Stephen Senior Applied Mathematics Researcher Thomas

Dr. Stephen Senior Applied Mathematics Researcher Thomas

National Renewable Energy Lab, Golden, CO 80401, USA

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
22 October 2020 in Numerical Linear Algebra with Applications
Verified
Journal Article
Closed access
PDF file
KŚKatarzyna Świrydowicz
Katarzyna Świrydowicz
JLJulien Langou
Julien Langou
SAShreyas Ananthan
Shreyas Ananthan
UYUlrike Yang
Ulrike Yang
Stephen Senior Applied Mathematics Researcher ThomasStephen Senior Applied Mathematics Researcher Thomas
Stephen Senior Applied Mathematics Researcher Thomas

28 December 2023 in International Journal for Numerical Methods in Fluids
Verified
Journal Article
Closed access
Stephen Senior Applied Mathematics Researcher ThomasStephen Senior Applied Mathematics Researcher Thomas
Stephen Senior Applied Mathematics Researcher Thomas
ACArielle Carr
Arielle Carr
PMPaul Mullowney
Paul Mullowney
KŚKatarzyna Świrydowicz
Katarzyna Świrydowicz
MDMarcus Day
Marcus Day

4 June 2026 in Numerical Linear Algebra with Applications
Verified
Journal Article
Open access
PDF file
Stephen Senior Applied Mathematics Researcher ThomasStephen Senior Applied Mathematics Researcher Thomas
Stephen Senior Applied Mathematics Researcher Thomas
Krylov subspace methods solve large sparse linear systems by building a sequence of polynomial approximations to from successive matrix‐vector products. In finite precision, the number of numerically independent directions that can be extracted from this sequence is bounded by the intrinsic information dimension , defined as the index at which the Krylov basis matrix becomes numerically rank‐deficient: , where is the accumulated finite‐precision noise floor. For sequences of elliptic PDE discretizations with spectral density, the normalized Gram matrix of any Chebyshev basis block decays off‐diagonal at a rate , established via two integrations by parts applied to the Chebyshev moment integral; this gives (monomial basis) or (Chebyshev basis), where is the spectral condition number. Two consequences follow directly from this bound. First, any restarted Krylov method requires restart length – to reproduce unrestarted convergence; the common choices or produce 8– overhead or practical stagnation for . Second, ‐step GCR with Chebyshev basis and Forward Gauss–Seidel Gram solve requires only blocks of inter‐block orthogonalization history, achieving storage and global synchronizations per iteration independent of iteration count; the 20– gap between this truncation depth and the restart minimum is explained by the same decay rate.

5 October 2026 in Applied Numerical Mathematics
Verified
Journal Article
Closed access
Stephen Senior Applied Mathematics Researcher ThomasStephen Senior Applied Mathematics Researcher Thomas
Stephen Senior Applied Mathematics Researcher Thomas

25 September 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
Stephen Senior Applied Mathematics Researcher ThomasStephen Senior Applied Mathematics Researcher Thomas
Stephen Senior Applied Mathematics Researcher Thomas
This paper produces an interval containing the chip temperature of a model, with no estimated constant, from any admissible temperature field and an equilibrated flux, exact in exact arithmetic and evaluated in binary64, on a copper heat spreader with $23{,}985$ nodes; the solver stops when its remaining contribution to the interval is a stated fraction of the width. The primal system and the Schur complement of the mixed system are solved by CD-CG, an $s$-step conjugate gradient in Christoffel--Darboux Gram coordinates, every solver accepted on the recomputed residual; at $s=8$ it reaches the budgets in $1.8$ to $4.7\times$ fewer global reductions than conjugate gradient under multigrid, Schwarz and Jacobi preconditioning, at $1.2\times$ the products, and the Christoffel--Darboux panel GMRES with ORAS in $1.3\times$ fewer than low-synchronization GMRES. The same interval certifies a field from a neural network on a thirty-two-mode Lanczos basis, trained by a polynomial filter on the spectral measure of the momentum's Gram matrix, reaching the comparator's final loss in $2.3\times$ fewer epochs: its interval is $0.02551$ K wide against $0.02546$ K, and the complete certificate takes $0.21$ s against $0.38$ s. Every number is reproduced by the accompanying scripts.

23 September 2026 in Journal of Computational and Applied Mathematics
Verified
Journal Article
Open access
Stephen Senior Applied Mathematics Researcher ThomasStephen Senior Applied Mathematics Researcher Thomas
Stephen Senior Applied Mathematics Researcher Thomas