The focus on confinement concept improvement will also mean following a multiple-concept development strategy to identify a more attractive reactor than that projected from the standard tokamak. Enhanced simulation capabilities will allow us to gain more physics insight out of our experimental facilities and theory research. Alternatives to the tokamak based on inertial or magnetic confinement principles will especially benefit from advanced simulations, since it will be impossible to fully test experimentally all approaches to fusion energy which show initial theoretical promise.

Addressing global climate change will require new technologies for power generation to be deployed in the early part of the next century. Fusion offers one of a relatively few acceptable technologies with the potential for reducing global climate change. However, the long lead time required to develop a commercial fusion power plant requires that we act now to accelerate the development and demonstration of this technology.

The next few years will be a period of important opportunities for both the US and world magnetic fusion programs. For the past decade, the US has been a partner in an international effort involving the European community, Japan and Russia, in an effort to design a burning plasma experiment based on the tokamak concept. During the next 2-3 years, an international design team will be working on reduced cost options. An accelerated US simulation and modeling effort will enable the US to maintain a strong scientific role in influencing this process, and will put us in a unique position to extract maximum scientific output from the operation of this multi-billion dollar experiment.

Another opportunity is in inertial fusion energy (IFE) which is a promising and fundamentally different approach from MFE. The inertial confinement fusion (ICF) program at DOE will begin operation of the National Ignition Facility (NIF) in 2003 under the auspices of Stockpile Stewardship. The physics of IFE and ICF has significant overlap, and definitive data on fusion target physics from NIF, aided by enhanced simulation and modeling of IFE systems will greatly accelerate progress in evaluating IFE as an attractive energy concept.

Besides these burning and ignited plasma experiments, the U.S. fusion program is supporting a broadening array of innovative confinement concepts (ICC). However it will not be possible to support each of these concepts through its complete development life from proof-of-principle to proof-of-performance. A validated simulation capability is a critical ingredient in supporting the development of these concepts and will aid in making the programmatic decisions that will have to be made in the next 5-10 years.

Marked progress has been made in recent years in the understanding and simulation of plasma turbulence and the associated plasma transport. This has been due in large part to the national Turbulence Task Force (TTF) and the Numerical Tokamak Turbulence Project (NTTP), a DOE high performance computing grand challenge. A significant innovation that has made quantitative simulation possible is the use of a reduced description of ion motions in a strong magnetic field, known as the gyrokinetic Vlasov approach. Scalable gyrokinetic codes exist, and have had considerable success in reproducing many aspects of experimentally observed core plasma transport. The past few years have also seen the development of promising simulation models for turbulence in the edge, a region complicated by the presence of both open and closed magnetic-field lines and by the importance of atomic processes.

3D Nonlinear MHD

The magneto-hydro-dynamic (MHD) description of MFE plasmas treats the plasma as an electrical conducting fluid. Significant progress is now being made in computing 3D static configurations and in developing 3-dimensional scalable nonlinear time-dependent MHD codes, including resistive and two-fluid effects needed to simulate the long-time behavior of high-temperature magnetic fusion devices. The development of closure schemes appropriate for long-mean-free-path plasmas is an active area of research. Idealized boundary conditions are also being replaced by more realistic models to better represent the physical configurations of interest.

Edge-modeling and material/wall interaction

Scalable 2-dimensional fluid codes have been successfully developed to treat both open and closed magnetic field line regions. These codes follow the evolution of the plasma dynamics parallel and perpendicular to the magnetic field, and include extensive neutral gas, atomic and radiation physics modules. A separate effort involves the modeling of plasma-material interaction to describe surface sputtering, reflection, redeposition and other important plasma surface phenomena under high heat and neutron flux conditions.

Much progress has been made in developing detailed models for the high-energy neutral beams and the high-power radio-frequency (RF) systems that heat the plasma to ignition and induce electrical current. The RF models include electromagnetic wave launching, wave propagation, wave-particle interaction, and plasma response. These models are critical for simulating present day experiments and projecting into the future.

Several "2D transport" codes exist that integrate together many important effects that combine to describe the physics of a complete MFE fusion experiment. It has become routine to couple together a transport time-scale description of the evolution of the plasma density and temperature profiles with those describing the evolution of the magnetic field while incorporating the plasma force-balance equilibrium constraint. These codes also contain models of plasma sources and sinks, including neutral beam heating, wave heating and current drive, etc.

Complex models have been developed to describe the dynamics of the intense beams (themselves non-neutral plasmas) for heavy-ion driven IFE, and the radiation-hydrodynamics of the targets for all IFE approaches. The dynamics of the beams in a heavy-ion IFE driver are dominated by space-charge effects (rather than by thermal pressure as in most accelerators), and so special codes have of necessity been developed for their study. Both moment-equation and discrete-particle methods are employed, but the latter are key to this endeavor because beam phase-space dilution (which must be avoided) is a kinetic phenomenon. For the target calculations, sophisticated codes exist in 2-D, but the 3-D codes that are needed for both laser and advanced heavy-ion targets are in their infancy.

Currently, the major computational resources used by fusion energy researchers are supplied by NERSC on the Cray C-90 and T3E computers supplemented by high-end workstations and the Cray J-90 computers. A facility with peak performance approaching ~100 TFlops and with memory of order 10 TB would enable a qualitative improvement in simulations for fusion energy research. Specifically,

(1) Turbulence simulations could realistically model electron dynamics (as well as ion dynamics which are being modeled now). They could also be extended from electrostatic to electromagnetic turbulence. Simulation space and time scales could be made to cover the full-radius of a fusion device with times approaching the energy replacement time.

(2) 3D nonlinear MHD stability simulations could approach realistic values of the magnetic Reynolds numbers (ratio of magnetic field diffusion time to Alfven wave transit time) under fusion conditions, and have sufficient numerical resolution to follow nonlinear 3-D macroscopic instability development in realistic geometry with realistic boundary conditions. MHD closures involving particle dynamics could be implemented

(4) High-resolution 2D and 3D full wave RF codes could be made capable of resolving the requisite range of spatial scales and it would become feasible to integrate these into the more general simulation codes.

We are now at a stage where computers are powerful enough so that we can consider refining and combining the above, largely independent models, to develop a few comprehensive models that will predict most essential aspects of the performance of actual fusion devices. One way to phrase the challenge is that of combining microscopic and macroscopic phenomena in a single simulation. The new computational resources should make integrated computational models feasible that contain all essential physical effects, and hence describe virtually the entire range of relevant space and time scales in a fusion device.

For both MFE and IFE, it will generally become feasible to greatly increase the degree to which it is possible to integrate disparate complex phenomena into a single simulation capability. This will lead to much more realistic and reliable comprehensive predictive tools which can simulate whole-device performance. The 3 major MFE integration projects and the IFE project that will be undertaken as part of this initiative are:

(i) 3D Kinetic Equilibrium Approach

Restricting the modeling to steady-state, or equilibrium, in the macroscopic sence, produces many simplifications, since it is no longer necessary to solve equations with a wide range of time-scales. One modeling approach, of special interest for steady state operation, is to build on the existing well-developed 3D plasma equilibrium capabilities and on the ability of the gyrokinetics codes to calculate turbulent transport.

Present production 3D equilibrium codes can calculate a complete magnetic configuration if two one-dimensional profiles are specified, together with appropriate boundary conditions. The profiles that need to be specified are the plasma pressure and the amount of electrical current carried by the plasma inside each magnetic surface. These quantities can be calculated by the gyrokinetic codes (GK) if the magnetic configuration is known. This suggests an iterative approach between solving for the magnetic field structure given the pressure and current profiles, and solving for the pressure and current profiles given a magnetic field structure.

Once particle and heat sources are specified, and the currents in the external field coils are given, this approach should be able to calculate the unique plasma pressure and current profiles and the associated magnetic field structure. Although not capable of calculating transient disturbances, this approach should yield an invaluable design and analysis tool.

(ii) Multi-scale Transport Approach

The objective of this integrated modeling activity is to greatly expand the present 2D models that predicts all the surface-averaged device-scale behavior of the plasma in toroidal magnetic fusion devices on the time-scale that the bulk profiles change. This includes predicting the evolution of the temperature, density, and current profiles, as well as modeling the coupling of the plasma core and edge regions, the first wall, the RF systems, and the rest of the fusion device. It would also approximate the effect of large-scale instabilities (such as sawtooth oscillations, Edge Localized Modes (ELMs) and disruptions) on the plasma profiles, etc.. This integrated modeling code would also be used to predict such things as the evolution of the plasma current profile, the accumulation of impurities within the plasma and the erosion of the first wall caused by the interaction between the plasma and the wall.

The gyrokinetic description utilized in the turbulence simulations can provide a long-mean-free-path closure scheme for the 3D MHD equations. This 3D-hybrid MHD/gyrokinetic particle approach, in principle, will be able to solve for MHD phenomena, turbulent transport and edge/physics phenomena. Unlike the previous 2 approaches, the extended-MHD approach would seek to temporally and spatially resolve all the relevant physical processes contributing to 3D plasma disturbances. The many disparate time scales involved make it unfeasible to use this technique to solve for the evolution of plasma configurations to the steady state. However, this approach is thought to be the only best way to solve for transient phenomena that are known to occur in otherwise steady-state confinement devices, such as "sawtooth oscillations", L-H transitions, pedestal formation, "edge localized modes" (or ELMS), and plasma disruptions.

The success of this initiative can be measured by the demonstrated use of the new computational capabilities in augmenting theory and experiment to advance the science and discover new concepts that have significant impact on fusion energy development. In particular, the appropriate physics, programmatic, and computational goals are as follows:

Small amplitude fluctuations in the electromagnetic field and in the plasma properties are always present in a fusion plasma. These fluctuations are due to micro-instabilities that saturate at some small amplitude due to a variety of non-linear processes. The saturation amplitude determines the level of turbulence, and hence rate at which the charged particles and energy are transported across the confining magnetic fields. One of the primary goals of this initiative is to better understand the physical processes and time and space scales involved in this turbulence.

There have been two primary approaches to simulating plasmas, the fluid approach and the particle in cell approach. The fluid approach, known as Magneto-Hydro-Dynamics (MHD), or more precisely, extended MHD, needs a "closure model" to provide a closed set of fluid equations, and several types of closure models are presently being investigated. This initiative will allow us to study the theoretical properties and limitations of several closure models, including their ability to describe long mean-free-path "kinetic" phenomena, and a variety of transient plasma events. It will also allow us to investigate the MHD model at much higher values of the magnetic Reynolds number that has ever been possible before.

The plasma edge region is unique in that the physics in play is an interaction between that of ionized plasma, neutral fluids, and atomic processes. The "edge" also represents a transition in the magnetic configuration between "closed" magnetic surfaces that exist in the interior and "open" field lines that eventually intersect material surfaces. This region is rich with physical phenomena, including the physics of the plasma/wall interaction itself.

This initiative will provide more complete physical models using qualitatively higher resolution enabled by more efficient computational methods and more powerful hardware. It will be considered a programmatic success if it helps in improving the attractiveness of fusion as a future energy source. This help is expected to come in several ways.

The fusion SSI will provide computational tools that will enable the program to extract the maximum output from existing and near-term experiments. As these devices are used to validate the simulations, it will not only stimulate development of appropriate diagnostics, but will provide a means for continuous interplay between theory and experiment that should help guide the experimental program itself. We can expect that this new simulation capability will lead to the discovery of new operational regime for MFE devices.

Of course the true value of the validated simulation models is to enhance our design and predictive capability, and allow much-improved optimizations of magnetic confinement configurations. The simulation program described here should substantially narrow the range of predictions in size of a fusion power plant, and will be invaluable for the design of an IFE driver.

To achieve the physics and programmatic goals, a number of computational goals must also be met. We must demonstrate the scalability of the algorithms used in the fusion codes on the massively parallel computers with upwards of 1 to 10 thousand processors. This scalability is needed both for the particle-in-cell codes, and also for the fluid codes that utilize implicit algorithms.

Many complex physical phenomena need to be linked together utilizing an appropriate computational framework. Such a framework must be such that it is efficient, documented, and efficient, and also allows effective remote collaborations.

The project described in this initiative requires high-performance super-computers in the 10-100 Teraflop range, with memory of order 10 TB. The fluid codes require systems that perform well at solving systems of time-dependent PDEs, including sparse linear solvers. The particle simulation models will work efficiently on most any modern computer architecture.

IV.2 Software

The IFE elements of this proposal are synergistic with the MFE elements in several ways. The studies of space charge dominated beams for a heavy-ion fusion driver requires the use of plasma simulation methods. Some parts of the HIF beam codes (e.g., field-solvers, systems for passing data among subdomains) are very similar to those needed for MFE plasmas.

The challenge in plasma sciences of modeling complex systems (e.g., reactor-relevant magnetically confined plasmas) involving the realistic simulation of multiple phenomena spanning multiple temporal and spatial scales is common to many ER disciplines (materials sciences, climate modeling, etc.). Other more specific cross-cutting issues include: advances in solving partial differential equations in complex geometry, advanced Monte Carlo methods to deal with the complicated collisional dynamics associated with plasma-wall interactions, selection of the best methods for adaptive mesh refinement in 3D, software management techniques, improved distributed computing environments, database software, remote data access, modular programming envorinments, and network connections requirements to support distant collaborations.

We also anticipate synergetic benefits between the numerical simulation of anomalous energy transport in plasmas and of global climate change because both fields must address the fundamental challenge of turbulence simulation—the interaction of many disparate space and time scales. Numerical algorithms are required which accurately represent the dynamics of eddies at large to medium scale, maintaining a large effective Reynolds number while minimizing the required space and time resolution (to minimize the computational requirements).

Similarly, we anticipate synergy between the numerical simulation of plasma scrape-off-layers and of combustion. In toroidal confinement systems, heat flows from the plasma core while the interaction with impurity ions, atoms, and molecules in the plasma scrape-off-layer provides an important energy sink, while in combustion the atomic and molecular interactions provide an energy source. In both cases the energy source or sink interacts with the underlying fluid dynamics to produce laminar flows and "flame fronts"—sharp layers dominated by a particular atomic or molecular process—or instabilities and turbulence.

The IFE elements of this proposal are also synergistic with the ASCI effort and other DOE programs. Considerable benefit can be expected to result from more widespread dissemination of computational methods and tools, initially developed for HIF beam simulation, into other accelerator applications. Emerging high-current research areas include the Accelerator Production of Tritium, the Spallation Neutron Source, Accelerator Transmutation of Waste, and radiographic hydrotest facilities for nuclear stockpile stewardship. IFE research benefits considerably from the DOE-DP’s stockpile stewardship programs in ICF and High Energy Density physics; Electromagnetic simulations of ion beam propagation in the fusion chamber have much in common with codes used for laser-plasma interaction studies for ICF.

The Fusion SSI, if successfully executed, will allow fusion to significantly improve its predictive capabilities and will harness large-scale simulation, alongside theory and experiment, in its quest for an attractive reactor. The responsibility for delivering on these important goals will be assigned to a new National Fusion SSI Center, established within one of the program’s research institutions, and reporting through the institution’s fusion management to DOE.

The National Center will be responsible for providing strong scientific leadership of the Fusion SSI, effective integration with the rest of the fusion program, and strong coupling with other fields such as climate-change modeling. One of the functions of the National Center will be to assure that the next generation of young computational fusion scientists will be trained to continue the program. Most importantly, the National Center will bear the responsibility for ensuring that high-value deliverables are produced for the research dollars spent on this initiative.

These project goals build upon existing efforts in the U.S. staff requirements to meet these goals (in addition to those researchers already supported by OFES) are estimated as:

Transport processes 8 FTEs

Macroscopic stability 8 FTEs

Edge and plasma-wall interaction 8 FTEs

Integrated Modeling of a magnetic confinement device 8 FTEs

Source to target IFE modeling 8 FTEs

Visualization Center 2 FTEs

Database group 2 FTEs

TOTAL 88 FTEs

Breakdown of FTEs: 44 computational physicist, 11 programmer, 33 junior computational physicist or physics post-doc (88 total)

A. Aydemir, R. Bangerter, G. Bateman, D. Batchelor, C. Bolton, B. Braams, J. Cary, V. Chan, B. Cohen, R. Cohen, W. Dorland, S. Eckstrand, A. Friedman, S. Jardin, S. Krasheninnikov, A. Kritz, W. Lee, W. Nevins, H. Neilson, S. Parker, D. Schnack, J. Van Dam, R. Waltz, K. C. Wilson, D. Stotler

(acknowledgments to P.Diamond, R. Goldston, T.Rognlien, W. Tang, S. Bodner, C. Verdon)