In this paper, a new methodology for effective process excursion monitoring using defect review/classification information is proposed. We introduce a new defect classification scheme, in which relevant defect types that are likely to be caused by the same mechanism or source are grouped into a "defect family". It is demonstrated that trending by the defect family drastically improves the detection efficiency of killer defect excursion by reducing or eliminating noise resulting from irrelevant benign defects. We compare the risks of missing critical excursions for monitoring by total defect count, killer defect count, and killer defect family, and illustrate the effectiveness of our methodology using data from actual fabline.

In-line monitoring of defects in VLSI manufacturing has become an indispensable tool in SPC (Statistical Process Control) and Yield Management. Wafer inspection for defects has two stages: optical wafer scanning to detect the presence of defects; and review/classification performed at the coordinates of the scanned defect to determine the defect type (for instance, shorts or opens). Typically, fabs use standard control charts, based on total defect count monitoring. However, many fabs have found this aggregate tracking of defects to be inadequate for efficient excursion or drift detection. In this paper, we demonstrate how defect type information can be utilized to optimize the excursion detection procedure. In addition, we also demonstrate how the defect type information can be useful for source identification

In this paper, new methodologies for effective process excursion monitoring and defect source isolation are proposed. We introduce a new defect classification scheme, in which relevant defect types that are likely to be caused by the same mechanism or source are grouped into a "defect family." We demonstrate that trending by the defect family drastically improves the excursion detection efficiency without suffering noise from irrelevant benign defects. Furthermore we have developed a methodology for identifying the source of the excursion using defect type Pareto. This is based on the fact that the signature of defect type Pareto leads to the defect source information and thus possibly indicates the origin of the problem. Thus both process control and excursion source identification can be achieved simultaneously by effective defect classification.