Large internal forces when LCALCEPS=.TRUE.
Posted: Fri Aug 15, 2014 1:14 pm
I'm trying to calculate polarization and optical properties using Berry's phase method. I am doing calculations on an insulating, non-magnetic perovskite.
My POSCAR has been relaxed such that internal forces are <2.5 meV/?. When I run this INCAR, these low forces are echoed in the OUTCAR:
Prec=Accurate
ISMEAR = 0
LRPA=.TRUE.
#LCALCEPS=.TRUE.
LCALCPOL = .TRUE.
ENCUT = 500
GGA=PS
Forces from the OUTCAR:
POSITION TOTAL-FORCE (eV/Angst)
-----------------------------------------------------------------------------------
0.17877 0.00000 0.00000 0.000145 0.000000 0.000000
2.47513 2.38674 2.38674 0.002007 0.000000 0.000000
2.20200 2.38674 0.00000 -0.002010 0.000000 0.000000
2.20200 0.00000 2.38674 -0.002010 0.000000 0.000000
0.14165 2.38674 2.38674 0.001868 0.000000 0.000000
-----------------------------------------------------------------------------------
total drift: 0.001167 0.000000 -0.000004
However, when I turn on LCALCEPS (set it to .TRUE.) then these are the forces in my OUTCAR:
POSITION TOTAL-FORCE (eV/Angst)
-----------------------------------------------------------------------------------
0.17877 0.00000 0.00000 0.699751 0.000000 0.000000
2.47513 2.38674 2.38674 -1.565424 0.000000 0.000000
2.20200 2.38674 0.00000 0.587409 0.000000 0.000000
2.20200 0.00000 2.38674 0.587409 0.000000 0.000000
0.14165 2.38674 2.38674 -0.309145 0.000000 0.000000
-----------------------------------------------------------------------------------
total drift: -3.962303 0.000000 -0.000004
The BECs I calculate using LCALCEPS=.TRUE. violate the acoustic sum rule badly. In both cases I'm using a 6x6x6 kpoint mesh.
Any words of advice would be greatly appreciated. Cheers.
My POSCAR has been relaxed such that internal forces are <2.5 meV/?. When I run this INCAR, these low forces are echoed in the OUTCAR:
Prec=Accurate
ISMEAR = 0
LRPA=.TRUE.
#LCALCEPS=.TRUE.
LCALCPOL = .TRUE.
ENCUT = 500
GGA=PS
Forces from the OUTCAR:
POSITION TOTAL-FORCE (eV/Angst)
-----------------------------------------------------------------------------------
0.17877 0.00000 0.00000 0.000145 0.000000 0.000000
2.47513 2.38674 2.38674 0.002007 0.000000 0.000000
2.20200 2.38674 0.00000 -0.002010 0.000000 0.000000
2.20200 0.00000 2.38674 -0.002010 0.000000 0.000000
0.14165 2.38674 2.38674 0.001868 0.000000 0.000000
-----------------------------------------------------------------------------------
total drift: 0.001167 0.000000 -0.000004
However, when I turn on LCALCEPS (set it to .TRUE.) then these are the forces in my OUTCAR:
POSITION TOTAL-FORCE (eV/Angst)
-----------------------------------------------------------------------------------
0.17877 0.00000 0.00000 0.699751 0.000000 0.000000
2.47513 2.38674 2.38674 -1.565424 0.000000 0.000000
2.20200 2.38674 0.00000 0.587409 0.000000 0.000000
2.20200 0.00000 2.38674 0.587409 0.000000 0.000000
0.14165 2.38674 2.38674 -0.309145 0.000000 0.000000
-----------------------------------------------------------------------------------
total drift: -3.962303 0.000000 -0.000004
The BECs I calculate using LCALCEPS=.TRUE. violate the acoustic sum rule badly. In both cases I'm using a 6x6x6 kpoint mesh.
Any words of advice would be greatly appreciated. Cheers.