Dear all
I'm trying to perform a bunch of calculations using the
Ag.pbe-dn-kjpaw.UPF PP contained in the pslibrary 0.2.5 and compiled
with the 5.0 QE version. Tests on the unit cell of Ag were fine; thus
I was moving to the next step: a (large) Ag slab, which I had
succesfully optimised in the past by using an USPP. After a first
succesful scf step, with HF forces on ions reasonably low, the
calculation (input below) stops complaining
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Error in routine mix_rho (1):
negative dr2
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
I've tried to mix every (reasonable) value of mixing_beta with the
available choices of mixing_mode, but I obtain always the same error.
As the calculations have been performed on the new fermi at cineca
bluegene machine (still plenty of linear algebra library problems,
cholesky errors, ... But not in this case!!!), I've tested the same
input on a more conventional intel/mkl linux cluster, obtaining the
same error at the same step.
Every suggestion will be appreciated...
Yours
Giuseppe
This is my inpfile
&control
calculation = 'relax'
restart_mode='restart',
prefix='ag100-k111s-dftd2',
pseudo_dir = '/fermi/home/userexternal/gmattiol/PP_UPF/',
outdir='/gpfs/scratch/userexternal/gmattiol/test_qe/pw_plain/paw/././run/tmp/',
nstep=200,
tprnfor=.true.,
/
&system
ibrav=6, celldm(1)=27.676853, celldm(3)=1.330000,
nat= 100, ntyp= 1,
ecutwfc = 55.0,
occupations='smearing', degauss=0.01,
nspin=1,
london=.true.,
/
&electrons
mixing_mode = 'plain'
mixing_beta = 0.1
conv_thr = 1.0d-6
electron_maxstep = 200
/
&ions
ion_dynamics='bfgs'
/
ATOMIC_SPECIES
Ag 107.868 Ag.pbe-n-kjpaw.UPF
ATOMIC_POSITIONS {angstrom}
Ag 3.807926690 12.302649620 0.000000000 1 1 0
Ag 6.151324080 14.060198150 0.000000000 1 1 0
Ag 2.635668595 4.100351315 0.000000000 1 1 0
...
K_POINTS {automatic}
1, 1, 1, 1, 1, 1
Giuseppe Mattioli
ISM-CNR
Italy
