Feature - Grid computing opens new line of studies in fusion
Things are heating up in the world of fusion as grid computing empowers an all-new approach to studing plasma behavior: simulations that extrapolate the individual trajectories of tens of millions of plasma particles.
Introduction: fusion in a magnetic nutshell
Magnetic confinement fusion is a promising energy source; however, its commercial exploitation requires an intricate understanding of the unusual properties of plasmas.
When matter is exposed to the very high temperatures required for commercial fusion—hundreds of millions of degrees—it enters a physical state with properties unlike any other: the plasma.
In the case of commercial fusion, the unusual properties of plasmas are further complicated by the presence of strong magnetic fields, which form protective magnetic “coils” around the plasma.
Understanding the way in which heat and particles propagate from the centre of this confinement device is crucial to plasma confinement and maintenance, and thus to progress towards commercial fusion.
New computing power; new powerful results
Scientists have traditionally treated plasmas as fluids with very special properties, studying them using a magnetized version of the Navier-Stokes equation. This has meant the calculations that must be solved are usually non-linear and require massive computational power.
Grid computing is now enabling another possible approach: the separate study of a large number of test plasma particles, whose behaviour can then be extrapolated to the whole plasma.
Ten to 100 million particles are typically required for this kind of calculation, and since each particle is independent, the computations can be distributed using grid computing techniques.
Fusion in parallel
In this approach, the trajectory of each individual particle is calculated under conditions similar to experimental conditions, and the macroscopic properties of the entire plasma are computed from the full population of particles.
This allows fusion scientists to calculate the particle and heat fluxes everywhere in the device, while relaxing important approximations usually customary in such studies. For instance, this new approach makes no assumptions regarding the typical size of trajectories, nor about the diffusive nature of transport.
New discoveries from grid-aided thinking
Thanks to this use of grid computing, a new set of collisional transport properties—such as the lack of symmetries and the non-diffusive character of transport—has been obtained. For example, researchers can now estimate the particle confinement time inside the plasma.
Further, this approach has allowed scientists to estimate the zones from which particles are more likely to escape. Strategies to avoid particle loss can thus be designed. For example, a strategy to deviate and intercept particles before they strike the container wall is under development in TJ-II, at CIEMAT in Madrid, Spain.
These studies will also provide a foundation for future stellarators, like the National Compact Stellarator Experiment, which will be built in the U.S.
Further innovations in store
Presently, particle trajectories are estimated using the effects of the magnetic field, the electric field, and collisions with the background plasma. Researchers are now working to introduce the effect of new phenomena, such as microwave heating and turbulence. These phenomena are also being studied for the ITER tokamak, using the same techniques.
This fusion research is currently operating on EGEE. The EU-funded EUFORIA project will provide further support for the fusion community with the help of the EGEE grid and the DEISA supercomputing centers.
- Francisco Castejón, Coordinator of Fusion Activities in EGEE, Asocaición euratom/CIEMAT para Fusión