[Pw_forum] convergence of pw.x

Paolo Giannozzi giannozz at nest.sns.it
Sat Dec 24 11:13:47 CET 2005


On Friday 23 December 2005 17:30, Konzern wrote:

> In the scf calculations, I set the parameter of conv_thr as 1.0d-12

this is very small. Typical values are 1.0d-6 to 1.0d-8. This number is
an estimate for the self-consistency error on the energy. Note that
the threshold for the phonon calculation is a different quantity (average
error on the potential or something like this)

> while the nscf just could not converge, there was always some
> " xx eigenvalues not converged."

as long as xx is small, it shouldn't be such a big deal, especially if you
have empty states (these are the state that tend to converge more
slowly)

> The threshhold I used in ph.x is 1.0d-14; I tried to use conv_thr=1.0d-10 in
> both scf and nscf while tr2_ph=1.0d-12 in ph.x, the resultant frequencies
> were "wrong" ( too many imaginary ones).

if you change all convergence parameters at the same time, it is
impossible to understand which are relevant and which are not.

P.

-- 
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



More information about the Pw_forum mailing list