Prediction of LPCVD silicon film microstructure from local operating conditions using numerical modeling

Abstract
Numerical modeling of LPCVD silicon film microstructure was performed by means of two combined numerical codes. First, a a two-dimensional kinetic reactor model was used to calculate the conditions at the gas-substrate interface. Next, these conditions were used as imput parameters of a semi-empirical two-dimensional statistical model in order to predict the cristallized content of the deposits. This model also gave an image of the space distribution of crystallites inside the layer. A lot of experimental knowledge was required in order to develop a realistic aigorithm, accounting for the most important events. Most of the coefficients of the empirical laws were obtained by fitting the calculations to various experiments. Although the capabilities of our statistical model are yet limited, numerical simulations succeeded in predicting several microstructural features of LPCVD silicon films from operating conditions.