The team noted that any candidate molecule had to meet a complex set of requirements to work. These include dissolving well in the battery’s existing electrolyte and participating in reactions without damaging the battery.
It had to also be highly compatible with various active materials and electrolytes. To this end, the team used machine learning to identify potential candidates by digitizing molecular properties and utilizing extensive organic chemistry, electrochemistry, and materials engineering datasets.
The result, CF3SO2Li, emerged as an ideal solution for the problem. It is relatively cheap, easy to make, and compatible with most mainstream batteries today.