Orr–Sommerfeld equation and transient growth

Nondimensionalize a steady parallel channel flow using a specified speed and length, so $\mathrm{Re}=UL/\nu$. For a two-dimensional streamfunction mode $\phi(y)e^{i\alpha(x-ct)}$, eliminate the pressure from the linearized viscous equations. With $D=d/dy$, the result is the Orr–Sommerfeld equation. The viscous term contains four derivatives of the disturbance profile. It therefore cannot generally be appended as a small algebraic damping correction to an inviscid eigenvalue while retaining the old boundary data.

At a stationary rigid wall, no penetration requires $\phi=0$ for nonzero $\alpha$, while no slip requires $D\phi=0$. Two walls provide four homogeneous conditions. In an unbounded problem, decay conditions replace these wall conditions. Changing the Reynolds-number length convention also changes quoted numerical critical values, so a number without its normalization is ambiguous. For three-dimensional disturbances, the wall-normal vorticity dynamics must also be included through the coupled Orr–Sommerfeld–Squire formulation.

Represent a discretized initial-value problem as $\dot{\mathbf q}=L\mathbf q$. If L is normal in the physically chosen energy inner product, orthogonal eigenvectors make the eigenvalue picture particularly direct. If it is non-normal, its eigenvectors can be strongly nonorthogonal. Contributions initially canceling in one component can decay at different rates and stop canceling later. Their sum can grow temporarily even while each eigenmode eventually decays. The energy norm and the allowed initial conditions are part of the question.

A transparent example has matrix entries $L_{11}=L_{22}=-1$, $L_{12}=K$, and $L_{21}=0$. Both eigenvalues are negative, but initial data $(0,1)$ produce a first component $Kt e^{-t}$. For sufficiently large K, the Euclidean norm grows far above its initial value before decaying. The repeated-eigenvalue example is deliberately simple; the same general finite-time issue occurs for nonorthogonal eigenvectors with distinct eigenvalues. It is the propagator’s singular values in an energy norm, rather than eigenvalues alone, that quantify optimal gain.

Shear flows can convert cross-stream motion into streamwise velocity variations by displacing fluid across the base gradient. This lift-up mechanism is one route to strong finite-time amplification. In other settings, disturbance tilting and pressure-mediated effects matter. A finite-time optimal perturbation depends on the target time and norm; the direction that maximizes one observable may not maximize another. Numerical calculations also require convergence tests, since a poorly resolved operator can manufacture sensitive eigenvalues or distorted optimal gains.

Transient growth is neither a proof that a particular experiment becomes turbulent nor a way to ignore nonlinear effects. It can bring a sufficiently large initial disturbance into a regime where nonlinear feedback matters, creating a possible route to transition. Whether that route sustains motion depends on the full dynamics and disturbance environment. Conversely, a linearly unstable mode can grow without immediately creating fully developed turbulence. Report modal instability, transient amplification, and observed nonlinear transition as distinct pieces of evidence.