On the Bloch decomposition based spectral method for wave propagation in periodic media

Zhongyi Huang, Shi Jin, Peter A. Markowich, Christof SparberWave 2008 · DOI

TL;DR

This paper introduces and validates a novel numerical method for simulating wave propagation in periodic media. By leveraging Bloch decomposition, their method achieves spectral accuracy in space and allows for significantly larger time steps than traditional approaches, making it highly efficient for complex wave phenomena.

What Problem Does This Paper Address?

Simulating wave propagation in materials with periodic microstructures, such as photonic crystals or composite materials, is computationally challenging. Traditional numerical methods often require extremely small grid sizes and time steps to accurately capture the fine details of the periodic structure and the short wavelengths involved. This leads to prohibitively high computational costs, especially for problems with oscillatory or discontinuous coefficients. Existing methods that simplify the problem by using homogenized equations lose crucial information about the microscopic dynamics. This paper aims to develop a numerical method that is both accurate and efficient, allowing for larger time steps and mesh sizes without sacrificing accuracy, even for complex periodic media.

Methods

The core of the paper is the Bloch decomposition based time-splitting spectral method. This approach decomposes the wave solution into a sum of "Bloch modes," which are eigenfunctions of the periodic operator. This decomposition allows for an exact treatment of the periodic coefficients in the spectral domain. The method then uses a time-splitting scheme to handle the time evolution:

  1. Bloch Decomposition and Spectral Evolution: The wave solution is projected onto the Bloch modes. The time evolution of the coefficients of these modes is then solved in the spectral domain, where the periodic operators become diagonal. This step is essentially exact for the periodic part of the equation.
  2. Standard ODE Solver: The remaining parts of the equation (e.g., source terms or nonlinearities) are handled by solving standard ordinary differential equations in physical space.

The paper uses a split-step algorithm, specifically Strang's splitting, to achieve second-order accuracy in time. A key advantage highlighted is that the preprocessing step (calculating Bloch bands and eigenfunctions) is independent of the spatial grid size used for the simulation, making it computationally inexpensive compared to the overall simulation cost.

Datasets

Not discussed in the paper. The paper focuses on numerical simulations of a specific class of equations using synthetic initial conditions and coefficients.

Key Insights and Ideas

  • Bloch Decomposition as a Game Changer: The paper demonstrates that by decomposing the wave into Bloch modes, one can effectively "diagonalize" the periodic operators, leading to an exact treatment of the medium's periodic structure in the spectral domain. This bypasses the need for extremely fine grids to resolve the microscopic periodicity.
  • Unconditional Stability and Spectral Accuracy: The method is shown to be unconditionally stable and achieves spectral accuracy in space, even for highly oscillatory or discontinuous periodic coefficients. This is a significant improvement over traditional methods that struggle with such features.
  • Larger Time Steps: A major practical advantage is the ability to use time steps that are independent of the small parameter $\epsilon$ (representing the ratio of microscopic to macroscopic scales), which is typically constrained by the CFL condition in other methods. This drastically reduces computational cost.
  • Handling Discontinuities: The method's robustness is highlighted by its ability to handle non-smooth (discontinuous) periodic coefficients, such as the Kronig-Penney model, while maintaining spectral accuracy.
  • Anderson Localization Insights: The paper extends the method to study wave propagation in disordered media with random coefficients, providing numerical evidence for Anderson's localization phenomenon.

Key Results

The paper presents extensive numerical simulations comparing the Bloch decomposition method (BD) with standard pseudo-spectral (SP) and finite-volume (FV) methods.

  • Example 3.1 & 3.2 (Linear & Oscillatory Initial Data): For smooth and non-smooth coefficients, the BD method consistently achieves spectral convergence in space, even with a single time step computation. In contrast, SP and FV methods show significantly poorer convergence rates, especially for non-smooth coefficients, and require much smaller time steps ($\Delta t \ll \epsilon$) to achieve reasonable results.
  • Example 3.3 (Nonlinear Wave Propagation): In a nonlinear setting, the BD method again demonstrates superior convergence rates compared to the pseudo-spectral method, even in weakly nonlinear cases.
  • Example 4.1 & 4.2 (Random Coefficients & Anderson Localization): The method is stable in the presence of random perturbations. Numerical evidence is provided for Anderson's localization, showing that the wave spread (measured by $A(t)$) exhibits linear growth and then saturates, indicating strong localization.

Limitations and Future Work

  • Computational Cost of Preprocessing: While the preprocessing (calculating Bloch bands) is stated to be independent of the spatial grid, it can still be computationally intensive for very complex periodic structures or a large number of bands.
  • Truncation of Bloch Bands: The method involves truncating the sum of Bloch bands at a certain number $M$. The paper mentions that $M$ needs to be chosen large enough to ensure mass conservation, but the optimal choice of $M$ for different problems might require further investigation.
  • Extension to Higher Dimensions: The current work focuses on one-dimensional problems. Extending the Bloch decomposition method to two or three dimensions, where the band structure can be more complex, is a natural and important direction for future research.
  • More Complex Nonlinearities and Disordered Media: Further exploration of the method's performance with more intricate nonlinearities and a wider range of disordered media models could be beneficial.

Summary updated

BibTeX
@inproceedings{huang2008bloch,
  author = {Zhongyi Huang and Shi Jin and Peter A. Markowich and Christof Sparber},
  title = {{On the Bloch decomposition based spectral method for wave propagation in periodic media}},
  booktitle = {Wave},
  year = {2008},
  doi = {10.1016/j.wavemoti.2008.07.001},
}
Figure