A quantitative structure-property relationship study of the photovoltaic performance of phenothiazine dyes

•A structure-photovoltaic performance relationship model for phenothiazine dyes has been established.•Robust and predictive power models were obtained using partial least squares regression.•Vibrational frequency based eigenvalue descriptors demonstrated the best predictive ability.•Analysed features were found to correspond with existing knowledge about DSSCs.

Quantitative structure-property relationships (QSPR) are developed for a set of phenothiazine dyes used in dye sensitized solar cells (DSSCs). To this purpose, we have employed a number of structural descriptor variables ranging from representations based on the atom's environment to those that exploit semiempirical molecular orbital information. We show that robust linear statistical regression models can be obtained for important photovoltaic parameters such as the power conversion efficiency (PCE), the open-circuit voltage (Voc) and the short-circuit current (Jsc). Overall, models based on eigenvalue (EVA) descriptors derived from vibrational frequencies, demonstrated the best predictive ability for both the calibration and independent test sets. The high predictive power of these models accompanied by the ease of computation can pave the way for rapid screening of new potential dyes and computer-aided materials design.