Shape Optimization of a Showerhead System for the Control of Growth Uniformity in a MOCVD Reactor Using CFD-based Evolutionary Algorithms

Abstract

A steady state, laminar flow coupled with heat transfer, gas-phase and surface chemistry, is numerically solved for the optimal design of a showerhead gas delivery system in an axis-symmetrical MOCVD reactor. The design method involves an evolutionary algorithm based on CFD simulations. A finite-volume CFD code for aluminum growth provides the numerical predictions of the growth rate and its spatial variation over the substrate. A multilevel evolutionary algorithm is used to continuously adjust the shape of the shower plate so as to minimize the spatial variation of the growth rate. A 5-variable parameterization of the shower plate is investigated and a near-optimal solution is proposed and compared to the original configuration of the shower plate.