[Q-e-developers] Probable bug in phonons

stefano de gironcoli degironc at sissa.it
Sat Nov 29 12:36:26 CET 2014


Dear  Daniele, Giovanni,
     the phonon corresponding to commensurate q-points are conceptually 
equivalent to a Gamma calculation in a supercell.
     As such the equations involved have a well defined expression that 
one could connect to a "frozen phonon" calculation with physical meaning.
     Phonons for q-points that are not commensurate are strictly 
speacking impossible to describe on a supercell so they are "unphysical" 
and one relies on a continuity argument to believe that the same 
formulas away from commensurate points give a smooth dispersion. In the 
early days I tested this continuity for semiconductors and simple metals 
but things have evolved a lot since and i don't know how much this fact 
has been monitored.
     Somehow the calculations you are performing challenge this 
continuity argument
     I would try to isolate the problem by performing similar 
calculations with 1) a norm-conserving pseudopotential 2) with/without 
nonlinear core correction 3) an uspp that requires significantly less 
cutoff (like 30-180).

stefano

On 11/28/2014 02:33 PM, Giovanni Pizzi wrote:
> Dear QE developers,
>
> we have been working with 'simple' ferromagnetic bcc iron and were 
> trying to calculate phonon dispersion with ph.x (5.1 QE version) on 
> different q meshes at the theoretical electronic equilibrium volume 
> and 'well' converged input parameters. Playing with the mesh of q 
> points we have noticed the following:
>
> *Apparently, the frequencies calculated at q points that do not belong 
> to the k-mesh of the underlying scf calculation are either negative or 
> apparently 'random'.*
>
> On the other hand, q points belonging to the k-mesh display expected 
> reasonable values. This means that using a q-mesh fully included in 
> the k-mesh, you are able to get reasonable phonon dispersions 
> comparable with experiments and you do not notice the problem.
>
> Trying to rationalize a bit more the problem, we have noticed that 
> this behavior can be systematically reproduced with different svn 
> versions of QE, back at least to revision 8500 and probably also 
> before (we tested versions from 8500 to 11000 at steps of 500, and all 
> provided the same "wrong" results, see data below). The test can be 
> reproduced in reasonably short time with the input files reported at 
> the end of this email (with pslibrary pseudos).
>
> We also provide in attachment an example of this behavior for the well 
> converged calculations of Fe (input&outputs of pw and ph; 
> kmesh=24x24x24, and phonons on 6x6x6 and 10x10x10 meshes).
> The attached figures show an example of the problem when the q-points 
> are chosen along the (0,0,z) line, and a comparison between the 
> "correct" interpolation for the 6x6x6 q-mesh (note that all phonons 
> are positive) and the wrong interpolation for the 10x10x10 q-mesh.
>
> We wonder if someone else has ever experienced such a problem. We also 
> think that this might be considered a severe problem that should be 
> addressed especially because no error or warning message is provided.
>
> Thank you in advance,
>
> Daniele Dragoni and Giovanni Pizzi
>
>
>
>
> Data from older revisions:
> (folder name: SVN revision; the 8 qpoints belong to a 4x4x4 mesh, and 
> only qpoints 1,3,8 are also in the 6x6x6 kmesh used for the SCF 
> calculation; input files for SCF and PH follow below)
> 8500/dynmat1:     omega( 1) =      -0.590377 [THz] =     -19.692841 [cm-1]
> 8500/dynmat1:     omega( 2) =      -0.590377 [THz] =     -19.692841 [cm-1]
> 8500/dynmat1:     omega( 3) =      -0.590377 [THz] =     -19.692841 [cm-1]
> 8500/dynmat2:     omega( 1) =     -18.238132 [THz] =    -608.358600 [cm-1]
> 8500/dynmat2:     omega( 2) =     -14.917870 [THz] =    -497.606579 [cm-1]
> 8500/dynmat2:     omega( 3) =      -5.360682 [THz] =    -178.813111 [cm-1]
> 8500/dynmat3:     omega( 1) =       5.116020 [THz] =     170.652049 [cm-1]
> 8500/dynmat3:     omega( 2) =       6.910189 [THz] =     230.499104 [cm-1]
> 8500/dynmat3:     omega( 3) =       9.840192 [THz] =     328.233473 [cm-1]
> 8500/dynmat4:     omega( 1) =     -17.878328 [THz] =    -596.356827 [cm-1]
> 8500/dynmat4:     omega( 2) =     -14.055078 [THz] =    -468.826953 [cm-1]
> 8500/dynmat4:     omega( 3) =       4.075287 [THz] =     135.936946 [cm-1]
> 8500/dynmat5:     omega( 1) =     -20.413217 [THz] =    -680.911617 [cm-1]
> 8500/dynmat5:     omega( 2) =     -20.309404 [THz] =    -677.448788 [cm-1]
> 8500/dynmat5:     omega( 3) =     -20.309404 [THz] =    -677.448788 [cm-1]
> 8500/dynmat6:     omega( 1) =     -18.153438 [THz] =    -605.533512 [cm-1]
> 8500/dynmat6:     omega( 2) =     -13.491302 [THz] =    -450.021393 [cm-1]
> 8500/dynmat6:     omega( 3) =       7.301284 [THz] =     243.544619 [cm-1]
> 8500/dynmat7:     omega( 1) =     -19.842645 [THz] =    -661.879388 [cm-1]
> 8500/dynmat7:     omega( 2) =     -19.842645 [THz] =    -661.879388 [cm-1]
> 8500/dynmat7:     omega( 3) =     -19.842645 [THz] =    -661.879388 [cm-1]
> 8500/dynmat8:     omega( 1) =      10.185086 [THz] =     339.737896 [cm-1]
> 8500/dynmat8:     omega( 2) =      10.185086 [THz] =     339.737896 [cm-1]
> 8500/dynmat8:     omega( 3) =      10.185086 [THz] =     339.737896 [cm-1]
> 9000/dynmat1:     omega( 1) =      -0.592712 [THz] =     -19.770745 [cm-1]
> 9000/dynmat1:     omega( 2) =      -0.592712 [THz] =     -19.770745 [cm-1]
> 9000/dynmat1:     omega( 3) =      -0.592712 [THz] =     -19.770745 [cm-1]
> 9000/dynmat2:     omega( 1) =     -18.237125 [THz] =    -608.324994 [cm-1]
> 9000/dynmat2:     omega( 2) =     -14.916639 [THz] =    -497.565517 [cm-1]
> 9000/dynmat2:     omega( 3) =      -5.358705 [THz] =    -178.747149 [cm-1]
> 9000/dynmat3:     omega( 1) =       5.117581 [THz] =     170.704131 [cm-1]
> 9000/dynmat3:     omega( 2) =       6.911330 [THz] =     230.537139 [cm-1]
> 9000/dynmat3:     omega( 3) =       9.841556 [THz] =     328.278961 [cm-1]
> 9000/dynmat4:     omega( 1) =     -17.877269 [THz] =    -596.321499 [cm-1]
> 9000/dynmat4:     omega( 2) =     -14.053790 [THz] =    -468.783972 [cm-1]
> 9000/dynmat4:     omega( 3) =       4.077984 [THz] =     136.026917 [cm-1]
> 9000/dynmat5:     omega( 1) =     -20.410496 [THz] =    -680.820848 [cm-1]
> 9000/dynmat5:     omega( 2) =     -20.306865 [THz] =    -677.364092 [cm-1]
> 9000/dynmat5:     omega( 3) =     -20.306865 [THz] =    -677.364092 [cm-1]
> 9000/dynmat6:     omega( 1) =     -18.152324 [THz] =    -605.496350 [cm-1]
> 9000/dynmat6:     omega( 2) =     -13.489915 [THz] =    -449.975145 [cm-1]
> 9000/dynmat6:     omega( 3) =       7.302901 [THz] =     243.598557 [cm-1]
> 9000/dynmat7:     omega( 1) =     -19.840424 [THz] =    -661.805304 [cm-1]
> 9000/dynmat7:     omega( 2) =     -19.840424 [THz] =    -661.805304 [cm-1]
> 9000/dynmat7:     omega( 3) =     -19.840424 [THz] =    -661.805304 [cm-1]
> 9000/dynmat8:     omega( 1) =      10.185938 [THz] =     339.766315 [cm-1]
> 9000/dynmat8:     omega( 2) =      10.185938 [THz] =     339.766315 [cm-1]
> 9000/dynmat8:     omega( 3) =      10.185938 [THz] =     339.766315 [cm-1]
> 9500/dynmat1:     omega( 1) =      -0.590362 [THz] =     -19.692343 [cm-1]
> 9500/dynmat1:     omega( 2) =      -0.590362 [THz] =     -19.692343 [cm-1]
> 9500/dynmat1:     omega( 3) =      -0.590362 [THz] =     -19.692343 [cm-1]
> 9500/dynmat2:     omega( 1) =     -18.238237 [THz] =    -608.362092 [cm-1]
> 9500/dynmat2:     omega( 2) =     -14.917871 [THz] =    -497.606611 [cm-1]
> 9500/dynmat2:     omega( 3) =      -5.360685 [THz] =    -178.813195 [cm-1]
> 9500/dynmat3:     omega( 1) =       5.116022 [THz] =     170.652121 [cm-1]
> 9500/dynmat3:     omega( 2) =       6.910191 [THz] =     230.499164 [cm-1]
> 9500/dynmat3:     omega( 3) =       9.840788 [THz] =     328.253358 [cm-1]
> 9500/dynmat4:     omega( 1) =     -17.878327 [THz] =    -596.356802 [cm-1]
> 9500/dynmat4:     omega( 2) =     -14.055044 [THz] =    -468.825819 [cm-1]
> 9500/dynmat4:     omega( 3) =       4.075301 [THz] =     135.937407 [cm-1]
> 9500/dynmat5:     omega( 1) =     -20.413130 [THz] =    -680.908710 [cm-1]
> 9500/dynmat5:     omega( 2) =     -20.309403 [THz] =    -677.448775 [cm-1]
> 9500/dynmat5:     omega( 3) =     -20.309403 [THz] =    -677.448775 [cm-1]
> 9500/dynmat6:     omega( 1) =     -18.153469 [THz] =    -605.534553 [cm-1]
> 9500/dynmat6:     omega( 2) =     -13.491302 [THz] =    -450.021402 [cm-1]
> 9500/dynmat6:     omega( 3) =       7.301280 [THz] =     243.544500 [cm-1]
> 9500/dynmat7:     omega( 1) =     -19.842973 [THz] =    -661.890323 [cm-1]
> 9500/dynmat7:     omega( 2) =     -19.842973 [THz] =    -661.890323 [cm-1]
> 9500/dynmat7:     omega( 3) =     -19.842973 [THz] =    -661.890323 [cm-1]
> 9500/dynmat8:     omega( 1) =      10.185085 [THz] =     339.737862 [cm-1]
> 9500/dynmat8:     omega( 2) =      10.185085 [THz] =     339.737862 [cm-1]
> 9500/dynmat8:     omega( 3) =      10.185085 [THz] =     339.737862 [cm-1]
> 10000/dynmat1:     freq (    1) =      -0.590365 [THz] = -19.692448 [cm-1]
> 10000/dynmat1:     freq (    2) =      -0.590365 [THz] = -19.692448 [cm-1]
> 10000/dynmat1:     freq (    3) =      -0.590365 [THz] = -19.692448 [cm-1]
> 10000/dynmat2:     freq (    1) =     -18.238245 [THz] = -608.362357 
> [cm-1]
> 10000/dynmat2:     freq (    2) =     -14.917869 [THz] = -497.606564 
> [cm-1]
> 10000/dynmat2:     freq (    3) =      -5.360684 [THz] = -178.813180 
> [cm-1]
> 10000/dynmat3:     freq (    1) =       5.116023 [THz] = 170.652156 [cm-1]
> 10000/dynmat3:     freq (    2) =       6.910189 [THz] = 230.499110 [cm-1]
> 10000/dynmat3:     freq (    3) =       9.840795 [THz] = 328.253573 [cm-1]
> 10000/dynmat4:     freq (    1) =     -17.878328 [THz] = -596.356839 
> [cm-1]
> 10000/dynmat4:     freq (    2) =     -14.055079 [THz] = -468.826956 
> [cm-1]
> 10000/dynmat4:     freq (    3) =       4.075245 [THz] = 135.935550 [cm-1]
> 10000/dynmat5:     freq (    1) =     -20.413139 [THz] = -680.909014 
> [cm-1]
> 10000/dynmat5:     freq (    2) =     -20.309404 [THz] = -677.448789 
> [cm-1]
> 10000/dynmat5:     freq (    3) =     -20.309404 [THz] = -677.448789 
> [cm-1]
> 10000/dynmat6:     freq (    1) =     -18.153472 [THz] = -605.534658 
> [cm-1]
> 10000/dynmat6:     freq (    2) =     -13.491303 [THz] = -450.021432 
> [cm-1]
> 10000/dynmat6:     freq (    3) =       7.301284 [THz] = 243.544606 [cm-1]
> 10000/dynmat7:     freq (    1) =     -19.842972 [THz] = -661.890313 
> [cm-1]
> 10000/dynmat7:     freq (    2) =     -19.842972 [THz] = -661.890313 
> [cm-1]
> 10000/dynmat7:     freq (    3) =     -19.842972 [THz] = -661.890313 
> [cm-1]
> 10000/dynmat8:     freq (    1) =      10.185084 [THz] = 339.737838 [cm-1]
> 10000/dynmat8:     freq (    2) =      10.185084 [THz] = 339.737838 [cm-1]
> 10000/dynmat8:     freq (    3) =      10.185084 [THz] = 339.737838 [cm-1]
> 10500/dynmat1:     freq (    1) =      -0.590364 [THz] = -19.692432 [cm-1]
> 10500/dynmat1:     freq (    2) =      -0.590364 [THz] = -19.692432 [cm-1]
> 10500/dynmat1:     freq (    3) =      -0.590364 [THz] = -19.692432 [cm-1]
> 10500/dynmat2:     freq (    1) =     -18.238223 [THz] = -608.361647 
> [cm-1]
> 10500/dynmat2:     freq (    2) =     -14.917870 [THz] = -497.606597 
> [cm-1]
> 10500/dynmat2:     freq (    3) =      -5.360684 [THz] = -178.813170 
> [cm-1]
> 10500/dynmat3:     freq (    1) =       5.116022 [THz] = 170.652112 [cm-1]
> 10500/dynmat3:     freq (    2) =       6.910191 [THz] = 230.499148 [cm-1]
> 10500/dynmat3:     freq (    3) =       9.840786 [THz] = 328.253285 [cm-1]
> 10500/dynmat4:     freq (    1) =     -17.878328 [THz] = -596.356829 
> [cm-1]
> 10500/dynmat4:     freq (    2) =     -14.055010 [THz] = -468.824685 
> [cm-1]
> 10500/dynmat4:     freq (    3) =       4.075346 [THz] = 135.938912 [cm-1]
> 10500/dynmat5:     freq (    1) =     -20.413145 [THz] = -680.909226 
> [cm-1]
> 10500/dynmat5:     freq (    2) =     -20.309403 [THz] = -677.448780 
> [cm-1]
> 10500/dynmat5:     freq (    3) =     -20.309403 [THz] = -677.448780 
> [cm-1]
> 10500/dynmat6:     freq (    1) =     -18.153472 [THz] = -605.534649 
> [cm-1]
> 10500/dynmat6:     freq (    2) =     -13.491302 [THz] = -450.021408 
> [cm-1]
> 10500/dynmat6:     freq (    3) =       7.301281 [THz] = 243.544530 [cm-1]
> 10500/dynmat7:     freq (    1) =     -19.842973 [THz] = -661.890327 
> [cm-1]
> 10500/dynmat7:     freq (    2) =     -19.842973 [THz] = -661.890327 
> [cm-1]
> 10500/dynmat7:     freq (    3) =     -19.842973 [THz] = -661.890327 
> [cm-1]
> 10500/dynmat8:     freq (    1) =      10.185084 [THz] = 339.737839 [cm-1]
> 10500/dynmat8:     freq (    2) =      10.185084 [THz] = 339.737839 [cm-1]
> 10500/dynmat8:     freq (    3) =      10.185084 [THz] = 339.737839 [cm-1]
> 11000/dynmat1:     freq (    1) =      -0.590368 [THz] = -19.692544 [cm-1]
> 11000/dynmat1:     freq (    2) =      -0.590368 [THz] = -19.692544 [cm-1]
> 11000/dynmat1:     freq (    3) =      -0.590368 [THz] = -19.692544 [cm-1]
> 11000/dynmat2:     freq (    1) =     -18.238164 [THz] = -608.359663 
> [cm-1]
> 11000/dynmat2:     freq (    2) =     -14.917869 [THz] = -497.606558 
> [cm-1]
> 11000/dynmat2:     freq (    3) =      -5.360686 [THz] = -178.813223 
> [cm-1]
> 11000/dynmat3:     freq (    1) =       5.116021 [THz] = 170.652098 [cm-1]
> 11000/dynmat3:     freq (    2) =       6.910193 [THz] = 230.499225 [cm-1]
> 11000/dynmat3:     freq (    3) =       9.840788 [THz] = 328.253370 [cm-1]
> 11000/dynmat4:     freq (    1) =     -17.878328 [THz] = -596.356825 
> [cm-1]
> 11000/dynmat4:     freq (    2) =     -14.055063 [THz] = -468.826451 
> [cm-1]
> 11000/dynmat4:     freq (    3) =       4.075210 [THz] = 135.934363 [cm-1]
> 11000/dynmat5:     freq (    1) =     -20.413252 [THz] = -680.912796 
> [cm-1]
> 11000/dynmat5:     freq (    2) =     -20.309404 [THz] = -677.448789 
> [cm-1]
> 11000/dynmat5:     freq (    3) =     -20.309404 [THz] = -677.448789 
> [cm-1]
> 11000/dynmat6:     freq (    1) =     -18.153474 [THz] = -605.534704 
> [cm-1]
> 11000/dynmat6:     freq (    2) =     -13.491302 [THz] = -450.021399 
> [cm-1]
> 11000/dynmat6:     freq (    3) =       7.301282 [THz] = 243.544563 [cm-1]
> 11000/dynmat7:     freq (    1) =     -19.842972 [THz] = -661.890317 
> [cm-1]
> 11000/dynmat7:     freq (    2) =     -19.842972 [THz] = -661.890317 
> [cm-1]
> 11000/dynmat7:     freq (    3) =     -19.842972 [THz] = -661.890317 
> [cm-1]
> 11000/dynmat8:     freq (    1) =      10.185085 [THz] = 339.737859 [cm-1]
> 11000/dynmat8:     freq (    2) =      10.185085 [THz] = 339.737859 [cm-1]
> 11000/dynmat8:     freq (    3) =      10.185085 [THz] = 339.737859 [cm-1]
>
>
> $ cat pw.in
>
> &CONTROL
>  calculation = 'scf'
>  outdir = './out/'
>  prefix = 'aiida'
>  pseudo_dir = './pseudo/'
>  restart_mode = 'from_scratch'
>  verbosity = 'high'
> /
> &SYSTEM
>  degauss =   1.0000000000d-02
>  ecutrho =   3.2000000000d+02
>  ecutwfc =   4.0000000000d+01
>  ibrav = 0
>  nat = 1
>  nspin = 2
>  ntyp = 1
>  occupations = 'smearing'
>  smearing = 'marzari-vanderbilt'
>  starting_magnetization =   3.6000000000d-01
> /
> &ELECTRONS
>  conv_thr =   1.0000000000d-10
>  mixing_beta =   2.0000000000d-01
> /
> ATOMIC_SPECIES
> Fe 55.845 Fe.pbe-spn-rrkjus_psl.0.2.1-tested-pslib030.UPF
> ATOMIC_POSITIONS angstrom
> Fe   0.0000000000       0.0000000000       0.0000000000
> K_POINTS automatic
> 6 6 6 0 0 0
> CELL_PARAMETERS angstrom
>  1.4167999999       1.4167999999       1.4167999999
> -1.4167999999       1.4167999999       1.4167999999
> -1.4167999999      -1.4167999999       1.4167999999
>
>
> $ cat ph.in
>
> AiiDA calculation
> &INPUTPH
>  fildyn = 'dynmat'
>  iverbosity = 1
>  ldisp = .true.
>  nq1 = 4
>  nq2 = 4
>  nq3 = 4
>  outdir = './out/'
>  prefix = 'aiida'
>  tr2_ph = 1.0000000000d-16
> /
>
>
> -- 
> Giovanni Pizzi
> Post-doctoral Research Scientist
> EPFL STI IMX THEOS
> MXC 340 (Bâtiment MXC)
> Station 12
> CH-1015 Lausanne (Switzerland)
> Phone: +41 21 69 31124
>
>
>
>
>
>
> _______________________________________________
> Q-e-developers mailing list
> Q-e-developers at qe-forge.org
> http://qe-forge.org/mailman/listinfo/q-e-developers

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/developers/attachments/20141129/ba2a244b/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/png
Size: 15212 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/developers/attachments/20141129/ba2a244b/attachment.png>


More information about the developers mailing list