[molpro-user] Default virtual orbitals used in Molpro

[Gerald Knizia]

On Monday 11 October 2010 11:52, Stephen Harrison wrote:
> I'm running a simple RHF calculation and am using the Occupied + a number
> of virtual orbitals for use in a further calculation in a different code.
> I just wondered what the default method for producing the virtual orbitals
> is? (as this will effect my further calculations). Are they simply the
> output of the Fock matrix? or are they something else such as MVO's?

No, nothing special. These should be the MOs diagonalizing the Fock matrix.

> I ask this as the coefficients for the virtual orbitals are different to
> when I use another code to produce RHF results (and yet the occupied
> orbital co-efficients match very well indeed).

There are two issues to consider:
1) You might need to increase the accuracy of the HF calculation to get equal 
numbers from both programs. For SCF, the energy depends only quadratically on 
the errors of the orbitals (and only on the occupied orbitals, not the 
virtual ones). Thus, typical default thresholds for the energy convergence 
(say, order 1e-8) will give you rather inaccurate MO coefficients (say, order 
1e-4).

2) In open-shell calculations there is considerable freedom in the choice of 
the Fock matrix definition. This might lead to different virtual orbitals and 
different orbital energies in different programs.
-- 
Gerald Knizia