In-Line Evaluation and Statistical Processing of Defect Density in Production SOI Substrates

SIMOX (Separation by IMplanted OXygen) is one of the most successful technologies for achieving total dielectric isolation of active semiconductor device regions from the substrate. Advances in SIMOX semiconductor fabrication technology include radiation hardened memories and gate arrays for space applications, ULSI DRAMs for low voltage operation, thin film SOI BiCMOS for low power circuitry, and extreme (400 degrees C) temperature aplications of SIMOX circuitry. Recent studies have brought attention to the role of the device silicon dislocations in gate oxide reliability and integrity. As such, a significant issue in the manufacturing of SIMOX substrates for advanced electronics is the ability to ascertain, in real time, the presence and statistical analysis of dislocations present in the active silicon layer prior to complete wafer fabrication and customer shipment. Commercially available systems for automatic defect counting are not sensitive to the type and wide range (1E3 to 1E8/cm squared) of scale for the thin silicon film etch pits which can be optically observed. An opportunity is presented for the development of a commercially available, in-line, intelligent dislocation analysis system using neural network information processing and state-of-the-art imaging. The proposed Phase I research incorporated a system design which addresses identified input parameters and present commercial limitations.