A high-order accurate algorithm for electrostatics of overlapping disks (Q1284867)

Random aggregates of overlapping objects are frequently studied geometries in the physics of disordered media: Objects with physical properties described by certain moduli are placed, at random, in a matrix with different moduli. Much work for random aggregates concerns effective properties. Of particular interest, and difficulty, is to predict effective properties of highly inhomogeneous aggregates close to the continuum percolation threshold, that is, close to the area fraction for which the randomly placed objects start to form a connected path through the material. Popular numerical methods for this problem are often based on statistical ideas and include biased diffusion, Monte Carlo or random walk simulations, ``blind-ant'' algorithms, resistor network approximations, and series expansion incorporating structural parameters. Most of these methods converge as \(1/\sqrt{N},\) where \(N\) is the computational work. A typical relative error for the effective moduli of strongly inhomogeneous large aggregates is on the order of 0.05. The author presents the algorithm for the numerical investigation of conductivity behavior for continuum percolation in two dimensions. The geometry can be arbitrary. As the examples, the disks are chosen because it is a standard choice. This paper demonstrates that the accurate numerical calculation of effective properties of strongly inhomogeneous large random aggregates of overlapping objects is not only possible, but also simple to perform on a regular workstation. The author specializes to overlapping disks and computes the effective conductivity for unit cells at two hundred different area fractions and with a relative error of 0.0005 at a modest computational cost. At percolation the unit cells contain around 3,500 disks. The key ingredients in the proposed algorithm are: a new integral equation formulation for the electrostatic partial differential equation on a doubly periodic domain, the fast multipole method for potential field evaluations, a recent algorithm for the evaluation of layer potentials close to their sources.