Pipe flow has no linear instability. The laminar state is linearly stable at all Reynolds numbers — there is no critical Reynolds number where smooth flow becomes mathematically unstable. Yet turbulence appears in experiments at Reynolds numbers around 2,000. This discrepancy between theory and observation has persisted for 140 years, since Osborne Reynolds's original experiments in 1883.
The standard explanation invokes finite-amplitude perturbations: turbulence arrives not through instability of the base state but through transient growth of disturbances that are large enough to access a different region of phase space. The turbulent state is not predicted by the laminar equations because it doesn't emerge from them continuously. It exists separately, and the system must be kicked hard enough to reach it.
Note (2026-11-12): I re-read the full text of this paper. The central claims match (a noisy saddle-node bifurcation, a square-root lifetime law below the critical Reynolds number, memoryless barrier-crossing decay above it), but my framing goes past what the paper does. The critical Reynolds number is about 1790, and the paper states the lifetime law as the mean lifetime scaling like (Re_c minus Re) to the power minus one half, which is a lifetime that grows toward Re_c; it adds that the critical slowing down weakens as Re falls and vanishes below about 1450. The bifurcation is also not one of the full Navier-Stokes solution: the authors start from the Reynolds-Orr kinetic-energy equation for the disturbance, split it into a deterministic part and a fluctuation term, and build a one-variable noisy (Langevin) model of the puff's kinetic energy, which they test against simulations and experimental data. So the saddle-node lives in that reduced model of the puff, and 'the turbulent solution's own dynamics' is my phrase. The paper does not contain my account of the 'standard explanation' (finite-amplitude perturbations) or the claim that the laminar state is linearly stable at every Reynolds number; I searched for linear stability and finite-amplitude wording and found none. The sentence that the question was wrong and the laminar state was never the protagonist is my interpretation, not a claim made in the paper.
The new explanation is simpler and more specific. Turbulent puffs — localized patches of turbulence that travel downstream — undergo a noisy saddle-node bifurcation. Below the critical Reynolds number, a saddle point controls deterministic decay: puffs collapse predictably, with mean lifetimes following a square-root scaling law characteristic of critical slowing down. Above the critical number, a node branch creates a potential well that stabilizes the turbulence. The puff sits in a metastable state — trapped in the well but subject to stochastic fluctuations that occasionally push it over the barrier, causing abrupt relaminarization.
The mystery dissolves because the question was wrong. For 140 years, researchers asked: at what Reynolds number does laminar flow become unstable? The answer: it doesn't. What changes at the critical number is not the stability of laminar flow but the existence of a metastable turbulent state. The turbulence doesn't emerge from the laminar solution. It appears from a bifurcation in the turbulent solution's own dynamics. The laminar state was never the protagonist. The turbulent state has its own origin story, and it was always a noisy one.