Stress matrix after NEB project (eV)

[madhavan_nampoothiri]

Hello,

I am using vasp 5.4 sol_VTST version for relaxing my 2d material. I am not using any solvation or transition state related commands in my INCAR file.

INCAR:

PREC= Normal
ALGO= Normal
LREAL= .FALSE.
ENCUT= 500
EDIFF= 1E-5
NELMIN= 5

ISMEAR= 0
SIGMA= 0.01

# Spin polarization setting
ISPIN= 2

LCHARG= .FALSE.
LWAVE= .FALSE.

# Parallelization options
NPAR= 8

# Exchange-correlation functional settings
GGA= PE
IVDW= 12
# Cell optimization details
IBRION= 1
ISIF= 3
EDIFFG= -0.01
NSW= 50


But still I am getting "stress matrix after NEB" printed in my OUTCAR file like below

stress matrix after NEB project (eV)
-0.01040 0.03977 -0.00053
0.03977 -0.07988 -0.00020
-0.00053 -0.00020 -0.04757
FORCES: max atom, RMS 0.102351 0.100539
FORCE total and by dimension 0.201078 0.089396
Stress total and by dimension 0.109162 0.079879

This value increases as the system size increases (increasing number of atoms). Does this mean that the system is under stress? Do I need to give any additional commands to avoid this stress matrix?

P.S : I tried the same job with normal vasp_std. With that, this matrix is not printed, yet the system energies are exactly same.

[martin.schlipf]

This is an additional output of sol_VTST so I cannot comment on the specifics.

More generally, VASP should report the stress tensor too and relax the positions and cell if you set IBRION = 1 and ISIF = 3. So for every ionic iteration, you should have one output for the forces and the stresses. VASP will inform you when the ionic relaxation is complete, so that should tell you when the stress is small enough. Of course you can also check manually, whether the stress is below your desired threshold. Note that the stress will (almost) never be zero, so even for a converged calculation, you expect to see a certain remaining stress. But when this stress is small enough it will not affect the properties of the crystal.

[madhavan_nampoothiri]

Hi Martin,

Thank you for the response. As I mentioned earlier the stress values are very less (less than 1 eV) for unit cell (4 atoms). However, if I am increasing the size of the system, this value goes up.

For example, when I have 200 atoms in the system the stress matrix looks like below. This is after achieving convergence.

stress matrix after NEB project (eV)
12.48228 -0.04786 0.00000
-0.04786 10.81736 0.00000
0.00000 0.00000 -3.01851
FORCES: max atom, RMS 0.007330 0.002928
FORCE total and by dimension 0.043818 0.007085
Stress total and by dimension 16.791023 12.482278
Quasi-Newton relaxation of ions (Broydens 2nd method)
g(Force) = 0.926E-04 g(Stress)= 0.000E+00

And when I increase the no. of atoms to 700, the value shoots up to 100 eV (after convergence). So this is my concern whether I am missing out some information in the input files. I am attaching the input and output files corresponding to 200 atoms.

supercell.zip

[martin.schlipf]

The stress (in eV) is an extensive quantity, i.e., it increases proportional to the system size like the energy. Hence, VASP considers the stress per atom and compares that with the specified cutoff (EDIFFG). To get the more physical meaningful quantity, you should look at the stress in kb which does not change with the system size.