6
$\begingroup$

What is an example of a compact manifold which does not admit a diffeomorphism with at least one dense orbit?

Moreover, is it true to say that every isometry of $\mathbb{C}P^n$ with the Fubini-Study metric do not possess any dense orbit?

$\endgroup$
4
  • 6
    $\begingroup$ No isometry of $\mathbb{CP}^n$ has a dense orbit, because the square of such an isometry is holomorphic and so can be regarded as an element of the compact Lie group $\mathrm{SU}(n{+}1)$ and hence is congugate to an element of the maximal torus, and no element of a maximal torus can have a dense orbit because it has to preserve a transverse collection of hyperplanes and so perserve the distances from these hyperplanes. A little more work takes care of the case of an anti-holomorphic isometry. $\endgroup$ Apr 25, 2018 at 22:10
  • 4
    $\begingroup$ For $C^1$ diffeomorphisms preserving volume, there is a residual set which is transitive (having a dense orbit). See Theorem 1.3 of mathscinet.ams.org/mathscinet-getitem?mr=2090361 So in some sense such diffeomorphisms are "generic". See also the comments after the statement of the theorem about the homeomorphism case. $\endgroup$
    – Ian Agol
    Apr 25, 2018 at 23:02
  • $\begingroup$ @RobertBryant Thank you very much for your answer. I try to understand its detail for $n>1$. My motivation was $S^2$ whose all isometries are non ergodic maps. $\endgroup$ Apr 26, 2018 at 14:35
  • $\begingroup$ My apology, I correct: The isometries are ergodic and without dense orbit. $\endgroup$ Apr 26, 2018 at 15:03

1 Answer 1

9
$\begingroup$

Dolgopyat and Pesin proved (ETDS 2002) that "every compact manifold of dimension $\geq 2$ admits a Bernoulli diffeomorphism with non-zero Lyapunov exponents". That is, on any such manifold $M$ there is a $C^\infty$ diffeomorphism $f$ that preserves volume $m$ and has the property that $(M,m,f)$ is a measure-preserving transformation that is measure-theoretically isomorphic to a Bernoulli shift. In particular, by the Birkhoff ergodic theorem, $m$-a.e. point $x\in M$ has the property that its orbit equidistributes with respect to $m$, meaning that for every continuous $\phi\colon M\to \mathbb{R}$, we have $\frac 1n \sum_{k=0}^{n-1} \phi(f^kx) \to \int \phi\,dm$ as $n\to\infty$. In particular, any such $x$ has a dense orbit.

$\endgroup$
0

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.