Restart KRMCSCF in Dirac 19 is not working

[jvi...@gmail.com]

Dear Dirac Experts,

I was trying to restart KRMCSCF in Dirac 19. From a finished KRMCSCF run, I exported DFPCMO, DFCOEF (it was asking for it separately) and *IND*. It did not copy me INDICES.0, just 4IND1XXXX0 and 4INDINFO. I also exported KRMCSCF and renamed it KRMCOLD.

And with the same inp.inp and mol.mol, I tried to rerun it with --put="KRMCOLD DFPCMO DFCOEF KRCI_CVECS.* *IND* AF*"  (the KRCI and AF* have not been used)

But even though the output notes reading the orbitals from KRMCOLD, it still started the same cycles as before and not from the last energy reached!

Old run:

grep " Total KR-MCSCF energy    " *.out
 Total KR-MCSCF energy        :    -28113.112361761126522
 Total KR-MCSCF energy        :    -28113.125394605609472
 Total KR-MCSCF energy        :    -28113.137158356683358
 Total KR-MCSCF energy        :    -28113.148795731693099
 Total KR-MCSCF energy        :    -28113.148892665234598
 Total KR-MCSCF energy        :    -28113.159426791746228
 Total KR-MCSCF energy        :    -28113.167254639451130
 Total KR-MCSCF energy        :    -28113.171805390698864
 Total KR-MCSCF energy        :    -28113.175001843981590
 Total KR-MCSCF energy        :    -28113.177225834660931
 Total KR-MCSCF energy        :    -28113.178761413157190
 Total KR-MCSCF energy        :    -28113.179054300391726
 Total KR-MCSCF energy        :    -28113.179744791079429
 Total KR-MCSCF energy        :    -28113.180569582793396
 Total KR-MCSCF energy        :    -28113.180967428823351
 Total KR-MCSCF energy        :    -28113.181092199196428
 Total KR-MCSCF energy        :    -28113.181095550113241
 Total KR-MCSCF energy        :    -28113.182457800605334
 Total KR-MCSCF energy        :    -28113.186341760123469
 Total KR-MCSCF energy        :    -28113.188204601661710
 Total KR-MCSCF energy        :    -28113.189043397836940
 Total KR-MCSCF energy        :    -28113.189500980857702
 Total KR-MCSCF energy        :    -28113.189698981710535
 Total KR-MCSCF energy        :    -28113.189802574495843
 Total KR-MCSCF energy        :    -28113.189923335707135
 Total KR-MCSCF energy        :    -28113.189938429812173
 Total KR-MCSCF energy        :    -28113.189938525651087
 Total KR-MCSCF energy        :    -28113.189938525621983

restart:

 grep " Total KR-MCSC energy    " *.out

 Total KR-MCSCF energy        :    -28113.112361760879139
 Total KR-MCSCF energy        :    -28113.125394602237066
 Total KR-MCSCF energy        :    -28113.137158335128333
 Total KR-MCSCF energy        :    -28113.148795583176252
 Total KR-MCSCF energy        :    -28113.148889419018815
 Total KR-MCSCF energy        :    -28113.159426513971994
 Total KR-MCSCF energy        :    -28113.167254123985913
 Total KR-MCSCF energy        :    -28113.171806769372779
 Total KR-MCSCF energy        :    -28113.174970647040027
 Total KR-MCSCF energy        :    -28113.177245320272050
 Total KR-MCSCF energy        :    -28113.178798594781256
 Total KR-MCSCF energy        :    -28113.179011064748920
 Total KR-MCSCF energy        :    -28113.179772575793322
 Total KR-MCSCF energy        :    -28113.180584389185242

...computation continues

It looks to me that it is starting from the same initial guess, not from the already reached converged point (it is just test, I would expect it to run one marcoiteration and to end...)

What am I doing wrong, please?

Could you include some KRMCSCF restart jobs in the tutorial with a detailed description of what to do, please?

I attached the output and script to run the job.

Best regards,

Jakub

[Hans Jørgen Aagaard Jensen]

Hi

 

You exported DFPCMO, and DFPCMO takes precedence. That is, MOs on DFPCMO is used and whatever was on KRMCOLD is not used. This explains why you got the same result every time you tried. If you put KRMCOLD but do not put DFPCMO then your KRMCSCF restart should work.

 

Best regards, Hans Jørgen.

--
You received this message because you are subscribed to the Google Groups "dirac-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dirac-users...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/dirac-users/6a3ee489-0593-4dce-9d3d-6dff16d9a18bn%40googlegroups.com.

[jvi...@gmail.com]

Hello Hans,

Thank you very much. I tried with only KRMCOLD, but the program gave me a weird error (inp_mol.out.KRMCOLD) (and similarly for KRMCOLD + 4IND* files, inp_mol.KRMCOLD4I).

Surprisingly, KRMCSCF alone (inp_mol.out.KRMCSCF) also did not work.

However, a combination of KRMCOLD + KRMCSCF (it is the same file, just copied into two) produced the expected result - single iteration and convergence. 

I did not found, but I also did not need the file "INDICES.0" mentioned in this group for an earlier version of Dirac.

Best regards,

Jakub

Dne čtvrtek 24. října 2024 v 14:10:17 UTC+2 uživatel h...@sdu.dk napsal: