Thank you very much for your rapid reply.
I run 3 simulations to come to the conclusions: the first simulation is using AIREBO (no LJ potential and torsional term, Pair_style airebo 3.0 0.0 0.0). The second simulation use LJ/cut/coul/cut only. The third one use the hybrid/overlay to combine the potentials in the Simulation one and two. The results turn out that the energy in simulation 3 is close to Simulation two.
This problem can be solved by using the following command instead:
pair_style hybrid/overlay airebo 3.0 1.0 1.0 coul/cut 10.0
!!!! (include the LJ potential in first airebo rather than in the second long-ranged potential)
pair_coeff * * airebo CH.airebo C
pair_coeff * * coul/cut
If you're thinking that the LJ term in AIREBO should produce
the same results as the LJ term in pair lj/cut/coul/cut, I don't
think there is any guarantee of that. You'd have to look
carefully to see what pairs are included/excluded with what
weighting factors in the 2 cases. The lj/cut/coul/cut will
exclude nothing, and I'm pretty sure AIREBO will exclude
a lot.