My Community

I put a H2 in a supercell .When I use ISPIN=1,I can get correct H-H bond lengh, but when I change ISPIN =2, the calculation will result in two hydrogen atoms not a hydrogen molecule . PAW-PBE and USPP-PW91 both gave such a result. I want to know if I can calculate H2 with ISPIN=2,because I will calculate the adsorption of hydrogen molecule on some surface with ISPIN=2.

the spins of the ground state of the H2 molecule are paired, and this ground state should be taken as reference for the adsorption. If you enforce spin-polarisation, you will necessarily end up in a non-ground state geometry.

If this is true, then how can one do a spin polarized calculation for an H2 molecule sitting in vaccum, say 5 to 10 angstroms above a "semi infinite surface" -in other words a slab, perhaps with bottom layer(s) frozen- all in the same supercell where there are atom(s) in the slab that require a spin polarized calculation to get the physics correct? This sort of set up is a reasonable ending point for a nudged elastic band calculation using VASP and UofT Austin's NEB code.

Head admin, When you said "enforce spin-polarisation" did you mean set NUPDOWN=2 for this problem, or did you just mean let NUPDOWN (end everything else) just be default and only set ISPIN=2?

Correct geometry and energy of H2 is obtained using ISPIN=2 and NUPDOWN=0.
When putting H2 on the surface pehaps you can tune the correct spin state using NUPDOWN.
[ Edited Mon Oct 01 2012, 11:27AM ]