[CPMD-list] (internal) stress tensor calculation

[luca]

Dear Everyone,
                            I am concerned in NPT calculations with a
liquid sample.
Working with a low cut-off (35Ry and smoothing parameters 40 6 25,
wishing an effective cutoff at 25Ry), I would have liked to verify the
pressure with a wavefunction optimization (NVT, fixed cell) after
equilibration.
It turned out that the external/internal pressure indicated in a NPT md
run is always higher than the converged value for static optimization in
NVT. Furthermore the value to which the pressure is converging in NPT
(increasing the cutoff) does not seem to be the same as NVT.

I do normally operate as this: I take some snapshots from the
equilibrated trajectory in NPT and I optimize the wavefuntion in those
configuration at different cutoffs, with the statement STRESS TENSOR 1.
A typical input file for that looks like:
&CPMD
  OPTI WAVE
  STORE
  10
  RESTFILE
  5
  TIMESTEP
  5
  EMASS
  1000
  MAXSTEP
  30000
  STRESS TENSOR
  1
&END

&SYSTEM
  SYMMETRY
  14
  CELL
  21.352719 1. 1. 0.5 0.5 0.5
  CUTOFF
  25.
&END

The behavior of calculated pressure vs. cutoff is roughfly always the
same, decreasing from 25 to 35 Ry , then increasing till 85, then it
starts to oscillate around a possible converged value. The distance from
the NPT value and the relative distances at different cutoffs stay, with
good approximation, the same. This last fact suggest me a sort of
sistematic error in the calculation of the stress tensor.
Furthemore, it appears that, when imposing an external pressure in NPT,
the system is equilibrating indeed toward some pressure, the exact value
of which is puzzling me at the moment.
Looking into the code, it seems that the calculation of the stress is
carried out always in the subroutine STRESS, but I am not able to
determine wether the parameter passed are the same in NPT or in NVT.

Have anyone noticed strange behaviors in calculating the pressure?
Why, optimizing the wavefunction in NPT, the stress tensor is not
calculated?
Looking into the code I found a (hidden) statement one can use in
&CPMD...&END, that is CLASSTRESS: how does it differ from STRESS TENSOR?

Regards,
Luca Ghiringhelli.