Error in eom-cc2 calculation - huge scratch

1

Hi,
I was trying to run a simulation on a molecule with 56 atoms (+1 charge), using eom-cc2 to get excitation energies and oscillator strength with the following input requirements,

set {
cachelevel 0
basis def2-tzvp
scf_type df
reference rhf
roots_per_irrep [3]
}
wfn = properties(‘eom-cc2’, properties=[‘oscillator_strength’], return_wfn=True)

I am getting a scratch which is larger than 5 TB, and the calculation aborted with following error,

RuntimeError:
Fatal Error: PSIO_ERROR: 12 (error writing to file)

Error occurred in file: /scratch/psilocaluser/conda-builds/psi4-multiout_1615417611393/work/psi4/src/psi4/libpsio/error.cc on line: 133
The most recent 5 function calls were:

psi::PSIO::rw(unsigned long, char*, psi::psio_address, unsigned long, int)
psi::DPD::file4_mat_irrep_wrt_block(psi::dpdfile4*, int, int, int)

Printing out the relevant lines from the Psithon → Python processed input file:
H 3.054521 4.941052 -0.812018
H 1.520790 6.813751 -0.249770
H -0.852075 6.318089 0.243932
“”“)
core.IO.set_default_namespace(”")
→ wfn = properties(‘eom-cc2’, properties=[‘oscillator_strength’], return_wfn=True)

!----------------------------------------------------------------------------------!
! !
! Fatal Error: PSIO_ERROR: 12 (error writing to file) !
! Error occurred in file: /scratch/psilocaluser/conda- !
! builds/psi4-multiout_1615417611393/work/psi4/src/psi4/libpsio/error.cc on !
! line: 133 !
! The most recent 5 function calls were: !
! psi::PSIO::rw(unsigned long, char*, psi::psio_address, unsigned long, int) !
! psi::DPD::file4_mat_irrep_wrt_block(psi::dpdfile4*, int, int, int) !
! !
!----------------------------------------------------------------------------------!

How can I run the simulation ? Where did I go wrong ? Kindly help.

Regards :slight_smile:

2

Sounds like you’ve run out of disk space. Can you tell how far the calculation progressed from the output file?

3

Hi, thank you so much for your response. The calculation aborted during the following steps.

Wfn Parameters:
--------------------
Wavefunction = EOM_CC2
Number of irreps = 1
Number of MOs = 1136
Number of active MOs = 1136
AO-Basis = NONE
Semicanonical = false
Reference = RHF
Print Level = 1

    IRREP   # MOs   # FZDC  # DOCC  # SOCC  # VIRT  # FZVR
    -----   -----   ------  ------  ------  ------  ------
     A         1136     0       111     0       1025            0
    Transforming integrals...
    IWL integrals will be deleted.
    (OO|OO)...
    Presorting SO-basis two-electron integrals.
    Sorting File: SO Ints (nn|nn) nbuckets = 14
    Constructing frozen core operators
    Starting first half-transformation.
    Sorting half-transformed integrals.
    First half integral transformation complete.
    Starting second half-transformation.
    Two-electron integral transformation complete.
    (OO|OV)...
    Starting second half-transformation.
    Two-electron integral transformation complete.
    (OO|VV)...
    Starting second half-transformation.
    Two-electron integral transformation complete.
    (OV|OO)...
    Starting first half-transformation.
    Sorting half-transformed integrals.
    First half integral transformation complete.
    Starting second half-transformation.
    Two-electron integral transformation complete.
    (OV|OV)...
    Starting second half-transformation.
    Two-electron integral transformation complete.
    (OV|VV)...
    Starting second half-transformation.
    Two-electron integral transformation complete.
    (VV|OO)...
    Starting first half-transformation.
    Sorting half-transformed integrals.

Traceback (most recent call last):

Is there any way we can reduce the disk space usage and converge the simulation for such a large calculation ? If yes what could be the ‘ideal’ memory and disk space allocation to be done ? - Thank you :slight_smile:

4

I note that you haven’t frozen the core orbitals. Is it necessary to keep all the MOs active?

This is a rather large calculations for the EOM-CC2 code, so it may not be possible in the end, but freezing core would help.

5

Thank you so much for your insight. I will try with frozen orbitals.