Consider we have a bulk water structure and use the NVT ensemble to set the temperature at 298 K and relax it a bit. Then, we use NPT ensemble; so the system should get the desired density 1 g/cm^3. Finally, we use NVE ensemble to find the average total energy of the system. I attached files and outputs here for more details.
But the problems here are:
The output result for density is 0.6 g/cm^3 when we use the NPT ensemble.
TotEng results fluctuates too much and not about constant (At third section outputs when we call NVE ensemble)
Could you please write any comments why don’t the density and TotEng tend to their desired value?
This is not a question about LAMMPS but a question about the science of your research and thus a topic for discussion with your adviser or collaborators or colleagues. It is not the job of this forum to do your work. I am seeing an emerging pattern that whenever you encounter something unexpected, you just dump your input here and let others do your work for you instead. This is not how science work and what this forum is intended for.
To address your questions,
If you don’t get the expected results from your simulation parameters, then either your simulation settings are bad or your parameters. It is not the job of LAMMPS to validate your input; it just takes your input and executes it. The rest is you job.
Fluctuations are normal in finite size system simulations where you discretize time. You can learn about this from text books on MD and on statistical mechanics. Whether the fluctuations are within the reasonable range is subject to proper simulation settings and system setup and that - again - makes it not a LAMMPS problem, but a user problem.
It is not just the settings, it is the entire simulation protocol. The system is equilibrated for a very long time at the “wrong” density. And then there is not enough simulation time, so the NPT simulation is not at all equilibrated. There is an obvious trend of the system to shrink. The brevity of the NPT run is ignoring the fact that it is easy to expand a condensed system, but very difficult to compress it on its own, so because of the choice of protocol a very much longer run with fix NPT would be required.
Or better yet a smarter simulation/equilibration protocol that doesn’t waste a lot of time with equilbrating the system toward an undesired state.
Thanks for the correction. I missed some of these details.
My point was that a quick glance knowing how to set up TIP4P model in LAMMPS already tells that this input is wrong. The simulation protocol itself is bad but as I said (advised from the best here), this is a discussion to be taken outside the forum to proper teaching conditions.
It is not so much a correction but an extension. I don’t want people when they read this discussion later to miss the big issues over a minor issue, draw the wrong conclusions and waste their time on trying to solve a collection of problems by correcting the least significant one first.
