Hello,
While I test a new form of attractive interaction made of negative gaussian well with barrier, I found something weird. For the test, I set up initial configuration and simulation script very simple and run the simulation shortly.
As a result, I found that two particles are getting close to each other, and stay for a while, and suddenly they are getting far away from each other extremely rapidly.
For your information, I upload simulation script and initial configuration here. Simulation is deterministic, and particles are bounced away from ~70000 steps. Have anyone experienced similar thing before where particle overlap is allowed by attraction interaction?
I uploaded all files on google drive here: https://drive.google.com/drive/folders/16r1axuNrG65JTpkMs1K3qwRQ0PIWjTaJ?usp=drive_link
Please take a look and thanks in advance for any comments or knowledges of what is going on here.
Your file share is set to limited access. It cannot be downloaded.
This is likely a consequence of your choice of simulation and potential parameters. The behavior you describe would be consistent with a potential that is either not steep enough at short distance or a time step that is too large for the steepness of the potential shape and thus atoms can get closer than expected and suddenly experience a very strong repulsion.
Woops, sorry for not changing an access authority before. Now you can download them.
Thank you very much for your comments. Perhaps I did not check the tabulated potential more thoroughly. Just in case if you are interested in what happened, I also change the access authority accordingly. Please take a look.
I still cannot. Perhaps you need to post an updated link?
I’m sorry. Here is a link.
I figured out just by coding custom potential and rebuild lammps. Therefore, I guess your comment is on the exact spot. However, in case if you are available of looking at the files, it would be highly appreciated.
Please let me know if you can download the files.
I think the problem is that in your simulation the energy is not conserved, so as the run continues, the particles gain sufficient energy to jump out of the well. Indeed you can see it from the etotal column in thermo output.
This can be caused by a too large timestep, or a potential that is not smooth. For your specific input, this can be fixed by using spline interpolation instead of linear interpolation, i.e.
pair_style hybrid zero 2.0 table spline 6000
With spline interpolation you can probably decrease N further.
2 Likes
Spline works perfect, now the particles are not jumping for sufficiently long time.
Perhaps you might have seen, in case of “linear” tabular potential, nve integrator tries its best to conserve the total energy, but meantime kinetic energy keep increasing and potential energy keep decreasing. Therefore, right before the bounce, two particles show almost like 1D oscillator motion until rapidly increasing kinetic energy breaks the balance.
Since the kinetic energy seems to increase, I assume it should mean that the tabular potential work as an energy(force) source to the system just as you mentioned.
The only thing left behind is that why it only happens where the distance between two particles is nearly zero. I don’t think there are not many people who wants to completely overlap particles, but if the interpolation is the reason, then it also can happen at other distances as well.
Anyway, I will investigate this issue further. Thank you so much!