rSQP Optimization of Large-Scale Reacting Flow Applications with MPSalsa

An rSQP optimization algorithm has been linked to the MPSalsa parallel reacting ows code. This goal is to develop SAND (Simultaneous Analysis aNd Design) methods for use with large-scale PDE simulations. MPSalsa is a unstructured grid nite element code that uses a fully coupled Newton method to solve the PDEs governing uid ow, heat transfer, and non-dilute mass transfer. In this paper, we present results for optimization of a Chemical Vapor Deposition reactor for growing thin lms of Gallium Nitride. In particular, we address issues of inexactness in the Jacobian matrix and of solution multiplicity.