Abstract:Numerical multiscale methods usually rely on some coupling between a macroscopic and a microscopic model. The macroscopic model is incomplete as effective quantities, such as the homogenized material coefficients or fluxes, are missing in the model. These effective data need to be computed by running local microscale simulations followed by a local averaging of the microscopic information. Motivated by the classical homogenization theory, it is a common practice to use local elliptic cell problems for computing the missing homogenized coefficients in the macro model. Such a consideration results in a first order error $O(\varepsilon/\delta)$, where $\varepsilon$ represents the wavelength of the microscale variations and $\delta$ is the size of the microscopic simulation boxes. This error, called "resonance error", originates from the boundary conditions used in the micro-problem and typically dominates all other errors in a multiscale numerical method. Optimal decay of the resonance error remains an open problem, although several interesting approaches reducing the effect of the boundary have been proposed over the last two decades. In this paper, as an attempt to resolve this problem, we propose a computationally efficient, fully elliptic approach with exponential decay of the resonance error.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the resonance error (resonance error) caused by boundary conditions in numerical assimilation methods. Specifically, when using local elliptic cell problems to calculate the missing effective material coefficients in the macroscopic model, a first - order error \(O(\varepsilon / \delta)\) related to the micro - scale variation wavelength \(\varepsilon\) and the micro - simulation box size \(\delta\) will be generated. This error is called "resonance error", and it usually dominates in multi - scale numerical methods.
To reduce the influence of this resonance error, the paper proposes a new fully elliptic method, which can make the resonance error decay exponentially. This method not only improves the computational efficiency, but also can achieve an arbitrarily high convergence rate without increasing additional computational costs.
### Paper Background
Numerical multi - scale methods usually rely on the coupling between macroscopic and microscopic models. The macroscopic model is incomplete because of the lack of data such as effective material coefficients or fluxes. These effective data need to be calculated by running local micro - scale simulations and performing local averaging. Based on the classical assimilation theory, the common practice is to use local elliptic cell problems to calculate the missing assimilation coefficients in the macroscopic model. However, this method will lead to a first - order resonance error \(O(\varepsilon / \delta)\), which originates from the boundary conditions in the micro - scale problem.
### Main Contributions
1. **Proposed a new elliptic cell problem**: The paper proposed a new elliptic cell problem. By adding a correction term, the influence of boundary values can be significantly reduced, thereby achieving exponential decay of the resonance error.
2. **Theoretical analysis**: The paper provided a detailed theoretical analysis, proved the convergence rate of the new method, and gave error estimates.
3. **Numerical strategy**: The paper discussed a numerical strategy based on Arnoldi decomposition for approximately solving the proposed cell problem, so as to achieve efficient computation in practical applications.
### Specific Methods
The new method in the paper is implemented through the following steps:
1. **Modify the elliptic cell problem**: For \(j = 1,\ldots,d\), the new cell problem is expressed as:
\[
-\nabla\cdot(a(x)\nabla\chi_j^{T,R}(x)) = g_j(x)-[e^{-A^T}g_j](x)\quad\text{in }K_R
\]
\[
\chi_j^{T,R}(x) = 0\quad\text{on }\partial K_R
\]
where \(A := -\nabla\cdot(a\nabla)\), \(g_j := \nabla\cdot(ae_j)\), and \(e_j\) is the standard basis vector.
2. **Calculation of assimilation coefficients**: The assimilation coefficients can be calculated by the following formula:
\[
e_i\cdot a_0^{T,R,L}e_j=\int_{K_L}\left(a_{ij}(x)+a_{ik}(x)\frac{\partial\chi_j^{T,R}}{\partial x_k}(x)\right)\mu_L(x)\,dx
\]
where \(\mu_L\) is a filter used to accelerate the convergence of non - periodic changes.
### Conclusion
The paper verified the effectiveness of the new method through theoretical analysis and numerical experiments. The new method not only theoretically realizes the exponential decay of the resonance error, but also has high computational efficiency in practical applications. This provides an important improvement direction for multi - scale numerical assimilation methods.