[lammps-users] about temperature control

Dear sir,
i use LAMMPS to do this simulation: fluid is confined between two solid wall, the wall atoms was fixed with a spring force to its original place with a langevin thermostat, and use a nve intergrator to the fluid. the starting temperature is higher than the wall T,so with the time going, the fluid will be cooled by the wall,and finally reaches the same T with the wall. My problem is that when the fluid is simple LJ model fluid, the finally fluid is consistent with the wall T, and the total energy decreases (this is reasonable);but when it comes to water, the water temperature decreases at the first stage, but it is always some degrees higher than the wall temperature, and the energy increases in the run, this seems unreasonable, i am confused about this problem for a long time. I am very grateful for any suggestion.
yours sincerely

Ge Song
PHD candidate
Engineering mechanics,
Tsinghua university,
Beijing, China

Dear sir,
     i use LAMMPS to do this simulation: fluid is confined between two
solid wall, the wall atoms was fixed with a spring force to its
original place with a langevin thermostat, and use a nve intergrator
to the fluid. the starting temperature is higher than the wall T,so
with the time going, the fluid will be cooled by the wall,and finally
reaches the same T with the wall. My problem is that when the fluid is
simple LJ model fluid, the finally fluid is consistent with the wall
T, and the total energy decreases (this is reasonable);but when it
comes to water, the water temperature decreases at the first stage,
but it is always some degrees higher than the wall temperature, and
the energy increases in the run, this seems unreasonable, i am
confused about this problem for a long time. I am very grateful for
any suggestion.

it is impossible to say anything without knowing any details.
what you describe sounds as it your water MD is not conserving
energy. this can have various reasons. most often it is too large
a time step, or improper treatment of coulomb interactions (which
are far more long range than LJ).

i would suggest to stick your nose in a text book on MD simulations
and look at the chapters where conservation of energy is discussed.

cheers,
   axel.