[lammps-users] questions on TC of graphene of small simulastion box

I’ve some problem in LAMMPS and I’d like to take opinions on it.

I’ve tried to obtain thermal conductivity of graphene plane in zigzag direction using LAMMPS and airebo potential.
It is supposed that 1000 W/mK should be theretical value in case of Brenner potential.

Airebo potential and Brenner potential are very simliar but I obtained 2~3 times smaller value of 300~400 W/mK.
I performed the test again with turning off both of LJ term and torsional options but it still gives the same result.

The most strange thing is as follows.
I tested 8X14 ,16X28, 8X56 systems (length of arm chair direction X length of zigzag direction) for aformentioned simulations.
I’ve reduced the system size to 6X10(two lengths should be around 25 A and its half length is larger than 3.4*3 = 10.2 of cutoff radius of LJ energy) and tested it.

This time, however, heat fluxs maintain almost the same upto 100,000 steps at least (commonly they decrease significantly after 3,000 steps) and consequently thermal conductivity become very large (more than 80,000 W/mK).
Small size results in smaller TC(thermal conductivity) commonly.
Furthermore, even though oppoiste PBC effect is assumed, TC is too large and it is quite strange.
I also tested the case that both of LJ and torsion options are turned off in airebo potential.

What should be wrong with it?
Is there anyone who observed simliar phenomena?
Should it be something like bug in LAMMPS concerned when airebo potential is treated?
Then, may it be also associated with this bug that smaller TC is obtained for 8X14, 16X28, and 8X56 systems?

Best regards,