Abstract:This article is a short introduction to numerical methods using the double exponential (DE) transformation, such as tanh-sinh quadrature and DE-Sinc approximation. The DE-based methods for numerical computation have been developed intensively in Japan and the objective of this article is to describe the history in addition to the underlying mathematical ideas.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to introduce and review the applications of the Double Exponential (DE) transformation in numerical computation, especially its high efficiency in numerical integration and function approximation. Specifically, the paper mainly focuses on the following aspects:
1. **History and Development of the Double Exponential Transformation**:
- The paper describes the development process of the DE transformation, especially the research progress in Japan.
- It emphasizes the important contributions of the double - exponential formula (such as the tanh - sinh quadrature method) proposed by Hidetosi Takahasi and Masatake Mori in 1972.
2. **Mathematical Principles of the Double Exponential Transformation**:
- It explains the basic form of the DE transformation and the mathematical ideas behind it.
- The formula \(\phi(t)=\tanh\left(\frac{\pi}{2}\sinh t\right)\) is widely used for variable transformation, making the derivative of the integrand show double - exponential decay:
\[
\phi'(t)=\frac{\pi}{2}\frac{\cosh t}{\cosh^{2}\left(\frac{\pi}{2}\sinh t\right)}\approx\exp\left(-\frac{\pi}{2}e^{|t|}\right)
\]
- This characteristic makes the DE transformation very effective in dealing with integral problems on infinite intervals.
3. **Optimality of the DE Formula**:
- By comparing with other transformation methods (such as \(\phi(t)=\tanh t\) and \(\phi(t)=\text{erf}(t)\)), it shows the advantages of the DE transformation in convergence speed.
- For some types of integrals, the DE formula can achieve a high precision with an approximate error of \(O(\exp(-C N / \log N))\), while other methods usually can only achieve a precision of \(O(\exp(-C\sqrt{N}))\).
4. **Extension of the Applications of the DE Transformation**:
- The paper also discusses the applications of the DE transformation in different types of integrals (such as Fourier - type integrals) and other numerical problems (such as initial - value problems of differential equations, boundary - value problems, etc.).
- It mentions the application of the DE - Sinc method in function approximation and shows its advantages over the traditional Sinc method.
In conclusion, this paper aims to comprehensively introduce the historical background, mathematical principles and wide applications of the DE transformation in numerical computation, emphasizing its superiority in improving computational efficiency and precision.