Design of molecular circuits from simulations and machine learning

This aim of this project is to develop high-throughput methods in order to design efficient single molecule circuits. This highly automated procedure will generalize the search for molecular conductors based on physical or chemical intuition to a data-driven approach. The project will develop and implement efficient methods to calculate single molecule conductance in metal-molecule-metal junctions for thousands of junction geometries. Then, machine learning methods will be used to analyze which parameters in the molecular geometry or electronic structure have the largest effect on conductance. Based on these results, new candidate molecules will be proposed by automatic substitutions of molecular components, and the conductance of these new candidate molecules will be calculated. This virtual screening of molecular materials is the first data-driven approach to molecular conductance and will provide design rules for single molecule circuits of tailored functionality.