Caculation of DOS and band structure need ionic relaxation?

Message

opvu · #1 Post by **opvu** » Fri May 12, 2006 2:33 am

Dear all,

According to VASP manual, for DOS and band structure , we first run a static self-consistent caculation -> DOS; then ICHARG=1 -> band structure, right?

However, I see some people calculate DOS/band struct. with ionic relaxation.
which one is the right one? and why?

Thanks you very much.

Franky · #2 Post by **Franky** » Fri May 12, 2006 10:00 am

you should perform a DOS/bandstr calculation based upon a previously converged SC calculation. With converged I mean the relxation of both electronic and ionic degrees of freedom (Unless you want to study effects of ionic relaxation on bandstructure.). Converged also implies a carfull check of properties (Latt. Const., Bulkmod., etc) in terms of parameters such as k-points, ENCUT, ENAUG, etc.
Take the charge dens. from the converged run and perfom a nonSC calculation (ICHARG=11) which does not include ionic relax. In this calculation you can choose any kpoints you need. DONT perform banstructure calculation along certain lines in the BZ in a SC run, this is nonsense since you dont sample the BZ correctly!

kristen · #3 Post by **kristen** » Tue Mar 02, 2010 5:22 am

Thank you for your comments. Exactly how do you check the lattice constant, bulk mod., etc. to ensure convergence? Do you check against experiment or does it say explicitly in the OUTCAR that they are the converged values?

physicalattraction · #4 Post by **physicalattraction** » Thu Mar 04, 2010 8:03 am

Neither of them. If you check for instance the lattice parameter with experiment, you will find in general that if you use a LDA functional, your lattice parameter is too small and if you use a GGA functional, your lattice parameter is too large (Errors in order 1-5%). However, stick to the lattice parameter you find with VASP if continuing building structures with it.
You check convergence of the lattice constant by changing input parameters which should not influence your calculation, like the number of k-points, ENCUT, SIGMA, and see if your lattice parameter is still the most stable structure. I guess the same will hold for bulk modulus, though I have never looked into that.

#5 Post by **admin** » Thu Mar 04, 2010 4:48 pm

please simply check the "external pressure" (written to OUTCAR if ISIF is set) and the remaining forces on the atoms to see whether a structure is fully relaxed.
(of course k-point and basis set convergence,... have to be checked before)
Unless you have a very good reason not to do so, please do all calculations at the equilibrium structure of the respective XC functional you use.