No of Ionic steps for Frequency Calculation

Message

kambiz · #1 Post by **kambiz** » Sat Mar 12, 2011 2:36 am

Hi dear all,

I want to calculate vibrational frequency (IBRION = 5) for adsorption of an atom on a catalyst surface.
I m using the optimized structure (using IBRION = 2) for this.

I m using this NSW = 2 in my INCAR, however, when I use "grep NSW OUTCAR" the result is 492 !!!!

I m using VASP 4.6, so I cannot set NSW = 0. Anyone please can tell me why it s like this.

Thanks

Kambiz

boris · #2 Post by **boris** » Mon Mar 14, 2011 3:08 pm

Hi

Try NSW= 1.

Boris

kambiz · #3 Post by **kambiz** » Tue Mar 15, 2011 3:55 am

Hi Boris,

I ve tried NSW = 1, 2 , ..., But still I have the problem.

alex · #4 Post by **alex** » Tue Mar 15, 2011 8:24 am

Some clarification: VASP is calculating _numerical_ 2nd derivatives. 2 movements in every direction for each atom means you have 82 atoms.
Yes, frequency calculations last loooong.

Hth,

Alex

boris · #5 Post by **boris** » Tue Mar 15, 2011 2:14 pm

@ Alex:

I think his problem is that he's setting NSW to 1 or 2 in the INCAR file but it prints 492 in the OUTCAR file. It seems like the keyword is not read by VASP.

Is that it Kambiz?

Boris

kambiz · #6 Post by **kambiz** » Tue Mar 15, 2011 11:28 pm

Thanks guys for your kindness,

Actually When I put IBRION = 2 or 1, the value of NSW in OUTCAR is exactly like that of INCAR.

I have 41 atoms, with NFREE = 2.

In my frequency calculations, surface atoms are frozen and the absorbate just can vibrate.

Another question I have, is that at least having one imaginary frequency is necessary in the results?

Kambiz

alex · #7 Post by **alex** » Wed Mar 16, 2011 7:44 am

It is not necessary to have imaginary modes. They mostly show up because of you are doing numerical derivatives (noise).
If you want to make it perfect, you have to set up a projection matrix which removes modes which are related to translation of the entire molecule.

Cheers,

alex

asaduzza · #8 Post by **asaduzza** » Tue Jun 05, 2012 2:06 am

I am trying to calculate vibrational frequency of a adsorb molecule on surface. The adsorption structure is optimized. For frequency calculation, I have frozen all surface atoms by putting F F F in POSCAR file. Although I set NSW=1, OUTCAR file always write NSW=1247. I haven't waited for 1247 ionic steps to finish, but checked up to 20 steps, and it continues to run. I even have tried by removing the surface atom from POSCAR and keeping only adsorbate. With only adsorbate, the NSW = 85 is written in the OUTCAR file. Could you guys indicate me the possible source of error. Here is the INCAR file I used:

NSW = 1
ENCUT = 450
EDIFF = 0.1E-06
EDIFFG = -0.02
IBRION = 5
NFREE = 2
POTIM = 0.015
ISIF = 2 (I tried 0 too)
ISMER = 0

alex · #9 Post by **alex** » Tue Jun 05, 2012 8:47 am

Typicalle you have 2 steps in each x,y,z (e.g. -x,+x,...) for every atom. So in total 6*NATOMS.

Neutrino · #10 Post by **Neutrino** » Tue Jun 05, 2012 11:43 pm

Check that you set selective dynamics in your POSCAR.

camilo · #11 Post by **camilo** » Fri Jun 08, 2012 1:56 am

About the imaginary modes, Kambiz:

These are not necessarily noise. You can visualize the mode, and thereby ascertain its nature. If it is a soft mode, experiment with a larger value of POTIM, if it is a hard mode, experiment with a smaller POTIM.

To get rid of noise, as such, make sure you do not have biasing errors in your calculation, and that the wavefunctions are highly converged prior to each ionic iteration.

Yes. These calculations represent a significant investment of resources, and hopefully your node won't crash at the next-to-last ionic iteration.