Automatic defect classification: A productivity improvement tool

The goal of this paper is to demonstrate a quantitative methodology for evaluating the effect that various in-line monitoring strategies have on the cost of defect excursions. An in-line monitoring case study using the IMPACT ADC from KLA-Tencor is the vehicle for demonstrating this data-driven methodology. Data was collected for this case study from wafers processed on a 64 Mb fabrication technology during a beta evaluation of IMPACT ADC at IBM Burlington. The overall benefit of an in-line monitor strategy that includes on-line ADC will be compared and proven superior to traditional line monitor strategies that include manual defect classification. Key advantages of on-line ADC that reduce the cost of excursions are high accuracy and a decrease in the overall time-to-results. Some qualitative factors such as operator skill level and training requirements will also be discussed