Table of Contents

Ion-Channeling Study of the SiC/Si/SiO2/Si Interface W. Jiang,(a,b) F. Namavar,(c) S. Thevuthasan, and W. J. Weber(b,d) Supported by EMSL Operations and BES/Materials. (a) Postdoctoral Research Fellow. (b) Environmental and Health Sciences Division. (c) Spire Corporation. (d) EMSL User. The cubic form of silicon carbide with the zinc blend structure (known as 3C-SiC or b -SiC) is an important wide-bandgap semiconductor material with outstanding properties, including high thermal conductivity and stability, high breakdown field, high saturated electron drift velocity, and small capture cross sections for neutrons. Electronic devices based on SiC are recognized as having great potential for applications involving high temperature, high radiation, high power, and high frequency. A procedure capable of preparing high-quality SiC films is a requisite for electronic device fabrication. Films of 3C-SiC, grown on Si wafers by chemical vapor deposition (CVD) methods, have been extensively studied since 1983 (Nishino et al. 1983). However, there is an increasing interest in the growth and study of SiC on Si/SiO2/Si (SIMOX) substrates. In this extended abstract, a summary of RBS/C results for 3C-SiC films grown on a SIMOX substrate is presented. A detailed description of this study was reported elsewhere (Jiang et al. 1999). Both thin (nominally 200 nm) and thick (nominally 2.4 mm) films of 3C-SiC for this study were grown by Spire Corporation using conventional CVD methods at a substrate temperature of 1350°C. The SIMOX substrate was prepared by high-dose, low-energy oxygen implantation followed by high-temperature annealing to regrow the top Si layer and redistribute the implanted oxygen (Namavar et al. 1993). The resulting SIMOX substrate consists of a top Si layer (nominally 180-nm thick) with a buried SiO2 layer (nominally 123-nm thick) on a Si (100) wafer. The RBS/C analyses were performed with a 3.4 MV tandem accelerator within the EMSL at PNNL. Experimental details have been described elsewhere. The RBS/C measurements along <100>- and <110>-axial directions were carried out using 2.0 MeV He+ beams at a scattering angle of 150°. A typical 2.0 MeV He+ RBS/C spectrum along <110> is shown in Figure 6.6a for a thin 3C-SiC film deposited on a SIMOX substrate. A random-equivalent spectrum, also shown in the figure, was obtained following sample rotations by 7° in polar and 3° in flip with respect to the <110> axis. The Si spectrum between channel numbers 300 and 580 consists of contributions from 1) SiC, 2) Si layer, 3) SiO2 layer, and 4) Si substrate. Due to different Si areal concentrations in the multiple layers and substrate, distinct Si front edges from layers 1, 2, and 4 appear, and are indicated by arrows in Figure 6.6a. Also marked in the figure is layer 3, based on the calculation of backscattering energy. The inserted schematic diagram illustrates the structure of the surface layers, the numbered surface/interfaces, and the scattering geometry of the experiment. A similar figure (not shown), which consists of <100> aligned and random spectra, was used to determine the minimum yields. The minimum yields cmin along <100> axis, obtained from the integral yield ratios of the channeling and random spectra in the small channel windows just behind the Si front edges, are ~28% for SiC, ~60% for the Si layer, ~100% for the SiO2 layer, and 63% for the Si substrate, respectively. It is usually expected that the dechanneling yield along the <110> axis is smaller than that along the <100> direction for face-centered cubic (FCC) structures (Chu et al. 1978) like 3C-SiC. However, the backscattering yields relative to random level show an "abnormality" for this sample. Such an enhanced dechanneling behavior along the <110> axis has been extensively studied for superlattice materials, both experimentally (Chu et al. 1982) and theoretically (Barrett 1983), and has been interpreted as due to small axial misalignment caused by strains at the interface. In order to study the lattice strain at the interface quantitatively, angular scans around the <110> axis were also performed for this specimen. The angular yield profiles are shown in Figure 6.6b for Si in the SiC, Si layer, and Si substrate. The dip curves for the Si layer and Si substrate coincide well with each other, but SiC shows an angular shift in the minimum yield position by ~0.16° with an accuracy of ±0.05°. This shift is an indication of a superlattice structure at the SiC-Si interface, and their <110> axes are misaligned by ~0.16°. These ion-channeling data clearly show the strained layer superlattice structure at the SiC-Si interface. Figure 6.6. (a) 2.0 MeV He+ <110>-aligned RBS/C spectrum for a 3C-SiC film (~160-nm-thick) deposited on a SIMOX (~160-nm-thick Si on ~107-nm-thick SiO2 on Si (100) wafer) substrate. Also included are a random-equivalent spectrum for the specimen and a <110>-aligned virgin spectrum for a 2.4-mm-thick 3C-SiC grown under the same conditions. (b) Angular yield profiles around <110> axis for Si element in the thin SiC film, Si buried layer, and Si substrate. Based on the models for the GexSi1-x/Si superlattice proposed by Bean et al. (1984), which describe a complete accommodation of the lattice mismatch by either strains or defects present at the interface, the following modified model is suggested in order to interpret our experimental observations. A schematic diagram is shown in Figure 6.7, where both defects and lattice strains are assumed to coexist at the SiC-Si interface. Due to the larger lattice constant of Si (a0=0.54307 nm), as compared to that of 3C-SiC (a0=0.43590 nm), there exist compressive strains in Si and tensile strains in SiC at the interface. The deformation of the crystal structure results in a directional change in the <110> axes for the two crystals, but the <100> axes can still remain in the same direction, as illustrated in Figure 6.7. This atomic configuration can qualitatively explain why the minimum yield positions coincide well along the <100> axis (not shown) and an angular shift emerges along the <110> axis (Figure 6.6b). The angular shift, determined from the experiment, can be directly applied to quantify the lattice parameters at the interface using an elasticity theory. To accommodate the large lattice mismatch (20%) between the SiC and Si crystal layers, the presence of defects, including misfit dislocations, can effectively reduce the lattice strain at the SiC-Si interface. Figure 6.7. Schematic illustration of the SiC-Si interface, showing the strained and defect-rich region. Lattice mismatch is exaggerated for clarity. References Barrett, J. H., Phys. Rev. B 28, 2328 (1983). Bean, J. C., L. C. Feldman, A. T. Fiory, S. Nakahara, and I. K. Robinson, J. Vac. Sci. Technol. A 2, 436 (1984). Chu, W.-K., F. W. Saris, C.-A. Chang, R. Ludeke, and L. Esaki, Phys. Rev. B 26, 1999 (1982). Chu, W.-K., J. W. Mayer, and M.-A. Nicolet, Backscattering Spectrometry (Academic Press, San Diego, 1978) p. 229. Jiang, W., S. Thevuthasan, W. J. Weber, and F. Namavar, Appl. Phys. Lett. 74(23), 3501-3503 (1999). Namavar, F., B. Buchanan, and N. M. Kalkhoran, Mat. Res. Soc. Symp. Proc. 284, 567 (1993). Nishino, S., J. A. Powell, and H. A. Will, Appl. Phys. Lett. 42, 460 (1983). William R. Wiley Environmental Molecular Sciences Laboratory Feedback: webmaster@emsl.pnl.gov Revised: June 12, 2001 Security & Privacy PNNL-13147