The American Physical Society (APS) has recognized a summer intern at the U.S. Department of Energy's (DOE) Princeton Plasma Physics Laboratory (PPPL) for producing an outstanding research poster at the world-wide APS Division of Plasma Physics (DPP) gathering last October. The student, Marco Miller, a senior at Columbia University majoring in applied physics, used machine learning to accelerate a leading PPPL computer code known as XGC as a participant in the DOE's Summer Undergraduate Laboratory Internship (SULI) program in 2019.
The modifications, which will enable the XGC code to calculate more quickly, could help expand the physics included in detailed simulations of the plasma that fuels fusion reactions. The poster, prepared under the mentorship of PPPL physicist Michael Churchill, showed how Miller used machine learning techniques in his research and was presented at the APS-DPP conference in Fort Lauderdale, Florida. "It felt great to get the award," Miller said. "It was an honor, especially in a room with a lot of very good posters."
Fusion, the power that drives the sun and stars, combines light elements in the form of plasma - the hot, charged state of matter composed of free electrons and electrically charged atoms, or ions - that generates massive amounts of energy. Scientists are seeking to replicate fusion on Earth for a virtually inexhaustible supply of power to generate electricity.
Miller was trying to upgrade a part of XGC known as the collision operator, which simulates particle collisions, to produce accurate simulations of plasma particle behavior more quickly than the current operator. Miller and Churchill gathered data that had been fed into XGC in the past, along with data the code produced. The researchers then used a machine learning algorithm to figure out how the new operator could produce the same simulations with fewer steps. "The work was quite successful," Churchill said.
Machine learning is a process that teaches computers how to accurately identify an object by providing lots of examples. For example, the technique could be used to show computers how to identify dogs in photographs, or how to correctly translate a word from one language to another.
"One of the big parts of this research was making sure that the new code not only used machine learning techniques, but also captured the physics involved in the collisions," Miller said. "It's important that the inner workings of the code match what is happening in the real world."
Miller and Churchill feel they have made significant progress although the project has not yet been completed. "We haven't yet ported the new code into XGC, but if we were to do so, I think we would notice a significant speed-up," Miller said.
PPPL physicist C.S. Chang, who heads the XGC program, agrees with that notion, particularly as it applies to ITER, the multinational tokamak being built in France to demonstrate the feasibility of fusion energy. ITER will produce a huge amount of data that will slow XGC's calculations, Chang said. "Miller's new collision operator could cut the computing time significantly."
The SULI internship was not Miller's first encounter with plasma physics. He is involved with the plasma physics lab at Columbia and spent a semester in France doing research supporting the construction of ITER, a multinational tokamak facility being built to demonstrate the feasibility of fusion energy.
Miller is now applying to graduate plasma physics programs. "It's looking pretty likely that I'll be doing more plasma physics in the future," he said. "It would be great to work at a lab like PPPL."
