DOS run warning to be worried or not?

[sankh]

Hi Admin,

I came to notice a warning in OUTCAR file while running a spin-polarized PBE DOS job statically reading in the charges from CHARGCAR, obtained through a prior single-point run with (higher K-mesh, say 15*15*15 gamma-centred, and added extra empty bands), and both single-pt runs were performed with Tetrahedron intergration methods with Bloechl's correction on the equilibrium geometry obtained through optimization of internal coordinates and cell shape (ISIF =4, ediff=1e-8, ediffg= -1e-3) with 9*9*9 mesh and with the equlibrium volume that was obtained through B-M EOS fit at higher cut-offs, 750eV. All geom opt and single-point runs are done with 520 eV. Eq vol, cell parameters and internal co-ordinates are in excellent agreement with the experimental and previously reported PBE data. And eq cell geometry had "0" external pressure, either, (re-checked with higher cutoffs, as well). However, during DOS run with LORBIT = 12, the OUTCAR (VASP 2.5. is printing the followig warning:

" use gradient corrections
WARNING: stress and forces are not correct
(second derivative of E(xc) not defined)
use of overlap-Matrix (Vanderbilt PP)
Fermi weights with tetrahedron method with Bloechl corrections"

Could you please throw some light to this issue? Why does it warn saying stress & forces are not correct?

Your co-operation will be highly appreciated and acknowleged. Thanks in advance.

Best regards,
Sankh

[admin]

usually this warning reminds you that the forces and stress tensors are not correct, if you read the charge density form a CHGCAR file, keep it fixed (ICHARG >10), but probably have changed the k-mesh (as eg done for calculating DOS and band-structures). Forces and stress are only correct for charge densities (and potentials) which are converged to self-consistency for a given k-mesh.
However, DOS-runs are typical post-processing runs, which are done after the geometries and electronic structures are converged. The general checks for k-point convergence (as for the convergence of any other input parameters) should be done for any system before the (electronic, structure) convergence iterations are started