A ML–DFT-based synergy to design new phenazine-based photovoltaic polymer designs with lowest exciton binding energies

Abstract

This study pioneers a machine learning-guided quantum design strategy to engineer phenazine-based photovoltaic polymers with record optimized exciton binding energies (E_b) 0.33–0.42 eV, a critical breakthrough for enhancing charge separation in organic solar cells. The E_b is defined and computed as the difference between the fundamental gap (energy of the lowest charge transfer state) and the optical gap (energy of the first excited singlet state), obtained from time-dependent DFT calculations. Leveraging DFT calculations on 618 polymers, we extracted 210 molecular descriptors and trained predictive models, identifying random forest as the optimal predictor (R² = 0.94, RMSE = 0.006). Crucially, feature importance analysis revealed electron mobility and BertzCT connectivity as dominant E_b regulators—insights impossible through traditional screening. We then designed novel 2,2’-((2Z,2′Z)-((4,4,9,9-tetrahexyl-4,9-dihydro-s-indaceno[1,2-b:5,6-b’]dithiophene-2,7-diyl)bis(methanylylidene)) bis (3-oxo-2,3-dihydro-1H-indene-2,1-diylidene)) dimalononitrile (IDIC)-derived chromophores by integrating ML-prioritized acceptors. Frontier orbital analysis confirmed enhanced charge separation, while photovoltaic characterization demonstrated unprecedented performance in chromophore 4 (V_oc = 1.71V, J_sc = 35.02 mA/cm², LHE = 95%). The achieved sub-0.01 eV E_b range is significantly lower than the typical 0.3–1.0 eV range reported for high-performance OPV polymers, marking a substantial advance. This work establishes a new paradigm for the rapid discovery of high-efficiency OPV materials, directly addressing the voltage/current trade-off in next-generation solar technologies.