[Pw_forum] HSE with USPP

[Xiaoming Wang]

Hello qe users and developers,

 

I am trying the new feature as in QE-5.2.0, which implement USPP with hybrid
functional calculations. However, as for my calculations, I found errors in
output which depend on number of cores I used.  The error reads " Error in
routine gk_sort (1):  array gk out-of-bounds" when I use 16 cores as 'mpirun
-np 16 pw.x', but it works for 4 cores. However, for NCPP, I found no such
problems, at least for my system. Does anyone know what the problem is? And
the following is my input:

----------------------------------------------------------------------------
------------------------------

&control

    calculation = 'scf'

    restart_mode='from_scratch',

    prefix='bp',

    tstress = .true.

    tprnfor = .true.

    etot_conv_thr = 1.0D-6

    forc_conv_thr = 1.0D-4

    pseudo_dir = ' /pseudo',

    outdir='./tmp'

    wf_collect=.true.

/

&system

    ibrav=8

    a=3.33826

    b=4.59076

    c=13.0

    nat=4

    ntyp=1

    ecutwfc=30

    ecutrho=200

    occupations='fixed'

    nosym=.true.

    noinv = .true.

    input_dft = 'HSE'

    nqx1=5, nqx2=4, nqx3=1

/

&electrons

    diagonalization='david'

    mixing_mode = 'plain'

    mixing_beta = 0.7

    conv_thr =  1.0d-8

/

ATOMIC_SPECIES

P  30.9738  p_pbe_v1.uspp.F.UPF

K_POINTS automatic

10 8 1 0 0 0

ATOMIC_POSITIONS (crystal)

P       0.500000      0.914274      0.002447

P       0.000000      0.585727      0.002447

P       0.000000      0.414273      0.167618

P       0.500000      0.085726      0.167617

----------------------------------------------------------------------------
-----------------------------------------

 

Best,

 

Xiaoming Wang

 

IAMDN, Rutgers

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20150701/974c79cf/attachment.html>