2p, work function, and diffraction pattern with those of the δ-7×7 phase prepared by ∼1 ML Na deposition on the clean 7×7 surface at room temperature. [13,17] Besides the experiments at the beamline, we have also performed in our laboratory reflection-high-energy electron diffraction (RHEED) experiments of Na deposition e-Journal of Surface Science and Nanotechnology  Fig. 2, the second possibility is denied. Therefore the Ag atoms in the √ 3 × √ 3-Ag substrate are notably affected by the additional Ag or Na adatoms, which may indicate that the Ag/Na adatoms adsorb directly on the Ag trimers, not on the Si trimers, as suggested by the first-principles calculations [22].  3-SBZ is adopted to describe the azimuthal directions, as followed after the earlier studies. [5][6][7][8] And we focused on a k -space region near the Γ 1 point along this direction since the region is crucial in the previous studies. [5][6][7][8] Six states, S 1 , S 2 , S 3 , S 4 , S 5 , and S 7 , are detected in the valence band spectra of measured energy range. Among them, two surface states, S 1 and S 4 , near E F are important for the discussion of the metallicity of the surface. Around emission angles (θ e ) of 32 • and 40 • , one can identify the S 4 state crossing E F . Then, it is obvious that the Si(111) √ 21 × √ 21-(Ag,Na) surface is metallic. The S 1 state is located at binding energy E B ∼ 0.9 eV at θ e = 32 • , and it disperses steeply upwards to a maximum energy of E B ∼ 0.5 eV at θ e = 36 • . But, instead of crossing E F , the S 1 seems to turn downwards beyond θ e = 37 • , implying that the state is not metallic and is fully occupied. At E B = 1 ∼ 3 eV, the other states, indicated by S 2 , S 3 , S 5 , and S 7 , are detected. Figure 4 (a) shows the experimental dispersion curves in gray scale for the spectral features observed along the Γ 0 -M -Γ 1 line, constructed from the spectra in Fig. 3. In this diagram, the intensities of the spectral features are approximately represented by the brightness in the gray scale by taking the second derivatives of the spectra. [21] The white solid curve is the edge of the projected bulk bands whose valence-band maximum was determined from shifts in photoemission lines of Si 2p core-levels and bulk valence bands. [18] Along the Γ 0 -M -Γ 1 line, the observed six states have their disper-sion curves within the projected bulk-band gap, indicating their surface-state nature. The overall band structure of the phase is described in detail elsewhere. [9] In Fig.4 (b), the present result is compared with the previous results on the noble-metal induced √ 21 × √ 21 phases. One can find perfect agreements of the surface band structure between the Na-induced phase and Au-or Ag-induced one. The band structures of the Ag-and Auinduced √ 21× √ 21 phases have been interpreted as modified states of the √ 3× √ 3-Ag substrate; the original states are energetically pulled down by electron doping from the Ag/Au adatoms. [6][7][8] Since Na is also monovalent as Ag and Au, and since Na has much lower electronegativity than that of Ag or Si, the observed band structure is reasonably ascribed to the same picture. The S 1 state of the √ 21 × √ 21 phase is assigned to an unoccupied surface state of the original √ 3 × √ 3-Ag substrate just above E F that has been pulled down below E F to be E B ∼ 0.9 eV.

III. RESULTS AND DISCUSSION
A simple electron density calculation confirms this picture as follows. Since the S 4 band is originally the S 1 band, its dispersion near E F is given by extrapolation of the S 1 band dispersion (Figs. 3 and 4). The density of states in a unit area of a two-dimensional free electron system is given by a constant D = m * /π 2 . Through a parabolic fit, E = 2 k 2 /2m * + E 0 , to the S 1 band dispersion, the effective mass and binding energy of the band bottom at the Γ point are given by m * = 0.25m 0 (m 0 is the free electron mass) and E 0 = −0.87 eV, respectively, which are similar to those of the reported noble-metalinduced √ 21 × √ 21 phases. [6,7] The effective mass becomes larger than that of the S √ 3 1 state of the original √ 3 × √ 3-Ag surface. [23] The charge density Q S1 filled in the S 1 band is then Q S1 = DE 0 = −9.2 × 10 13 e/cm 2 , where e is the elementary charge, and it corresponds to 0.12 ML of electrons. According to our adatom coverage estimation, the √ 21 × √ 21 phases are formed at 0.1 ∼ 0.2 ML, implying that there seems 3∼ 4 adatoms in a √ 21 × √ 21 unit cell (corresponding to the adatom coverage of 0.14 ∼ 0.19 ML). If each Na adatom provides about one electron to the substrate, we, then, find Q S1 ∼ Q Na . Despite the rough estimation, we can confirm semi-quantitatively the band filling picture with electron doping into the surface state by the adatoms. As a consequence, two bands, S 1 and S 4 , which are thought to be originated from an unoccupied band of the original √ 3 × √ 3-Ag surface, may be ascribed to band splitting at the zone boundary (ZB) of √ 21 × √ 21 SBZ, as shown in Fig. 4 [5,9]. Furthermore, with the first approximation of one electron transfer from a Na atom, the √ 21 × √ 21 phase seems to form at the coverage of 0.14 ML. A proper theoretical calculation is highly required to determine the √ 21 × √ 21 surface structure and the accurate adatom coverage.
The S 2 and S 3 surface-state bands have also been pulled down compared to the corresponding bands of the original √ 3 × √ 3-Ag surface. The downward shift is evident when compared with the previous ARPES results of the √ 3 × √ 3-Ag surface [6,7,14,19,20]. The S 5 band, on the other hand, appears at binding energy of around 1 eV, which is similar to that of the corresponding band of the pristine √ 3 × √ 3-Ag surface [6,7,14,19,20]. These facts seem to indicate that the surface states of the √ 3 × √ 3-Ag surface split into two groups, downward-shifted ones and non-shifted ones. This is due to partial occupation of the √ 3 × √ 3 site by the adatoms in the √ 21 × √ 21 unit cell. [22] There simultaneously exist perturbed sites and unperturbed ones, which produce the shifted states and non-shifted states, respectively. This is probable because of a small coverage (∼ 0.13 ML) of Na adatoms in the √ 21 × √ 21 phase.
Concerning the S 7 state, the band has its maximum at E B ∼ 1.8 eV at the Γ 1 point. The same band has been observed previously at the same E B for Si(111) √ 21× √ 21-Ag. A corresponding band is observed from the original √ 3× √ 3-Ag around E B ∼ 1.4 eV under the same measurement conditions. [7,24] The energy position of this state at Γ 1 corresponds to the uppermost Si(111) bulk band of Λ 3 symmetry at Γ 0 point, and the dispersion is also similar to the bulk band. [25] Thus, the S 7 band dispersion could arise from surface √ 3 × √ 3 umklapp scattering of bulk direct transition. The difference in binding energy of the S 7 state between the √ 3 × √ 3-Ag and √ 21 × √ 21 phases comes from the change in band bending, described above. Therefore, the observed √ 21 × √ 21 surface band structure is essentially interpreted to have the same origins as those of the Si(111) √ 3 × √ 3-Ag surface.

IV. CONCLUSION
Through LEED, RHEED, XPS, and ARPES measurements, we found a √ 21× √ 21-(Ag,Na) surface superstructure with Na deposition of 0.1 ∼ 0.2 ML coverage on a Si(111) √ 3 × √ 3-Ag surface below 250 K. The surface was found to be metallic, and the surface band structure was very similar to those of √ 21 × √ 21-Ag and √ 21 × √ 21-(Ag,Au) surface superstructures induced by Ag and Au adsorptions on the √ 3 × √ 3-Ag surface. At least, six surface-state bands are observed within the bulk-band gap, and they may be derived from those of the original Si(111) √ 3 × √ 3-Ag structure. The surface is expected to possess high electrical conductivity and its transport measurement is highly requested. [1,26,27]