How are eigenvalues related to linear stability analysis?

In stability studies in fluid mechanics, the following two premises are usually taken for granted:

While both principles are well justified for Newtonian flows, the situation is much more uncertain for viscoelastic flows, while involve hyperbolic partial differential equations. Indeed, abstract counterexamples where linear stability is not determined by the spectrum have been known since the 1950s, and our recent work has shown that this happens for quite natural problems such as a lower order perturbation of the wave equation. In Renardy's paper, a positive result is obtained for flows of viscoelastic flows with constitutive laws of Jeffreys type, under the assumption that the Weissenberg number is sufficiently small. In this case, it is shown that both 1 and 2 hold. In Renardy's paper, the abstract issue of spectrally determined growth is investigated from a new angle. While in general the growth rate of solutions is not determined from the spectrum, one may ask whether spectrally determined growth is the rule or the exception. One may thus consider a family of evolution problems du/dt=(A+B)u, where B varies over a class of operators, say all bounded operators. The result is that, if A is the infinitesimal generator of a C_0-semigroup in a Hilbert space, then the set of all operators B for which spectrally determined growth fails is of first category.


Return to Research Group homepage