Re: Are there restrictions to the number of active/inactive orbitals in GUGA?


[ Molcas user's WWWBoard ]

Posted by Per Ake Malmqvist on September 09, 2008 at 15:39:21:

In Reply to: Are there restrictions to the number of active/inactive orbitals in GUGA? posted by Evgeniy on July 01, 2008 at 18:01:43:

Dear Evgeniy,
Sorry for the late answer!
There is a limit to the number of orbitals, but this is not the reason that your
calculation failed.

First of all, GUGA is concerned only with the correlated electrons, and only
with the inactive and active orbitals. The number of virtual orbitals becomes interesting
only in a subsequent CI or CPF calculation. Your calculation askes to correlate 84
electrons, which are in a reference space of 41 inactive and 2 active orbitals. The
calculation is not really large. This gives rise to 'internal' configurations: Valence states(946), doublet-coupled singles(3528), etc, altogether a little more than 9000.

The reason that the calculation failes is just one single array, unfortunately not
dynamically allocated, which was dimensioned for 14000 items, but now needed almost the
double. You can increase the parameter MXCASE to 28000 (say) in integ.inc, and the
calculation will pass after recompilation. I will make a similar patch, but will try and
see if I can make the array dynamical.

The code is very old, and is difficult to maintain properly. Also, it is not written
to be very effective with large incative spaces: The so-called internal coupling
coefficients are computed and put on a file, to be used by the MRCI or MR-ACPF: this file
is (in your case) about 730 MBytes. However, it is also a matter of whether a calculation such as yours is very meaningful. The size consistency errors will be large, and the correlation contribution from external excitations will be very modest given the small basis set.

Best regards,

NUMBER OF VALENCE STATES 946
NUMBER OF DOUBLET COUPLED SINGLES 3528
NUMBER OF TRIPLET COUPLED DOUBLES 2542
NUMBER OF SINGLET COUPLED DOUBLES 2666
: Dear All!

: First, I'd like to thank Luis Serrano for replying in detail to my previous post (about CASPT2). I might return to the CASPT2 problems in next poster but now I'd like to ask about the performance of GUGA. Namely I wonder whether there are some restrictions to the number of active and correlated (inactive) orbitals in GUGA. More precisely I would like to perform CISD for a 84 electron system and I am interested in the two lowest states, S0 and S1. Unfortunately my input (given below) does't work, crashing at the GUGA step. So, I wonder whether this is due to internal limitations or there is some error in the input. Many thanks!

: Best,
: Evgeniy

: &SEWARD &END
: Title
: pCTA(H2O)2 Alpha twisted structure
: coord
: geom.xyz
: basis
: sto-3g
: group
: c1
: End of input
: &SCF &END
: Charge
: -1
: End of input
: &MOTRA &END
: Frozen
: 19
: End of input
: &GUGA &END
: Electrons
: 84
: Spin
: 1
: Active
: 2
: Inactive
: 41
: End of input




Follow Ups:



Post a Followup

Name:
E-Mail:

Subject:

if B is 1s22s22p1, what is Li?

Passfield:

Comments:


[ Follow Ups ] [ Post Followup ] [ Molcas user's WWWBoard ]