Singlet and Triplet states Calculations

Message

Neutrino · #1 Post by **Neutrino** » Sun Jul 10, 2011 2:44 pm

I'd like to compute the total energy of an oxygen atom in the open shell singlet, closed shell singlet, and triplet states.

I assume that if I run non-spin polarized calculations, I will obtain closed the shell singlet state. Also, I assume if I run a spin polarized calculations, I will obtain the triplet state.
My questions are:
1-Are my assumption right? If not what is the right scenario?
2-How to compute the open shell singlet state?

Thanks in Advance!

Neutrino · #2 Post by **Neutrino** » Tue Jul 12, 2011 6:54 pm

I'm also curious to compute the three solutions when the oxygen atom is inserted as interstitial in an Oxide. Would it still be possible to compute the three solutions.

Neutrino · #3 Post by **Neutrino** » Tue Jul 12, 2011 7:34 pm

So, I have just figured out that one can set NUPDOWN to force certain state.

So, if I understood correctly
NUPDOWN =0 should correspond to a singlet state
NUPDOWN=2 should lead to a triplet state

The question now is how to distinguish between the open shell and the closed shell for the singlet state.
Your hel is appreciated!

mrossi · #4 Post by **mrossi** » Thu Jul 14, 2011 12:06 am

I recently dealt with a problem similar to this, however, I was not using VASP. What you need is some way to turn on multi-configurational electronic states, such as using MCSCF or the like. I was studying ZrI3+ZrI3 in the open and closed shell singlets, and had to use MCSCF. I'm not sure if there's a way to set up this kind of calculation in VASP, but that's what you need.

If VASP isn't capable, I recommend using GAMESS, which is free, and has quite a comprehensive user manual.

Neutrino · #5 Post by **Neutrino** » Thu Jul 14, 2011 4:08 pm

It would be appreciated if the admin or any VASP guru can comment on the possibility of using VASP to compute the open shell case.

boris · #6 Post by **boris** » Thu Jul 14, 2011 7:04 pm

Hi

You can try to force a specific configuration using the magmom tag.

For instance, if you want to compute the triplet state, you set MAGMOM=2, which is the number of unpaired electrons on the 2p orbitals. Together with NUPDOWN=2, you should get the correct answer.

The magmom tag only acts as an initial guess though, so the algorithm may change it.

I don't think there is a way in vasp to set electronic occupancies by hand.

A useful trick to check the result is to activate the LDA+U approximation and to set U and J to zero (so you end up with a regular LDA or GGA calculation). Then you ask vasp to print the occupation matrices for p orbitals and you'll be able to check in what state you're in.

Regards

<span class='smallblacktext'>[ Edited Thu Jul 14 2011, 08:39PM ]</span>

Neutrino · #7 Post by **Neutrino** » Thu Jul 14, 2011 9:41 pm

Hi boris,

Thank you for your response. I like the LDA+U trick to trigger VASP to print the occupation matrices.

However, regarding MAGMOM, it cannot really help in obtaining the sinlet open shell. Because "both" the open shell and closed shell singlet will correspond to MAGMOM =0 , so there should be another way to distinguish between the two solutions.

Regarding whether VASP can distinguish between the two cases, I found the paper:
A. Sokol, A. Walsh, C. Catlow, Chemical Physicas Letters, 492 (2010) 44-48.
in which the authors mentioned using VASP to compute all the three states I mentioned above.Also, they did not mention applying any modifications to the code.

So, the remaining question is how to use VASP to compute the singlet both open shell and closed shell?

boris · #8 Post by **boris** » Thu Jul 14, 2011 10:22 pm

What do you mean by open shell and closed shell?

What would be the corresponding electronic occupancies?

Neutrino · #9 Post by **Neutrino** » Thu Jul 14, 2011 10:28 pm

A neutral oxygen atom (whether free or interstitial) will have the electron configuration:
1s2 2s2 2p4
The 4 electrons in the p shell can be in three states:

Triplet: (up-down) (up- ) (up- )
Singlet open shell also called spin unrestricted: (up-down) (up-) (down-)
Singlet closed shell also called spin restricted: (up-down) (up-down) ( - )

The total magnetic moment for both of the singlet states is zero.

boris · #10 Post by **boris** » Thu Jul 14, 2011 10:31 pm

Oh ok I see.

Then a non spin polarized calculation will probably give you the spin restricted result.

As for the other one, I have no idea. There might be a keyword in vasp that allows you to set by hand the band occupancies for each point (only one in your case). Have you searched the manual?

Neutrino · #11 Post by **Neutrino** » Thu Jul 14, 2011 10:38 pm

Yes, I tend to believe that ISPIN=1 calculations lead to spin restricted results.

So, far I could not find any hint in the manual about the singlet open shell and I hope to get some help here!

boris · #12 Post by **boris** » Thu Jul 14, 2011 11:34 pm

If you don't switch symmetries off in the calculation, the ISPIN=1 calculation will give (up-dn)(up-dn)(up-dn) with up=0.666 electrons.

VASP will automatically fill the three p orbitals with the same amount of electrons because of the degeneracies.

Neutrino · #13 Post by **Neutrino** » Fri Jul 15, 2011 12:09 am

By switching symmetries off do you mean setting ISYM=0 ? I'm aware of the trick of using as asymmetric as possible cell for the oxygen atom. But this does not work for bulk calculations, so do you suggest ISYM=0?

Neutrino · #14 Post by **Neutrino** » Fri Jul 15, 2011 12:18 am

And when symmetry is switched off, does that lead to (up-dn) (up-dn) (-)?
Thank you very much for help!

alex · #15 Post by **alex** » Fri Jul 15, 2011 11:16 am

Hi Neutrino,

the way to go for the open shell singlet is to do a triplet calculation, save WAVECAR and CHGCAR, then put NUPDOWN = 0 into INCAR and restart reading the wavefunction. You'd have to check the DOS if you really ended up with an open shell singlet, or if it converges to degenerated alpha and beta anergies.

Cheers,

alex