pair style hybrid

Dear all

I want to simulate a system containing graphene and carbon nanotube, which both are defined with tersoff potential; but the interaction between these two is just LJ and not tersoff. I have used something like below and I am not sure whether that is what I mean or not.

pair_style hybrid tersoff lj/cut 10
pair_coeff * * tersoff SiC.tersoff C C
pair_coeff 1 2 lj/cut 0.00248 3.4

please guide me.
thanks.

Azizi.

Dear all

I want to simulate a system containing graphene and carbon nanotube, which
both are defined with tersoff potential; but the interaction between these
two is just LJ and not tersoff. I have used something like below and I am
not sure whether that is what I mean or not.

pair_style hybrid tersoff lj/cut 10
pair_coeff * * tersoff SiC.tersoff C C
pair_coeff 1 2 lj/cut 0.00248 3.4

please guide me.

finding out whether a model works in principle is simple:

build some example system that has so few atoms,
that you can compute the energies (and forces) by
yourself and compare the result with what lammps
computes. i am not certain, but i would suspect that
with your setup, you cancel the two body interactions,
but not the manybody terms between the two parts
that are represented by the tersoff potential. i would
expect that you have to specific tersoff twice to make
sure this doesn't happen, but only an experiment
like i described above can prove it.

axel.

Hi Azizi,

I agree with Axel that the manybody terms between the two C that are
represented by the tersoff potential may not be turned off with your
setup. The second pair_coeff only adds additional LJ interaction
between the two C, instead of turning off the tersoff between them.

To properly turn off tersoff interaction between two C, you need to
make some changes to the first pair_coeff command and the SiC.tersoff
potential file. First change two C into C1 and C2 in the first
pair_coeff command, then duplicate the "C C C" entry in the potential
file into "C1 C1 C1" and "C2 C2 C2". Most importantly, set
coefficients A and B in the potential file for C1-C2 mixed interaction
to zero.

Best,
Ray

Ray is correct, except that using by hybrid instead of hybrid/overlay, the
LJ pair coeff does not "add" LJ interactions, it tries to replace the
Tersoff 1-2 interaction with LJ. This cannot work b/c Tersoff is
a many-body potential.

The pair hybrid doc page discusses this in great detail and gives
you several work-around options. See the IMPORTANT note.

Steve

Hello All,

I am doing similar simulations. Let us say we have 2 nanotubes around 3 A apart. Now suppose i declare
all the atom types under REBO, ‘pair_coeff * * rebo’ and replace the interactions between the two tubes as
LJ with ‘pair hybrid’. Will this give me wrong results? am i interrupting the manybody potential?

The way i see it, the two tubes never come closer than 1.7A which is the cutoff in REBO so i need not
worry about it. Will it make any difference in the computing time?
If what I am doing does the script below make any sense?

pair_style hybrid lj/cut 15 rebo
pair_coeff * * rebo CH.airebo C C
pair_coeff 001 002 none
pair_coeff 001 002 lj/cut constA constB

Thanks,
Freddie.

Hello All,

I am doing similar simulations. Let us say we have 2 nanotubes around 3 A
apart. Now suppose i declare
all the atom types under REBO, 'pair_coeff * * rebo' and replace the
interactions between the two tubes as
LJ with 'pair hybrid'. Will this give me wrong results? am i interrupting
the manybody potential?

rather than guessing and depending on the say-so
of random people on the internet, i strongly recommend
to set up a small but large enough test system, that
would cover the worst case scenario and determine
properly, if the interactions are treated the way you
expect. you can just use neighbor list exclusion tricks
to only compute the interaction energy and forces for
each part of the system separately, then the cross-term
and finally check if the sum of the term matches with
everything enabled.

The way i see it, the two tubes never come closer than 1.7A which is the
cutoff in REBO so i need not
worry about it. Will it make any difference in the computing time?

again, the best way to find out, is to test this empirically.
why guess, when you can measure? :wink:

axel.

Dr.Axel,

Thanks for the reply. I tested this against a in-house code and I got the same results.
I compared the interaction energies, forces due to LJ and REBO and found that “for my case” using neigh_modify for the manybody interactions between two nanotubes is not necessary and gives same results as the case where I directly overwrite the interactions with LJ potential where needed. Though this technique does seem non- physical.

Also coming to think of it, I think it will be better for me to use an hybrid/overlay style as this would facilitate for any cross-linking between the nanotubes, thought that is unlikely to happen in my case. But why take chances!

Thanks,
Freddie.

If you have set the atoms in one CNT to be type
1 and in the other to be type 2 (both still Carbon),
then what you have will work no matter how
close together the 2 tubes get. See the pair hybrid
doc page and its discussion of using many-body
potentials. The example it gives iwth Tersoff and
2 surfaces is conceptually the same as your 2 CNTs.

Steve