Maybe the nonlinear core correction is slowly converging, especially in Sodium which has s and p semicore electrons and a very localized (pseudo-)core charge. Just guessing.
P. On Fri, 2014-09-19 at 14:20 -0400, Joshua Davis wrote: > I was performing a convergence test for the energy cut-off for the > density for a set of PAW potentials just to make sure, but I have some > results that confuse me a bit. > > > The compound is a NaB5C (space group: Pm-3m) so the potentials I used > were: > > B.pbe-n-kjpaw_psl.0.1.UPF > C.pbe-n-kjpaw_psl.0.1.UPF > Na.pbe-spn-kjpaw_psl.0.2.UPF > > > The cell was also completely relaxed each time with cell and ion > dynamic being "bfgs" algorithm > > > > Already did a ecutwfc and kpoint convergence test for each and I found > and ecutwfc of 35 Ry and a kpoint mesh of 4x8x8 to be sufficient (I am > calculating a cell doubled in one direction). > > > When I did the ecutwfc convergence test I went from 10 to 50 Ry in > increments of 5 Ry, and I kept the default ecutrho=4*ecutwfc for each > run. > > > so when I did the ecutrho convergence kept ecutwfc at 35 Ry and the > kpoints at 4x8x8 and started at 140 and went to 400 Ry. > > > The graph of my convergence is shown below: > Inline image 1 > > > and if the fourm scrubs the image here is the raw data: > ecutrho(Ry),TotalEnergy(ry) > 140,-370.0979858 > 150,-370.0980023 > 160,-370.0995505 > 170,-370.1019804 > 180,-370.1047998 > 190,-370.1072663 > 200,-370.1086284 > 250,-370.1048218 > 300,-370.1049835 > 400,-370.1060986 > > > > > The convergence shows a density cut-off of 140 Ry and 150 Ry to be > pretty similar in calculated energy, but after that the calculated > energy quickly drops until a cut-off of 200 Ry. After that the > calculated energy shoots up at a cut-off of 250 Ry by 0.004 then > tappers down by 0.002 Ry as it gets to a cut-off of 400 Ry. > > > I thought the energy cut-off for density would be the easiest after > ecutwfc and kpoint, but I want to make sure my results would be fine > based on what I read for he input documentation for ecutrho. It says > ecutrho=4*ecutwfc is usually good (which is pretty standard for the > PAW method), but I wanted to make sure. I thought the calculated > energy would not have changed or the calculated energy would have > converged at a reasonable value. > > > Am I doing something wrong? Is the ratio of ecutwfc and ecutrho too > off at the higher energy ecutrho cut-off? Other posts suggest > lowering ecutwfc and keeping ecutrho cut-off constant for proper > convergence, should I have done this? I figured sneaking up ecutrho > would be okay. Am I dumb and this is just an artifact of the PAW > method at a large ratio of ecutwfc to ecutrho? Is this an artifact of > how the FFT mesh is implemented in QE 5.1? Should I have not gone > that high in my density cut-off? > > > Any help would be much appreciated. > > > Other considerations: > I used QE 5.1 compiled with ifort 13 and mkl 11. > > > &CONTROL > calculation = 'vc-relax', > etot_conv_thr = 1.0D-5, > forc_conv_thr = 1.0D-4 > / > > > &SYSTEM > ibrav = 0, > nat = 14, > ntyp = 3, > nosym = .TRUE., > nspin = 1, > occupations = 'fixed', > > ---------------------------------------------------------------------------------------------------------------- > Joshua D. Davis > > Graduate Assistant > Department of Chemistry > Michigan State University > 578 S. Shaw Lane, room 432 > > > East Lansing, MI 48824 > > ----------------------------------------------------------------------------------------------------------------- > > _______________________________________________ > Pw_forum mailing list > Pw_forum at pwscf.org > http://pwscf.org/mailman/listinfo/pw_forum
