Yield improvement using statistical analysis of process dates

Rapid yield improvement is necessary in modern wafer fabrication facilities to ensure profitability of huge investments. Statistical analysis of data is a valuable method for improving yield. Usual methods like statistical process control (SPC), designs of experiments (DOE) and mapping analysis are necessary, but not sufficient because of process complexity. This paper presents a new statistical approach for solving yield issues when the root cause comes from a failure at a single process stage. The advantage of this approach is the only data required are the process dates of the lots at each process stage and the probe results. Three different methods have been tested to solve these issues. Only one of them is detailed in this paper. The efficiency of this method is demonstrated on two real yield issues where the defective stage is known.