Production data based optimal etch time control design for areactive ion etching process

摘要

This paper addresses the issue of end point detection and etchntime control for a reactive ion etch process. Our approach involves thenuse of neural networks to model the functional relationship between annend point detection signal, as well as various in situ measurements, andnthe resulting film thickness remaining. An optimization algorithm isnthen employed to determine the optimal etch time based on the neuralnnetwork model in order to achieve the desired level of film thicknessnremaining. This circumvents the need for monitoring and operating onnnoisy end point detection signals typically associated with conventionalndetection schemes. Simulation studies based on production data arenpresented to further demonstrate the associated design procedures andnthe feasibility of the algorithm. Tested on data from 89 randomlynselected wafers, our controller yields a film thickness distributionnwith the standard deviation of 6.42 Å, a 50% improvement over thenscheme currently implemented in production