Rattle affects pressure upon restart


Quick question. When using rattle, when restarting an NVT or NPT simulation of a 293K/1 atm liquid benzene system, I see the pressure is different before vs after the restart. The magnitude of the pressure-change-upon-restart varies between a few tenths of an atm up to ~10-15 atm (rattle tolerance of 1E-8). These changes are obviously smaller than the actual fluctuations in the system pressure during a run, which reaches 100s or 1000s of atms.

All other thermo quantities (kin energy, potential energy terms, temperature, cell size) are identical before and after the restart.

I know use of Shake will cause this change in pressure upon restart (due to half- vs full-time step velocities), but I’m not clear if Rattle will as well. Is this the case? It looks to be, unless I’m doing something wrong unintentionally

Also, I’m not seeing the pressure change during a restart when restarting an NVE run, for whatever reason. And the changes in pressure upon restart appear to be greater when restarting an NPT run than an NVT run.

It is very difficult to discuss such matters in the abstract. Best take one of the example or benchmark systems in the LAMMPS distribution and modify their input as needed to demonstrate your observation. People can then experiment and check out whether this is genuine or whether there are issues with your inputs and whether this are ways to address this or this is the expected/documented behavior.