What is the relation between topological entropy and Lyapunov exponent?

325 Views Asked by At

In this post, topological entropy is defined. In this book, the notion of Lyapunov exponent for, let's say a self-diffeomorphism from a smooth, closed Riemannian manifold to itself, is defined (in terms of the eigenvalues/singular values of the Jacobian matrix).

How are they related? Is the topological entropy of a self-diffeo $T: M \to M$ simply the maximum Lyapunov exponent over all $x \in M$?

Thanks in advance.

UPDATE: Is this true: The topological entropy of $T = \tilde{\Phi}_{t=1}$, and therefore $\xi$, is the sup over all measures on $M$ that $T$ leaves invariant of the integral over $M$ of the sum of the positive Lyapunov exponents of $T$, $\displaystyle \sup_{\nu \in \text{ Invar}(T)} \int_{M} \sum_{i=1}^{k}\lambda_k(x,v)\ d\nu$?

1

There are 1 best solutions below

0
On

Trying to understand what your notation means, in general you only have $$ h(f)\le \sup_{\nu} \int_{M} \sum_{i=1}^{k(x)}\lambda_{k(x)}(x)\ d\nu(x) $$ (really it is a bad idea to use $h$ for the map). This follows from the Margulis–Ruelle inequality.

The supremum runs over all $f$-invariant probability measures and the sum is over all positive Lyapunov exponents $\lambda_1(x)\le\cdots \le\lambda_{k(x)}(x)$ (unfortunately $k(x)$ may indeed depend on $x$).

But if $f$ is a $C^{1+\alpha}$ diffeomorphism and $\nu$ is absolutely continuous with respect to the volume, then $$ h_\nu(f)= \int_{M} \sum_{i=1}^{k(x)}\lambda_{k(x)}(x)\ d\nu(x). $$ This is Pesin's entropy formula. Therefore, under the same assumptions, $$ h(f)\ge \int_{M} \sum_{i=1}^{k(x)}\lambda_{k(x)}(x)\ d\nu(x), $$ although you cannot add a supremum in general. However, if $\nu$ is also a measure of maximal entropy you obtain $$ h(f)= \int_{M} \sum_{i=1}^{k(x)}\lambda_{k(x)}(x)\ d\nu(x) $$ and so also $$ h(f)= \sup_{\nu} \int_{M} \sum_{i=1}^{k(x)}\lambda_{k(x)}(x)\ d\nu(x). $$

Of course, all depends exactly on what you are assuming. Up to some technicalities on the $C^{1+\alpha}$ assumption this is the most general answer that you can get under general assumptions.