High Quality InSb Films Grown on Si(111) Substrate via InSb Bi-Layer ∗

The heteroepitaxial growth of InSb ﬁlms via InSb bi-layer (Si(111)-2 × 2-InSb surface reconstruction) was studied. The InSb bi-layer was able to be prepared by 1 monolayer Sb adsorption onto Si(111)- √ 7 × √ 3-In surface reconstruction, as well as the results with 2 × 2-In or √ 3 × √ 3-In. The InSb ﬁlm grown via the √ 7 × √ 3-In surface reconstruction was fully rotated by 30 ◦ with respect to Si substrate. Because by using the √ 7 × √ 3-In surface reconstruction with higher In coverage, the area covered by the InSb bi-layer increased, and the area covered by 2 × 1-Sb surface phase which caused by desorption of In atoms from the InSb bi-layer decreased. Due to the decrease of the InSb crystals without rotation, which have poor crystal quality and electric properties, the electric properties of the ﬁlms improved than those of the samples grown via 2 × 2-In. The cross-sectional scanning transmission electron microscope image clearly showed that the InSb ﬁlm grown via the √ 7 × √ 3-In surface reconstruction has lower dislocation density than the sample directly grown on Si(111) substrate. [DOI: 10.1380/ejssnt.2009.145]


I. INTRODUCTION
Because of its highest electron mobility and saturation velocity of all semiconductors, InSb has been widely noticed as candidate materials for ultra-fast and very low power device [1]. The heteroepitaxial growth of InSb on Si has attracted much interest from a viewpoint of integration of InSb-based devices and Si-LSI. However, it is very difficult to achieve the heteroepitaxy of InSb on Si, because of the large lattice mismatch of about 19.3% between them. The other materials such as GaAs have often used to solve this difficulty [1][2][3].
We have studied the surface reconstruction by In and Sb atoms on Si substrate at the initial stage of the growth of InSb films [4][5][6][7], and found that InSb bi-layer (Si(111)-2×2-InSb surface reconstruction), prepared by adsorption of 1 monolayer (ML) Sb onto In-induced surface reconstruction, may be good solution of the lattice mismatch problem [8]. We have reported the heteroepitaxial growth of InSb films on Si(111) substrate via InSb bi-layer prepared by √ 3× √ 3-In and 2×2-In surface reconstruction [8].
The InSb films grown on the InSb bi-layer were rotated by 30 • with respect to Si(111) surface, and have good crystal quality and electric properties. In this case, the lattice mismatch between InSb and Si nominally improves down to about 3.3%. However, the InSb films included InSb crystals without rotation which leads to lower electron mobility. These unrotated InSb crystals are grown on 2 × 1-Sb surface reconstruction which formed during preparation process of InSb bi-layer. Sb atoms on Ininduced surface reconstruction on Si tend to substitute for In atoms from In-Si bonds [6,7,9]. The In atoms displaced by Sb are forced to locate on top of Sb layer, and easily desorb from the surface at lower temperature. The 2 × 1-Sb phase is formed due to the desorption of In atoms during the substitution process. The area covered by the 2 × 1-Sb phase also depends on the In coverage of the initial In-induced surface reconstruction. Because the sticking coefficient of In atoms on Sb surface phase is very low [10], the growth rate of InSb crystals on 2 × 1-Sb phase is also low, meaning the preferential growth of the rotated InSb. So, the increase of In coverage by using √ 7 × √ 3-In surface reconstruction may leads to increase the area covered by InSb bi-layer, and be effective to grow the fully 30 • -rotated InSb films on Si(111) substrate.
In this study, we report the heteroepitaxial growth of InSb films via √ 7 × √ 3-In surface reconstruction with higher In coverage than 2 × 2-In. To confirm the improvement of crystal quality by insertion of the InSb bi-layer, the cross-sectional scanning transmission microscope (STEM) images of the InSb films grown on the InSb bi-layer are also compared with the sample directly grown on Si(111) substrate.

II. EXPERIMENTAL
Si substrates with a dimension of 5× 15 × 0.6 mm 3 were cut from mirror polished p-type Si(111) wafers. The resistivity of the substrates at room temperature is about 20 Ωcm. After loading the substrates into a molecular beam epitaxy (MBE) chamber (base pressure: 2 × 10 −8 Pa) equipped with reflection high-energy electron diffraction (RHEED), they were flush annealed at 1250 • C to prepare a clean 7 × 7 surface. High purity (6N) indium and antimony were evaporated from each PBN K-cell. The substrate temperature (T s ) was monitored by an infrared pyrometer. The preparation process of InSb bi-layer is followed. First, 0.33ML-In atoms were deposited at 450 • C on the clean 7 × 7 surface to make √ 3 × √ 3-In surface reconstruction. After cooling down the T s to RT, the √ 7 × √ 3-In surface reconstruction was prepared by ad-sorption of additional In atoms of about 0.87ML (total In coverage: 1.2ML) onto the √ 3 × √ 3-In surface. There are two types surface phases with the same √ 7× √ 3 periodicity, but different structures, namely, the hexagonal phase with 1.0ML-In and the rectangular phase with 1.2ML-In [11]. In this work, we used the rectangular phase, because the √ 7 × √ 3-In-hexagonal phase is usually the mixed surface with 2 × 2-In. Then 1ML-Sb atoms were evaporated at 180 • C onto the √ 7 × √ 3-In surface reconstruction to prepare the InSb bi-layer. The InSb films were grown by the two-step growth procedure. In this procedure, the 30nm-thick InSb layer was grown on the InSb bi-layer at 200 • C as the first layer. The second layer was then deposited at 350 • C. The total film thickness of the samples in this work was about 1.1 µm. The grown InSb films were characterized by RHEED, X-ray diffraction (XRD) and Hall measurement, and compared with the samples grown via √ 3 × √ 3-In and 2 × 2-In surface reconstruction. In the XRD measurement, we also took φ scan patterns to know in-plane orientation of the films. In this case, the χ axis was tilted by about 70.5 • to detect the equivalent (111) planes ((111), (111), (111)), while keeping 2θ/ω axis to the bragg angle of InSb (111) or Si(111) peaks.

III. RESULTS AND DISCUSSIONS
The In coverage of the and this is larger than the saturation coverage of In of InSb bi-layer. To clear a question whether it is possible to prepare the InSb bi-layer via the √ 7× √ 3-In phase with higher In coverage, we compared RHEED patterns of the sample before and after the growth of InSb film. Figure 1 shows the RHEED patterns of the surface (a, b) after adsorption of 1ML-Sb onto the √ 7×  Fig. 1, it should be noticed that the relation between the line spacing of the first order streaks and beam incident azimuth became reverse before and after the deposition of InSb film. Because the angle between 1 10 and 2 11 azimuth is 30 • , the InSb film on the InSb bi-layer, which prepared via the √ 7 × √ 3-In phase, was rotated by 30 • with respect to the Si(111) substrate. This rotation of the InSb film is a peculiar feature of the growth of InSb films on the InSb bi-layer. So, these results indicate that the InSb bi-layer can be prepared by adsorption of 1-ML-Sb onto the √ 7 × √ 3-In surface reconstruction.
Because of very low sticking coefficient of In atoms on Sb surface phase [10], the growth rate of InSb crystals on the Sb surface phase is also low. This means the preferential growth of the rotated InSb crystals. So it is expected that the reduction of the area covered by 2 × 1-Sb surface phase lead the growth of the InSb films with good crystal quality and electrical properties. The preparation of InSb bi-layer via √ 7 × √ 3-In surface reconstruction with higher In coverage may be useful from a viewpoint of the reduction of 2 × 1-Sb surface phase. The XRD pattern (2θ/ω scan) of the InSb film grown via √ 7 × √ 3-In surface reconstruction is shown in Fig. 2.
For comparison purpose, that of the samples grown via √ 3 × √ 3-In and 2 × 2-In surface reconstruction is also shown in Fig. 2. These patterns look like very well. There are only three InSb peaks related with (111), (222) and (333) planes in these patterns, implying the epitaxial InSb. The full width at half maximum (FWHM) of the InSb(111) peak is about 300 arcsec in the Fig. 2(c). This value is narrower than that of the other samples (about 350 arcsec). This result implies the reduction of InSb crystals grown on the 2 × 1-Sb, which has poor crystal quality.
To confirm the reduction of InSb crystals grown on the 2 × 1-Sb (un-rotated InSb crystals), we measured the φ scan pattern of the (111) reflection of the InSb films.  Fig. 3(c). The origin of these weak peaks is InSb crystals grown on the 2 × 1-Sb surface reconstruction [8]. These weak peaks and broad peaks weakened with increase in In coverage of the initial surface reconstruction. Especially, those peaks disappeared in Fig. 3(c). This result indicates the reduction of 2 × 1-Sb phase.
The results of the XRD measurements indicated that the InSb film grown via √ 7 × √ 3-In surface reconstruction was fully 30 • -rotated with respect to Si substrate. This means the growth of high quality InSb films on Si substrate. To check the improvement of crystal quality by inserting the InSb bi-layer, we took the cross-sectional scanning transmission electron microscope (STEM) image of the samples grown via √ 7 × √ 3-In surface reconstruction. Figure 4 shows the STEM images of the InSb films with and w/o InSb bi-layer. Black and white layers on the upper side of the images are protection layers of tungsten and amorphous carbon for focused ion beam (FIB) process. As shown in Fig. 4(a), a large number of dislocations can be seen in the sample grown without InSb bi-layer, indicating lower crystal quality of the film. On the other hand, STEM image of the InSb film grown on the InSb bi-layer shows lower dislocation density than that in Fig. 4(a). This result clearly shows the improvement of crystal quality due to the decrease in the large lattice mismatch between InSb and Si by inserting the InSb bi-layer.
The STEM image and the φ scan pattern of the InSb film grown via √ 7 × √ 3-In surface reconstruction encourages us to measure the electric properties of the film. Hall measurement was performed by using the van der Pauw method. The maximum electron mobility and carrier concentration of the InSb film were about 18,000 cm 2 /Vs and 7.1 × 10 16 /cm 3 at 270K, respectively. The electron mobility is higher than that of the sample grown via 2 × 2-In surface reconstruction [8]. This improvement of the electric properties is caused by the reduction of InSb crystals grown on the 2 × 1-Sb surface reconstruction.

IV. CONCLUSIONS
The heteroepitaxial growth of InSb film on Si(111) substrate via √ 7 × √ 3-In surface reconstruction was shown and discussed. The InSb bi-layer, which leads to rotate the InSb film on it, was able to be prepared by adsorption of 1ML-Sb onto the √ 7 × √ 3-In surface reconstruction, as well as the results via 2×2-In and √ 3× √ 3-In. The φ scan pattern of the sample grown via √ 7 × √ 3-In surface reconstruction indicated that the existence of ±30 • -rotated InSb crystals, and the absence of InSb crystals without rotation. The STEM image showed lower dislocation density by insertion of InSb bi-layer. These results originate the reduction of the area covered in 2 × 1-Sb surface reconstruction due to the increase of the initial In coverage. The maximum electron mobility and carrier concentration of the InSb film grown via √ 7 × √ 3-In surface reconstruchttp://www.sssj.org/ejssnt (J-Stage: http://www.jstage.jst.go.jp/browse/ejssnt/) tion were about 18,000 cm 2 /Vs and 7.1 × 10 16 /cm 3 at 270K, respectively.