In a collaboration aimed at overcoming one of the toughest hurdles in clean energy research, Google Research and the private fusion company Tri Alpha Energy have developed a computer algorithm that dramatically accelerates experiments on plasma, the superheated gas at the heart of nuclear fusion reactions. The algorithm, named the Optometrist, was tested on Tri Alpha Energy's C2-U machine, where it compressed a month's worth of experimental work into a few hours and cut system-induced energy losses by half.
The findings, published in the journal Scientific Reports, represent a practical step in the long-running quest to harness fusion—the process that powers the Sun—as a viable energy source on Earth. While fusion has long been hailed for its potential to produce abundant clean energy with minimal radioactive waste, sustaining a controlled reaction has proven extraordinarily difficult, largely because of the challenge of manipulating plasma, which must be held at temperatures exceeding 30 million degrees Celsius.
“The whole thing is beyond what we know how to do even with Google-scale computer resources,” said Ted Baltz, a senior software engineer on Google’s Accelerated Science Team, in a blog post describing the work. He explained that the team boiled the problem down to a simple goal: “Let’s find plasma behaviors that an expert human plasma physicist thinks are interesting, and let’s not break the machine when we’re doing it.” The approach, he added, was “a classic case of humans and computers doing a better job together than either could have separately.”
The Optometrist algorithm works by running computational models in parallel with human experiments, allowing researchers to explore a wide range of plasma configurations without the risk of damaging equipment. By identifying promising behaviors quickly, the algorithm shortens the time needed to test hypotheses, a bottleneck that has slowed fusion research for decades.
In the C2-U experiments, the algorithm not only reduced experiment time but also improved plasma energy retention. Baltz noted that the system achieved a 50 percent reduction in energy losses, which increased the total energy of the plasma—though only for about two milliseconds. “It was only for about two milliseconds, but still, it was a first!” he wrote.
Why Fusion Research Is Gaining Momentum
Fusion has attracted growing attention from governments, private investors, and research institutions because of its potential as a nearly limitless and clean energy source. A single fusion event can release about 200 MeV of energy—roughly four times the energy produced by a fission event—making it an attractive alternative to conventional nuclear power. The promise of abundant energy without long-lived radioactive waste has led some to call fusion the “holy grail” of energy research.
However, achieving a stable, self-sustaining fusion reaction remains elusive. Researchers are exploring multiple avenues, including building more advanced reactors and experimenting with alternative plasma fuels. For instance, physicists at the Princeton Plasma Physics Laboratory have been investigating the use of argon-based plasma instead of the more common hydrogen, deuterium, or helium.
Google’s algorithm could accelerate these efforts by significantly reducing the time required for each experiment. “Results like this might take years to solve without the power of advanced computation,” Baltz said.
Next Steps for Tri Alpha Energy
Tri Alpha Energy has already moved beyond the C2-U machine, having replaced it with a more advanced device called Norman, which achieved its first plasma earlier this month. The company plans to build a demonstration power generator if further experiments with Norman prove successful.
While the two-millisecond plasma burst is a far cry from a working fusion reactor, the collaboration demonstrates how machine learning and human expertise can combine to tackle complex scientific challenges. As fusion research continues to evolve, such computational tools may become indispensable in the race to unlock a clean energy future.
