technical question about frequency calculation

Message

bademoor · #1 Post by **bademoor** » Mon Apr 06, 2009 8:31 am

Dear VASP users,

I am doing calculations on quite big systems (>200 atoms) and as frequencies are calculated numerically, those calculations (even for partial hessians) take a long time.
I was wondering whether someone can give me advice for
(i) splitting up calculations (I don't know whether this is possible)
(ii) restarting calculations from the point they've crashed (what sometimes happens, unfortunately)

Thx for the help
Regards
Bart

#2 Post by **admin** » Fri Apr 10, 2009 4:22 pm

unfortunately, you cannot split up or continue from a crashed run if you want to calculate the frequencies with the standard version of vasp. However, please check if it is necessary to have ALL the vibration modes (at the same time). (Without knowing the details of your investigations): Usually, if you do eg the frequencies of an adsorbate on a surface, it is sufficient just to allow for vibrations of the adorbate molecule itself and its NN atoms in the surface to have a very good accuracy.

bademoor · #3 Post by **bademoor** » Fri Apr 17, 2009 7:36 am

Thanks for the answer.
I'm looking into the partial hessian approach already, but still I sometimes have quit a lot of atoms (I'm dealing with zeolite unit cells). Anyway, I'll just assign some more cpu-power to speed up my calculations ;-)

Regards
Bart

alex · #4 Post by **alex** » Tue Apr 21, 2009 3:06 pm

Hello Bart,

I wrote myself an extension to finite_diff.F and main.F for vasp 4.6 to tackle this problem.
It's still alpha-stage, it's not really maintained, it does not always work, but most of the time. And it's tricky, if you have more than one restart to do.

Interested?

Cheers

alex