Karthik, I am without a computer at the moment and can not access my
dreiding example files.
If my memories are correct, then I am under the impression that the
hbond/dreiding/lj is numerically stable and was conserving energy. I
think that I tested it using a timestep between 0.5fs and 2fs. (I
can't remember exactly.)
1) I am not familiar with "fix_modify 2 energy yes" , but try
removing it and running the simulation at NVE. (If you want the total
energy to be constant, why would you use nvt?)
2)
Make sure the equilibrium length of the donor-accepter bond is not
less than 1.0 (angstroms)
3)
a) Minimize the system before running the simulation.
b) After minimization, give the atoms a non-zero velocity
(For example by using the velocity command:
http://lammps.sandia.gov/doc/velocity.html)
c) Then start your test simulation
What to expect:
There may be a fluctuation in total energy in the first few timesteps,
especially if the system has not been minimized or equilibrated
beforehand. Ignoring that, in a short simulation (100000 timesteps),
the total energy should not change by more than 2%. Aim for that.
Please try this and report back to us if it worked.
Cheers
Andrew