vdW-DF2 job

[serenity_3]

Hello,
I am attempting to study adsorption of O atoms on a metal surface using the vdW-DF2 functional. While my jobs for the the surface+O and surface have converged, the job involving the O atom placed in a box needed to calculate the binding energy of O on the surface gives me unreasonable positive energy ( free energy TOTEN = 1.14957393 eV) . Please let me know what is wrong. Below is my INCAR file:

SYSTEM =0
NWRITE = 1
ISTART = 0

PREC = HIGH
ENCUT = 400.0
EDIFF = 1E-05
EDIFFG = -1.0E-03

IBRION = 2
ISPIN = 2
ISIF = 2
NSW = 20
LCORR = T
ISMEAR = 0
SIGMA = 0.01
LORBIT = 0

IALGO=38
LDIAG = T
LREAL = F
LPLANE = T
NSIM = 4
NPAR = 1

NBANDS = 12
MAGMOM = 2.0

LWAVE = F

GGA = ML
LUSE_VDW = .TRUE.
ZAB_VDW = -1.8867
AGGAC = 0.0
LASPH = .TRUE.

[serenity_3]

Hello,
I am attaching the input and output files of my unsuccessful run.
I appreciate your guidance on how to run this job.
Thanks!

[pedro_melo]

Dear serenity_3,

I have some suggestions regarding your calculation. To begin with, since you are running a single atom in a box, it is probably easier that you:

• Make all side equal

• Remove the van der Waals interactions

• Deactivate the ionic-minimisation options

The first suggestion just makes the calculation easier, since the system is now symmetric, while the others remove unnecessary complexity. The box should be big enough so that there are no fictitious forces between replicas, thus there is no need to optimise lattice parameters or atomic positions. Also adding a long range term like vdW will only force you to use bigger boxes until this interaction is negligible.

I attach an example of the modified INCAR and POSCAR files that I used. With it I obtained a total energy of -0.01247611 eV.

Let me know if this helps. Kind regards,
Pedro

[serenity_3]

Pedro,
Thank you for your guidance. However, I am still a bit confused about the issue.
My goal is to calculate binding energy of O atoms to a metal surface. I thought it's prudent to apply the same vdW correction method to both the isolated oxygen atom and the oxygen-metal adsorption system to maintain consistency and accuracy across my calculations. You write that the calculations involving isolate oxygen atom does not need to be treated at the vdW level. I understand that these interactions are typically weak to significantly affect the system's energy, but should it still be included? What about O2 in a box?

Also, I ran a calculations for an oxygen atom in a symmetric box without the vdW correction. However, the energy I get is -1.52565277 eV, thus significantly different from the energy you provided. I am attaching the OUTCAR file for more information. Can you please let me know how you got -0.01247611 eV?

Thank you so much for your help.

[pedro_melo]

Dear serenity_3,

I understand your point about wanting to have the same level of approximations, but I think that including van der Waals interactions on a system that does not have them will only make your calculation more complicated. You will need to increase the size of your cell until they vanish completely, i.e. till the fictitious interaction between different replicas is zero. In the end the result should be the same, but you will have spent more resources than necessary. So I advise you to keep the same functional and pseudo potential, but turn off the vdW interactions. The same goes for your O2 molecule.

The example I provided is by no means to be taken as the "correct calculation" to the ground state of the isolated O atom, just as a starting point. Also, both calculations used different INCAR files, with yours having

ISPIN = 2
ISIF = 2
ISMEAR = 0
SIGMA = 0.002

which are different in the INCAR I provided as an example.

Also, as I mentioned on my previous message, you do not need to use ISIF=2, since it is just an isolated atom in a box. If you want to make the cell size increase until the total energy converges, you can use ISIF=7 or do it manually by running calculations where the lattice parameter is changed in the POSCAR. This should just increase/decrease the lattice parameter of your cubic unit cell. Again, this is probably not necessary if you use a cell that is already large enough. And if you check your OUTCAR, you will see that VASP did not even ran more than one SCF cycle because the forces and stress were already small enough.

Note that for your O2 molecule yes, you will want to run coordinate optimisation, since the bond length (and thus the total energy) will depend on the exchange-correlation functional that you chose. In this case yes, do use ISIF=2 (or ISIF=8 if you think the cell is too small).

Let me know if this helps. Kind regards,
Pedro

Kind regards,
Pedro