Reliability of ISIF=8 for System Optimization

[Neeta Bisht]

Dear community,

I am working on a 2D monolayer of FeBr₂ with a reduced stoichiometry, i.e., Fe₁Br₁ (7 Fe atoms and 7 Br atoms). My aim is to observe the change in the ground state lattice parameter for an Fe₁Br₁ system compared to FeBr₂.

To achieve this, I performed potential energy surface calculations by manually varying the lattice parameter of Fe₁Br₁. However, I encountered an issue: the system does not converge to a minimum energy value (i.e., no minimum point in the potential energy curve was found).

I then noticed the addition of a new ISIF setting, ISIF=8, which operates similarly to ISIF=2 but allows for changes in volume as an additional degree of freedom. Although I did not expect my Fe₁Br₁ system to converge to a ground state with this setting, it surprisingly did, providing a minimum geometry with an optimized lattice parameter and reaching the required accuracy.

My question is: Can I rely on the optimization results obtained using ISIF=8 for this system?

Thank you for your insights.
Regards
Neeta Bisht

[svijay]

Dear Neeta Bisht,

For the same change in lattice vectors, volume and positions I would expect manually changing these parameters and using the corresponding ISIF value to be very similar. One reason for a mismatch may be due to the basis (see here: https://www.vasp.at/wiki/index.php/Ener ... lay_stress for more details). Often this issue can be fixed by increasing ENCUT.

Somewhat related: if you would like to just relax the lattice vectors for the 2D system, you may consider using LATTICE_CONSTRAINTS = .TRUE. .TRUE. .FALSE. as listed here: https://www.vasp.at/wiki/index.php/LATTICE_CONSTRAINTS

Sudarshan

[Neeta Bisht]

Dear Sudarshan,
Thank you for the advice. I looked into the LATTICE_CONSTRAINTS tag and its a new addition in the VASP which is available from vasp 6.4.3. I will definitely use this for the 2D material optimization. About the pully stress article, I read the article carefully and implemented the suggestions. However, my PES curve (manual change in volume and plotting it with energy) still looks bad (Please see the attached plot and the INCAR for this run). I doubt that this reduced stoichiometry FeBr2 monolayer is stable. Because, I ran series of AIMD calculations and the resultant structure looks heavily distorted. But, I would still think naturally, that ISIF=8 gives me a optimized structure that looks fine and stable.

IVDW = 11 # (DFT-D3 method of method with no damping)
GGA=PE
ISTART = 0
ICHARG = 2
PREC = Accurate
ISPIN = 2 # (Spin polarised DFT)
LREAL = .FALSE. # (Projection operators: automatic)
ENCUT = 400 # (Cut-off energy for plane wave basis set, in eV)
LWAVE = .FALSE. # (Write WAVECAR or not)
LCHARG = .FALSE. # (Write CHGCAR or not)
ADDGRID= .TRUE.

#Electronic Relaxation

ISMEAR = 1 # (Gaussian smearing, metals:1)
SIGMA = 0.2 # (Smearing value in eV, metals:0.2)
NELM = 300 # (Max electronic SCF steps)
NELMIN = 6 # (Min electronic SCF steps)
EDIFF = 1.0E-5 # from 0.001 to 0.0000001
MAGMOM=7*5.0 7*0.6
#Ionic Relaxation

NSW = 150 # (Max ionic steps)
IBRION = 1 # (Algorithm: 0-MD, 1-Quasi-New, 2-CG)
ISIF = 8 # (Stress/relaxation: 2-Ions, 3-Shape/Ions/V, 4-Shape/Ions)
EDIFFG = -0.001 # (Ionic convergence, eV/AA)
ALGO=FAST
NCORE=36 # No. of processor per node in use/2

Regards
Neeta Bisht

[svijay]

There is nothing immediately incorrect that I can see with the INCAR file. Maybe the only thing to check is if ENCUT is converged for the calculations that you are doing for the different volumes. For anything more specific I would need to see the OUTCAR files for all the different runs.

Sudarshan