Hi Abraham, I have run into these convergence issues with the OrderN approach in the past as well for force constant calculations. I suggest switching to Diagon for your force constant calculations and keeping a decent k-point grid in your kz direction. This should help the calculations converge so you can get meaningful force constants.
Best regards, Derek ######################### Derek Stewart, Ph. D. http://www.people.cornell.edu/pages/das248/ Scientific Computation Research Associate 250 Duffield Hall Cornell Nanoscale Facility Ithaca, NY 14853 > Hello all, > > I'm trying to use SIESTA to calculate the phonon spectrum of two different > 1D chains using the method perscribed in the vibra routine. I cannot, however, get past the calculation of the force matrix. For each system that > I study, I use the relaxed coordinates (per each specific > exchange-correlation functional) as input, create the FC.fdf with fcbuild > with supercell_3=1 (1 above, 1 below in the z-direction, have also tried supercell_3=2 with the same results) and then run this file in SIESTA like > the tutorials say to do: > > SystemLabel LDA_111_68_vibra_PD > NumberOfSpecies 2 > %block ChemicalSpeciesLabel > 1 14 Si > 2 1 H > %endblock ChemicalSpeciesLabel > PAO.BasisSize DZP > MeshCutoff 200.0 Ry > %include FC.fdf > > I would use this file for the relaxed geometry corresponding to the stationary point found with an LDA xc-functional + PP and the > PAO.BasisSize > DZP in the original calculation. For other systems like a BLYP > xc-functional > and user-defined basis I would use those parameters in the input file (but > with roughly the same results, which follow). Defaults are used for everything else, including the OrderN method since the full unit cell for > the FC calculation is about 200 atoms. > > The FC calculation seems to run smoothly until I get something like this in > the standard output: > > > ********************************************************************************** > cgwf: iter = 984 grad = -0.000241 Eb(Ry) = > -203.356707 > cgwf: iter = 985 grad = -0.000241 Eb(Ry) = > -203.356720 > cgwf: iter = 986 grad = -0.000241 Eb(Ry) = > -203.356734 > cgwf: iter = 987 grad = -0.000243 Eb(Ry) = > -203.356748 > cgwf: iter = 988 grad = -0.000248 Eb(Ry) = > -203.356762 > cgwf: iter = 989 grad = -0.000257 Eb(Ry) = > -203.356775 > cgwf: iter = 990 grad = -0.000260 Eb(Ry) = > -203.356788 > cgwf: iter = 991 grad = -0.000263 Eb(Ry) = > -203.356801 > cgwf: iter = 992 grad = -0.000257 Eb(Ry) = > -203.356814 > cgwf: iter = 993 grad = -0.000256 Eb(Ry) = > -203.356827 > cgwf: iter = 994 grad = -0.000250 Eb(Ry) = > -203.356840 > cgwf: iter = 995 grad = -0.000253 Eb(Ry) = > -203.356853 > cgwf: iter = 996 grad = -0.000252 Eb(Ry) = > -203.356866 > cgwf: iter = 997 grad = -0.000252 Eb(Ry) = > -203.356880 > cgwf: iter = 998 grad = -0.000251 Eb(Ry) = > -203.356893 > cgwf: iter = 999 grad = -0.000244 Eb(Ry) = > -203.356906 > cgwf: iter = 1000 grad = -0.000238 Eb(Ry) = > -203.356920 > > cgwf: Maximum number of CG iterations reached > > denmat: qtot (before DM normalization) = 765.6153 > ordern: qtot (after DM normalization) = 546.0000 > > siesta: iscf = 2 > Eharris(eV) = -12344.2602 E_KS(eV) = -12020.4605 dDmax = 0.2649 > > > ordern: enum = 546.0000 > cgwf: iter = 1 grad = -67.320865 Eb(Ry) = > -216.535666 > cgwf: iter = 2 grad = -149.630916 Eb(Ry) = > -224.149328 > cgwf: iter = 3 grad = -165.080016 Eb(Ry) = > -229.107766 > cgwf: iter = 4 grad = -122.780882 Eb(Ry) = > -235.802701 > cgwf: iter = 5 grad = -128.457462 Eb(Ry) = > -240.351352 > cgwf: iter = 6 grad = -107.815766 Eb(Ry) = > -244.409131 > cgwf: iter = 7 grad = -105.320678 Eb(Ry) = > -248.300033 > cgwf: iter = 8 grad = -92.022410 Eb(Ry) = > -251.093176 > cgwf: iter = 9 grad = -88.452386 Eb(Ry) = > -254.170805 > cgwf: iter = 10 grad = -202.420032 Eb(Ry) = > -257.806315 > cgwf: iter = 11 grad = -2466.490254 Eb(Ry) = > -222.354649 > cgwf: iter = 12 grad = -163.364902 Eb(Ry) = > -255.045915 > cgwf: iter = 13 grad = -1168.957839 Eb(Ry) = > -232.144564 > cgwf: iter = 14 grad = -161.959255 Eb(Ry) = > -254.849595 > cgwf: iter = 15 grad = -1102.828538 Eb(Ry) = > -232.803424 > cgwf: iter = 16 grad = -157.425714 Eb(Ry) = > -254.979384 > cgwf: iter = 17 grad = -194.373222 Eb(Ry) = > -255.299976 > cgwf: iter = 18 grad = -182605.315981 Eb(Ry) = > -100.555913 > cgwf: iter = 19 grad = -217036.659307 Eb(Ry) = > -99.721836 > cgwf: iter = 20 grad = -249.395730 Eb(Ry) = > -249.824155 > cgwf: iter = 21 grad = -244.560067 Eb(Ry) = > -253.941014 > cgwf: iter = 22 grad = -152.248315 Eb(Ry) = > -257.760953 > cgwf: iter = 23 grad = -232.419799 Eb(Ry) = > -261.955244 > cgwf: iter = 24 grad = -96401477.679874 Eb(Ry) = > -911.330200 > cgwf: iter = 25 grad = -107154809.385566 Eb(Ry) = > -912.719590 > cgwf: iter = 26 grad = -1438.764936 Eb(Ry) = > -262.289834 > cgwf: iter = 27 grad = -1408.809784 Eb(Ry) = > -262.325692 > cgwf: iter = 28 grad = -186592.532696 Eb(Ry) = > -109.282399 > cgwf: iter = 29 grad = -217965.896792 Eb(Ry) = > -108.576682 > cgwf: iter = 30 grad = -302.647048 Eb(Ry) = > -254.260707 > cgwf: iter = 31 grad = -194.085227 Eb(Ry) = > -259.234987 > cgwf: iter = 32 grad = -148740.557077 Eb(Ry) = > -279.409429 > cgwf: iter = 33 grad = -2536890.389917 Eb(Ry) = > -319.726178 > cgwf: iter = 34 grad = -5780230.992347 Eb(Ry) = > -331.297794 > cgwf: iter = 35 grad = -132426997.296064 Eb(Ry) = > -642.016407 > cgwf: iter = 36 grad = -354656197.135856 Eb(Ry) = > -740.210175 > cgwf: iter = 37 grad = ****************** Eb(Ry) = > -3046.280084 > cgwf: iter = 38 grad = ****************** Eb(Ry) = > -3113.114594 > cgwf: iter = 39 grad = -2197752344.447697 Eb(Ry) = > 19491.918692 > cgwf: iter = 40 grad = -402.146008 Eb(Ry) = > -261.851463 > cgwf: iter = 41 grad = -430786270.291137 Eb(Ry) = > 4171.581613 > cgwf: iter = 42 grad = -430958482.835069 Eb(Ry) = > 4171.584019 > cgwf: iter = 43 grad = -82331696.109239 Eb(Ry) = > -24934.797112 > cgwf: iter = 44 grad = -2565001.988477 Eb(Ry) = > 2255.634962 > cgwf: iter = 45 grad = -80853.063402 Eb(Ry) = > -174.667315 > cgwf: iter = 46 grad = ****************** Eb(Ry) = > 1621007.748789 > > cgwf: CG tolerance reached > > denmat: qtot (before DM normalization) = ************ > ordern: qtot (after DM normalization) = 546.0000 > > siesta: iscf = 3 > Eharris(eV) = -7881.8204 E_KS(eV) = -12044.3075 dDmax = 21.6098 > > > ordern: enum = 546.0000 > cgwf: iter = 1 grad = ****************** Eb(Ry) = > ************** > cgwf: iter = 2 grad = ****************** Eb(Ry) = > ************** > cgwf: iter = 3 grad = ****************** Eb(Ry) = > ************** > cgwf: iter = 4 grad = ****************** Eb(Ry) = > ************** > cgwf: iter = 5 grad = ****************** Eb(Ry) = > ************** > cgwf: iter = 6 grad = ****************** Eb(Ry) = > ************** > cgwf: iter = 7 grad = ****************** Eb(Ry) = > ************** > cgwf: iter = 8 grad = NaN Eb(Ry) = > ************** > cgwf: iter = 9 grad = NaN Eb(Ry) = > NaN > cgwf: iter = 10 grad = NaN Eb(Ry) = > NaN > etc. > ****************************************************************************** > > > The end result is that I get a force constants matrix full of NaN's, no matter which 1-D chain I run. I'm not sure what I'm doing wrong... has anybody run into this problem before? Could you suggest additional simulation parameters that help to converge this type of calculation? > > > Thank you, wise SIESTA gurus. > > Abraham Hmiel > Research Assistant > College of Nanoscale Science and Engineering at SUNY Albany >
