2
$\begingroup$

Let $T\colon R^n\to R^n$ be a linear map. If we want to study the behavior of $T^kx$ for some $x\in R^n$ as integer $k$ grows, we usually look at the eigen structure of $T$.

Now let $S\colon R^n\to R^n$ be a linear map plus a nonlinear perturbation. And I want to study the behavior of $S^kx$ for some $x\in R^n$ as integer $k$ grows. I am wondering if there exists a theory that discusses this kind of problem.

$\endgroup$

2 Answers 2

1
$\begingroup$

Yes, there is such a theory, and you already gave yourself the answer in the tag. For particular shape of your map $S$, i.e. a linear hyperbolic map (meaning: the spectrum is disjoint from the unit circle) $T$ plus a small (in a sense to be precised) Lipschitz perturbation , the Hartman-Grobman tells you that $S$ is conjugated with $T$ by a Hölder continuous homeomorphism. The proof is very elementary and follows from the contraction principle; see e.g. M.Shub's book, Global stability of Dynamical Systems.

$\endgroup$
0
$\begingroup$

Suppose there is a linear $T$ with $\| Sx - Tx \| = o(\|x\|)$ as $x \to 0$, and all eigenvalues of $T$ have absolute value $< 1$. Then there exist positive integer $N$ and positive reals $\delta$ and $\epsilon$ such that for $\|x\| < \delta$, $\|S^N x\| \le (1 - \epsilon) \|x\|$. It follows that for $\|x\|$ sufficiently small, $S^k x \to 0$ as $k \to \infty$. Was that the sort of result you were looking for?

$\endgroup$
2
  • $\begingroup$ The nonlinear map $S$ that I am interested in is equal to a linear map $T$ plus a polynomial in $x$. I want to know for what kind of $x$ $S^kx$ will converge as $k$ grows and where it will converge to. $\endgroup$
    – silkrain
    Apr 17, 2011 at 16:04
  • $\begingroup$ The general theory (and in particular the Hartman-Grobman theorem, as mentioned by Pietro) can tell you about what happens in a neighbourhood of 0. So if $T$ has $k$ eigenvalues (counted by algebraic multiplicity) with absolute value $< 1$ and none with absolute value $=1$, there will be a $k$-dimensional manifold near 0 on which the iterates will converge to 0. If there are fixed points other than 0, the linearizations around those fixed points will tell you about convergence to them. $\endgroup$ Apr 17, 2011 at 18:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.