| 2:41:20 AM PDT - Tue, Jul 29th 2014 |
|
I would guess that different programs become more sensitive to implementation choices under "difficult" conditions, meaning anything other than neutral closed shell systems near equilibrium geometries that are well-described by a single reference. A 0.1 hartree difference is indeed large. It is good to hear that you have at least one calculation where the HF energies are the same, because that clears up worries about basis sets, at least for that calculation. I would encourage you to look at the calculations where HF energies are the same more closely, to figure out where results begin to significantly diverge between programs. I would also wonder about calculations where the HF energies are not close. That will be a problem if you want to continue using both programs for a single line of research.
If you want to try to compare Molpro and NWChem results against a third program, Psi4 is pretty capable and well-documented: https://github.com/psi4/psi4public
I don't know if that well help resolve your puzzle or only deepen it, though.
|