Temporal homogenization of linear ODEs, with applications to parametric super-resonance and energy harvest (Q261300)

The paper is concerned with the time-dependent non-homogeneous linear ordinary differential equation \[ \dot x=Ax+\varepsilon P(t)x+f(t)\leqno(1) \] on \(\mathbb{R}^n\), where \(\varepsilon\) is small, \(A\) is a constant \(n\times n\) real matrix, \(P:\mathbb{R}\to {\mathcal M}_n(\mathbb{R})\) is a square-integrable periodic function with period \(2\pi/\omega\), \(f(t)\) is a vector-valued function with \(e^{-At}f(t)\) integrable on \([0,\hat C\varepsilon^{-1}]\) for some \(\hat C>0\) and \(0<\varepsilon\ll 1\). By using a \(2\)-scale expansion approach, the authors obtain the long-time approximation of solutions for equation \((1)\) over a \({\mathcal O}(\varepsilon^{-1})\) timescale, without resolving oscillations of \(P(t)\) over that interval of time. Then they apply this method of approximation to control the oscillation amplitude of a damped oscillator and couple oscillators for minimizing the threshold on fluctuation amplitude needed for harvesting energy.