Using yield models to accelerate learning curve progress

The authors describe the application of yield models at SEMATECH to define the relative contribution of unit process steps and equipment to particle-limited yield. After validation, these detailed models can simulate the yield effects of process and equipment improvement plans. Yield models, used with short-loop defect monitors, allow immediate feedback of experiment effects to yield improvement efforts. Rapid feedback accelerates learning by bypassing normal processing cycle time delays. SEMATECH experience is used to outline methods for developing a detailed yield model, collecting manufacturing parameters, and validating results. Detailed yield models use the relationship between the learning curve and the process improvement cycle to accelerate the rate of performance-price improvement.<<ETX>>