Should I use a different cut-off scheme for non-bonded interactions when simulating a system with MD in vacuum?

Question

I am currently doing a simulation of the behaviour of various sizes of water droplets (~100-1000 water molecules) in vacuum. I am using NAMD with the spherical boundary conditions to keep the droplet in place, and using the CHARMM parameters for simulation.

The CHARMM parameter stream files contain a list of options which contain the recommended non-bonded cut offs for vdW forces and electrostatics:

NONBONDED nbxmod  5 atom cdiel shift vatom vdistance vswitch -
cutnb 14.0 ctofnb 12.0 ctonnb 10.0 eps 1.0 e14fac 1.0 wmin 1.5
                !adm jr., 5/08/91, suggested cutoff scheme

(Explanation—pair-list distance= 14 Å, cut-off distance= 12 Å, dielectric const.= 1, 1,4-interaction scaling= 1)

However, these values are only read by the CHARMM software, NAMD manual mentions that NAMD will ignore them, and set the values according to what is set inside the NAMD configuration file.

If I understand correctly, force fields are usually parameterised considering a liquid system (i.e. with 3D periodic boundary conditions) therefore, these cut offs are also likely optimised for a liquid simulation.

What cut-offs should be used if a system in vacuum has to be simulated with CHARMM parameters? (and why?)

My guess here is that because the CHARMM non-bonded parameters were developed with that cut-off, the cut-off should be kept the same, even if the simulation is done in vacuum. However, a more-experienced colleague has told me that I should use a very large cut-off distance (much larger than the droplet, say ~500Å).

So I am confused about what cut-off I should use, and how much it would affect the system if I use a very large cut-off.

(Note that in the simulation I am not trying to measure any dynamic property of the system like diffusion or timescales etc. I am only interested in overall positions of the water molecules, their distributions, effect of a different molecule in the droplet etc.)

Answer 1

I think the effect of larger cutoffs will not affect your simulation. Shorter cutoffs will be more problematic and can see artifacts than the larger cutoffs. Even if you have larger cutoffs, the interaction strength will be very very weak as you go outward and eventually die. This will just increase your computational cost (neighbour list calculation) and will not affect observable properties in the simulation.

In biomolecules, with shorter cutoffs, the conformations of the proteins deviated far away from the initial structure and were significantly distorted, but for longer cutoffs the structure was quite stable. [1]

If you use WCA potential instead of LJ-potential, there can be issues with varying cutoff distances.[2]

Earlier works have shown that with WCA potential, there is little effect of cutoffs in determining the equilibrium structure of fluid in NVT ensemble. Whereas in NPT ensemble, cutoff distances play a key role in fluid equilibrium structure, density and self diffusion coefficient.

I will be looking for more brainstorming on this topic. Feel free to comment.

References:

[1] Norberg, Jan, and Lennart Nilsson. "On the truncation of long-range electrostatic interactions in DNA." Biophysical journal 79.3 (2000): 1537-1553.

[2] Huang, Cunkui, et al. "Effect of cut-off distance used in molecular dynamics simulations on fluid properties." Molecular Simulation 36.11 (2010): 856-864.