No abstract

Gd-silicide phases were investigated by x-ray diffraction. The results showed that not only one phase exists in a Gd thin-film and silicon substrate reactions. The first phase formed was hexagonal GdSi≊1.7, the second orthorhombic GdSi2. The ratio of the two phases depends on temperature of the heat treatment, and at a given temperature and time of annealing, a dependence of the thickness of the evaporated Gd layer was found. At ∼100-nm Gd thickness the dominant phase was orthorhombic GdSi2, at ∼250 nm hexagonal GdSi≊1.7. In the 300–1000-nm interval orthorhombic GdSi2 was the main component again. Rutherford backscattering analysis showed that the phases were found mixed within the layer. This thickness-dependent formation could be described with a simple model proposed by Gösele and Tu [J. Appl. Phys. 53, 3252 (1982)].

Epitaxial orthorhombic GdSi2 was grown by in situ vacuum annealing of a 50-nm Gd layer on 〈100〉 silicon. The epitaxy was proved by x-ray diffraction, electron diffraction, and ion channeling measurements. The lattice mismatch between the orthorhombic GdSi2 and 〈100〉 silicon substrate was found to be 4%.

Several-parameter fitting of multiple-angle-of-incidence ellipsometry data is developed to characterize near-surface layers on semiconductors damaged by implantation. The damage depth profiles were described by either rectangular, trapezoid-type, or coupled half-Gaussian (realistic) optical models. The rectangular model has three parameters: the average damage level and effective thickness of the implanted layer, plus the thickness of the native oxide. The trapezoid-type model is enhanced with a fourth parameter, the width of the back (a/c) interface. The realistic optical model consists of a stack of layers with fixed and equal thicknesses and damage levels determined by a depth profile function (presently the coupled half-Gaussians). Five parameters were used: the center, the height, and two standard deviations of the profile, plus the thickness of the native oxide. The complex refractive index of each layer is calculated from the actual damage level by the Bruggeman effective medium approximation. The method was tested on Ge-implanted silicon layers (at a wavelength of 632.8 nm) and was cross checked with high depth resolution Rutherford backscattering spectrometry and channeling.

No abstract