Skip to main content

Stephen Thomas

Researcher IV-Applied Mathematics

| 303-275-3949

Throughout his research and consulting career, Dr. Thomas has focused on the intersection of high-performance computing and iterative solvers for large sparse linear systems with applications in climate, geoscience, and renewable energy.

Prior to joining NREL, Dr. Thomas relocated to Boulder, Colorado, from Canada to join the National Center for Atmospheric Research (NCAR), working on high performance and scalable climate models. Dr. Thomas is a licensed professional engineer in Canada. His projects at NREL are focused on solvers for wind turbine turbulence simulations, as part of the U.S. Department of Energy (DOE) Exascale Computing Initiative.

Novel contributions for NREL and DOE include low communication iterative linear solvers and preconditioners suitable for GPU acceleration. These novel approaches have also resulted in faster and more accurate restarted Arnoldi eigenvalue algorithms in collaboration with the University of Colorado Denver mathematics department.

Research Interests

Iterative algorithms and preconditioners for the solution of large sparse linear systems of equations arising in physics-based simulations

Krylov methods, ILU preconditioners, algebraic multigrid and smoothers

Stable numerical linear algebra algorithms for massively parallel computation

Eigenvalue and singular value decompositions for large sparse problems


Ph.D., Computer Science, University of Montreal  

M.S., Electrical and Computer Engineering, McGill University  

B.S., Applied Mathematics, University of Waterloo 

Featured Work

S. J. Thomas, S. Ananthan, S. Yellapantula, J. J. Hu, M. Lawson, M. A. Sprague (2019). A comparison of classical and aggregation-based algebraic multigrid preconditioners for high-fidelity simulation of wind-turbine incompressible flows. SIAM Journal of Scientific Computing. Special Issue. Copper Mountain Conference on Iterative Methods, in press.

K. Swirydowicz, J. Langou, S. Ananthan, U. Yang, S. Thomas (2019). Low synchronization Gram-Schmidt and GMRES algorithms. Numerical Linear Algebra with Applications; accepted. 

Invited Talks

Low-Synch Gram-Schmidt Projection Schemes for GMRES-AMG and for Moving Mesh Solvers. Exascale CEED third annual meeting. Virginia Tech, August 2019.

Low-Synch Gram-Schmidt and GMRES Algorithms. PIMS Applied Mathematics Seminar. University of British Columbia, Canada, August 2018.