Accuracy of Spin-Component-Scaled CC2 Excitation Energies and Potential Energy Surfaces.

Attila Tajti,Péter G. Szalay

doi:10.1021/acs.jctc.9b00676

Attila Tajti, Péter G. Szalay

Open Access

https://doi.org/10.1021/acs.jctc.9b00676

Copy DOI

Abstract

Benchmark calculations with the Spin-Component-Scaled CC2 variants SCS-CC2 and SOS-CC2 are presented for the electronically excited valence and Rydberg states of small- and medium-sized molecules. Besides the vertical excitation energies and excited state gradients, the potential energy surfaces are also investigated via scans following the forces that act in the Franck-Condon region. The results are compared to the regular CC2 ones, as well as higher level methods CCSD, CCSD(T)(a)*, and CCSDT. The results indicate that a large fraction of the flaws of CC2 revealed by earlier studies disappears if spin-component scaling is employed. This makes these variants attractive alternatives of their unscaled counterparts, offering competitive accuracy of vertical excitation energies of both valence and Rydberg type states and reliable potential energy surfaces, while also maintaining a low-power-scaling computational cost with the system size.

Highlights

The application of quantum chemical methods for electronically excited states is subject to wide scientific interest due to their role in spectroscopy, biological processes, photovoltaics, and many other fields
Statistics on the error of the vertical excitation energies are presented in Table 1 for the different spin-component-scaled CC2 variants as well as for the regular CC2, CCSD(2), and CCSD methods, evaluated against the CC3 reference values. (The full set of results for the excitation energies is available in the Supporting Information.) The latter two doubles methods, as already shown in ref 16, have a general tendency to slightly overestimate the vertical excitation energy of both valence and Rydberg states
The statistical findings on the accuracy of vertical excitation energies confirm that the valence−Rydberg misbalance of CC2 could be reduced by spin-component scaling, the biggest improvement in this aspect is seen when setting Css to zero

Summary

Introduction

The application of quantum chemical methods for electronically excited states is subject to wide scientific interest due to their role in spectroscopy, biological processes, photovoltaics, and many other fields. As the performance of such approximate methods is highly inconsistent, many of them have been extensively benchmarked in the past years by our group[14−17] and others,[18−26] addressing the quantification of errors for each method for vertical excitation energies, transition moments, and potential energy surfaces. In a recent study by our group,[15] the popular CC2 method[7] was found to perform very badly for Rydbergtype excited states, despite being accurate for the valence-type ones. We found[17] that the accuracy observable for the vertical excitation energies of valence states does not extend to the associated surfaces and energy derivatives either. The CC2 level excited state gradients and surface plots that follow these forces showed a surprisingly bad performance when compared to high-level reference data for many states where otherwise a very good vertical excitation energy is obtained

Objectives

Methods

Results

Conclusion