Systematic Prediction of Dyes for Dye-Sensitized Solar Cells: Data Mining via Molecular Charge-Transfer Algorithms
Jacqueline M. Cole
Cavendish Laboratory, University of Cambridge
J. J. Thomson Avenue
Cambridge, CB3 0HE, UK
and
Department of Chemistry, University of New Brunswick
P.O. Box 4400
Fredericton, NB E3B 5A3, Canada
jmc61@cam.ac.uk
Abstract
Graph theoretical algorithms and classification tests are combined with quantum chemical calculations and data mining tools to present successful predictions of high-performance dyes for dye-sensitized solar cells (DSCs). The construction of molecular charge-transfer algorithms is described, featuring recursive depth-first, back-tracking, graph traversal algorithms with classification test formalisms. These algorithms are employed to search through a representative set of organic chemical space (120000 chemical molecules) to identify compounds that have the required structural attributes to act as high-performance dyes for DSCs. The first results of these predictions are validated by comparing predicted structural motifs to existing well-known dyes that are currently in use for DSC devices. Three chemical motifs are shown to form the chemical backbone of three popular dyes, thereby validating the predictions. Further work is described that includes the DSC fabrication and testing of the new classes of unknown dyes; this pertains to the ultimate goal of systematically designing new dyes for use in DSC devices.
