Neuroevolutionary Learning of Particles and Protocols for Self-Assembly.

Area of Science:

• Computational chemistry
• Materials science
• Artificial intelligence

Background:

• Molecular self-assembly is crucial for materials design.
• Traditional methods often rely on physical principles like thermodynamics.
• Discovering novel self-assembled structures can be challenging.

Purpose of the Study:

• To demonstrate neuroevolutionary learning for designing self-assembling molecules.
• To explore the capability of AI in designing materials without physical constraints.
• To enable both targeted material properties and novel structure discovery.

Main Methods:

• Simulations of molecules deposited on a surface.
• Utilizing a neuroevolutionary learning algorithm.
• Implementing directed and exploratory design modes.

Main Results:

• Successfully designed particles and time-dependent protocols for self-assembly.
• Achieved self-assembly without input on thermal equilibrium or mechanical stability.
• Demonstrated AI's ability to explore beyond low-energy, kinetically inaccessible states.

Conclusions:

• Neuroevolutionary learning offers a powerful, knowledge-free approach to designing self-assembling materials.
• This method expands the possibilities for discovering novel material structures and properties.
• AI can effectively navigate complex design spaces for advanced materials.