A PCA-Based PCM Data Analyzing Method for Diagnosing Process Failures

In an IC process line, after a process is finished, the process control module or monitor (PCM) is tested and the data are examined so that the status of the process quality is known. In the case of a process failure, the root cause of the failure must be analyzed, and relevant actions must be taken to correct it. In this paper, we describe a novel way of diagnosing process failures. First, the tested PCM parameters, which are correlated to each other, are analyzed and transformed to a new set of independent parameters using principal component analysis (PCA). In the second step, the most important eigenvectors from PCA calculation are identified, and the causes of the process failures can therefore be extracted. Furthermore, using the PCA eigenvectors as a coordinate base, the state space of a process can be constructed. As a result, the process states from different lots of wafers can be compared; thus, it is possible to trace the IC processes, or even to predict a possible process failure, before it happens