[Pw_forum] question on electron_dynamics and orthogonalization

Paolo Giannozzi giannozz at nest.sns.it
Thu Mar 30 11:55:38 CEST 2006


On Wednesday 29 March 2006 22:26, Nichols A. Romero wrote:

> I am struggling with trying to get PWSCF and the CP code to agree. 
> [...] I'm attaching a model problem I've been playing with for the
> last two days.

thank you for providing simple, useful examples.

* as mentioned by Nicola M., the agreement between energies and
  forces in PW and CP in your test is very good, once you take into
  account the different units

* there is a difference in the definition of V(0) that affects, as a rigid
  shift, the eigenvalues: see attached notes. I don't know which one
  is "better" (i.e. converges faster to the limit of infinite cell size).
  I like the PW definition better: the CP one is in my opinion arbitrary.

* The convergence of  eigenvalues with cutoff may be quite slow,
   expecially for higher-energy states, and different in CP and PW
  (because of the "small boxes" in CP, as mentioned by Silviu)

* two rather exotic bugs in stress calculation in PW:
  - gamma point + l=0 nonlocality only + parallelization
  - gamma point + LSDA
  have been fixed only very recently (8 and 16 march)

* there is a known problem with stress calculation in CP, related to
  l>0 nonlocality. The error is rather small but not negligible. Sooner
  or later we'll find it. This is the bug Nicola M. was referring to.

* Stresses calculated in CP with the 'cg' option in your example are
  completely bogus. Most likely, stress calculation is not implemented
  in the 'cg' case. This is a different bug from the one above

Paolo
-- 
Paolo Giannozzi             e-mail:  giannozz at nest.sns.it
Scuola Normale Superiore    Phone:   +39/050-509876, Fax:-563513 
Piazza dei Cavalieri 7      I-56126 Pisa, Italy
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ewald.tex
Type: text/x-tex
Size: 15178 bytes
Desc: not available
Url : /pipermail/attachments/20060330/c1890b30/attachment.tex 


More information about the Pw_forum mailing list