Mission Areas

This project addresses two key areas: (1) Plasma physics, focusing on the transition to turbulent plasma, and (2) Magnetohydrodynamics (MHD) modeling through high-order positivity-preserving methods and sub-grid models for high Lundquist number flows.

Center Introduction

Turbulence is a critically important problem in plasma physics, both as a phenomenon of fundamental interest and for its impact on applications such as pulsed-power devices (where it strongly impacts rates of energy transport, magnetic reconnection, and plasma instabilities). Turbulence in plasmas is fundamentally different than hydrodynamic turbulence due to the presence of magnetic fields and their interactions with charged particles. Magnetic fields in plasmas allow physically separated regions to interact with each other which, in the context of turbulence, allows both spatially nonlocal energy transfer as well as an “inverse cascade” of energy from smaller to larger spatial scales. This makes plasma turbulence a complex, nonlinear phenomenon, and studying it can provide insights into nonlinear dynamics and chaos as well as potentially providing insight into other turbulent systems in fluid dynamics. Furthermore, the complexity of plasma turbulence fundamentally requires large-scale simulations as a tool for discovery and understanding, which in turn motivates the need for high-order numerical methods and training for the next generation of computational scientists to study these phenomena.

Techical Approach

High-order methods are critical for the direct numerical simulation of turbulence, doubly so for the case for plasmas where magnetic fields allow for transfer of energy and information between physically disparate scales. Scalable state-of-the-art magnetohydrodynamic (MHD) models used to study these effects are, in general, second-order methods. Members of this team have comprehensively demonstrated the advantages of high-order methods in the context of plasma dynamics, including problems with strong shocks. Some team members have developed scalable open source community codes and frameworks. In addition, team members have developed structure-preserving machine learning (ML) surrogates that identify and enforce key mathematical properties and can be hybridized with traditional computing methods, yielding stable integrated simulations. This FIC brings these threads together in the development of open source frameworks for plasma problems that undergo strong shocks and have a range of multi-scale and multiphysics turbulent behaviors, and which include synthetic diagnostics for direct comparison with plasma experiments. We will do so using DOE-supported open source frameworks that are developed in part by our team members.

Goals

High order numerical methods and scientific machine learning (SciML) are fundamental enabling technologies for problems equiring exascale computing. While their benefit has been demonstrated in some circumstances, we will develop these methods in the context of complex plasma phenomena. Specifically, we will simulate high energy density plasmas in fluid and fluid-adjacent regimes, with a focus on plasma turbulence and radiative magnetic reconnection. In the latter case we will make direct connections to the Magnetically Ablated Reconnection on Z (MARZ) platform on Sandia’s Z Machine. Our primary goal in choosing these applications is to ensure that the methods and tools that we develop are maximally useful for a broad range of high energy density problems of DOE interest.