Dear Axel,
Thanks a lot for your consideration and reply. As a brief reminder of my problem: I have a system consisted of two groups of atoms: an admolecule and a substrate. I run my simulation using fix nve while I ONLY froze the rotational motion of the admolecule using “fix momentum 1 angular” command. My time step was 1 fs. Although people might expect there must be some “energy leakage” in my simulations, but exact monitoring of energies indicates that my total energy (as well as PE and KE) is conserved during long time run (10,000,000 and beyond). So, my question was: why the energy is conserved in my simulation (which of course is a good event for me)?!
You replied me:
“there are always three explanations for things like this:”
So, let’s analyze your answer from bottom to top:
“3) your system doesn’t behave the way you expect or the contributions that you expect to be significant are not.”
No! I have precisely checked the movies, and I found that the system behaves as I expected: The molecule has a translational diffusive motion on the substrate without any rotation. (For your information, without applying fix momentum angular command, the natural diffusive motion of my admolecule also consisted of rotational motion around its center of mass).
- what you consider conserved energy is not that well conserved after all,
I have precisely monitored the thermodynamics outputs in my log files. The total PE, KE and total energy of my system remains constant with an acceptable fluctuation around the equilibrium average. (Though I will back to this fluctuations in the next section).
1)you do have an error and it is being compensated by a second error that goes
into the other direction,
Maybe! So, please see whether my explanation is reasonable for you:
I have a system consisted of about 1000 atoms. Using fix momentum 1 angular, I slightly manipulated the velocities of about 50 atoms in my “admolecule” group at every time step to freeze its rotational motion. Since I eliminate 3 dimensional rotational motion of the admolecue, I “kill” 3 degrees of freedom in my system. Consequently, based on the equipartition theorem, the “turbulence” I impose to my system by adjusting the velocities of admolecule is in the order of “31/2kBT”. On the other hand, in my simulations, I observed that the standard deviation of the total kinetic energy (and potential energy) of my system, which is due to the existence of energy fluctuations raised form Verlet integration technique, is more than one order of magnitude higher than 31/2kBT. So, since the imposed turbulence in the kinetic energy (according to applying fix momentum angular) is much smaller than the numerical uncertainties of the energy calculations in fix nve, no accumulative leakage is observable during my simulations. (It is also noteworthy that in practice, the actual turbulence due to “fix momentum angular” can be much smaller than 31/2kBT, because the rotational motion of the admolecule may not have enough time to be thermalized).
So, is my description above reasonable/convincing for you to describe “why there is no energy leakage” in my simulations?
Thanks a lot for your all your passionate considerations and help,
Mehdi