[lammps-users] Different results : fortran LAMMPS (2001) vs C++ LAMMPS (July 07)

Thanks but I don’t think lj will matter in both systems as I set EpsilonLJ equal to zero

I don't see anything obviously different between the 2 scripts.
Setting epsilon = 0.0 means there are no interactions
other than between bonded (and angled) monomers. I.e. the
chains do not interact with each other?

Do the 2 simulation equlibrate to the same state point, i.e.
temp and pressure?


Yes they equilibrate at the same state points. I changed the timestep of
each script from 0.002 to 0.01 and noted that both <R^2(n)> for fortran and
c++ runs converged to the theoritical value of <R^2(n)> for semiflexible
polymer chain. This tells me that I need to run simulations longer and
sample data points at least five Rouse relaxation times to get reliable
data. I believe the initial deviation was because fortran and c++ have
different random number generators so the initial trajectories are different
but will eventually average out during longer runs.

In fortran lammps, I noted that I need to specifiy non bonded or pair_style
interactions explicitly otherwise I get an error. I did a simulation in C++
Lammps where I did not specify pair_style and I didn't get an error. Is this
fine? Is setting Epsilon_LJ for pair_style equal to zero equivalent to not
defining it in the input script?


[mailto:[email protected]] On Behalf Of Steve

C++ LAMMPS does not require a pair style be defined (or bond, angle, etc).
Should be the same as LJ with epsilon = 0, but a lot cheaper.

If you have no pair_style, you do need to insure the atoms in bonds
are communicated from far enough away. Since there is no pair
cutoff, you'd need to set the neighbor skin big enough to include them.