Google's Nuclear Fusion Project Is Paying Off

[frigidmagi]

popular mechanics

Nuclear fusion, the process the sun has used for billions of years to fuse atoms of hydrogen into atoms of helium, could be the pot of gold at the end of the clean energy rainbow. If we could engineer a reaction to snowball but remain contained, nuclear fusion reactors could supply virtually unlimited clean energy here on Earth. Yet, the technology seems perpetually just around the corner.

Google and nuclear fusion company Tri Alpha Energy, which operates fusion reactor projects in California, just took us one step closer to rounding that corner. The two companies began working together in 2014, and they just released their first major research results. Google and Tri Alpha Energy developed a new process to sift through the enormous amounts of data that detail plasma's behavior in fusion reactors. The process involves humans who input preferences into an advanced Google machine learning algorithm, and so far the system has successfully achieved a 50 percent reduction in energy loss. The results were recently published in the journal Scientific Reports.

Tri Alpha's plasma generators use magnetic confinement, meaning they trap the plasma that is to undergo fusion using a magnetic field, but it is unique from other magnetic confinement reactors such as tokamak reactors. The Tri Alpha reactors use what is known as a field-reversed configuration, which takes advantage of eddy currents in the plasma itself to reverse the magnetic field, rather than relying entirely on external magnetic coils on the machine. The result is a self-stabilized, rotating cylinder of particles held in place by magnetism, similar in structure to a smoke ring. The major advantage to this technique is that as the energy of the plasma grows higher, the magnetic confinement gets stronger and more stable as a response.

Creating this magnetic field-reversed configuration and maintaining it by injecting protons, electrons, and boron fuel into the reactor is incredibly complicated. The number of variables is almost endless, which is why Tri Alpha Energy looked to Google for computing help in the first place. But even Google's renowned supercomputers couldn't handle the job.

"The reality is much more complicated," said Ted Baltz of Google's Accelerated Science Team. "The ion temperature is three times larger than the electron temperature, so the plasma is far out of thermal equilibrium. Also, the fluid approximation is totally invalid, so you have to track at least some of the trillion plus individual particles, so the whole thing is beyond what we know how to do even with Google-scale computer resources."



The solution was to input some human deduction back into the problem, something the team is calling an "Optometrist Algorithm." The number of variables to account for is simply too high, so human plasma technicians told the computer what specific behaviors to look for. If the particles are acting weird in a specific way, or the magnetic field at large is losing strength, the computer can be programed to sift through the data and search for just the relevant causes.

"We boiled the problem down to 'let's find plasma behaviours that an expert human plasma physicist thinks are interesting, and let's not break the machine when we're doing it'," said Baltz. "This was a classic case of humans and computers doing a better job together than either could have separately."

To further the study of nuclear fusion even more, Tri Alpha Energy built a new, larger plasma generator. The machine, called "Norman" after the late co-founder of Tri Alpha Energy, Norman Rostoker, heats plasma with a beam of high-energy neutral particles, a process known as neutral beam injection, and traps the plasma with a field-reversed configuration.

A 50 percent reduction in energy loss is huge, and the improved magnetic system generates more ion heat and plasma energy to begin with as well. We can trap plasma and spark nuclear fusion in a lab today, but the reaction requires more energy input than it releases. With advances in efficiency from data scouring processes like Google and Tri Alpha's "Optometrist Algorithms," it's possible we actually see an energy-positive fusion reactor constructed in the coming years.

How expensive the infrastructure will be, and whether or not nuclear fusion will usher in a brave new world of unlimited power, is yet to be seen.

nature article