Re: [cclib-devel] CI and TDDFT

[Noel O'B. <bao...@gm...>]

I've just done the equivalent calculation with PCGAMESS and checked it
in. It's not only an order of magnitude faster (or more), but the
squares of the triplet contributions add to around 1. Is this a bug in
GAMESS-US then? Who knows? (Just to be clear, the TD-DFT code in
PCGAMESS and GAMESSUS is completely different)

Noel

On 15/11/2007, Karol Langner <kar...@kn...> wrote:
> On Wednesday 14 November 2007 11:37, Noel O'Boyle wrote:
> > I'm wondering whether either of you know whether the etsecs from a
> > restricted calculation of triplet excited states is likely to add to 1
> > (after squaring). I'm getting:
> >
> >    ---------------------------------
> >      R-TDDFT CALCULATION CONVERGED
> >    ---------------------------------
> >  NUMBER OF USED VECTORS / MAX =  49 / 200
> >  NUMBER OF ITERATIONS   / MAX =   6 / 100
> >  NUMBER OF SINGLE EXCITATIONS =   875
> >  NUMBER OF ATOMIC ORBITALS    =  60
> >  NUMBER OF TOTAL ELECTRONS    =          70.0035208
> >
> >           -------------------
> >           TRIPLET EXCITATIONS
> >           -------------------
> >
> >  STATE #   1  ENERGY =    3.027228 EV
> >  OSCILLATOR STRENGTH =    0.000000
> >         DRF    COEF       OCC      VIR
> >         ---    ----       ---      ---
> >          35 -1.105383     35  ->   36
> >          69 -0.389181     34  ->   37
> >         103 -0.405078     33  ->   38
> >         137  0.252485     32  ->   39
> >         168 -0.158406     28  ->   40
> >
> > The squares add to 1.63 or so.
>
> I've never seen a coefficient larger than one - I don't know what that means.
> Notice the oscilator strength is zero (or really very small).
>
> Can you attach the entire output file (i think only the singlet version is in
> teh trunk)? Also, how does this compare to Gaussian output? From my
> experience, GAMESS is buggy in some places when it comes to excitations. I've
> had some problems using soem of the CIS options.
>
> > It's not a big deal for users of cclib as we just extract whatever
> > figure is given here. Still, if anyone has any insight it would be
> > good to know. For comparison, here is a similar transition for the
> > SINGLET EXCITATIONS:
>
> True, although it would be good to know what's happening and maybe warn the
> user. I routinely use scripts to check output files for errors/warnings and
> generally strange output. This might be a good thing to add to cclib n the
> future.
>
> > > The handling of the etsecs is likely to have changed for CI due to my
> > > latest checkin. I hope this is not a problem. I am currently trying to
> > > standardise the values of etsecs for all calculations (using tests),
> > > but am only halfway. I need to carry out some TD-DFT calculations on
> > > unrestricted calculations and sort these out at the same time as
> > > handling the CI calculations.
>
> I'll have closer look over the weekend.
>
> Karol
>
> --
> written by Karol Langner
> Thu Nov 15 14:38:10 EST 2007
>