[LIVE] AI in an Academic Cleanroom Applied to Nanoelectromechanical Systems: Challenges and Rewards
An academic cleanroom is often confronted with the use of last-generation equipment used by untrained student users, used by a wide variety of users interested in very different devices, and the institutional duty to train future users of semiconductor technology. At the same time, the federal dollars appropriated for R&D are limited, and getting working devices completed on time would lead to faster convergence to discoveries and commercialization. AI-based identification of time-varying equipment performance, and the effects of the previous recipe used on the outcome of the current desired methods, are just some of the ways AI can be used to reduce the process variability. At Cornell, we have applied AI approaches to optimize lithography and etching processes involved in the development of an RF wake-up NEMS switch that needs a well-controlled gap between a moving shuttle and a contact. We report on a decision tree based AI model for predicting lithography outcomes. Current work is taking this work and extending to the application to plasma etching and the combined prediction of lithography and etching, using CD-SEMS for learning data.