Statistical fracture modelling of silicon with varying thickness

The strength of silicon is a statistical property and can differ significantly between specimens processed under similar conditions. This necessitates the development of different statistical distribution models, which, amongst others, includes Weibull and log-normal models to determine the statistical fracture strength of material like silicon. This paper describes the development of statistical approach to model fracture strength of silicon of varying thickness. An application of classic weakest link model was done to correlate the number of links undergoing failure and the Weibull modulus. The Weibull modulus for different die thickness was found to be lying in the range of 3-5. Particularly for thin silicon samples, the log-normal model gave a relatively lower value of complexity criterion. It was also observed that the fracture strength of 525 μm thick silicon could be described using a normal distribution. Pooled strength data analysis was also performed to interpolate the fracture strength of specimens of different thickness.