Abstract:
Density Functional Theory (DFT) has become increasingly popular for calculating linear and nonlinear optical (NLO) properties of various molecular materials in recent decades due to its favourable accuracy-to-computational-cost ratio. A critical comparison of different DFT functionals against experimental data or accurate wavefunction-based calculations is generally required to choose the best functional for specific systems or properties. Despite a plethora of DFT studies reporting the NLO calculations, such comparisons of different DFT method for calculating optical properties are sparse. Herein, the performance of sixteen different DFT methods (BLYP, BP86, PBEPBE, SVWN, M11L, B3LYP, PBE1PBE, M06, M062X, BHandHLYP, O3LYP, CAM-B3LYP, LC-BLYP, wB97X, LC-PBEPBE and M11) in the calculations of static first hyperpolarizabilities (βtot) and electronic excitations of p-nitroaniline (PNA), 4-amino-4ˈ-nitrobiphenyl (DANB), 4-amino-4ˈ-nitrodiphenylacetylene (DANA), 4-amino-4ˈ-nitrostilbene (DANS) and 4-amino-4ˈ-nitroazobenzene (DANAB) was assessed. The computed data were compared against the MP2 hyperpolarizabilities and experimental absorption maxima. For any molecule, the HF hyperpolarizability is less than the DFT data. However, the mean absolute error MAE (compared to MP2) of HF is much lower than that of pure and some hybrid methods, with pure functionals overestimating βtot for relatively large molecules. Hybrid functionals showed considerable improvement over the pure functionals, particularly two methods M062X and BHandHLYP. CAM-B3LYP was found to outperform the rest of DFT methods. The magnitudes of the computed hyperpolarizability strongly rely on the amount of HF exchange in the functional than correlation effects. A similar trend can be seen in the computed excitation energies. From MAEs, the long-range corrected functionals considerably improve the excitation energies compared to their pure counterparts, especially for large molecules. Again, the lowest MAE was obtained for CAM-B3LYP. All the hybrid methods are superior to the pure ones. The long-range corrections are important for calculating NLO properties and excitation energies of large push-pull organic molecules, with the lowest MAE given by CAM-B3LYP. While pure functionals overestimate the optical data, the incorporation of a fixed amount of HF exchange in the hybrid functionals improve the pure data significantly.
