Dear Axel and Steve,
I hope all is well.
I have used a LAMMPS code to calculate the elastic constants for different materials using different potentials: https://github.com/lammps/lammps/tree/5e485657611c6056e736989740744f81b6bedc03/examples/ELASTIC
However, when I use it to calculate the elastic constants for Mg2Si, sometimes I get “nan” for therdynamics outputs in the middle of the calculation, or sometimes I get very high elastic constants, by repeating the simulation several times:
Mg2Si.lmp (545 Bytes)
init.mod (1.13 KB)
in.elastic (6.54 KB)
Jelinek_2012_meam.alsimgcufe (2.25 KB)
displace.mod (3.83 KB)
Jelinek_2012_meamf (1.29 KB)
potential.mod (546 Bytes)
restart.equil (1.77 KB)
log.Mg2Si (110 KB)
nobody will debug your input for you, and none of the developers will take a serious look until you have evidence, that there is a problem with LAMMPS itself (and not your input system or parameters) and can prove convincingly that this issue exists with the latest LAMMPS version (as of this writing it is 6 August 2019).
axel.
Dear Axel,
Sorry about sending the reply to a wrong email address. I repeated the possible reason for “nan” problem below.
I asked the developer of the potential about this issue, and that guy mentioned that this might be because of changes made to meam style in lammps, i.e., from meam to meam/c. I tried an older version of lammps (2016), and I didn’t see the symptom. Please let me know if this is the lammps problem, or the potential should be redesigned by the developer compatible with the new versions of lammps.
Bests,
Reza.
Dear Axel,
Sorry about sending the reply to a wrong email address. I repeated the possible reason for “nan” problem below.
I asked the developer of the potential about this issue, and that guy mentioned that this might be because of changes made to meam style in lammps, i.e., from meam to meam/c. I tried an older version of lammps (2016), and I didn’t see the
meam and meam/c have both been tested thoroughly to yield identical results for a variety of inputs, IF used with an up-to-date LAMMPS version. that is why we removed the fortran version after the december 2018 stable release.
symptom. Please let me know if this is the lammps problem, or the potential should be redesigned by the developer compatible with the new versions of lammps.
again, it is your job to prove that LAMMPS may be at fault. i doubt that there is something at fault with the published parameters. please note, that older versions of MEAM had bugs, too, that may be hiding errors or omissions in your input.
in summary, i see no convincing evidence that would exclude a user mistake. your provided logs are from using an old version of LAMMPS so they have no value. BTW, we just today released a new stable version of LAMMPS. you have to provide convincing evidence, that this version is giving incorrect results versus the MEAM implementation in version 12 December 2018 with detailed description and logs on how to reproduce it, otherwise we will assume, that this is a user error as with the vast majority of this kind of report here where people insist, that they have done everything correctly (only most of the time, there is still some mistake).
what you have here is vague and lacking lots of detail.
axel.