[Pw_forum] LDA+U in md calculations

Matteo Cococcioni matteo at umn.edu
Fri Apr 2 18:56:46 CEST 2010

Dear Jiayu,

are you talking about Car-Parrinello MD or Born-Oppenheimer MD (run with 
In any case lda+U MD is implemented in the sense that the code computes 
the forces and the stress
as derivatives of  the lda+U energy functional. However it is not able 
to account for the variation
of U with atomic position or with the cell. So it is a dynamics at fixed U.
To account for this variation one would need to evaluate dU/dR or 
dU/d\epsilon (R and \epsilon being atomic
positions and strains). Some work in this direction has been recently 
done (not sure it's
published yet) by Kulik and Marzari.
If the structure you are interested in doesn't change dramatically 
during the MD run I would expect
that keeping the same U is not too bad. but this is just my guess.



Jiayu Dai wrote:
> Dear developers,
> I have some doubts about the molecular dynamics calculations using LDA+U. First, is LDA+U performed in MD calculations? Because i did not find the Hubbard energy in the output file of MD. Second, i think the value of U be dependent on the temperature and density, also on the ionic structures. Therefore, since ionic positions are different in each time step, is U different at each time step? If so, the performance of LDA+U should be very difficult in MD.
> I did not know if what i said is right. Hope experts comment on this problem.
> Thanks a lot.
> Jiayu
> -------------------------------------------
> Jiayu Dai
> Department of Physics
> National University of Defense Technology, 
> Changsha, 410073, P R China
> -----------------------------------------
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum

More information about the users mailing list