Automatic classification of wafer defects: status and industry needs

This paper describes the Automatic Defect Classification (ADC) beta site evaluations performed as part of the SEMATECH ADC project. Two optical review microscopes equipped with ADC software were independently evaluated in manufacturing environments. Both microscopes were operated in bright-field mode with white light illumination. ADC performance was measured on three process levels of random logic devices: source/drain, polysilicon gate, and metal. ADC performance metrics included classification accuracy, repeatability, and speed. In particular, ADC software was tested using a protocol that included knowledge base tests, gauge studies, and small passive data collections.