Various issues with Berry phase calculations

Message

camilo · #1 Post by **camilo** » Wed Oct 22, 2008 9:22 pm

Hi,

I am currently attempting to do these calculations for a lone HF molecule in a 12x12x12 box. As per the instructions in the online manual, I do this in 2 steps:
First step is simply obtaining the CHGCAR, which I do using the following INCAR:

SYSTEM = HF
NWRITE = 3
ISPIN = 1
ENCUT = 300
PREC = HIGH
ALGO = NORMAL
ISMEAR = 0
SIGMA = 0.10

and the following KPOINTS:

Comment
0
M
5 5 5
0 0 0

I then proceed to copy the resulting CHGCAR to the Berry directory, along with the IBZKPT, and with the following INCAR:

SYSTEM = Berry's phase
NWRITE = 3
ISPIN = 1
ENCUT = 300
PREC = HIGH
ISMEAR = 0
SIGMA = 0.10
LBERRY = .TRUE.
IGPAR = 1
NPPSTR = 4
DIPOL = 0.500 0.500 0.500

This run crashes as soon as VASP starts calculating the <R>bp portion, with the following complaint printed out in the last few lines of the OUTCAR file:

Berry-Phase calculation of electronic polarization

IGPAR = 1 NPPSTR = 4

Error in subroutine BERRY: problem with k-mesh

NKPTS = 18 NSTR = 4 NPPSTR = 4

I am currently running on Mandrake Linux. The code has been compiled using the makefile for P4 systems, compiled using ifort 10.1

I am willing to post the makefile, should it be needed.

Thanks for any and all help.

#2 Post by **admin** » Wed Nov 12, 2008 1:21 pm

1) This is not a bug of the program:
The error message appears if NSTR*NPPSTR is not equal to NKPTS. If the appropriate kmesh is used, this condition is always fulfilled.
please do not use IBZKPT, but KPOINTS to calculate Berry phases. The k-point mesh has to be generated again, because the symmetry of the cell is automatically reduced in order to anticipate the displacement of an ion (usually, polarizations modes are calculated for bulk systems).
In your case, this gives a 36-kpoints mesh running without problems, for 9 k-points (5x5x5 MP) and NPPSTR=4 instead of the fixed 18-points mesh taken from IBZKPT.

2) just a remark: for a free molecule, please use an 1x1x1 k-mesh.

camilo · #3 Post by **camilo** » Tue Jan 06, 2009 2:16 am

Thank you.

Concerning remark #2, this will still result in an error because the variables NSTR & NPPSTR would each have to be 1 for the satisfaction of the NSTR*NPPSTR=NKPTS condition. Setting NPPSTR=1 in the INCAR does result in NSTR=1, but the calculation does not proceed beyond the computation of the expectation value term.

#4 Post by **admin** » Thu Jan 08, 2009 1:18 pm

NKPTS is re-scaled for that reason if you calculate Berry phases:
even if just the Gamma point is used, you will find (eg for NPPSTR=6)
NKPTS = 6 NSTR = 1 NPPSTR= 6
in OUTCAR, just before the polarization terms are written

jlbettis · #5 Post by **jlbettis** » Thu Jul 29, 2010 4:31 am

Simple question... What is the definition of "c" regarding the Berry-Phase term R.BP

R.BP = ( 0.00000, 0.00000, 1.63007 ) electrons Angst
+( 0.00000, 0.00000, 4.00100 ) *c

Most examples depict R.BP as a 1x3 matrix. I don't know how to treat the second term. What is the function of "c" ? Thanks in advance for your reply.