In the rawest form, the electrostatic potential required calculations per evaluation. Particle Mesh Ewald can we be used to reduce the complexity to but is complicated to implement.
Cutoffs are a simpler approach. At distances of about 1 nm and greater, the electrostatic forces are small enough to ignore without significant effects on accuracy. With an abrupt truncation, the electrostatics are simply ignored beyond a cutoff radius :
However, the resulting potential isn’t continuous, resulting in a failure to conserve energy. A switching function can be used to smoothly switch the potential to zero in a boundary region () to produce a continuous function:
Cutoffs with a switching function also happen to be the fastest method, requiring only calculations per evaluation.
Performance improvements are not the only reason to use switching and cutoffs, however. When I first tried to simulating water boxes, I failed to use any form of cutoffs. As a result, the waters formed a rigid lattice with no diffusion activity. (This was less of an issue with solvated proteins since the protein disrupted the periodicity of the potential.) After several weeks of frustrated effort, I noticed that most popular MD software packages employ cutoffs by default. Once I changed my forcefield settings to switch the electrostatics, I was able to recover proper hydrogen bonding networks and diffusion rates.