read_data velocity rapid increase


I am simulating diffusion at a given temperature between two binary metallic liquids of different composition. I need to begin the simulation at 2700k with a perfect composition profile, i.e liquid one side of the interface has 0% solute, and liquid on the other side has a uniform distribution of solute.

I have equilibrated both sides of the interface to 2700k separately (to stop any diffusion below this temperature), thus finding equilibrium volume at this temp. I then manually combined the xyz co-ordinates and vx vy vz velocities, into a data file.

Reading this data file, using an nvt, npt or nve ensemble gives initial conditions (timestep 0) that are just what I am looking for. T=2700k, with two liquid structures either side of an interface. But after one timestep (0.001ps) the temperature raises to ~4,000K and before 0.1 ps it is around 60,000K! The atoms that have the highest velocities that bring up the temperature are equally distributed through the sample, so there I believe that there is no problem with the interface or boundaries (e.g atoms overlapping). But just to check I ran the simulation with gaps at the interfaces, which yielded very similar results.

As there is no velocity input whatsoever (in nve, npt and nvt have a fix at 2700) I have no idea how this huge temperature is being generated. Any help as to why this is happening, or advice on a method of separating atoms, equilibrating them to 2700k and then removing the separation to allow diffusion, would be greatly appreciated.


Jack Shuttleworth

I don’t see how what you describe could happen if there is indeed
no overlap. You don’t say what you are using for potential(s).
Presumably the same potentials work when the systems are
separate? If you are using a new hybrid/alloy potential
for the first time for the alloy system, then maybe that is the

You can remove the velocities from the combined data file,
or just use the velocity command after reading it to
reset the velocities of the combined system. That would
eliminate any chance that something about the initial

combined velocities is causing a problem.