What does it mean for a linear approximation to be reliable?

[Dusty912]

Homework Statement

In regards to linearization of a nonlinear system in differential equations. What does it mean for a linear approximation to be reliable to describe the long term behavior of the non-linear system around the equilibrium point?

Homework Equations

jacobian matrix

The Attempt at a Solution

General question.

 

[soarce]

Linearized equations are "reliable" when the equations are linearly stable, i.e. the time dependent solution of the linearized system do not diverge from the nonlinear solution. The linearized solution won't capture all features of the nonlinear solution but at least it gives you a rough idea about the time evolution. This is equivalent to saying that the equations are linearly stable.

To study the linear stability you replace, roughly speaking, the nonlinear solution as following \Phi(x)\rightarrow\Phi_0(x) + \delta(x)e^{\lambda t} where $\Phi_0(x)$ is a time independent solution of the nonlinear equation. After plugging $\Phi_0(x)+\delta(x)e^{\lambda t}$ into the nonlinear equation one has to determine the eigenvalues $\lambda$. If $Re\{\lambda\}>0$ then perturbation $\delta$ will grow with time and the solution $\Phi_0(x)$ it is said to be linearly unstable.
Very important: keep in mind that the linear stability depends on the (is associated with a) time independent solution of the nonlinear equation. It may happened that a solution may be linearly stable while others not.
See for instance the wikipedia page.

However, the linear stability is a weak criteria when deciding whether a system is stable or not. This means that even if solution is linearly stable don't imply that it will follow the long time behavior of the nonlinear equation. Aside from the linear perturbations there are other types of perturbations which may set in and affect the time development. The stability chain is as following Energetic\: stability \Rightarrow Dynamical\: stability\Rightarrow Linear\: stability The linear stability is used to rule out the stability, is the system is not linearly stable then it won't be neither dynamical nor energetic stable. The energetic and dynamical stabilities are in general cumbersome to undertake, one should study the Hamiltonian structure and, something like, the Lyapunov stability (related directly to the time dependent evolution of the solution). They are performed only for simple nonlinear systems and solutions.

LE: You can follow, for instance, this notes as guide on linear stability analysis.