Accurate prediction of band gaps in neutral heterocyclic conjugated polymers

Heterocyclic π-electron polymers such as polythiophene, polypyrrole, and polyfuran attract wide interest on both experimental and theoretical levels. While the optical properties of these materials are dominated by the band gap, no accurate, reliable computational method currently exists to predict the band gaps of large oligomers. Six computational methods, including ZINDO/CIS, ZINDO/RPA, HF/CIS, HF/RPA, TDDFT/ TDA, and TDDFT are compared here for a set of 60 structurally well-defined heterocyclic oligomers of varied structure. All six methods are compared using both AM1 semiempirical and B3LYP DFT predicted geometries. Among the methods, the semiempirical ZINDO/CIS method applied to DFT-predicted geometries affords the best agreement between computed and experimental band gaps, yielding an RMS error of 0.31 eV over the data set considered. Analysis of the computed band gaps provides a simple, straightforward empirical correction that significantly improves the accuracy of all six methods, with RMS errors between 0.23 eV and 0.44 eV for TDDFT using DFT-predicted geometries and for ZINDO/RPA using AM1-predicted geometries, respectively.