Simple coulomb cutoff problem

Hi Jan,

The coulomb cutoff method will not give physical results for a liquid simulation. Cutting off the coulombic energy at 14.0 Angstroms results in a large step change in the energy as atoms move across the cutoff. I ran the diethylether example with (a) Ewald and (b) cutoff. Using Ewald, the average change in energy for attempting to translate one molecule was 1334 +/- 4102 atomic units, resulting in an acceptance ratio of 36%. Using the cutoff, the average energy changes of translation attempts were much larger, 12225 +/- 8982 atomic units and the acceptance ratio dropped to 5%.

The size of the downhill energy attempts also differs between the two methods in the diethylether example. Using Ewald, the average downhill energy attempt for translation moves was -338 +/- 831. Using the cutoff, -3053 +/- 3174.

So, most translation attempts move molecules beyond the cutoff resulting in high changes in energy and low acceptance, making the simulation appear frozen. When a move brings two molecules within the cutoff, the change in energy is also large, but negative. These downhill energy moves all get accepted, causing the energy to keep decreasing.

In short, I recommend against using the coulomb cutoff for a liquid simulation.

Ryan

Correct, there is no option to turn off periodic boundary conditions in Cassandra. For a single molecule simulation, using a cut-off larger than your molecule but smaller than half your box will make it so that the PBCs are never enforced during the energy calculation -- the energies will be identical to the same forcefield without PBCs.