diiferent Lammps version gives different for force of the simple problem of two charged rods

Hi All,

I checked the results of a code using two different version of Lammps. I have a system of two rods and I increase the separation of two charged rods. The force on each rod has to follow 1/r in which r is the separation between two rods. Lammps May 31, 2013 does not give the 1/r behavior while Lammps Oct 30 2013 shows the behavior of 1/r.

I attached the input file and the plot of the results. It seems that new lammp version does not give the correct results at least in short rod separation.



ts.txt (4.36 KB)


in2.simple (2.79 KB)

Your plot says the old version is Oct 2010, not Oct 2013.

If you post the 2 log files from the 2 runs, perhaps
Stan can look at the differences, since this is
a Kspace calculation. Do the step 0
energies and pressures (forces agree)?


This has to do with changes to the kspace error estimator. If you use a kspace accuracy of 1e-9 instead of 1e-4, then the current version will give a similar answer as the Oct 30, 2010 version. With such a small system of 40 atoms, the error estimator was only giving you 1 kspace vector… Paul has previously checked the error estimator for Ewald and confirmed it is working correctly and does give the requested error.