This research is supported by NASA's Fluid Physics Program

Nonlinear Instabilities

Systems described by nonnormal evolution operators (operators with non-orthogonal eigenfunctions) often display rather surprising dynamics. For instance, turbulence in shear flows often develops for Reynolds numbers where the basic laminar flow is still linearly stable. The critical Reynolds number at which the transition occurs depends rather sensitively on the geometry of the system and the roughness of the boundaries. The onset of turbulence can be linked to a nonlinear (or finite amplitude) instability arising due to the interaction between the nonlinearity in the Navier-Stokes equation and strong transient amplification of disturbances caused by the nonnormality of its linearization about the laminar flow with strong shear. The idea of a nonlinear instability can also be used to explain the disagreement between the predictions of linear stability analysis and experimental data for the contact line instability in driven spreading of thin liquid films and drift wave instabilities in plasma. Thin films represent a particularly nice system for studying the effects of transient amplification due to easy experimental accessibility (click here to see some movies) and the availability of a simple analytical description in terms of the lubrication theory.

Another application of the theory of nonlinear instabilities is to control of spatially extended systems. Nonnormality, it turns out, can also arise in the absence of mean flow as a result of localized feedback control. For instance, the dynamics of all scalar reaction-diffusion equations in one spatial dimension, controlled at one or both boundaries, displays pronounced transients (see, e.g., figure below), characteristic of strong nonnormality.

Transient amplification leads to extreme sensitivity of the dynamics to noise, which increases exponentially fast with the size of the system. As a result, localized control of a system with large spatial extent can fail for a very small level of noise. This failure can be explained using the idea of a nonlinear instability: nonlinear terms distort the action of linear control, producing a positive feedback loop amplifying stochastic disturbances.



Go back to Home Page