3
$\begingroup$

I did a vc-relax calculation using HSE by Quantum ESPRESSO and after 22h time WALL the job was complete and JOB DONE was mentionned in the end of the output file but I didn't find the optimized coordinates. I was just able to display the latest achieved coordinates by XCRYSDEN. So even fi the Job was done the convergence wasn't achieved.

Here are the stress and pressure achieved, as you can see the pressure is still so high, but normally it should be close to zero.

enter image description here

What can be the causes for not achieving optimized coordinates, please?

Edit: Input file

&CONTROL
.
.
.
verbosity='high',
  tprnfor=.true.,
  tstress=.true.,
  etot_conv_thr =1e-5
  forc_conv_thr =1e-4

/
&SYSTEM

  occupations='fixed'
  input_dft='hse',
  nqx1 = 3, nqx2 =3, nqx3 =3,
  x_gamma_extrapolation = .true.
  exxdiv_treatment = 'gygi-baldereschi'
  input_dft='hse',
/
&ELECTRONS
  conv_thr=1d-07,
  mixing_beta=0.1d0,
  electron_maxstep = 3000
  mixing_mode = 'plain'
/

&IONS

/
&CELL
  cell_dofree='ibrav'
  press=0.d0,
  press_conv_thr=0.2d0
/
.
.
.

```
$\endgroup$

1 Answer 1

4
$\begingroup$

Optimization does not guarantee that you have achieved a well-converged geometry. It only means that the differences in energy, cell-volume and forces calculated between two successive steps are within the convergence criteria that you have defined. As an example, it is possible to converge benzene molecule to a C$_{2v}$ symmetry, while the global minima is D$_{2h}$ symmetry.

$\endgroup$
10
  • $\begingroup$ Ok thank you. But in my case my coordinates didn't get optimized what should I do? $\endgroup$
    – Camilla
    Jun 10 at 13:09
  • $\begingroup$ You can take the last coordinates and restart the optimisation run. $\endgroup$ Jun 10 at 14:33
  • $\begingroup$ @HemanthHaridas shouldn’t the job continue until the target pressure has been achieved? Maybe in this case nstep has been been completed! Or it could also be the case (unlikely though) that the target pressure was mistakenly set to very high! $\endgroup$ Jun 10 at 14:50
  • $\begingroup$ @AbdulMuhaymin No, Optimization is usually a gradient descent algorithm, which means that if the difference between the two successive steps is less than a tolerance, the optimizer does not find a new point to move to. $\endgroup$ Jun 10 at 15:41
  • $\begingroup$ @Camilla, If your question is sufficiently answered, please accept the answer. $\endgroup$ Jun 10 at 17:04

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .