I am having problems using the wall/region harmonic fix, and closer
inspection reveals that the wall-particle interaction energies are
much larger than expected.
I may be misunderstanding how the wall potential is defined. Say I
have a spherical wall with an r_cutoff of 20 A. Based on the potential
function in the manual ...
E = \epsilon * (r - r_cutoff)^2, for r < r_cutoff
... I expected that the wall–particle energy when an atom reaches
within 20 A of the wall would be zero (since r - r_cutoff = 0), and
increase as the particle gets closer to the wall (due to decreasing r,
and therefore increasing (r - r_cutoff)^2).
However, what I find is that the wall–particle energy (as output by
f_wall-ID) when an atom is at the cutoff distance (i.e. just entering
the wall potential), is not zero, but actually equals ...
E_cutoff = \epsilon * r_cutoff^2
... causing the atom to bounce off violently (often blowing up the system).
Can anyone confirm this? Is this a bug, or — much more likely, I'm
sure — am I missing something?
Note: The E_cutoff values reported here are seen only in the
21-Dec-2010 Windows build of LAMMPS. The 15-Mar-2011 version on a
Linux machine returns "nan" for the energy, but a similar trajectory
to the Windows run. As expected after the 6-Feb-2011 bug fix, the wall
forces reported by the later version are different, but still high for
a particle just entering the wall potential — if this is a bug, I
suspect that the 6-Feb-2011 fix for the forces did not fix it.