My Community

Dear all,

I am trying to calculate the adsorption energies of an N atom, a N2 molecule and an H2 molecule to the surface of metals (relative to N2, H2 in the gas phase). I am comparing with results obtained using DACAPO and find good agreement for N2 and H2, but very poor agreement for the N adatom. Does anyone have any experience with these types of calculations and can lend any insight on why the two programs are yielding such different results? I will post the contents of the INCAR file below and more details can be provided upon request.

Thanks in advance!

# Regular
SYSTEM = test
ISYM = 0
NSW = 1000
EDIFF = 0.1E-4
EDIFFG = -0.01
ENCUT = 408
GGA = RP
LWAVE = .FALSE.
LCHARG = .FALSE.
PREC = medium
ISIF = 0

# Speedup
NPAR = 12
NSIM = 4
IALGO = 48
LPLANE = .TRUE.
LREAL = Auto

# Optiizers
IBRION = 2
POTIM = 0.2
IOPT = 2
MAXMOVE = 0.15
FDSTEP = 5E-3

Hi annagarden,

could you please tell us, where exactly the difference(s) are?
What are the reference systems (e.g. like N-atom, ...)

Cheers,

alex

Hi Alex,

Thanks for your reply. The reference systems are N2 (or H2) in a 10x10x10 angstrom box and a clean metal slab (supercell approach, 12 angstrom vacuum). For an N adatom the reference energy is half that of N2. I calculate the energies by simply taking the optimised energy (sigma->0) of the system with the adsorbate and subtracting the sum of the clean slab and the free molecule. Using VASP I find that the N adatom is bound roughly 0.5 eV stronger than using DACAPO. I am using PAW but the same problem exists when using the ultrasoft pseudopotentials instead. I have tried different optimisation algorithms and the problem persists. The binding of an H atom to the same metal slab is in good agreement with the two programs, so I suspect the problem lies in the description of the nitrogen. I am a relatively new VASP user so I was wondering if there was some simple flag that I am missing in the input.

Cheers,

Anna

Hi Anna,

one thing which likes to occur is a wrong occupation for 'easy' molecules and atoms. You may switch to a non-cubic box to destroy a possible symmetry. One other thing is to check the DOS if the actual occupation is the expected one. For a non-equal distribution of electrons you might have to reduce the smearing to sth. like 0.001 eV.

Cheers,

alex

Hi Alex,

Great, thanks, I'll look into those things today.

Cheers,

Anna

Hi Alex,

The asymmetrical box didn?t affect the energies of the isolated molecules and the occupations look fine. Furthermore, I have looked a little closer at the results and I believe that the problem is in the description of the adsorption process, rather than the reference state. For example, I have built two systems. In the first, N2 is adsorbed to the surface. In the second, two separate N atoms are adsorbed. In this way the same reference is used to calculate the binding energy. Using DACAPO, I find that the energy difference between the two adsorbed structures is 0.42 eV, whereas in VASP it is only 0.02 eV. In plain words, DACAPO calculates that N2 binds considerably stronger to the surface than two N atoms whereas in VASP the binding energies are similar. This would be unaffected by any choice of reference state. The situation is very similar when I use a fixed geometry, ruling out any difference in physical structure as predicted by the two methods. I have some confidence in the DACAPO results, as I have been able to reproduce published data on similar systems (slightly different slab structure, but same adsorbates).

I hope that this makes some sense, but if you have any thoughts regarding this I would really appreciate it!

Cheers,

Anna

Hi Anna,

sorry for the delay.

Two ideas to think about:
- are the calculations at the numerical limits (at least VASP isn't)
- if I remember correctly, DACAPO provides PPs of Vanderbilt quality, isn't it?

Cheers,

alex

Hi Alex,

No problem with the delay. Firstly, no I don't think that the calculations were at the numerical limit, but I thought that the difference between the two results was so great that this would be unlikely to cause the problem. I have done a little fiddling with convergence criteria and it seemed to not affect the results. Secondly, yes the DACAPO PPs are of Vanderbilt quality.

Thanks for your help in this, I guess it still remains a mystery at this stage! I will write back here if I discover anything else.

Cheers,

Anna