[Pw_forum] Relaxation of AlN surface

[Paolo Giannozzi]

lan haiping wrote:

> How do you think about the thread for electron scf : conv_thr = 1.D-6 ?

I think it is quite good. Consider that it will be reduced during
structural optimization. What is meaningless (way too small) is
the following:
    etot_conv_thr = 1.d-8
This is however irrelevant for scf convergence.

> And only gamma k-point sampling for your cell seems not enough .

gamma may or may not be sufficient; I would try a denser grid in
the xy direction. The main problem in this case is however that
the system is treated as an insulator. It is sufficient to add
a few more bands and a broadening, as explained by Eyvaz

Paolo
(apologies if this message arrives twice: there was some problem
on the mailing list this morning)
-- 
Paolo Giannozzi, Democritos and University of Udine, Italy