9503988 Prasad The object of the proposed research is to develop a parallelized multizone adaptive scheme for accurate and efficient simulations of materials processes of industrial importance. The final software will be able to simulate three-dimensional transient processes involving diffusion and convection of heat, mass, and species, and radiation heat transfer together with melting/solidification, flows induced by buoyancy and capillary forces, and effects of electrical and magnetic fields. The use of a generalized governing equation will allow consideration of many different materials with various phases in a single computational domain. The numerical scheme will employ (a) the multizone adaptive grid generation (MAGG) technique for the discretization of physical domains of arbitrary shape, and (b) the curvilinear finite-volume(CFV) approach for the discretization of the govening partial differential equations and development of the finite difference equations. An efficient parallel algorithm suitable for multiphase systems will be developed based on the massively-parallel distributed-memory MIMD techniques and will take advantage of the physics of the problem to develop a domain-decomposition strategy. Load balance, communication, and migration of elements will be given special attention in the development. The immediate implementation of this model will be made to the CZ and CCZ growth of silicon single crystals (a project sponsored by NSF and Ferrofluidics Corporation) and one-step in-situ synthesis and high pressure MLEK growth of indium phosphide crystals (an AFOSR project). The proposed parallel computer model will be able to simulate many kinds of complex materials processes involving free and moving boundaries that are not possible by the present techniques.
