I find a problem when using fix GCMC with water molecules.
I have noticed that both the command “molecule” and “fix GCMC” have an option for offsets of atom types. For “molecule”, we can set the offset value by the keyword “toff”; for “fix GCMC”, we can specify the offset by the keyword “type” for inserted molecules. Assuming the offset value to be 2, I guess the following two cases should be the same:
“molecule ID file offset 2 0 0 0 0
…
fix ID group-ID gcmc N X M 0 seed T mu displace keyword values …”
“molecule ID file offset 0 0 0 0
…
fix ID group-ID gcmc N X M 2 seed T mu displace keyword values …”
However, the results of simulation just proves that I am wrong and I don’t know why.
I have attached the input file and the molecule file. I was trying to model the adsorption behavior of water molecules in a slit surrounded by SiO2.
I find a problem when using fix GCMC with water molecules.
I have noticed that both the command “molecule” and “fix GCMC” have an option for offsets of atom types. For “molecule”, we can set the offset value by the keyword “toff”; for “fix GCMC”, we can specify the offset by the keyword “type” for inserted molecules. Assuming the offset value to be 2, I guess the following two cases should be the same:
“molecule ID file offset 2 0 0 0 0
…
fix ID group-ID gcmc N X M 0 seed T mu displace keyword values …”
“molecule ID file offset 0 0 0 0
…
fix ID group-ID gcmc N X M 2 seed T mu displace keyword values …”
However, the results of simulation just proves that I am wrong and I don’t know why.
I have attached the input file and the molecule file. I was trying to model the adsorption behavior of water molecules in a slit surrounded by SiO2.
I want to perform the simulations of a rod ( modelled by a group of 20 beads in a line) in water using fix rigid command with langevin keyword.
The command is like:
fix 1 rod rigid single langevin 1.0 1.0 1.0 428984
Howevery, as we know, the rod is asymmetrical. That means the temperature damping parameter is anisotropic(it is different along the long axis and short axis) .
Is it reasonable to use the same damping parameter in the fix command for rod-like object moving in water.
I need some comments and suggestions to deal with the above.
I want to perform the simulations of a rod ( modelled by a group of 20 beads
in a line) in water using fix rigid command with langevin keyword.
The command is like:
fix 1 rod rigid single langevin 1.0 1.0 1.0 428984
Howevery, as we know, the rod is asymmetrical. That means the temperature
damping parameter is anisotropic(it is different along the long axis and
short axis) .
what kind of model/potential are you using for the beads?
for conventional point particles, your moment of inertia for rotation
around the long axis would be infinity and thus the rotations cannot
happen at all around that axis.
OOOOOOOOOOOOOO
Like the above, the rod-like rigid body is a group of beads in a line.
The frictional drag is different for the translational motion when it moves along the long axis and perpendicular to the long axis of the rod. So, we may need the anistropic damping parameter for center-of-mass motion.
Trung can comment, but I believe the thermostats
for rigid bodies (langevin or nvt) are simply thermostatting
the 6 DOF for a rigid body (translation of COM, rotation),
and I don’t think there is anything special that needs
to be done for bodies of different shapes.
Trung can comment, but I believe the thermostats
for rigid bodies (langevin or nvt) are simply thermostatting
the 6 DOF for a rigid body (translation of COM, rotation),
and I don't think there is anything special that needs
to be done for bodies of different shapes.
this can be easily rationalized by considering that due to the
thermalization of the rotational DOFs, the shape to be considered for
translation should be the average over all orientations and that would
be determining the parameter.
another option to consider would be to use the regular fix langevin
which would thermalize individual beads and that should thus be shape
agnostic as well.
Hi Wade
I don't know how LAMMPS handles rigid bodies, but either of the
methods that Axel and Steve mentioned are adequate to maintain the
system at constant temperature. (As long as the damping is not too
rapid, and you can wait long enough, the simulation should eventually
generate conformations that can be used for realistic thermodynamic
averages.)
But, generally speaking, I would not expect to get realistic
dynamics (as opposed to thermodynamics) using fix langevin. When you
have molecules, some of the atoms interact with the surrounding
solvent. Interior atoms do not. And friction forces on different
atoms is correlated. (Atoms that push on the solvent create currents
which push on other atoms.) There are probably other issues I am not
even thinking of. I've never had to worry about this. If these
effects are important to you, you may have to use a different
approach.
Hopefully it's not necessary to worry about these effects to answer
question you are pursuing with your simulation.
I'm curious to hear what direction this thread goes...
I found that there was indeed an error in how the gcmc offset is
handled. I wish I had seen this before I responded to Chen's repeat
post. One more reason why questions should only be posted once.
looking at the source code (fix_rigid.cpp:893-910), I can see that the Langevin thermostat is coupled with the rotational DOF of the rigid bodies through random and viscous torques. The magnitude of these torques are proportional to the corresponding moment of inertia and the angular velocity, so for the rods you’re modeling, whose the moment of inertia along the director is zero, the added torques are zero along the director. I think this is a reasonable approximation as to capture the asymmetric damping and noise components of the Langevin thermostat coupled to the body as a whole, though it would be more helpful if some references were given for the formulae involving the rotational DOF.
In my research, I often apply the Langevin thermotat to individual beads, one of the approaches Axel mentioned. For shapes more complex than rods, like patchy spheres, tetrahedra, cubes, etc, the Langevin thermostat is applied to the beads on the body surface only. In my studies I was considering thousands of rigid bodies and interested in the equilibrium behavior, not the dynamics, of the system. In such circumstances, the Langevin thermostat works well to maintain the system temperature at set values. And if I recall correctly, for sufficiently large systems and long simulation runs, the body translational and rotational velocities obtained also conform to canonical distributions.
If you are interested in the Brownian motion of a single body, I would suggest you go through the literature with keywords like Brownian dynamics and rotational diffusion. From that, you may, or may not, want to modify/extend LAMMPS source code to suit your needs.