I'm running simulations for a bead-spring polymer networks using LAMMPS. I need my simulations to reach biologically relevant timescales of at least a few milliseconds. However each simulation would take upwards of a few weeks given my current use of Langevin dynamics and velocity verlet integration. I need to collect data on at least 50 such simulations, so this is prohibitive. Is there any way to improve performance? Are there absolutely stable, implicit integration schemes for molecular dynamics integrators?
Hello Brian Adam,
I do not have an idea on your simulation configuration; however, I may suggest you modifying the neighbour lists to reduce the complexity of your simulation if you have not already. You may find more information about neighbour lists in LAMMPS manual. Even though this will not be critical as the neighbour lists, you may also consider decreasing the cutoff distances defined for the pairwise interactions.
Both of them are commonly used to decrease the complexity and the simulation length by giving up on some accuracy of the simulations; so, you need to be careful while assigning some values to them.
Regards
- Badges
- Science topic