hi dears , for simulating poiseuille water flow in nanochannel with copper walls, is the berendsen tharmostat good for hold temperature at 300K ?
so entirely what does the berendsen thermostat?
can berendsen thermostat interrupt the velocity profile?
in my simulation(poiseuille water flow in nanochannel),not langevin,not another way,whatever you tell,none of them can not hold the temperature of system at 300K EXCEPT berendsen thermostat , but with berendsen the velocity profile became smaller than correct value.
whether the berendsen thermostat can decay velocity profile?
The simple solution is to apply the thermostat only to velocity
components perpendicular to the flow direction. Equipartition
principle will ensure that the velocity component in the flow
direction will be close to the right temperature i.e. even though the
streaming velocity will vary spatially, the temperature will not. This
works with all the standard thermostat fixes, but you need to add in a
bias temperature compute:
http://lammps.sandia.gov/doc/compute_temp_partial.html
A more complicated solution is to use a time-averaged bias velocity
profile in the temperature calculation:
http://lammps.sandia.gov/doc/compute_temp_profile.html
Aidan
I will just add that you should be really worried by the fact that
your results depend strongly on the thermostatting method. Thermostats
should only modify slightly the dynamics of the system in order to
prevent long-term drift of the temperature... It is really strange
that different thermostats are not able to perform the task they have
been designed to on your system. If your system does not behave
reasonnably without thermostat, then you should solve the bad behavior
before applying a thermostat. Otherwise the thermostat may hide the
problem, but it will not solve it.
That is good general advice. However, under Poisseuille flow, which
presumably involves an external field doing work on the system, the
simulation will fail without a thermostat to remove energy. Also, if
a standard thermostat is applied in the direction of the field, the
behavior is also not well-defined, and will be different for different
thermostatting methods.
If the thermostat is applied correctly using a suitable bias
temperature, then yes, the results should not depend on the details of
the thermostat. Generally this will be true only if the maximum
streaming velocity is small compared to the thermal velocities of the
atoms.
Having some experience with Poiseuille flow simulations in
nanochannels, I agree that the detail of the thermostat will affect
the flow profile. However, my experiences have told me that it is
usually possible to reach a sufficiently small external forcing that
all thermostating methods provide the same flow profile. Of course one
can see it the other way, and wonder what thermostating method affects
the flow less when going to large external forcing (to get a better
signal/noise ratio), and I fully agree there with your suggestions.
When dealing with simple flows along one given direction, I got the
best results (in terms of linear response up to large external
forcing) when using a Nosé-Hoover thermostat, applied only to the
degrees of freedom perpendicular to the flow (when dealing with more
complex, 3D flows, and after thorough tests, I switched to DPD
thermostating with suitable parameters).
But here Ilia Iliana is writing that other thermostating methods are
not even able to maintain the temperature! To me this indicates a bad
dynamics, but it is hard to be specific in the absence of more
details.
Best,
Laurent