Scientific Achievement
Foundry staff and users developed a machine-learning algorithm based on the principles of evolution and used it in molecular simulations to self-assemble materials with user-defined properties.
Significance and Impact
This study shows the potential of an evolutionary approach to materials design, furthering progress toward automated materials discovery or “synthesis by design”.
Research Details
- Optimizing the synthesis of a given material is a challenging problem that has previously required significant human input and trial and error.
- The research team developed an AI that can design both the molecular-scale building blocks and the time-dependent protocols required for their self-assembly. All the human user needs to do is specify the desired material and the AI uses simulated evolution to satisfy this goal.
- The team demonstrated the assembly of some of the more complex planar structures reported in the literature using neural networks that control only a small handful of system variables. The ability of neural networks to deal with many variables suggests that this approach can be used to address problems of considerable complexity.
