T in NVE run increases indefinitely after equilibration

Dear all,

I am experiencing a weird problem with a Si nanowire.

- I run an equilibration run where I rescale the velocities.
- I switch off the velocity rescaling and let the system evolve microcanonically in an NVE run

When I do so, however, the temperature, i.e. the kinetic energy, starts increasing indefinitely and the potential energy conversely goes down.

Notice that the observed behavior seems to be independent:

- on the length of the velocity rescaling run
- on the timestep (up to 0.1 fs)
- on the potential used.
- on switching on/off the fix momentum constraint to avoid rotation and translation
- on the type of box (I like 'p p p' with explicit vacuum, but no changes if I use 's s p')

Notice also that:

- using NVT instead of velocity rescaling doesn't change things
- if use exactly the same script with bulk it works nicely

Any suggestion would be greatly appreciated!

Riccardo

The box is defined in the coordinate file and is:

  -50.00000 50.00000 xlo xhi
  -50.00000 50.00000 ylo yhi
    0.00000 197.52390 zlo zhi

The input script follows below:

# Defining the units

  • I run an equilibration run where I rescale the velocities.
  • I switch off the velocity rescaling and let the system evolve
    microcanonically in an NVE run

When I do so, however, the temperature, i.e. the kinetic energy, starts
increasing indefinitely and the potential energy conversely goes down.

It’s likely the system isn’t fully equilibrated before you start NVE.

Steve