Torus translation is ergodic if and only if the components of the translation vector are rationally independent.

972 Views Asked by At

I'm reading Ergodic Theory and Differential Dynamics by Ricardo Mane. There is a theorem in the book that states the following: If x $\in$ $R^n$, the translation L $_{\pi(x)}$: $T^n \rightarrow T^n$ is ergodic if and only if (k,x) $\notin$ Z for every k $\in Z^n$.

I was hoping someone could either give me or direct me to, a comprehensible proof of this result, as Mane's does not give very much explanation.

1

There are 1 best solutions below

2
On BEST ANSWER

Actually this is not quite right: you need to exclude $k = 0$.

If $(k,x) \in \mathbb Z$ and $k \ne 0$, then $f(t) = \exp\left(2\pi i (k, t)\right)$ is invariant under $L_{\pi(x)}$ and is not a.e. constant.

Conversely, if $f \in L^1(\mathbb T^n)$ is invariant under $T_{\pi(x)}$, look at the Fourier transform of $f$: $\widehat{f}(k) = \int_{\mathbb T^m} \exp(-2\pi i (k,t)) f(t)\; dt$ (where $dt$ is normalized $m$-dimensional Lebesgue measure on $\mathbb T^m$). Since $\widehat{L_{\pi(x)} f}(k) = \exp(2\pi i (k,x)) \widehat{f}(k)$, we must have $(k,x) \in \mathbb Z$ for any $k$ such that $\widehat{f}(k) \ne 0$. If there are no such $k$ except $0$, then (by the fact that Fourier transform is one-to-one on $L^1$) $f$ is a.e. constant.