[Pw_forum] calculating U

Matteo Cococcioni matteo at umn.edu
Thu Jan 8 16:22:49 CET 2009


Dear Naol.

you can find many details about this in this paper: PRB 71, 035105 
(2005). I will try to answer to your questions below.

Naol Regassa wrote:
> Dear all
>
> I was calculating Hubbard U for strongly correlated perovskite like
> SrMnO3. I can run scf and scf+u, perturbed for different
> Hubbard_alpha. My question is
>   

those with alpha are just "fake" lda+U calculations: lda_plus_u needs to 
be set to .true. (so that the code will write occupations on the output 
file)
but U is very small (practically 0; normally a value of 1.d-10 eV is 
chosen) because we usually perturb LDA or GGA ground states.

> 1- Do we check the convergence of alpha? If yes with what parameter?
>   

I'm not sure what you mean by convergence of alpha. It's a good habit to 
make sure that the total occupation of each site has a linear dependence 
on alpha around alpha=0.

>  2- How do we find the value of U? What are the important parameter
> there(in the perturbed outputs)?
>
>   

This requires constructing response matrices and a little bit of 
processing of the results you obtain from the code. U is not something you
directly read from any ourput file of the code.

> 3 - Is it needed to construct super cells so that we can check the
> convergence with them?
>
>   

Yes. again please check the reference above.

keep in mind that in the perturbed run diago_thr_init needs to be set 
equal to the threshold of the iterative (davidson) diagonalization of 
the last
electronic iteration of the unperturbed run. see the example input 
attached for details.

> great thanks for your response
> Thesis student
>   

would you please add your affiliation?

regards,

Matteo

> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>   

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: run_Fe_ucalc_example
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20090108/7b4b867f/attachment.ksh>
-------------- 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/20090108/7b4b867f/attachment.vcf>


More information about the users mailing list