Hello Francesco, Apart from the observations made by Salvador, it is well known that the LDA/GGA approximations do not give good values for band gaps and lattice constant for various semiconductor systems (they work fine with others).
In case you check what Salvador pointed out, and the "bad" results continue, I suggested you to use other approximations like LDA+U or LDA-1/2. Regards, Camps On Fri, Oct 19, 2012 at 9:57 AM, Salvador Barraza-Lopez <[email protected]>wrote: > Hi Francesco, a couple of quick observations: > Your meshcutoff is quite low. > > You are working with an insulator, so the number of k-points does not > need to be excessively large, but are you sure 4x4x4 is sufficient for your > purposes? > > Your tolerance is too small, 10-5 may be enough. > > You also mention your a0 is not the equilibrium one but the expt one, yet > should I assume you are comparing to a planewave code to get your gap? if > not, to what you are contrasting? You could examine if the gap is smaller > as you wish when you get the lattice constant in equilibrium. > > Having said the above, the procedure you are following may be risky... I > would start with a comprehensive study of convergency of meshcutoff, your > pao.basis parameter, your a0, your k-points. Only when all of those are > converged I would go for an estimation of the band gap. The idea is to get > everything converged and then essentially trust your final result (unless > the pseudos are off, but that would be the last thing to check on your > list). > > Hopefully the suggestions are helpful. > > Best regards, > -Salvador > > ------------------------------ > *From:* [email protected] [[email protected]] on behalf of > Francesco Tassone [[email protected]] > *Sent:* Friday, October 19, 2012 7:07 AM > *To:* [email protected] > *Subject:* [SIESTA-L] zincblend CdSe bands > > Hello, I' m trying to calculate ZB CdSe bands using TM pseudos downloaded > from the > SIESTA web. I' m doing the calculation at the exp lattice constant of > 6.05 A. > > I get > 1. a large gap around 1.2eV (using standard CA LDA should be 0.7eV), > reducing only a bit by enlarging pao basis (e.g. energyshift to 10meV) , > so it's not a problem of basis > 2. 5 'fake' bands around -8 eV below valence band top at Gamma. > (should be deeper by about 4 eV, guess from Cd d orbitals, and quite > flatter) > > > LatticeConstant 6.05 Ang > %block LatticeVectors > 0.500 0.500 0.000 > 0.000 0.500 0.500 > 0.500 0.000 0.500 > %endblock LatticeVectors > > MeshCutoff 100.0 Ry > > MaxSCFIterations 50 > DM.MixingWeight 0.1 > DM.NumberPulay 3 > DM.Tolerance 1.d-8 > > %block kgrid_Monkhorst_Pack > 4 0 0 0. > 0 4 0 0. > 0 0 4 0. > %endblock kgrid_Monkhorst_Pack > > SolutionMethod diagon > ElectronicTemperature 25 meV > > HELP NEEDED > > what am I getting wrong here, TM pseudos should be used with CA-LDA which > is default, correct ?? > > PS: everything ok with Si, so I guess it' s not a compilation problem. > >
