Spectral theory of effective transport for discrete uniaxial polycrystalline materials

TL;DR

This paper develops a discrete spectral framework for analyzing effective transport properties in uniaxial polycrystalline materials, leading to efficient numerical algorithms for computing bulk transport coefficients and physical fields.

Contribution

It introduces a discrete matrix analysis that parallels continuum theory, deriving spectral measures and resolvent representations for effective conductivity and resistivity matrices.

Findings

Derived discrete spectral measures for effective transport coefficients.

Developed a projection-based algorithm for efficient spectral measure computation.

Numerically demonstrated the method on 2D and 3D polycrystalline models.

Abstract

We previously demonstrated that the bulk transport coefficients of uniaxial polycrystalline materials, including electrical and thermal conductivity, diffusivity, complex permittivity, and magnetic permeability, have Stieltjes integral representations involving spectral measures of self-adjoint random operators. The integral representations follow from resolvent representations of physical fields involving these self-adjoint operators, such as the electric field $\boldsymbol{E}$ and current density $\boldsymbol{J}$ associated with conductive media with local conductivity $\boldsymbol{\sigma}$ and resistivity $\boldsymbol{\rho}$ matrices. In this article, we provide a discrete matrix analysis of this mathematical framework which parallels the continuum theory. We show that discretizations of the operators yield real-symmetric random matrices which are composed of projection matrices. We derive discrete resolvent representations for $\boldsymbol{E}$ and $\boldsymbol{J}$ involving the matrices which lead to eigenvector expansions of $\boldsymbol{E}$ and $\boldsymbol{J}$. We derive discrete Stieltjes integral representations for the components of the effective conductivity and resistivity matrices, $\boldsymbol{\sigma}^*$ and $\boldsymbol{\rho}^*$, involving spectral measures for the real-symmetric random matrices, which are given explicitly in terms of their real eigenvalues and orthonormal eigenvectors. We provide a projection method that uses properties of the projection matrices to show that the spectral measure can be computed by much smaller matrices, which leads to a more efficient and stable numerical algorithm for the computation of bulk transport coefficients and physical fields. We demonstrate this algorithm by numerically computing the spectral measure and current density for model 2D and 3D isotropic polycrystalline media with checkerboard microgeometry.