Re: GUGA Errors


[ Molcas user's WWWBoard ]

Posted by pawel wielgus on December 21, 2005 at 17:44:46:

In Reply to: Re: GUGA Errors posted by Per Åke Malmqvist on December 21, 2005 at 13:55:30:

Thank you for the answer. I followed your suggestions, you were right, it helped. Anyway I still have a problem since I cannot reproduce Dunning's results. The reference energy by Dunning should be E=-75.777984. It was calculated with Molpro92 using the following procedure:
a)generate HF orbitals
b) use HF orbitals for MCSCF: use full valence active space (OCC,3,1,1,0,3,1,1,0), correlate all electrons (CORE,0,0,0,0,0,0,0,0), force 1sigma(g) and 1sigma(u) to be doubly occupied but optimize them (CLOSED,1,0,0,0,1,0,0,0)
c) use MCSCF orbitals for MRCI, but freeze 1sigma(g) and 1sigma(u) orbitals (OCC,3,1,1,0,3,1,1,0), (CORE,1,0,0,0,1,0,0,0), (CLOSED,1,0,0,0,1,0,0,0).

I was able to reproduce the cited energy hihgly accurately with the latest version of Molpro.
But my MOLCAS energy is unfortunately different: E=-75.780113, and, for me surprisingly, lower.
So one of the following happens: I don't understand something, I miss something in my input file, or Molcas uses different algorithm for MRCI.

That's the stage I would like to ask for help more experienced Molcas users...
Why I get different energy?

My input looks right now like the following:
(cut)
HF is OK, I checked the energy

&RASSCF &END
Title
C2 Dimer R=1.2520 BS=cc-pVTZ
OUTOrbitals
canonical
Symmetry
1
Spin
1
NActEl
8 0 0
Inactive
1 0 0 0 1 0 0 0
Ras2
2 1 1 0 2 1 1 0
LumOrb
End of input
*The above RAS part gives me a good result

&MOTRA &END
Title
C2 Dimer R=1.2520 BS=cc-pVTZ
JobIph
FROZen
1 0 0 0 1 0 0 0
End of input
*I freeze 1sigma(g) and 1sigma(u) on the stage of MOTRA

&GUGA &END
Title
C2 Dimer R=1.2520 BS=cc-pVTZ
Electrons
8
Spin
1
Symmetry
8
CIALl
1
Active
2 1 1 0 2 1 1 0
End of Input

&MRCI &END
Title
C2 Dimer R=1.2520 BS=cc-pVTZ
SDCI
NRROOTS
1
ROOTS
1
MAXIterations
40
End of input

: Hi,
: You are getting too many 'internal configurations', and the error message
: is very uninformative. But this should be easily fixed for you.

: GUGA/MRCI is a very old pair of programs, and the usage and input is a little
: bit different from most of our other programs. You probably do not want to
: correlate also the core electrons. Therefore, you should actually freeze the
: core orbitals already in the MOTRA, and the GUGA code -- which concerns itself
: solely with correlated electrons -- should correlate 8 electrons, and use the
: input active orbitals 21102110, or maybe inactive 10001000 and active 11101110.
: Note that the electrons, active, inactive etcetera inputs serve to define the
: *reference* only. Then, all single and double excitations out from this reference
: will comprise the actual CI space in the MRCI program.

: I do suspect that there may be also bugs in the code; I will certainly have a hard
: look at it *some time*, fixing up usability, documentation and messages, and I am
: thankful for any user reports; but I do believe that this change of input will solve
: your problem.

: Happy Christmas to all,
: PÅ




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 ]