Optimization Solvers in Run-to-Run Control

Semiconductor manufacturing processes with multiple settings or multiple targets, such as furnace applications or processes controlling uniformity, may not have a single algebraic solution, in which case optimization methods are required. While there are many optimization algorithms available, most are not guaranteed to find the global optimum, take longer than an analytic solution, and produce a less precise result. Despite this, in practice people use optimizers even when they are not needed. This paper explains when optimizers are needed and examines how the structure of the process model can be used to speed up the optimization process and ensure that the global optimum is found.