MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $T_t$ be the geodesic flow on a surface $S$ of constant negative curvature, and let $M(f,t) := \langle \bar f \cdot (f \circ T_t) \rangle$, where $\langle f \rangle := \int_S f(x) d\mu(x)$ and where $\mu$ is the natural invariant measure.

A 1984 paper by Collet, Epstein, and Gallavotti (PDF here) shows (prop. 5, p. 90) that for $f$ nice (in a sense defined at the bottom of page 71), $\lvert M(f,t) \rvert < \lVert f \rVert^2_\xi C( \xi) \cdot t^{b(\xi)}\exp(-t/2)$, where $\lVert \cdot \rVert^2_\xi$ is a certain rather complicated norm (defined in equation 3.5 of the paper) and $C$, $b$ do not depend on $f$.

I have two related questions about this result which hopefully someone here already knows (the paper is quite technical and I really don't need to know its details if I can get a bit of clarification here):

  • This result seems to imply that the rate of mixing is 1/2. How can this be? (see also this question)

  • How does this result (for which the "decay is not exponential") square with the results of Dolgopyat and Liverani that give exponential decay of correlations for reasonably nice Anosov flows?

share|cite|improve this question
In the first paragraph do you mean to say $M(f,t):=\langle \bar f \cdot(f \circ T_{t}) \rangle$? because that's how you use it in the second paragraph. In other words, what is A$? – drbobmeister Dec 28 '10 at 7:27
up vote 5 down vote accepted

Dear Steve,

In general the "precise" rate of mixing depends on the class of functions one considers. In particular, the authors of the paper you quoted treat only "analytic" functions (with some fixed band of analyticity) and they show that you get a rate of $t^{b(\xi)}e^{-t/2}$. In this sense, the rate of mixing is exponential and essentially equal to 1/2: I said essentially because of the factor $t^{b(\xi)}$ can't be removed in general, so that's why the authors said that the rate of mixing is not "genuinely exponential" (i.e., the sharp bound that they obtain is not exactly an exponential function of time).

On the other hand, the results of Dolgopyat and Liverani concerns smooth but non-analytic functions and contact Anosov flows (and not only the ones coming from constant negative curvature surfaces). In this context, they show "exponential mixing" in the sense that the correlations can be bounded by some exponential function of $t$ (say $e^{-\sigma t}$ for some $\sigma>0$). Of course, they don't know whether their "rate" (i.e., the number $\sigma$ they obtain in the end of the calculations) is optimal (for instance, there is a subsequent work of Tsujii improving on Liverani's work). In particular, although Dolgopyat and Liverani obtain "exponential mixing", it is not in the same "sense" of Collet, Epstein and Gallavotti.

In resume, I guess that the confusion comes from the distinct employments of the nomenclature "exponential mixing" in the works of Collet, Epstein, Gallavotti, and Dolgopyat and Liverani (and Tsujii also).



share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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