Singular value decomposition for deconvolution and physical insight into reflectance anisotropy in zincblende semiconductors

Abstract

Reflectance Anisotropy Spectroscopy (RAS), also known as Reflectance Difference Spectroscopy (RDS), is an optical probe with high surface specificity for cubic semiconductors. In this paper, we demonstrate the use of Singular Value Decomposition (SVD) for the deconvolution of the RA spectrum of GaAs(001) surfaces with an As-rich reconstruction into its constitutive spectral components. SVD analysis allows the decomposition of the RA spectrum in terms of a small number of linearly independent spectra, which in turn allows the determination of their physical origin. We identify three main components: a first component that is associated with the upper atomic layers, a second component that originates from the subsurface layer that is orthorhombically strained by the reconstruction of the GaAs surface, and a third component associated with the surface roughness due to the surface steps. Moreover, these obtained SVD components are well compared with both first-principles calculations and experimental reflectance anisotropy spectra reported in the literature. The results presented here are relevant for the use of RAS/RDS as a surface probe for cubic semiconductors. In particular, for the real-time in situ monitoring of epitaxial growth of zincblende semiconductors.