Query about seamless resumption

Dear Lammps users,

I am seeing that a history-dependent fix command such as ‘fix ave/time’ could give something unreal value when we restart using a binary restart file. To say it more specific, I calculated msd and wrote a file using the ‘fix ave/time’. At a certain time passed (10000000 timestep), I obtained a restart file, and then resumed the simulation with it. In this case, I obtained following unrealistically huge numbers (O(12)) in the very first line of the msd-written file:

Time-averaged data for fix 4

TimeStep c_msd[1] c_msd[2] c_msd[3] c_msd[4]

10001000 5.35876e+12 100.233 89.4598 5.35876e+12

10002000 5.35876e+12 100.164 89.4114 5.35876e+12

10003000 5.35876e+12 100.207 89.5413 5.35876e+12

So, there is gap in msd value between the first and next resumed simulation. Do you have any suggestion to make it ‘seamless’?

Thanks in advance!


From the fix ave/time doc page:

No information about this fix is written to binary restart files.

So you cannot do a "seamless" restart with respect to
values this fix writes to its output file. If the fix is averaging
over the last 10,000 timesteps, there is no info about those
9999 previous values in the restart file when the 2nd run