Phase function methods for second order inhomogeneous linear ordinary differential equations (Q6184278): Difference between revisions

The paper introduces a numerical method for solving second order inhomogeneous linear ordinary differential equations of the form \[ y^{\prime \prime }(t)+q(t)y(t)=f(t),\qquad a<t<b \] whose solutions are highly oscillatory when \(q\) is positive and of large magnitude; for this reason standard numerical algorithms perform poorly. The paper takes under consideration the case in which \(f\) is real-valued and slowly-varying and \(q\) is positive and slowly-varying and introduces a numerical method which, remarkably, exhibits high-accuracy and runs in time independent of frequency. The algorithm works by building a nonoscillatory phase function representing the solution of the corresponding homogeneous equation. Then, a set of auxiliary nonoscillatory functions are built, which, together with the phase function, allow the inhomogeneous equations to be solved subject to essentially any reasonable boundary conditions. Numerical examples are reported to enlighten the properties of the proposed algorithm.