On-line process optimization and control using the sequential design of experiments

A modular framework is being used for the implementation of process control in VLSI fabrication. The function of process control has been divided into three core modules: the flexible recipe generator, the run by run controller and the real-time controller. These three modules act to divide up the control function on the basis of the time scale of the response and the range of operating space encompassed. The flexible recipe generator has the task of taking the best first guess about where to begin operations, resulting in a recipe for use in the first run of a new product. The run by run controller starts from the recipe provided by the flexible recipe generator and then updates the recipe between product runs based on the post-process measurements. The real-time controller accepts in-situ measurements as inputs and modifies the recipe during a process step in order to bring the result as close as possible to target. Here, the authors focus on the run by run controller. The run by run controller covers a smaller range of space than the other modules and is invoked after each run of product. As part of a system for process control, the run by run controller implements a form of adaptive control based on the sequential design of experiments. The run by run controller was tested in application to an equipment simulator of polysilicon low-pressure chemical vapor deposition (LPCVD). Growth rate uniformity from wafer to wafer was achieved. Optimization and control speed were further enhanced by an improved modeling approach