Let $G$ be a discrete countable infinite group acting on a compact metric space $X$ via homeomorphisms preserving a probability measure $\mu$.

A function $\lambda\colon G\to \mathbb C$ is an eigenvalue of the action of $G$ if there exists a function $f\in L^2(X,\mu)$ such that for every $g\in G$ one has $\lambda(g)\cdot f=f\circ g$.

In this paper: Ergodicity of the Cartesian product by E. Flytzanis, Trans. Amer. Math. Soc. 186 (1973), 171-176 (freely available link at AMS site), there are two results concering the product of two dynamical systems given by $\mathbf Z$-action:

  • a sufficient condition for the ergodicity of the product,
  • a description of the spectrum of their cartesian product.

It is also written there that the above results hold also for $G$-action, but it is not specified what does it exactly mean. Am I right that:

  • A product of $(X,G)$ and $(Y,G)$ is ergodic if the function constantly equal to 1 is the only common eigenvalue for $(X,G)$ and $(Y,G)$?
  • The set of eigenvalues of $X\times Y$ equals the set of all functions of the form $f\cdot g$ (pointwise multiplication), where $f$ is an eigenvalue for $X$ and $g$ is an eigenvalue for $Y$?

I need also two other properties, but I could not find the appropriate references for them (maybe I am wrong that they are true?):

  • The set of eigenvalues of a factor of a dynamical system $(X,G, \mu)$ is contained in the set of eigenvalues of $(X,G,\mu)$?
  • Every eigenvalue of an ergodic dynamical system is simple (beware: I do not assume that $G$ is abelian).
  • 1
    $\begingroup$ For your first question the answer is "no". This is already seen when considering the regular representation of a finite group with trivial abelinzation. As a general rule of thumb, it is convenient to replace "eigenvalues" with "finite dimensional subrepresentations" when considering non-commutative groups. Upon such a replacement you'd get a positive answer. Your second question is badly formulated. Are you looking for eigenvalues or eigenvectors? $\endgroup$
    – Uri Bader
    May 19, 2020 at 9:44
  • 1
    $\begingroup$ The last two properties that you ask about are correct. For the first, a representation of a factor is a subrepresentation, and for the second, in the particular case of a compact group, every ergodic representation is a factor of the regular. Note that when asking about eigenvalues, the question always reduces to the case of a compact group. $\endgroup$
    – Uri Bader
    May 19, 2020 at 9:45
  • $\begingroup$ Thank you, Uri Bader! Your answer is really helpful. For the first question - could you recommend some reference for this topic? In the second question I meant eigenvalues. If I understand correctly, an eigenvalue is a function from $G$ to $\mathbb C$ so I can consider pointwise multiplication (that is $f\cdot g(x)=f(x)\cdot g(x)$). $\endgroup$ May 20, 2020 at 10:46
  • $\begingroup$ I am sorry for the confusion - I used $f$ and $g$ for eigenvalues which is probably not standard. $\endgroup$ May 20, 2020 at 10:58
  • $\begingroup$ oh... this is a horrible notation you adopted there. The answer is negative here as well. I'll write it in an answer format. $\endgroup$
    – Uri Bader
    May 21, 2020 at 7:58

1 Answer 1


The answer to both question is negative.

Take $G=S_3$, the symmetric group of the set $X=\{1,2,3\}$. Then $L^2(X)$ is decomposed to the trivial representation and a another two dimensional irreducible representation. As a $G$-space, $X\times X\simeq X \cup G$ where $X\subset X\times X$ is the diagonal and $G$ corresponds to the rest. This shows that $X\times X$ is not ergodic, providing a counter example to the first question. It also follows that $L^2(X\times X)$ contains a sub-representation isomorphic to the regular representation $L^2(G)$ and in particular it contains the one dimensional sign representation. This provides a counter example to the second question.

  • $\begingroup$ Thank you for the answer :) $\endgroup$ May 21, 2020 at 11:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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