Asymptotic analysis of numerical steepest descent with path approximations (Q604683): Difference between revisions
From MaRDI portal
Added link to MaRDI item. |
Removed claim: reviewed by (P1447): Item:Q278551 |
||
Property / reviewed by | |||
Property / reviewed by: Denis Nikolaevich Sidorov / rank | |||
Revision as of 12:54, 12 February 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Asymptotic analysis of numerical steepest descent with path approximations |
scientific article |
Statements
Asymptotic analysis of numerical steepest descent with path approximations (English)
0 references
12 November 2010
0 references
The authors study Fourier-type highly oscillatory integrals \[ I[f] = \int_{-1}^1 f(x) e^{i\omega g(x)} \,dx, \] where \(\omega\) is a large parameter, \(f(x)\), \(g(x)\) are smooth functions. The numerical method of steepest descent for oscillatory integrals using a low-cost explicit polynomial approximation of the paths of steepest descent is proposed. A disadvantage of this approach is that the method is asymptotic. It is to be noted that in many cases it is sufficient to have only a rough approximation of the exact steepest-descent path. The proposed scheme requires the computation of derivatives, which may not always be available. This is the case for most numerical schemes for oscillatory integrals with high asymptotic order, because the asymptotic behaviour of the integral depends precisely on those derivatives. However, when aiming for a high asymptotic order, it is sufficient to know the derivatives only approximately, as long as the accuracy of the approximation scales with \(\omega\) in a suitable fashion. Results regarding the asymptotic error behaviour are formulated in a theorem and have confirmed using the numerical experiments.
0 references
oscillatory quadrature
0 references
Gauss-Laguerre quadrature
0 references
Fourier-type highly oscillatory integrals
0 references
method of steepest descent
0 references
asymptotic error behaviour
0 references
numerical experiments
0 references