Fast, numerically stable computation of oscillatory integrals with stationary points

From MaRDI portal
Publication:2379359





The author derives a new numerically stable method for the approximation of the oscillatory integral \(\int_{-1}^1 f(x) \exp(i \omega g(x)) dx\) where \(f \in C^1[-1,1]\), \(g \in C^r[-1,1]\), and \(g\) may have a single stationary point of order \(r-1\) at the origin. The algorithm does not require the evaluation of \(f'\). The computational cost of the method decreases as the frequency \(\omega\) increases. The location of the nodes is independent of the frequency, hence the function evaluations of \(f\) can be reused if the same function \(f\) is to be integrated with different values of \(\omega\). In the case when \(g\) has no stationary points, the method can be interpreted as a version of the Levin collocation scheme. A detailed analysis of the method is given with a statement of its strengths and weaknesses. As a numerical example, the author computes values of certain Bessel functions.



Cites work


Cited in
(38)


Describes a project that uses

Uses Software






This page was built for publication: Fast, numerically stable computation of oscillatory integrals with stationary points

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2379359)