[Pw_forum] Exceptionally large charge density cut-off

[Paolo Giannozzi]

Subhradip Ghosh wrote:

> [..] I find that phonon frequencies at Gamma and [1/2 1/2 1/2] points 
 > do not converge even after a large charge density cut-off although with
> respect to k-mesh size and kinetic energy cut-off,
> the convergence was achieved with reaonable values of these parameters.
> [...] The lowest lying acoustic mode is not yet converged to within 1% 
> while the other modes have converged.
> 
> This is very surprising because when I looked into the archive and the
> relevent literature I didn't find any reference to such a behavior.

you are using ultrasoft pseudopotentials, right? the calculation of
phonon frequencies - especially acoustic phonons - with USPP is quite
tricky. Large cutoffs for the charge density are needed, sometimes
so large to make the calculation impractical. Unfortunately I am not
aware of any remedy. Last time I tried to use USPP in a phonon
calculation, I gave up. I am trying again right now...

Acoustic modes are sensitive to small violations of the Acoustic Sum
Rule (i.e. of translational invariance), and this is well known (even
in the case of norm-conserving PP, where the violation of the ASR comes 
from the discretization on a grid of the exchange correlation
potential). At Gamma this is typically fixed by imposing the ASR on
the dynamical matrix, but this may not be sufficient.

Paolo
-- 
Paolo Giannozzi, Democritos and University of Udine, Italy