Paper detail

A computationally efficient approach for the removal of the phase shift singularity in harmonic resolvent analysis

The recently introduced harmonic resolvent framework is concerned with the study of the input-output dynamics of nonlinear flows in the proximity of a known time-periodic orbit. These dynamics are governed by the harmonic resolvent operator, which is a linear operator in the frequency domain whose singular value decomposition sheds light on the dominant input-output structures of the flow. Although the harmonic resolvent is a mathematically well-defined operator, the numerical computation of its singular value decomposition requires inverting a matrix that becomes exactly singular as the periodic orbit approaches an exact solution of the nonlinear governing equations. The very poor condition properties of this matrix hinder the convergence of classical Krylov solvers, even in the presence of preconditioners, thereby increasing the computational cost required to perform the harmonic resolvent analysis. In this paper we show that a suitable augmentation of the (nearly) singular matrix removes the singularity, and we provide a lower bound for the smallest singular value of the augmented matrix. We also show that the desired decomposition of the harmonic resolvent can be computed using the augmented matrix, whose improved condition properties lead to a significant speedup in the convergence of classical iterative solvers. We demonstrate this simple, yet effective, computational procedure on the Kuramoto-Sivashinsky equation in the proximity of an unstable time-periodic orbit.