[Pw_forum] Phonon crash

[Andrea Dal Corso]

Presently the maximum number of k points is npk    = 40000, but you need
at least 64000 for this q point. Increasing this parameter in
Module/parameters.f90 and recompiling should work.

Andrea




On Wed, 2007-03-14 at 10:54 +0100, Miguel Martínez Canales wrote:
> Dear fellows,
> 
> I have found that, when running a phonon calculation under certain 
> conditions, it crashes. More concretely, on fcc systems (I haven't checked 
> diamond structures) if the scf run uses a 20 20 20 monkhorst-pack grid and 
> the phonon calculation uses ldisp=.true. for a 8*8*8 regular grid the 
> phonon run will crash at dyn23 (q = 0.625 -0.375  0.875). The last lines of 
> the phonon output are:
> 
>       Calculation of q =   0.6250 -0.3750  0.8750
> 
>  
> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
>       from set_kplusq : error #     32000
>       too many k points
>  
> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
> 
>       stopping ...
> 
> It really puzzles me that it stops on the 23rd point, and not sooner. Maybe 
> it's related to the star of q?
> 
> If I specify this point by hand, doing a nscf run and then a phonon run at 
> that qpoint, it will also crash. It depends neither on the lattice 
> parametre, nor the the pseudpotential (I've done it with Na LDA, Li LDA and 
> Li PBE). It won't crash if I reduce the k-point grid, and it won't crash 
> with the same grid on a bcc lattice.
> 
> I am using espresso-3.2.1 plus the patch that Paolo submitted to make 
> -npool work with ph.x. It's been compiled with ifort 8.1 and mkl 9.0 in an 
> itanium 2 machine. Any ideas?
> 
> Thanks for your comments,
> 
> Miguel
> 
> In case the input is relevant, I attach it:
> 
> 
> # self-consistent calculation
> cat > na.scf.in << EOF
>   &control
>      calculation='scf'
>      restart_mode='from_scratch',
>      tstress = .true.
>      tprnfor = .true.
>      prefix='na',
>      pseudo_dir = '$PSEUDO_DIR/',
>      outdir='$TMP_DIR/'
>   /
>   &system
>      ibrav=  2, celldm(1) = $alat , nat=  1, ntyp= 1,
>      ecutwfc = 30.0
>      occupations='smearing', smearing='m-p', degauss= 0.010
>   /
>   &electrons
>      conv_thr =  1.0d-8
>      mixing_beta = 0.7
>   /
> ATOMIC_SPECIES
>   Na  22.989768  Na.pz-n-vbc.UPF
> ATOMIC_POSITIONS
>   Na 0.00 0.00 0.00
> K_POINTS {automatic}
>   20 20 20   1 1 1
> EOF
> $ECHO "  running the scf calculation...\c"
> $PW_COMMAND < na.scf.in > na.scf.out
> $ECHO " done"
> 
> # phonon calculation
> cat > na.ph.in << EOF
> phonons of Na
>   &inputph
>    tr2_ph=1.0d-16,
>    prefix='na',
>    ldisp=.true.,
>    nq1=8, nq2=8, nq3=8
>    amass(1)=22.989768,
>    outdir='$TMP_DIR/',
>    fildyn='na.dyn',
>   /
> EOF
> $ECHO "  running the phonon calculation ...\c"
> $PH_COMMAND < na.ph.in > na.ph.out
> $ECHO " done"
> 
> 
-- 
Andrea Dal Corso                    Tel. 0039-040-3787428
SISSA, Via Beirut 2/4               Fax. 0039-040-3787528
34014 Trieste (Italy)               e-mail: dalcorso at sissa.it