Automated molecule design through machine learning helps scientists identify and synthesize a new polymer electrolyte for lithium-ion batteries.
Machine learning is becoming increasingly ubiquitous in our quest to solve the scientific problems of the day. From protein folding to image editing, machine learning is a well-established tool for making new discoveries.
One such area that would benefit from material discovery is battery technology. Machine learning has been applied to predict the performance of rechargeable batteries, but was limited in dataset size and the model was unsupervised. And other challenges remain.
Now, in an article recently published in Fast macromolecular communicationsresearchers at Waseda University, Tokyo, in collaboration with Fujitsu, have used machine learning to find candidate polymeric materials for Li+ batteries much more efficiently.
“AI can predict new material structures with preferable parameters,” said Kan Hatakeyama-Sato, a researcher at Waseda University’s Department of Applied Chemistry and first author of the paper. “However, most of the candidate structures are known to be junk. They do not meet the criteria for practical applications, such as ease of synthesis, stability and processability.”
The authors argue that using AI alone to evaluate candidate materials is difficult because the AI does not have the tacit knowledge that experts do in terms of what the preferable parameters are for the material.
AI with the help of experts
Hatakeyama-Sato and the team looked at some limitations to automating the discovery process for new battery materials and thought about how they might be overcome with a little expert help.
“Instead of teaching each material scanning criteria to the AI,” said Hatakeyama-Sato, “the system was trained to learn what practical materials look like. AI learned its essential features by inputting the molecular structures of existing materials (unsupervised machine learning). Inspired by existing species, our system could create unprecedented material structures with superior performance. This process is somewhat like the design of new materials by expert researchers, using the vast experience of materials. [that they have].”
Once the machine presented its list of candidate materials, these were then screened using a new piece of hardware called digital annealing, a powerful computer system that allows researchers to efficiently screen candidate materials for suitability as battery materials.
“The AI system designed a new ion-conducting polymer,” said Hatakeyama-Sato. “The material offered good conductivity while maintaining solid-state electrolyte standards such as chemical stability, thermal stability, mechanical strength and adequate solubility.”
Future challenges for machine learning
Materials science would greatly benefit from more efficient and sophisticated automation of the structure design process, particularly for organic materials. The researchers believe their system brings us one step closer to this goal. The next step could be for the AI to tell us not only the ideal materials for the final product, but also the synthetic pathways needed to make them.
“Organic materials are difficult to design computationally due to their complex structures,” Hatakeyama-Sato concluded. “However, as these results demonstrate, new algorithms and hardware are gradually opening paths towards fully automated material design.”
Reference: Kan Hatakeyama-Sato et al., ‘Automated Design of Li+-Conductive polymer by annealing inspired by Quantum‘, Rapid Macromolecular Communications (2022) DOI: 10.1002/marc.202200385
Image Credit: Tyler Lastovich on Unsplash
