By a dynamical system I understand a pair $(K,G)$ consisting a compact Hausdorff space and a subgroup $G$ of the homeomorphism group of $K$.

We say that a dynamical system $(K,G)$

$\bullet$ is topologically transitive if for every non-empty open set $U\subseteq K$ its orbit $GU=\{g(x):g\in G,\;x\in U\}$ is dense in $K$;

$\bullet$ has dense orbit if for some point $x\in K$ its orbit $Gx$ is dense in $K$.

It is easy to see that a dynamical system is topologically transitive if it has a dense orbit. If the space $K$ is metrizable and nonempty, then the converse also is true.

On the other hand, under $\mathrm{non}(\mathcal M)<\mathfrak c$, there exists a subgroup $G\subset S_\omega$ of cardinality $|G|\le\mathrm{non}(\mathcal M)<\mathfrak c$ that induces a topologically transitive action of the Stone-Cech remainder $\omega^*=\beta\omega\setminus\omega$. The dynamical system $(\omega^*,G)$ does not have dense orbits since the space $\omega^*$ has density $\mathfrak c>|G|$. I am interested if such an example can be constructed in ZFC.

Problem. Is there topologically transitive dynamical system without dense orbits?

  • $\begingroup$ Does $S_\omega$ have a dense orbit on $\omega^*$? (Its action is obviously topologically transitive.) I thought I knew the answer to be negative, but I don't see right now. $\endgroup$
    – YCor
    Apr 2, 2020 at 17:44
  • $\begingroup$ By the way the action on the empty set is trivially topologically transitive without dense orbits, so the assertion for $K$ metrizable should assume $K$ non-empty. $\endgroup$
    – YCor
    Apr 2, 2020 at 18:27
  • $\begingroup$ @YCor The action of $S_\omega$ on $\omega^*$ is minimal: all orbits are dense. $\endgroup$ Apr 2, 2020 at 18:37

1 Answer 1


The answer is yes:

there is a topologically transitive dynamical system without dense orbits.


let X be a topological space that is not separable. Let $\ K=X^{\Bbb Z},\ $ and let $\ G\ $ be the group of homeomorphism of $\ K,\ $ induced by shifts $\ s_n\ (n\in\Bbb Z)\ $ of $\ \Bbb Z:\ $

$$ \forall_{n\in\Bbb Z}\forall_{x\in\Bbb Z}\quad s_n(x):= x+n $$

Let $\ p:=(p_n)\in K\ $ be arbitrary, and let

$$ P:=\{p_n:n\in \Bbb Z\} $$

Then there exists non-empty open $\ U\ $ in $\ X,\ $ disjoint with set $\ P.\ $ Then non-empty open in $\ K\ $ set $\ W,$

$$ W\ :=\ \pi_0^{-1}(U)\ $$

is disjoint with the orbit $\ p.\ $

On the other hand, let $\ \emptyset\ne G\subseteq K, $ where $\ G\ $ is open in $\ K.\ $ Then there exists non-empty $\ H\ $ and integer $ a\ge 0\ $ such that $\ H\ $ is an open subset of $\ X^{(-a)..a}\ $ (Perl notation "s..t") and

$$ \emptyset\ \ne\ \pi_{(-a)..a}^{-1}(H)\ \subseteq G $$

Obviously, the orbit of $\ \pi_{(-a)..a}^{-1}(H)\ $, hence of $\ G,\ $ is dense in $\ K.$


  • 1
    $\begingroup$ Thank you for the great answer. And what about the case of subgroups $G\subseteq S_\omega$ acting on $\omega^*$? $\endgroup$ Apr 2, 2020 at 18:32
  • 2
    $\begingroup$ By $\omega^*$ I denote the remainder of the Stone-Cech compactification of $\omega$ and this is the principal space I am interested in. In this case the group $\mathbb Z$ (and any other countable group) does not work, unfortunately. $\endgroup$ Apr 2, 2020 at 18:43
  • 2
    $\begingroup$ @TarasBanakh at this point if you're specifically interested in $G\subset S_\omega$ acting on $\omega^*$ it would be worth a separate question. $\endgroup$
    – YCor
    Apr 2, 2020 at 19:17
  • 1
    $\begingroup$ @YCor Thank you for the suggestion. I will then accept the answer of Wlod AA and write this specific question separately. $\endgroup$ Apr 2, 2020 at 19:23
  • 1
    $\begingroup$ Taras, thank you (and to YCor :) ) for accepting. $\endgroup$
    – Wlod AA
    Apr 2, 2020 at 19:33

Your Answer

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

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