Hi Marcos,I redid the calculation by increasing the mesh cutoff to 250Ry.It seems that the resulsts are improved.Attached please find the output file.Do you have any other good suggestions? Thank you very much.
Mu
Quoting Marcos Verissimo Alves <[EMAIL PROTECTED]>:
Looking at the output, I would initially increase the mesh cutoff to at least 250 Ry (N has a rather hard pseudo) and see if anything improves. MarcosHi Marcos, Attached please find the input file for generation Ga and N PP, the PP files for Ga and N, the input file for seista and the outfile file(out). For wurtize GaN, three parameters should be optimized(a,c,u),so the fractional coordinate can't be fixed. Really thank you very much. Mu Quoting Marcos Verissimo Alves <[EMAIL PROTECTED]>:Hi Mu, I don't really get what you mean by single-point calculation - I'll assume it is just geometry relaxation. You should do both, one as a check of the other, in my opinion. In principle, if your pseudos are ok and your parameters (namely k-point sampling, mesh cutoff, basis set size and electronic temperature) are ok, then I guess there should not be too much difference between the results of an E vs a curve and a variable cell calculation. Would you mind posting the pseudopotentials and input so we can look at it? Siesta has no way of maintaining symmetry. I have seen a few suggestions on how to do this on the list, like entering fractional coordinates and fixing them, but I don't know if this would work or give sensible results at all - mainly because I haven't tried. Cheers, MarcosHi Marcos, Thank you very much. I have generated a N pseudo using LDA since I want to use this method in the following calculations. I checked it for NO molecule.The optimized bond length is in excellent agreement with experiments. I want to optimize the lattice parameter for wurtzite GaN. Should I do single point calculation or define MD.TypeOfRun cg and MD.Variable-Cell T?If I choose the latter, the shape of the cell is changed after optimization and the symmetry is completely broken.Is there any way to keep symmetry during calculation?The difference between calculations and experiments is large, about 10% error for a and c. It seems that the compiler is ok.I will try to include the 3d states of Ga. Once again, thank you very much. Mu Quoting Marcos Verissimo Alves <[EMAIL PROTECTED]>:Hi Mu, There is a good N pseudo on the siesta website, but it's gga - however uses partial core, so it might be a good idea to include partial core also for N. There is one thing that is always worth reminding - check what compiler you are using for compiling atom. I remember last year (maybe) someone wrote to the list complaining that her pseudo did not work, giving ghost states which did not exist, since the pseudo had already been used in previsous work, and it was the compiler's fault. When she changed the compiler (or compiling options, I don't remember exactly now) her pseudo worked fine. Search the list archives for that issue. If the compilation is not really a problem, then it might be good to include the 4d states for Ga. I agree with Alexander that you might see a big difference when you include 4d states as semicore - this is my experience with Pb in PbO, it only really gave sensible results when I included the semicore states of Pb. However, I'm not sure I agree completely that DZ would definitely not be enough. DZ might not give you the best results ever, but would not be the worse ever, either. How much is the difference of lattice parameter from experimental value you obtain? Remember that ~2% would be ok for LDA. Hope this helps, Marcoshello, I want to generate the pseudopotential of Ga and N for GaN bulk. For Ga the input file is as follows: pe Ga, LDA, rcore=0.80 tm2 0.0 n=Ga c=car 0.0 0.0 0.0 0.0 0.0 0.0 6 2 4 0 2.00 0.00 # 4s2 4 1 1.00 0.00 # 4p1 2.10 2.50 3.00 0.00 0.00 0.80 For N the input file is as follows: pg Nitrogen tm2 0.0 N car 0.0 0.0 0.0 0.0 0.0 0.0 1 2 2 0 2.00 0.00 2 1 3.00 0.00 1.14 1.14 1.14 0.00 0.00 0.00 The reference electronic configuration and s/p/d cutoff radius (except the parameter for core correction)are cited from PRB,70,235209, where the lattice constant of GaN agrees well with experiments.In their work it has been tested that the 3d of Ga can be left in the core. But however I adjust the r(pc)for core correction, the optimized lattice constant differs greatly from experiment.(DZ basis set is employed) Is anyone who have good idea on how to generate good pseudopotential for GaN bulk?Thank you very much. Mu-- Dr. Marcos Verissimo Alves Post-Doctoral Fellow Condensed Matter and Statistical Physics Sector International Centre for Theoretical Physics Trieste, Italy -------- I have become so addicted to vi that I try to exit OpenOffice by typing :wq!-- Dr. Marcos Verissimo Alves Post-Doctoral Fellow Condensed Matter and Statistical Physics Sector International Centre for Theoretical Physics Trieste, Italy -------- I have become so addicted to vi that I try to exit OpenOffice by typing :wq!-- Dr. Marcos Verissimo Alves Post-Doctoral Fellow Condensed Matter and Statistical Physics Sector International Centre for Theoretical Physics Trieste, Italy -------- I have become so addicted to vi that I try to exit OpenOffice by typing :wq!
out.fdf
Description: application/vnd.fdf
