My Community

Dear All,

It seems for IBRION=5, the symmetry of the dynamics matrix is NOT used. For example, to get correct vibrational frequencies, each atom has to be displaced in all the three directions even for cubic crystals.

Could someone tell me:
What "selective dynamics" is used for?
Is there any way to displace in only one direction while correct vibrational frequencies are obtained?
Thanks a lot.

Yes, if I understand your question, use selective dynamics: for example if you want an atom to displace along the z-axis only while the x and y fixed use F F T .
It is clearly written in the manual, have a look at


http://cms.mpi.univie.ac.at/vasp/vasp/node101.html

<span class='smallblacktext'>[ Edited Fri Mar 10 2006, 07:49AM ]</span>

Thanks for the suggestion.
I tried that before, but the results were not correct.
e.g. for diamond, there should be 2*3=6 frequencies.
However, if only displacements along z are allowed,
there will be only 2 frequencies printed in OUTCAR.
In other words, the numbe of frequencies is always the number of degrees of freedom.

This is not unexpected. To calculate the frequencies, the Hessian matrix must be calculated using finite differences, and then diagonalized (i.e. find eigenvalues/vectors of Hessian matrix). If only 2 degrees of freedom are allowed, the Hessian matrix can only be a 2x2 matrix, and hence you only get 2 vibrational frequencies.

In principle, the full Hessian matrix can be reconstructed using the symmetry. Because I'm new to VASP, I was wondering if there is such option hidden somewhere. Thanks.

Ah- I see what you want. I'm guessing that what you want is not supported in VASP, as per the manual.

I'm guessing you can use one of two methods. 1) Calculate the full Hessian by setting all atoms to "T T T". 2) Use selective dynamics, and create a Hessian by hand by using your knowledge of the symmetry. You'll then have to manually get the eigenvectors/values from the Hessian.