5
$\begingroup$

Consider large countable ordinals (e.g. $\epsilon_0$ which is not "large", but still interesting).

  • These are countable sets, so they inject into the complex plane ( or even the real line).
  • They contain limit ordinals.

Can we construct such an embedding in such a way that the only limit points in the image are images of limit ordinals?

For ordinals like $\omega^n$ and even $\omega^\omega$ the answer is yes.

I guess the answer should be yes, because these ordinals are none else but well orderings of integers.

$\endgroup$
0

2 Answers 2

6
$\begingroup$

Yes, in fact every countable ordinal embeds into the rational numbers in this way, an order-preserving map as a closed set of rational numbers.

Let me give three proofs.

First, one can easily prove this by induction on ordinals. Having a closed copy of $\alpha$ in $\mathbb{Q}$, we can scale it down to a bounded interval and then add another point on top to get a copy of $\alpha+1$. And if we have copies of every $\alpha<\lambda$ for a countable limit ordinal, we can similarly scale to successive intervals in such a way that the combined embedding contains its limit points.

Second, alternatively, one can also prove it using the universal property of the rational order. For any countable ordinal $\alpha$, extend the order of $\alpha$ to a dense linear order by adding a copy of $\mathbb{Q}$ between every ordinal and its successor. This gives a countable dense linear order, which must be order isomorphic to an interval of $\mathbb{Q}$. The original order is closed in the larger order, so this gives the ordinal as a closed subset of $\mathbb{Q}$.

Third, here is another soft proof. Cantor proved that the rational order $\mathbb{Q}$ is universal for all countable linear orders. So every countable ordinal $\alpha$ admits an order-embedding into $\mathbb{Q}$. The image of this embedding might not be closed in the way you have asked. But let us simply add the limit points. The new set (the old copy of $\alpha$ plus the limits) will still be well ordered and thus be a copy of some ordinal $\beta\geq \alpha$. So taking the first $\alpha$ many elements of this set gives a closed copy of $\alpha$ in $\mathbb{Q}$.

$\endgroup$
5
  • 1
    $\begingroup$ But if one asks about an explicit embedding? Presumably embeddings if non-recursive ordinals cannot be explicitly described, or? $\endgroup$ Commented Sep 14, 2023 at 5:09
  • 1
    $\begingroup$ The embedding is an isomorphism to a closed countable well-ordered set of rational numbers. This makes them quite explicit by many accounts of what this word means. (For example, the world of Borel sets and functions is sometimes described as the realm of explicit mathematics, and closed sets are Borel.) If you want something more explicit than this, then I'd want you to state clearly what the requirements are. $\endgroup$ Commented Sep 14, 2023 at 11:09
  • $\begingroup$ Fair enough. No, I cannot formulate any rigorous requirement. Actually I am rather wondering how would an obvious failure look like. Let me say it this way: surely one cannot do it for $\omega_1$ as no explicit well-ordering of the continuum is possible in any sense, right? So one might expect the same for some countable ordinals "so big they are as bad as $\omega_1$"... $\endgroup$ Commented Sep 14, 2023 at 14:42
  • 1
    $\begingroup$ If one is thinking of computability, then you will land at the Church-Kleene ordinal, the supremum of all ordinals representable by a computable relation on the natural numbers. en.wikipedia.org/wiki/Nonrecursive_ordinal $\endgroup$ Commented Sep 14, 2023 at 14:46
  • $\begingroup$ The computable ordinals are also exactly those that support a computable copy as a subset of the rationals. $\endgroup$ Commented Sep 14, 2023 at 14:58
7
$\begingroup$

Here is another construction derived from a solution a student submitted to a final exam for my undergraduate set theory course some years ago.

Let $f:\alpha \hookrightarrow \mathbb{N}$ be an injective function from $\alpha$ to the natural numbers. (This exists as $\alpha$ is countable.) Define a measure $\mu$ on $\alpha$ by assigning the point masses $\mu(\{ \beta \}) = 2^{-f(\beta)}$ for $\beta \in \alpha$. Then $\mu:\alpha \to \mathbb{R}$ is an order preserving embedding (if $\beta < \gamma$ with $\gamma \in \alpha$, then $\beta \subsetneq \gamma$ and because every point has positive measure, $\mu(\beta) < \mu(\gamma)$). Moreover, with the exception of $\mu(\alpha)$ itself when $\alpha$ is a limit ordinal, the only limits in of the image of $\mu$ are the images of the limit ordinals in $\alpha$. In the case that $\alpha$ is a limit, replace $\mu$ with $h \circ \mu$ where $h:[0,\mu(\alpha)) \to [0,\infty)$ is an order preserving homeomorphism.

$\endgroup$
3
  • 2
    $\begingroup$ Did that student become a set theorist or a topologist, or something more reasonnable? $\endgroup$
    – nombre
    Commented Sep 14, 2023 at 6:16
  • $\begingroup$ @nombre Are you suggesting that it's unreasonable to become a set theorist or a topologist? $\endgroup$ Commented Sep 15, 2023 at 13:48
  • $\begingroup$ @TimothyChow I'm suggesting it is less reasonnable! $\endgroup$
    – nombre
    Commented Sep 17, 2023 at 7:56

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .