Hello,
In my recent work I was simulationg Coarse-Grained polymer chains (one 1000-mer composed of 3000 CG atoms) with implicit solvent using Langevin dynamics. Surprisingly, I realised that the center of mass of the whole simulation system (i.e. COM of all atoms involved) remained at the same position. I tried to read through the manual but didn’t find anything decribing this.
fix constnve all nve
fix langDyn all langevin 294.13 294.13 1000 194312 zero yes
compute langTemp all temp/com
fix_modify langDyn temp langTemp
velocity all create 294.13 638479 mom yes dist uniform
This is a section of my LAMMPS input file. I was told that I need to use “zero yes” in the command line to remove COM velocity bias. That’s why I did it.
The problem is the dynamic of chain in simulation was considerablly slower than what previous literature described and what we expected. Although I was told that the dynamic of system is not so important as we were interested in the equilibrium structure of polymer chain, I was kind of worry about the fact that the COM was fixed for such single chain system would actually have an impact on our results. Honestly some of the simulated chain conformations behaved weird according to our knowledges. I wonder if there is any way to solve this problem or COM is always fixed for a Langevin Dynamic simulation?
Thanks for your time.
Regards
Ming
Hello,
In my recent work I was simulationg Coarse-Grained polymer chains (one 1000-mer composed of 3000 CG atoms) with implicit solvent using Langevin dynamics. Surprisingly, I realised that the center of mass of the whole simulation system (i.e. COM of all atoms involved) remained at the same position. I tried to read
why is this surprising? usually, people are concerned if it is
not the case since it can lead to the so-called flying icecube
syndrome (= all kinetic energy being converted into a COM
drift at the expense of relative velocities and ultimately the
system freezing with a large COM velocity).
through the manual but didn’t find anything decribing this.
fix constnve all nve
fix langDyn all langevin 294.13 294.13 1000 194312 zero yes
compute langTemp all temp/com
fix_modify langDyn temp langTemp
velocity all create 294.13 638479 mom yes dist uniform
This is a section of my LAMMPS input file. I was told that I need to use “zero yes” in the command line to remove COM velocity bias. That’s why I did it.
The problem is the dynamic of chain in simulation was considerablly slower than what previous literature described and what we expected. Although I was told that the dynamic of system is not so important as we were interested in the equilibrium structure of polymer chain, I was kind of worry about the fact that the COM was fixed for such single chain system would actually have an impact on our results. Honestly some of the simulated chain conformations behaved weird according to our knowledges. I wonder if there is any way to solve this problem or COM is always fixed for a Langevin Dynamic simulation?
there are a ton of other possible explanations for why you are not seeing
expected behavior. i doubt that you can blame it on the langevin thermostat
and what you describe seems like it is working correctly.
it is much more likely that you have some messed up
potential parameters or some other error in your input.
axel.
Also, check that you have enough chains. Even if you’re just looking at structure, a small box relative to the chain length will have strange conformations if the chains see themselves. And the dynamics will definately be effected. Typically the box edges are some number of Rg.
Tim
The "zero yes" option on the fix langevin command is an option.
If you use it then the total force applied each timestep will
be adjusted to be a net force of 0.0. Hence the COM will
not drift. If you don't use the option that the sum of a bunch
of random forces will not be exactly 0.0 and the system will
tend to drift (randomly, slowly). This is explained on the doc
page. Is neither of these what you want?
Steve