Title: Stability implies robust convergence of a class of preconditioned parallel-in-time iterative algorithms

Abstract: Solving evolutionary equations in a parallel-in-time manner is an attractive topic and many algorithms are proposed in recent two decades. The algorithm based on the block $\alpha$-circulant preconditioning technique has shown promising advantages, especially for wave propagation problems. By fast Fourier transform for factorizing the involved circulant matrices, the preconditioned iteration can be computed efficiently via the so-called diagonalization technique, which yields a direct parallel implementation across all time levels. In recent years, considerable efforts have been devoted to exploring the convergence of the preconditioned iteration by studying the spectral radius of the iteration matrix, and this leads to many case-by-case studies depending on the used time-integrator. In this paper, we propose a unified convergence analysis for the algorithm applied to $u'+Au=f$, where $\sigma(A)\subset\mathbb{C}^+$ with $\sigma(A)$ being the spectrum of $A\in\mathbb{C}^{m\times m}$. For any one-step method (such as the Runge-Kutta methods) with stability function $\mathcal{R}(z)$, we prove that the decay rate of the global error is bounded by $\alpha/(1-\alpha)$, provided the method is stable, i.e., $\max_{\lambda\in\sigma(A)}|\mathcal{R}(\Delta t\lambda)|\leq1$. For any linear multistep method, such a bound becomes $c\alpha/(1-c\alpha)$, where $c\geq1$ is a constant specified by the multistep method itself. Our proof only relies on the stability of the time-integrator and the estimate is independent of the step size $\Delta t$ and the spectrum $\sigma(A)$.