The value of nbnd is particularly important for magnetic system since in priniciple one should have a larger nbnd for majority bands than for minority bands. We found that a safer value is nelec/2*1.5 In some cases it can really lead to convergence problem, for slabs or clusters.. Cyrille
________________________________________ De : [email protected] [[email protected]] de la part de Paolo Giannozzi [[email protected]] Envoyé : jeudi 6 novembre 2014 18:47 À : PWSCF Forum Objet : Re: [Pw_forum] Non Convergence On Thu, 2014-11-06 at 17:12 +0000, BARRETEAU Cyrille wrote: > I would also advise to increase nbnd since for magnetic system > the default value is often too small. I usually take nbnd=nelec/2*1.5 > instead of the default value of nelec/2*1.2 > > Therefore you should take at least nbnd=80 instead of 60. the original input had a larger nbnd. I reduced it just to speed up the calculation. I agree that it is on the lower bounds of what one can safely use with metals. Paolo > Cyrille > -------------------------------------------------------------------------- > Cyrille Barreteau > CEA Saclay, IRAMIS, SPEC Bat. 771 > 91191 Gif sur Yvette Cedex, FRANCE > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > DTU Nanotech > Ørsteds Plads, building 345E > DK-2800 Kgs. Lyngby, DENMARK > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > +33 1 69 08 29 51 / +33 6 47 53 66 52 (mobile) (Fr) > +45 45 25 63 12/ +45 28 72 55 18 (mobile) (Dk) > email: [email protected] / [email protected] > Web: http://iramis.cea.fr/Pisp/cyrille.barreteau/ > > ------------------------------------------------------------------------------- > > ________________________________________ > De : [email protected] [[email protected]] de la part de > Paolo Giannozzi [[email protected]] > Envoyé : jeudi 6 novembre 2014 17:45 > À : PWSCF Forum > Objet : Re: [Pw_forum] Non Convergence > > Metallic slabs are sometimes difficult to converge. > I tried with mixing_mod='local-TF' (works better for slabs), > a smaller mixing_beta (0.1), a larger degauss (0.05) and a > denser k-point grid (8 8 1) that should yield smoother results, > and it converges, at least for the reduced cutoff (30/300) > I tried. > > Paolo > > > Try > On Wed, 2014-11-05 at 15:41 +0000, Elliot Menkah wrote: > > Hello Everyone, > > I'm running a geometry optimization calculation(relax) of an 11-layer > > Nickel slab and I used the deafult nstep (nstep=50) and electron_maxstep > > (100). > > > > My system did not converge with the parameters and configurations set as > > it run through 100 iterations. > > > > Could there be something wrong with the configuration or would I have to > > increase the electron_maxstep value?. > > > > Please find attached both input and out files in that regard. > > > > Thank you > > > > > > Kind Regards, > > > > Elliot > > > > > > > > > > _______________________________________________ > > Pw_forum mailing list > > [email protected] > > http://pwscf.org/mailman/listinfo/pw_forum > > -- > Paolo Giannozzi, Dept. Chemistry&Physics&Environment, > Univ. Udine, via delle Scienze 208, 33100 Udine, Italy > Phone +39-0432-558216, fax +39-0432-558222 > > _______________________________________________ > Pw_forum mailing list > [email protected] > http://pwscf.org/mailman/listinfo/pw_forum -- Paolo Giannozzi, Dept. Chemistry&Physics&Environment, Univ. Udine, via delle Scienze 208, 33100 Udine, Italy Phone +39-0432-558216, fax +39-0432-558222 _______________________________________________ Pw_forum mailing list [email protected] http://pwscf.org/mailman/listinfo/pw_forum _______________________________________________ Pw_forum mailing list [email protected] http://pwscf.org/mailman/listinfo/pw_forum
