Under the assumption that the static correlation can well be described by selected configuration interaction, to the extent that the residual dynamic correlation can be accounted for accurately by a second-order perturbation theory (PT2), a family of multireference PT2s (iCI-MRPT2) based on the iterative configuration interaction with selection are proposed and compared. The first step is to partition a large complete active space (CAS) P into the direct sum of a primary (Pₘ) and a secondary (Pₛ) subspace via a parameter Cₘᵢₙ. The former serves as the actual reference space, whereas the latter is treated via Epstein-Nesbet PT2. An MRPT2 is then invoked to account for the dynamic correction of the $Q=1-P$ space to the Pₘ space. Different MRPT2s are distinguished by the underlying zeroth-order Hamiltonians. To circumvent the prohibitive computational cost and memory footprint associated with high-order reduced density matrices in the fully internally contracted scheme and the exponential growth of external functions (XF) in the uncontracted scheme, a partial contraction strategy is employed for constructing the XFs: those involving one or two inactive (core or virtual) orbitals are kept uncontracted, whereas those involving three or four inactive orbitals are internally contracted. They span the uncontracted (Qᵤ) and contracted (Qc) portions of Q, respectively. The former is further split into two subspaces Qᵤᶜ and Qᵤᵘ via a parameter Qₘᵢₙ, which are to be treated in coupled and uncoupled manners, respectively. Extrapolations to zero Qₘᵢₙ and Cₘᵢₙ can be performed, so as to approach CAS-based MRPT2. Calculations with active spaces as large as CAS(56e,58o) can now be performed. The size-consistency of the iCI-MRPT2s is also analyzed in detail.
No takes yet. Share an insight, caveat, or question.
Zhang et al. (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: