Benchmarking Coupled Cluster Methods on Valence Singlet Excited States

Abstract

In this paper, benchmark results are presented on Coupled Cluster calculation of singlet excitation energies and the corresponding oscillator strength. The test set of Thiel et al. (Schreiber, M.; Silva, M. R. J.; Sauer, S. P. A.; Thiel, W. J. Chem. Phys. 2008, 128, 134110) has been used, and the earlier results have been extended by CC3 oscillator strength for the whole set, CC3 excitation energies for larger molecules, and CCSDT results for some small molecules. Accuracy of the members of the hierarchy CC2-CCSD-CC3-CCSDT has been analyzed. The results show that both CC2 and CCSD are quite accurate and the difference to CC3 excitations energies is typically not larger than 0.2-0.3 eV. While the mean deviation of the CC2 results is close to zero, CCSD systematically overshoots the CC3 results by about 0.2 eV. The standard deviation is, however, somewhat smaller for CCSD, that is, the latter method provides more systematic results. Still, only a few cases could be identified were the absolute value of the error is over 0.3 eV in case of CC2. The results are even better for CCSD, with the exception of uracil, where surprisingly large error of the excitation energies have been found for two of the four lowest n-π* transitions. Both LR (Linear Response) and EOM (Equation of Motion) style oscillator strengths have been calculated. The former is more accurate at both CC2 and CCSD levels, but the difference between them is only 1-2% in case of CCSD. The error of the CC2 oscillator strength are substantially larger than that of CCSD but qualitatively still correct.