University of Alberta scientists have achieved a world’s first—they’ve employed a machine learning process using artificial intelligence to automate and refine atomic-scale manufacturing.
The process holds an array of efficiencies: it’s a greener, faster, and smaller technology. The development can satisfy modern day digital demands while reducing its influence on the climate.
“Most of us thought we’d never be able to automate atomic writing and editing, but stubborn persistence has paid off, and now Research Associate Moe Rashidi has done it,” says University of Alberta Professor of Physics Robert Wolkow, who along with Rashidi, published a paper on the research.
“Until now, we printed with atoms about as efficiently as medieval monks produced books,” says Wolkow. “For a long while, we have had the equivalent of a pen for writing with atoms, but we had to write manually. So we couldn’t mass produce atom-scale devices, and we couldn’t commercialize anything. Now that has all changed, much like the disruption following the arrival of the printing press for those medieval monks. Machine learning has automated the atom fabrication process, and an atom-scale manufacturing revolution is sure to follow.”
Wolkow has been on the hunt for climate-friendly, atomic-scale, low-power electronics during his career. With technology’s energy consumption and contributions to global carbon emissions, Wolkow felt the need to form a new basis for electronic designs through atomic-scale fabrication and mass manufacturing, which is now a possibility due to this recent breakthrough.
“Fabrication at the ultimate small scale not only lets us do things better, but we can also create entirely new functions that conventional technology simply cannot do. Combining that with a practical path to manufacturing will be game changing. This allows us to create a new, extremely efficient basis for computing using the natural properties of individual atoms,” Wolkow adds.
To learn more, read “Autonomous Scanning Probe Microscopy in Situ Tip Conditioning through Machine Learning,” published in the journal ACS Nano.
