I am trying to prove an analytic result for gesodesic flows on negatively curved manifolds and I encountered the following dynamical-system porblem.

Let $B^n$ be $n$-dimensional balls and $h:B^{n-k}\times B^{k} \to U$ a $k$-dimensional foliation chart with $C^\infty$ leaves. i.e. For $x\in B^{n-k}$, $W_x := h(\{x\}\times B^k)\subset U$ are $C^\infty$ submanifolds. Furthermore let $y\in B^{n-k}$ such that $L=h(\{y\}\times B^{n-k})$ is a smooth transversal. Then according to Brin-Stuck "Introduction to Dynamical Systems" the foliation is absolutely continuous if there is a measurable family of positive measurable functions $\delta_x: W_x\to\mathbb R$ such that for all measurable sets $A\subset U$ $$\int_U \mathbf 1_A(x,y) dm_{euc} = \int_L \int_{W_x} \mathbf 1_A(x,y) \delta_x(y) dm_{W_x}(y)dm_L(x).$$ Here $dm_{euc}$ is the Eucildean measure on $U$ and $dm_{W_x}$ and $dm_L$ the measures obtained by restricting $dm_{euc}$ on the smooth submanifolds.

If the foliation ist the stable foliation of an Anosov differomorphism then I found that it is absolutely continuous. It think it is clear that one cannot hope that $\delta_x(y)$ depends $C^\infty$ on $x$ unless the foliation is smooth.

My question is: Under which conditions on the foliation can one obtain, that for any $x$ one has $\delta_x \in C^\infty(W_x)$(ideally I would also like to have, that $\delta_x$ depends continuousely on $x$ w.r.t. to the $C^\infty$-topology, but maybe I can also handle the case if it is only measureable)? And more precisely: Is such a statement always true for the strong stable foliation of a $C^\infty$ Anosov flow. If it doesn't hold for general Anosov Flows: What are the weakest known assumption that it holds? (If I understand it correctly then $C^1$ foliations implies this statement via Fubinis theorem).


In general, for the geodesic flow of a compact negatively curved manifold for example, the strong stable foliation is Holder continuous; in particular, even if each leaf is (locally) a smooth manifold, transversally, the foliation (and therefore the map $x\to\delta_x$) is no more than Hölder continuous.

In the case of geodesic flows in constant negative curvature, it is smooth, and if I remember well, it is $C^1$ in the case of surfaces.

I guess that for general Anosov flows, it is similar: you cannot expect better regularity than Hölder, except in very particular cases.

  • $\begingroup$ Dear Barbara, thanks for your reply. I understand, that foliations are in general only Hölder and that I cannot expect $\delta_x(y)$ to be smooth in $x$. My question is rather if for given $x$ the conditional measure is smooth along the sheat $W_x$ $\endgroup$ – twch May 19 '15 at 21:05

Due to helpfull discussions with several dynamical systems experts I am now able to give an answer to the question, that I asked about one year ago and provide some references:

The short answer is, that for the smoothness of the conditional measures it is sufficiant to know the smoothness of the holonomy maps with respect to a smooth transversal foliation. A detailed explanation of this argument can for example be found in Appendix A of arxiv-version but the techniques are rather standard, see e.g. Brin-Stuck "Introduction to dynamical systems" Section 6.2.

Concerning the stable and unstable foliations of an $C^\infty$-Anosov flow, a statement on the smoothness of the holonomy maps can for example found in Giulietti-Liverani-Pollicott Ann.Math (2013) (Appendix E). Thus one obtains the smoothness of the conditional densities for $C^\infty$-Anosov flows. A precise formulation of this statement is now given in Theorem 7 arxiv-version.

If one is interested in an analogouse statement for Anosov map, there is a remark in Chernov "Invariant measures for hyperbolic dynamical systems" (Handbook of Dynamical systems 1A), stating the smoothness of the conditional densities and giving a rough idea how to prove it.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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