molecular degrees of freedom


Thanks for your reply.
You and Steve told me that one needs to explicitely define all the bonds.
Ray replied that it depends on the potential, whether it is able to
determine bonds by itself or not, that is whether it is of reactive type or
So i am still confused !!
I am still wondering if, using the REBO potential, because it is of reactive
type, I do need to define all bonds !
What do you think about that. ?

I think you are right. You don't need to define a list of BONDS (or
angles or dihedrals, etc..) for your system if you use AIREBO/REBO.
   The AIREBO/REBO potentials for hydrocarbons that Ray recommended
are one of the many potential styles available in LAMMPS. It sounds
like AIREBO might be useful for the simulation you want to run, and
it's certainly cheaper than running a quantum mechanical simulation.
However there are probably many other ways of doing this. (You may
want to try running a quantum chemical simulation?) (I confess I
don't know the best way to run a simulation containing radicals.)
Hopefully you can discuss the simulation with someone who can advise
you what type of simulation you need to run.

   The approach Steve and I were proposing is more common. If you
want to link a atoms together in a strong way which prevents the bond
from breaking, then you would create a DATA file containing a "Bonds"
section which has list of pairs of atoms which are bonded. These atom
pairs can not be separated later (unless you use the fixbond/break
command LAMMPS Molecular Dynamics Simulator).
This is useful because the forces acting on atoms which have been
bonded together this way can be highly customised (finely tuned) using
many different 3-body angle and 4-body dihedral/improper forces,
including tabular (spline) forces.

It sounds like perhaps REBO/AIREBO approach might be better for you.

Incidentally. LAMMPS you can also use multiple force field styles
together (using pair_style "hybrid",
LAMMPS Molecular Dynamics Simulator), using REBO/AIREBO for
the atoms which need it, and using unbreakable bonds for the atoms
which don't. (This kind of simulation is a little more difficult to
set up.)

Good luck


a great thank to all of you.
It's much more clear now !

Everything Andrew says is correct. I'll just
add that you can use atom_style atomic with
(AI)REBO. You only need atom_style bond
or molecular if you define explicit bonds
in your data file.


Dear Andrew,
I am currently doing some simulations on interaction
between DLC layer with its substrate with LAMMPS.
In the process, we intend to heat up the DLC using
a thermostat layer of atoms. However, with the fix
heat command, the temperature of the thermostat
layer may increase to an extent where atoms may be
lost from the thermostat layer which will cause
the simulation to abruptly stop.

I saw in one of the LAMMPS archive that you mentioned
about using pair_style hybrid where one is the
Airebo and one is unbreakable bonds as shown in
the following link.
LAMMPS Molecular Dynamics Simulator

I am sorry. What I wrote here was misleading.
You don't need to use pair_style hybrid if you are only
trying to add bonds to a simulation (unbreakable or otherwise).

To use bonds, you must create a LAMMPS data file for your system (if
you have not done this already) and add a "Bonds" section to your data

--- details regarding the "Bonds" section ---

Each line of the "Bonds" section contains 4 integers, and defines a single bond.
The format for each line is:
  bondid bondtype atomid1 atomid2
The first number (bondid) is a unique ID number. This number begins
at 1 and counts the number of bonds in your system. For a detailed
example, read the "examples/peptide/date.peptide" file which comes
with LAMMPS.

Also see the documentation for the "read_data" command:

I hope this helps.