Dear all, my question is related is about using restart files. I have done a series of ab-initio NVT md-simulations (a standard inputfile is added (thf.inp)) of 64 molecules of tetrahydrofurane. Every 'frame' consists of 350 steps of 1 fs (the total number of steps for now is 24 times 350 steps). I have taken a look at the total energy and the conserved quantity of every frame and did the same exercise when combining all the output of every frame into one big output file. I have added the plots of 3 frames and called them 'basic_analysis_1' for frame 1, ... and 'basic_analysis_tot' when all output is combined. I'm a bit baffled by the result of this latter one: the conserved quantity for the separate frames is conserved but when they are combined there appears to be some kind of drift (although this might not be the exact term for this phenomenon). Has anyone experienced this before? If yes, does this has severe consequences for the quality of the simulation? I have also done a NVE-simulation for a random frame (with a bit longer simulation time however; see basic_analysis_NVE) and there seems to nothing wrong there, so I think the problem arises because of the use of restart files, but however I'm not sure about this. I have also added the restart file after the first 350 steps as an example (thf-1.restart). thanks in advance cheers & beers marc