Numerical Simulation of Neutron Transport in Spallation Source Target/Moderator Systems
B. Heuser,* T. Li
U.S. Department of Energy, ANL 970392401/LM ORNL 19X-SX702V

The design of the Spallation Neutron Source (SNS) is currently underway. Line-item construction funding for this 1-MW neutron source was passed by Congress in the fall of 1998. The primary purpose of the SNS is to provide intense neutron beams for a variety of scattering techniques. Our involvement in this project centers on the optimized coupling of long-wavelength, elastic scattering instruments (small-angle and reflectometry instruments) to the target/moderator system. We are simulating neutron moderation and transport within the complex spatial geometry of the target/moderator system using MCNP.

Development of Advanced Computational Methods for Smoke Dynamics
R. Uddin,* F. Wang
Computational Science and Engineering Program

A relatively large proportion of fire-related deaths in large buildings are caused by inhalation of combustion products away from the combustion sites. Hence, accurate prediction of smoke movement is important for improved building designs and for reliable evacuation procedures. In order to extend the building size over which smoke movement can be accurately predicted, we are developing advanced computational methods for implementation on large multiprocessor computers for a special set of equations specifically developed to study smoke dynamics.

Energy-Conservative Solver for Multidimensional Fokker-Planck Problems
G. H. Miley,* L. Chacon, D. Barnes
Daimler-Benz Aerospace AG; Los Alamos National Laboratory

The Fokker-Planck equation for plasma modeling is a highly nonlinear PDE, being its coefficients integral functionals of its solution. To ensure numerical energy conservation-granted theoretically-the nonlinearity has to be resolved using Newton's method, resulting in a dense, nonsymmetric matrix equation. Although intractable using standard solvers, this system is dealt with efficiently with Newton-Krylov iterative techniques (GMRES). A matrix-free implementation avoids forming the Jacobian matrix. Convergence is accelerated with a multigrid preconditioner, resulting in a constant number of iterations with grid refinement. Typical CPU times of the energy-conservative solver are comparable to those of nonconservative approaches.

A Nodal Computational Scheme for the 2-D Burger's Equation
R. Uddin,* B. L. Westcott
DOE Center for Simulation of Advanced Rockets

To demonstrate its feasibility in solving the Navier-Stokes equations, the modified nodal integral method is being applied to solve the 2-D Burger's equation at high Reynolds numbers. Performance of the nodal method is being compared with other competing methods.