Many countries are now invested in the race for nuclear fusion, and they all hope to bring about an energy source that is both stable and low in carbon emissions. Recently, US-based TAE Technologies has leveraged Google’s machine learning application to shorten a two-month-long mission down to a few hours to accelerate the progress of research.

Nuclear fusion research is long and complex. Scientists have given maximum effort in constructing an energy equipment on earth that can rival a star. Such equipment would transform hydrogen and other relatively light atoms under high temperature and high pressure, through nuclear fusion, into heavier atomic nuclei.

Established in 1998, TAE received funding from the late Microsoft founder Paul Allen, the Rockefeller family, and Goldman Sachs. With an investment of more than US$880 million, TAE has developed technologies that are a drastic departure from existing nuclear fusion technologies used in donut-shaped tokamaks. TAE’s nuclear fusion equipment features 30 m long cylindrical shape and is called C2W “Norman”.

Nuclear fusion requires precise systems that can micro-adjust and control plasma millions of degrees Celsius. TAE CEO Michl Binderbauer indicates that TAE’s design is more advantageous than other nuclear fusion technologies. The company has currently been optimizing its nuclear fusion equipment thanks to Google’s machine learning applications. Generally, after updating the relevant hardware equipment, it takes as much as two months for optimization and calibration to wrap up, but all of this can be done within an afternoon via machine learning.

Binderbauer explains that machine learning can be used for either reproducing nuclear fusion reactions or sorting massive amounts of data. As this presents a compute-intensive challenge, few made the attempt in the past. Binderbauer credits the Google partnership with shortening TAE’s estimated timeframe to completion by one year and hopes to perform commercial nuclear fusion testing in 2030.

TAE’s nuclear fusion technology quite differs from that used in ITER’s tokamaks. Composed of deuterium and tritium, ITER’s fuel yields nuclear fusion energy that measures several tens of millions of degrees Celsius. However, not only is tritium limited in terms of availability, but it also emits radiation that can damage the internal equipment of the reactor. Norman, on the other hand, leverages more usual hydrogen and deuterium fuels, thereby making it a more stable, albeit less performant, choice. In the future, Norman will likely transition towards hydrogen-boron fuels as these materials generate neither neutrons nor radioactivity, in addition to making the reactor easier to maintain, despite a minor hike in operating temperature.

At the moment, Norman operates at a temperature of 70 million degrees Celsius, but the utilization of hydrogen-boron fuels will raise this temperature by 20-30 times, well into the billions. Regarding this, Binderbauer is optimistic.

(Image: TAE)