The Wikipedia article on Beth numbers defines $\beth_\alpha$ such that $\beth_{\alpha} =\begin{cases} |\mathbb{N}| & \text{if } \alpha=0 \\ 2^{\beth_{\alpha-1}} & \text{if } \alpha \text{ is a successor ordinal} \\ \sup\{\beth_\beta: \beta<\alpha\} & \text{if } \alpha \text{ is a limit ordinal} \end{cases}$
As it happens, for any ordinal $\alpha$, the statement $\exists \mathbb{S} | \beth_{\alpha}<|\mathbb{S}|<\beth_{\alpha+1}$ is independent of ZFC. In fact, according to an answer I received elsewhere, any statement or negation of that form is independent of any other under ZFC.
Now, to define what I'm calling the union-power sequence $\mathcal{UP}_n$ on a set $\mathbb{S}$ as
$$\mathcal{UP}_n(\mathbb{S}) = \begin{cases} \mathbb{S} & \text{if } n=0 \\ \mathcal{P}(\mathcal{UP}_{n-1}(\mathbb{S})) \cup \mathcal{UP}_{n-1}(\mathbb{S}) & \text{if } n \ge 1\end{cases}$$
With that definition, I can trivially prove that $\beth_n = |\mathcal{UP}_{n}(\mathbb{N})|$.
I would expect that behavior to continue when extending the sequence into infinite ordinals. The most logical way to handle $\omega$ is, in most cases, the limit behavior as $n \to \infty$. In this case, that's
$$\mathcal{UP}_{\omega}(\mathbb{S})=\bigcup_{n\in\mathbb{N}}\mathcal{UP}_{n}(\mathbb{S})$$
And if I extend to all ordinals, it looks like:
$$\mathcal{UP}_\alpha(\mathbb{S}) = \begin{cases} \mathbb{S} & \text{if } \alpha=0 \\ \mathcal{P}(\mathcal{UP}_{\alpha-1}(\mathbb{S})) \cup \mathcal{UP}_{\alpha-1}(\mathbb{S}) & \text{if } \alpha \text{ is a successor ordinal} \\ \bigcup_{\beta < \alpha}\mathcal{UP}_{\beta}(\mathbb{S}) & \text{if } \alpha \text{ is a limit ordinal}\end{cases}$$
Now, here's what throws me:
It would seem intuitively that $\beth_\omega=|\mathcal{UP}_{\omega}(\mathbb{S})|$. But according to the above definition, $\beth_\omega$ is just the smallest cardinality larger than any $\beth_n$ for any natural number $n$.
The definitions are equivalent, if and only if $\nexists \mathbb{S}\forall n \in \mathbb{N}:|\mathcal{UP}_n(\mathbb{N})|<|\mathbb{S}|<|\mathcal{UP}_\omega(\mathbb{N})|$.
That is, if there's no cardinality strictly greater than any $\beth_n$ for $n\in\mathbb{N}$ but strictly smaller than $|\mathcal{UP}_{\omega}(\mathbb{N})|$. Since if any intermediate cardinality does exist, then the supremum of $\{\beth_n: n\in\mathbb{N}\}$ would be the smallest such cardinality.
In fact, for $\beth_\alpha=|\mathcal{UP}_\alpha(\mathbb{N})|$ to hold in general for ordinal $\alpha$, you'd have to have $\nexists \mathbb{S}\forall \beta < \lambda:|\mathcal{UP}_\beta(\mathbb{N})|<|\mathbb{S}|<|\mathcal{UP}_\lambda(\mathbb{N})|$ for any limit ordinal $\lambda$.
This seems intuitively true, but it looks enough like the Continuum Hypothesis to give me pause.
So, to recap, the statement $$\nexists \mathbb{S}\forall \beta < \alpha:|\mathcal{UP}_\beta(\mathbb{N})|<|\mathbb{S}|<|\mathcal{UP}_\alpha(\mathbb{N})|$$ is $$\begin{cases} \alpha=0 & \text{False but not interesting; since } \nexists \beta<0 \text{(}\beta\text{ being an ordinal), it amounts to }\nexists\mathbb{S}\forall\emptyset:\phi\\ \alpha \text{ is a successor ordinal} & \text{Undecidable in ZFC, and the whole point of the Generalized Continuum Hypothesis} \\ \alpha \text{ is a limit ordinal} & \text{Unknown to me, and what I'm asking about}\end{cases}$$
Is that statement provable, disprovable, or undecidable for limit ordinals in ZFC? And if it's not provable, is there a more compact way to write cardinalities of the form $|\mathcal{UP}_\alpha(\mathbb{N})|$ that holds for infinite ordinal values of $\alpha$?
Or does Wikipedia have the definition wrong? The definition for limit ordinals listed there reads more like what I'd expect for a $\aleph_\alpha$ definition.