Impact of the choice of model spaces and basis sets on the performance of the valence-universal coupled-cluster method: Energies for Be and C2+

Abstract

To study the impact of the choices of complete model spaces (CMS's) and of orbital basis sets on the performance of valence-universal (VU) coupled-cluster (CC) methods, extensive calculations have been performed for Be and ${\mathrm{C}}^{2+}$ by means of the atomically oriented form of these methods accounting for one(S)- and two(D)-electron radial (R) excitations (VU-CCSD/R method). In addition to the commonly used complete model space defined by the orbital set (2s,2p) we have employed the CMS's defined by the (2s,3s) and (2s,2p,3s) sets. Despite the anticipated increase of intruder state difficulties, for the new CMS's, no problems have been encountered when solving the nonlinear VU-CCSD/R equations. The results for the total energies and energy differences disclosed a weak dependence on the CMS chosen. We have used four Slater-type orbitals basis sets partially balanced for an accurate description of the ground and excited states. It turned out that the energies of excited states strongly depend on the basis set used, which emphasizes the importance of developing methods for constructing well balanced basis sets. It has been demonstrated that the VU-CCSD/R method can be used for the accurate localization of the position of the 2${\mathit{p}}^{2}$ $^{1}$S resonance with respect to the ionization limit. Some attention is also paid to a singular behavior of the energy obtained with the linearized version of the VU-CCSD/R method.