Yield prediction in semiconductor manufacturing using an AI-based cascading classification system
Peter Stich, M. Wahl, Peter Czerner, Christian Weber, Madjid Fathi
Abstract
The production of a wafer is a highly complex process with hundreds of individual, sequential process steps and thousands of parameters, where each of them can have an influence on the final product. One approach for yield prediction before or during production is based on the divide-and-control principle. The individual process steps are considered independently and based on the analysis of these steps a prediction for the overall yield is calculated by a master system. In this article, a novel concept for predicting the yield is presented, which combines a cascading prediction algorithm, based on the sequential process steps, with an artificial intelligence (AI) focused master system. The system also proposes a recommended selection for optimal yield if the next process step can be executed on different machines/ chambers.