I am asked whether or not there exists an ordinal $\alpha$ such that $\omega_{\alpha}$ where we define:

1). $\omega_{0} = \omega$

2). $\omega_{\alpha+1} = \gamma(\omega_{\alpha})$

3). $\omega_{\lambda} = sup\{\omega_{\alpha} \mid \alpha < \lambda\}$ for a non-zero limit ordinal $\lambda$

Where, if I remember correctly, $\gamma(\alpha)$ is the least ordinal that cannot be injected into $\alpha$

Given these definitions, and the question of finding such an ordinal $\alpha$, the natural thought is something along the lines of $\alpha = \omega_{\omega_{\omega_{\ddots}}}$, but I'm not even sure this object is a well defined ordinal.

Trying to define it in terms of the recursive relationship, I just end up with it inside of its own definition.

Given this, does there then exist such an ordinal $\alpha$? I feel as if there can't be anything else, but experience has told me that maths can have some bizarre counter examples, so I don't really know what to think anymore.

  • 1
    $\begingroup$ en.wikipedia.org/wiki/Fixed-point_lemma_for_normal_functions $\endgroup$ Apr 25 '18 at 21:11
  • $\begingroup$ As @HagenvonEitzen says, this question apply any other normal function. I recommend you to study epsilon numbers. This site has some questions about that. I also recommend study a little Veblen functions $\endgroup$
    – Dog_69
    Apr 25 '18 at 22:13
  • $\begingroup$ One could say that the notation $\psi(I)$, where I is the least inaccessible cardinal, in a system of ordinal collapsing functions such that C(n) is the smallest ordinal not defined using values of C(m) for m < n, with the functions of addition, multiplication, exponentiation and $f: \alpha \mapsto \Omega_\alpha$, would define the ordinal you speak of, which I should just the fixed point of $f: \alpha \mapsto \Omega_\alpha$. $\endgroup$ Apr 8 '19 at 21:49

The map $$\alpha\mapsto\omega_\alpha$$ has two important basic properties: it's increasing $(\alpha<\beta\implies\omega_\alpha<\omega_\beta$) and continuous at limit ordinals (for limit $\lambda$ we have $\omega_\lambda=\sup\{\omega_\beta:\beta<\lambda\}$). Your question is then answered by the following fundamental result about infinitary combinatorics:

Suppose $F$ is increasing and continuous. Then $F$ has a fixed point - that is, there is some $\gamma$ such that $F(\gamma)=\gamma$.

Proof sketch: First, we show by induction that $F(x)\ge x$ for all $x$. Then, consider the supremum of the sequence $$0,F(0), F(F(0)), F(F(F(0))), ...$$


The answer is yes. The mapping function $\alpha\mapsto\omega_\alpha$ is both continuous and increasing, which means it must have a fixed point. We can find this fixed point by taking the supremum - $\sup\{0,\omega,\omega_\omega,\omega_{\omega_\omega},...\}$ and in fact if we start with any two ordinals $\alpha_{1,2}$ which aren't fixed points of the function and we know that there exists an ordinal $\alpha_1<\gamma<\alpha_2$ which is a fixed point of the function, then if we take the two supremums $\sup\{\alpha_1,\omega_{\alpha_1},\omega_{\omega_{\alpha_1}},...\}$ and $\sup\{\alpha_2,\omega_{\alpha_2},\omega_{\omega_{\alpha_2}},...\}$ we will in fact get different fixed points. If we take a function $F(\alpha)$ which maps $\alpha$ to the least $\omega$-fixed point above it, then we take the set $\{F(\alpha)|\alpha\in\text{Ord}\}$ and since we already know that for different vaues of $\alpha$ with fixed points inbetween we get different fixed points, then it follows that there are as many ordinals satisfying $\alpha=\omega_\alpha$ as there are ordinals.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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