Question: compute stress/atom

Dear LAMMPS-Users:

I could not find any posts on this topic archived. I understand from the LAMMPS user guide that the 6 stress components are in pressure*volume units.

My question is is the “volume” in the formula “supercell volume” or "atomic volume” ?

Thank you

Srini Srivilliputhur
University of North Texas

1 Like

It is precisely because of this confusion that the per/atom stress components are not normalized by volume and hence are in pressure*volume units. If you choose to divide by volume yourself, then you should use the atomic volume, but this is not an easy thing to obtain, in general.

Thanks Aidan for clarifying. I agree with you on the difficulty of getting atomic volume. I’m stressing a 1-layer material and do not know what to do now!

Srini Srivilliputhur

You can look at compute voronoi/atom which estimates

a local per-atom volume.

Steve

You can look at compute voronoi/atom which estimates
a local per-atom volume.

​in bulk areas. for atoms on surfaces, the estimated volume depends on the
size of ​simulation cell or the distance to the next object.

axel.

1 Like

And here is another angle:

Resolving stress on the level of single atoms is usually not a good idea, unless you are at T=0, because the thermal fluctuations are huge. At T>0, it is a good idea to group the atoms into larger volume elements (rectangular bins), summing the per/atom stress for each bin and dividing by the bin volume. You can do this in a post-processing step, or on-the-fly using the handy LAMMPS chunk/atom bin command.