How to specify the magnetic moment of each atom?

Matt W MattWa... at
Thu Feb 26 10:16:42 UTC 2015


you can get what you are after by using the broken symmetry (&BS) keyword 
within the &KIND block. It is a bit fiddly to use, but can be used to 
specify the atomic guess used to generate the initial density matrix. The 
are some threads in the forum and some stuff online if you look for it.

But, my preferred way of _trying_ to localize the spins is to use the 
&CORE_CORRECTION keyword, again in the &KIND section. So in this case pick 
the two Ce that you want to localize the electrons onto and make them a new 
kind in your geometry file (Ce3 or something). Then in &KIND Ce3 add 
&CORE_CORRECTION 0.5 or 1.0.  This increases the effective nuclear charge 
of these atoms and makes it very likely the electrons will localize on 
them. You can then do a rough geometry relaxation to get the lattice 
distortion that might stabilize the localized electrons. Then you can 
restart but setting the &CORE_CORRECTION back to 0.0. If you are using an 
appropriate density functional then you have a decent chance of getting the 
localization you want. For the O vac in CeO2 you can just run a triplet 
calculation with the modified core charge on a couple of Ce and you should 
get what you want without bothering with the broken symmetry guess.


On Thursday, February 26, 2015 at 6:23:01 AM UTC, stclair wrote:
> Dear CP2K user,
>     I am working on CeO2 (111) surface with oxygen vacancy. Based on the 
> previous studies, two extra electrons should locate on two next-nearest Ce 
> atoms (as shown in the attached pictures). But my calculation converges to 
> a different electronic configuration where the extra electrons locate on 
> the Ce atoms closest to vacancy. Is there any keywords I can use to specify 
> the magnetic moment for the atoms? Some keywords are similar to MAGMOM in 
>    Thanks in advance.
> Best,
> Lei
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>

More information about the CP2K-user mailing list