(Invited) Precipitation of Suboxides in Silicon

Abstract

Investigations of the precipitation of interstitial oxygen in Czochralski (CZ) silicon have been carried out since decades. Nevertheless, there are still open issues like the getter efficiency of oxide precipitates for metallic impurities and the recombination activity of oxide precipitates both in relation to their composition.Falster et al. found a discrepancy between the total loss of Oi and the visible loss of Oi in the form of precipitates observed by transmission electron microscopy (1). They explain this assuming that strained oxide precipitates, which are able to getter metallic impurities, and unstrained oxide precipitates so-called “ninja” precipitates, being not able to getter metallic impurities can exist (1). Until now, several phenomena are explained by the conversion from so-called unstrained to strained oxide precipitates. Murphy et al. (2) found that recombination at strained oxide precipitates is much stronger and rather depends on their density and less on their size. Typically, the density of strained oxide precipitates was assumed to be the density measured after preferential etching and the density of unstrained precipitates was always determined indirectly from the loss of interstitial oxygen.An important recent finding, obtained by electron energy loss spectrometry carried out by scanning transmission electron microscopy, was that the center of oxide precipitates consists of oxygen-rich SiO x being in most cases SiO2 surrounded by suboxide with decreasing x towards the edges. However, it is not yet clear if oxide precipitates always possess such a suboxide shell. It was also discovered that the suboxides surrounding oxide precipitates act as gettering sinks for Cu impurities (3). Therefore, the issue, if in any case oxide precipitates are surrounded by suboxide, could become an important criterion for efficient gettering.In order to further clarify these issues, we carried out a theoretical investigation of the phase composition of oxide precipitates and the corresponding emission of self-interstitials at the minimum of the free energy and their evolution with increasing number of oxygen atoms in the precipitates.First, we applied the model for spherical oxide precipitates. It was found that for low numbers of precipitating oxygen atoms the composition is SiO2 accompanied by a high emission of self-interstitials to relieve the strain. The emission of self-interstitials just slightly decreases up to a critical number of precipitated oxygen atoms. Then, the minimum free energy suddenly drops to low values of x, the stoichiometry changes into a suboxide, and the emission of self-interstitials becomes very low. The critical number of precipitated oxygen atoms increases with increasing temperature and decreases with increasing density of oxide precipitates. This means that the availability of suboxides for gettering of Cu impurities depends on both the annealing temperature and the density of oxide precipitates.If we apply the model to oblate spheroidal precipitates for low numbers of precipitating oxygen atoms the composition is also SiO2, the self-interstitial emission relieving the strain is very high, and the morphology remains spherical. However, when a critical number of oxygen atoms is reached the precipitate changes its shape from sphere to platelet, stops the emission of self-interstitials while maintaining a composition of SiO2. The optimum aspect ratio of the platelets lies between 0.02 and 0.03.The modeling results refer to homogeneous oxide precipitates under quasi-equilibrium conditions. In reality, the kinetics of morphology change is much more slowly than a change of stoichiometry. This means, when the octahedral morphology gets unstable suboxides can be expected to form as well. It was found by STEM investigations that the morphology change starts at octahedral morphology. Then the growth into [110] directions starting at the edges of the octahedron is enhanced until a platelet evolves from the octahedron and the original octahedron remains as fins on the platelet before it vanishes completely.The formation of suboxides at the edges of oxide precipitates after reaching a critical size can explain several phenomena like gettering of Cu by segregation to the suboxide region, lifetime reduction by recombination of minority carriers in the suboxide, and improved detectability by preferential etching. It thus provides an alternative explanation, based on minimized free energy, to the theory of strained and unstrained plates.(1) R. Falster, V. V. Voronkov, V. Y. Resnik, and M. G. Milvidskii, Proc. Electrochem. Soc., High Purity Silicon VIII (2004), Electrochem. Soc., Pennington, NJ, USA, Vol. 2004-05, p.188.(2) J. D. Murphy, K. Bothe, M. Olmo, V. V. Voronkov, and R. J. Falster, J. Appl. Phys., 110, 053713 (2011).(3) G. Kissinger, D. Kot, M. Klingsporn, M. A. Schubert, A. Sattler, and T. Müller, ECS J. Solid State Sci. Technol., 4, N124 (2015).