We are exploring new methods of interpreting turbulent transport of molecular species or particles that describe the field as resulting from a distribution of sources and sinks. These studies are carried out with a supercomputer simulation of turbulent flow in a channel and are made possible by the development of particle-tracking routines. Of particular interest is the effect of molecular transport on turbulent mixing and the effect of gravity on particle transport in dispersed flows. Direct numerical simulations of turbulent heat transfer in a channel have been carried out from Prandt number varying between 0.05 and 10.

The modeling, simulation, design, and optimization of complex chemical processes, steady- or unsteady-state, can be done much more effectively using advanced computer architectures, especially those using some form of parallel processing. However, since current methods for solving such problems were developed for use on conventional serial computers, they usually cannot take much advantage of the power of parallel computing. Thus, solution strategies must be completely rethought. The goal of this project is to develop and apply new strategies for exploiting the power of parallel computing in process engineer- ing. Some of the techniques developed are already in use in supercomputer versions of commercial simulation programs.

The goal of the proposed research is to develop robust spectral boundary integral algorithms for three-dimensional transport problems in realistic geometries. The spectral boundary elements combine the high-order convergence associated with spectral methods and the versatility of boundary element methods. To realize the potential of this method, each of the four major components must be optimized: (1) formulation of the integral equation, (2) discretization, (3) numerical integration, and (4) solution of the algebraic linear systems.