Internally contracted multireference coupled-cluster (icMRCC) theory is extended to the computation of first-order properties (expectation values). We use the previously defined Lagrange formulation of the energy functional to derive the required equations for the Lagrange multipliers and arrive at an expression for first-order properties according to the generalized Hellmann-Feynman theorem, analogous to single-reference coupled-cluster theory. The present formulation does not include orbital relaxation, but in line with previous experience in coupled-cluster theory, the single-excitation cluster operator can recover a significant portion of orbital relaxation. Further aspects of the theory that arise from the internal contraction approach are discussed. Using automated derivation techniques, we have implemented a pilot code for icMRCCSD and icMRCCSDT for testing the method numerically. We find good agreement with full configuration interaction for several properties of boron monohydride and dipole moment curves of hydrogen fluoride and chromium hydride. A particular focus is given to spin-dependent properties: The hyperfine coupling tensors of Σ and Π radicals have been computed and compared to experiment and previous computations. We discuss the problem of describing spin polarization with properly spin-adapted wavefunctions, which requires either including pseudo-triple excitations or employing sufficiently flexible reference functions.
REFERENCES
There is some confusion in the literature regarding the occurrence of the fine structure constant in the expression for PK, as this depends on the chosen unit system. In true Gaussian units, one factor α is absorbed in each of the Bohr magneton and the nuclear magneton, respectively. Atomic units, however, are based on the electrostatic system of units (esu), so the α2 should appear explicitly.
A subsequent discussion with one of the authors of Ref. 58 (D. D.) revealed that upon decontraction one should remove the extra s functions of the core-valence correlation set. This was not done in the present work (the impact on the results is small) but is recommended in general as it serves for numerical stability.