MRCI thresholds or contractions
H. -J. Werner
werner at theochem.uni-stuttgart.de
Mon Nov 3 07:58:30 GMT 2003
The problem is due to the redundancy check in the internally contracted
mrci. Pairs are deleted if either the norm becomes very small (threshold
PNORM, default=1.d-8) or if an eigenvalue of the overlap matrix becomes
very small (threshold THRDLP, default=1.d-7). In the present case, the
norm of one pair becomes small at long distance. Using your input,
this can be avoided using
acpf;
thresh,pnorm=0
Still, the number of pairs changes at shorter distances from 24 to 25.
Should this still cause a non-negligible step in the potential, you can
try
acpf;
thresh,pnorm=0,thrdlp=1.d-8.
The problem with lowering this threshold is that you might get numerical
problems due to near linear-dependency (not in the present case). Even
though the number of pairs still changes at some distance, the energy step
is of the order of 0.1 microH only.
Another possibility is to fix the pair list using save/start
acpf;
save,5000.2;start,5000.2
thresh,pnorm=0,thrdlp=1.d-8
In this case you would have to determine the pair list (first calculation)
at very long distance and then compute the curve for smaller distances.
This should give a smooth potential (constant pair list), but some accuracy
may be lost at small distances because one pair is missing.
H.-J. Werner
On Fre, 31 Okt 2003, Fabio Mariotti wrote:
>
>
>
>
> Dear Molpro users,
>
> I have the following problem:
> I'm running an ACPF calculation on HF molecule
> close to dissociation:
> namely i'm computing 4 points. Here the distancies
> and ACPF energies:
>
> VRDFH EACPF
> 4.95 -100.0478137
> 5.00 -100.0478125
> 5.05 -100.0477949
> 5.10 -100.0477939
>
> as you can see from 5.00 to 5.05 there is a "jump"
>
> After a while I realised that the main difference
> in the two points (5.00/5.05) is the number of
> contractions in the MRCI/ACPF part.
>
> So for the 5.00 geometry I have:
>
> Total number of contracted configurations: 2588
> Total number of uncontracted configurations: 8642
>
> While for 5.05 geometry:
>
> Total number of contracted configurations: 2368
> Total number of uncontracted configurations: 8642
>
>
> For reference here is my current test input:
>===========================================================
>***,nh3 !A title
>!
>print,ref,pspace;
>gprint,basis,orbitals=8;
>!
>rdfh=1.00000000;
>geometry={angstrom;
>F;
>H, 1, rdfh;
>}
>!
>gthresh,oneint=1.0E-14,twoint=1.0E-14,zero=1.0E-14,preface=1.0E-20;
>!
>basis=aug-cc-pVDZ;
>!
>i=0;
>do ist=4.90,5.1,0.05;
>!
>i=i+1;
>rdfh=ist;
>vrdfh(i)=rdfh;
>!
>hf;accu,20;core;
>!
>casscf;
>acpf;
>eacpf(i)=energy;
>!
>enddo;
>!
>text,START TABLE FOR get_molpro_table.pl;
>table,vrdfh,eacpf;
>plot,FILE=plot.dat;
>text,END TABLE FOR get_molpro_table.pl;
>!
>---
>!
>===========================================================
>
>
>I've been playing around with all the cutoff/thresholds I know
>about but with no results!
>
>Does anybody knows how I can solve this problem?
>
>The CAS before the MRCI looks identical (beside the small
>changes due to the different geometries).
>
>Which other parameters the number of contractions can
>depend on?
>
>(I'm using molpro 2002.3 and 2002.6)
>
> Thanks in advance
> Fabio
>
>
>--
>=============================================
>Dr Fabio Mariotti - Tel.: +49 (0)208 306 2157
>Max Planck Institute - Kaiser-Wilhelm-Platz 1
>Muelheim an der Ruhr, 45470 Germany
--
Prof. Hans-Joachim Werner
Institute for Theoretical Chemistry
University of Stuttgart
Pfaffenwaldring 55
D-70569 Stuttgart, Germany
Tel.: (0049) 711 / 685 4400
Fax.: (0049) 711 / 685 4442
e-mail: werner at theochem.uni-stuttgart.de
More information about the Molpro-user
mailing list