A Spin-Adapted Size-Extensive State-Specific Multi-Reference Perturbation Theory with Various Partitioning Schemes. II. Molecular Applications.

Shuneng Mao,Lan Cheng,Wenjian Liu,Debashis Mukherjee
DOI: https://doi.org/10.1063/1.3672085
IF: 4.304
2012-01-01
The Journal of Chemical Physics
Abstract:Following the theoretical development of a spin-adapted state-specific multi-reference second-order perturbation theory (SA-SSMRPT2) as expounded in the preceding publication, we discuss here its implementation and the results of its applications to potential energy curves (PECs) of various electronic states of small molecules. In particular, we illustrate its efficacy in states of various spin multiplicities and varying multi-reference character. Both Møller-Plesset (MP) and Epstein-Nesbet (EN) type of partitions have been explored. Also, a straightforward Rayleigh-Schrödinger (RS) and Brillouin-Wigner (BW) version of the SA-SSMRPT2 have been studied. Ground state PECs were computed for singlet states of HF, BH, and H(2)O molecules as well as the doublet state of NH(2) and BeH radicals and compared to corresponding full configuration interaction numbers, which serve as benchmark results. As an extensive application on a production level, the ground state PECs of N(2), a classic example of multiple-bond breaking, were calculated using cc-pVXZ (X = 3,4,5) basis and then extrapolated to obtain estimates of the complete basis set limit. Vibrational energy levels were extracted from these N(2) PECs, which compare favorably to the experimental values. In addition, extensive studies were also carried out on PECs of the seven low-lying excited states of the N(2) molecule. Finally, it is shown that the flexibility to relax configuration coefficients in SA-SSMRPT2 helps to provide good descriptions for the avoided crossing between the two lowest (1)Σ states of the LiF molecule. Our results indicate (1) that more studies are needed to draw firm conclusions about the relative efficacies of the MP and EN results and (2) that the RS version works so well as compared to the BW version that the extra computational expenses needed in the later formalism is not warranted.
What problem does this paper attempt to address?