Source
Trefethen, Trefethen, Reddy & Driscoll, “Hydrodynamic Stability Without Eigenvalues”, Science 261 (1993), LIT-932, read in full in NOTE-839: eqs. 5–13, Table 1, Figs. 2 and 4–9.
What was actually shown
Computation on the linearized Navier–Stokes operator L about laminar plane Couette and plane Poiseuille flow, Fourier-transformed in x and z and discretized in y, with amplification measured in the energy norm.
- The anomaly. Eigenvalue analysis gives instability only above R = 5772 for Poiseuille flow and never for Couette flow. Transition is observed near R = 1000 and R = 350.
- Leading-order scalings as R → ∞ (Table 1).
- Transient growth sup_t ‖exp(−itL)‖: R/29.1 (Couette), R/71.5 (Poiseuille), reached at t = O(R).
- Maximum resonance over real frequencies: (R/8.12)² and (R/17.4)².
- Minimal destabilizing operator perturbation ε_min, the inverse of the maximum resonance: O(R⁻²) (Fig. 8). At R = 5000 a perturbation of norm 1.2 × 10⁻⁵ makes Poiseuille flow eigenvalue-unstable. For a normal operator with the same spectrum the growth would be at most 1 and the resonance O(R).
- The amplified structures are three-dimensional. The maxima occur at streamwise wavenumber α = 0, or nearly, and spanwise β between 1.18 and 2.04: streamwise vortices that become streaks. For two-dimensional perturbations, the traditional focus under Squire’s theorem, “some amplification can still occur, but it is far weaker”.
- The three quantities are linked. ‖L⁻¹‖ ≤ ∫‖exp(−itL)‖dt, within a factor of about 2 here, so the O(R²) resonance is O(R) growth over an O(R) time. The pseudospectra give a lower bound on transient growth that falls short by a factor of about 1.4.
The scalings could have come out bounded in R; they did not. Their lower bounds, transient growth at least cR, resonance at least cR², and a destabilizing perturbation of norm at most CR⁻², follow from the streamwise-independent scaling alone. That derivation is CLAIM-383, demonstrated and Active since 2026-10-19. The constants, and the fact that growth is no more than order R, are the computation’s. The paper credits the discovery of large transient growth to Gustavsson, Henningson et al. and Butler & Farrell, and its foundation to Butler & Farrell and Reddy & Henningson.
What this does not say
- It does not say this is how transition to turbulence happens. The paper’s last section opens “we speculate”. From a 2 × 2 model with energy-neutral quadratic mixing, whose threshold amplitude scales as R⁻³, it conjectures that transition in eigenvalue-stable shear flows is “essentially linear” with threshold O(R^γ), γ < −1. It lists how the Navier–Stokes equations differ from the model. The conjecture is not filed here.
- It does not say which mechanism operates in a given experiment. Resonance to vibration, growth from inlet disturbances, and destabilization by imperfect geometry are offered as possible triggers for distinct experiments, as speculation.
- It does not give operator perturbations a physical meaning. The authors leave open whether they describe transition or are “a symptom of the nonnormality of the unperturbed problem”.
- It is not about other flows. Pipe flow and boundary layers are named as cases where eigenvalue analysis fails, but are not computed.
- The figures are in the energy norm. Non-normality is norm-relative (THEORY-323).
Connections
No relations are declared.
- THEORY-323. The general account; this is its best-measured physical case.
- CLAIM-186. Its note says shear flows “turn turbulent because a transient amplification hands the disturbance to nonlinear dynamics”. This account holds the first half. The “because” is the conjecture this account excludes, and the claim should cite it as conjecture.