Electronic States on Bi2Te3 Studied by Angle-Resolved Photoelectron Spectroscopy Using Synchrotron Radiation∗

We have studied the electronic structure within the topmost two quintuple layers (QL) on vacuum-cleaved nand p-type Bi2Te3(111) at 10 K by angle-resolved photoelectron spectroscopy. Tunable synchrotron radiation maximized the surface sensitivity of photoelectrons. The gapless surface-state band (SSB) in the angle-resolved photoelectron spectra clearly demonstrates the topological nature of the samples. Below the SSB, the bulk valence bands (BVB) is observed. Against the three-fold symmetry of the crystal structure, the k∥ dispersion of the BVB is highly symmetrical around the Γ̄, which indicates the six-fold symmetry. Moreover, most of the BVB show almost flat k⊥ dispersion perpendicular to the surface. These facts are the experimental evidence of a strong modification of the valence bands within the topmost two QLs into highly two-dimensional states by the advent of the surface. [DOI: 10.1380/ejssnt.2012.117]

As shown in Fig. 1, Bi 2 Te 3 is a layered material. The primitive unit cell is a rhombohedron which contains two Bi atoms and three Te atoms [26]. The conventional hexagonal unit cell is 30.497Å high and the edge of the basal plane is 4.386Å long. As in Fig. 1, one Te(2) layer is sandwiched by Bi and Te(1) layers on the both side, where Te(1) and Te (2) indicate crystallographically inequivalent sites. These five atomic layers form a quintuple layer (QL) and the QLs stack in the [111] direction. The interaction between the QLs are weak van der Waals force. Therefore, Bi 2 Te 3 readily cleaves between the Te(1) layers [14]. However, the interaction between the QLs must be appreciable since Bi 2 Te 3 is not a twodimensional but a three-dimensional topological insulator. Accordingly, relatively large k ⊥ dispersion perpendicular to the surface is predicted by many theoretical calcula-tions [17-20, 23, 24], where k ⊥ is the wavevector perpendicular to Bi 2 Te 3 (111). Upon cleavage, the topmost QLs must be strongly perturbed by losing the interacting partner QL, and as a consequence the SSB advents and its electronic structure must be strongly modified from the bulk one. Then, the electronic structure within the topmost two QLs is particularly important because the SSB localizes within them [27]. In these terms, we have studied the electronic structure of Bi 2 Te 3 (111) by ARPES tuning photon energy (hν) so as to maximize the surface sensitivity. We used hν ranging from 47.8 to 66.9 eV. At these hν, the inelastic mean free path for electrons, in other words, the escape depth of photoelectron for nearly normal emission becomes minimized to about 5Å according to Seah and Dench [28]. This means that 98% of the observed photoelectrons from the valence-band region come from the topmost two QLs, assuming that the photoelectron intensity decreases exponentially with travelling distance. Therefore, the surface sensitivity of the present study is sufficiently high.

A. Apparatus
All the measurements were conducted at the Saga-University beamline BL13 in Saga Light Source [29]. At BL13, synchrotron radiation, which is linearly polarized in the horizontal plane, is available from a recently upgraded planar undulator. The incident angle of the synchrotron radiation is 45 • . Because the hν from 40 to 800 eV is available, the observation of photoelectrons from not only valence bands but also shallow core levels is possible. The photon flux is in the range of 10 8 -10 11 pho-  tons/s/300 mA. The beam spot on sample is about 100 µm (horizontal)×30 µm (vertical) in size. The electron spectrometer used in the present experiment was MB Scientific A-1 analyzer. The total energy resolution was determined from the Fermi edge of polycrystalline Au and was estimated to be as low as 13 meV at hν = 47.8 eV and 17 meV at hν = 66.9 eV. The acceptance angle is ±7 • and the angle resolution is 10 mrad. The sample temperature is kept at 10 K during the measurements.

B. Sample preparation
We grew Bi 2 Te 3 single crystals in Yamagata University as in the following procedures. Commercial Bi 2 Te 3 powders were melt at 900 • C and then were crystallized several times followed by slow cooling with a rate of −10 K/h. As usual, thus obtained sample is p-doped with excess Bi substituting Te sites [1]. In order to change the carrier type, we added Te powders to the host p-type Bi 2 Te 3 and melt them at 900 • C again, followed by crystallization at the cooling rate of −10 K/h. The grown sample is indeed n-doped with excess Te substituting Bi sites.
Just before measurements, the sample was vacuum- cleaved in-situ and was transferred to another UHV chamber for photoelectron spectroscopy without exposing to air. The base pressure of the analyzing chamber was 2×10 −8 Pa. The crystallinity and cleanliness were checked by low-energy electron-diffraction (LEED) patterns and photoelectron spectra taken at hν = 670 eV, respectively. Actually, very sharp LEED patterns were observed, which indicates the high quality of the cleaved Bi 2 Te 3 (111). Figure 2 is the x-ray photoelectron spectrum for vacuum-cleaved n-Bi 2 Te 3 taken at hν = 670 eV, which shows that contaminations such as C and O are negligible whereas distinct C1s and O1s peaks are observed in the spectrum for air-cleaved n-Bi 2 Te 3 . Figure 3 shows the photoemission intensity plots of p-Bi 2 Te 3 along theΓ-M direction of the surface Brillouin zone (BZ) in the angle-resolved mode taken at (a) hν = 60.4 eV and (b) 66.9 eV. The Fermi level taken at hν = 60.4 eV and 66.9 eV corresponds to the Γ and Z points in the bulk BZ, respectively, assuming V 0 = 10 eV as the inner potential [8,9,30]. In the figure, the SSB and BVB are clearly resolved as explained below. One can see that the gapless SSB is clearly observed in the binding-energy range of 0.0-0.2 eV centered atΓ [10] (See Fig. 4 also). The presence of the gapless SSB is the direct experimental evidence that Bi 2 Te 3 in the present paper is surely a topological insulator. Below the SSB, there are observed the BVB. Most of the BVB below E Bin = 1.1 eV show symmetrical k ∥ dispersion around theΓ whereas others above E Bin = 1.1 eV are slightly asymmetrical. Here, k ∥ is the wavevector parallel to the surface. In Fig. 3 (b), for example, the band at 1.0 eV at theΓ point disperses upwards towards theM andM ′ points, however the dispersion to- wards theM ′ point is more steep. The symmetrical dispersion is unreasonable because we probe the topmost two QLs which has a three-fold symmetry and therefore thē Γ −M line is not equivalent to theΓ −M ′ line. The threefold symmetry was actually shown in the calculated and experimental BCB of Bi 2 Te 3 by Chen et al. [10] They observed the three-fold BCB with hν = 19 eV, which is more bulk-sensitive than our condition. Moreover, Noh et al. observed the BVB with hν = 23.0 eV, which is still bulksensitive than our condition, and the obseved BVB are more asymmetrical than the present result [9]. According to Larson et al. [17], the BVB have a larger Te(1) 5p character than Bi 6p or Te(2) 5p character. At the topmost surface, therefore, originally three-fold symmetrical BVB of mainly Te(1) 5p character get to have a six-fold symmetry, exhibiting the symmetrical k ∥ dispersion around theΓ point. The spectra at these two hν in Fig. 3 are very similar to each other and is in very good agreement with the spectrum at hν = 47.8 eV (Point Γ) for n-Bi 2 Te 3 (not shown). This fact indicates that the observed peaks in the spectra show no shift in energy and thus no dispersion perpendicular to the surface. This is against the fact that Bi 2 Te 3 is a three-dimensional topological insulator [10] and many theoretical predictions [17-20, 23, 24]. These observations indicate a strong modification of the valence bands within the topmost two QLs into highly two-dimensional states by the advent of the surface. Figure 4(a) shows the photoemission intensity plot for the SSB of p-Bi 2 Te 3 along theΓ-M direction in the angleresolved mode taken at hν = 60.4 eV. Figure 4 [10]. The band velocity for n-Bi 2 Te 3 is slightly smaller than that for p-Bi 2 Te 3 . Figure 5 compares the normal-emission photoelectron spectra of n-and p-type Bi 2 Te 3 taken at hν=60.4 eV (Point Γ). The intensity is normalized to the maximum height. The spectrum of p-Bi 2 Te 3 is shifted by 0.09 eV toward higher binding energy for comparison. The shift by 0.09 eV is smaller than the reported energy gap of ∼0.15 eV [3] and the direction of the shift is consistent with the p-type nature. The shift of the BVB and SSB by 0.09 eV is very close to that of the core levels by 0.10 eV. Other than the SSB, there are observed 10 peaks as numbered from zero to nine. Although the two spectra are almost identical to each other, there are some differences between the spectra of n-and p-type Bi 2 Te 3 . One of them is at the foot of the SSB. For n-Bi 2 Te 3 , the peak #0 is clearly found, however the peak is missing for p-Bi 2 Te 3 . We attribute the peak #0 as the impurity band as Greanya et al. [8] That is, the impurity peak #0 of n-Bi 2 Te 3 is the donor states due to excess Te which are barely excited at low temperature as 10 K. On the other hand, http://www.sssj.org/ejssnt (J-Stage: http://www.jstage.jst.go.jp/browse/ejssnt/) the acceptor states of p-Bi 2 Te 3 is almost unoccupied and therefore cannot be observed, which is the case. The other differences are found for the peaks #2, 4 and 7. These peaks for p-Bi 2 Te 3 are slightly smaller than those for n-Bi 2 Te 3 . These peaks might originate from Te 5p and the smaller concentration of Te in p-Bi 2 Te 3 might reduce the photoemission intensity.

IV. CONCLUSION
We studied the electronic structure within the topmost two QLs on vacuum-cleaved n-and p-type Bi 2 Te 3 (111) at 10 K by angle-resolved photoelectron spectroscopy. Tunable synchrotron radiation maximized the surface sensitivity of photoelectrons. The gapless SSB in the angleresolved photoelectron spectra clearly demonstrates the topological nature of the samples. Below the SSB, the BVB were observed. Against the three-fold symmetry of the crystal structure, the k ∥ dispersion of the BVB is highly symmetrical around theΓ, which indicate the sixfold symmetry. Moreover, most of the BVB show almost flat k ⊥ dispersion perpendicular to the surface. These facts are the experimental evidence of a strong modification of the valence bands within the topmost two QLs into highly two-dimensional states by the advent of the surface.