Wafer Classification Using Support Vector Machines

Abstract

Increasing yield is a primary concern to integrated circuit manufacturing companies as it dictates the readiness of a new process for high volume manufacturing. In order to expedite the process of discovering yield issues, companies have started looking for ways to perform early prediction for such issues. This paper suggests the use of the support vector machines (SVMs) for early wafer classification. The choice of SVM is motivated by the model's ability to effectively classify multivariate, multimodal, and inseparable data points. This model uses multidimensional hyperplanes to separate and classify wafers into low-yield and high-yield classes. This paper includes a proposal for how the classification model can be applied for yield classification and how it can be adaptively updated in a manufacturing environment. We show how the values for the SVM parameters can be selected for best yield classification. Furthermore, performance evaluation is conducted on real manufacturing data, comparing the proposed SVM classifier to state of the art. Results show that in all cases, SVM consistently outperforms other methods with and without adaptive model updates. The experiments also show that all classifiers' performances depend on yield thresholds. It is also shown that the classification model can be built and executed using a reduced set without compromising its accuracy.