1
$\begingroup$

I'd like to show that the class of all limit ordinals, LIM, is cofinal in the class of all ordinals ON.

I thought that I could do this by contradiction and assume that it wasn't cofinal in ON. Then there is an $\alpha \in$ ON such that for all $\beta \in $ LIM: $\beta < \alpha$, i.e. $\beta \in \alpha$ and hence LIM $\subset \alpha$.

This means that LIM is a set so we can consider $\bigcup$ LIM. Then we have $\beta < \bigcup$ LIM for all $\beta \in$ LIM by definition of $\bigcup$ LIM. So LIM $\subseteq \bigcup$ LIM.

$\bigcup$ LIM is either a limit or a successor ordinal.

If it is a limit ordinal then $\bigcup$ LIM $\in$ LIM and by transitivity of ON, $\bigcup$ LIM $\subseteq$ LIM, so $LIM = \bigcup$ LIM in this case which would mean that $\bigcup$ LIM $\in$ LIM $= \bigcup$ LIM which would be a contradiction because sets cannot contain themselves. Hence $\bigcup$ LIM cannot be a limit ordinal.

If $\bigcup$ LIM is a successor ordinal then $\bigcup$ LIM = $\beta \cup \{ \beta \}$ for some $\beta \in $ ON.

Now I'm not sure how to proceed from here. I'm also not sure about whether I'm on the right track with this proof because I think there should be a shorter way to prove this.

Thanks for your help.

2 Answers 2

4

Trivially, let $\beta$ be an ordinal, then $\beta+\omega$ is strictly larger than $\beta$, and it is a limit ordinal. Therefore the limit ordinals are cofinal in the class of ordinals.

Why is $\beta+\omega$ a limit ordinal? Well, recall that $\beta+\omega = \sup\{\beta+n\mid n<\omega\}$. Suppose it is not a limit ordinal then $\beta+\omega=\gamma+1$ therefore for some $n<\omega$ we have that $\beta+n+1=\beta+\omega$, which is a contradiction.


Just as well, and perhaps even nicer: given $\beta$ we can take $\beta^+=\min\{\gamma\mid\forall f\colon\beta\to\gamma, f\text{ not surjective}\}$, the "next initial ordinal" is also a limit ordinal by definition.

  • 0
    Is showing that every cofinal class of ON is proper also a one liner?2011-12-29
  • 0
    @Matt: Of course. Let $M\subseteq\mathrm{Ord}$. If it a set then $M\in V_\alpha$ for some $\alpha$ and thus bounded by $\alpha$; and if it not a set then for every $\alpha$ we have $M\cap\alpha\subsetneq M$, so there is $\beta\in M$ such that $\alpha<\beta$, therefore $M$ is cofinal.2011-12-29
  • 0
    Can this proof be modified into not using the von Neumann universe?2011-12-30
  • 0
    @Matt: Why? What other definition of ordinals do you want to use?2011-12-30
  • 0
    I thought the von Neumann universe was the universe of *all* sets. So it's bigger than ON.2011-12-30
  • 1
    @Matt: It is, but I'm just talking about the rank of the set $M$, and a set of ordinals of rank $<\alpha$ is a subset of $\alpha$. This is the point, that $M$ is bounded by $\alpha$.2011-12-30
  • 0
    I was quite confused yesterday. Now I think that in your first comment the second part is superfluous. To show that a cofinal subclass is proper it's enough to assume that it's a set and produce a contradiction.2011-12-30
  • 0
    Where do you use cofinality in the first part of your proof?2011-12-30
  • 0
    I think I have figured it out now: Assume $M$ is a cofinal subset of ON. Then for every $\alpha$ in ON there is a $\beta$ in $M$ such that $\alpha \leq \beta$. Then consider $\bigcup M + 1$. This is an ordinal. But there is no $m \in M$ such that $\bigcup M + 1 \leq m$, therefore $M$ cannot be a set.2011-12-30
  • 0
    @Matt: I am not bloody, of course. My proof shows that sets are bounded and proper classes are cofinal. Your proof shows that every cofinal subclass is a proper class.2011-12-30
  • 0
    : ) But that's what I wanted to show!2011-12-30
  • 0
    @Matt: $\varphi\iff\psi$ is equivalent to $\lnot\varphi\iff\lnot\psi$. You should know that we both showed the same thing.2011-12-30
1

If $\mathbf{LIM}$ is not cofinal in $\mathbf{ON}$ then there is an ordinal $\alpha $ such that $\forall \lambda \in \mathbf{LIM}: \lambda \le \alpha$. This implies that $\mathbf{LIM}$ is a set so that we can construct the smallest ordinal exceeding all limit ordinals: $\bigcup \mathbf{LIM}$. $\bigcup \mathbf{LIM}$ is itself a limit ordinal (easy to show) but it is not in $\mathbf{LIM}$ which constitutes a contradiction.