Cohesion energy calculation of Cr23C6

[tracy]

Dear All,
I just did a coarse calcualtion of Cr23C6 for non-spin and spin cases. Thanks again to forsdan for his help of the carbides structures!
Now I'm wondering how to calcualte the cohesive energy of Cr23C6 for the non-spin and spin cases.
For example, for the non-spin, I got a energy (sigma->0)in the OUTCAR which is -275.409eV. Is this the cohesive energy of Cr23C6? Or the cohesive energy should be the remainder that I subtract the energy of the references atoms contained in the cell (which should be 23-Fe atoms and 6 C atoms, am I right? )from this energy(-275.409eV)?
How about the cohesive energy of Cr23C6 for the spin pola rized calculation? Thanks for any comments! Have a nice day!
Best,
tracy

[forsdan]

The energy given in VASP is with respect to the non-polarized atoms. The cohesive energy should however always be evaluated with respect to the ground state of the free atoms, so if the free atoms have spin-polarized states then you'll need account for those. In your case both C and Cr atoms are polarized in their free states so you must perform separate calculations for the atoms. The cohesive energy is then defined as

Ecoh = - ( E(Cr23C6) - 23*E(Spin-polarized free Cr atom) - 6*E(Spin-polarized free C atom) )

Then it's possible to evaluate E(Cr23C6) using non-polarized or spin-polarized settings and get the cohesive energy in respective case, but the reference level should still be the spin-polarized atoms in both cases.

PS. If you only want to know the difference in the cohesive energy between the non-polarized and spin-polarized Cr23C6 system (and not the absolute cohesive values), then you of course don't need to perform the separate calculations since the reference level will cancel out DS.

Best regards,
/Dan
<span class='smallblacktext'>[ Edited Wed Jun 10 2009, 09:49PM ]</span>

[tracy]

Thanks a lot to Dan! Could I ask several further questions?
1) where can we find the information about the occupancies of the free atoms (such as the electronic ground state configuration, magnetic moments etc.)?
2) Should we devide the Ecoh = - ( E(Cr23C6) - 23*E(Spin-polarized free Cr atom) - 6*E(Spin-polarized free C atom) )
by 29 (the number of atoms in Cr23C6) for the cohesive energy of Cr23C6?
3) Is it right that we can use the following INCAR file for the free atom energy calculation? we used only one kpoint and the box size is (10,10.25.10.5) For C(s2p2) and Cr(3d54s1).
INCAR: ISYM = 0
ISPIN = 2
VOSKOWN = 1
ISMEAR = 0
SIGMA = 0.00001
AMIX = 0.2
BMIX = 0.0001
NELM = 20
ICHARG = 1
4) Have you ever calculated the spin-polarized Cr23C6 system, I just found out I got the same results for non-polarized and spin-polarized Cr23C6 system, because the spin-polarized case converged to a nonmagnetic state even I set the MAGMOM=23*4 6*0 in the INCAR files. I don't know the reason for this, it seems very weird.
Thanks a lot! Have a nice day and nice weekend!

[forsdan]

1. If you don't have any book containing tables over the elements, you can have a look at http://www.webelements.com , where the ground state configurations and term symbols are given. Please keep in mind that there is a possibility that the ground state configuration for the pseudopotential you're using can differ from the experimental established one. It depends on if another configuration was chosen when generating the potential. I don't remember if this is the case for Cr, so it might be good to check different configurations in addition to the experimental one to see which one gives the lowest energy.

2. Personally I would present the cohesive energy per Cr23C6 unit since the atoms are not the same species, but there are several people that present the average cohesive energy per atom. I guess this is a matter of personal preference. But as long as you explicitly state how you have derived the presented results it shouldn't be a problem.

3. Basically you can use that INCAR file, but the convergence can be quite slow so I suggest to increase NELM. To look at specific configurations can you also use the NUPDOWN tag.

EDIT: You might also need to play around with the mixing tags as convergence can be hard sometimes especially at lower SIGMA. You can try to change AMIX_MAG and BMIX_MAG to linear mixing as well (see the manual). In some cases I found that pre-converging the density at a slightly higher SIGMA first and then restarting with a lower SIGMA can help. Maybe you don't need to go as low as SIGMA=0.0001. Just keep in mind that the energy levels are discrete so check that the filling of the energy levels are integers. Please also read the corresponding section in the manual (if you haven't done so already):

http://cms.mpi.univie.ac.at/vasp/vasp/node230.html

4. A ferromagnetic starting configuration will for Cr23C6 relax to the nonmagnetic solution. This is nothing weird.

Best regards,
/Dan Fors




<span class='smallblacktext'>[ Edited Sat Jun 13 2009, 10:19AM ]</span>

[tracy]

Thanks a lot! I checked the V_RHFIN for paw_GGA of Cr, the configuration is 3d54s1, it's consistent with that one in the Periodic table of the elements. Have a nice day!