My Community

Hi everyone,

Has anyone been able to produce a density of states plot for a standard gas phase molecule such as benzene?

I am following the outline found here:

http://cms.mpi.univie.ac.at/vasp/guide/node102.html

Using the CHGCAR from a previous converged run, I used the following files for the run:

KPOINTS:

molecule
0
GAMMA
7 7 7
0 0 0

INCAR:

SYSTEM = Benzene
NWRITE = 4
ISTART = 0
ISMEAR = -5
ISPIN = 1
SIGMA = 0.05
ENCUT = 300.0
EDIFF = 1.0E-06
LREAL = AUTO
PREC = ACCURATE
ALGO = NORMAL

NELM = 200
DIPOL = 0.500 0.500 0.500
IDIPOL = 3
LDIPOL = .TRUE.

NBANDS = 50
NEDOS = 1000
NPAR = 1
LORBIT = 2
ICHARG = 11
RWIGS = 0.863 0.302

The problem is that the p states appear to be distributed rather evenly over a rather wide range (about 1.5 eV total), and that there is a significant d-population as well. My main concern is that this, being a gas phase molecule, a properly performed DOS calculation should show discrete peaks in the final plot.

Any suggestions/insights?

Camilo

What are your lattice vectors? If you have a molecule the cell should be large, and the bands essentially flat, which give you large peaks the in DOS.

Greetings!

The POSCAR is as follows:


C H
1.00000000000000
20.0000000000000000 0.0000000000000000 0.0000000000000000
0.0000000000000000 20.0000000000000000 0.0000000000000000
0.0000000000000000 0.0000000000000000 20.0000000000000000
6 6
Direct
0.5000000000000000 0.4302298156365565 0.5000000000000000
0.5603808840462569 0.4651341647080757 0.5000000000000000
0.5603808840462569 0.5348658352919243 0.5000000000000000
0.5000000000000000 0.5697701843634434 0.5000000000000000
0.4396191159537430 0.5348658352919243 0.5000000000000000
0.4396191159537430 0.4651341647080757 0.5000000000000000
0.5000000000000000 0.3757603875449354 0.5000000000000000
0.6075652945069621 0.4379113406185772 0.5000000000000000
0.6075652945069621 0.5620886593814227 0.5000000000000000
0.5000000000000000 0.6242396124550644 0.5000000000000000
0.3924347054930378 0.5620886593814227 0.5000000000000000
0.3924347054930378 0.4379113406185772 0.5000000000000000

These lattice vectors result in a nearest H-H distance of 15 Ang. Additionally, I've actually used IDIPOL=4, not IDIPOL=3 in the INCAR.
<span class='smallblacktext'>[ Edited Fri May 25 2012, 07:58PM ]</span>

I am curious why is Gamma is used, instead of Monkhorst Pack in the KPOINTS?

Perhaps it might give a more accurate badn structure using the GW potentials?

In addition pull up ENCUT to at least 500eV (you are using the standard C potentials are you?)

Since you are doing a molecule shoudn't you set your k-points to 1x1x1 ??

Are you sure your preconverged calculation converged, an not just stopped after 40/60 steps?

and finally, have you tried centering the molecule on the origin?

Danny

Hi Danny,

It boiled down to using a single point located at the gamma. The states are discretized, and appear to be in correct locations.

Note: The POSCAR I pasted above is, in fact, centered in the box; it also corresponded to a well-optimized geometry.


Thanks!

hmm, interesting to know.

The suggestion to move the molecule to the origin was with the idea that VASP did something wrong with the symmetry. (I was just covering a few bases at once;) )
Danny