Spin-polarized Calculations (and Charge)

[Brian Day]

Hello everyone,

I am curious about the use of spin-polarized calculations in cp2k. My primary question is: Are spin-polarized calculations (UKS TRUE) mandatory to even start a run in CP2k for an odd number of electrons? 

My experience would tell me no, as I have successfully run calculations on systems with an odd number of electrons with this flag off (see attached). However, the reason I ask is because I still occasionally have codes with the error message "ABORT: Use LSD for an odd number of electrons". Ignoring the issue of accuracy for now, if anyone could provide some insight on what yields this error message, that would be greatly appreciated!

Now speaking to the issue of accuracy, is it correct to say that as long as a system is not magnetic, the calculated energies in both cases should be very similar? Almost all of the literature I saw only addressed this type of calculation in regard to magnetic (FM and AFM) systems. Some forums also seem to suggest that this is a correct statement (https://cms.mpi.univie.ac.at/vasp-forum/viewtopic.php?f=4&t=199).

Codes attached:

Mix of geometry optimizations/single point energy evaluations for the xyz file with zero charge (*-engn), plus one (*-ecgn), and minus one (*-eagn) charge. The system ran successfully both with and without the LSD flag set to TRUE.

Thanks!

-Brian

[Krack Matthias (PSI)]

Hi Brian

 

CP2K allows for non-spin-polarized calculations with an odd number of electrons only if smearing is activated. If such a setting makes sense for non-metallic systems is another issue. You perform in that way a kind of constraint ROKS-type calculation, since you have only one set of spatial wave functions. I doubt, that the results, especially for non-metallic systems, will be similar with and without LSD for most of the cases.

 

Matthias

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.

[Brian Day]

Thanks for the reply Matthias! So far this explains everything I'm seeing, I just have to check on whether or LSD False type calculations make sense for my systems. 

-Brian