Predicting Ion Beam Tuning Success in Semiconductor Manufacturing

Advanced Process Control (APC) systems monitor semiconductor manufacturing processes continuously via equipment internal sensors. The logged data enables data-driven predictive maintenance approaches. Integration of APC-derived constraints into scheduling has the potential to improve the overall equipment effectiveness (OEE). Therefore, we introduce a real-world semiconductor manufacturing case study: Ion Implantation equipment accelerates dopants in an electric field onto wafers, to change electrical properties of defined layers. Every recipe change necessitates ion beam tuning, to meet specifications under varying equipment conditions. In order to avoid expensive timeouts of unsuccessful tuning, a prediction model estimates the tuning success and scheduling is (to be) optimized accordingly. Our preliminary results show that more than half of unsuccessful ion beam tuning can be correctly predicted and thus avoided.