Critical area based yield prediction using in-line defect classification information [DRAMs]

Optically measured in-line defect data is used for critical area analysis based yield prediction. Because this data can be noisy, however, data can be filtered using kill ratios established from in-line defect to bitmap correlation by mask layer on arrayed devices. This paper reports results from increased granularity of the kill ratio analysis: in-line defect classifications are considered and individual kill ratios for each classification are calculated and used for yield modeling. Furthermore, performing automatic signature classification on the bitmaps and signature to defect correlation adds valuable insight into yield loss mechanisms and improves the accuracy of the yield model