13
$\begingroup$

Suppose $\kappa$ is a measurable cardinal and $j:V\to M$ is the ultrapower by a normal measure on $\kappa$. Let's say, for instance, that $2^\kappa=\kappa^{++}$ (note that this assumption has consistency strength greater than just a measurable cardinal). It is easy to see that $j(\kappa)$ has size $\kappa^{++}$ in $V$. So the cofinality of $j(\kappa)$ in $V$ can be either $\kappa^+$ or $\kappa^{++}$ (it cannot be $\kappa$ because $M$ is closed under $\kappa$-sequences). Can each of these possibilities be realized in some model of set theory? More generally, what can be said about the cofinality of $j(\kappa)$ under various GCH assumptions?

As Asaf Karagila notes in the comments below, we can reformulate the question to ask what are the possible cofinalities of the order $\kappa^\kappa/U$ for a normal measure $U$ on $\kappa$. This is equivalent because elements of $j(\kappa)$ are precisely the equivalence classes of functions $f:\kappa\to\kappa$.

$\endgroup$
13
  • 2
    $\begingroup$ That's a very nice question! You can transform it into just asking what is the cofinality of the order $\kappa^\kappa/U$. This should reduce it to a model theoretic question, methinks. $\endgroup$
    – Asaf Karagila
    Oct 20, 2016 at 10:11
  • 1
    $\begingroup$ It's a good thing that I stopped by, then. :) $\endgroup$
    – Asaf Karagila
    Oct 20, 2016 at 10:18
  • 1
    $\begingroup$ @AsafKaragila Done! $\endgroup$ Oct 20, 2016 at 12:31
  • 2
    $\begingroup$ Suppose $\kappa$ is supercompact and Laver indestructible. Given any regular $\lambda > \kappa$ do a $<\kappa$-support iteraton where at successor steps we add a function $f: \kappa \to \kappa$ where $f$ is eventually above any ground model function. $\kappa$ remains supercompact in the extension and the functions introduced during the iteration should witness the cofinality is equal to $\lambda$ (with respect to any normal measure on $\kappa$). $\endgroup$ Oct 20, 2016 at 13:49
  • 2
    $\begingroup$ @MohammadGolshani: The length of the iteration can be an ordinal. By controlling the cofinality of the length of iteration you will get any desirable cofinality (above $\kappa$) for $j(\kappa)$. For example, if the length of the iteration is $\kappa^{++} + \kappa^{+}$ you will get $2^\kappa = \kappa^{++}$ and $\text{cf } j(\kappa) = \kappa^{+}$. This way you can get full independence between $2^\kappa$ and $\text{cf } j(\kappa)$. $\endgroup$
    – Yair Hayut
    Oct 20, 2016 at 21:12

1 Answer 1

10
$\begingroup$

As Mohammad Golshani remarked, it is possible to control the cofinality of $j(\kappa)$ by iterating the forcing that adds a function $f\colon \kappa \to \kappa$ which is eventually larger than any ground model function.

The conditions of the forcing notion are pairs of the form $(s, g)$ where $s{\in} ^{<\kappa}\kappa$ and $g\colon \kappa \to \kappa$, where $(s, g)$ is stronger than $(t, h)$ if $s\supseteq t$, $g \geq h$ everywhere and $\forall \alpha \in \text{dom } s \setminus \text{dom }t$, $s(\alpha) \geq h(\alpha)$.

Assuming $\kappa^{<\kappa} = \kappa$, this forcing is $\kappa$-centred and $\kappa$-directed closed. Let $\mathbb{P}_\alpha$ be the iteration of adding dominating function for $\alpha$ many steps with support ${<}\kappa$. Using the $\kappa$-closure of iteration and standard $\Delta$-system arguments - this iteration is $\kappa^{+}$-c.c. Therefore, it doesn't collapse cardinals.

Let $\langle f_i \mid i < \alpha\rangle$ be the sequence of the generic dominating functions.

Lemma: If $\text{cf }\alpha \geq \kappa^{+}$ then the true cofinality of $\kappa^{\kappa} / J^{bd}$ is $\text{cf }\alpha$. Namely, there is a cofinal, increasing sequence of functions in $\kappa^{\kappa}$ of order type $\text{cf }\alpha$.

Proof: Let $\lambda = \text{cf }\alpha$. Let $\langle \gamma_i \mid i < \lambda\rangle$ be a cofinal sequence. The sequence of functions $g_i = f_{\gamma_i}$ is increasing (modulo bounded error) and by the chain condition of $\mathbb{P}_\alpha$, every function in the generic extension is bounded by one of them. $\square$

Let $\kappa$ be a measurable cardinal. If $\text{tcf } \kappa^{\kappa} / J^{bd} = \lambda$, then for every $\kappa$-complete measure $\mathcal{U}$ on $\kappa$, $\text{cf }j_{\mathcal{U}}(\kappa) = \lambda$. Therefore, in order to construct a model in which $\text{cf }j(\kappa) = \lambda$, $2^\kappa = \mu$ where $\kappa < \lambda \leq \mu$, $\lambda$ regular, $\text{cf } \mu \geq \kappa^{+}$, we may start with indestructible supercompact $\kappa$ such that $2^\kappa = \kappa^{+}$ and force with $\mathbb{P}_{\mu + \lambda}$.

$\endgroup$
2
  • $\begingroup$ Is it provable that $\kappa^\kappa/\mathcal J^{bd}$ has a true cofinality in the case of a measurable cardinal? $\endgroup$
    – Asaf Karagila
    Oct 22, 2016 at 18:15
  • $\begingroup$ It's even enough for the non stationary ideal to have true cofinality in this case, when you think about it. $\endgroup$
    – Asaf Karagila
    Oct 23, 2016 at 8:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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