Slow RASPT2 convergence on CASPT2 energies

Subject:

if B is 1s22s22p1, what is Li?

Passfield:

Comments:
: Hello all, : I have been comparing RASPT2 and CASPT2 excitation energies for free base porphyrin and regular metalloporphyrins (ZnP, MgP). I notice that the convergence of RASPT2 values on those obtained with CASPT2 is quite slow. I am using an active space of 16 electrons in 14 orbitals, of which 3 orbitals are in RAS1 and 4 orbitals in RAS2 (i.e. the RAS2 correlates 10 electrons in 7 orbitals). Whilst RASSCF/CASSCF excitation energies are in agreement to ~0.01 eV when allowing up to triple excitations between RAS1 and RAS3, the agreement between RASPT2 and CASPT2 is significantly worse, differing by ~0.2 eV. Increasing the highest order of allowed excitation improves things, but only slowly. Even when allowing up to quintuple excitations, excitation energies differ by around 0.1 eV. Is this to be expected? And if so, is it due to an implicit difference in the CASPT2 and RASPT2 methodologies? : Thanks and best wishes, : Andy Kerridge