Science Success Story

« Back

Machine Learning Helps Plasma Physics Researchers Understand Turbulence Transport

XSEDE-Allocated Simulations Reveal New Insights that Could Advance Fusion Energy Use

By Kimberly Mann Bruch, SDSC Communications

This snapshot of vorticity and turbulence density from a simulation using the XSEDE-allocated Comet supercomputer at the San Diego Supercomputer Center illustrates a notable physics concept: the formation of zonal (i.e. y-direction) structures, which have important consequences for magnetic confinement devices. Credit: Robin Heinonen, University of California San Diego Center for Energy Research

For more than four decades, University of California, San Diego, Professor of Physics Patrick H. Diamond and his research group have been advancing our understanding of fundamental concepts in plasma physics. Most recently, Diamond worked with graduate student Robin Heinonen on a model reduction study that used the Extreme Science and Engineering Discovery Environment (XSEDE)-allocated Comet supercomputer at the San Diego Supercomputer Center at UC San Diego to showcase how machine learning produced a new model for plasma turbulence.

Plasmas have many applications, including fusion energy. When light nuclei fuse together, the mass of the products is less than that of the reactants, and the missing mass becomes energy – hence Albert Einstein's famous E=mc2 equation. In order for this to occur, temperatures must literally reach astronomical levels, such as those found in the Sun's core. At those temperatures, hydrogen changes from a gas to a plasma, an extremely high-energy state of matter where electrons separate from their atoms.

As for societal benefits, fusion energy is viewed as a longer-term solution to meeting the world's overall energy needs as the global population and the resulting demand for electricity increases. Fusion has the capability to provide large-scale, emissions-free energy wherever needed, while being a viable complement to intermittent renewables and battery storage. Although fusion energy shows great potential, some obstacles remain, one of them being how to manage plasma turbulence which drives the transport of heat and particles toward the wall of the fusion device, preventing adequate confinement.

"Turbulence and its transport is chaotic in a sense, but this chaos is ordered and constrained," said Heinonen, who co-authored Turbulence Model Reduction by Deep Learning with Diamond in the Physical Review E journal. "Moreover, in certain turbulent systems, the chaos conspires to spontaneously form large, long-lived coherent structures and, in many cases, we only have a tenuous understanding of why and how now. There are definitely aspects of structure formation and self-organization which we do understand, but it's still an active area of research."

The authors say that advances in machine learning (ML), notably deep learning techniques, have provided them with the tools they needed to develop their new model for turbulent transport. The power of ML lies in the ability to learn patterns from the data instead of being explicitly programmed and the ‘deep' in deep learning refers to the many layers of interconnected processing units in the model.

"This deep learning approach allowed us to run many simulations and then train a neural network," explained Heinonen. "The neural network then outputs the average turbulence-driven fluxes as a function of a handful of physical variables of interest, which reduces the dimensionality of the model from two-dimensional to one-dimensional, making it more tractable and easier to understand."

The researchers used a particularly simple ML model and applied it to a simplified, yet physically rich, model of plasma turbulence. This approach sacrifices some predictive power for the sake of interpretability, which can lead to new insights. Previously unreported results include a non-diffusive flux driven by the flow and higher-order corrections to the fluxes that are difficult to calculate or understand.

"After we received our Comet allocation, the XSEDE tech support team was extremely helpful from beginning to end," said Robin Heinonen, a UC San Diego physics graduate student. "Installing and running software on a supercomputer can be a major challenge, and I am grateful for the XSEDE support team."

Heinonen first learned about XSEDE through Kevin Smith, who works with the information technology program within the UC San Diego Physics Department. Heinonen said that Smith helped him with the start-up program proposal and within a few days he was granted an XSEDE allocation on SDSC's Comet supercomputer, which is part of XSEDE's portfolio of available resources.

"After we received our Comet allocation, the XSEDE tech support team was extremely helpful from beginning to end," said Heinonen. "Installing and running software on a supercomputer can be a major challenge, and I am grateful for the XSEDE support team."

The work on Comet was supported by XSEDE allocation (TG-PHY190014), which is supported by the National Science Foundation (ACI1548562). The research was supported by the U.S. Department of Energy, Office of Science, Office of Fusion Energy Sciences (DE-FG02-04ER54738).

 

At a Glance:

  • UC San Diego physics researchers used XSEDE-allocated Comet supercomputer at the San Diego Supercomputer Center to develop a new model for plasma turbulence that uses recent advances in machine learning techniques.

  • Those new techniques use computational models that automatically learn patterns from data instead of having to be programmed.

  • These types of fundamental physics studies have the potential to greatly influence the future of fusion energy research, which in turn assists with the future of generating electricity in novel ways.