Progress through computation

The Quasi Poloidal Stellarator was designed with the help of a computer optimization model that used an extensive suite of physics design tools. The QPS is expected to be built by 2008. Full size image available here.

If we continue to burn fossil fuels for energy at the current rate, they will last only another few hundred years. In the context of civilization, the fossil fuel era is drawing to a close. In addition, it would be wise to reduce our combustion of oil, gas, and coal because the process produces pollutants that are bad for our health and carbon dioxide that could change our climate in undesirable ways. One possible future source of electricity for the world is fusion energy.

When the fusion process that produces the energy of the sun and stars is harnessed on the earth, we will have a virtually unlimited supply of clean energy.

Achieving this goal will occur when a hot plasma gas containing heavy hydrogen atoms--deuterium obtained from the ocean and tritium bred by the fusion process itself--is heated to a high enough temperature, squeezed to a high enough density, and held in this state by magnetic fields for a long enough time to sustain fusion reactions. The heat produced from these reactions could then be used to drive a turbine and generate electricity.

Which geometry should external magnetic fields have to confine the plasma in a way that ensures sustained fusion reactions? Can radiofrequency (rf) heating--high-frequency radio waves generated by oscillators outside the doughnut-shaped plasma vessel--possess the right set of frequencies to heat and control the plasma? Heating involves transferring rf energy to the charged particles in the plasma, which in turn collide with other plasma particles, boosting the temperature of the bulk plasma. Controlling the plasma involves suppressing its instabilities so fewer particles crash into the plasma vessel wall and cool the plasma. The most efficient way to answer these questions and to identify all the parameters needed for heating and controlling fusion plasmas in an economic fusion device is through fusion simulation on supercomputers and validation of these simulations by experiments.

In the past five years, the theory group in ORNL's Fusion Energy Division in collaboration with computer scientists and applied mathematicians in ORNL's Computer Science and Mathematics Division, has developed powerful computer codes to simulate fusion plasmas in two and three dimensions and predict their responses to simulated magnetic fields and radiofrequency heating. During that time, the Department of Energy's Center for Computational Sciences at ORNL obtained powerful supercomputers, including the IBM Power4 (Cheetah), which is a primary production resource for DOE's Scientific Discovery through Advanced Computing (SciDAC) program, and the SGI Altix, which has the extra advantage of having a large memory that holds 2 trillion bytes of data.

ORNL's fusion theorists have been using these two supercomputers for their SciDAC-sponsored research and have also ported their codes to the Cray X1 system at CCS. They are glad DOE is providing funding to build a supercomputer at ORNL that will exceed the computational speed record currently held by Japan's Earth Simulator supercomputer. A supercomputer with this much capability will hasten the solution of design problems related to future fusion power devices.

ORNL and ITER

One DOE goal is to demonstrate a sustained, self-heated (burning) fusion plasma, in which the plasma is maintained at fusion temperatures by the heat generated by the fusion reaction itself. To attain this ambitious goal, DOE recently declared the International Thermonuclear Experimental Reactor (ITER) to be its number one near-term-science facility of the future. ITER, the first facility capable of producing a sustained burning plasma, will demonstrate the scientific and technological feasibility of fusion energy. To fully exploit ITER's capabilities and to guide the design and development of subsequent fusion power devices will eventually require a predictive capability for fusion systems. To accomplish this goal, DOE will launch a major effort to advance state-of-the-art computational modeling and simulation of plasma behavior. ORNL's fusion simulation work will contribute to this effort and ORNL's involvement with ITER will allow fusion researchers to use data from the ITER project to validate simulation programs and reliably design future fusion reactors. The fusion community at ORNL in collaboration with other major U.S. research institutions, such as DOE's Princeton Plasma Physics Laboratory (PPPL), will leverage American successes in fusion simulation to regain a U.S. leadership role in this $5 billion international project.

Simulation--A Shortcut to Solutions

Fusion energy research has a long history of employing supercomputers to solve highly complex mathematical equations. Physicists know the equations, and applied mathematicians and computer scientists bring expertise in solving them. In addition to fusion specialists,

ORNL is home to a team of world-class physicists, mathematicians, and computer scientists who collaborate with their peers from other national laboratories and international organizations. To solve complex scientific problems, the research community must develop new algorithms, optimize computer codes for modern computer architectures, and help vendors design and build ultrascale supercomputers.

ORNL's All Orders Spectral Algorithm computational model for analyzing plasma wave interactions is a success story in integrating physics and software. Full size image available here.

The ultimate goals of fusion simulation are quite ambitious--to predict, reliably, the behavior of plasma discharges in a toroidal magnetic fusion device on all relevant time and space scales. This effort would bring together into one framework all the codes and models that presently constitute separate disciplines within plasma science. Scientific and numerical issues awaiting solutions are those associated with the coupling of codes with different natural time- and length-scales.

In the near term, ORNL scientists will be carrying out massively parallel computations on the IBM Power3 supercomputers at DOE's National Energy Research Scientific Computing Center (NERSC) and on the IBM Power4 and Cray X1 supercomputers at CCS. These fusion simulation efforts, including collaborative projects funded by SciDAC, focus primarily on grand challenge problems in four areas: transport processes in plasmas, wave/plasma interaction, advanced stellarator designs, and simulation of plasma evolution.

Advanced stellarator designs, for example, have the potential for leading to a fusion reactor that is smaller and more economically attractive than existing stellarators and would eliminate the potentially damaging plasma disruptions that plague conventional research tokamaks (including ITER). Stellarators employ 3-D magnetic fields as opposed to the 2-D fields used for tokamaks. This added complexity has historically limited our ability to design optimized systems. New software and faster computers have lifted this constraint. ORNL, in collaboration with PPPL, has developed a computer optimization model that employs an extensive suite of stellarator design tools. These tools enabled the design of compact stellarator facilities valued at $100 million, including the Quasi-Poloidal Stellarator that DOE plans to build at ORNL later this decade.

Simulation of plasma evolution will help validate recent evidence that non-linear processes called self-organized criticality exists for fusion plasmas. This approach could provide insight into the vulnerability of complex systems such as power grids and communication networks.

Fusion Energy Institute

In the past, tough physics problems could be attacked only piecemeal. Now, thanks to advanced computer simulations, more powerful supercomputers, and improved data storage, these once intractable problems are now within the realm of being solved systematically. To speed this process, ORNL experts in physics, applied mathematics, and computer science have formed a Fusion Energy Institute in CCS. The Institute will draw on the expertise of the world's best fusion scientists to identify grand challenges in fusion that can be solved only by using the leadership-class computing capabilities available at CCS. Because CCS will become a site for the world's fastest scientific supercomputer, it is more likely that today's intractable fusion problems will be solved, enabling a more rapid design of cost-effective fusion devices for power production.