[Pw_forum] Geometry optimization does not converge

Lorenzo Paulatto paulatto at sissa.it
Tue Sep 22 10:33:16 CEST 2009


In data 21 settembre 2009 alle ore 14:56:48, Gunther Jegert  
<guntherjegert at mytum.de> ha scritto:
> I varied the energy cutoff between 40 – 60 Ry and kept the
> density ten times higher. I added some additional unoccupied states and a
> smearing. I also increased the conv_thr for the electrons up to 1D-11. I
> even tried different starting geometries. However, I always end up with a
> too high force in the order of 1D-2 (forc_conv_thr = default).


Dear Gunther,
your input looks ok to me, there are only a couple of point you could  
improve:
1. the values you've choosen for nbnd is lower than the default one, it  
may be too low
2. I'm not sure pot_extrapolation is supposed to work with bfgs
3. wfc_extrapolation is not implemented with ultrasoft pseudopotentials,  
it should be harmless though
4. you could increase upscale from 10 to 1.d+4 or even more, this would  
allow the code to tighten the convergence threshold if necessary
5. you can always try the damp algorithm instead of bfgs...

I hope it helps,
best regards



-- 
Lorenzo Paulatto
SISSA  &  DEMOCRITOS (Trieste)
phone: +39 040 3787 511
skype: paulatz
www:   http://people.sissa.it/~paulatto/

     *** save italian brains ***
  http://saveitalianbrains.wordpress.com/



More information about the users mailing list