Group virtual fixed problems in nonequilibrium molecular dynamics(NEMD)

Dear all,
I want to try to simulate the heat transport process under nano-confinement. The model is shown in the figure below. Graphene nanotubes (grey) are blocked by 4 graphene slabs (gold). Now I want to apply heat source and heat sink (red part in the picture) on the left and right nanotubes respectively to realize heat transport. My difficulty is that I found that if the graphene slab is not fixed, the model can easily collapse (for example, carbon nanotubes become skewed or lose atoms, etc.). But the heat transport process cannot be achieved by fixing the graphene plate, because there is no temperature in this area.

The way I can think of is:
1.fix rigid handles graphene slabs, but during the actual operation, I found that this cannot guarantee the stability of the model (maybe graphene slabs and carbon nanotubes are no longer connected);
2. Use fix spring/self to fix a spring force, but I found that the temperature of the graphene plate is no longer determined by the heat source and heat sink alone.

None of the above worked very well for my problem! I’m at my wits end, any advice would be greatly appreciated!

There is too little tangible information (input file, data file, LAMMPS version, command line options) here to give any specific advice. If your system is set up correctly according your description with periodic boundaries and proper force field settings, the structure should be stable and no additional stabilization modifications required.

Using fix rigid is definitely not a good choice. But if additional position restraining would be needed, it can be applied to just a few atoms near the two ends of the CNT and/or in the graphene sheets. This could be done by simply excluding those atoms from time integration (so they are completely immobile) or using fix spring/self on them (so they are pulled back to their original position, if displaced too much). No need to restrain all atoms (that would modify your energy transfer too much).

Dear Axel,
Yes, I’m adding the complete information now, I’m using LAMMPS (22 Dec 2022). Since I’m a new user, I uploaded a connection (with inout and data)

There’s a lot of complexity in your model, since heat is passing through (1) the CNT (2) the CNT-graphene junction (3) the graphene-electrolyte (?) Interface (4) the electrolyte. Getting the interactions wrong at any of these could jeopardize the rest of your results, especially if you don’t separately know what the “correct” heat conductivities should be across each of these individually.

Additionally, the complexity of your system means you’ve went with a small box size, which makes the thermal features of your system even less realistic because you’re missing long-wavelength vibrations.

So it would help if you make a simpler working model of what you’re actually trying to test. It’s better to have many simple models of different parts of a nanosystem, rather than One Grand Unified Model, so the influences of each part are clear and easily analysed.

Dear stree,
Yes, I know the complexity of my model stems from my interest in the nature of heat transport in such complex scenarios. You are right, I would split to check the heat transport of each part. I still want to try this model when I feel confident enough. I now think that maybe a virtual wall can be used to fix the graphene plates and nanopores. In this way heat transport is no longer disturbed.

Case solved! Maybe a little fix recenter command solved this tricky problem!

I don’t think so. Fix recenter does not change the dynamics of your system, only the point of view.

Moreover, when looking at a replicated system, there are certainly issues with the periodic continuation of the graphene sheets and the connection between the CNTs and the sheets as shown by the visualization below (bonds are determined by a 1.5 angstrom cutoff between carbon atoms).

Dear Axel,
I hope it is best to use fix without changing the dynamics of the system, but keep the graphene sheet in the middle so that the CNT is not very tilted. Can’t fix center work at this time?
In addition, thank you for pointing out the loopholes in the model, I will continue to improve it~!
Thank you!

Hope is not a good foundation for research. If your model is set up properly, there should be no need to apply fix recenter. And if you read its documentation, it will clearly warn you about applying it to only a part of the system. In general, you should not change an input until its trajectory looks like what you expect, but rather each individual command should have a proper reason to be there motivated by the physics of the model. Otherwise you are not doing computer simulation, but computer animation.

Yes, you are right, I totally agree! After I optimized the model, everything changed, and I changed to a different model and no longer needed any fixes! Thanks for the helpful help! Thank for the wonderful lammps exchange community!