(435d) Camd.Jl: A Generalized Computer-Aided Molecular Design Framework

The increasing need for application-specific molecules has underscored the indispensable role of computer-aided molecular design (CAMD) methodologies. CAMD typically necessitates two integral components: a surrogate model for the prediction of pertinent thermodynamic properties and a molecular design toolkit. Although Machine Learning (ML) approaches have garnered considerable popularity in furnishing such surrogates, they are often constrained by their capacity to fulfill singular design objectives. Conversely, Clapeyron.jl, an advanced open-source thermophysical property estimation library, offers a notably generalized and predictive framework, capable of encompassing a wide array of properties relevant to molecular design endeavors. Capitalizing on the capabilities of this robust library, we introduce CAMD.jl, a CAMD framework adept at tackling diverse molecular design challenges of significance. In this talk, employing a graph-based genetic algorithm methodology, we showcase the versatility of CAMD.jl by its application in various contexts, including:

• Design of non-ozone-depleting refrigerants to satisfy particular operating conditions.
• The development of innovative surfactants aimed at minimizing interfacial tension between aqueous and oil phases.
• The formulation of battery electrolytes employing ionic liquids to optimize conductivity.

This framework holds a great deal of promise to be applicable to other important problems in the community such as carbon capture and pharmaceutical solvent-selection.