SciDAC-5 : Development of High-Fidelity Simulation Capabilities for ELM-free Design Optimization
CETOP: A Center for Edge of Tokamak OPtimization (ELM-SciDAC)
Abstract:
For the advanced tokamak concept, the focus of the US fusion community is based on two major objectives: 1) sustained operation at the pressures and temperatures to achieve the conditions needed for the fusion reaction to occur and 2) closing the technological gaps needed to integrate a high performance plasma into a viable fusion power plant (FPP) system. One of the key challenges to these objectives is the occurrence of periodic instabilities at the plasma edge. Transient Edge-Localized Modes (ELMs) are one of the most concerning phenomena in burning plasmas as they could pose a risk to plasma facing components (PFCs). We propose to develop the simulation capability and to perform extended MHD and kinetic simulations of non-ELMing (and some ELM-ing) regime operating points in an FPP, projected from multi-machine databases. High performance computation (HPC) using the extended MHD codes, NIMROD and M3D-C1, as well as the XGC, and MARS-K/Q codes, together with existing multi-machine databases of ELM-free scenarios will be used for AI/ML based design extrapolation. This will enable optimization of future devices for ELM-free operation. Our physics goals are to obtain new high-fidelity-based ELM stability boundary maps by including non-ideal and multi-species physics, as well as stability analyses and nonlinear simulations of non-ELMing regimes due to plasma shaping (negative triangularity). In addition to AI/ML contributions, SciDAC Institute research necessary for the high fidelity MHD simulations also includes advance time-integration methods and transition to GPU-accelerated architectures.
Purpose:
The project overarching objective is to develop the simulation capability and to perform extended MHD and (drift-gyro) kinetic simulations of non-ELMing (and some ELMing) regime operating points to close gaps in understanding, prediction, and optimization of edge stability for an FPP.
To achieve our overarching goal, our ASCR objectives are therefore,
to develop advanced time discretizations (via time integration methods) and transition to GPU accelerated architectures for our MHDcodes (NIMROD and M3D-C1) to enable higher-fidelity multi-species simulations.
to apply ML techniques for extracting reduced-order models, data reduction, and feature extraction to the existing non-ELM database (based on interpolation of data), and to extrapolate to new parameter regimes (such as coil currents for negative-triangularity shaping) for ELM-free optimization.
Approach:
Due to multi-physics and multi-scale challenges, a multi-hierarchy approach is essential. In this SciDAC, for both understanding and reliable prediction of ELMing and non-ELMing discharges, we therefore utilize a hierarchy of models and methods, from comprehensive first-principle/high-fidelity to reduced linear models, as described below. Our physics goals are to obtain new high-fidelity-based ELM stability boundary maps by including non-ideal and multi-species physics, as well as stability analyses and nonlinear simulations of non-ELMing regimes due to plasma shaping (negative triangularity). We focus on physics challenges of 1-ELM onset prediction, 2- Reproduction of complete ELM cycles, 3- Prediction of ELM-induced heat loads, and 4- Optimization for ELM-free plasmas.
Linear MHD modeling using MARS-Q, M3D-C1 and NIMROD: To meet the physics challenge of ELM onset prediction, we plan to include non-ideal effects using the linear MARS-K/Q and extended MHD NIMROD/M3D-C1 codes to perform extensive benchmarking.
Extended MHD studies for plasma shaping and multi-species: We will employ nonlinear NIMROD and M3D-C1 simulations to investigate non-ELMing regimes, in particular via plasma shaping (negative triangularity, NT).We plan to also include multi-species collisionality (enabled by the GPU/time-integration developments).
Coupling of extended MHD, XGC, drift kinetic: To meet challenge of reproduction of complete ELM cycles, drift-, gyro- and fully kinetic codes will be utilized to provide the MHD calculations with perpendicular (neoclassical and turbulent) transport information, and closures for the parallel heat transport. We plan to perform hybrid high-fidelity simulations via loose coupling. Neoclassical and kinetic turbulent transport coefficients obtained during ELM-free (or inter-ELM) region using XGC simulations are used in the extended MHD to correctly model the onset and decay for tens of ms.
Finally, to meet challenge of prediction of ELM-induced heat loads, full nonlinear calculations also are needed for understanding and reliable prediction and optimization for non-ELMing regimes. Magnetic reconnection effects as well as multiple species in the SOL region will be studied via full nonlinear extended MHD and kinetic simulations.
SciDAC team members and institutions:
SciDAC-5 Principal Investigator : Fatima Ebrahimi (PPPL)
Princeton Plasma Physics Laboratory: PI: Fatima Ebrahimi. Co-PIs: Robert Hager, Andreas Kleiner, Seung-Hoe Ku, Alexei Pankin, Benjamin Sturdevant
Columbia University: Institutional lead PI: Chris Hansen. Co-PIs: Carlos Paz-Soldan
Fiat Lux: Institutional lead PI: Jake King
General Atomics: Institutional lead PI: Yueqiang Liu.
Utah State University: Institutional lead PI: Eric Held. Co-PIs: A. Spencer
Lawrence Livermore National Laboratory: Institutional lead PI: David Gardner. Co-PIs: Carol Woodward, Cody Balos
National Renewable Energy Laboratory: Institutional lead PI: Marc Day
Oak Ridge National Laboratory: Institutional lead PI: Prasanna Balaprakash
Brookhaven National Laboratory: Institutional lead PI: Thomas Flynn. Co-PIs: Shinjae Yoo
Argonne National Laboratory: Institutional lead PI: Sandeep Madireddy
University of Tennessee (UTK): Institutional lead PI: Natalie Beams