[Pw_forum] Problems with EXX calculation and q-points

Lorenzo Paulatto lorenzo.paulatto at impmc.upmc.fr
Thu Jan 14 13:35:46 CET 2016


On Wednesday, January 13, 2016 07:05:22 PM Francesco Macheda wrote:
> Dear all,
> I'm new to Quantum Espresso. I'm trying to get the energy bands from
> Silicon bulk with hybrid functionals. I've read in previous posts and in
> quantum documentation that "Their value (of the q-points) must be
> compatible with the k-points used (that is k+q must be equivalent to some
> other k in the k-points list)". I'm not certain to understand why it has to
> be done this way. 


The "correct" way to do EXX is that the q grid is identical to the k grid, in 
this case every k+q point is equal to another k' point eventually plus a G 
vector of the reciprocal lattice. Often a reduced grid of q points can be 
used, but in order to have a reasonable self-consistency procedure you still 
need k+q = k'+G. Try to do it, and you will see why :)

> Moreover, and this is the main problem, I don't know how
> to make them compatible. Once I have chosen a certain dimension of the
> Monkhorst-Pack grid (nk1=nk2=nk3=K) which values can I give to
> nqx1=nqx2=nqx3 ? Is there an a-priori argument?

You can set nqx* to any integer Q divisor of K, i.e. nqx1=Q with Q*N = K with 
N integer.
 
> Another error message that I get is "Error in routine electrons (1):
>      dexx is negative!   Check that exxdiv_treatment is appropriate for the
> system" in the calculation with B3LYP, while HSE seems working well up to a
> certain cutoff, after which the error appears . I read that somebody solved
> it changing exxdiv_treatment, but I don't if or why I should do that, or if
> it is related to my "manual" choice of q-grid. Any suggestion?


You will have to try both, I guess. 

HTH


-- 
Dr. Lorenzo Paulatto
IdR @ IMPMC -- CNRS & Université Paris 6
+33 (0)1 44 275 084 / skype: paulatz
http://www.impmc.upmc.fr/~paulatto/
23-24/4é16 Boîte courrier 115, 
4 place Jussieu 75252 Paris Cédex 05




More information about the users mailing list