1
$\begingroup$

Suppose $(X,\mathcal A,\mu,T)$ is a finite measure-preserving system. Then we define a new measure system $(X^{(K)},\mathcal A^{(K)},\mu^{(K)},T^{(K)})$ defined by $X^{(K)}=X\times \{1,2,...,K\}$ for any positive integer $K$, $\mu^{(K)}(A\times \{k\})=\dfrac{\mu(A)}{K}$ for $1\leq k\leq K$ and $A\in\mathcal A$, and define $T^{(K)}:X^{(K)}\to X^{(K)}$ by $T^{(K)}(x,n)=(x,n+1)$ if $n<K$ and $(Tx,0)$ if $n=K$.

Then one can check that even if $T$ is ergodic, $T^{(K)}$ is not totally ergodic.

Now suppose we have a finite measure preserving system $(Y,\mathcal C,\nu,S)$ which is not totally ergodic.

Question: Find if possible a measure system $(X,\mathcal A,\mu,T)$ such that $S$ is measurably isomorphic to the system $(X^{(K)},\mathcal A^{(K)},\mu^{(K)},T^{(K)})$ for some positive integer $K$.

I get it that my $K$ will be the least positive integer $n$ for which $S^n$ is NOT ergodic. But I cannot define my $X$. Is it true that we will always get such an $X$?

$\endgroup$
2
$\begingroup$

Assume $(Y,\nu,S)$ is ergodic (but not totally ergodic), let $T = S^K$, where $K$ is the least positive integer $n$ for which $S^n$ is not ergodic, and let $Z$ be an ergodic component of $(Y,\nu,T)$. Since $S$ is ergodic, the ergodic decomposition of $(Y,\nu,T)$ is $Y = Z \cup S Z \cup \cdots \cup S^{K-1} Z$. So it is easy to see that $(Y,\nu,S)$ is isomorphic to $(Z^{(K)}, \mu^{(K)}, T^{(K)})$.

$\endgroup$
1
  • $\begingroup$ It's worth adding that this is not unique. There may be larger $n$ values for which the decomposition becomes finer. $\endgroup$ Jun 30 '16 at 19:48

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.