Predicting with confidence the efficiency of new dyes in dye sensitized solar cells

We ask whether it is possible to predict the efficiency of a new dye in dye sensitized solar cells (DSSCs) on the basis of the known performance of existing dyes in the same type of device. We evaluate a number of computable predictors of the efficiency for a large set of dyes whose experimental efficiency is known. We have then used statistical regression methods to establish the relation between the predictors and the efficiency. Our predictions are associated with a rigorously determined confidence level. For a new dye of the same family we are able to predict the probability that its efficiency in a DSSC is larger than a certain threshold. This method is useful for accelerating the discovery of new dyes and establishing more rigorously the existence of specific correlations between structure and properties. Within the properties considered we find that the dye efficiency correlates more strongly with its oxidation potential and reorganization energy.