[Pw_forum] how to get U in DFT+Hubbard U

[Matteo Cococcioni]

Dear Agostino,

I will try to answer to your questions.

the fact that the electron never localized is not due to a "real" 
electronic coupling between the two Fe (which are important at short 
distances of course), but mostly to self-interaction.

Yes, you should use two different U ideally self-consistently computed 
with the water molecules relaxed in the configuration with the electron 
localized on the 2+ Fe. However it may turn out difficult to get this 
result with LDA+U only because LDA+U is not designed to cure 
self-interaction problems. You may need Us as large as 10 eV to get that 
result.
It could be useful for you to read this paper:

Phys. Rev. Lett. 97, 028303 (2006)

Hope this will help.

Matteo


Agostino Migliore wrote:
> Hello
>
> I am working on electron transfer in a solvated Fe2+/Fe3+ system. The
> electron is (or "should be") always (i.e. for each water configuration)
> essentially localized on one site, but a small portion (increasing near
> the transition state) must be on the other site, just because of a nonzero
> (although small) electronic coupling between the two sites. So, I wonder
> and ask
> if I should use two different U numbers for the two ions (e.g., that
> obtained for the isolated oxidized (3+) site and the one for the reduced
> site (2+), after check, without U, about the site that should turn out to
> be essentially reduced/oxidized) or if I should use an average U for both
> (neglecting the differences due to different water arrangements) and
> "leave the transferring electron to reach its state", without biasing its
> localization through different U's.
>
> Regards,
> Agostino Migliore
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>   

-------------- next part --------------
A non-text attachment was scrubbed...
Name: matteo.vcf
Type: text/x-vcard
Size: 294 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20071017/364bc49d/attachment.vcf>