Re: [cclib-devel] atombasis in ADF

[Adam T. <a-t...@st...>]

> I added a code snippet that parses the atom indices of BAS  
> functions into
> atombasis. That is probably not what is needed, though, since the  
> MOs in ADf
> are built from SFOs. Since SFOs can in general involve many atoms,  
> I'm not
> sure how this should be parsed?

 From the ADF manual (2004.01):

Create Runs:
BAS => (core-orthogonalization) => CBAS => (symmetry) =>  
(orthonormality) => LOW => (Fock diagonalization) => MO

Fragment mode:
FO (= MO from Fragment file) => (symmetry) => SFO => etc.

I take this to mean the following. During the create run, the BAS  
functions (primarily valence) are made orthogonal to core functions,  
which results in the CBAS functions. These are made symmetric and  
orgthogonal using the Lowdin approach. I believe at this point they  
are diagonal in the fock representation (as they are atoms). These  
(ie FOs) are then combined in the main calculation to make SFOs.

It's probably not trivial to go from SFO mocoeffs back to BAS  
functions because of this complicated procedure. I'm also not sure if  
we can assume, for example, that the primary '2 Px' BAS function is  
the same one that is referred to in the SFO section because of what  
goes on during the Create run. But this is all speculation, so maybe  
we should check it out in more detail.

Since the only reason we care about atombasis is generating molecular  
orbitals and density functions, it may make more sense to handle the  
TAPE* (checkpoint) files instead of trying to calculate this info  
from atombasis printed in logfiles. TAPE41, for instance, already  
contains electron density functions. If we decide to follow this  
route, we may want to consider also handling gaussian cube files, as  
well as any other volume data files.

Adam