problem with 5.3.2 LOPTICS=.TRUE. and KPAR\=1

[hamelseb]

Hi,

I've been running a comparison between the static calculation without local field correction (LEPSILON=.TRUE.) and the 0-freq limit of LOPTICS=.TRUE.

I find that they agree as long as KPAR = 1
I also find that the value LEPSILON=.TRUE. returns is the same with KPAR\=1
However the value that LOPTICS=.TRUE. returns is different when KPAR\=1

Any thought?

Here's the INCAR file:
SYSTEM = lif
ISTART =1;ICHARG=11
ENCUT = 900
LWAVE = .TRUE.
LCHARG = .TRUE.
PREC = Accurate
NELM = 50
NELMIN = 4
IALGO = 38
KPAR =1
LREAL= .FALSE.
EDIFF=1.0E-8
ISMEAR = -1; SIGMA = 0.0861734300601
WEIMIN=0.0
LOPTICS=.TRUE.
NEDOS=12000
NBANDS=256

OUTCAR with 32 cpu KPAR=1:
frequency dependent REAL DIELECTRIC FUNCTION (independent particle, no local field effects)
E(ev) X Y Z XY YZ ZX
--------------------------------------------------------------------------------------
0.000000 2.213640 2.213640 2.213640 0.000000 0.000000 0.000000

OUTCAR with 32 cpu KPAR=4:
frequency dependent REAL DIELECTRIC FUNCTION (independent particle, no local field effects)
E(ev) X Y Z XY YZ ZX
--------------------------------------------------------------------------------------
0.000000 1.479533 1.479533 1.479533 0.000000 0.000000 0.000000

OUTCAR with 16 cpu KPAR=4:
frequency dependent REAL DIELECTRIC FUNCTION (independent particle, no local field effects)
E(ev) X Y Z XY YZ ZX
--------------------------------------------------------------------------------------
0.000000 1.479533 1.479533 1.479533 0.000000 0.000000 0.000000

Here is the comparison with LEPSILON=.TRUE. with either KPAR=1 or KPAR=4

HEAD OF MICROSCOPIC STATIC DIELECTRIC TENSOR (independent particle, excluding Hartree and local field effects)
------------------------------------------------------
2.214524 0.000000 0.000000
0.000000 2.214524 0.000000
0.000000 0.000000 2.214524
------------------------------------------------------

KPOINTS:
K-Points
0
Gamma
5 5 5
0 0 0

POSCAR:
Li F
3.00000000000000
1.1509157108062773 0.0000000000000000 0.0000000000000000
0.0000000000000000 1.1509157108062773 0.0000000000000000
0.0000000000000000 0.0000000000000000 1.1509157108062773
Li F
4 4
Direct
0.0000000000000000 0.0000000000000000 0.0000000000000000
0.0000000000000000 0.5000000000000000 0.5000000000000000
0.5000000000000000 0.0000000000000000 0.5000000000000000
0.5000000000000000 0.5000000000000000 0.0000000000000000
0.5000000000000000 0.5000000000000000 0.5000000000000000
0.5000000000000000 0.0000000000000000 0.0000000000000000
0.0000000000000000 0.5000000000000000 0.0000000000000000
0.0000000000000000 0.0000000000000000 0.5000000000000000

[hamelseb]

By the way I tried this with NPAR=1, NPAR=ncpu and NPAR*KPAR=ncpu with KPAR=4 and the results are the same

[hamelseb]

Oh and if KPAR = 2 we get:

frequency dependent REAL DIELECTRIC FUNCTION (independent particle, no local field effects)
E(ev) X Y Z XY YZ ZX
--------------------------------------------------------------------------------------
0.000000 1.597323 1.597323 1.597323 0.000000 0.000000 0.000000

which is again different.

[admin]

we cannot reproduce this behaviour with the latest release of VASP