[Pw_forum] problem with conjuguate gradient in cp

[Merlin Meheut]

Dear PWSCF users,

I am trying to realize a molecular dynamics using cp.x, and I find some
difficulties. In particular, the kinetic energy is slightly deriving, and I
would like to cool down the electrons, using the cg electron_dynamics
option, as proposed in the espresso-5.1/CPV/Doc/INPUT_CP.def:

             'cg'      : conjugate gradient is used to converge the
                         wavefunction at each ionic step. 'cg' can be used
                         interchangeably with 'verlet' for a couple of ionic
                         steps in order to "cool down" the electrons and
                         return them back to the Born-Oppenheimer surface.
                         Then 'verlet' can be restarted again. This
procedure
                         is useful when electronic adiabaticity in CP is
lost
                         yet the ionic velocities need to be preserved.


So I do molecular dynamics (with electron_dynamics = 'verlet' ) then I
change to 'cg'.
Unfortunately, the calculation is systematically stuck .  The last lines of
the output files being:

-------------------------------------------------------------
...
   reading restart file: /tmpdir/mmeheut/Ca++/CaCl2-62wrl_56.save
   restart file read in    2.062 sec.


   formf: eself=  1290.97722
   formf:     vps(g=0)=  -0.0010930     rhops(g=0)=  -0.0007593
   formf: sum_g vps(g)=  -0.7938444 sum_g rhops(g)=  -0.8983221
   formf:     vps(g=0)=  -0.0011291     rhops(g=0)=  -0.0005315
   formf: sum_g vps(g)=  -1.8515153 sum_g rhops(g)=  -0.6288255
   formf:     vps(g=0)=  -0.0005757     rhops(g=0)=  -0.0004585
   formf: sum_g vps(g)=  -1.5145512 sum_g rhops(g)=  -1.0525002
   formf:     vps(g=0)=  -0.0000296     rhops(g=0)=  -0.0000770
   formf: sum_g vps(g)=  -0.0789760 sum_g rhops(g)=  -0.7182652
   Delta V(G=0):   0.071172Ry,    1.936684eV
 PERFORMING CONJUGATE GRADIENT MINIMIZATION OF EL. STATES
-------------------------------------------------------------

In the log file of the job I have the following message:
 forrtl: severe (174): SIGSEGV, segmentation fault occurred

Here is my input:


 &control
       calculation = 'cp',
      restart_mode = 'restart', ! 'reset_counters',
            prefix = 'CaCl2-62wrl',
           disk_io = 'default' ,
    pseudo_dir     = '/tmpdir/mmeheut/Ca++',
    outdir         = '/tmpdir/mmeheut/Ca++',
    tprnfor        = .true.,
    tstress        = .true.,
    dt             =  5.0, ! 0.2 fs
    ndr = 56,   ! 51: relaxe ini ; 51-52: heat50K; 52-53: heat150K ; 56:
simmer10
    ndw = 57,
    nstep          =  800,
    iprint         =  10,
    isave          =  100,
    etot_conv_thr = 3.d-5,
    ekin_conv_thr = 1.d-6,
    forc_conv_thr = 3.d-4,
/&end
 &system
    ibrav = 0, celldm(1)=23.4752,
    nat = 189, ntyp = 4, ecutwfc = 40.0, ecutrho=200.0,
    nr1b=20, nr2b=20,nr3b=20,
/&end
 &electrons
    emass = 400.d0,
    emass_cutoff = 2.5d0,
    orthogonalization = 'ortho',
    ortho_eps = 5.d-8,
    ortho_max = 25,
    electron_dynamics = 'cg',
    electron_velocities='default',
    electron_temperature='not_controlled',
 /
 &ions
    ion_dynamics = 'verlet',
    ion_radius(1) = 1.0d0,
    ion_radius(2) = 1.0d0,
    ion_radius(3) = 0.8d0,
    ion_radius(4) = 0.5d0,
    ion_velocities = 'default',
    ion_temperature = 'not_controlled',
    tempw=300.0,
    fnosep = 100.0,
    nhpcl  = 1,
    nhptyp = 0,
 /
 &cell
    cell_dynamics = 'none',
    cell_velocities = 'zero',
    press = 0.0d0,
    wmass = 70000.0d0
 /

ATOMIC_SPECIES
  Ca   39.9625   ca_pbe_v1.uspp.F.UPF
  Cl   34.9689   cl_pbe_v1.4.uspp.F.UPF
  O    15.9949   o_pbe_v1.2.uspp.F.UPF
  H     1.0079   h_pbe_v1.4.uspp.F.UPF

CELL_PARAMETERS (alat= 23.47520000)
   1.000000000   0.000000000   0.000000000
   0.000000000   1.000000000   0.000000000
   0.000000000   0.000000000   1.000000000

ATOMIC_POSITIONS (angstrom)
Ca       2.246613456   8.592421018   8.991482765
...

I tried with versions 5.1. and 5.3. and acheved the same result. Would you
have any hints to help me move on?

Thank you in advance for your help,

Best regards,

-- 
Merlin Méheut (associate professor)
adresse labo:
GET - OMP   - Université Paul Sabatier
14 avenue Edouard Belin
31400 Toulouse
FRANCE
tel: (+33) 5 61 33 26 17
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20170904/603065eb/attachment.html>