If $\lambda, \delta \in \widehat{G}$, where $G$ is compact abelian and $A$ is ergodic, then $(U_A^n \lambda, \delta) = 0$ eventually

101 Views Asked by At

I'm going through Peter Walters's Introduction to Ergodic Theory (pg. 50) and I'm having trouble understanding the proof of Theorem 1.28, which says that for an endomorphism of a compact group strong mixing, weak mixing and ergodicity are equivalent.

In particular, the proof starts off with the assumption $G$ is abelian. It then says it suffices to show that if the endomorphism $A: G \to G$ is ergodic, then $A$ is strong-mixing, which is clear. However, why is the following true?


If $\lambda, \delta \in \widehat{G}$, then $(U_A^n \lambda, \delta) = 0$ eventually unless $\lambda = \delta \equiv 1$.


Here, for our purposes $U_A: L^2(m) \to L^2(m)$ is the induced operator defined by $U_A f = f \circ A$, where $f \in L^2(G, \mathscr{B}, m)$. It is known that this operator is linear and unitary, and that $U_A(fg) = U_A(f) U_A(g)$ and $U_A c = c$ when $c$ is constant.

$\widehat{G}$ is the character group of $G$, and we know that the members of $\widehat{G}$ form an orthonormal basis for $L^2(m)$, where $m$ is the Haar measure.

There is also an earlier theorem (Theorem 1.10), which says that if $G$ is a compact abelian group equipped with normalized Haar measure $m$ and $A: G \to G$ is a continuous surjective endomorphism of $G$, then $A$ is ergodic iff the trivial character $\lambda \equiv 1$ is the only $\lambda \in G$ that satisfies $U_A\lambda =\lambda \circ A^n = \lambda$ for some $n > 0$.

So basically we know that $\lambda, U_A \lambda, U^2_A \lambda, \dots$ are all distinct, but how does it follow that eventually $(U_A^n \lambda, \delta) = 0$? I get that this isn't true for $\lambda = \delta \equiv 1$, but why does it hold for all other choices of characters?