I am working with InN using Stillinger-Weber (SW) potential. First, to check the parameters I calculated the elastic and lattice constants of bulk InN and got very good results. Then I generate the parameters for two-body and three-body interactions with total of 8 possible combinations. To generate these parameters I followed similar procedure as done in LAMMPS for GaN. and also tried two more possible way of generating using antiemetic and geometric means for the parameters “Lambda” and “eps”.

The problem comes when using NVT ensemble , "fix 1 all nvt temp 300 300 1 " . I always get deformed structures after 50000 and 100000 iterations. ! I checked the initial structure is ok and also tried to change the time step, the Tdamp parameter and the velocity distributions but its all the same !

Any idea or suggestions to solve this problem would be of great help .

Aidan can comment on SW for alloys, but in general this is not
a LAMMPS problem, but a force-field development/debugging problem,
which is a whole field unto itself.

Also, NVT should have nothing to do with it. NVE should show the
same problems if you have a well-equilibrated system.

As Steve says, this is a problem with your potential, not with LAMMPS.

This behavior is pretty common when generating potentials for crystalline materials. It is relatively easy to construct a potential with a local minimum in energy for the equilibrium bulk crystal structure, with correct lattice and elastic constants. However, this may not be the global minimum. Moreover, lower energy structures may exist separated by small energy barriers that are quickly surmounted during MD simulations at modest temperatures.

The most direct strategy for fixing this is to lower the bulk InN structure relative to the deformed structure. Of course, this must be done in a way that does not degrade the other desired properties.