Thanks, Eric, for the insight. That core correction can be vicious. Turning up the density cutoff does help. There is still oscillation in the total energy at the 100 micro-Ry (meV) level for density cutoff 16 times the wave function cutoff, but I guess one shouldn't expect too much stability at that level.
https://wiki.alcf.anl.gov/wparker/images/6/68/ZnO_ecut_convergence_rho4_10_16_lda_ncpp.png --William On Oct 19, 2011, at 9:13 PM, Eric J. Walter wrote: > > Hi, > >> Just to double-check. Is charge density cutoff 4 times that planewave cutoff? > > I think that this hits on the issue... The Zn.pz-nc.UPF has a sharp > core correction. The FFT/Density grid > needs to be increased significantly, here is a plot of from 4x to 20x > ecutwfc for the case with and without > a core correction in Zn. I think this is the cause of the odd *ecutwfc* > shown by William Parker. > > http://dl.dropbox.com/u/8094696/Znplot.pdf > > So, in this case, no, the density cutoff needs to be higher: > > > Eric > > > > > On 10/19/2011 08:47 PM, Nichols A. Romero wrote: >> The behavior here is what I might expect for a real-space DFT code, >> but V_NL is evaluated in reciprocal space right (we don't have >> real-space projectors)? >> >> Just to double-check. Is charge density cutoff 4 times that planewave cutoff? >> >> >> On Wed, Oct 19, 2011 at 2:51 PM, William Parker<wparker at anl.gov> wrote: >>> Dear Yun Song and interested parties, >>> >>> The possibility of a phonon test case in ZnO piqued my curiosity so I >>> decided to have a go at it. I find a very odd convergence pattern with >>> these two pseudopotentials in this system--the total energy has a damped >>> sinusoidal decay with increasing cutoff: >>> (selecting the maxima and minima at 10 Ry sampling) >>> >>> E_cutoff (Ry) E_total (Ry) >>> 100 -338.741 >>> 160 -338.660 >>> 210 -338.675 >>> 290 -338.665 >>> 370 -338.668 >>> 400 -338.667 >>> >>> Figure here: >>> https://wiki.alcf.anl.gov/wparker/images/b/b5/ZnO_ecut_convergence_lda_ncpp.png >>> Details here: >>> https://wiki.alcf.anl.gov/wparker/index.php/Zinc_Oxide_Parameter_Convergence >>> >>> I have never encountered this kind of behavior with norm-conserving >>> pseudopotentials before. At worst, I've seen little steps in an otherwise >>> smooth convergence to a value. The eigenvalues at the various k-points >>> don't appear to have any ghost states that come and go with cutoff. Has >>> anyone observed this kind of oscillation in the total energy with >>> increasing plane-wave cutoff before? What gives rise to it? >>> >>> --William >>> >>> ********************************************************* >>> William D. Parker >>> Computational Postdoctoral Fellow fax: (630) 252-4798 >>> MSD-212, Rm. C-215 >>> Argonne National Laboratory >>> 9700 S. Cass Ave. >>> Argonne, IL 60439 >>> ********************************************************* >>> >>> >>> >>> >>> _______________________________________________ >>> Pw_forum mailing list >>> Pw_forum at pwscf.org >>> http://www.democritos.it/mailman/listinfo/pw_forum >>> >> >> > > _______________________________________________ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum ********************************************************* William D. Parker Computational Postdoctoral Fellow fax: (630) 252-4798 MSD-212, Rm. C-215 Argonne National Laboratory 9700 S. Cass Ave. Argonne, IL 60439 *********************************************************
