The dipole expansion technique for determining the grid charge density and the electrostatic forces due to a system of extended charges is presented. In this scheme a charge (monopole) and a dipole moment, both extended, are assigned to the nearest grid point. The corresponding calculations of the force require the determination of the electric field and its gradient. The approach aims to minimize the number of operations per particle at the expense of more operations per grid point. The technique is appealing on physical grounds, since it helps one to relate numerical approximations to physical concepts. We also present a modified version, the subtracted dipole scheme (SUDS). In this version the field calculations do not take significantly longer than in nearest-grid-point calculations. This is particularly important in two- and three-dimensional simulations. We also discuss an optimization of the coding, particularly for the operations that must be done for each particle (computing its contribution to the charge density and updating its position and velocity). For example, this optimization has given a one-dimensional, nearest-grid-point code (lowest order in the multipole expansion scheme) that takes ∼1.8 μsec/particle on the IBM 360/91. We assess the value of the dipole correction by comparing dipole and nearest-grid-point simulations. Finally the relation between the dipole expansion (or SUDS) and charge sharing is discussed.
Read full abstract