Running Lammps


I ran lammps for silicon with free surface (minimizing energy) once through lmp_win and once by lmp_mpi (the same script). But I got two completely different atomic stress profile!

I am wondering if there is a difference to obtain results through executable for windows?


there is not enough information here to make any specific statement. you are not even showing how different is “completely different”.

atomic simulations are high-dimensional problems and thus minimizations will find the next significant minimum, but rarely the global minimum. if your initial geometry is on a local maximum or saddle point in the potential hypersurface which is the “next” minimum can depend on the tiniest difference, which in turn can depend on choice of compiler, number of processes, time of day, or phase of the moon.

as a rule of the thumb, it is more likely that there is some misconception or misunderstanding or mistake or oversight on your side than an issue in the software, especially when using commonly used features. but it is impossible to tell from remote without any tangible information.