[Pw_forum] Cutoff convergence

[Francesco Macheda]

Dear all,
I have a question about cutoff convergence of total energy. I run some QE
calculations on Silicon bulk with hybrid functional, and I found that,
fixing the geometry and the number of k/q-points, total energy doesn't
monotonically decrease with the increase of the cutoff. Indeed, there are
some cutoff values which give a poorer description, in terms of total
energy, then some lower cutoff values. I didn't change default convergence
parameters. I stress that each simulation is different ONLY for the cutoff
value. Now my question is: in your opinion, which is the main source of
error? Is it the 30% default mixing value, the finite number of k/q-point,
the really high number of operations involved in the calculations that
afflict the precision, or something else? I ask this because hybrid
calculations are pretty long, and to test every possible source of error
would require a great amount of time.

Thanks for the attention,
Francesco Macheda
Student at Pisa University
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20160222/f0085ce3/attachment.html>