Vacancy Self-trapping During Rapid Thermal Annealing of Silicon Wafers

The density and spatial distribution of oxide precipitates within a crystalline silicon wafer is of paramount importance for microelectronic device yield. In this letter, the authors show how the formation of previously unconsidered, very small vacancy aggregates can explain macroscopic spatial variations in the oxide precipitate density, which are observed following certain rapid thermal annealing conditions. The formation of these nanometer-sized voids is predicted on the basis of their recent model for vacancy aggregation that accounts for high temperature entropic effects. Comments Postprint version. Published in Applied Physics Letters, Volume 89, Issue 19, Article 191903, November 2006, 3 pages. This journal article is available at ScholarlyCommons: http://repository.upenn.edu/cbe_papers/82 Vacancy self-trapping during rapid thermal annealing of silicon wafers Thomas A. Frewen and Talid Sinno Department of Chemical and Biomolecular Engineering, University of Pennsylvania, Philadelphia, Pennsylvania 19104 Received 20 June 2006; accepted 21 September 2006; published online 6 November 2006 The density and spatial distribution of oxide precipitates within a crystalline silicon wafer is of paramount importance for microelectronic device yield. In this letter, the authors show how the formation of previously unconsidered, very small vacancy aggregates can explain macroscopic spatial variations in the oxide precipitate density, which are observed following certain rapid thermal annealing conditions. The formation of these nanometer-sized voids is predicted on the basis of their recent model for vacancy aggregation that accounts for high temperature entropic effects. © 2006 American Institute of Physics. DOI: 10.1063/1.2385069 The distribution of oxide precipitates also known as bulk microdefects or BMDs within a crystalline silicon wafer is of paramount importance for microelectronic device yield. While oxide precipitates are harmful if present in the near-surface device active region, they provide a critical metal gettering function in the wafer bulk. Oxygen precipitates also provide mechanical toughness, which is important for thermal annealing of large-diameter wafers. It is now well established that under most commercially relevant conditions, oxide precipitation is strongly dependent on the presence of single vacancies, which lower the thermodynamic cost of forming compressively stressed precipitates. It has been elegantly demonstrated that relatively straightforward manipulation of vacancy populations by rapid thermal annealing RTA of Czochralski-grown wafers can be used to precisely tailor the distribution of oxide precipitates in order to create a “denuded zone” near the wafer surface corresponding to low vacancy concentration while producing a high precipitate density in the wafer bulk high vacancy concentration . This profile results from the combination of point defect diffusion to and from the wafer surface and point defect recombination in the wafer bulk. Following RTA, regions of the wafer that have a residual vacancy concentration above a critical value, CV * 1–3 1012 cm−3, exhibit rapid formation of oxide precipitate nuclei in subsequent nucleation-growth thermal treatments. Recently, RTA experiments under certain annealing protocols, namely, high temperatures and cooling rates, demonstrated unusual BMD density distributions in which the maximum BMD density is observed to lie somewhere in between the wafer edge and the center, leading to a so-called m-like profile. Given the importance of BMD density distributions in silicon technology, we specifically address this density variation here by applying a recently developed model for defect diffusion and aggregation. The approach taken in this work is to first consider the predictions from a point defect-only model and then analyze the effects of including vacancy aggregation. In both cases, it will be assumed that the local BMD density following a nucleation-growth anneal is closely related to the local vacancy concentration at the beginning of the anneal. In other words, the BMD density is not explicitly considered but rather implicitly modeled via the vacancy distribution. In the point defect picture, only the diffusion and recombination of self-interstitials and vacancies are considered explicitly, so that