van der Waals functional gives positive total energy to Na and K bulk

[zhaohui]

Dear all,

I am using van der Waals functional to calculate formation energy of alkali metal graphite intercalation compounds, for example KC8. The reaction is simple, like K + 8C--> KC8. For my understanding, I need to use the same functional for all compounds in the equation to get formation energy. When I calculate K (bcc structure), it gives positive total energy. I've try optPBE and revPBE, they all return positive energy. It can't be correct. My INCAR file is as follows, does anyone know what could be the reason? Did I set anything wrong? (The cutoff energy is very high because of using C_h potential).

I need to mention, when I switched off van der Waals function and used GGA_PBE, the total energy became negative.

General:
SYSTEM = K_bulk
ISTART = 0 # 0 start from scratch, 1 use old WAVECAR
ICHARG = 2 # 1 use old CHGCAR, 2 superposition atoms
# 10+ Fix CHG
ISMEAR = 2 # -5 tetra+Blochl, 0 Gaussian, >=1 MP
SIGMA =0.01 # width semearing --> not used with Tetra+Blochl

LWAVE = .FALSE. # do not write the WAVECAR file
LCHARG = .FALSE. # write the CHGCAR file
LVTOT = .FALSE. # do not write the potetial file
LREAL = .FALSE. # real space projection yes/no
SYMPREC = 1E-05
# ALGO = fast # stable Davidson algorithm for the initial five iterations
# followed by the faster RMM-DIIS algorithm
NPAR = 4

Electronic relaxation:
PREC = accurate
ENCUT = 910 # kinetic energy cutoff
# for 1meV covergency in energy difference
EDIFF = 1.0E-7 # electronic convergence criterium

Ionic relaxation:
ISIF = 3 # full geometry relaxation

IBRION = 2 # -1 fix atoms, 2 conjGrad for relax
NSW =60 # number of ionic steps (static: NSW=0)
EDIFFG = -1.0E-3

vdw functional:
LUSE_VDW = .TURE. # These 2 lines are needed for all following exchange functional
AGGAC = 0.0000 # remove PBE correlation correction

GGA = RE # revPBE functional

Any help would be much appreciated!

Best regards,
Zhaohui

[TopolM]

Hi, all,
I have similar problem:
positive total energy for bulk Ag
I took the parametrs from the paper: J. Klime?, D. R. Bowler, and A. Michaelides, Phys. Rev. B 83, 195131 (2011).
for PBE I obtained correct lattice constant
but for revPBE, optB88 and other +vdW VASP gives positive total energy
I use vasp.5.3.2
why this can happen?
could anyone answer?

<span class='smallblacktext'>[ Edited Tue Feb 05 2013, 04:40PM ]</span>

[kelum]

Hi,

the energy written out by VASP is not the true total energy but energy with respect to isolated reference atom (or atoms when you have more in POSCAR) for which the POTCAR was calculated. To put it simply, VASP just shifts the zero of the energy so that we don't have to deal with huge numbers. Therefore for an isolated atom one obtains energy around zero, although the true total energy will be some negative number.

If you obtain a positive energy for an atom it just means that the atom is higher in energy than the reference atom. This can happen even for PBE. In case you calculate atomisation energy, you should always do the reference atom calculations by yourself.
(And then do E_atomis = E_VASP(bulk) / n - E_VASP(reference atom) )
It is not correct to use do E_atomis = E_VASP(bulk) / n .

For the vdW-DF calculations the vdW correction is added. It is a positive number and then for the atoms the energy written out by VASP is often positive (up to few eV). This is not a problem. This just says that the energy of the reference vdW-DF atom is a bit higher than the energy of the PBE atom.

Since the binding in Ag or K is quite weak, the energy for the bulk can come out positive as well. (If the "energy" of K atom is 1.5 eV and the cohesive energy is -0.8 eV, then the "total" energy given by VASP will be just 0.7 eV per atom.) Again, no problem, just calculate the reference atom and subtract it to get E_atomis=0.7-1.5=-0.8 eV .

To sum up, the total energy given out by VASP does not make sense to use by itself, only energy differences are sensible (at least in these situations).

To obtain the cohesive energy for bulk Ag, do the bulk calculation and the reference atom calculation and subtract them. That should give you the right number.

To get the intercalation energy of K, do the K, C, and KC8 calculations and subtract properly. That will give you the correct answer, although some of these energies might be positive. The INCAR looks all right to me.

best,
jik

[TopolM]

kelum, thanks a lot
now it's clear

regards,
tm

[nv]

Hi,
I have similar problems (but without vdw correction). I am trying to converge cell energy of SiO2 with respect to energy cutoff, but the results look very strange.
E_cut E_cell
200 0.96465370E+03
300 -.65450925E+02
400 0.10919453E+04
500 0.86967192E+03

Any help and suggestions would be appreciated!
mfg,
nv