Phonon dispersion with supercell

[jeong9090]

Hi,

I'm currently working on phonon calculation.

I'm using VNL for acedemic user.

I can have phonon dispersion and phonon density of state for two atom in unit cell.

I wonder whether I can obtain phonon dispersion and phonon density of state results for graphene supercell having many atoms not only two atom in unit cell.

Thanks,

JY

[Jess Wellendorff]

yes, just create the desired supercell in VNL Builder and run the DynamicalMatrix with only few or no repetitions.

[Petr Khomyakov]

If you want to calculate the phonon dispersion and DOS for pristine graphene only, there is no need to use a graphene supercell structure as an initial structure for phonon calculations. To increase accuracy of the phonon band structure calculations, one only needs increasing the number of repetitions for calculating the dynamical matrix (DM). This has an advantage over the direct supercell approach, as the method implemented in the ATK for DM calculation takes into account symmetries of the system of study, i.e., graphene's lattice symmetry in your case. 

If your graphene structure has defects, adatoms or ripples, you then certainly have to start from a graphene supercell structure that includes this kind of structural imperfections.

[jeong9090]

Hi, Petr Khomyakov

If I can only use the unit cell of graphene having two atoms to calculate the phonon dispersion and DOS, do you mean that there is not any big different between the phonon dispersion for two atoms and that for many atoms?

Finally, I want to check the phonon dispersion of graphene on Cu as function of distance between both materials.
In this case also, can I only use unit cells for both materials? Do I need to increase the number of repetitions for calculating the dynamical matrix to increase accuracy?

I don't use any defects, doping, ripples.

Because I only use VNL for academic version, there is BZ folding when I use many atoms to calculate phonon band structure.
Finally, the phonon dispersion is so many straight lines.

For VNL academic version, is there any method to avoid BZ folding?

Thanks,

JY

[Petr Khomyakov]

If you have no imperfections in the structure, you can then build this structure (whatever it is) with a minimal unit cell and calculate the dynamical matrix for this cell, using some repetitions along the three lattice vectors of your minimal unit cell.

More repetitions you set for this calculation - more accurately the phonon band structure will be computed. Even so, you adopt repetitions, i.e., de-facto doing supercell calculations, the phonon bands will be computed and visualized for the minimal unit cell, i.e., the bands are given in the Brillouin zone of the minimal unit cell adopted, not the supercell.

If the minimal unit cell of the graphene-Cu system is not a primitive cell of a graphene sheet, then you have a natural folding of the phonon bands of the Cu-graphene hybrid system and that cannot be avoided.  It might be that you could do an (approximate) unfolding using the spectral function to get an effective band structure, as done for electronic bands in Phys. Rev. B 85, 085201 – Published 2 February 2012, but I am not aware of any code doing it for phonon band structures.