This thread is locked.Only browsing is available.
Top Page > Browsing
Dimer binding energy
Date: 2006/12/11 09:15
Name: Alexandre Tkatchenko

Dear developers:

I have made some comparison of binding energies of simple diatomic
systems with openmx and plane-wave codes.

Using well-tested pseudopotential for iodine atom in the plane-wave
code and converging the results in respect to cell size, cutoff energy, etc.,
I get the binding energy of an iodine dimer of 0.083 Hartree at 2.66 A
interatomic distance.

Using the same system parameters (cell size, scalar relativity, etc.)
in openmx, with the pseudopot taken from the 2006 database and "s3p3d3" basis set,
I get -0.057 Hartree.
With larger "s5p5d5" basis set, the result is -0.067 Hartree. However, the
"s5p5d5" seems to be the largest basis set available.
The pseudopotential doesn't seem to be the problem (the data from the database
seem to indicate that the pseudopot has a good transferability).
It seems quite strange to me that using a large basis set such as "s5p5d5"
does not yet give a converged dimer binding energy.

Do you think I need to use contracted orbitals or what would be the way
to increase the basis set size ?

Best Regards,
Alexander Tkatchenko

Page: [1]

Re: Dimer binding energy ( No.1 )
Date: 2006/12/18 19:37
Name: T.Ozaki


Have you checked the convergence property of the binding energy
with respect to the cutoff radius of basis functions ?
The cutoff radius of basis function is also an important parameter
to obtain the convergent result as well as the number of basis funtions.


Re: Dimer binding energy ( No.2 )
Date: 2006/12/20 07:34
Name: Alexandre Tkatchenko


I have tried the PAO's with 9 and 10 A cutoff with practically the same results.
I get around 0.067 Hartree for the binding energy. Could it have something to do with the system temperature (do I need to do extrapolation to 0K ?).

Re: Dimer binding energy ( No.3 )
Date: 2006/12/20 19:26
Name: T.Ozaki


> Could it have something to do with the system temperature
> (do I need to do extrapolation to 0K ?).

In this case, the electric temperature is not important unless
a too high temperature is used.

If the basis set convergency is properly taken into account for
both the calculations of the dimer and the atom in OpenMX and
the same functional is used in OpenMX and the PW code, the
difference should be attributed to the pseudopotential you used.

To assess which one is more close to the all electron calculation,
only the reference is the all electron calculation in which the
same functional and a basis set enough to be close to the
basis set limit are used.

Do you know the all electron result ?
I have not found any reference by DFT on this issue, but
found the experimental result in

(1) K.P.Huber and G.Herzberg, Molecular Spectra and Molecular
Structure IV. Constants of Diatomic Molecules; Van Nortrand:
Princeton, NJ, 1979.
(2) J.Phys.Chem. 95, 8044 (1991).

The experimental value is 35.6 kcal/mol (=0.056 Hartree).
Thus, we see that the OpenMX result you obtained is close to
the experimental result, while the true value by the LDA (or GGA?)
seems to be unknown.

One thing I can say is that the pseudopotential for iodine in OpenMX
contais the 4d state (I wonder that the 4d state is excluded for the
PW pseudopotential).

Also, I wonder how the dissociation energy is changed by inclusion
of f-orbitals.

If you know another reference, please let us know.


Re: Dimer binding energy ( No.4 )
Date: 2006/12/21 02:07
Name: Alexandre Tkatchenko


I had some references on coupled cluster and CI calculations on the iodine
dimer, but that is not comparable to DFT anyhow. It seems to be a very difficult system to describe. Even the spin-orbit effect contributes largely to the dimer binding energy. Therefore, any agreement of a scalar-relativistic PBE-DFT with the experimental results would be just spurious.

The pseudopotential for PW calculations is well calibrated for many different systems, not only the dimer (I have used it for surface adsorption, etc. and it is in agreement with other recent PAW calculations and experiment). I also have performed some gaussian localized basis set calculations for iodine dimer before and had binding energy values close to PW results for a large basis set. Finally, a 17 valence electrons pseudopot for PW
gives similar results for a large (60 Ry) cutoff in the PW code.
Therefore, I believe that something could be wrong with the pseudopotential in the openmx database. I will try to do calculations for other atoms and see if the problem is specific to the iodine pseudopot.

I will keep you posted on this issue. Anyhow, thank you very much for releasing the code, I found it very easy to compile/use and very complete.

Re: Dimer binding energy ( No.5 )
Date: 2006/12/23 12:59
Name: Alexandre Tkatchenko

Dear Prof. Ozaki:

Little update on the things:

I have performed the same comparison between PW and OpenMX on the carbon dimer and I found very good agreement for the binding energy between the two approaches. However, I had to use a "s4p4d4" basis set for the carbon atom to get it.

However, I still have the same problem for the iodine dimer. I have tried to increase the basis set to "s5p5d5f5", but still get -0.065 Hartrees for OpenMX, which is still too low compared to PW. Therefore, either the OpenMx iodine pseudopotential is not accurate or one needs to go to a very large basis set (which I doubt).

Re: Dimer binding energy ( No.6 )
Date: 2006/12/25 20:23
Name: Vasilii Artyukhov

Actually, it would be very interesting to know how well-tested the database is...
Re: Dimer binding energy ( No.7 )
Date: 2006/12/26 13:54
Name: Alexandre Tkatchenko

Yes, it would be definitely interesting to make a systematic study for the validity of the database. I guess it would be the best to take some large published database of all-electron calculations and compare it to OpenMX (for both distances and binding energies). I do actually have the computational resources for such a study. If someone else is interested, we could probably do that.
Re: Dimer binding energy ( No.8 )
Date: 2006/12/28 00:22
Name: T.Ozaki

Hi Dr. Alex,

Do you know such a large well-organized and reliable database
by all-electron calculations within LDA or GGA ?
If available, it would be very helpful to check the validity
of the database (2006) and polish it up.

Best regards,


Page: [1]