Author Topic: RI-UHF error without error message  (Read 59 times)

JakubV

  • Jr. Member
  • **
  • Posts: 14
  • Karma: +0/-0
RI-UHF error without error message
« on: November 11, 2017, 08:13:24 pm »
Dear Turbomole Experts,

I am trying to calculate RI-UHF for excited state of uranyl pentaaqua complex (I use ECP on uranium and add pointcharges simulating water solvent around the studied complex) further continuation should be CCSD or CCSD(F12*) with that reference state), but I faced strange error.

The RI-DFT output file ends in the middle:

Quote
BLOCKING OF BETA-MOS FOR RI-K


 occupied MOs partitioned into blocks of maximum length 47
 this requires 258 MiB core memory.

            MULTIPLE PASS Statistics
 ----------------------------------------------------
  pass       irrep     lambda      batch : MO  to  MO
 ----------------------------------------------------
   1       1  ->  1    1  -> 47             1  ->  47


and there is no other error message or information. Except for message from queueing system:

Quote
*** Error in `ridft': double free or corruption (!prev): 0x00000000075cedd0 ***
/var/spool/pbs/mom_priv/jobs/3658104.arien-pro.ics.muni.cz.SC: line 18:  7654 Aborted                 ridft > ridft.out
 ccsdf12 ended abnormally
program stopped.

For RI-RHF for the same system everything works fine for me, even the afterward CCSD(F12*) calculation (I am thinking about some Delta-CCSD(F12*) calculation of excitation energies).
Also non-RI UHF with the same configuration works on my system (I must aplogize for a side-question- would it be possible to do non-RI SCF reference based CCSD calculation, please?)

It might be error of the queueing system (PBSPro at Metacentrum computational grid), but maybe not and therefore I would be curious about and would appreciate your hints on this topic.

Best regards,
JakubV

uwe

  • Global Moderator
  • Sr. Member
  • *****
  • Posts: 409
  • Karma: +0/-0
Re: RI-UHF error without error message
« Reply #1 on: November 12, 2017, 09:57:12 pm »
Hi,

was that a parallel job or a serial one? And which version of Turbomole did you use? The error message seems to stem from the ridft module. Is that crash reproducible, or did it just happen once?

I'd recommend to send this issue to the Turbomole support team (turbomole@cosmologic.de) to find out what happened and how to fix that.

Regards,

Uwe