NUPDOWN in spin polarized calculations

[KWH]

I am doing calculations for desorption of a cobalt porphyrine from graphite. The porphyrine should be a doublet in the gas phase. I find that the energy of the spin polarized calculation on the surface depend strongly on what parameters I default to. I have appended INCAR, CONTCAR, OSZICAR, and OUTCAR files for a few cases.

Spin unpolarized -1972.3696
Spin 1/2 -1972.3657 (Output and INCAR in All_3 folder) ISPIN=2 MAGMON set NUPDOWN = 1
No Magmom -1972.3742 ISPIN = 2 NUPDOWN =1 (No Magmom specified)
No NUPDOWN -1972.4873 ISPIN = 2 MAGMOM set (NO NUPDOWN specified) (OUTPUT and INCAR as NO_UPDOWN in folder)

For the free molecule, the spin results are all very similar and much more negative than the unpolarized case (probably as expected)

Spin unpolarized -466.5715
Spin 1/2 -466.9651 ISPIN=2 MAGMON set NUPDOWN = 1
No Magmom -466.9649 ISPIN = 2 NUPDOWN =1 (No Magmom specified)
No NUPDOWN -466.9651 ISPIN = 2 MAGMOM set (NO NUPDOWN specified) (OUTPUT and INCAR as freeMolecule in folder)

Not only are the energies different for the surface case, the magnetic moments are also quite different. Since I wish to compute a desorption energy and they vary considerably with which type of calculation I choose, , I need to know which type should give the best result.

How should I be running these calculations?

Best wishes

[manuel_engel1]

Hi,

Thanks for reaching out. This is a rather complex scenario, since the difference in numbers that you report could stem from various sources. I would ask you to upload also the other input and output files according to our forum posting guidelines in a zip archive.

[KWH]

I have attached the files requested. When you look them over, please also consider he fact that adsorption appears to be quenching the effect of magnetic moment.

Best wushes

[manuel_engel1]

Thanks for uploading the files. It seems that this could be a convergence problem. First of all, you perform ionic relaxation in each of these cases so you end up with a slightly different structure each time. You could try to compare the energies with the same structure. You can also try to run a more accurate calculation without ionic relaxation to get more accurate total energies. Finally, it may just be that you end up in different magnetic groundstates in each case that results in a small difference in total energies.

Here are a few settings you can try to improve accuracy of your calculation

• Use PREC=Accurate
• Use a more accurate (smaller) EDIFF
• Increase ENCUT
• Vary POTIM
• Vary EDIFFG
• Use ALGO=All which can help with magnetic calculations

[KWH]

Thank you for your helpful suggestions. Using the parameters you suggested, and I would like to make sure my input files and results are appropriate. The system I am studying is a cobat(II) porphyrin on hopg with 2 molecules per unit cell. Since each molecule has spin 1/2, the ground state of the cell might be either spin paired (S=0) or unpaired (S=1).

I ran both calculations (files attached) and found a 200 mV difference in energy with the singlet laying lowest. Since this calculation does NOT include LS coupling, is it reasonable to have such a large difference based only on spin state?

Since the other parts of the adsorption energy (CoOEP in gas) and HOPG (only) are the same, this represents a 200 mV difference in adsorption energy. Can I believe the paired energy result and the implication that the surface has no net magnetic moment?

Thank you for considering my question.

[marie-therese.huebsch]

Hi,

I haven't looked at all the details of the calculation and I cannot comment on the interpretation of the result. However, I did notice that you are not using dipole corrections. These are generally necessary for the desorption of an adsorbent from a slab. In simple words, the dipole causes an electric field across the vacuum that causes long-range interactions and instead of simulating an isolated surface, you are simulating a stack of dipolar surfaces. This yields different potentials, total energies, forces, etc.

You can/should start from the converged calculatuions you have now and switch on dipole corrections. See Electrostatics, DIPOL, IDIPOL, LDIPOL. You may consider centering the system in the vacuum or be extra careful when setting the center of the dipole.

Let me know if this helped!
Cheers,
Marie-Therese