<p dir="ltr">I have a personalised version of the random_matrix routine that fills it with normal-distributed numbers. This in theory should cause the eigenvalues to be uniformly spaced. I have never uploaded it because I have no proof that it works better than the usual one, but if someone wants to try it in some difficult corner case, please contact me </p>
<p dir="ltr">-- <br>
Lorenzo Paulatto<br>
Written on a virtual keyboard with real fingers</p>
<div class="gmail_extra"><br><div class="gmail_quote">On 15 Nov 2016 2:07 p.m., "Antoine JAY" <<a href="mailto:antoine.jay@polytechnique.edu">antoine.jay@polytechnique.edu</a>> wrote:<br type="attribution"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">To avoid this error message and to be able to do the phonon calculation<br>
without modifying the file "random_matrix.f90", I slightly modified the<br>
last decimals of the input atomic positions and lattice parameters.<br>
This does not change significantly the the inter atomic forces which are<br>
still quite null (<10^-5 Ry/au) nor the stress tensor, but this noise is<br>
enough to separate non-degenerated modes.<br>
<br>
Antoine Jay<br>
<br>
<br>
<br>
Answering myself: later, for a specific k-point. Quick example showing the<br>
problem attached.<br>
<br>
<br>
Paolo<br>
<br>
On Tue, Nov 8, 2016 at 7:12 PM, Paolo Giannozzi <<a href="mailto:p.gianno...@gmail.com">p.gianno...@gmail.com</a>><br>
wrote:<br>
<br>
> When does this happen? at the beginning of the phonon calculation or later?<br>
><br>
> Paolo<br>
><br>
> On Tue, Nov 8, 2016 at 1:22 PM, Antoine JAY <<a href="mailto:antoine....@polytechnique.edu">antoine....@polytechnique.edu</a><br>
> > wrote:<br>
><br>
>> Hello community & developers,<br>
>><br>
>> Performing phonons on gamma-boron at 30 GPa (using GGA), I have this<br>
>> message:<br>
>> from set_irr_sym_new : error # 2111<br>
>> wrong representation<br>
>><br>
>> I send down the atomic positions as I know from previous mail (Lorenzo<br>
>> Paulatto 2012-07-23) that you need some tests to find a reliable way to<br>
>> generate the initial random pseudo-dynamical matrix (random_matrix.f90).<br>
>> Lattice and atomic positions are already converged.<br>
>><br>
>> Increasing the threshold 10^-6 to 10^-4 line 156 and 158 of<br>
>> set_irr_sym.f90<br>
>> does not solve this issue.<br>
>><br>
>> Do you have any other suggestion?<br>
>><br>
>> Thank you very much<br>
>><br>
>> Antoine Jay<br>
>><br>
>> &control<br>
>> calculation = 'scf',<br>
>> prefix = '$name.$Pression',<br>
>> tstress = .true.,<br>
>> tprnfor = .true.,<br>
>> pseudo_dir = '$PSEUDO_DIR/',<br>
>> outdir = '$TMP_DIR/',<br>
>> etot_conv_thr=1.0d-6,<br>
>> forc_conv_thr=1.0d-4,<br>
>> nstep=1400,<br>
>> wf_collect=.true.<br>
>> /<br>
>> &system<br>
>> ibrav =8<br>
>> celldm(1) = 9.213227<br>
>> celldm(2) = 1.113897<br>
>> celldm(3) = 1.365520<br>
>> nat= 28,<br>
>> ntyp= 5,<br>
>> ecutwfc =80,<br>
>> /<br>
>> &electrons<br>
>> diagonalization='david'<br>
>> mixing_beta = 0.7<br>
>> conv_thr = 1.0d-13<br>
>><br>
<br>
______________________________<wbr>_________________<br>
Pw_forum mailing list<br>
<a href="mailto:Pw_forum@pwscf.org">Pw_forum@pwscf.org</a><br>
<a href="http://pwscf.org/mailman/listinfo/pw_forum" rel="noreferrer" target="_blank">http://pwscf.org/mailman/<wbr>listinfo/pw_forum</a><br>
</blockquote></div></div>