I'd like to speed up a simulation of a polymer brush in good solvent (FENE bonds on the same chain, Lennard-Jones potential interchain) by allowing bonds to cross, since I'm only interested in the final configuration of the chains, e.g. density profile and brush height, and not the dynamics. Right now, my input script contains the standard non-crossing coefficients:
bond_coeff 1 30.0 1.5 1.0 1.0
From reading the Kremer/Grest 1990 reference on the FENE documentation, it appears k, the spring constant, can be decreased to allow bonds to cross. I can't find further details on their study of different values of k and R_0. I'm not sure how much to decrease k, whether to change R_0, and if anything else in my script would need to be altered, e.g. the size of the time step, which is currently at 0.01tau. I'd appreciate any suggestions.
thanks in advance,
I'd like to speed up a simulation of a polymer brush in good solvent
(FENE bonds on the same chain, Lennard-Jones potential interchain) by
allowing bonds to cross, since I'm only interested in the final
configuration of the chains, e.g. density profile and brush height, and
i am not at all an expert in this, but have you looked at
the "fix bond/swap" command? what you describe sounds to me
a lot like that this does.
Bond/swap is a good tool - it greatly speeds equilibration.
Reducing the FENE spring constant will change the equilibrium bond length and hence the structure. Probably not what you want.
Another possible way to do is to temporarily core-soften the LJ (at small r) s. t. crossings occur. But you want to go back to the LJ+FENE at the end, right? In other words, you just want to allow crossings to equilibrate the system, right?
Right, I want to allow crossings to equilibrate the system. Will bond/swap only reduce the number of time steps it takes to reach equilibrium, or will it also affect the amount of time it takes to execute each time step?
Robert Hoy wrote:
Using fix bond/swap
should be relatively low overhead. Note that it is not allowing
crossings per se, rather it is swapping the ends of 2 chains at specific
bond locations on both chains, as described on the doc page.
Also, the thing with the bondswap procedure is that the best way to do it is to start off at a low k_FENE (which increases swapping), then increase k_FENE up to its “equilibrium” value (usually 30) over a couple stages.