Spin polarized calculation

[xhsh]

Hi,

It is very interesting to see that even if I set initial_scaled_spin=[0.0]*32 in electronDensityParameters( ), i.e., when I set the initial of the spin of each atom to zero, it still performs a spin polarized calculation.

It seems that, if we want to do a non spin polarized calculation, we should not set the option initial_scaled_spin at all, even if it is 0.0.

Is it correct?

[Nordland]

This is correct. However be warned that initial spin of 0.0 on all atoms is a fixed point in respect to spin, hence if you start with 0.0 on all atoms, you will end with 0.0 on all atoms.

[zh]

However be warned that initial spin of 0.0 on all atoms is a fixed point in respect to spin, hence if you start with 0.0 on all atoms, you will end with 0.0 on all atoms.

I think that this may be not true for all cases.  The initial spin of 0.0 on all atoms does not mean the fixed spin calculations. The proper setup for the initial spin of each atom is important, it may affect the convergence speed of self-consistent calculations, and the obtained magnetic state (e.g., ferromagnetic or anti-ferromagnetic states) of system.

[Nordland]

If the initial scaled spin is zero for all atoms, then the Hamiltonian is the same for up and down electrons. If this is the case, the eigenvalues and eigenstates are the same for up and down electrons, and hence the density matrix will be the same for up and down electrons. Therefore will the real space density also be the same for up and down electrons. This will produce an effective potential for up and down electrons that are exactly the same. And then the Hamiltonian will again be the same for up and down, corresponding to a initial scaled spin of 0.0

Therefore unless numerical noise destroy this cycle, it will never become move away from zero polarization.
However sometimes numerical noise does this, but in principal it should be considered as a fix point.