[Pw_forum] how to determine the value of U in a lda+U calculation

[Matteo Cococcioni]

Dear Du,

the input you found below is to extract the U values from a series of 
scf calculations.
You can give a look (besides the reference cited below) to the tutorials 
you find in this webpage:
http://www.quantum-espresso.org/wiki/index.php/QESB09.
Unfrtunately the value of U is generally not portable and changes 
depending on the invironment of the atom in the
specific crystal you are considering. the one you cite is a good 
example: Fe in bulk Fe and in FeO.
So in principles we need to recompute it each time.

hope this helps.

Matteo


duchl06 wrote:
> dear pw users
> i want to do a calculation of V chain absorb on a plane with DFT+U, 
> but i donot know how to determine the value of U.
> i found a PPT on the pwscf's media site, they said the value of U 
> could be decised by a linear response approach.
> but i really donot understand that like the input file resp_mat.in
>
> &input_mat
> ntyp = 1 # number of types of atoms
> na(1) = 2 # number of atoms per type
> if we have different type atoms with different number, how we set these?
>
> nalfa = 5 # number of perturbations (alpha)
> filepos = 'pos_sc1' # file containing the atomic positions and the 
> vectors of the unit cell
> back = 'neutral' # to add a neutralizing ¡°background¡±. see PRB 71 
> 35105 (2005)
> what does this mean
> filednda = 'file_sc1' # file containing the names of the files dn*dat
> n1 = 4
> n2 = 4 # number of unit cells in each direction to construct the 
> supercell for
> n3 = 4 # extrapolation (in this case a 4x4x4 supercell of the SC one)
> &end
>
> and they get the U is about 2.4ev for the iron, and they use 4.3ev for 
> the FeO calculation,
> we always need a linear response approach to get the U for every system
> or just a U for a certain pseudopotential (i mean if we get a U for V 
> bulk, then if i could use this value for another system with V atom, 
> like V2O3 etc.)
> any suggestions or comment would be OK.
> thanks
>
> Du
> xiaguanying university
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>