[QE-users] [SPAM] Pw2wan unexpectedly exceeding memory limit

[Stefan Velja]

Dear QE and W90 users,


I've used QE for several months running multiple calculations on very large MoSe2 supercells (including 200+ atoms) without trouble. Recently, I have been trying to run Wannier90 on such a configuration with 96 atoms. I ran the NSCF calculation easily on a "standard96" HLRN Göttingen mode (362 GB of usable memory) and tried to run a corresponding pw2wan calculation, resulting in the following error:


 Case  : isolated

  AMN: iknum =        1

===================================================================================
=   BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
=   RANK 0 PID 3970 RUNNING AT gsn2002
=   KILLED BY SIGNAL: 9 (Killed)
===================================================================================


This seems to be a "memory limit exceeded" error message. As seen above, I used only 1 k-point to simplify the calculation (after attempts with more k-points failed) and used the "huge96" partition (1522 GB of usable memory). The same calculation with a pristine MoSe2 cell (3 atoms) ran easily, even with SOC. What steps could I take to solve this?


Please find the following excerpts from my NSCF, W90 and pw2wan input files and let me know if you need any additional data. I excluded some of the flags for brevity.


nscf.in:

&CONTROL
calculation = 'nscf'
pseudo_dir = '/home/bepsvelj/pseudopotentials'
prefix = 'mose2'
outdir = 'tmp'
disk_io = 'low'
verbosity = 'high'
/

&SYSTEM
nbnd = 1088
/// Other sistem properties listed here
/

&ELECTRONS
conv_thr = 0.000001
/

ATOMIC_SPECIES
Se     79.      Se_ONCV_PBE-1.2.upf
Mo     96.      Mo_ONCV_PBE-1.2.upf

ATOMIC_POSITIONS (angstrom)
/// Atom positions listed here

K_POINTS crystal
1
  0.00000000  0.00000000  0.00000000  5.000000e-02


wannier90.in:
num_bands = 1088
num_wann = 1088
num_iter = 100
num_print_cycles = 10

iprint = 3

auto_projections = .true.

bands_plot = true

begin kpoint_path
G 0.0 0.0 0.0 G 0.0 0.0 0.0
end kpoint_path

mp_grid = 1 1 1
begin kpoints
  0.00000000  0.00000000  0.00000000
end kpoints

begin atoms_cart
/// Atom positions listed here
end atoms_cart

begin unit_cell_cart
/// Cell properties listed here
end unit_cell_cart


pw2wan.in:
&inputpp
  outdir = 'tmp'
  prefix = 'mose2'
  seedname = 'w90'
  write_amn = .true.
  write_mmn = .true.
  scdm_proj = .true.
/

Thank you very much in advance!

Kind regards,
Stefan Velja
PhD Student,
Electronic Structure Theory group (Prof. Dr. Caterina Cocchi)
Carl von Ossietzky University of Oldenburg
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20230304/a444004a/attachment.html>