[molpro-user] DF-LCCSD(T) problem: crash with larger basis set
Jörg Saßmannshausen
j.sassmannshausen at ucl.ac.uk
Tue Feb 12 13:34:35 GMT 2013
Hi Joachim,
thanks for the feedback.
Is there any way I could check whether or not the HF is converged to the
correct state? Maybe I miss something here but I thought as it is a closed
shell molecule I only have paired electrons and hence only one spin state?
I am not using Cr or a transition metal but Al so I don't really see how I
could converge to a different state? I know Gaussian has something like the
'stable' keyword which is testing whether or not the SCF is 'correct'. Is
there a similar feature in Molpro?
As for (ii): as I am currently doing a LCCSD calculation that possible problem
should be eliminated I hope.
Thanks for your sentiments!
All the best from London
Jörg
On Tuesday 12 February 2013 13:26:13 you wrote:
> I have no idea what's wrong, but possibilities are:
>
> (i) wrong state, perhaps HF converged to a wrong state and there is a lower
> one? (ii) insufficient fitting basis
>
> Other remark: Using DF-HF is probably much faster than direct HF, in
> particular for larger basis sets.
>
> Best regards
> Joachim Werner
>
> > Dear all,
> >
> > I am currently trying to perform some single point DF-LCCSD(T)
> > calculations on an organometallic compound.
> >
> > At the DF-LCCSD(T)/cc-pVDZ level of theory, the calculation runs smooth
> > without any problems and I get my single point energy.
> >
> > However, when I am trying to run the same calculation at the
> > DF-LCCSD(T)/cc- pVTZ level, the calculation crashes in the CCSD part:
> > Minimum memory for LCCSD: 38.40 MW, used: 200.00 MW, max: 389.90
> > MW
> >
> > ITER. SQ.NORM CORR.ENERGY TOTAL ENERGY ENERGY CHANGE
> > DEN1 VAR(S) VAR(P) DIIS TIME
> >
> > 1 2.00665126 -2.99296600 -941.60036694 0.05725124
> >
> > -2.92306557 0.11D-01 0.16D-01 1 1 4454.17
> >
> > 2 2.00663902 -5.22692889 -943.83432983 -2.23396289
> >
> > 0.00001279 0.53D-03 0.65D+00 2 2 4521.51
> >
> > 3 2.00603281 -5.22667989 -943.83408083 0.00024900
> >
> > 0.00001475 0.57D-03 0.65D+00 3 3 4585.54
> >
> > 4 2.00602946 -5.25966260 -943.86706354 -0.03298271
> >
> > 0.00002130 0.10D-02 0.66D+00 4 4 4649.57
> >
> > 5 2.00442950 -5.43585382 -944.04325475 -0.17619122
> >
> > 0.00008619 0.55D-03 0.71D+00 5 5 4713.17
> >
> > 6 2.00450932 -5.15168470 -943.75908563 0.28416912
> >
> > 0.00002367 0.10D-02 0.65D+00 6 6 4780.77
> >
> > 7 12750.36999404 -5.75382440 -944.36122534 -0.60213970
> >
> > 0.47490011 0.73D-03 0.84D+00 6 1 4850.68
> >
> > UNREASONABLE NORM. CALCULATION STOPPED
> >
> > Norm of t1 vector: 13.76970764 S-energy: 0.00001516
> > Norm of t2 vector: 112.07035801 P-energy: -2.99387560
> >
> > ?ERROR: NO CONVERGENCE. TRIPLES WILL NOT BE DONE. THIS CAN BE AVOIDED
> > USING THE NOCHECK OPTION.
> >
> > RESULTS
> > =======
> >
> > LCCSD singles energy -0.003213334121
> > LCCSD singlet pair energy -2.848042961511
> > LCCSD triplet pair energy -3.097983757676
> >
> > LCCSD strong pair energy -5.750611063722
> > LMP2 close pair energy -0.110540018435
> > LMP2 weak pair energy -0.082516374676
> > LMP2 distant pair energy -0.002359262354
> >
> >
> > Reference energy -938.607400937737
> >
> > LCCSD correlation energy -5.949240053308
> >
> > !LCCSD total energy -944.556640991046
> >
> > Program statistics:
> >
> > Available memory in ccsd: 198929165
> > Min. memory needed in ccsd: 38401396
> > Max. memory used in ccsd: 199999901
> >
> > ? Error
> > ? No convergence. This error exit can be avoided using the NOCHECK option
> > ? The problem occurs in ccsd_loc
> >
> > My input file is like that (minus the coordinates):
> > memory,200,m
> > SET,CHARGE=1 !charge of the molecule is +1
> > ! SET,ZSYMEL=NOSYM ! turn off symmetry
> > geomtyp=xyz
> > geometry
> >
> > 45
> >
> > [ ... ]
> > gdirect !direct calculations
> > hf,maxit=300
> > hf,direct
> > basis=cc-pVTZ !cc-pVTZ basis set for all elements
> > hf !Perform HF calculation
> > DF-lccsd(t) !Perform L-CCSD(T) calculation with density
> > fitting
> >
> >
> > I am currently trying to run a pure LCCSD with the same basis set but I
> > was running out of scratch space. :-(
> >
> > Is there something wrong with my input or is there a 'problem' with the
> > molecule. The geometry has been calculated at the M11/6-311G(d,p) level
> > of theory and a frequency calculation indicates a minimum on the PES.
> > However, I got to admit, I am not really trusting the DFT calculations
> > here and hence the CCSD(T) calculation. It is a closed shell compound
> > with Al as the metal.
> >
> > Does anybody got an idea here?
> >
> > All the best from a cold London
> >
> > Jörg
--
*************************************************************
Jörg Saßmannshausen
University College London
Department of Chemistry
Gordon Street
London
WC1H 0AJ
email: j.sassmannshausen at ucl.ac.uk
web: http://sassy.formativ.net
Please avoid sending me Word or PowerPoint attachments.
See http://www.gnu.org/philosophy/no-word-attachments.html
More information about the Molpro-user
mailing list