Dear lammps developers,
I have encountered an unusual error when restarting a simulation using read_restart. The first restart step matches exactly the last step from the previous simulation. However, the system temperature blows up right at the second step using the same integrator (nve verlet), thermostat (langevin) and a fix with the same ID (spring/self). At the end of the first simulation, I wrote both restart and data files. Restarting the simulation with the data file works well, and so restarting the simulation with a converted binary restart file via -r switch. That leaves us with three possibilities. There is:
1- a bug in the read_restart routine or
2- a bug in the restart routine under some very specific circumstance.
3- a very silly mistake that I am making
Problems with the configuration of the system are out of the question since I was able to continue the simulation normally. The fix spring/self is not critical to prevent the temperature blow up as I had no problems in redefining the positions of the tethering springs (although it is desirable to keep the original positions).
My simulation runs on a GPU plus one CPU core with option neigh no. It has molecules, bonds and charged atoms. The pair_style is born/coul/wolf and the bond_style is table. I created the restart files using the 30-Apr-15 version and tried to restart it with the same version and with the latest version. Based on my past simulations, this issue is recent because I was restarting simulations with no problems using the 6-Jan-15 version.
I intend to send a small input script that reproduces the error but I think the information above should help you for now.
Any thoughts on this?
Best regards,
Luis Goncalves