(709a) Computational Molecular Engineering for Advanced Materials

In the dynamic domain of material science, the potential for fabricating advanced materials is vast, promising transformative advancements for sectors such as electronics, energy, and environmental science. These advanced materials, characterized by their significant diversity and adaptability, offer a virtually limitless array of structural variants by manipulating different molecular components. Exploring this extensive spectrum of potential structures necessitates employing cutting-edge computational methods to illuminate uncharted or overlooked regions of the chemical space and accelerate the discovery of materials with superior properties.

A particularly effective strategy to meet this need is the computational molecular engineering of materials. By virtue of advancements in molecular simulations and machine learning, this approach allows for the optimization and precise generation of structures targeted for specific applications. In our presentation, we will share our latest efforts in computational molecular engineering, specifically highlighting a machine learning strategy that combines Monte Carlo tree search with a recurrent neural network.

For practical demonstrations, we've applied our tool in virtual experiments to design 1) high-performance Metal-Organic Frameworks for methane storage and carbon capture, 2) Ionic Liquids with potential for flue gas (CO2/N2) and syngas (CO2/H2) separation, and 3) sustainable polyamides with enhanced properties. Our method has proven highly effective in the creation of promising materials.

Our approach can be easily adapted to other contexts by modifying the reward function to align with the desired performance property.