[lammps-users] hybrid airebo + class2

Dear Mr. Plimpton,

I'd like to use airebo with class 2 forcefields. Instead of hybrid, I
have decided to use hybrid/overlay and set the redundant force
co-efficients in class2 to zero. This is basically a roundabout method
to apply hybrid, and i find it more convenient to adjust those class 2
parameters (which are being considered by REBO) to zero. But when
applying a hybrid pair_style i came across an unstable system, even
though the corresponding system with pure AIREBO is stable.

Essentially, I'm trying to establish an equivalence between
'airebo+zero_parameter_class2' and pure airebo. I should be able to
get the exact energies for identical systems.

I am doing that because i'm studying functionalized CNTs, for which
airebo does well for the CNT and class2 is my best option for studying
the functional group. But i'm unable to achieve the above equivalence
owing to discordant energy profiles in the two cases. I have attached
my input scripts. One of the input scripts (mixed) has been generated
by MedeA.

Please advise me how i should implement such a hybrid scheme.


CH.airebo_real (770 KB)

input_mixed.dat (3.33 KB)

structure_mixed.dat (32.7 KB)

input_pure.dat (608 Bytes)

structure_pure.dat (5.39 KB)


i recently noticed a similar problem with airebo in lammps.
i have not been able to track it down completely, but for now
it seems that only systems that use the atomic atom style are
currently giving correct results with airebo.


I think this is an issue that has come up before for Tersoff.
None of the many-body potentials (AIREBO, Tersoff, SW, EAM, etc)
are expecting to operate with neighbor lists which have bond exclusion
terms in them, due to use of special_bonds weighting factors.
So currently they will do something bogus if this is the model you have setup.
Some logic could be added to allow for this or to throw an error, but ...

The real question to me is what are you expecting to model?
In LAMMPS, the manybody potential (AIREBO, Tersoff, etc)
is operating on a set of atoms (could be all the atoms or a subset).
(a) Within that (sub)set are you also wanting to define fixed bonds
and compute things like a harmonic bond? (b) And are you
expecting some kind of weighting factor to be applied to
the manybody potential to account for the presence of the bond?

To me, neither of these makes sense, since the point of the manybody
potential is to determine whether there is a bonding effect
dynamically. So you'll have to explain your model.

But (a) is possible if we add the logic. Or it would work now
if you just defined special_bonds with all the 3 weighting
factors = 1.0. (b) is not possible without considerable
more work, and I also think it makes even less sense in
a model.


If so, then either
you need to use special_bonds with all 3 params set to 1.0,
or the logic mentioned above needs to be added