Identification for complicated shape objects by using stochastic fractal variables - categorizing dust particles on LSI wafer surface

Improving the yield and stability of LSI wafer production requires that the number of dust particles during every process of the production be decreased. To do this, we must develop a method of identifying dust particles. However, the current method of inspecting LSI wafers depends on human judgment; this makes it difficult to develop an algorithm to identify dust particles with complex shapes because mathematical models of dust particles have not been created. With this study, we propose the so-called “stochastic fractal variable” to make the mathematical model and identify the dust particles by using the maximum-likelihood method. Furthermore, we confirm the effectiveness of the identification method we propose by categorizing actual data.