Re: [lammps-users] MD simulations in hexagon prism unit cell
Andrew Jewett <jewett@...1937...>
Tue, 19 Sep 2017 12:01:40 -0700
Fair enough. Now I see why you are confused. A "hexagonal" unit cell
is not a hexagon.
Take a look at this web link:
According to popular convention at least, a "hexagonal" unit cell is a
specific kind of triclinic (rhombohedral) unit cell with 60 and 120
degree angles in the XY plane. You can use this kind of
diamond-shaped unit cell to build a periodic system with hexagonal
symmetry. For this reason, nobody bothers to write software which
supports boundary conditions which are actually shaped like hexagons,
but I did not get the impression you are asking for this feature. It
sounded like you just want to find a way to run your simulations of
graphene/graphite. Just try using rectangular boundaries, and make
your box large enough that it encloses an integer number of
diamond-shaped unit cells. Here's that picture from the example I
mentioned earlier again:
As this discussion is no longer specific to LAMMPS, perhaps we
should continue in private.
(Incidentally, I recall some simulation software programs support
"truncated octahedron" shaped periodic boundaries, but they do it for
a very specific reason, and I always thought it was not worth the
On Tue, Sep 19, 2017 at 6:20 AM, Axel Kohlmeyer <akohlmey@...24...> wrote:
> please note that you are about to set yourself up for public
> embarrassment and ridicule.
> a) the kind of changes you would need to do to LAMMPS to support a
> different kind of cell representation are **substantial**.
> it is not just a couple of subroutines. it is all over the place.
> check out the changes required for a triclinic cell. i counted over
> 1000 places where this is tested for and where special subroutines are
> implemented. ...and a lot of the triclinic code can utilize the
> orthogonal cell code by switching between fractional coordinate
> representation (which is orthogonal) and conventional coordinates.
> specifically the parallel communication for setting up the domain
> decomposition and communication of ghost atom information (forward and
> reverse) and the border communication for particle exchange with
> neighboring cells is a major effort. major effort meaning of the order
> of years of your time spent programming and debugging.
> b) on top of that, what you claim you need to do is _not at all_ needed.
> you can **easily** represent your hexagonal system with a triclinic
> cell that is tilted in the xy plane, and if you build a supercell from
> two primitive triclinic cells, you can even use an orthogonal cell. i
> recommend the latter, as it makes many things easier.
> since this is essentially a 2d problem, i suggest you pick up a piece
> of quad-ruled paper, draw it out, and you'll see. draw four hexagons
> (one over two over one) and connect the centers, you will get a
> rhombus, that covers the same area as one hexagon. add more hexagons
> and more rhombi you will see clearly that you cover the same volume
> with hexagons and rhombi and that both are a space-filling lattice.
> or pick up a book about crystallography, read about bravais lattices,
> and you'll see.
> or google it for a bit, and you'll see.
> On Tue, Sep 19, 2017 at 12:08 AM, surendra jain <jainsk.iitkgp@...24...> wrote:
>> Dear Andrew,
>> I am using hexagon prism unit cell. The boundaries of the simulation
>> cell is below :
>> 2 1
>> 3 4
>> Thus 1 to 6 marks the corners of my simulation cell.. The PBC in this
>> system is different than orthogonal or hexagonal system. I have to
>> change the neighbors calculation routine and PBC routine.
>> Best Regards,
> Dr. Axel Kohlmeyer akohlmey@...24... http://goo.gl/1wk0
> College of Science & Technology, Temple University, Philadelphia PA, USA
> International Centre for Theoretical Physics, Trieste. Italy.