New public-private partnership comes to PPPL through a novel program to speed the development of fusion energy

Written by
John Greenwald
Jan. 27, 2022

Expertise in artificial Intelligence (AI) and the design of twisty magnetic fields in fusion facilities called “stellarators” has brought a new public-private collaboration to the U.S. Department of Energy’s (DOE) Princeton Plasma Physics Laboratory (PPPL). The year-long collaboration is one of eight recent selections of the DOE’s Innovation Network for Fusion Energy (INFUSE) program to speed harvesting on Earth the fusion energy that powers the sun and stars. The collaboration brings to 13 the number of PPPL partnerships sponsored under INFUSE since the DOE launched the novel program in 2019. 

“Some of these recent awards are focused on technology development, material testing, and machine learning,” said Ahmed Diallo, deputy director of INFUSE and a fusion scientist at PPPL.  “This focus highlights the industry’s needs to accelerate the development of carbon-free fusion energy on the electricity grid – one of the main objectives of the COP26 United Nations Climate Change Conference.”

Renaissance Fusion

The new collaboration brings together three PPPL scientists with the U.S. affiliate of Renaissance Fusion, a start-up in Grenoble, France. Founding and heading Renaissance and its affiliate is physicist Francesco Volpe, who has taught at Columbia University and the University of Wisconsin and received recognitions that include a DOE Early Career Award.  

Renaissance America, the U.S. affiliate, is tapping the laboratory’s AI machine learning capabilities to accelerate predictions of the loss of key particles called alphas from fusion reactions in stellarators. Such predictions are now quite slow and developing machine learning software to speed them up should enable designers to quickly enhance the shape, or geometry, of stellarator magnetic fields to improve alpha particle confinement. 

Fusion combines light elements in the form of plasma — the hot, charged state of matter composed of free electrons and atomic nuclei, or ions, that makes up 99 percent of the visible universe — to generate vast amounts of energy. The high-energy alpha particles — helium ions produced and released in fusion reactions — transfer their energy to the main ions to sustain the fusion process in what is called a “burning plasma.” Confining alpha particles is essential to keeping fusion reactions going. 

“What Renaissance wants is to look into speeding up the calculation of fast ion losses that physicists use in stellarator optimization,” said physicist Michael Churchill, a machine-learning expert and principal PPPL investigator in the collaboration. “The idea will be to use machine learning to learn to take in different magnetic geometries and instantly predict what the fast ion losses will be.”

Better stellarators

“The results will help not only Renaissance but the entire fusion community to design better stellarators,” said Chris Smiet, Chief Scientific Officer for Renaissance Fusion and a former researcher at PPPL. “The big picture will be to use machine learning to predict for any stellarator how fast the alpha particles are leaving, and we’ll be able to do that lightning fast.” 

Software now requires thousands of computer hours to calculate losses since the extremely high energy of the alpha particles substantially slows tracking their trajectories. PPPL’s goal is to develop software to create a near-instant prediction.

These PPPL scientists are engaged in the collaboration: 

Chang Liu, a Theory Department researcher who has developed software to study the dynamics of energetic particles and has implemented his tools in comprehensive fusion plasma simulations. 

Harry Mynick, a world-renowned expert in stellarators, will advise the collaboration as an emeritus principal research physicist at PPPL. 

Michael Churchill, principal PPPL investigator for the project, will write the machine-learning software with input from Mynick and Liu. 

Successful completion of the project will generate an open-source dataset of stellarator configurations that scientists around the world can use to train their own models and advance AI research in stellarators, according to the company’s partnership proposal.

The latest collaboration further demonstrates the wide-ranging scientific expertise in fusion and plasma science that large and small companies seek from PPPL. Previous corporate partners include software giant Microsoft, which is using AI to improve predictions of dangerous tokamak disruptions, and spinoffs Tokamak Energy from the Culham Centre for Fusion Energy in the United Kingdom, and Commonwealth Fusion Systems (CFS) from the Massachusetts Institute of Technology (MIT). Tokamak Energy aims to develop a spherical tokamak that will confine plasma with high-temperature superconducting (HTS) magnets and CFS is developing a compact tokamak called SPARC that also will feature high-temperature superconductors.

PPPL, on Princeton University's Forrestal Campus in Plainsboro, N.J., is devoted to creating new knowledge about the physics of plasmas — ultra-hot, charged gases — and to developing practical solutions for the creation of fusion energy. The Laboratory is managed by the University for the U.S. Department of Energy’s Office of Science, which is the single largest supporter of basic research in the physical sciences in the United States and is working to address some of the most pressing challenges of our time. For more information, visit