On $\mathbb{T}^n$ with a Riemannian metric, the stable norm is defined as $$\Vert h\Vert=\inf \sum |r_i| \cdot \mathrm{length}(\sigma_i),$$

where $h\in H_1(\mathbb{T}^n,\mathbb R)$ and $\sum_i r_i\sigma_i$ is a Lipschitz cycle representing $h$. The $\inf$ is taken over all such cycles.

The question is, if we consider

(1)$h$ is completely rational, i.e. there are n-1 linearly independent integer vectors perpendicular to $h$,

(2) The $\inf$ is achieved on a set supporting a measure that is uniquely ergodic. (This excludes Hedlund's counterexample)

then is it correct that the $\inf$ is always achieved by a cycle, i.e. the measure is supported on a closed orbit?

This is true in 2-dim case, but not sure about the high dim case.


Let $\mathbb T^n=\mathbb R^n\pmod {2{\cdot}\pi}$. Consider the embedding $f\colon\mathbb T^2\hookrightarrow \mathbb T^3$ defined as $$f\colon(u,v)\mapsto (\sin u,\cos u, v).$$

It is easy to find a metric on $\mathbb T^3$ so that for $h=(0,0,1)$, the infimum is achieved on a geodesic which runs in the embedded $\mathbb T^2$ in an irrational direction.

Is that what you wanted?

  • $\begingroup$ Thanks very much. This is a nice example. I am also trying to exclude the possibility of Denjoy minimal set, i.e. if we take a section to the measure, we get a Cantor set. In general, orbits in Denjoy minimal set has hyperbolicity, so it is different from your example. Is it possible to have a Denjoy minimal set as support of measure with rational rotation vector? $\endgroup$ – John Galt Aug 13 '13 at 14:39

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.