It’s incomplete without that data file…
I am attaching the data file. It is reasonably large, with about 7000
particles. The data file contains information about particles that were
sheared with the same rate as of the previous input file without the fix
addforce for a reasonable amount of time.
data.restart (1.76 MB)
I’ll try it out tomorrow, but just to make a little test to prove what’s going on. Change the direction of the y-forcing on your “group down” particles (from negative to positive), see what happens to vx in the box. (should go negative - and you might even think why that might happen. I’ll give you a hint, it’s all about fluxes.)
Just thought I would give away what was happening, because I can’t think of a way to compute confirmation of what’s going in lammps.
You are forcing particles that are on average moving faster in the streaming direction into layers of particles that are moving slower. This will on average speed up the particles in the slower layers. Over time this would accelerate particles in the streaming (as well as shearing) direction. To confirm this compute the average velocity difference, (ie slip) between the forcing particles and it’s neighbors. With your forcing model the slip should be nonzero, but zero if there was no forcing.
In conclusion, lammps is doing what you are telling it to do, whether or not it’s doing what you mean it to do.