Thanks so much for your reply and suggestion. I did find out there is still one water molecule bond has incredibly large distance. However, after I removed these atoms from bond and I rather defined them as single atoms in the modal still turned out the NAN results.
this modification is a bad idea. you are changing your model in an
unphysical way. however, it has nothing to do with the NaN.
When you mentioned my water model interactions are not consistent with popular water models, do you mean there are something wrong with the geometry coordinate or it's the in appropriate potential input.
popular water potentials like TIP3P, TIP4P, SPC, SPC/E and their many
variants have a combination of coulomb and lennard jones interactions.
you were only having lennard jones.
Besides, I tried to change the pair_style hybrid/overlay to pair_style hybrid. It displayed error as "Pair hybrid sub-style is not used". Base on my limited understanding so far. I cannot figure out how to deal with this situation. Could you provide ways to solve it?
that error is because your
pair_coeff * * lj/cut ...
statement will wipe out all previous pair_coeff settings.
In addition, probably is not really relate with Lammps, but I have yours advice that my input potential is way too low. I have potential information like following table for example. In my case for units real, I change the parameter obtained from eV to kcal/mol and I put them into my input script. Do you have any command for this step?
my recommendation is that you have to go back to learn the basics. you
obviously have too little background knowledge, experience and
training to do a simulation of a complex system as you are trying to
do. i suggest you start practicing setting up MD simulations with
something much more simple (like liquid argon and then liquid water)
and keep working with those simple and well understood systems with
lots of published data until you are actually able to make informed
choices. then you could try something more complex (e.g. a mixture of
water and argon) which you can do with a single pair style, but then
practice using a hybrid pair style to see if you can reproduce the
this is best done under the supervision of somebody local with
sufficient experience in MD (doesn't have to be with LAMMPS for the
at any rate, there are far more problems with your existing input,
than we can resolve via a few e-mail exchanges.