Ionic relaxation problem (VASP 5.2.11)

Message

superyoyo · #1 Post by **superyoyo** » Thu Feb 24, 2011 8:56 am

Dear All,
I am not sure if anyone have found this behavior of vasp5.2.11 yet. For some defect configurations in supercell calculations, I found that vasp5.2.11 give unphysical forces acting on atoms. This cause the problem that the calculation will never converge.

I have tested by setting two calculations in which the POSCAR file is exactly the same; 1) by vasp5.2.2 2) by vasp 5.2.11

my INCAR is like:
ISPIN=2
ISMEAR=0
SIGMA=0.05

ENCUT = 400 ! ENMAX Oxygen 400
NELECT = 866 !neutral 868
NSW=60
IBRION=2
NELMIN=4
EDIFF=1E-5
EDIFFG=-0.01
LREAL=.TRUE.

The calculation by vasp5.2.2 give the converged result within 35 iterations while vasp5.2.11 cannot. When I check the forces, it turns out that there are > 1eV force acting on several atoms!!, and the total energy cannot decrease.

If this is not a bug please let me know how to set INCAR correctly for vasp5.2.11

PS. I use PBE potential.

Thanks in advance!

superyoyo · #2 Post by **superyoyo** » Sat Feb 26, 2011 11:37 am

In addition to above problem. I also find this error occur very often in different materials:
ZBRENT: fatal error in bracketing
please rerun with smaller EDIFF, or copy CONTCAR
to POSCAR and continue

It does not matter if I follow the instruction it tells. It will happen again and again. Note that it does not happen that often in vasp4.6

Thanks.

superyoyo · #3 Post by **superyoyo** » Thu May 12, 2011 8:56 am

Dear Admin,
As far as I use VASP5.2.11, the problem I raised here still occurs and it is very annoying.
Could you please shed me some light on how to deal with it. Is it whether because of the VASP compilation or a bug of VASP5.2.11? Thanks a ton.

mrossi · #4 Post by **mrossi** » Fri May 13, 2011 2:32 am

Do you necessarily need ISPIN=2? That is, you require spin polarization?

I only ask, because when I was initially starting to use VASP, I left ISPIN=2, instead of changing to ISPIN=1, and kept having the ZBRENT error. However, once I turned off spin polarization, I achieved convergence rather quickly.

superyoyo · #5 Post by **superyoyo** » Fri May 13, 2011 10:34 am

Hi mrossi,

Even in the system that I use ISPIN = 1, the problem still occurs, just today.

25 F= -.78260631E+03 E0= -.78254989E+03 d E =-.117534E-05
curvature: 0.00 expect dE= 0.706-309 dE for cont linesearch 0.689-309
ZBRENT: fatal error in bracketing
please rerun with smaller EDIFF, or copy CONTCAR
to POSCAR and continue
Fri May 13 15:17:41 JST 2011

Any idea? Thanks

#6 Post by **admin** » Fri May 13, 2011 6:33 pm

we have never encountered any problems with the ionic relaxation in vasp.5.2.11 in all our tests.
please note: you charged your cell by +2, please consider what you are doing if you charge an infinitely repeated cell! If you cell is cubic, please try the multipole corrections as described in the online manual
the ZBRENT error indicates that vasp cannot interpolate a reasonable new guess of the next ionic configuration based on the position/force data obtained by the run so far. It seems that your system is converged ionically already.

superyoyo · #7 Post by **superyoyo** » Sat May 14, 2011 7:15 am

[quote="'smallblacktext'>[ Edited Mon May 16 2011, 11:40AM "]</span>

sabbem · #8 Post by **sabbem** » Wed May 25, 2011 4:07 pm

We encountered very similar problems while compiling VASP 5.2.11 using the latest Intel compiler (ifort/icc 12.0): wrong forces lead to a steady energy increase during ionic relaxations with IBRION 1.
Compilation using the former ifort/icc 11.1 solves the issue, and yields accurate forces.

superyoyo · #9 Post by **superyoyo** » Mon Jul 25, 2011 6:04 am

Hi Sabbem,

Thanks for your reply. However, the problem still exists even we compiled VASP 5.2.11 with ifc11.1.

asaduzza · #10 Post by **asaduzza** » Thu Feb 16, 2012 7:05 pm

Any solution to the above problem?