Self-Replication of Turbulent Puffs: On the edge between chaotic saddles

Abstract: Pipe flow is a canonical example of a subcritical flow, the transition to turbulence requiring a finite perturbation. The Reynolds number ($Re$) serves as the control parameter for this transition, going from the ordered (laminar) to the chaotic (turbulent) phase with increasing $Re$. Just above the critical $Re$, where turbulence can be sustained indefinitely, turbulence spreads via the self-replication of localized turbulent structures called puffs. To reveal the workings behind this process, we consider transitions between one and two-puff states, which dynamically are transitions between two distinct chaotic saddles. We use direct numerical simulations to explore the phase space boundary between these saddles, adapting a bisection algorithm to identify an attracting state on the boundary, termed an edge state. At $Re = 2200$, we also examine spontaneous transitions between the two saddles, demonstrating the relevance of the found edge state to puff self-replication. Our analysis reveals that the process of self-replication follows a previously proposed splitting mechanism, with the found edge state as its tipping point. Additionally, we report results for lower values of $Re$, where the bisection algorithm yields a different type of edge state. As we cannot directly observe splits at this $Re$, the self-replication mechanism here remains an open question. Our analysis suggests how this question could be addressed in future studies, and paves the way to probing the turbulence proliferation mechanism in other subcritical flows.