Silicon oxynitride films: Ion bombardment effects, depth profiles, and ionic polarization, studied with the aid of the Auger parameter

Abstract

Thin films of Si3 N4, SiO2, and of silicon oxynitrides with compositions in the range 0.3≤O/N≤3.6 were deposited on silicon substrates to thicknesses between 20 and 150 nm by low-pressure chemical vapor deposition. Auger parameters of the films were measured during the study of ion bombardment effects and depth profiling, and as a function of the O/N ratios. The particular parameter used was based on the Si 2s photoelectron, and bremsstrahlung-excited Si KLL Auger lines. When normalized to constant ion dose, no significant differences in the effects of bombardment with argon ions in the energy range 0.5–4.0 keV could be found, and the equilibrium O/N ratio measured by x-ray photoelectron spectroscopy was close to that measured by Rutherford backscattering and elastic recoil detection analysis. During depth profiling no reduction of the oxynitride to elemental silicon was observed, in contrast to previous work using the low-energy Si LVV Auger peak with electron excitation; it is therefore recommended that to perform depth profiling of these materials while retaining artefact-free compositional and chemical information, the high kinetic energy Si spectral features should be used. At the oxynitride/silicon interfaces in some of the films additional features were found corresponding to intermediate Si chemical states. The widths of the Si 2s, 2p and KLL spectral features were measured at the stages in depth profiling at which equilibrium composition had been reached. In every case there was an increase in width as a function of oxygen fraction O/(O+N), from Si3 N4, through a maximum at about O/(O+N)≊0.5, followed by a decrease to SiO2. The maximum increase for the KLL Auger peak was ≊0.9 eV, but for the 2s and 2p only ≊0.35 eV, and the peak broadening was uniformly distributed about the peak centroid. The broadening is attributed to the appearance of a number of additional chemical states of Si too closely spaced to be resolvable, and probably arising from a defective film structure. The Auger parameters of the various films, measured also once equilibrium composition had been reached during profiling, showed what appeared to be a monotonic dependence on O/(O+N), thus supporting the random bonding model in which nitrogen atoms are substituted randomly by oxygen in the SiNz O4−z (z=0, 1, 2, 3, 4) tetrahedra. However, the Auger parameter versus composition plot was displaced slightly from the theoretical one in the direction of higher parameter than expected. When the Auger parameter was plotted against the function (n2−1 )/n2, proportional to the polarization energy, n being the measured refractive index of a film, a linear relationship was found within experimental error, but with a slope different from that for the standards SiO4/2 (silica), SiO6/3 (stishovite), Si3 N4, and Si. This divergence is interpreted in terms of a lower O ionic polarizability in the oxynitrides than in the oxides.

Comment on ‘AES and SAM microanalysis of structure ceramics by thinning and coating the backside. Yu and Jin’

“…In particular, when E Ar C ¾ 1-4 keV, the Si-Si bonding states appear to be confined chiefly at the outermost layers of the sample, because Si LVV spectra (electron inelastic mean free path ¾ 0.4 nm) 4 show a damage component at KE ¾ 92 eV, whereas the more 'bulksensitive' Si KLL transition ( ¾ 3 nm) 4 remains virtually unaffected. 3 (Similar conclusions also were drawn by this author in considering the synchrotron radiation photoemission Si 2p spectra of Ar C -etched silica surfaces carried out at the same surface sensitivity as that of Si LVV spectra.) 5 As a consequence, it has been recommended that use of the Si LVV line, both as a chemical indicator and as a measure of silicon concentration, should be avoided in the Auger analysis of such Ar Cetched materials.…”

“…E-mail: paparazzo@ism.rm.cnr.it successful in avoiding surface charging phenomena, it should be remembered that Ar C ions are quite likely to produce artefacts. Indeed, there is ample evidence in the literature 3 that Ar C ions produce silicon reduction as a result of preferential removal of nitrogen or oxygen in silicon nitrides, silicon oxynitrides and silica. In particular, when E Ar C ¾ 1-4 keV, the Si-Si bonding states appear to be confined chiefly at the outermost layers of the sample, because Si LVV spectra (electron inelastic mean free path ¾ 0.4 nm) 4 show a damage component at KE ¾ 92 eV, whereas the more 'bulksensitive' Si KLL transition ( ¾ 3 nm) 4 remains virtually unaffected.…”

Critical evaluation of the state of the art of the analysis of light elements in thin films demonstrated using the examples of SiOXNY and AlOXNY films (IUPAC Technical Report)

“…Although the quantitative analytical data were derived from the peak integrals, a nonlinear correlation between composition and Auger intensity was observed in [34], similar to [97]. The spectra show the decomposition of the oxygen-rich AlO X N Y film.…”

“…In [126], the authors concluded from the AES spectra that oxygen and nitrogen were mixed on an atomic scale in the oxynitride films, not forming a mixture of SiO 4 and Si 3 N 4 tetrahedra. This work was also mentioned in [97], where the authors indicate the problems of the investigations, such as specimen charging, electron beam-induced damage and sample decomposition, the inherent high, sloping, inelastic background, etc. The authors of this publication thus prefer XPS.…”

“…In a very comprehensive publication [97], the authors describe the advantages of XPS over AES for the analysis of SiO X N Y films as immense, especially if information on the coordination of the atoms is wanted. The beam damage is avoided and chemical shifts, line broadening, and fine structure are easier to study.…”