Numerical Computation of Wave-Plasma Interactions in Multi-dimensional Systems
Principal Investigator – D. B. Batchelor
(
L. A. Berry, J. A. Carlsson, M. D. Carter, E. F. Jaeger
Oak Ridge National Laboratory – Fusion Energy Division
E. D’Azevedo, L. Gray, T. Kaplan
Oak Ridge National Laboratory – Computer Science and Mathematics Division (Science Application Pilot)
C. K. Phillips
Princeton Plasma Physics Laboratory
P. T. Bonoli
Massachusetts Institute of Technology
D. N. Smithe
Mission Research Corp.
R. W. Harvey
CompX
D. A. D’Ippolito, J. R. Myra
Lodestar Research Corporation
Executive Summary for SciDAC Kickoff Meeting
September 24-25, 2001
Background and project vision
Electromagnetic waves play a fundamental role in the dynamics of plasmas either as externally-driven waves or as self-generated instabilities. Plasma waves are an essential process in natural systems ranging from the solar corona, to planetary magnetospheres, to the earth's ionosphere, and in laboratory and commercial devices. However, it is in the complex phenomena that arise in fusion-relevant laboratory plasmas that understanding of wave dynamics poses perhaps the most significant scientific challenge and opportunity of any area of wave physics. The plasmas in fusion devices are bounded and inhomogeneous, having essential variations in at least two spatial dimensions and in many cases three dimensions. Due to the confining magnetic field, these plasmas are also anisotropic. Within these plasmas there can be many distinct wave modes propagating in a single plasma region at a given frequency. The different wave modes can coexist at widely separated wavelength scales, or can interact when spatial variation of the plasma bring the wavelengths close together – a process called mode conversion. There are numerous mechanisms by which the waves can be absorbed – collisional damping, collisionless Landau and cyclotron damping, and damping by collisionless stochastic particle interactions. The waves can produce nonlinear modifications of the plasma medium, thereby altering the wave propagation characteristics and changing the macroscopic properties of the plasma, which in turn affect other processes occurring in the plasma such as stability and transport. These modifications can be at the microscopic level of energy distribution functions. It is important to the future advancement of wave-plasma interaction physics that all of these wave and plasma effects be self-consistently analyzed in the appropriate, multi-dimensional geometry.
The overarching goal of the project is to obtain quantitatively accurate predictive understanding of electromagnetic wave processes that support important heating, current drive, and stability and transport applications in fusion-relevant plasmas. Several specific areas of wave/plasma physics can be identified which are critical to the understanding of near term experiments and to realization of fusion program needs for plasma control in which progress is now limited by the available computing power. Massive parallelization will provide access to new wave physics of the following kinds:
• Higher dimensionality – Computation of important wave features in 2D and 3D that can now only be calculated in 1D or marginally in 2D
• Higher resolution – Power to resolve short wavelength structures arising from mode conversion, high plasma dielectric constant or multi-wave interference effects, but which are presently infeasible to calculate
• Inclusion of phenomena in the plasma wave current which are presently omitted or subject to restrictive approximations– Ability to represent arbitrary, non-Maxwellian distribution functions, retention of high cyclotron harmonics, inclusion of non-local effects in conductivity operator, inclusion of non-linear effects
• Self-consistency of the distribution function as used in the wave solution –Self-consistent calculation of production of tails, and effects of non-thermal populations on wave propagation and absorption
In this project we will use massively parallel processing to address each of the above physics areas.
Elements of the Research Program
The project establishes a multi-institutional topical center, in a partnership between plasma physicists and computational scientists, to create simulation codes of wave-plasma interaction which take full advantage of terascale computers. The research program will consist of 3 elements:
1) Physics generalization, computational scale-up, and interconnection of the elements of a complete wave/plasma modeling capability
• Development of generalized plasma conductivity operator modules and macroscopic source modules with increased physics content and that are structured using modern programming techniques for flexibility of application and ease of extension in capability.
• Creation of a core suite of wave solution codes which are complementary and cover the various needs in physics capability. The core suite will require development of new wave solution codes in higher dimensions and restructuring and interfacing of a small number of existing codes to take advantage of terascale computers and to incorporate new physics effects.
• Computationally efficient interface of the conductivity operator modules and wave solvers with Fokker-Planck solutions, macroscopic source calculations and antenna solvers.
2) Demonstration of the enhanced simulation capabilities by exploring four major unsolved problems in wave-plasma interactions having immediate ramifications in ongoing fusion experiments:
• To what extent do realistic 2D and 3D equilibrium variations modify the local deposition of wave energy and momentum in the plasma?
• To what extent does the presence of non-Maxwellian particle velocity-space distributions modify local deposition of wave energy and momentum in plasma?
• What is the mechanism by which lower hybrid waves, launched with a phase velocity several times the electron thermal speed, are able to couple strongly to electrons and drive substantial currents?
• What is the effect of global plasma modes on the wave fields produced by launching structures?
3) Exploratory research into completely reformulating wave-plasma problems and solution methods to permit further extension of physics scope, such as to non time-harmonic or highly non-linear phenomena, and to examine possibilities for increases in computational efficiencies such as through alternate field representations like wavelets.
Major Milestones
• Year 1
Accelerate existing codes in preparation for increased computational load by restructuring for MPP and incorporation of parallel solvers. Carry out initial scoping studies for physics thrust areas.
Develop 3D code with existing physics, test with numerical equilibria
Determine domain of validity for finite Larmor radius expansion models by comparison with all orders codes
Develop generalized conductivity operators (non-Maxwellian and non-local separately at first), validate and examine physics implications in 1D
• Year 2
Develop generalized dielectric tensor operator in 2D. Test 1D and 2D versions with Fokker Planck derived distributions (open loop).
Couple full wave fields into Fokker Planck solvers (open loop)
Carry out full-wave solution for LH wave in 2-D and use model to study quantum chaos effects on filling spectral gap problem in lower hybrid current drive.
Develop advanced matrix solvers – iteration, physics-based preconditioning, out-of-core, fast moment methods
• Year 3
Complete code development – closed loop self-consistency of distribution function with dielectric operator, incorporation of generalized modules into core suite of wave codes
Complete studies of physics thrust areas