A MONITOR WAFER BASED CONTROLLER FOR SEMICONDUCTOR PROCESSES

A monitor wafer based controller is described. The controller can be a pplied to equipment with or without in-situ sensors. The controller in corporates a novel multivariable adaptation methodology for the feedba ck controller that employs a layered process/equipment model. The laye red model consists of an intrinsic component that corresponds to the i nitial settings to outputs model and an extrinsic component that trans forms the inputs and the outputs of the intrinsic model. The adaptatio n strategy tunes the extrinsic model only and thus the adaptation stra tegy is independent of the intrinsic model form. The controller determ ines whether the process and equipment have changed state by using mod el based SQC to compare product parameter measurements with the compos ite model predictions. If a change in state is deduced, a model tuner is activated which adapts the extrinsic model to reflect the new state . To adapt the model, a local experiment design technique is applied t hat perturbs the equipment settings. Finally, a stepwise optimization technique that permits the specification and utilization of user prefe rence toward changing some process inputs over others is used for dete rmining the new process receipe. We report the controller's applicatio n to the plasma enhanced chemical vapor deposition of silicon nitride (PECVD Nitride) process run on Applied Materials Precision Reactor (AM T 5000). The controller has been tested in two ways. First, single and multiple faults were introduced in the process equipment. Second, the controller performance was observed during an extended period of rout ine use. These evaluations indicate that the controller is able to det ect process state change and to adjust the process recipe to keep the process on target.