[lammps-users] Dislocation Loops... advice?

Hi Erin,

I wonder if you have some atoms missing at the boundaries of your cell
before you insert the extra plane of atoms.

I am using the same lattice rotation in a fcc lattice and I found that
depending on how I set the boundaries of the simulation cell, I get
different numbers of atoms created. This does not seem to be a problem with
non-rotated lattices. I have appended a script below to show what I mean.
The number of atoms should be the same in each case because the cell volume
does not change, and the cell is periodic in all dimensions, but the number
of atoms created in each case is:
corner_zero: 209
center_zero: 216
side_zero: 210
small_offset: 216

These missing atoms are really hard to see in projections, because they are
hidden by other atoms in each plane/row, but when you run the simulation,
those unexpected vacancies really mess things up! I think the problem is
related to atoms falling near the boundaries of the simulation cell.
LAMMPS developers, is this a bug?
Apologies if this is an unrelated issue...

Have you tried minimizing your cell before you insert the extra plane of
atoms to check that the perfect lattice is stable? If that doesn't look
right either, try hunting for missing atoms at the cell boundaries. Maybe
when your dislocation loop migrated through the cell in the simulation you
ran before you froze atoms, it moved one of those pesky vacancies to a
surface step where it wouldn't cause so much trouble.
Also, does your minimization converge correctly? Intermediate steps in a
minimization can look pretty strange so if your minimization is stopping
prematurely, that could explain your visualization.

regards,
Rebecca.

LAMMPS version: 7Sep09

in.missing_atoms:

These missing atoms are really hard to see in projections, because they are
hidden by other atoms in each plane/row, but when you run the simulation,
those unexpected vacancies really mess things up! I think the problem is
related to atoms falling near the boundaries of the simulation cell.
LAMMPS developers, is this a bug?

The generic problem here is matching a rotated lattice to a periodic
box. LAMMPS attempts to do the right thing at the periodic boundary
and not omit atoms or put in extra ones, but it may not always be correct.
So the user needs to insure the atom counts at the boundaries are what
is wanted. Slightly adjusting the box size can help, as round-off can be
an issue. The lattice doc page discusses this.

Steve