Computational homogenization of parabolic equations with memory effects for a periodic heterogeneous medium Note: The work was supported by the Russian Science Foundation (grant No. 24-11-00058). Journal: arXiv

P.N. Vabishchevich · 2025

Read Computational homogenization of parabolic equations with memory effects for a periodic heterogeneous medium Note: The work was supported by the Russian Science Foundation (grant No. 24-11-00058). Journal: arXiv by P.N. Vabishchevich (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-07111.txt.

Introduction

Vabishchevich presents a paper from twenty twenty-five on computing averaged models for diffusion in patterned materials. Fine-scale models are too costly to solve across whole parts. Averaging builds macro models from one repeating cell's solution. With weakly conducting inclusions, the macro model gains memory terms. The memory kernel comes from a time-varying cell problem. An effective spread tensor follows from standard finite element work. The kernel is then matched by a sum of decaying exponentials.

That match turns the nonlocal model back into local form. Extra helper problems carry the memory effects cheaply. Stability of the time stepping is proved. A two-dimensional test case illustrates the whole method.

Catalog record

CREATORP.N. Vabishchevich
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE36 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-07111

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Computational homogenization of parabolic equations with memory effects for a periodic heterogeneous medium Note: The work was supported by the Russian Science Foundation (grant No. 24-11-00058). Journal: arXiv P.N. Vabishchevich Abstract In homogenization theory, mathematical models at the macro level are constructed based on the solution of auxiliary cell problems at the micro level within a single periodicity cell. These problems are formulated using asymptotic expansions of the solution with respect to a small parameter, which represents the characteristic size of spatial heterogeneity. When studying diffusion equations with contrasting coefficients, special attention is given to nonlocal models with weakly conducting inclusions. In this case, macro-level processes are described by integro-differential equations, where the difference kernel is determined by the solution of a nonstationary cell problem. The main contribution of this work is the development of a computational framework for the homogenization of nonstationary processes, accounting for memory effects. The effective diffusion tensor is computed using a standard numerical procedure based on finite element discretization in space. The memory kernel is approximated by a sum of exponentials obtained from solving a partial spectral problem on the periodicity cell. The nonlocal macro-level problem is transformed into a local one, where memory effects are incorporated through the solution of auxiliary nonstationary problems. Standard two-level time discretization schemes are employed, and unconditional stability of the discrete solutions is proved in appropriate norms. Key aspects of the proposed computational homogenization technique are illustrated by solving a two-dimensional model problem. MSC 35B27 , 45K05 , 65M12 , 65M60 1 Introduction Modern challenges in mathematical modeling of transport processes in inhomogeneous media require developing effective methods for analysis and numerical solution, particularly for media with pronounced microstructures and complex physical properties. Typical examples include composite materials, porous structures, biological tissues, and geological formations, where transport coefficients may vary significantly in space and time. Direct numerical modeling of such systems faces substantial computational difficulties due to the need to resolve small-scale inhomogeneities, making the application of multiscale methods [ 1 , 2 , 3 ] particularly relevant. For periodic structures, fundamental results in homogenization theory have been obtained using two-scale convergence. This concept provides rigorous justification for transitioning from microscopic to effective macroscopic descriptions (see [ 4 , 5 ] ). The effective diffusion tensor coefficient is computed through solutions of auxiliary cell problems. Special attention is given to high-contrast diffusion coefficients, where standard homogenization methods require significant modification. For weakly conducting inclusions, memory effects become crucial, and the diffusion process is described by integro-differential equations [ 6 , 7 ] . The memory kernel is determined by solving nonstationary cell problems on periodic microstructures. The central challenge addressed in this work involves numerical solutions to integro-differential equations [ 8 ] . Conventional approaches requiring complete solution history storage lead to exponential growth in memory and computational resource demands. More promising computational approaches for memory-effect problems involve transitioning from nonlocal to local formulations [ 9 ] . This transformation can be achieved through exponential approximations of memory kernels [ 10 , 11 ] . Memory kernel approximation by exponential sums deserves particular attention [ 12 ] , with fundamental approaches noted in [ 13 ] . This paper presents a novel computational homogenization technique for nonstationary processes incorporating memory effects. The technique involves three key steps: (i) solving the periodicity cell problem for effective diffusion tensors, (ii) approximating memory kernels through spectral problems on cells, (iii) constructing extended macroscopic equation systems. This approach solves both memory kernel approximation and macroscopic nonlocality problems, showing strong potential for practical applications. The organization of our work is as follows. Section 2 formulates the homogenization problem for nonstationary diffusion equations in periodic media, focusing on memory effects. Section 3 describes the nonlocal-to-local transition, determining memory kernels from spectral problems on periodicity cells, enabling direct exponential-sum approximations and extended macroscopic systems. Section 4 details the computational algorithm using finite element spatial approximations and two-level temporal schemes, with stability estimates in appropriate spaces. Section 5 presents numerical solutions for two-dimensional diffusion problems, demonstrating the algorithm’s effectiveness for nonstationary process homogenization. Section 6 concludes with study findings. Figure 1: The computational domain \Omega with periodic inhomogeneities of properties and a periodicity cell for a two-component medium. 2 Problem statement We consider the problem of nonstationary diffusion in a domain \Omega\subset\mathbb{R}^{2} (Fig. 1 ). The medium is assumed to be periodic with period \varepsilon . The substance concentration u^{\varepsilon}(x,t) is determined from the equation \frac{\partial u^{\varepsilon}(x,t)}{\partial t}=\nabla\cdot\left(d^{\varepsilon}(x)\nabla u^{\varepsilon}(x,t)\right),\quad x\in\Omega,\ t>0. (2.1) The periodically oscillating diffusion coefficient d^{\varepsilon}(x) , which depends on the small parameter \varepsilon>0 , has the form: d^{\varepsilon}(x)=d\left(\frac{x}{\varepsilon}\right), where d(y) is a periodic function of the variable y=x/\varepsilon . Equation ( 2.1 ) is supplemented by initial and boundary conditions: u^{\varepsilon}(x,0)=u_{0}(x),\quad x\in\Omega, (2.2) u^{\varepsilon}(x,t)=0,\quad x\in\partial\Omega,\ t>0. (2.3) We consider two-component media. In this case, the periodicity cell Y=[0,1]^{2} is divided into two non-overlapping domains: Y=Y_{1}\cup Y_{2} (Fig. 1 ). The diffusion coefficient d(y) takes constant values in Y_{1} and Y_{2} . Different homogenization models are used depending on the contrast ratio \delta of the diffusion coefficient in Y_{2} to that in Y_{1} . Analysis of different limiting values of \delta leads to qualitatively different averaged models. Integro-differential homogenization models are associated with the most interesting case of weakly permeable inclusions when \delta\to 0 . We consider the approximation with \delta\sim\varepsilon^{2} : d(y)=\begin{cases}d_{1},&y\in Y_{1},\\ \varepsilon^{2}d_{2},&y\in Y_{2},\end{cases}\quad\frac{d_{1}}{d_{2}}=\mathcal{O}(1). (2.4) Under the assumptions ( 2.4 ), the approximate solution to the initial-boundary value problem ( 2.1 )–( 2.3 ) for averaged values was formulated in [ 6 , 7 ] . The average concentration u(x,t) satisfies the following integro-differential equation: \frac{\partial u(x,t)}{\partial t}+\chi(t)*\frac{\partial u(x,t)}{\partial t}=\nabla\cdot\left(D\nabla u(x,t)\right),\quad x\in\Omega,\ t>0. (2.5) Here * denotes the convolution operator with respect to t : \chi(t)*f(t)=\int_{0}^{t}\chi(t-s)f(s)\,ds. The boundary and initial conditions for equation ( 2.5 ) are formulated similarly to ( 2.2 ), ( 2.3 ): u(x,0)=u_{0}(x),\quad x\in\Omega, (2.6) u(x,t)=0,\quad x\in\partial\Omega,\ t>0. (2.7) The diffusion tensor D and the memory effects kernel \chi(t) in equation ( 2.5 ) are determined by solving boundary-value problems on the periodicity cell Y . To determine D , we use a standard procedure (see, e.g., [ 14 ] ) accounting for the strong contrast in the diffusion coefficient. Let \bm{e}_{i} be the unit vector in the i -th direction, i=1,2 . The auxiliary functions \theta_{i}(y) are defined as solutions to the equation on the part Y_{1} of the periodicity cell: \nabla_{y}\cdot\left(d(y)\left(\bm{e}_{i}+\nabla_{y}\theta_{i}(y)\right)\right)=0,\quad y\in Y_{1},\quad i=1,2. (2.8) Periodic boundary conditions are imposed on the external boundary \partial Y , while Neumann conditions are specified on the internal boundary: \frac{\partial\theta_{i}(y)}{\partial n}=-(\bm{e}_{i},\bm{n})=0,\quad y\in\partial Y_{2},\quad i=1,2, (2.9) where \bm{n} is the outward normal. The unique solution is determined by the normalization condition: \int_{Y}\theta_{i}(y)\,dy=0,\quad i=1,2. (2.10) The effective diffusion tensor D is determined from the solution of the cell problem ( 2.8 )–( 2.10 ): D_{ij}=|Y_{1}|^{-1}\int_{Y_{1}}d(y)\left(\delta_{ij}+\frac{\partial\theta_{i}(y)}{\partial y_{j}}\right)\,dy,\quad i,j=1,2,\quad|Y_{1}|=\int_{Y_{1}}dy, (2.11) where \delta_{ij} is the Kronecker symbol. This tensor is symmetric and positive definite. For weakly permeable inclusions, we formulate an auxiliary non-stationary problem. The function \psi(y,t) is the solution to the initial-boundary value problem: \frac{\partial\psi(y,t)}{\partial t}=\nabla_{y}\cdot\left(d_{2}\nabla_{y}\psi(y,t)\right),\quad y\in Y_{2},\ t>0, (2.12) \psi(y,0)=1,\quad y\in Y_{2}, (2.13) \psi(y,t)=0,\quad y\in\partial Y_{2},\ t>0. (2.14) The memory kernel \chi(t) is determined from the solution of problem ( 2.12 )–( 2.14

Read the complete text