Bork, T.: New approaches to the III-V on silicon heteroepitaxy. , Berlin, Humboldt-Univ., Diss., 2007, 2007
Abstract:
Different stages of the III-V nucleation occurring during the metamorphic growth of InP on Si(100) surfaces were studied. In particular, the first stages were investigated in detail, i.e. the deoxidation and the As-termination, and finally the growth of thick InP films on the nucleation layer. The heteroepitaxial growth of the MOCVD-prepared III-V compound InP on Si (100) surfaces was characterized by optical in-situ spectroscopy, benchmarked with various surface science tools, and studied with transmission electron microscopy. During the MOCVD preparation C and O contaminations were removed from Si(100) as it was shown by XPS and UPS. For the first time, tertiarybutylarsine was employed as non-gaseous, much less-toxic precursor compared to arsine to prepare clean arsenic-terminated Si(100). Cleaning and deoxidation occurred at temperatures of around 900°C, which is significantly lower than it is known for UHV (MBE) preparation. The thickness of the passivating As-termination was about one atomic monolayer. Even and odd-numbered atomic steps occurred on the surface always resulting in a (2x1)/(1x2) twodomain surface. The domain structure of the Si(100) surface was independent from the angle and the degree of the misorientation, also including exactly oriented Si(100). Different, non-bulk-like double row structures were observed on the As-terminated Si(100) surface via scanning tunneling microscopy (STM). These double row structures have been modeled and described theoretically as so-called ¿5-7-5¿ atomic ring structures at the National Renewable Energy Lab by McMahon et al., where concomitant studies have been performed in perfect agreement. These atomic ring structures do not appear on Si(100) surfaces treated with pure arsenic. They are energetically favorable compared to MBE-prepared bulk-like steps and flat surfaces and occur at steps, trenches, and ridges. In-situ reflectance difference (RD) spectroscopy signals taken of the As-terminated Si(100) surface showed the same fingerprint as single-domain, (2x1)-reconstructed surfaces obtained in UHV via MBE by Kipp et al. (PRL 76 (1996) 2810), although LEED images of samples prepared in this work displayed a (2x1)/(1x2) two-domain surface. For the first time, temperature-resolved RD spectra down to 20 K were measured and did not show any energetic shift of the RDS peaks. Thus, the origin of the RD spectra appears to be different from the attribution made by Kipp et al. in their theoretical calculations, i. e. electron phonon interaction and reduced lattice vibration has been excluded to be the origin of the RDS signals in the measurements of this thesis. The clean and well-defined As-terminated Si(100) substrate led to a reduction of defects for the heteroepitaxially grown InP layer. A termination of the Si(100) surface with phosphorous instead of As turned out to be unfavorable. On the two-domain, As-terminated Si(100) surface a 40 nm InP nucleation layer and, finally, a >1 µm P-rich state-of-the-art InP film was grown that displayed a (2x1), single-domain structure. AFM displayed that the nucleation layer was formed by about 20-400 nm crystallites with orientations parallel to the crystal axis. The defect density of the InP film was dependent on the size of the crystallites. For the first time, the entire growth process was monitored with in-situ RDS. Some approaches have also been made to grow GaAs on Si(100) employing a SrTiO3 buffer layer as it was proposed by Motorola.