vibrational frequency

Message

apple · #1 Post by **apple** » Thu Feb 17, 2011 8:37 pm

hi,
I am doing vibrational frequency calculations for NH3 using PAW GGA potentials. To optimize it, I enclosed it in a box of 15x16x17 A3. The optimized N-H bond length is 1.0282 A and the angle is 107.5493 deg (actually one of the three angles is 107.5242deg...tiny difference).
When doing the frequency calculations, I get 12 modes, of which 3 are imaginary. The first six are (3615.124965, 3613.049218, 3471.404748, 1594.448852, 1594.251445, 946.393054)--note some of them are degenerate. The experimental values are (3333, 1628, 968).
(1) Can I say that my values in a good agreement with the experimental values? How close should they be, or what is the typical error in VASP?

The rest of the values are (20.034039, 10.07646, 0.533104,25.826477 img, 103.264461 img ,119.813231 img ).
While the first four are small and can be ignored, I suppose, I do not know what to think about the remaining two. They seem to be too high. (2) What is the source of this? Are my calculations wrong?
I played with the POTIM flag (tested 0.02, 0.01, 0.05) and EDIFFG (-1E-03, -1E-04), but did not notice any difference.

I would appreciate if someone could comment on my findings.
I am new to VASP, and this is my first vibrational freq. calculation.

Thanks a lot.

alex · #2 Post by **alex** » Fri Feb 18, 2011 12:32 pm

Hi apple,

I'm guessing your gradients are still too bad, e.g. decrease EDIFF(G) further. It might also be a good idea to reoptimise the structure first with the better settings. However, imaginary modes might not dissappear, but the absolute value should decrease and should be closer to zero.
About the additional mode in comparison with experiment you might look in the character tables of some decent group theory book (Raman active modes and so on).
6 modes are due to translation and rotation of the entire molecule.
DFT freqs are possibly wrong by up/down to 10% or more.

Cheers,

alex

apple · #3 Post by **apple** » Tue Feb 22, 2011 5:14 am

Thanks Alex. I reoptimized the structure with better setting and ran a vibrational freq. calc again, but did not notice any difference, i.e. my values are still a bit too high. Any other suggestions? Thanks for help. I appreciate it.

alex · #4 Post by **alex** » Tue Feb 22, 2011 7:39 am

Hello apple,

the high values are ok, probably the DFT error. Did the large imaginary modes decrease?

Cheers,

alex

apple · #5 Post by **apple** » Wed Feb 23, 2011 5:12 am

hi Alex,
the large imaginary modes decreased by ~ 1/2 after making a very careful calc. It seems like reoptimizing the structure with better settings was the key! Thanks for help.

alex · #6 Post by **alex** » Wed Feb 23, 2011 9:21 am

ok, I see. But 60cm-1 for an imaginary mode is still a little large. Try the analytic 2nd derivative in addition, please (IBRION = 6 instead of 5). And make a comparison.
How large is POTIM now? Should be rather small, 0.01 or 0.02 is good. It's harmonic approximation.

Cheers,

alex