Quadrature methods for highly oscillatory linear and nonlinear systems of ordinary differential equations. I (Q1014894)

Consider the highly oscillatory system of ordinary differential equations (ODEs) \[ y'=A_{\omega }y+f(t,y),\quad y(0)=y_o\in{\mathbb{R}}^d,\;t\geq 0, \] where \(A_{\omega}\) is a constant non-singular \(d\times d\) matrix with large eigenvalues, \(\sigma (A_{\omega })\subset\) i\({\mathbf{R}}\), \(\parallel A_{\omega}\parallel\gg 1,\omega\gg 1\) is a real parameter and \(f:\mathbb{R}\times{\mathbb{R}}^d\rightarrow{\mathbb{R}}^d\) is a smooth vector-valued function. Its implicit solution has the form \[ y(t)=e^{tA_{\omega}}y_o+ \int_0^te^{(t-\tau )A_{\omega}}f(\tau,y(\tau ))d\tau=e^{tA_{\omega}}y_o+I[f]. \] The paper deals with quadrature methods for a numerical approximation of these linear and nonlinear systems of \(ODE\)s. First, approximations of highly oscillatory integrals of the form \(I[f]=\int_a^bf(x)e^{i\omega g(x)}dx\) are considered, where \(f,g\in\mathcal{C}^{\infty}\) are smooth, \(g\) is strictly monotone in \([a,b],a\leq x\leq b\) and the frequency \(\omega\gg 1\). The author shows how an approximate selection of quadrature rule improves the accuracy of the approximation with the growing frequency of oscillation. Then, asymptotic and Filon-type methods are derived for highly oscillatory linear systems of \(ODE\)s. Waveform relaxation methods with Filon-type quadrature are used for the approximation of highly oscillating nonlinear systems. Numerical examples are supplied.