[QE-users] Non-converging vc-relaxation calculation

Lorenzo Paulatto paulatz at gmail.com
Thu Jul 16 10:11:25 CEST 2020


> 
> However, I'm not sure on how to do either. Do i just increase the nbnd 
> number and increase the degauss value?
> 

Hello, the value of smearing you are using is quite large, I would try 
to reduce it first. Also, as you have not done it already, reduce 
mixing_beta (maybe 0.3 or 0.1) and try a Thomas-Fermi type mixing 
(mixing_mode="local-TF")

However, if the magnetic state you are asking for is not the ground 
state one, converge may be very difficult. This is because you are 
trying to converge to a local minimum, which may be shallow. If your are 
scanning all the possible magnetic states, you can just throw this away. 
If you want this specifically, and you fail to converge it no matter 
what, you may have to use constrained magnetization or DFT+U (but then, 
comparing the energies of different magnetic states will be very tricky).

Please note that "crystal_b" is not a valid option for the atomic 
positions, I think the code just interprets it as "crystal", but you may 
want to check.

cheers


-- 
Lorenzo Paulatto - Paris


More information about the users mailing list