The Bessel Expansion of Fourier Integral on Finite Interval

Yongxiong Zhou, Zhenyu ZhaoSymmetry 2019 · DOI

TL;DR

This paper introduces a novel Bessel function expansion method to efficiently compute highly oscillatory Fourier integrals. By leveraging complex analysis and extending Filon-type quadrature, the method achieves significantly reduced errors, especially at high frequencies, offering a more robust and computationally cheaper alternative to traditional numerical integration techniques.

What Problem Does This Paper Address?

Calculating Fourier integrals of the form $\int_{-1}^{1} f(x)e^{i\omega x} dx$, especially when the frequency parameter $|\omega|$ is large, presents a significant challenge in many scientific and engineering fields, including asymptotic analysis and electromagnetics. These "highly oscillatory integrals" require extremely small step sizes with standard numerical methods like Gaussian quadrature to achieve acceptable accuracy, leading to prohibitively high computational costs. While Filon-type methods have been developed to address this, this paper aims to further refine and extend these techniques by introducing a Bessel function expansion, providing a more systematic and analytically grounded approach.

Methods

The core of this paper is the extension of Filon-type quadrature to a Bessel function expansion. The authors utilize complex analysis to derive theoretical results and error bounds. Key steps include:

  • Hermite Series Expansion: They first express the function $f(z)$ as a Hermite series, which is a polynomial approximation based on derivatives at the interval endpoints.
  • Cauchy Integral Kernel: They leverage the Cauchy integral formula and approximate the kernel using geometric series expansion, leading to a representation of $f(z)$ as an infinite series of polynomials.
  • Bessel Function Expansion: The main contribution is the derivation of the Fourier integral using a Bessel function expansion. This involves transforming the integral into a form that can be expressed as a sum of Bessel functions, weighted by coefficients derived from the function's analytic properties.
  • Error Analysis: Rigorous error analysis is performed using complex analysis, providing upper bounds for the truncation error of the Bessel expansion. These bounds demonstrate that the error decreases with increasing order of the expansion and, crucially, decreases rapidly with increasing oscillation frequency.
  • Handling Branch Points: The paper also extends the method to handle functions with symmetric branch points at $z = \pm 1$, which are common in certain physical problems.

Datasets

Not discussed in the paper. The paper focuses on theoretical development and numerical verification using example functions.

Key Insights and Ideas

  • Bessel Expansion for Oscillatory Integrals: The paper demonstrates that Bessel functions are a natural and effective basis for approximating highly oscillatory Fourier integrals, offering a more structured approach than traditional polynomial approximations.
  • Frequency-Dependent Error Reduction: A key finding is that the proposed Bessel expansion method's error not only decreases with the order of the expansion but also exhibits a rapid decrease as the oscillation frequency ($\omega$) increases. This is a significant advantage over methods that might struggle with higher frequencies.
  • Unified Framework for Analytic and Branch Functions: The method is shown to be applicable to both analytic functions and functions with symmetric branch points, broadening its utility.
  • Complex Analysis as a Powerful Tool: The paper highlights the power of complex analysis in deriving accurate error bounds and understanding the convergence properties of numerical integration methods for oscillatory problems.

Key Results

The paper presents theoretical results and numerical experiments that validate the Bessel expansion method.

  • Theoretical Derivations: Equations (1) and (2) provide specific Bessel function representations for integrals involving $(1-x^2)^\alpha e^{i\omega x}$ and $x(1-x^2)^\alpha e^{i\omega x}$.
  • Error Bounds: Theorems 1, 2, 3, and 4 provide explicit error bounds for the Bessel expansion of Fourier integrals for analytic functions and functions with symmetric branch points. These bounds show convergence and error reduction with increasing expansion order and frequency.
  • Numerical Verification: Examples 1 and 2 showcase the effectiveness of the method. Figure 2 and Figure 3 plot the absolute errors of the expansion against the error estimates, visually confirming the theoretical findings and demonstrating the rapid error decrease with increasing frequency. For instance, in Example 1, the error estimates for Equations (16) and (17) show a significant reduction in error as the expansion order increases.

Limitations and Future Work

  • Limitations: The paper focuses on integrals over the finite interval $[-1, 1]$. While the method is shown to be effective for highly oscillatory integrals, the specific implementation and computational cost for extremely high frequencies or complex functions might still be a consideration. The paper also assumes the function $f(x)$ is analytic or has specific types of branch points.
  • Future Work: The authors state that they will continue to study frequency analysis and calculation, suggesting ongoing research in this area. Potential future directions could include extending the method to different integration intervals, more general types of oscillatory integrands, or exploring adaptive order selection for optimal efficiency.

Summary updated

BibTeX
@article{zhou2019bessel,
  author = {Yongxiong Zhou and Zhenyu Zhao},
  title = {{The Bessel Expansion of Fourier Integral on Finite Interval}},
  journal = {Symmetry},
  year = {2019},
  doi = {10.3390/sym11050607},
}
Figure