EDDIAG warning when using LCALCEPS

Problems running VASP: crashes, internal errors, "wrong" results.


Moderators: Global Moderator, Moderator

Locked
Message
Author
dahvyd.wing
Newbie
Newbie
Posts: 3
Joined: Tue Jun 27, 2017 2:53 pm
License Nr.: 4

EDDIAG warning when using LCALCEPS

#1 Post by dahvyd.wing » Thu Jun 14, 2018 12:29 pm

Hi All,

I'm getting the following error when I calculate the macroscopic static dielectric tensor (inlcuding local field effects) using lcalceps = .true. for a hybrid functional:

WARNING in EDDIAG: sub space matrix is not hermitian 5 -0.575E+05

I'm using an 11x11x11 kgrid and this problem only happens for some of the materials I am calculating. Does anyone know what is causing this warning and what I should do to remove it?

Thanks,
Dahvyd

Here is an example of the INCAR that has this problem:

System = AlP
ENCUT = 300
EDIFF =1.e-7
ALGO = D
LHFCALC = .TRUE.
HFSCREEN = 0.2
ISMEAR = 0; SIGMA = 0.01
KPAR = 4
NBANDS = 12
GGA_COMPAT = .FALSE.
LRPA = .FALSE.
LCALCEPS = .TRUE.
EFIELD_PEAD= 3*0.001

admin
Administrator
Administrator
Posts: 2921
Joined: Tue Aug 03, 2004 8:18 am
License Nr.: 458

Re: EDDIAG warning when using LCALCEPS

#2 Post by admin » Fri Aug 31, 2018 10:28 am

the most possible reason for the error is that
One or more components of EFIELD_PEAD is too large, such that you are too near to the onset of Zener tunneling
please try to reduce
-- the EFIELD components
and/or the density of the k-grid.

Locked