[Pw_forum] Relaxation with Spin-Orbit Coupling

Julen Ibanez Azpiroz julen.azpiroz at gmail.com
Tue Jul 13 11:07:19 CEST 2010


Hello

I would like to make a question to anyone who has experience in using the
spin-orbit coupling (SOC). My problem is as follows:


I relax a structure WITHOUT including SOC by the method damped dynamics:


/
&IONS
 ion_dynamics = "damp",
 pot_extrapolation = "second_order",
 wfc_extrapolation = "second_order",
/
 with the time interval dt=20, and I get a structure where the forces are
<1.0D-4Ry/au. Lets call this structure A.

Now, I include the SOC, and I run an scf calculation with structure A. It
turns out that there are quite big forces, of the order of 0.01Ry/au. The
energy of the unit cell with structure A and SOC included is -207.10202515
Ry.

Then, I try to relax the forces with SOC included with the same method as
before. It relaxes (forces<1.0D-4), but in a structure quite different form
A; lets call this new structure B. The surprising thing is that now, in
structure B, the energy is -207.08231311 Ry, higher than with structure A.
To summarize:

Calculations with SOC included:

          Structure A->  -207.10202515 Ry, forces of the order of 0.01Ry/au
          Structure B->  -207.08231311 Ry, forces <1.0D-4Ry/au

Besides, if I plot the energies of the different structures that go from A
to B when relaxing with SOC included, I see that the energy increases in
each step going from A to B, and the energy of A seems to be in the botttom
of a parabola. So, everything suggests that structure A is the most stable
structure even when SOC is included, except there exist forces. I have
checked the convergence with respect to the cutoff and the k point mesh, and
I have used the same parameters in the calculations without SOC and with
SOC.

I do not know how the forces are calculated when SOC is included, and I do
no know who to trust, the energy or the forces. I will apreciatte very much
any help, thank you in advance


Julen Ibañez Azpiroz

University of the Basque Country
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.democritos.it/pipermail/pw_forum/attachments/20100713/9ca44b40/attachment.htm 


More information about the Pw_forum mailing list