[Pw_forum] Huge imaginary and positive frequencies

[Lorenzo Paulatto]

Hello,
> K_POINTS automatic
>   27 27 3 0 0 0


If I understand correctly your material is a monolayer, in this case I 
would not use more than one k-point in the z direction, as there isn't 
interaction between periodic copies

Also, if your material has a Dirac cone or similar feature in the 
electronic bands dispersion, I would use a grid that avoids including 
it. I.e. in the case of graphene, a grid size which is NOT divisible by 3.

>    ! tr2_ph=2.0d-8, ! for the stubborn dyn2

This will just give random results, if it does not converge with a 
stricter threshold there is some underlying problem.

Apart from changing the k-points grid, you may have a shot at trying 
different pseudopotentials. The AK ones (from Axel Kohlmeyer) aren't the 
most commonly used and if any of them give some problems there is a 
chance it went undetected. You may try the GBRV library 
https://www.physics.rutgers.edu/gbrv/
or have a look at the recommendations from the Marvel group 
http://materialscloud.org/sssp/

hth


-- 
Lorenzo Paulatto - Paris