1. ADVANCED COMPUTATION IN PLASMA PHYSICS Forty-Third American Physical Society Division of Plasma Physics Annual Meeting Long Beach, California W. M. TANG Princeton University, Plasma Physics Laboratory, 2 November 2001

2. PERSPECTIVE GOAL: Reliable predictions of complex properties of high temperature plasmas –Acquire scientific understanding needed for predictive models superior to empirical scaling Plasma Science is both utilizing and contributing to the exciting advances in Information Technology and Scientific Computing. Advanced computation in tandem with theory and experiment is powerful new tool for scientific understanding and innovation in research Focus of present talk: Magnetically-Confined Plasmas (Fusion Energy Sciences )

3. Fusion Plasma Science is in Pasteur’s Quadrant Prof. Donald Stokes, Dean, Princeton Woodrow Wilson School Considerations of Use? No – Yes Quest for Basic Understanding? No – Yes Bohr Edison Pasteur Tight coupling of understanding and innovation. Strong commitment to both!

4. Plasma Science Challenges NRC Plasma Science Committee Macroscopic Stability –What limits the pressure in plasmas? Astrophysical accretion disks Wave-particle Interactions –How do particles and plasma waves interact? Solar coronal heating Microturbulence & Transport –What causes plasma transport? Accelerator collective dynamics Plasma-material Interactions –How can high-temperature plasma and material surfaces co-exist? Materials processing

5. Challenge to Theory & Simulations Huge range of spatial and temporal scales Overlap in scales often means strong (simplified) ordering not possible 10 -6 10 -4 10 -2 10 0 10 2 Spatial Scales (m) electron gyroradius debye length ion gyroradius tearing length skin depth system size atomic mfp electron-ion mfp 10 -10 10 -5 10 0 10 5 Temporal Scales (s) electron gyroperiodelectron collision ion gyroperiod Ion collision inverse electron plasma frequencyconfinement Inverse ion plasma frequency current diffusion pulse length

6. Scientific Computing Critical to Discovery in Many Scientific Disciplines Subsurface Transport Global Systems DOE Science Programs Need Dramatic Advances in Simulation Capabilities To Meet Their Mission Goals Health Effects, Bioremediation Fusion Energy CombustionMaterials

7. Plasma Physics in DOE Advanced Scientific Computing Programs New DOE Office of Science Program: “Scientific Discovery through Advanced Computing” ---- FES is an active member of this broader scientific portfolio with access to new resources Plasma Science Advanced Computing Institute (PSACI) –Lead role for coordinating Plasma Science component of DOE’s new SciDAC Program –Peer-reviewed projects include FES Collaboratory, Magnetic Reconnection, Wave Heating, Atomic Physics, Turbulent Transport, and MHD Simulations –Program Advisory Committee (with distinguished members from outside & within FES) provides excellent advice/guidance

8. COMPUTATIONAL CHALLENGES IN FUSION ENERGY SCIENCES IMPORTANT FOR MOST AREAS (Cross-Disciplinary Opportunities) ______________________________________________________________ Enhance physics models & develop more efficient algorithms to better address scientific issues  Multi-scale physics  e.g. Kinetic (electromagnetic) dynamics  Improved algorithms  e.g. Adaptive mesh refinement for higher dimensionality phase-space  Scalability of codes  e.g. Efficient implementation of codes on most powerful MPP supercomputers Improve analysis/interpretation of greatly increased volume of simulation data  New diagnostic & visualization tools, improved data management/analysis

9. Advanced Scientific Codes --- a measure of the state of understanding of natural and engineered systems Theory (Mathematical Model) Applied Mathematics (Basic Algorithms) Computational Science (Scientific Codes) Computer Science (System Software) Problem with Mathematical Model? Problem with Computational Method? Computational Predictions Agree* w/ Experiments? No Yes Speed/Efficiency? Inadequate Adequate Use the New Tool for Scientific Discovery (Repeat cycle as new phenomena encountered ) *Comparisons: empirical trends; sensitivity studies; integrated measurements (spectra, correlation functions, heating rates …)

10. Single Fluid Resistive MHD Two Fluid MHD (electrons and ions) Two Fluid MHD plus energetic gyro- particles Gyro- particle ions and fluid electrons Full orbit particle ions and fluid electrons Less complex model, valid for high- collisionality, strong fields, long times More computationally demanding. Required to describe many important but subtle phenomena. External kink modes Neoclassical tearing mode (including rotation) Collisionless reconnection MHD modes destabilized by wave- particle resonance with energetic species Kinetic stabilization of internal MHD modes by ions Tilting and interchange modes in FRC MACROSCOPIC (MHD) SIMULATIONS: DIFFERENT LEVELS OF ANALYSIS CAPABILITY PPPL, SAIC, MIT, LANL, NYU, GA, U.Wisc., U. Texas, U. Colorado

11. Neoclassical Tearing Mode (NTM) Analysis Capability Self-consistent closure for Neo-classical Fluid Eq.’s being developed & applied [e.g., NIMROD] Results to be cross-benchmarked & validated against experimental results Enable assessment of NTM impact on beta limit for long-pulse, high-performance tokamaks

12. UNSTABLE INTERNAL KINK (LEFT) EVOLVES (RIGHT) MHD SIMULATION OF INTERNAL RECONNECTION EVENT

13. Hot Inner Region Interchanges with Colder Outer Region via Magnetic Reconnection

15. MPP Supercomputers Provide Access to New Plasma Wave Physics ORNL, PPPL, MIT Mission Research, Lodestar, CompX Improved Physics and All Orders Spectral Algorithm (AORSA-2D) Field solutions for conversion of fast ion cyclotron waves to ion Bernstein waves in 2D for a tokamak – collaboration with Computer Science and Math division at ORNL Contours of wave electric field strength for mode conversion using DIII-D tokamak parameters Patterns shown here are not revealed in a 1D treatment Extension to shorter wavelength and to 3D will be possible with new generation computers

17. UNDERSTANDING TURBULENT PLASMA TRANSPORT  An important problem:  -- Size of plasma ignition experiment determined by fusion self-heating versus turbulent transport losses  -- Dynamics also of interest to other fields (e.g., astrophysical accretion disks)  A scientific Grand Challenge problem   A true terascale computational problem for MPP’s

18. PLASMA MICROTURBULENCE SIMULATION CODES HAVE MADE EXCELLENT PROGRESS LLNL, PPPL, GA, U. Maryland, UCLA, U. Colorado Builds on National Turbulent Transport Project -- multi- institutional “Grand Challenge” Realistic Geometry –Full Torus (3D) –Flux Tube Codes Efficient Algorithms –Gyrokinetic --- PIC –Gyrokinetic --- Vlasov Continuum Demonstrated scaling beyond 100’s of processors

21. Full Torus Simulations of Turbulent Transport Scaling Large-scale full torus gyrokinetic particle simulations for device-size scans Global field-aligned mesh saves factor ~100 in computation Efficient utilization of new 5 TF IBM SP @ NERSC (just available 8/01) -- fastest non-classified supercomputer in world Most recent simulations used 1 billion particles (GC), 125 M spatial grid points, and 7000 time steps --- leading to important (previously inaccessible) new results

22. Full Torus Simulations of Turbulent Transport Scaling Transport driven by microscopic scale fluctuations (ITG modes) in present devices can change character: transition from Bohm-like scaling ~ (  i v i ) to Larmor-orbit-dependent “Gyro-Bohm” scaling ~ (  i v i )(  I / a ) “Rollover” is good news ! (since simple extrapolation is pessimistic)

23. 0.01 0.1 1 10 100 1 10 100 1000 10000 number of processors computing power IBM SP CRAY T3E 3D Gyrokinetic Toroidal Code (GTC) Scalable on Massively Parallel Computers Y-axis: number of particles (in millions) which move one step in one second

24. Plasma Edge Turbulence Studies: Experiment and Simulation Comparisons S. Zweben & J. Terry et al. + B. Rogers & K. Hallatschek, et al. + D. Stotler (Paper UI1.004) Gas Puff Imaging (GPI) Experiments on Alcator C-Mod Tokamak interpreted with MPP neutrals code (DEGAS 2)  GPI results compared vs. 3D EM fluid code local (flux tube) simulations of plasma 0.5 cm outside separatrix k (cm -1 ) Fluctuation amplitude Simulation GPI Images normalized to same total amplitude Initial k-spectrum Comparisons

25. New Cross-Disciplinary Opportunities for Diagnosing and Understanding Turbulence Z. Lin, GTC Simulation G.J. Kramer, E. Valeo, R. Nazikian, Full Wave Simulation of  -wave Reflection S. Klasky, I. Zatz, Visualization Target plasmaGrowth of Radial structuresZonal flows and decorrelation Break-up and scattering of microwaves from plasma turbulence

27. THE NATIONAL FUSION ENERGY SCIENCES COLLABORATORY (involves 40 US sites in 37 states) Collaboratory Goals: -- enable more efficient use of experimental facilities by developing more powerful between pulse data analysis -- enable better access by researchers to analysis & simulation codes, data, and visualization tools -- create standard tool set for remote data access, security, and visualization Collaboratory Partners: D. Schissel, et al. : -- 3 large fusion experiments* * C-MOD, DIII-D, NSTX -- 4 computer science centers ** ** ANL, LBNL, Princeton U., U. of Utah

28. STELLARATOR DESIGN STUDIES Optimization of Stability, Transport, and Constructability for Designing National Compact Stellarator Experiment (NCSX) Utilization of MPP Computations Essential for Optimizations

29. •Collaboration on Magnetic Reconnection Simulations U. Iowa, U. Chicago, U. Texas "FLASH CODE: R. Rosner, et al., U. Chicago " Solves fully compressible Navier Stokes Equations (explicit viscosity, implicit dissipation, single-fluid MHD) " Fully parallel and uses Adaptive Mesh Refinement " Supercomputing 2000/Gordon Bell Prize winner " Large and diverse scope of applications Cellular detonations Compressed turbulence Helium burning on neutron stars Richtmyer-Meshkov instability Laser-driven shock instabilities Nova outbursts on white dwarfs Rayleigh-Taylor instability Flame-vortex interactions

30. Relation to other scientific disciplines Space Physics –reconnection in Earth’s magnetosphere, solar corona, astrophysical plasmas –dynamos, collective phenomena, ……. High Energy Physics –Collective dynamics impacting advanced accelerator design Industrial Applications –Plasma Processing, Xerography, Flat Panel Display, …. Computational Physics -- issues common to many areas –advances in solving partial differential equations in complex geometry, –adaptive mesh refinement in 3D, –parallel methods for inverting sparse matrices –etc.

31. Dipole “surface mode” can be destabilized with introduction of background electron component [BEST Code : 3D PIC for ions and electrons] Electron-Proton Two-Stream Instability Growing from Initial Noise Two-stream instability can be stabilized by a modest axial momentum spread. [unwanted electrons in LANL Proton Storage Ring and the Spallation Neutron Source Project] t=0t=200

32. Particle simulations with 70 M particles and 20 M grid points Development of turbulence in 3-D model –two-stream instabilities –anomalous resistivity 3-D Magnetic Reconnection and Anomalous Resistivity U. Maryland, Max Planck, Dartmouth Zeiler, Swisdak, et al. GM1.003 Drake, et al., GM1.007

33. Generation of “Electron Holes” (possible relevance to satellite observations) Intense electron beam generates two-stream instability –nonlinear evolution into “electron holes” localized regions of intense anti- parallel electric field –strong electron scattering x z EzEz f vzvz ions electrons

34. Driving Applications Science/Engineering Scalable Services Princeton University’s PICASso Program Program in Integrative Computer and Application Sciences Integrative Research and Training in Entire Computational Pipeline CS PPPL Astro Geo Bio Eng. GFDL Genomics Finance ModelsMethodsSoftware Networks and Distributed Systems Scalable Systems Data Management Visualization The Computational Pipeline Internet Services Biology, Genomics Astrophysics Plasma Physics Geosciences Mobile Services Information Archives

35. CONCLUSIONS Advanced Computations is cost-effectively aiding progress toward gaining the physics knowledge needed to harness fusion energy by making crucial contributions to all areas of Plasma Science. Advanced Computations is a natural bridge for fruitful collaborations between Plasma Science and other scientific disciplines. Plasma Science is both utilizing and contributing to the exciting advances in Information Technology and Scientific Computing. Computational Plasma Science is helping to attract, educate, & retain young talent essential for the future.