[lammps-users] atoms out-of-bounds with no error in triclinic box?

Answers below.


Dr. Plimpton,

Thank you the prompt and generous help.

If I set the neighbor table to update every 1 timestep, would that ensure that no atoms are outside bounds at any time?

yes, that should work - but that will likely be very inefficient
a better question would be, why do you care if the dump file has
every atom inside the box? If they are slightly outside, you can put
them back in.

If you'll pardon my reiteration: do the values reported in dump represent the maximum/minimum atom coordinates, or the maximum/minimum POTENTIAL atom coordinates?

I assume you mean the BOX BOUNDS, printed at the top of the dump file.
That is the physical box. You can set the physical box to be larger than
the actual atoms (e.g. free surface), though it's usually a bad idea to do so.
If you are using shrink-wrap boudaries (in some dim)
then they will be slightly larger than the physical atom positions. Although
again, atoms may have moved outside those bounds since the last reneighboring.

Is there a pure shear mode that does not result in an unintentional rigid body rotation?

If you are speaking of fix deform, then it shears the box. The atoms
inside the box
respond however they wish to. They are not constrained. I'm not sure what you
mean by keeping a rigid body from rotating.

If I have a tilt xz and wanted to do a manual unwrapping of period boundary updating, could I either adjust the x coordinate by (xhi-xlo) for that snapshot or (xhi-xlo-xz)? Or is this one of the shear directions that requires rotation as well?

If you are talking about dump file coords, then a periodic tilted box
is a periodic image of
the system. So you can create as many periodic images as you like of
an individual
atom. It's not as simple as subtracting xhi-xlo-xz or the like in any
dimension. You
have to apply the 6-element H matrix, in general terms.