We recently learned about the many types of MCSCF, which treat "static" electron correlation. There are then many ways to treat further "dynamic" electron correlation to try and bridge the gap between the MCSCF and FCI. But these methods can vary drastically:

  • Some require 4-body RDMs (e.g. NEVPT2), others only 3-RDMs (e.g. cummulant methods).
  • Some are size-consistent (e.g. NEVPT2), some are not (Davidson's +Q).
  • Some are available everywhere (Davidson's +Q), some are only in in-house codes (AC0/AC1).
  • ENPT2 can be parallelized w/ 100s of cores, NEVPT2 can be split into 8 parallel calculations.

I seek descriptions of all post-MCSCF dynamic correlation methods, including the following:

  • formula for how the energy and/or wavefunction is calculated,
  • N-Body RDM requirements,
  • size-consistency and size-extensivity (or lack thereof),
  • ability to Parallelize (or lack thereof),
  • whether or not spin purity is guaranteed,
  • invariance under orbital rotations (or lack thereof),
  • absence (or risk) of intruder states,
  • which software packages implement them.

Some post-MCSCF dynamic correlation methods I know include:

  • NEVPT (usually truncated at 2nd order: NEVPT2),
    • PC-NEVPT (partially contracted NEVPT),
    • SC-NEVPT (strongly contracted NEVPT),
  • Multi-reference Davidson correction (+Q),
  • Multi-reference Pople correction,
  • ENPT (Epstein-Nesbet Pertubration Theory, often used with SHCI),
  • Multi-reference Møller–Plesset Perturbation Theory,
  • CASPT / RASPT / GASPT (usually truncated at 2nd order: CASPT2 / RASPT2 / GASPT2),
  • AC0 and AC1 (Adiabatic Connection extrapolation schemes),
  • TCC (Tailored Coupled Cluster),
  • Externally Corrected Coupled Cluster,
  • MPCDFT (Multiconfigurational Pair Density Functional Theory) [potentially described here too].
  • MPSPT (matrix product state perturbation theory)

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.