Error in EOM-EE-CCSD and EOM-EE-CC(2,3) calculation

How To
#1

Hello,

Our Q-Chem version is 5.1.2.

We are trying to calculate the low-lying singlet and triplet excited state energies and the SOC constants by using the EOM-EE-CCSD and EOM-EE-CC(2,3) methods of a molecule consisting of 79 atoms.

Input

$rem
Method eom-ccsd (or eom-cc(2,3))
BASIS sto-3g #This basis set is to check, need a def2-SVP or 6-31G basis set
jobtype sp
MEM_STATIC 2000
cc_symmetry false
ee_singlets [1]
ee_triplets [2]
n_frozen_core fc
CC_TRANS_PROP = 1
state_analysis=true
CC_STATE_TO_OPT [1,3]
calc_soc true
CC_MEMORY 200000
cc_single_prec=1
eom_single_prec=1
cc_sp_dm = 1
eom_aresp_single_prec = 1
cc_sp_t_conv = 4
cc_sp_e_conv = 6
cc_erase_dp_integrals = 1
$end

output error (eom-ccsd)
Processor 1, Threads 4 (-nt 4)

Point group: C1 (1 irreducible representation).
A All

All molecular orbitals:

  • Alpha 295 295
  • Beta 295 295

Alpha orbitals:

  • Frozen occupied 63 63
  • Active occupied 121 121
  • Active virtual 111 111
  • Frozen virtual 0 0

Beta orbitals:

  • Frozen occupied 63 63
  • Active occupied 121 121
  • Active virtual 111 111
  • Frozen virtual 0 0

Import integrals: CPU 0.00 s wall 0.00 s
Import integrals: CPU 349.44 s wall 182.55 s

mpirun noticed that process rank 0 with PID 23627 on node node33 exited on signal 6 (Aborted).

remove work dirs /scratch/qchem23614.0 – /scratch/qchem23614.0
rm -rf /scratch/qchem23614.0
rm -rf /scratch/qchem23614

output error (eom-cc(2,3))
-nt 6

BASIS ORBS = 295 MOL ORBS = 295
NAUXBASIS = 0
FROZEN OCC = 0 FROZEN VIR = 0
CORR ORBS = 295 CORR SP ORBS = 590
NUM ALP ELEC = 184 NUM BET ELEC = 184
NUM ALP EXPL = 184 NUM BET EXPL = 184
NUM SO OCC = 368 NUM SO VIR = 222
NUM RESTR DOCC= 63 NUM RESTR DVIRT= 0
ORBS PER BLCK = 16 RESTRICTED_REF = 1
BLOCKING PARAMETERS:
NUM ROCC BLOCKS = 4 NUM AOCC BLOCKS= 8
NUM AVIRT BLOCKS= 7 NUM RVIRT BLOCKS= 0
ORBITALS/BLOCK = 16 16 16 15 16 15 15 15 15 15 15 15 16 16 16 15 16 15 15 15 15 15 15 15 16 16 16 16 16 16 15 16 16 16 16 16 16 15
BIRREP = A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A1
Q-Chem fatal error occurred in module libmdc/newfileman.C, line 384:
FileMan error: Error writing 4380160 bytes to file FILE_HALF_TRNS1
Path: /scratch/qchem15812/65.3
Please submit a crash report at Q-Chem Crash Reporter

mpirun noticed that process rank 0 with PID 15827 on node node34 exited on signal 6 (Aborted).

Can you help me with this issue? We actually want to explain the double-excitation configuration in our system.

#2

Looks like maybe (?) insufficient memory or disk for the 4-index transformation (FILE_HALF_TRNS1).

Are you trying to use MPI Q-Chem? That’s not really supported anymore.

#3

Yes, tried with MPI but not working.

#4

I do not recommend doing EOM(2,3) – it is based on the old code (ccman1 instead of ccman2) and is hard to execute efficiently. As far as EOM-CCSD is concerned:

  • you should make sure that CC_MEMORY is no more than 80% of memory available on the node;
  • check the scratch space (sometimes jobs crash because scratch is filled up by some leftovers from failed jobs;
  • if memory is a bottleneck, you can use CC_BACKEND = XM (instead of default VM);
  • note that ccman2 jobs are not mpi-parallel, so you should not attempt to distribute them across nodes.

Overall, EOM-CC job should run very effectively on a machine with the memory you specified or even smaller.

#5

Thank you ma’am and noticed that after removing

CC_STATE_TO_OPT [1,3]
calc_soc true

it terminates finely with all the properties (eom-ccsd). But are those right for doing SOC at the same time?
I will run the calculation at a higher basis set with your advice.

#6

It’s still not clear to me whether you are using MPI or OpenMP parallelization. You say MPI but then mention using -nt to specify the number of threads, which is OpenMP nomenclature. Are you running mpiexec to use Q-Chem? If not, this is not the MPI version (and anyway it’s the OpenMP version that’s supported at present).

#7

I am confused what you mean. Did you get the SOCs in your calculation or not?

SOCs are transition properties and you need to specify between which states they will be computed (and possibly add CC_TRANS_PROP).

I recommend you follow the examples in the manual.

#8

No. I did not get the SOCs. But without the keywords for the SOCs the job ran properly. I mean that transition orbitals, energies, oscillator strengths etc.

#9

This is good! It should now be easy to add keywords for SOCs – I believe we have a couple of example in the manual, when we discuss transition properties for EOM-CC.