[lammps-users] Temperature becoming zero

Dear all
I am a beginner in lammps. I have been trying to do uniaxial tension test in a vertically stacked heterostructure of graphene-WS2-MoS2-graphene using KC potential for their interlayer actions (REBO II for graphene), along with SW potential for WS2, MoS2 and REBO II for graphene for intra-layer interactions. First I tried it for a small structure, size of 60×50 Angstrom, and 30 A in z direction, with r cut off value of 24 A. But the problem is, after the WS2 and MoS2 layers break, the temperature becomes zero.I used fix spring command between WS2 and MoS2 layers and it worked. But when I applied it to a bigger structure, 130×100 A, with the same z and r cut values, it’s showing the same problem. All the atoms freeze and temperature becomes zero. I’ve attached the input file, KC potential file and log file.
If anyone can please help me to identify what’s going wrong and how to solve them, I’ll be very grateful.

in.txt (42.8 KB)

ILP.KC (1.75 KB)

log.lammps (209 KB)

Dear all
I am a beginner in lammps.

it is not recommended for a beginner to start with simulations of such a complex system. it is near impossible to say whether any unexpected behavior is due to a mistake you made in the input or setup or whether it is genuine or a combination of both.

it also looks like your setup is having far too many hybrid substyles. it is my understanding that you need to apply the inter-layer potential for all atoms in a layer simultaneously instead of creating an instance of the pair style for all permutation of pairs.

again, my recommendation is for first set up a uniaxial tension simulation for a system with single atom type and compare your results against published results to make certain that part of the input is correct. then repeat this with just a simple two-layer system with the same kind of atoms and then gradually make it more complex while at each step validating/confirming that the results are meaningful.

one can only speculate about what is causing the temperature drop, but the fact that you have told LAMMPS to ignore lost atoms is not a good idea. you should not lose atoms for such a simulation. if this happens, most likely some simulation parameters are not chosen well. but that is impossible to tell for certain from just looking at such a convoluted input.
best you take a step back and start over and build your expertise and experience in steps as I have outlined above.


Dear Axel

Thanks a lot for your advice.

I have done uniaxial tension tests for the monolayers of MoS2, WS2 and graphene; vertically stacked graphene-MoS2-graphene (validation of a published paper), graphene-WS2-graphene using the same input setup and they worked without any problem. But when I tried it for the bilayer of WS2 and MoS2, it showed the same problem mentioned before. I tried the bilayer simulation with “fix 1 all spring/self 10 z” instead of the fix spring command I used before and it didn’t show any problem and no atom was lost, the obtained Young’s modulus was also very close to the value mentioned in a published paper. And this command worked fine with multiple combinations of r cut off values and z length of simulation box values. I am mentioning this because the fix spring command also worked for the bilayer with r cut 14 and z 20 A, but showed problems when these values were increased, like r cut 24 and z 30 A.
I observed that the problem arose in these cases from when atom or atom chunks got free from their respective layer after failure and moved arbitrarily in the simulation box. Since in the case of using fix spring/self command, no such thing happened due to the constraint in z direction, no problem arose either.
I have attached the input file for the bilayer simulation with the fix spring/self command.
Is it alright to use this fix in my case study? Please help me.

in.txt (31.6 KB)

the research specific details of this are outside my area of expertise, but in general adding forces to a system with fix spring fixes has the potential to change the physics, so I would doubt that either spring fix is something that you want to use. your description sounds very much like it is a duct tape solution to suppress a correct but inconvenient behavior.
that is assuming that the model is built correctly and the force field choices are adequate. I already mentioned my doubts about the use of the interlayer potential with excessive instances of hybrid sub-styles. for certain, it is irrelevant whether I agree or disagree with your choices on a preference basis as it will be your name on the results and you have to justify your choices. there is no additional help I can provide.


Thanks a lot for your help.