Restart molecules, shear forces are zero

Dear all,
I have been having an issue that my shear forces are zero when restarting my simulations for a molecule sample. This is affecting my stresses within the sample and also leading to a peak in coordination number which seems to recover quite quickly.
I wanted to ask if anyone has an idea how I can fix this issue. I read here that in the past (2012) people had issues with the shear history. However, since I have only been using LAMMPS for the past 3 years I never encountered this issue only using spheres.

Hi Tara,

Would you be able to provide an input script and restart file ?



Institute for Computational Physics
University of Stuttgart

Hi Simon,

Thank you for your reply.
I realised yesterday that the same issue also occurs for spheres. My shear force becomes zero when restarting. This only happened in the newer version of LAMMPS that I have recently been using. So I do not think it is related to my script.


Hi Tara,

Alright. Well, if you can provide a minimal example reproducing this issue, I am glad to have a look and see if there is an issue with the last LAMMPS version or not.


While that is likely true, you are missing an important point: without a minimal input example that reproduces the situation, chances are much lower that this will be looked into (almost zero), since you put the burden of figuring out and construct such an input on the person try to debug this, too, while you have already figured it out. Developers are usually very busy with many different tasks on their TODO list, and like other human beings, they favor tasks that require less effort.
I have some hunch which change to LAMMPS could have triggered the different behavior.

This kind of issue should best be reported as a “Bug report issue” on GitHub: Issues · lammps/lammps · GitHub

Dear LAMMPS representative,

I am very much aware of this. And I appreciate Simons help and really want to provide him with a script an restart file. However, I am unable to upload my script here as it states that new users can not upload attachments. Which why I initally posted it as a general question.


You can always upload files to Dropbox, Google Drive, Microsoft One drive or similar and add the link.

Apart from that GitHub will get you attention faster and more reliable as an open issue must be fixed or explicitly closed.

Thank you for this information. I did not think about this.

Dear all,
Please find attached two example scripts for this issue:

I provided also an example of the issue as you can see for the first timestep my stresses are zero in the shear_mean_stresses_iso.txt. In dump8000000 you can see that my shear foces which correspond to the 3 first entries are zero.

Please not that the following subdirectories should be created before running the script: atom connectivity contact stress velocity

Furthermore, I am using fix multistress which should be ignored.

Thank you very much.

This is undebuggable.

  • you only provide a restart file, and that for a rather old version of LAMMPS. There should be a (small) data file and then two small input with only the essential commands to create a restart and continue the calculation without restarting and then a second to continue from the created restart.
  • the systems are too big
  • you are using a non-canonical pair style (that is the main deal breaker)
  • your input contains lots of things that have nothing to do with the issue at hand.

You will not see any of the LAMMPS developers take a proper look unless you can reproduce this with the standard LAMMPS features. There is a very good chance that the culprit is your custom pair style and not anything else.