Random defect limited yield using a deterministic model

For success in the competitive semiconductor industry, the need to reduce cost per die is necessary. One way to accomplish this is to move to the next generation, or shrink technology to produce more die per wafer. Similarly, it is just as important to produce better die per wafer by minimizing the cycle time to detect and fix yield problems associated with the current technology. Wafer sort yield (good chips/wafer) can be broken into three components: random defect limited yield, systematic yield, and repeating yield loss. Random defect limited yield is due to defects, primarily caused by process equipment and their byproducts. Defects, usually randomly distributed, can also be localized to one or multiple die on a wafer. In-line QC inspection tools can detect most defects. Systematic yield losses are process-related and can affect all or some die on a wafer, or die by wafer region. Systematic yield losses are not detectable by in-line QC defect inspection tools. Repeating yield loss is due to reticle defects occurring on the same die within a reticle field. Repeating defects are caused by contamination on the stepper lens, by contamination of the pellicle that protects the reticle, or by reticle contamination. Reticle defects are sometimes detectable by in-line QC inspection tools. The paper focuses on calculations of random defect limited yield using a deterministic yield model. The model is used to prioritize defect problems and drive yield improvements. Actual examples are used to demonstrate the benefits and strengths of the deterministic model. We also discuss the methodology, assumptions, and limitations of this model.