frequency calculation

[tssunrise]

When doing frequency calculation （IBROIN=5）,it is suggested to build a supercell large enough.However,if my cell is very small,building a large supercell(eg 3*3*3) would induce too many atoms in the system,and slow down the calculation significantly.What should I do ? Thanks!

[alex]

Why do want to calculate a supercell if your system does _not_ require one?

[forsdan]

@alex: I do not understand your answer. If you want to evaluate the dynamical matrix by using finite difference, you will have to use a larger supercell. Otherwise your force constants will be effected significantly by the displacements in the periodic cells.

@ tssunrise: I would suggest that you first do the calculation using a 1*1*1 cell and a 2*2*2 cell. Then if there is a significant difference and/or to check for convergence, you will have to accept that it is necessary and perform the 3*3*3 supercell calculation as well. (Alternatively you can gradually extent the supercell in one direction at the time to estimate the effect, 2*2*2 -> 2*2*3 -> 2*3*3 -> 3*3*3. )

EDIT: If you have access to Vasp 5.2, I recommend to use IBRION = 6 instead to reduce the number of needed displacements. Otherwise if you are using Vasp 4.6, you can do a corresponding symmetry analysis, and then assemble the dynamical matrix by hand.

Cheers,
/Dan






<span class='smallblacktext'>[ Edited Wed Jul 21 2010, 04:09PM ]</span>

[alex]

@forsdan: right. I was not considering tiny cells