[Pw_forum] vc-relax

Giuseppe Mattioli giuseppe.mattioli at ism.cnr.it
Fri Oct 20 13:51:37 CEST 2017

Dear Ibrahim
There is a further thing you can do. I see that after 100 steps the  
scf is not converged yet but it is not so far from convergence. You  
can carefully try to force the geometry step by using this in your input


When things work well you displace the atoms along forces that are not  
highly accurate, but you are still so far from convergence that the  
accuracy for this first step is quite reasonable. In a couple of BFGS  
steps you usually obtain a structure that converges (more or less)  
smoothly within 100 scf steps.


Let me add one last word on the convergence of wavefunctions. As a  
rule of thumb you can consider 1.0d-7 a good threshold for energy and  
eigenvalues (scf calculations), 1.0d-8 for forces (relax), 1.0d-9 for  
stress (vc-relax). However, remember that conv_thr is a *starting  
value* that is decreased down to two orders of magnitude below the  
provided value when BFGS is approaching convergence. This avoids that  
too tight criteria create problems such as yours in the very first  
steps. For "normal" vc-relax calculations conv_thr=1.0d-7 is usually a  
sensible choice, but you can use up to 1.0d-5 and restart the  
calculation with a lower value after a few steps.

Best Wishes

Quoting Ibrahim Yahaya Muhammad <ibrahimyahayamuhammad at gmail.com>:

> Dear Giuseppe,
> Thanks a lot for answering. I include all the correction and I successfully
> solved for bilayer MoS2 but I'm still having problem with 'vc-relax'
> calculation for 7 layers MoS2. Attached is the input and output.
> Best Regards,
> Ibrahim Yahaya Muhammad

Via Salaria Km 29,300 - C.P. 10
I-00015 - Monterotondo Scalo (RM)
Mob (*preferred*) +39 373 7305625
Tel + 39 06 90672342 - Fax +39 06 90672316
E-mail: <giuseppe.mattioli at ism.cnr.it>

More information about the users mailing list