H2 dissociation energy

[pavel]

I tried to calculate the dissociation energy of H2 using PAW_GGA potential for H (version 07Jul1998) using the following INCAR files:

Code: Select all

SYSTEM = isolated H2 molecule

ISMEAR = 2
SIGMA = 0.2
ENCUT = 450

NPAR = 4

GGA = 91

ISPIN = 1

IALGO = 48
LREAL = .FALSE.
NSIM = 4

NSW = 20
IBRION = 2
POTIM = 0.3

NELMIN = 4
EDIFF = 1E-2
EDIFFG = 1E-4

and

Code: Select all

SYSTEM = isolated H atom

ISMEAR = 2
SIGMA = 0.001
ENCUT = 450

NPAR = 4

GGA = 91

ISPIN = 1

IALGO = 48
LREAL = .FALSE.
NSIM = 4

NSW = 20
IBRION = 2
POTIM = 0.3

NELMIN = 4
EDIFF = 1E-2
EDIFFG = 1E-4

I have got a dissociation energy about 6.7 eV, while experimental value is about 4.52 eV.

Then I realized that I have forgotten spin. I recalculated both values with spin, but the dissociation energy changed only in the forth digit after comma.

What could be the reason for this discrepancy?

[pavel]

Dear Ojwang,
Certainly, I must be wrong and I am trying to understand where.

The value of 6.7 eV I have got as a difference between the energy of two free hydrogen atoms and the energy of a molecule. You are right, it should be called formation energy, not dissociation energy. But then dissociation energy defined as the energy necessary to separate H2 molecule should be higher as the height of the activation barrier should be added to the formation energy. However it is lower (4.5 eV). Something is still wrong with my understanding.

Could you also explain what is 1.1188 eV?

[radny]

There are two ways of doing the calculations.
1. Calculate the energies of H2 and H+H in the same box seperately and then calculate the B. energy as E(H2) - E(H+H), or
2. Calculate the energies of H2 and H in the same box seperately and calculate the energy from E(H2) - 2E(H).

For large systems these two ways of calculating the b. energetics are not the same.

[Franky]

Dear Pavel,

please switch on spinpolarization for the isolated atom using:
ISPIN = 2

If you say you did so and the binding energy didnt change, something is very wrong !
The total energy of the H-Atom changes by 1 eV after turning on spinpol.,
so the binding energy changes by 2 eV !
{ E_total(H2) - 2*E_total(H; Spin) = E_binding ~ - 4.6eV}

@radny:

I would be very worried if the calculation using on H-atom per cell results in a
different total energy than 0.5*E_total(H+H) using a cell that is twice as
big as the first on containing two seperated H-atoms!

If you use the same setup as for one H-atom per cell, you might want to think
about changing the occupation numbers manually for the 2 H-atoms. Otherwise, you
are not calculation the total energy of two individual H-atoms!

[yzt102]

I calculate the energy for a sigle H and it comes out to about -0.9eV.
How did you guys got 1.1188eV, Here is my INCAR

SYSTEM = Hydrogen absoption
LWAVE = .F.
LCHARG = .FALSE.
ISMEAR = 0
SIGMA = 0.1
IALGO = 48
LPLANE=.TRUE.
#NPAR = 2
#NSIM = 2
ENCUT=500 eV
NELMIN = 6 ! do a minimum of ten electronic steps
EDIFF = 1E-6 ! default accuracy for electronic groundstate
EDIFFG = -0.02 ! tolerance for ions
NSW = 0 ! 80 ionic steps should do
SMASS=0.5
IBRION = 3
POTIM=0.05
ISPIN=2
MAGMOM=1

[Franky]

I used this INCAR:
System = H-Atom
GGA = 91
PREC = High
LREAL = .FALSE.
ENCUT = 520
ENAUG = 1000
ISPIN = 2
NBANDS = 4
ISMEAR = 0
SIGMA = 0.10
ALGO = Normal
NSW = 0
NELM = 60
EDIFF = 1E-7

KPOINT:
Automatic
0
Gamma
1 1 1
0 0 0

POSCAR:
cubic H
1
10.0 0.0 0.0
0.0 10.0 0.0
0.0 0.0 10.0
1
cartesian
0.0 0.0 0.0

[Franky]

OUTPUT:

FREE ENERGIE OF THE ION-ELECTRON SYSTEM (eV)
---------------------------------------------------
free energy TOTEN = -1.120607 eV

energy without entropy= -1.120607 energy(sigma->0) = -1.120607

[wlyim]

Dear all,

I used PAW-PBE for hydrogen (planewave cutoff = 500 eV, charge density cutoff = 800eV) and the calculated bond dissociation of H2 was 4.54 eV. The trick is among the occupation of the atomic species. Try to use "ISMEAR = -2", "SIGMA = 0.0", and "FERWE" and "FERDO" tags. This trick is also crucial to get correct BDE's for other diatomics.

[Franky]

Using FERWE and FERDO to choose a magnetic groundstate is crucial for certain atoms (see manual).
But not for hydrogen since there is only 1 electron.
For H2 it is certainly not an issue because of the unpolarized
groundstate. The H2 / H results (atleast mine) are independent of the ISMEAR tag with an appropriate SIGMA. The crucial point is to closely watch the occ. numbers being: up=1 0 0 ... down=0 0 0 ... for H.
ISMEAR = 0 and SIGMA = 0.1 was sufficient to accieve this.

And of course to do convergence tests using different ISMEAR and SIGMA.

By the way, in order to compare your results to the exp. value (~4.52eV) you have to include zero point vibration of H2.
These weaken the theo. bond by ~0.27eV !

[wlyim]

Hi all,

I am sorry that my previous reply may be confusing. Anyway, for some cases, like Cl atom, it is crucial to check the occupation and "SMEAR=-2" can help.

Referring to Pavel's question, I have tried to reproduce the results. To calculate the dissociation energy of H2, H2 should be calculated using spin-unpolarized, while a spin-polarized calculation for the H atom.

It should be mentioned that using spin-polarized calculation for H2 with smearing, it would run into a wrong electronic state (both spin up) and the hydrogen atoms will be separated. I and my co-worker have experienced this before, no matter what kind of basis set and functional. So, one has to check the electronic occupation very carefully when using open-shell calculation for H2. If I understand Pavel correctly, Pavel got a dissociation energy of 0.000x eV, which was a consequence of the open-shell H2 I mentioned.

[yzt102]

I calcualted the H-H bond energy with the INCAR I posted here on last wednesday. It turned out that the bond energy for PBE is 4.5 eV and LDA is 4.85 eV. Could different XC function resulted in such a huge difference?

SYSTEM = Hydrogen atom
LWAVE = .F.
LCHARG = .FALSE.
ISMEAR = 0
SIGMA = 0.1
IALGO = 48
LPLANE=.TRUE.
#NPAR = 2
#NSIM = 2
ENCUT=500 eV
NELMIN = 6 ! do a minimum of ten electronic steps
EDIFF = 1E-6 ! default accuracy for electronic groundstate
EDIFFG = -0.02 ! tolerance for ions
NSW = 0 ! 80 ionic steps should do
SMASS=0.5
IBRION = 3
POTIM=0.05
ISPIN=2
MAGMOM=1

POSCAR

H atom
1.000000000000000
20 0 0.0000000000000000
0 20 0.0000000000000000
0 0.0000000000000000 20
1
Direct
0.5 0.5 0.5

[Franky]

yes. But a diff. <10% isn't huge I think.

(see eg. D. C. Patton et. al., PRB 55, 7454 (1997) )

[tak]

Dear Admin and VASP user,

I thought here is the better place to post my question rather than establish new thread.

I’m dealing with H2 and H for the part of my study, formaldehyde reaction on Fe surface. I have total energies of -6.677 and -1.117eV for non-spin polarized H2 and spin polarized H, respectively using PAW-PBE. Based on these values, I came up with bond dissociation energy of 4.44eV. It looks these energies are not far from other DFT and experimental studies, however my calculation does not include zero point energy, which is ~0.266eV with calculated frequency of 4295cm^-1.

I wonder if other studies, such as Kresse and Hafner (Admin?) (Sur. Sci. 459 (2000) 287, 4.58eV) and Jiang and Carter (PRB 70, 064102 (2004), 4.54eV) included zpe.

I appreciate anyone’s comment.

Thank you very much.

[admin]

please let me add 2 comments
tak: no, zero point energies are not included in vasp results
wlyim: please have a look into some basic quantum mechanics textbook: the ground state of H2 is the spin singlet state, therefore you will always obtain the stable ground state solution for H2 with a non-spinpolarized run, the spin-polarized solution (triplet state) is the ground state of 2 H atoms in the box, not matter which functional you choose.