Automatic statistical determination of dislocation density in production SOI substrates

Abstract

Summarizes a successful prototype demonstration of an automatic etch pit counting system which employs a neural network program for dislocation identification over a wide exponential range required for SOI material analysis. Overall results indicate that the automatic dislocation counting system is feasible to employ in SIMOX manufacturing. The neural network system exhibited sufficient capability for accurate dislocation density analysis of both standard and thin BOX SIMOX material, with clear recognition and classification of enhanced silicon defects.