PPPL physicists win supercomputing time to simulate key energy and astrophysical phenomena

Thursday, Jan 10, 2013

 Visualization by Prof. Kwan-Liu’s group, University of California-Davis

Three teams led by scientists at the U.S. Department of Energy’s (DOE) Princeton Plasma Physics Laboratory (PPPL) have won major blocks of time on two of the world’s most powerful supercomputers. Two of the projects seek to advance the development of nuclear fusion as a clean and abundant source of energy by improving understanding of the superhot, electrically charged plasma gas that fuels fusion reactions. The third project seeks to extend understanding of a process called magnetic reconnection, which is widely believed to play a critical role in the explosive release of magnetic energy in phenomena like solar flares that can disrupt cell phone service and black out power grids.

“This is great for the Laboratory,” PPPL Director Stewart Prager said of the highly competitive, three-year awards. “Getting this kind of computing time allows the solution of complex equations and critical issues that wouldn’t be possible otherwise.”

The three-year awards come from a DOE program to accelerate scientific discovery called Innovative and Novel Impact on Computational Theory and Experiment (INCITE). The awards provide allocations of millions of computer core—or processor—hours.  For example, 100 million core hours would equal roughly100 million hours—or 11,000 years—on a desktop computer powered by a single processor.

The PPPL recipients:

A nationwide center headed by PPPL physicist C.S. Chang that is developing computer codes to simulate the dazzlingly complex conditions at the edge of magnetically confined plasmas in donut-shaped devices called tokamaks. Chang’s team, the Center for Edge Physics Simulation (EPSI), won 100 million core hours a year on Titan, a Cray XK7 machine that is housed at the DOE’s Oak Ridge National Laboratory and has been proven to perform over 17 quadrillion—or million billion—calculations a second, making it the world’s fastest supercomputer, according to the November, 2012, TOP500 list.

The allotted time could go surprisingly quickly. “If we utilized one million cores an hour we would have the machine for 100 hours or less than five days,” said Chang.

Learning to control the turbulent plasma edge will be vital for future fusion facilities such as ITER, an international tokamak of unprecedented size and power that is under construction in France. Uncontrolled turbulence can disturb the rest of the plasma and prevent fusion from taking place, while controlling the edge conditions can make the fusion reaction more efficient. “This is definitely among the most critical issues facing fusion plasma physics,” said Prager, “and one that is perfect for an INCITE award.”

The INCITE award will enhance a five-year, $12.5 million project to simulate the plasma edge that Chang’s EPSI team has already begun. Funds for that project come from the DOE’s Scientific Discovery through Advanced Computing (SciDAC) program supported by the Department’s Office of Science. Participants in EPSI include physicists, mathematicians and computer scientists from 11 U.S. research institutions, together with PPPL staffers Stephane Ethier; Seung-Hoe Ku; Jianying Lang; and Darren Stotler, and postdoctoral fellow Robert Hager.

A PPPL-led international team that is studying the rapid loss of plasma confinement caused by growing turbulence as fusion facilities become larger and more powerful. Such losses can significantly decrease the power output of fusion systems but have been shown to level off when facilities reach a certain size—a development that bodes well for future tokamaks. “This is very good news for ITER,” said project leader William Tang, a PPPL physicist and Princeton University lecturer with the rank of professor in the Department of Astrophysical Sciences.

Tang’s project, called “Kinetic Simulations of Fusion Energy Dynamics at the Extreme Scale,” won 40 million core hours on Mira, an IBM Blue Gene/Q supercomputer at the DOE’s Argonne National Laboratory. Mira can calculate 10 million billion times a second, a speed that will be needed to simulate the complex processes that cause the turbulence to grow to a certain level as the plasma size increases, only to stop growing when the dimensions of the system increase further. “The question is a very basic one,” said Tang. “What’s the physics behind this favorable trend that is expected to occur in large plasmas such as ITER? No one can presently answer this question, which will require the efficient engagement of computing at the extreme scale to properly address.”

Joining Tang in the quest for answers will be PPPL researchers Stephane Ethier and Weixing Wang, together with Bei Wang from the Princeton Institute for Computational Science and Engineering (PICSciE) at Princeton University, and co-investigators from Columbia University, the Max Planck Institute for Plasma Physics in Garching, Germany, and the DOE’s Oak Ridge and Lawrence Berkeley national laboratories.

Researchers investigating magnetic reconnection, an astrophysical phenomenon that gives rise to the northern lights, solar flares and geomagnetic storms. A team led by Amitava Bhattacharjee, head of the Theory Department at PPPL and a professor of astrophysical sciences at Princeton University, won 35 million core hours on the Titan supercomputer at Oak Ridge. 

Reconnection takes place when the magnetic field lines in merging plasmas snap apart and explosively reconnect, a process seen throughout the universe and in disruptions of plasma during fusion experiments. New insight into reconnection could lead to better predictions of geomagnetic storms and other space weather, and to greater control of experimental fusion reactions.

Plans call for Bhattacharjee’s team to develop codes to simulate the reconnection that occurs when researchers focus high-powered laser beams on tiny spots of foil. This creates fast-expanding plasma bubbles with magnetic fields that trigger reconnection when the bubbles come together. The recently observed phenomena open up “a new regime of reconnection study of great interest to laboratory and plasma astrophysics,” said Bhattacharjee, who launched the code-development project while a professor of physics at the University of New Hampshire (UNH) before joining PPPL and Princeton in August, 2012.

Work on the reconnection project will continue under Bhattacharjee with collaboration by UNH physicists Naoki Bessho; William Fox; Kai Germaschewski; and Yi-Min Huang.

The supercomputing awards to PPPL were among five 2013 INCITE awards to researchers at Princeton University. Also winning supercomputing time were Emily Carter, the Gerhard R. Andlinger Professor in Energy, and a professor of mechanical and aerospace engineering & applied and computational mathematics; and Jeroen Tromp, a professor of geosciences, applied and computational mathematics, and Blair Professor of Geology. Carter and collaborator Lin-Wang Wang of Lawrence Berkeley National Laboratory won 25 million core hours on Titan to simulate processes inside nanosystems—systems measured in billionths of a meter. Tromp, who also is director of PICSciE, and co-investigator Olaf Schenk of the University of Lugano, Switzerland, received 100 million core hours on Titan to develop computer models of the interior of the Earth.