Negative frequencies

[madhu_menon2]

I am working with s 2D structure whose phonon dispersion curves contain tiny negative frequencies around gamma point. They appear to be within the error margin. Is there a way to get rid of them? Thanks.

[ferenc_karsai]

That's hard to tell from the little information you gave.
First of all check if your calculational parameters in the INCAR file are converged (ENCUT, number of k-points, EDIFF, EDIFFG, LREAL=A, PREC=Accurate, etc.).
Also maybe you need to relax the structure (IBRION=2, ISIF=3), since the unrelaxed one is maybe vibrationally instable.

If that don't help try another XC functional.

[madhu_menon2]

Here are the details of my calculation.
First, I did a full ISIF=3 relaxation using ENCUT=520 eV,
PREC=Accurate; ALGO=Fast; EDIFF=1.0E-08; LREAL=.FALSE.; EDIFFG = -5E-04; IBRION=2; SIGMA=0.2 and ISMEAR=1,
I kept the kmesh of 5X3X1 using Gamma centered scheme.

After one full relaxation calculation converged, I again did the next ISIF=3 calculation taking the CONTCAR from the first calculation.

After my second ISIF=3 got converged, I took CONTCAR from that calculation and did ISIF=2 calculation with everything exactly same as in ISIF=3.

After my ISIF=2 calculation converged, I took the converged CONTCAR and did the DFPT calculation. For DFPT calculation, I took IBRIION=8, NSW=1, ISIF=3 and kept other parameters same as before. Please see the attached bans structure.

Thank you.

[alex]

Hi madhu_menon2,

is it possible that you are observerving the translational modes (in total 3) there? They typically show up with wavenumbers around zero.

Hth,

alex

[madhu_menon2]

Hi Alex, Thanks for the suggestion. But, how do I verify if they are translational? Do I need to look at EIGENVAL file or OUTCAR?

[ferenc_karsai]

The translational modes are the last three modes in the OUTCAR file. Type "grep THz OUTCAR" and you will immediately regcognize them.

[alex]

Hi,

they are last not by default, but very often. In case you want to be sure, check the normal modes for their deltax,y,z (e.g. with some visualisation tool).

Cheers,

alex

[ferenc_karsai]

Thank you Alex, good suggestion.

[madhu_menon2]

Thank you Ferenc_Karsai and Alex.

[madhu_menon2]

Thanks for your suggestions. I am using phonopy to get the frequency plots. Is there a way to have the plots without including the translational modes?

[SKM]

Hi Madhu Menon,

i am also facing the same situation exactly for my 2D material.
did you resolve the issue? if so please let us know.
i am trying with VASP/Phonopy with finite displacement method.
attached two figures with LDA and GGA based on SCF runs. The LDA based has slightly improved but not removed them.
Regards.

[alex]

Hi,

the trick is that you have to project the translational modes away (following Wilson, Decius, Cross in one of their books). They will disappear after applying the projection scheme. On top you'll get your thermodynamics right, because the low lying translational modes will add to your entropy based correction to enthalpy and free energy.

I'm not sure if e.g. phonopy considers this somehow.

Cheers,

alex

[SKM]

Hi Alex,

is this the book you suggesting?
Molecular Vibrations: The Theory of Infrared and Raman Vibrational Spectra (Dover Books on Chemistry) Revised ed. Edition
by E. Bright Wilson Jr. (Author), J.C. Decius (Author), Paul C. Cross (Author)

by the way what do you mean by "....project the translational modes away .."? you mean find and delete the data from the data file plotting phonons?
Regards

[alex]

Hi,

it's one of the two they published. But I don't remember, which book covers this topic.

You need to not just remove these frequencies, but to treat that physically by projection. Otherwise your entropy will be wrong.

Cheers,

alex