FOR years, magnetic fusion energy (MFE) scientists have dreamed of an integrated, easy-to-use, and comprehensive family of computer codes that would simultaneously simulate all of the important physics processes that take place in a magnetic fusion reactor. Such a package would be valuable for enhancing the understanding of the extremely complex phenomena observed in experiments. It would also provide a tool to optimally design future MFE experiments.
Although this goal of virtual experiments is still years away, researchers at Lawrence Livermore have made important advances in developing such a comprehensive simulation package. In designing such a code, called Corsica, they have developed techniques to efficiently couple separate physics processes. These techniques plus continuing advances in high-performance computer hardware and software offer the prospect of achieving the goal. Corsica is only one part of a widespread effort, evident throughout Livermore research programs, to simulate to unprecedented levels of accuracy the physical phenomena taking place on scales ranging from atomic particle interaction to global weather patterns.
In fusion, two light nuclei (such as hydrogen) combine into one new nucleus (such as helium) and release enormous energy in the process. One approach to fusion uses a powerful magnetic field to confine a plasma (a gas consisting of charged ions and electrons), generating energy in a controlled manner. To date, the most successful approach for achieving controlled fusion is in a donut-shape configuration called a tokamak.
Future experimental facilities will be much larger than today's research tokamaks and much more expensive, costing as much as several billion dollars each. Although advanced simulations will never replace experimental work, they are needed to optimize the design of future tokamaks and experiments, which will save millions--maybe billions--of dollars. Simulations are also needed to analyze and optimize alternative concepts to the tokamak, such as the small spheromak device that Livermore is now constructing. Scientists consider such simulations essential to resolving several important physics issues, such as how to structure the magnetic fields to produce the maximum pressure and what processes drive electric currents and magnetic fields.





The Need for Integration
"Simulations of individual phenomena--physics 'packages'--now exist as essential tools for analyzing fusion experiments," says Livermore's Keith Thomassen, Deputy Associate Director for MFE. Phenomena such as the equilibrium of the plasma, turbulent transport, stability, and heating, are examples of such processes and are interdependent. Thus, codes that simultaneously describe all these phenomena have these packages "hard wired" together, and the codes are extremely complex. A contributor to this complexity is the disparate time and spatial scales of these phenomena.
MFE processes span a wide range of time scales, from turbulent fluctuations on the microsecond scale to transport processes with scales of seconds to hours. For example, particle velocities parallel to the magnetic field are orders of magnitude greater than those moving perpendicular to the magnetic field. Additionally, MFE models must reflect a range of spatial scales that extend, for example, from the spiral orbit size of an electron to the several-meter-wide tokamak device.
David Baldwin, vice president for fusion research at General Atomics (GA) in San Diego, was Associate Director for Energy Programs at Livermore in the early 1990s when Corsica development began. "The MFE community had done a very good job in developing discreet physics packages but had little experience integrating them. We wanted to provide that capability for the first time but in modular units," he says, "so that, as individual physics packages improved, they could be exchanged. But the entire code would not have to be rewritten."
The inspiration for Corsica, says Baldwin, was Livermore's long-standing LASNEX integrated code for laser fusion that simulates interactions between laser light and its targets. He also notes that Corsica's objective of integrating all the physics phenomena is analogous to that of the Accelerated Strategic Computing Initiative (ASCI), the Department of Energy's effort in the Stockpile Stewardship Program to develop full-system simulations running on new generations of high-performance computers. Indeed, several Livermore Corsica developers are also contributing to ASCI.
Corsica I, released in 1994, flexibly coupled one-dimensional calculations of particle, energy, and magnetic-flux transport in the core, or confined region, of the plasma to a calculation of a two-dimensional magnetic configuration. Corsica II, released for testing by sophisticated users in 1995, coupled the one-dimensional core transport calculation to a simulation of the two-dimensional "edge" where magnetic-field lines intersect material surfaces.
A still more advanced version, Corsica III, is being developed as funding permits. Its ambitious goal is to couple the evolution of the fusion process to plasma turbulence models for a more comprehensive simulation. The turbulence coupling effort benefits from the sheer computing power of the newest supercomputers as well as from the experience in turbulence of Livermore's MFE theory group. The group participates in the Department of Energy's Numerical Tokamak Turbulence Project, designated a DOE grand challenge.
By handling a wide range of physics with disparate time and spatial scales, Corsica permits more complete modeling of toroidal (donut-shaped) plasmas and other magnetic fusion concepts than previously possible. As a result, the software has been adopted at both national and international research sites. For example, Corsica is used to model plasma confinement in the DIII-D experimental tokamak at GA, where a team of Lawrence Livermore scientists is working. Corsica was used on simulations of the Tokamak Fusion Test Reactor that operated at the Princeton Plasma Physics Laboratory. Corsica has also been used to model the design for the International Thermonuclear Experimental Reactor and Livermore's Sustained Spheromak Physics Experiment.

Corsica Keys on Flexibility
Livermore physicist Tom Casper uses Corsica to replicate experiments conducted on GA's DIII-D tokamak. He says Corsica helps him obtain a better understanding of past experiments and plan future experiments. He points to the code's flexibility as one of its strong suits. "I can modify the code and manipulate variables, capabilities we don't find in other codes."
In addition, Casper notes that budgets and space constraints limit the number of diagnostic instruments that can be used on the DIII-D tokamak. With Corsica running on a powerful workstation, Casper can add as many so-called synthetic diagnostics as he desires and thereby gain a more complete picture of an experiment.
Corsica's flexibility is provided in part by BASIS, a code steering system developed by Livermore computational scientist Paul Dubois in the 1980s. BASIS also served as the underlying system for LASNEX. Leaders of both the Corsica and ASCI programs have recently chosen Python as the successor to BASIS because of its greater flexibility and its ability to run on many different computers.
Baldwin says Corsica today is "darn good" but still far from the complete simulation code that he and other leaders of the MFE community envision. For that to happen, he says, other major MFE centers need to join forces with Livermore and GA and integrate their own specialized codes, as well as tap the experience of other comprehensive code efforts such as ASCI.
"We want to create a national project that benefits from our experience here," says Ron Cohen, leader of Livermore's MFE theory and computation program. Cohen and others have proposed a "national transport code," a modern code that would draw upon the concepts in Corsica and the vast wealth of physics simulation "building blocks" available nationally. Cohen notes that because the family of MFE simulation codes was created at different research centers in different software languages and on different computers, the effort to modify them and then combine them into a seamless, integrated package will demand strong cooperation among researchers.
In the meantime, Livermore scientists are collaborating with colleagues at GA to combine the best features of GA's transport code with those of Corsica. They hope to carry out their project within the context of the national transport code project.
If the national effort is successful, users will be able to select from many different physics "plug-in" modules to build custom simulations of various--or all--aspects of a magnetic-fusion experiment (tokamak or other configuration). The integrated code would also allow a user to modify experimental conditions and theory parameters or even whole models in real time, without having to use elaborate commands.
"We want to have an integrated, 'living' code that keeps changing and growing as computer capabilities and ideas advance," says Cohen. While it could never serve as a substitute for an actual experiment, such an integrated simulation would undoubtedly save money by enabling better-designed and more aggressive experiments. "By combining data from this code with hardware experiments, we'd get the most out of the research dollar," he says. -- Arnie Heller

For further information contact Ronald H. Cohen (925) 422-9831 (cohen2@llnl.gov).


Back to May 1998