IBRION=5 : from Hessian matrix to phonon dispersion; how?

[lahaye]

Hi,

I would like to produce a phonon dispersion relation.
Can someone give me some hints where to start how to
proceed from the VASP generated output.

So far I have understood following:

1. IBRION=5 in INCAR is needed, together with ISYM =0
POTIM=0.03 (for example) and NSW=1.

2. For accurate results I need a large supercell and many k-points,
although I don't yet really understand why.

3. A third party software, called "PHONON", seems to produce a
phonon dispersion relation with VASP, although this was with older
versions of VASP. Can VASP version 4.6 do this without extra software?

Thank you,
Rob.

[tjf]

You need a large supercell because the "direct method" can only resolve phonons of a wavelength that fit inside the supercell. These wavelengths of course vary a lot with the material. For the crystals I deal with (complex metal hydrides) cells of edge length of the order of 10 Ang are sufficiently large.

You need accurate forces. I'm not convinced that you necessarily need many k-points, but then I usually converge my energies w.r.t. cutoff and k-points to sub-meV levels.

IBRION=5 will allow you to calculate phonons at the gamma point only. There are standard integrals to go from these to the dispersion relations. (See, for example, Maradudin et al., "Theory of lattice dynamics in the harmonic approximation" in Solid State Physics Supplement 3, Academic Press, New York, 1971.)

Phonon, available (at a price) either as part of the Materials Design suite or directly as a binary from Krzysztof Parlinski (http://wolf.ifj.edu.pl/phonon/), does the full k-space integration for accurate properties.
<span class='smallblacktext'>[ Edited Thu Aug 03 2006, 04:07PM ]</span>

[bandy]

About the requirement of a large supercell, what I understand is that large supercells better reproduce the interactions(particularly long range) than their smaller counterparts. I do not understand how it is related to wavelength of phonon without much elaboration possibly with some refs.
<span class='smallblacktext'>[ Edited Thu Aug 03 2006, 08:33PM ]</span>

[tjf]

You could try Parlinski, Li and Kawazoe, PRL 78 (1997) 4063. Though in my opinion it's not expressed very well there. The Phonon manual is better.

The phonons are only accurate for all wavevectors within your zone if the interactions with regions outside your supercell are negligible.

[lahaye]

Thank you 'tjf' and 'bandy' for your comments.

I still have some basic questions:

1) When I use in my INCAR file

Code: Select all

 NSW = 1
 IBRION = 5
 POTIM = 0.05
 NFREE = 2
 ISYM = 0

then is the corresponding output only in the OUTCAR file,
or is there more elsewhere?

2) I don't fully understand the related output in OUTCAR, neither
is it explained in the VASP guide. The relevant parts of the OUTCAR,
which I mostly do not understand, I have put here:
http://surfion.snu.ac.kr/~lahaye/vasp/ Also the corresponding INCAR and POSCAR files are there.

For example, what do the numbers in the DYNMAT header mean:

Code: Select all

 DYNMAT
 ------
     1   32    3
 12.011
    1    1  0.0500  0.0000  0.0000

which corresponds to

Code: Select all

? number-of-atoms ?
mass-of-C-atom
 ? ? POTIM  ? ?

Thanks,
Rob.

[admin]

the output lines are
DYNMAT
------
number of atom types (1) , number of atoms (of each type) (32), number of degrees of freedom (3)
atomic mass (es for each type)
#of the degree of freedom (1), #of the displacement (1) , displacement stepwidth (x,y,z) (POTIM 0.0 0.0)

[bandy]

Thanks, tjf.

[Danny]

[quote="admin"]the output lines are
DYNMAT
------
number of atom types (1) , number of atoms (of each type) (32), number of degrees of freedom (3)
atomic mass (es for each type)
#of the degree of freedom (1), #of the displacement (1) , displacement stepwidth (x,y,z) (POTIM 0.0 0.0)[/quote]

and the rest of the lines? They differ from the TOTAL-FORCES given in the OUTCAR file...