Accurate DOS and Band-structure calculations

Message

bth20 · #1 Post by **bth20** » Wed Jul 12, 2006 12:21 am

Hi there,

I'd be very grateful if someone could clear this up for me:

In the online guide
(http://cms.mpi.univie.ac.at/vasp/vasp/node211.html) on how to generate accurate DOS and band-structure calculations it states

"as a first step one generates a high quality charge density using a few k-points in a *static* selfconsistent run." (my emphasis)

My question is: why does this step need to be static. Why can you not relax in this step? After all, the output will still be a self-consistent solution. Is this just a mistake in the online guide?

Essentially I don't see the difference between

1. Relax to get self-consistent relaxed solution (and CHGCAR and WAVECAR)
2. Go on to the final non-selfconsistent run (using CHGCAR and from step 1.).

and this:

1. Relax to get self-consistent relaxed solution (and CHGCAR and WAVECAR)
2. Perform a static self-consistent run (using the charge density etc from the relaxation run)
3. Go on to the final non-selfconsistent run (using CHGCAR and from step 2.).

I've been using the first method (ie without the static run) and I've been getting perfectly sensible results.

Any help appreciated

Thanks

Ben

#2 Post by **admin** » Wed Jul 12, 2006 9:14 am

this applies to the case that the cell volume and shape (c/a) have been optimized automatically and differ from the input geomtery after geometry optimization: as the FFT-grids are not adjusted at each geometry step (even if the cell volume and/or shape changes), one should do one final static run with the NGX,.... parameters determined for the converged unit cell in order to have optimum accuracy for the fully relaxed cell

bth20 · #3 Post by **bth20** » Wed Jul 12, 2006 10:45 am

Many thanks but I don't quite understand your reply.

What exactly is it that "applies to the case that the cell volume and shape (c/a) have been optimized automatically"?

and what do you mean by the cell volume and shape (c/a) being *optimized automatically*?

Do you mean that the additional static step is only necessary if the cell volume and/or shape have been relaxed in the first run (that is with ISIF =3,4,5,6,7) or do you mean it is only necessary if the cell volume and/or shape have NOT been relaxed in the first run (that is with ISIF =0,1,2)? Or is the static step necessary for both cases (ie for all values of ISIF)?

Furthermore how serious are the problems with accuracy of results if you have NOT performed the static step? I used ISIF=2 and did not run the static step. The results seem very believable but this issue is giving me doubts (there were a lot of simulations and I don't want to have to re-run them all!)

Ben

#4 Post by **admin** » Fri Jul 21, 2006 8:14 am

the additional static run only is necessary IF the volume or/and shape have changed (ISIF=3,4,5,6,7).
A general answer to your question cannot be given, because it depends on how large the volume and shape changes were, how PREC had been set,....
for ISIF=2, the additional static step can be omitted, as NGX,... NGXF,... stay the same if only the ions' positions are relaxed.