Error in calculating the Optical Properties for 1D system

[jjhskang]

Hi!

When we try to calculate the optical properties of the one-dimensional system,
we get an error emessage even in the lowest level, i.e., neglect of the local field effect.

I believe that this isbecause DO LOOP assumes that it has at least two k-points even along each of two aperiodic directions.
Long time ago fir the older version of VASP, I manually made corrections to the error and found that it worked afterwards.
However, it will be better if you fix this problem in the newest version.

[admin]

Could you be more specific? Which message have you got?
Which version of vasp do you use, and in which version
have you made corrections, which made it working?

[jjhskang]

We are now using vasp 5.3.5 for calculating the optical properties of one dimensional (1D) semiconducting polymers. To this aim, we tried to calculate the "frequency dependent dielectric functions through independent particle approximation. As you know, it needs to start with a standard SCF calculation using a proper DFT ( HSE06 hybrid or PBE functional) calculation.

In the next step, we ran a job for the optical calculation using INCAR file below:

SYSTEM =
PREC = Normal ! standard precision
ENCUT= 400
ISMEAR = 0 ; SIGMA = 0.01
NELM = 100
ALGO = Exact
NBANDS = 504
LOPTICS = .TRUE.
CSHIFT = 0.1 # this is the default

After a few minutes, the job crashed with the following message:

reading WAVECAR
random initialization beyond band 169
the WAVECAR file was read successfully
entering main loop
N E dE d eps ncg rms rms(c)
--------------------------------------------------------------------------
mpirun noticed that process rank 21 with PID 12040 on node jnode03 exited on signal 9 (Killed).
--------------------------------------------------------------------------

The tail of obtained OUTCAR file is also as follows:

First call to EWALD: gamma= 0.098
Maximum number of real-space cells 2x 3x 5
Maximum number of reciprocal cells 5x 3x 2

FEWALD executed in parallel
FEWALD: cpu time 0.00: real time 0.00


----------------------------------------- Iteration 1( 1) ---------------------------------------


POTLOK: cpu time 0.30: real time 0.30
SETDIJ: cpu time 0.01: real time 0.01
....................................................

Please note that we got similar messages for both PBE and HSE06 calculations.

In addition, a similar calculation for a 2D system does not encounter this kind of problem. Namely, the problem occurs only for 1D system!!


It was a long time ago that I modified VASP source file to make it work for 1D system. I guess that it was version 5.3.3. At the moment, I do not keep the complete modification. However, you will find that two subroutines in “optics.F” was modified as an example.
(1) SUBROUTINE CALC_NABIJ was modified:
! loop over all (current) k-points, spin component and cartesian directions
DO NK=1,KPOINTS%NKPTS
DO ISP=1,INFO%ISPIN
! HSK
! DO IDIR=1,3 (I changed IDIR from 3 to 1)
DO IDIR=1,1
! get the matrix element of the nabla-operator in eV/Angstroem units
CALL NABIJ_SOFT(NABIJ,IDIR,NK,ISP,W,WDES,GRID_SOFT,LATT_CUR)
CALL NABIJ_AUG_ADD(NABIJ,IDIR,NK,ISP,W,WDES,P,T_INFO)
……………………………………………

(2) SUBROUTINE SET_NABIJ_AUG was modified:
! intel efc compiler workaround (required for 6.X versions)
! P(NT)%NABLA=0._q ! old command
! HSK
! DO IDIR=1,3 (I changed IDIR from 3 to 1)
DO IDIR=1,1
DO LL=1,P(NT)%LMMAX
DO MP=1,P(NT)%LMMAX
P(NT)%NABLA(IDIR,LL,MP)=0._q
ENDDO
ENDDO
ENDDO

[admin]

We have not found any reason for the behaviour you have described.
Please report the complete input data set that we can reproduce
the crash for the 1D system.