Abnormal binding energy for adsorption of charged particle by 3D materials

[xinru_peng]

Dear VASP developers and users,

I have learned that the energy of charged system should use NELECT and some correction keywords. I am calculating an adsorption model with a charged molecule and the neutral 3D material (It's not cubic and I use unit cell) using vasp-6.5.0. I got a positive binding energy (1.1 eV) and very small interaction energy (-0.08 eV) which is abnormal for me. I also check the interaction energy using Gaussian16 (same structure as unit cell, PBE0/def2SVP-D3(BJ)) and it shows a much more negative energy (-5.7 eV).

For the charged system, I use "NELECT" and "LMONO=.T.". Therefore, my question is what may be the reason that I got these energy results and what I can do make it more accurate.

Thank you for your time and looking forward to your reply.

The INCAR file (this is for optimization, and I deleted the Ionic part for interaction energy):
#starting parameters
ISTART = 0
ICHARG = 2
LWAVE = .TRUE.
LCHARG = .TRUE.
NPAR = 4

#electronic SCF
LREAL = auto
PREC = Accurate
ADDGRID = .TRUE.
ALGO = All

ENCUT = 520
ISMEAR = 0
SIGMA = 0.01
EDIFF = 1E-5
NELM = 500
NELMIN = 6
ISPIN = 1
ISYM = 0
IVDW = 12

# charged system
NELECT = 408
LMONO = .TRUE.

# Ionic
EDIFFG = -0.01
IBRION = 2
POTIM = 0.2
NSW = 500
ISIF = 2

[xinru_peng]

I am not sure if this is helpful, but I can add more information:
In my understanding, using LMONO=.TRUE. will decrease the energy of the whole system (E_AB).
However, actually, the energy of the whole system (E_AB) is higher than only using NELECT.

[jonathan_lahnsteiner2]

Dear Xinru Peng,

From you explanation I guess you have some bulk material with a molecule absorbed to it's surface. For surfaces one should switch on dipole corrections. You could do this by setting

Code: Select all

LDIPOL = .TRUE.
IDIPOL = 3
DIPOL = x y z  # set to the center of your system

IDIPOLE should be set to the direction of your surface.

Because you are reporting you might get more reasonable binding energy values with guassian. Gaussian does not use periodic boundary conditions whereas VASP does. So the reason for the nonphysical binding energy might arise due to artificial interactions of the periodic images of your system. To determine this you could do a finite size analysis or try increasing the vacuum thickness.

All the best Jonathan

[xinru_peng]

Dear Jonathan,

Thank you for your reply. And sorry for my unclear description.

My 3D material is a porous material, and it is periodic in all direction, for which I guess DIPOL correction may be not helpful. And I tested it with IDIPOL=4, but there is no difference in energy. Hopefully I didn't misunderstand this situation.

Best Regards,
Xinru

[xinru_peng]

Dear Jonathan,

Here is my progress and now I agree that IDIPOL correction will be helpful.
I read VASP wiki more and realized my setting of DIPOL is not correct, and I also use KERNEL_TRUNCATION/LTRUNCATE for the charged molecules, then its energy becomes higher, so the binding energy may be lower. Now I will use IDIPOL, DIPOL and EPSILON for the whole system, hopefully I can get a negative binding energy.

Thank you!
Best Regards,
Xinru