Dedicated to George Papanicolaou on the occasion of his 70th birthday.Abstract. The analytic continuation method of homogenization theory provides Stieltjes integral representations for the effective parameters of composite media. These representations involve the spectral measures of self-adjoint random operators which depend only on the composite geometry. On finite bond lattices, these random operators are represented by random matrices and the spectral measures are given explicitly in terms of their eigenvalues and eigenvectors. Here we provide the mathematical foundation for rigorous computation of spectral measures for such composite media, and develop a numerically efficient projection method to enable such computations. This is accomplished by providing a unified formulation of the analytic continuation method which is equivalent to the original formulation and holds for finite and infinite lattices, as well as in continuum settings. We also introduce a family of bond lattices and directly compute the associated spectral measures and effective parameters. The computed spectral measures are in excellent agreement with known theoretical results. The behavior of the associated effective parameters is consistent with the symmetries and theoretical predictions of models, and the computed values fall within rigorous bounds. Some previous calculations of spectral measures have relied on finding the boundary values of the imaginary part of the effective parameter in the complex plane. Our method instead relies on direct computation of the eigenvalues and eigenvectors which enables, for example, statistical analysis of the spectral data. ).
825
826
SPECTRAL MEASURE COMPUTATIONSThis formulation demonstrated that the measures underlying these integral representations are spectral measures associated with the random operators, which depend only on the composite geometry. These measures contain all the information about the mixture geometry, and provide a link between microgeometry and transport. Local geometry is encoded in "geometric" resonances in the measures [46], while global connectivity is encoded by spectral gaps [58,46] and the presence of δ-components in the measures at the spectral endpoints [58]. A remarkable feature of the method is that once the spectral measures are found for a given composite geometry, by the spectral coupling of the governing equations [13,56,14,18], the effective electrical, magnetic, and thermal transport properties are all completely determined by these measures.The integral representations yield rigorous forward bounds on the effective parameters of composites, given partial information on the microgeometry [7,53,33,8,10]. One can also use the integral representations to obtain inverse bounds, where data on the electromagnetic response of a sample, for example, is used to bound its structural parameters, such as the volume fractions of the components [51,52,16,13,17,75,9,15,21,32], and even the separation of the inclusions in matrix particle composites [59]. Furthermore, the spec...