3
$\begingroup$

I am afraid this question might be very naïve, but I find it hard to locate a reference that does not answer a slightly different question.

Consider the Cantor set $C$ and a continous map $f: C\to C$ with finite topological entropy. Is it true that there must exist a $k$ such that the one-sided full shift on an alphabet of $k$ letters $\sigma : \{1,\dots, k\}^{\mathbb{N}} \to \{1,\dots, k\}^{\mathbb{N}}$ contains an embedded copy of $f$?

$\endgroup$

1 Answer 1

5
$\begingroup$

No.

Your condition is called being a (topological) subshift. If $(C,f)$ is a topological subshift, then there exists a finite clopen partition $P$ of $C$ such that the family $(f^{-n}P)_{n\ge 0}$ separates the points of $C$. (Indeed this is obvious of the shift on $k$ letters: take the partition into $k$ clopen subsets defined by evaluation at 0).

An example with entropy zero that is not a topological subshift is the odometer $(\mathbf{Z}_p,+1)$. Indeed, at the opposite, it satisfies that for every $P$, $(f^{-n}P)_{n\ge 0}$ can be refined to a finite partition (classes modulo $p^n$ for some $n$, since the latter partition is $f$-invariant). Also this shows that its entropy is (obviously) zero.

That $(C,f)$ is a topological subshift is also referred as "$f$ is expansive", but I find this terminology counter-intuitive (at least to me).


Algebraic insight: A pair $(C,f)$ is the same, through Stone duality, as a pair $(A,f')$ consisting of a Boolean algebra endowed with an endomorphism. The free objects on $k$ generators in this category (Boolean algebra endowed with an endomorphism) correspond to one-sided full shift on $k$ letters. Being a topological subshift thus corresponds to being a quotient of such a free (Boolean algebra with one endomorphism) on finitely many generators. The odometer is a typical case where the corresponding (Boolean algebra with one endomorphism) is not finitely generated; it is indeed infinite and locally finite (increasing union of finite Boolean subalgebras that are stable under the endomorphism).

$\endgroup$
2
  • 1
    $\begingroup$ PS I realize that the identity map already answers the question. Anyway, my answer provides a little more context. $\endgroup$
    – YCor
    Dec 22, 2017 at 17:03
  • $\begingroup$ Yes, I realized this night that having too many fixed (or periodic) points was an obstruction, and was going to adjust my question accordingly. Your answer closes the whole thing, thanks. $\endgroup$ Dec 23, 2017 at 6:44

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.