Algebraic-diagrammatic construction scheme for the polarization propagator including ground-state coupled-cluster amplitudes. I. Excitation energies

Abstract

An ad hoc modification of the algebraic-diagrammatic construction (ADC) scheme for the polarization propagator is presented. Within this approach, all first-order Møller-Plesset correlation coefficients occurring in the second-order ADC secular matrix are replaced by amplitudes obtained from a coupled cluster doubles (CCD) calculation. This new hybrid method, denoted CCD-ADC(2), has been tested on a series of small diatomic and triatomic molecules and benchmarked with respect to Thiel's benchmark set of medium-sized organic molecules. For the latter, the calculation of 134 singlet and 71 triplet states has shown that CCD-ADC(2) exhibits a mean error and standard deviation of 0.15 ± 0.34 eV for singlet states and 0.0 ± 0.17 eV for triplet states with respect to the provided theoretical best estimates, whereas standard ADC(2) has a mean error and standard deviation of 0.22 ± 0.30 eV for singlet and 0.12 ± 0.16 eV for triplet states. The corresponding extended second-order schemes ADC(2)-x and CCD-ADC(2)-x revealed accuracies of -0.70 ± 0.32 eV and -0.76 ± 0.33 eV for singlet states and -0.55 ± 0.20 eV and -0.67 ± 0.22 eV for triplet states, respectively. Furthermore, the investigation of excited-state potential energy curves along the dissociation of the N 2 molecule has shown that the higher reliability of the ground-state CCD method as compared to MP2 is also inherent to the excited states. While the curves obtained at the ADC(2) level break down at around 2 Å, the ones obtained at CCD-ADC(2) remain reasonable up to about 3.5 Å.