Abstract:We discuss a phenomenological method which allows to determine the singular asymptotic behaviours for a pure fluid at equilibrium, when the liquid-gas critical point and the tangent plane to the characteristic surface of this point are localized experimentally, in the pressure, density and temperature coordinates.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: how to numerically approximate the Fourier sine and cosine transforms efficiently and accurately. Specifically, the paper focuses on integrals of the following forms:
\[
f_c(t) = \int_0^\infty f(x) \cos(tx) \, dx,
\]
\[
f_s(t) = \int_0^\infty f(x) \sin(tx) \, dx.
\]
Although there are extensive Fourier transform tables, numerical approximation of the Fourier transform is still required in practical applications. This paper proposes a new method to achieve efficient numerical calculation of the Fourier sine transform \(f_s(t)\) by converting the integral interval from \((0, +\infty)\) to \((- \infty, +\infty)\) and then using the trapezoidal rule to approximate the converted integral.
### Main Contributions
1. **New Method**: The paper introduces a novel single - exponential transformation and combines it with the trapezoidal rule to achieve efficient numerical approximation of the Fourier sine and cosine transforms.
2. **Error Analysis**: Through the method of contour integration in the complex plane, the discretization error and the truncation error are analyzed in detail, and error estimates are given.
3. **Comparison and Improvement**: A comparison is made with the existing double - exponential transformation method, showing the characteristics and advantages of the new method.
### Method Overview
- **Transformation**: A specific variable substitution \(x=\frac{m}{t}\phi(u)\) is introduced, where \(\phi(u)\) is a function mapping from \((- \infty, +\infty)\) to \((0, +\infty)\).
- **Trapezoidal Rule**: The trapezoidal rule is applied on the new integral interval to obtain the approximate formula:
\[
T_m=\frac{\pi}{m}\sum_{k = - \infty}^{\infty}F_m\left(k\frac{\pi}{m}\right),
\]
where \(F_m(u)=f\left(\frac{m}{t}\phi(u)\right)\sin(m\phi(u))\frac{m}{t}\phi'(u)\).
- **Error Control**: By choosing appropriate parameters \(m\) and the number of truncation terms \(n\), the discretization error and the truncation error can be effectively controlled, thus obtaining high - precision results.
### Conclusion
The paper verifies the effectiveness of the new method through detailed theoretical analysis and numerical experiments, especially when dealing with functions with singularities and slow - decay characteristics. Compared with the existing double - exponential transformation method, although the convergence speed is slightly slower, the new method has a simpler implementation and higher stability.
In general, this paper provides a new and efficient tool for the numerical calculation of the Fourier transform, especially suitable for scientific and engineering fields requiring high - precision calculations.