Realize bonded interactions Softening or Failure when bonds are shortened

Science Talk
,
1

Hello all,

I am executing a compression simulation on a multi-layered graphene-shaped model under a customized coarse-grained (CG) force field.

  • The CG force field is simplified with almost bonded-interactions, such as harmonic bonds and angles.
  • The compression simulation will lead to the in-plane deformation, shorten bond length and buckling.
    image
    image1174×414 89.9 KB

1. My goal and Problems facing to

I would like to fit the all-atom level compression strain-stress curve, which the strength will sharply reduce due to the compression failure.

However, I found the CG force field can’t reproduce all-atom level softening or failure behaviour under compression. A schematic of the performance is shown below:

  • All-atom level model strength plunges after ultimate strain.
  • Stress of CG model continues to increase, and no failure happens due to no bond cutoff or failure mechanisms for harmonic bonded interactions during compression.

image
image906×706 11.5 KB

2. Questions

I recognize the original harmonic bonded style is too simple to reproduce all-atom level performance, especially the failure behaviour. However, I can’t easily change the present CG framework. Based on this, I would like to ask:

  • If there is any recommended command to change the bond behaviour by setting a threshold value. For example, when the bond length is shorter than the value, the bond would be deleted or softened, even though it may not follow real physics?
  • (A similar command is bond/break, which can delete bonds going beyond a length. But it doesn’t work for shortened bonds)

3. About Attachment

I can provide more details if necessary.

Thanks for your help!

2

@YetionMars This is not really a question about LAMMPS but about your research. Thus I have moved it to the Science Talk category where such topics belong. Although, the general experience is that these kinds of “conceptual” inquiries do not gather many responses, if any when posted in a public forum, as it would require somebody that has run into the same problem before or is willing to spend significant time to investigate the issue. I thus suggest to reach out for a potential collaborator that has the expertise you are looking for. Given that people use machine learning for all kinds of science problems these days, I would also suggest you research if there is any precedent along those ways for setting up a coarse grain model like you want to.

3

Thank you for your pointing out and help.