Making Materials from Scratch
Robots Get Creative in the Lab
Estimated Reading Time: 3 minutes
For the first time ever, a team of materials scientists and engineers have developed autonomous robots that can discover and create completely new materials on their own. Dubbed the "A-Lab," these crafty robots designed and successfully produced 41 new solid materials in just over two weeks without any human intervention after being programmed.
This breakthrough technology mimics how human researchers leverage experience, data, and trial-and-error to select the raw ingredients and steps to produce novel, usable materials. By removing the human element, the A-Lab system developed by scientists at Lawrence Berkeley National Laboratory accelerated the typical pace of new material creation from months or years down to creating over two per day on average.
Learning from the Past to Inform the Future The ingenious programming that drives the autonomous robots draws on both theoretical physics-based simulations containing data for over 290,000 potential but undiscovered compounds as well as text analysis of 33,000 published studies on synthesizing solid-state materials like alloys. This hybrid data- and literature-driven approach attempts to predict which raw materials and chemical reactions are most promising for trying to produce a particular new material.
Over 17 consecutive days in 2023, the tireless A-Lab robots assessed 58 predicted but previously unmade inorganic materials composed of earth-abundant and non-toxic metals paired with oxygen alongside common ingredients like phosphorus, silicon, or sulfur. They successfully produced 41 that ranged from modified forms of nickel oxide to iron-based silicon-oxygen compounds in the form of stable solid powders ready for further research.
Practice makes perfect, but it was not smooth sailing the whole way—initially only 13 out of the first 26 attempts yielded a completely new material. Much like a chemist honing a difficult new recipe after failed batches, the hands-off robotic system amended ingredients and processing steps by continuously updating its material database and reactions list to achieve a 79% eventual success rate.
When the standard heating method came up short, the autonomous lab applied specialized techniques gleaned from past discoveries that overcame challenges with ingredients evaporating prematurely or products ending amorphous and uncrystallized rather than nicely structured solids.
Other research kinks arose from limitations in state-of-the-art physics simulations, which struggle to perfectly model certain elements like lanthanides with complex electron behaviors. But the study authors note that even identifying inaccuracies helps improve computational methods for future predictions.
Accelerating Discovery with Tireless Robots
This breakthrough provides a template for vastly speeding up the discovery of novel, usable materials by removing resource-intensive and slow human effort from key parts of the innovation process.
Immediate applications could expand families of low-cost metal oxides for affordable battery components in electric vehicles and efficient solar-energy converters. And exercising the knowledge learned by artificial intelligence software promises to unlock even more new materials at staggering rates.
The autonomous approach allows a small team of scientists to shepherd material innovation that once required years of grueling experimentation by dozens of graduate students and postdoctoral fellows. Taking humans out of the critical thinking also reduces costly errors and helps focus innovation targets on sustainable products with positive societal impacts.
So while the prospect of clever robots toiling away on inventing and producing new stuff with no human supervision may seem a bit scary, it promises to solve global challenges like climate change faster than was ever conceived possible.
Source Papers:
Merchant, A., Batzner, S., Schoenholz, S.S. et al. Scaling deep learning for materials discovery. Nature 624, 80–85 (2023). https://doi.org/10.1038/s41586-023-06735-9 https://www.nature.com/articles/s41586-023-06735-9
Szymanski, N.J., Rendy, B., Fei, Y. et al. An autonomous laboratory for the accelerated synthesis of novel materials. Nature 624, 86–91 (2023). https://doi.org/10.1038/s41586-023-06734-w https://www.nature.com/articles/s41586-023-06734-w