[Pw_forum] could pwscf do a Geometry Optimization calculation in LDA+U Scheme

[Stefano de Gironcoli]

Dear Du,


duchl06 wrote:
> dear pw users
> could any one tell me which method is right?
>   1)relax in GGA scheme, then use the  Geometry do a GGA+U scf calculation
> or
>   2) relax in GGA+U scheme, got the Geometry, then a GGA+U scf calculation
>   
The consistent way of proceeding is 2).

consider however that structural minimization may suffer from the 
problem of having many local minima and typically one is interested in 
the lowest energy one.
the two different results may therefore represent two different minima 
in your system landscape, reached because the history of the relaxation 
has been different for some random reasons.
To check on that you might start a new GGA+U relaxation from the 
converged GGA geometry and see whether the system evolves toward the 
geometry you have found in the GGA+U relaxation, suggesting a real 
difference between the minima of the two methods, or remains in the 
neighborhood of the GGA minimum, suggesting that both minima are 
actually local ones for GGA+U.
The ultimate decision on which minimum is better would then be taken on 
the basis of their energy .. paying attention to compare apples with 
apples, that is comparing GGA+U OR GGA energies among them and NEVER a 
GGA energy with a GGA+U one ...

hope this helps,

stefano


>  
> I ask this question because i did a Geometry Optimization GGA scheme and  GGA+U (other parameters are almost the same), and i got a very pwdifferent Geometry with the two method, why does this happen?
>  
> best regards!
>  
> Du
> xiaguanying university
>   
> ------------------------------------------------------------------------
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>