[Pw_forum] accuracy of stress determination of vc-relax

[Uri Argaman]

Dear QE users and developers
I do vc-relax with BFGS algorithm as follows:

&CONTROL
  calculation = "vc-relax",
  pseudo_dir  = '/home/uriargaman/espresso-5.0.3/pseudo/',
  outdir      = '/home/uriargaman/tmp/tmp28898/',
  tstress=.true.
  verbosity='high',
 /
&system
    ibrav=  4, celldm(1) =5.548454954360, celldm(3) =1.581979631777413 ,
nat= 2, ntyp= 1,
    ecutwfc =80, ecutrho=1200,
    occupations='smearing', smearing="methfessel-paxton", degauss=0.01,
 /
&ELECTRONS
  conv_thr    = 1.D-9,
  electron_maxstep=1000
  mixing_beta = 0.7D0,
/
&IONS
   ion_dynamics="bfgs",
/
&cell
  cell_dynamics="bfgs",
  press=0.6,
  press_conv_thr=0.01,
/
ATOMIC_SPECIES
  Ti   47.9  Ti.pbe-sp-van_ak.UPF
ATOMIC_POSITIONS (crystal)
   Ti      0.666666666    0.333333333    0.250000000    0  0  0
   Ti      0.333333333    0.666666666    0.750000000
K_POINTS {automatic}
 24 24 24 0 0 0

In the last iteration (after begin final coordinates) I got stress:
          total   stress  (Ry/bohr**3)                   (kbar)     P=
0.71
   0.00000485  -0.00000000  -0.00000000          0.71     -0.00     -0.00
  -0.00000000   0.00000485   0.00000000         -0.00      0.71      0.00
  -0.00000000   0.00000000   0.00000481         -0.00      0.00      0.71

and the stress in the iteration before (last iteration before begin final
coordinates):
  total   stress  (Ry/bohr**3)                   (kbar)     P=    0.73
   0.00000499  -0.00000000   0.00000000          0.73     -0.00      0.00
  -0.00000000   0.00000499   0.00000000         -0.00      0.73      0.00
   0.00000000   0.00000000   0.00000495          0.00      0.00      0.73

I did'nt reach the pressure threshold although I have very good convergence
with respect to all numerical parameters. The threshold is achieved only
for hydrostatic condition but not for the value of the stress (P=0.73 kbar
instead of 0.6 kbar). Is the parameter "press_conv_thr" is just for the
hydrostatic condition and not for the absolute value of the stress? How can
I get more accurate stress?
Thank you very much
Uri Argaman
Ben-Gurion University
Israel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20141211/cb208543/attachment.html>