[Pw_forum] spin splitting

[Gabriele Sclauzero]

vtmtrinh at caltech.edu wrote:
> Hi Gabriele,
> 
> Thanks for your reply. I am sorry for the unclear question.
> The question is to calculate the splitting between the spin up and spin
> down states in the conduction band (CB) of Si. For equlibrium Si bulk,
> there should be no spin splitting. Therefore, the splitting can be induced
> by moving the atoms a long phonon modes. The spin splitting is probably of
> the order ~1e-7eV.
> 
> Yes, you are right.  I need to set nspin=2.  Since I also need to include
> the spin-orbit coupling in this calculation, I have to use
> noncolinear=.true.

Now the situation is much more clear :-)


> 
> I run a calculation yesterday, in which I set lspinorb=.true. and
> noncolinear=.true. By setting this I think that spin-polarized calculation
> was aslo done.  However, in the output, I don't see the eigen values for
> spin up and spin down, but only one (like for the non-polarization case). 

You're right, in a noncolinear calculation you don't see SPIN UP and SPIN DOWN bands as in 
a LSD calculation.
Anyway, as you also noticed, the number of bands is doubled when you set noncolin=.true.
If you don't include SO coupling, you should get (for each band of the non spin-polarized 
calculation) a spin up and a spin down band which coincide with those obtained with a 
collinear LSD calculation (i.e. with noncolin=.false., nspin=2). Also the final total 
magnetization should be the same (you can check from the output).
You can plot the bands (for instance using bands.x and then plotband.x) and compare the 
splittings.
When you include SO the situation is more complicated (you should label your bands with 
the irreps of the (magnetic) double group of the system), but in principle more close to 
reality.
You can plot bands also in this case and see what changes and how much (tyr lsym=.true. 
when running bands: it can label bands according to the irreps and it may be very useful).
The SO splitting would be small, though. In the Si atom the SO splitting of 3p electrons 
should be of the order of few mRy (which is only slightly larger than the error from PPs, 
so be careful...), and in the bulk they are usually lower.


Cheers,

GS


> I am not sure if I did not set the right parameters.  Attached is my
> input.  Could you let me know what wrong with it.
> 
> Thanks,
> 
> MyTrinh
> 
-- 


o ------------------------------------------------ o
| Gabriele Sclauzero, PhD Student                  |
| c/o:   SISSA & CNR-INFM Democritos,              |
|        via Beirut 2-4, 34014 Trieste (Italy)     |
| email: sclauzer at sissa.it                         |
| phone: +39 040 3787 511                          |
| skype: gurlonotturno                             |
o ------------------------------------------------ o