0318 Leveraging Reduced Descriptions to Accelerate Kinetic Simulations

  • Lee Ricketson, Lawrence Livermore National Laboratory
  • William Taitano

Kinetic partial differential equations are regarded as the first principles model for various physical systems, including plasmas, rarefied gases, and radiation. The numerical solution of these equations is complicated considerably by their high dimensionality and multiple scales. Frustratingly, physically observable quantities of interest are typically lower-dimensional and evolve on coarser scales, but their accurate evaluation often requires a full system solution. This tension motivates many researchers to explore the use of reduced descriptions for the observables.
Classically, these usually take the form of asymptotic fluid limits of the kinetic system, while more recently, modern data-driven techniques have been used to find such models. This mini-symposium will bring together researchers from various application areas to discuss how these reduced descriptions are used to accelerate the numerical solution of the full kinetic system. These reduced descriptions may be used as preconditioners for implicit solvers; in domain decomposition when an asymptotic limit is valid in a portion of the domain; or to accelerate the arrival at a quasi-steady state, among other approaches. The discussion of these fundamental considerations across application areas is expected to promote the cross-pollination of ideas, giving researchers fresh perspectives to bring back to their respective fields.

At this stage, we have six speakers that have agreed tentatively to present at the mini-symposium. Below are the speakers, their standings in the community, and recent work relevant to the mini-symposium:

Dr. Luis Chacon is a senior research scientist at the Los Alamos National Laboratory. He has made many high-impact and seminal contributions to multiscale methods applied to kinetic plasma and radiation transport problems. He is also a current fellow of the American Physical Society and is regarded as one of the leading authorities in the field. He has recently developed a reduced-order Gaussian-mixture-based, unsupervised machine learning algorithm for accelerating the evaluation of the Coulomb interaction model, which will fit nicely with the symposium’s theme.

Mr. Iman Datta is a Ph.D. candidate at the University of Washington. His recent work in developing a continuum, physics-based hybrid fluid-kinetic boundary interaction algorithm fits nicely into the symposium theme.

Dr. Jiequn Han is a junior researcher at the Princeton University, working on applying modern data-driven (machine learning) approaches to tackle the curse of dimensionality in numerically solving high dimensional partial differential equations. He is highly active in the topic and has made seminal contributions in recent years to solve the Boltzmann equation.

Dr. Benjamin Sturdevant is a research scientist at the Princeton Plasma Physics Laboratory. He is active in studying multi-scale and conservative simulations methods for kinetic plasma simulations. His work on equation-free projective integration techniques to accelerate long time-scale simulation of such systems is particularly relevant to this minisymposium.

Dr. Paul Tranquilli is a research scientist at Lawrence Livermore National Laboratory with a background in implicit and multiscale time integration techniques. His recent and ongoing work on multi-model parallel-in-time simulation holds great promise for accelerating large-scale kinetic simulations.

Dr. Tomohiko Watanabe is a senior scientist at the Nagoya University with a track record of developing novel and practical methods for solving the gyrokinetic-Vlasov equation for studying fusion plasmas Tokamaks. His recent publication in the use of a low-order moment system to accelerate the simulation of the full Vlasov system has made a significant contribution in advancing the state-of-art in simulating fusion plasmas.

© WCCM-APCOM 2022. All Rights Reserved.