Recently I’ve been running energy minimization on different structures with different versions of LAMMPS. I have a crystalline structure containing a grain boundary. When I use an earlier version, for example, October 2020, to run energy minimization using conjugate gradient descent, everything works fine. However, when I try to use a later version, for example, October 2021, to run the exact same script, I get an error message saying that some atoms are lost during minimization. Are there anyone else that encountered the same problem?
There is no enough information here to make a proper assessment of the situation. In general, I trust more recent versions of LAMMPS more than older ones because of the systematic way that we now test for and eliminate bugs. The older version probably just does not crash by chance, not by design.
Hi Dr. Kohlmeyer,
Thank you for your prompt reply. Since my structure is very simple and periodic boundary conditions are applied, I can’t figure out why the simulation crashes in later versions. I attached my input file along with the interatomic potential that I used for reference (though the simulation also crashes when other potentials are used). If you could take a look, I would really appreciate it. Thank you!
ni1.set (83.8 KB)
cylinder-cg.in (828 Bytes)
There are no lost atoms, when I run this input deck with the latest LAMMPS version.
log.lammps.txt (6.0 KB)
Hi Dr. Kohlmeyer,
Thank you for checking.