Hi forum,
I am simulating a polymer melt. I want to maintain the temperature of the system constant.
Using the Langevin thermostat resulted in considerable slowdown in the dynamics of the
system. So I decided to thermostat the wall instead. In my view “fix spring/self” command
with an appropriate spring constant is the best choice. But just wondering whether there are
better ways maintain the constant temperature without slowing down the actual dynamics of the melt.
Srikanth
Re: thermostatting the wall (wasn't this a Pink Floyd album?)
You can thermostat only the mobile wall atoms if you wish
by putting them in a "group" and giving that as the argument
to fix langevin or some other thermostat fix. But if you use
fix langevin on the melt and it stays at the right Temp, then
I don't know what you mean by slowing down the dynamics.
The dynamics is what it should be at T = Temp.
Steve
Re: thermostatting the wall (wasn't this a Pink Floyd album?)
You can thermostat only the mobile wall atoms if you wish
by putting them in a "group" and giving that as the argument
to fix langevin or some other thermostat fix. But if you use
fix langevin on the melt and it stays at the right Temp, then
I don't know what you mean by slowing down the dynamics.
yeah, i would suspect that this is not a thermostatting
issue, but rather could be related to finite size effects.
cheers,
axel.
Sorry, probably I didn’t phrase the problem properly. With fix langevin though
the temperature of the system remains constant the system takes long time
to equilibrate and this was expected. So I was looking for other means of maintaining
constant temperature which can speed up the equilibration process.
Cheers
Srikanth
If you're asking how to equilibrate/relax a polymer melt
faster due to slow polymer dynamics, then that's a hard
problem. You can look at fix bond/swap which does MC
moves to attempt this for a specific kind of bead-spring
system.
Steve
Hi Srikanth
If you are using fix langevin you can speed up the dynamics by increasing
the damping parameter. You can do this as long as your total energy
remains constant ( the system does not blow up) or as long as your
timestep allows.
for my equilibration process, I use LJ units and I use 7.0 tau for my
damping constant in fix langevin with a timestep of 0.01. And, if you use
Lennard Jones atoms you can set sigma smaller or set the interaction
parameter epsilon_lj to zero so that the atoms in the melt will just pass
trough each other making the equilibration process faster. You should have
a simulation run of at least one Rouse time (approx. N^2 tau where N is
the degree of polymerization of your polymer and 1 tau is equal to
timestep * MD step) and then check if your chain in the melt is Gaussian.
After the equilibration process, you can use fix nve/limit to grow back
your atoms (adjust sigma to the correct sigma in LJ) and change the
interaction parameters as well. With fix nve/limit, you could have some
overlapping atoms but the simulation will continue. Run as long as needed
and then you can unfix nve/limit and change to your desired integrator
during your production runs.
Jan-Michael