calculations of a charged molecule with spin

[imcnab]

Dear All,

I would greatly appreciate advice.

I wish to use VASP to calculate the potential energy curve of HCl^2+, which is a molecule with charge +2, and a triplet ground state (2S+1 = 3).

Charge was set at +2 using NELECT=16

To set a multiplicity of 3, I think I need to set ISPIN=|2|, and therefore
NUPDOWN = +2 or NUPDOWN = -2.

I tried calculating both NUPDOWN = +2 and NUPDOWN= -2, and to my suprise, I get different energies.

NUPDOWN= +2 gives 56.178107 eV
NUPDOWN= -2 gives 56.338249 eV

Calculations were made in a cubic supercell, with IDIPOL=4, and LDIPOL=.TRUE.

Is NUPDOWN positive or negative to be preferred? Why?

Are there any other crucial settings that I have missed?

All help will be most gratefully received.
best regards... Iain

[ehftz943315]

If NUPDOWN is positive, unit cell (= molecule in your case) multiplicity is kept FIXED. Therefore, for NUPDOWN = +2 there are always 2 alpha electrons more, than beta electrons at each iteration.

It seems like for negative NUPDOWN its absolute value might set only INITIAL value for difference of the number of alpha and beta electrons. Then VASP will allow optimization of number of electrons with different spin projections.
Alternative variant is that NUPDOWN = -2 is treated the same as NUPDOWN=-1, just because NUPDOWN<0 are treated the same way in the code. The final result ideally should be the same: the difference of occupations for alpha and beta spins gets optimized during iterations (= full relaxation of electron wavefunction).

I would suggest to compare spin states at the end of your calculations. You might want to run the same calculation with NUPDOWN=-1 to see effect.

[imcnab]

Thank you very much for the swift reply.

I will check the results with NUPDOWN=-1 to check the effect.
As the triplet state IS the ground state, it would be nice if that was what VASP found automatically.

best ... Iain