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

  • bkwx97
  • Topic Author
  • Offline
  • Senior Member
  • Senior Member
More
4 years 8 months ago #719 by bkwx97
Dear developers

I am running a CCSD(T)/PVQZ and CCSDT/PVQZ computation on Fe. I'm experiencing something very odd wrt scf convergence. The CCSD(T)/PVQZ scf does not converge while the scf portion of CCSDT/PVQZ does. Does the hand coded CCSD code call a different scf solver? I'm attaching the inputs of the two. This has been ran on a binary of the Feb 2019 version.



occ=6,1,1,1,0,2,2,2/5,0,0,0,0,2,2,2
basis=cc-pvqz
calc=CCSD(T)
mult=5
mem=60000MB
scftol=9
scftype=uhf
scfmaxit=100
ccmaxit=100
geom=xyz
1
stuff
Fe 0.0000000000 0.0000000000 0.1040823614


occ=6,1,1,1,0,2,2,2/5,0,0,0,0,2,2,2
basis=cc-pvqz
calc=CCSDT
mult=5
mem=60000MB
scftol=9
scftype=uhf
scfmaxit=100
ccmaxit=100
geom=xyz
1
stuff
Fe 0.0000000000 0.0000000000 0.1040823614

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

More
4 years 8 months ago #720 by Nike
Hi bkxw97,
You can force the post-scf integrals to be the same in multiple runs, by using iface=cfour.

This tricks mrcc into thinking that the integrals are from cfour, so instead of doing another scf, it just reads the fort.55 file in the folder.

There's another way but it doesn't work for excited states, whereas this method works always.

With best wishes,
Nike

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

  • kallay
  • Offline
  • Administrator
  • Administrator
  • Mihaly Kallay
More
4 years 8 months ago #721 by kallay
Sorry, I cannot reproduce this problem, the SCF converges in neither of the cases for me. You shouldn't use that tight convergence criterion, it converges with the default criterion. You can also use level shifting, which improves the convergence.

Best regards,
Mihaly Kallay

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

  • bkwx97
  • Topic Author
  • Offline
  • Senior Member
  • Senior Member
More
4 years 8 months ago - 4 years 8 months ago #722 by bkwx97
edit: fixed the issue
Last edit: 4 years 8 months ago by bkwx97. Reason: Adding attachement

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

  • kallay
  • Offline
  • Administrator
  • Administrator
  • Mihaly Kallay
More
4 years 8 months ago #723 by kallay
For the Ni atom mult=3 should be set.

Best regards,
Mihaly Kallay

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

  • bkwx97
  • Topic Author
  • Offline
  • Senior Member
  • Senior Member
More
4 years 8 months ago #725 by bkwx97
I have had a V atom CCSD(T) computation fail with the following error

forrtl: severe (37): inconsistent record length, unit 908, file /tmp/local/welchbr2/29700288/fort.908
Image PC Routine Line Source
ovirt 0000000000F7860C Unknown Unknown Unknown
ovirt 00000000004A6940 Unknown Unknown Unknown
ovirt 0000000000482486 Unknown Unknown Unknown
ovirt 0000000000400336 Unknown Unknown Unknown
ovirt 0000000001014775 Unknown Unknown Unknown
ovirt 0000000000400229 Unknown Unknown Unknown


Input is the following

basis=cc-pvtz
calc=CCSD(T)
mult=6
occ=5,1,1,1,0,2,2,2/3,0,0,0,0,2,2,2
mem=4000MB
scftype=uhf
geom=xyz
1
UCCSD(T)/CC-PVQZ ENERGY=-967.63978739
V 0.0000000000 0.0000000000 0.00000000

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

Time to create page: 0.045 seconds
Powered by Kunena Forum