× If you have problems during the execution of MRCC, please attach the output with an adequate description of your case as well as the followings:
  • the way mrcc was invoked
  • the way build.mrcc was invoked
  • the output of build.mrcc
  • compiler version (for example: ifort -V, gfortran -v)
  • blas/lapack versions
  • as well as gcc and glibc versions

This information really helps us during troubleshooting :)

Strange scf convergence between CCSDT and CCSD(T) scf for Iron

More
1 month 4 hours ago #757 by kallay
Please note that 18 electrons are correlated by mrcc, while molpro correlates just 10 electrons. You should freeze the lowest 4 orbitals in the mrcc run. However, even in this case the molpro and mrcc energies will still differ slightly as standard and the semicanonical orbitals are frozen by the two programs, respectively. So you should use standard ROHF orbitals with mrcc as well (it is only possible for iterative CC methods).
That is, to reproduce molpro's CCSD set
core=4
rohftype=standard

Best regards,
Mihaly Kallay

Please Log in or Create an account to join the conversation.

More
2 weeks 2 days ago #761 by bkwx97
I'm having a difference in energies in the ROHF energy between molpro and MRCC. Attached is input but here's the gist. It seems as though MRCC is landing on a higher energy solution even with the occ specified.

for CCSD(T)/PVTZ

MOLPRO ROHF:-1043.355682034293
MRCC ROHF: -1043.35224088

MOLPRO input

***,title
memory,800,m
mass,iso
geomtyp=xyz
basis=cc-pvtz
geometry={
1
stuff
Cr 0.0000000000 0.0000000000 -0.0167749442
}

{rhf,MAXIT=400;WF,24,1,6;occ,6,2,2,1,2,1,1,0;closed,3,2,2,0,2,0,0,0}
{uccsd(t),MAXIT=400}

MRCC INPUT

basis=cc-pvtz
mult=7
occ=6,1,1,1,0,2,2,2/3,0,0,0,0,2,2,2
scftype=rohf
rohftype=semicanonical
calc=CCSD(T)
mem=16000MB
geom=xyz
1
optimized structure
Cr 0.0000000000 0.0000000000 -0.0167749442

Please Log in or Create an account to join the conversation.

More
2 weeks 1 day ago #762 by kallay
Thank you for pointing out this. The cc-pVTZ basis for Cr seems to be wrong.
You should download the basis from the Basis Set Exchange ( www.basissetexchange.org/ , use CFOUR format), and replace the cc-pVTZ basis in the BASIS/Cr file.

Best regards,
Mihaly Kallay

Please Log in or Create an account to join the conversation.

More
2 weeks 1 day ago - 2 weeks 1 day ago #763 by bkwx97
So moving onto the boride I'm getting asinine results for the ROHF wavefunction. See attached. The SCF will not converge and if I add a damping factor I get an energy hundreds of hartrees off.

This is the current binary with the patched ucssd and ovirt compiled with Intel 2019.
Attachments:
Last edit: 2 weeks 1 day ago by bkwx97.

Please Log in or Create an account to join the conversation.

More
2 weeks 1 day ago #764 by kallay
You should try to restart the ROHF from UHF orbitals with scfiguess=mo but do not use scfdamp.

Best regards,
Mihaly Kallay

Please Log in or Create an account to join the conversation.

More
2 weeks 22 hours ago - 2 weeks 21 hours ago #765 by bkwx97
Edit: I reran with a UHF/pvdz and used that as a guess for ROHF/pvtz. The scf still did not converge. I increased the scfmaxit to 100 to check and see if it just needed a few more iterations. It did not converge.

What result did you get when you restarted with uhf orbitals?

edit: I used scfiguess=mo and the SCF converged as well as the CCSD(T) portion. The odd part is the CCSD(T) is more positive than MOLPRO's energy even when considering the standard rohf orbitals vs. semicanonical. The SCF is .25 more positive in MRCC.
Last edit: 2 weeks 21 hours ago by bkwx97.

Please Log in or Create an account to join the conversation.

Time to create page: 0.058 seconds
Powered by Kunena Forum