# [lammps-users] potential equation

Dear All,

I had one potential question and could not figure out a way that LAMMPS can solve it. Hope someone can help me out this problem.

I have a unit cell that has periodic boundary conditions in all dimensions. This unit cell contains two groups of atoms(say group1 and group2). Each group contains more than one atom and has interaction with the other group (so group1 has interaction with group2 but no potential exists within one group).

By this definition, I define the potential like
pair_style xxxx
pair_coeff 1 2 parameters

However, if I define potential in this way, my group1 won’t have interaction with group 1 in the neighbor unit (My intention is each group has interaction with other groups, so if group1 in neighbor is close enough it should have effect on group1 in base unit). Any suggestion is greatly appreciated.

Regards,
ML

“However, if I define potential in this way, my group1 won’t have interaction with group 1 in the neighbor unit (My intention is each group has interaction with other groups, so if group1 in neighbor is close enough it should have effect on group1 in base unit). Any suggestion is greatly appreciated.”

This is not clear at all. Please explain in more detail

Dear All,

I had one potential question and could not figure out a way that LAMMPS can
solve it. Hope someone can help me out this problem.

I have a unit cell that has periodic boundary conditions in all dimensions.
This unit cell contains two groups of atoms(say group1 and group2). Each
group contains more than one atom and has interaction with the other group
(so group1 has interaction with group2 but no potential exists within one
group).

this sounds like a _very_ strange model. it would be
much easier to provide any advice, if you could provide
some more tangible motivation for why you would
want to do something like that.

By this definition, I define the potential like
pair_style xxxx
pair_coeff 1 2 parameters

However, if I define potential in this way, my group1 won't have interaction
with group 1 in the neighbor unit (My intention is each group has
interaction with other groups, so if group1 in neighbor is close enough it
should have effect on group1 in base unit). Any suggestion is greatly
appreciated.

do i understand it right, you want to have type 1 atoms only interact
with its periodic images, but not with itself?

i am not certain how this would work with any MD code.
i also cannot imagine a scenario where this would be
a good model. you may perhaps get it figured out by
modeling the periodic images as explicit copies and
then giving them a new atom type each and then define
the pairs of interactions explicitly.

cheers,
axel.

Thank you all for reply. I can detail my model here.

There are two CNTs in the base cell, each of which has LJ potential with the other but not within itself (I have Tersoff potential within CNT). This model has periodic boundary conditions. Axel is correct, I would like to have LJ potential between two CNTs if they are close enough so type1 atoms might have LJ potential with type1 in neighbor cell.

This model is from some reference so I think it is reasonable.

Regards,
ML

Thank you all for reply. I can detail my model here.

There are two CNTs in the base cell, each of which has LJ potential
with the other but not within itself (I have Tersoff potential within
CNT). This model has periodic boundary conditions. Axel is correct, I
would like to have LJ potential between two CNTs if they are close
enough so type1 atoms might have LJ potential with type1 in neighbor
cell.

it has been argued several times on this list (and a few times
very recently, you should really have a look at the archives),
that coupling objects that are treated via a many-body potential
with a pair-wise additive potential is introducing inconsistencies
into your model.

the nature of a many-body potential requires that the interaction
between atoms in each CNT is affected by the proximity of atoms
in another CNT. with the model you describe, this is ignored.
if you don't care about those many-body contributions, you could
represent your CNT as easily by a simple LJ pairwise potential
and a (harmonic) bond potential. that would make the calculation
much faster, too.

This model is from some reference so I think it is reasonable.

a lot of stuff gets published, not all of it stands the test of times.
it always depends on the justification of for the given application,
whether a model is considered adequate.

at any rate, it will be very difficult to implement what you are
proposing and considering the questionable benefits of implementing
it, it would recommend looking for an alternative.

if you would explain what you actually what to achieve, and
particularly what you intend to gain by replacing the tersoff
potential with LJ for the inter-CNT interactios, perhaps
somebody can make a suggestion here.

cheers,
axel.

If group 1 = atoms of type 1, and ditto for group 2.
Then you can define a pair style like lj/cut and set the
coeffs for the 1-2 interactions. You can use neigh_modify exclude
to turn off 1-1 and 2-2 interactions. Thus only 1-2 interactions
will be computed. Not clear if that is what you want but it
appears to be.

Steve