Excited state ACPF

Peter Knowles P.J.Knowles at bham.ac.uk
Thu Aug 8 07:20:05 BST 2002

In principle, ACPF for excited states can be OK, but you should realise
that all the code can do is look for stationary values of the ACPF
energy functional as defined in Gdanitz and Ahlrichs' 1988 paper.  The
way that it does this gives some chance of success: a solution is sought
starting from the corresponding eigensolution of the hamiltonian matrix,
and thus should be close to that CI solution.  But there are no
guarantees or variational principle (except for the ground state), and
there is always the possibility of converging to the wrong root, or even
a stationary point that does not correspond to a physical wavefunction.
It is, of course, your responsibility to worry about this.

Sometimes, as you observe, the code is not robust enough to converge
accurately to a stationary point during the iteration process, usually
because of numerical ill conditioning. If it converges in the end, then
it has converged!


> -----Original Message-----
> From: owner-molpro-user at stchem.bham.ac.uk 
> [mailto:owner-molpro-user at stchem.bham.ac.uk] On Behalf Of Shai Ronen
> Sent: 04 August 2002 17:30
> To: molpro-user at stchem.bham.ac.uk
> Subject: Excited state ACPF
>  Hello,
>  I'm interested in doing ACPF calculations for multiple 
> states, that is, ACPF for a few roots following SA-MCSCF. 
> This is quite general so the details are not important (I 
> think). I had some problems with this before. Sometimes it 
> didn't converge (MRCI did), and I got a communication from 
> H.-J. Werner that it is not possible to do that. However more 
> recently, for my calculations of interest, although I 
> encounter some converegence warning about the higher roots 
> during intermediate iterations, the final energies seem to 
> converege normally and the program exits normally.  My 
> question is, in view of the above, are such results trustable?
>  Many thanks in advance,
>  Shai Ronen
>  Shai Ronen        sronen at post.tau.ac.il
>  School of Chemistry
>  Tel Aviv University
>  Tel Aviv 69978
>  Israel

Prof Peter J. Knowles
Email P.J.Knowles at bham.ac.uk  Phone +44-121-414-7472 Fax
School of Chemical Sciences, University of Birmingham, Birmingham B15

More information about the Molpro-user mailing list