Thank you all for your suggestions, you've given me a lot to work on! I'll let you know about the results.
Cheers, Sergi 2015-02-03 15:00 GMT+01:00 xiaochuan Ge <[email protected]>: > There is another easier solution but I have never tested. The default > startingwfc is "atomic + random", instead one could try to use "random". > > =================== > Dr. Xiaochuan Ge (Giovanni) > Center for Functional Nanomaterials > Brookhaven national laboratory > =================== > > On 3 February 2015 at 08:49, Pu ZHANG <[email protected]> wrote: > >> Dear all, >> >> I'm just curious about Xiaochuan's solution. It would be Interesting and >> useful to be able to specify the initial configuration in such a flexible >> way. As I understand from experience in other codes, it's trivial to set up >> the occupation for the whole system, however rather than for separate >> parts. >> >> Is there a way to make this setting? >> 2015 年 2 月 3 日 下午 9:38于 "xiaochuan Ge" <[email protected]>写道： >> >> Dear Sergi Vela, >>> >>> I have encountered similar problems before. To solve your trouble I >>> think the main question is: what is the converged electronic configuration >>> you are expecting? >>> >>> I notice that your system is made by a gold slab plus a C2S3N3 molecule. >>> The latter being isolated contains odd number of electrons. I believe if >>> you run the ground-state calculation for only the C2S3N3 molecule without >>> special care you will end up with similar problems with convergence. Using >>> smearing does not solve your problem is probably because the band gap of >>> C2S3N3 is too large for the smearing parameter you used to make a >>> difference. >>> >>> The trick is, if somehow C2S3N3 can take one electron from the slab, >>> then an even electron configuration can be achieved for the molecule, and >>> most probably you will achieve the convergence. However, the initial >>> electron configuration is generated based on the overlap of atomic >>> orbitals. Since the molecule is pretty far away from the metal slab, a >>> large barrier may hinder the electron from metal to transfer to the >>> molecule. Therefore during the SCF iteration the system can not find its >>> energy minimum. >>> >>> I have some suggestions that you may want to try. First, you may want to >>> put the molecule closer to the slab, so the barrier between the initial >>> electronic configuration and the reasonable final configuration is low >>> enough for SCF to overcome. Second, you could try to set the initial >>> configuration by hand, say let the system start with ( C2S3N3 - ) + ( Slab >>> + ), I believe this is possible in QE. Third, maybe you can add a little >>> magnetic filed to breakdown the degeneracy of spin up and down. >>> >>> Hope this can solve your problem! >>> >>> =================== >>> Dr. Xiaochuan Ge (Giovanni) >>> Center for Functional Nanomaterials >>> Brookhaven national laboratory >>> =================== >>> >>> On 3 February 2015 at 08:02, Sergi Vela <[email protected]> wrote: >>> >>>> Dear Guido and Giuseppe, >>>> >>>> Thank you both for your replies. >>>> >>>> The structure should be ok, it represents an organic molecule between 2 >>>> gold electrodes. You should get something like the image I attach (do >>>> you?). It is a rather standard scheme, as you may also see in: >>>> >>>> http://pubs.acs.org/doi/abs/10.1021/ja2090096 >>>> >>>> although my structure is simplified for testing. >>>> >>>> I don't know whether ibrav=0 is strictly necessary but, anyway, it >>>> should not affect the outcome of the calculation, right? >>>> >>>> I have tested other k-point meshes without improvement. Anyway, I'll >>>> try denser meshes together with Giuseppe's suggestions and let you know the >>>> outcome. >>>> >>>> Cheers, >>>> Sergi >>>> >>>> 2015-02-03 13:56 GMT+01:00 Giuseppe Mattioli < >>>> [email protected]>: >>>> >>>>> >>>>> Dear Sergi >>>>> It seems that you bounce between more than one electronic >>>>> configuration at every scf step. This is far from uncommon in the case of >>>>> molecular open >>>>> shell systems. :-( >>>>> I can add one or two things to Guido's good advice. >>>>> >>>>> > I've tried several tricks to try to improve electronic structure's >>>>> > convergence. Changing the 'mixing_beta', the 'type of mixing', >>>>> removing the >>>>> > smearing, >>>>> >>>>> Use always the more robust gaussian smearing (at least 0.01 Ry, but >>>>> also more if you struggle for convergence) for geometry optimizations. If >>>>> it does >>>>> not converge, then m-p and m-v will also fail 99.9% . You can use >>>>> other schemes after you tamed the system. Reduce also the mixing_beta down >>>>> to 0.05 >>>>> or 0.02 (with davidson). >>>>> >>>>> Maybe your starting configuration is horribly unlucky. Try (carefully) >>>>> to use something like >>>>> >>>>> conv_thr=1.0d-5~1.0d-6 >>>>> electron_maxstep=50~100 >>>>> scf_must_converge=.false. >>>>> >>>>> and see if your convergence problem improves after few bfgs steps >>>>> without showing unphysical behavior. >>>>> >>>>> If your SOMO is higher than the Ag Fermi energy, then there is >>>>> probably no way to have an integer value for the magnetization of the >>>>> system because of >>>>> the molecule-to-surface charge transfer. You may try not to bias the >>>>> tot_magnetization variable. >>>>> >>>>> Finally, ecutwfc=25.0 seems a bit too low. It could be a source of >>>>> instability. I would use something around 35/280 Ry for ecutwfc and >>>>> ecutrho. >>>>> >>>>> HTH >>>>> Giuseppe >>>>> >>>>> >>>>> On Tuesday, February 03, 2015 10:57:23 AM Sergi Vela wrote: >>>>> > Hi all, >>>>> > >>>>> > I'm having a lot of troubles in modeling the adsorption of molecules >>>>> on >>>>> > surfaces. I've quite a lot of experience with QE when working with >>>>> crystals >>>>> > made of open-shell molecules but I'm not familiar with the >>>>> description of >>>>> > metal surfaces. >>>>> > >>>>> > First of all, I'm using Qespresso v.5.1.1 >>>>> > >>>>> > I succeed in simulating 4 layers of a 111 Gold surface with Quantum >>>>> > Espresso (QE). However, using the same surface and unit cell, when I >>>>> > include a magnetic (open-shell) molecule on top of that surface, >>>>> even if >>>>> > its far away from it, I cannot reach convergence. I'm 99% sure that >>>>> the >>>>> > input is fine in terms of the unit cell definition (see input file >>>>> > attached) and the problem is just the convergence of the electronic >>>>> > structure. >>>>> > >>>>> > I believe the problem is the fact that the system is highly >>>>> inhomogeneous >>>>> > as it has a metal surface and an open-shell molecule. I take as an >>>>> example >>>>> > the smearing, since, although it is mandatory when modeling the metal >>>>> > alone, it seems to hinder the description of an open-shell molecule. >>>>> > Originally, I used methfessel-paxton smearing with a 'degauss' of >>>>> 0.2 eV. >>>>> > The convergence seems to be even worse as the value is increased, and >>>>> > changing the type of smearing doesn't help neither. >>>>> > >>>>> > I've tried several tricks to try to improve electronic structure's >>>>> > convergence. Changing the 'mixing_beta', the 'type of mixing', >>>>> removing the >>>>> > smearing, switching to 'cg' algorith, increasing number K points ... >>>>> all of >>>>> > them with identical output (take Summary.out as an example of the >>>>> typical >>>>> > evolution of the SCF energy, you'll see that is awful). >>>>> > >>>>> > Anyone has experience on modeling such king of systems? >>>>> > >>>>> > Any help is appreciated, thanks in advance >>>>> > >>>>> > Dr. Sergi Vela, >>>>> > University of Strasbourg, France. >>>>> >>>>> ******************************************************** >>>>> - Article premier - Les hommes naissent et demeurent >>>>> libres et ègaux en droits. Les distinctions sociales >>>>> ne peuvent être fondèes que sur l'utilitè commune >>>>> - Article 2 - Le but de toute association politique >>>>> est la conservation des droits naturels et >>>>> imprescriptibles de l'homme. Ces droits sont la libertè, >>>>> la propriètè, la sùretè et la rèsistance à l'oppression. >>>>> ******************************************************** >>>>> >>>>> Giuseppe Mattioli >>>>> CNR - ISTITUTO DI STRUTTURA DELLA MATERIA >>>>> v. Salaria Km 29,300 - C.P. 10 >>>>> I 00015 - Monterotondo Stazione (RM) >>>>> Tel + 39 06 90672836 - Fax +39 06 90672316 >>>>> E-mail: <[email protected]> >>>>> http://www.ism.cnr.it/english/staff/mattiolig >>>>> ResearcherID: F-6308-2012 >>>>> >>>>> _______________________________________________ >>>>> 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 >>>> >>> >>> >>> _______________________________________________ >>> 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 >> > > > _______________________________________________ > 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
