Firefly and PC GAMESS-related discussion club

Learn how to ask questions correctly  
We are NATO-free zone

Re^7: Orbitals sorting with firefly

Alex Granovsky

Dear Alex,

I'm sorry for delay with my reply.

Normally, I would recommend use of ASIS=.t. only together with the
extended orbital printout:

 $contrl wide=.t. $end

I.e. the first run should be run with wide=.t. and will punch
orbitals using extended (full) precision. The second run,
also with wide=.t., will read in orbitals in full precision.

Otherwise, ASIS=.t. is potentially dangerous as (inexact) orbitals
are not exactly orthogonal and will not be orthogonalized.

Kind regards,
Alex Granovsky

On Tue Oct 7 '14 8:20pm, alex wrote
>Dear Alex,
>Your comment helps a lot (specially the info about 2 runs; I am surprising if this important detail is available beyond of your comment, it might be a good idea to update the corresponding section of the manual). Unfortunately I had been trying to run JORDER/IORDER as a one run and that didn't work (evidently now). Currently I am using orbitals from NBO module with IORDER only and ASIS=.T. The convergence rate is high (after gathering enough statistics I'll update my old criticism for NBO) but may I ask you a question to be on a safe side - is the approach OK or there are things I had missed?
>On Sat Oct 4 '14 6:51pm, Alex Granovsky wrote
>>Dear Alex,

>>   As to JORDER, the thing one must remember is that if one moves all
>>active orbitals to the beginning of the orbital list using JORDER
>>instructions, one cannot avoid moving some occupied inactive orbitals
>>beyond the limits of the initial list of inactive + active orbitals.
>>This is because there are some "unoccupied" active MOs at the
>>beginning of the list. The best one can do in this situation is to
>>move these "extra" orbitals to the very beginning of the list of
>>virtual orbitals.

>>   This means that at the initial MCSCF iteration one now need to use
>>the extended active space to include extra orbitals and to set MAXIT
>>to 1, presumably with FOCAS converger turned on. One can reduce the
>>maximum possible level of excitation though.

>>   After completion of the initial MCSCF iteration, the natural
>>orbitals will be computed and punched. Note, as a result of orbital
>>manipulations there will be some natural orbitals having population
>>very close to zero. One now need to restore the initial active space
>>in $det or $drt as well as exclude these extra orbitals from the
>>list of active orbitals by picking up NOs from the punch file and
>>reading them in using proper $guess IORDER reordering instructions.

>>Hope this helps.

>>Kind regards,
>>Alex Granovsky

[ Previous ] [ Next ] [ Index ]           Thu Oct 16 '14 8:21pm
[ Reply ] [ Edit ] [ Delete ]           This message read 397 times