periodic BC in Lammps and triclinic box

Hey dear all,
I’ve tried to simulate a thermal shock through a triclinic geometry supercell.
Basically, I want to constantly heat rapidly one region and let the heat flow to the rest of the material (slab of 20x2x2 unit cells).
Obviously, I need to keep the X-direction fixed (i.e not periodic, as opposed to Y, Z directions) - but here comes the problem
it says “Triclinic box must be periodic in skewed dimensions”.
Also, i want to put a repulsive wall on X - direction surrounding the material, so the X direction should NOT be periodic (fix wall demand).

Anyone encountered such a problem ?
I’d be glad to hear your suggestions !

I'm not clear on what you are asking with regard to the shock, but
this issue:
"Triclinic box must be periodic in skewed dimensions"
means the following:

A skewed system means not only that the lattice of atoms
is tilted (e.g. when you look at a snapsshot), but that the
periodicity across the boundary is shifted by a delta in
the transverse direction (the tilt factor that the LAMMPS doc
discusses). That shift has no meaning if the system is
not periodic in that direction.

If you just want a lattice of tilted atoms with no periodicity,
you can use free surfaces. But then you don't need a triclinic
box (in that dimension). You can still have an oriented (tilted)
lattice if you want via the lattice orient command.


Hey Steve,
What i'm trying to do is to follow the protocol of this paper:
But, my material is monoclinic, which means I have to set the X direction
periodic too! (which is bad, because it's the direction I want to follow
during heat loss - it is what differs my simulation from the paper btw)
Even without that difference, Say, I wanted to do exactly the same thing as
in that paper, I couldn't have done it because of monoclinic geometry of my
crystal and it's lammps restriction to have periodic BC on every tilted face

I hope my problem is clear now,
Thanks alot

Sorry - I don't have to time to read the paper. Please
explain why you can't create a non-periodic system
without a triclinic box (in the non-perioidic dim).
If it's non-periodic with free surfaces, you can put
the atoms whereever you want.


Please keep the list in the loop.

If I understand, here is a 2d description of your model.
You want non-periodic in x (free surfaces), and periodic in y. You want
the crystallographic orientation of the atom lattice to be tilted,
so that (e.g) the 100 direction in the crystal is at some angle
with respect to both the x and y axes.

In that case, you don't need a triclinic box, you can use a rectangular
box that is periodic in y, but not in x. Simply fill it with atoms
using the lattice orient command to get the tilt you want. And make
sure the box length in y is commensurate with the periodicity.

That will give you an infinite system (in y) that has the orientation
you want, will it not?


I can't use the lattice orient command, because my material structure comes
from an .xyz file which I convert into
a proper data file for read_data command.

Can I do the same "trick" via lattice orient to data that comes from xyz or
cif filetypes ?


2011/8/24 David Furman <[email protected]>:

I can't use the lattice orient command, because my material structure comes
from an .xyz file which I convert into
a proper data file for read_data command.

you could pre-process the orientation into the .xyz
coordinates. e.g. with some vmd scripting. where you
can apply rotation/translation/scaling to coordinates,
define a box and then pbc wrap the coordinates into the box
and write out a data file based on that info.


When you read a data file, LAMMPS will remap atoms that
are out of the box, back into the periodic box. So if you xyz
coords (put into the data file) are for a tilted box, but you
use an orthogonal box in the data file (like I suggested in
the earlier email, for periodicity in y), and you choose
the y box length appropriately, it will just work. I.e. the
tilted atom coords out of the orthog box will be remapped
into the orthog box at a correct image-atom location.


Thanks for your suggestion !
Where can I read more about this option ?
What would be the correct Y box length to choose ? How do I obtain it ?

You said the system should be periodic in y, so it
is whatever that periodic length is, for your original xyz data.
The periodic length is the same, whether it is an orthogonal
box, or a triclinic box (with an xy tilt).