Direct and Inverse Design of Functional Molecules
Driven by algorithm but steered by chemistry
Functional materials can have multiple key properties that intercorrelate with each other. To decorrelate the properties and to achieve efficient automated optimization, I have developed Molecular Genetic Algorithm (MolGA), a open-source Python code specialized at globally optimizing molecular systems.
The method has been applied to decorrelate intercorrelating molecular properties, such as the basicity and CO2 binding strength of CO2 capturing agents, and reduction potential and reaction energies of molecular electrocatalysts for CO2RR.
We are aware that the search suffers limitations of the predefined chemical space and the level of theory used. Hence, we always go back and look into the chemistry of the top-scoring candidates after the search converges, so as to understand the origin of the decorrelating behaviors and to propose straightforward designing principles.