Assume GCH. Suppose $j: V\to M$ is an elementary embedding such that $\mathrm{crit}(j)=\kappa$, ${}^\kappa M \subset M$, $M\supset V_{\kappa+2}$. We can assume $j$ is defined from some extender of length $\kappa^{++}$. Hence, in particular, we know $\kappa^{++}<j(\kappa) < \kappa^{+++}$.

For each $\delta<j(\kappa)$, let $U_\delta$ be an ultrafilter on $\kappa$ be defined such that $A\in U_\delta$ iff $\delta\in j(A)$. Is it true that for any $\delta <\delta' < j(\kappa)$, $U_{\delta}\neq U_{\delta'}$? It is true if $\delta<\kappa^{++}$ since $\mathrm{Ult}(V, U_\delta)$, $[id]_{U_\delta}=\delta$. Of course if the length of the extender is too long then this is not true since the number of ultrafilters on $\kappa$ is $\kappa^{++}$. In particular with appropriate hypothesis, it is possible that $j=j_E$ where $E$ is some extender on $[\delta]^{<\omega}$ where $\delta\in \kappa^{+++}\cap \mathrm{cf}(\kappa^{++})$ and some $\alpha\neq\alpha'<j(\kappa)$ satisfy that $U_\alpha=U_{\alpha'}$. But $\delta$ may not be $\kappa^{++}$.


1 Answer 1


TLDR: Yes, all the ultrafilters are different.

Suppose $M$ is a transitive class and $j : V\to M$ is an elementary embedding.

Some notation.

  • For any $x\in M$, let $H_x = \{j(f)(x) : f\text{ is a function}\}\prec M.$
  • Let $j_x : V\to M_x$ be the ultrapower by the derived ultrafilter $U_x = \{A\subseteq X : x\in j(A)\}$ where $X$ is any set such that $x\in j(X)$.
  • Let $k_x : M_x\to M$ be the unique elementary embedding with $k_x\circ j_x = j$ and $k_x([\text{id}]_{U_x}) = x$.

Equivalently, $M_x$ is the transitive collapse of $H_x$, $k_x : M_x\to M$ is the inverse of the transitive collapse, and $j_x = k_x^{-1}\circ j$.

A nonstandard definition. An ordinal $\xi$ is a weak generator of $j$ if $\xi\neq j(f)(\alpha)$ for any function $f$ and any $\alpha < \xi$. (The difference between a generator and a weak generator is that in the definition of a generator one requires that $\xi\neq j(f)(\vec \alpha)$ for any $\vec \alpha \in [\xi]^{<\omega}$. Thus a generator is the same thing as an additively indecomposable weak generator.)

A useful fact about elementary embeddings. Suppose $x\in M$ and $\xi$ is the least ordinal such that $x \in H_\xi$. Then $$H_x = H_\xi$$ Thus $j_x = j_\xi$ and $k_x = k_\xi$. Moreover $\xi$ is a weak generator of $j$. We call $\xi$ the weak generator associated to $x$.

Key point: In your situation, for any distinct weak generators $\xi,\xi'$ of $j$, $j_\xi\neq j_{\xi'}$. (Also weak generators are the same thing as generators in your situation, but this is not really relevant.) Since $V_{\kappa+2}\subseteq M$, if $\xi < (2^\kappa)^+$ is a weak generator of $j$, then $$\xi < ((2^{\kappa})^+)^{M_\xi} = \min (\text{Ord}\setminus H_\xi)$$ Hence $((2^{\kappa})^+)^{M_\xi}$ is the least weak generator of $j$ above $\xi$. Therefore for distinct weak generators $\xi,\xi'$, $((2^{\kappa})^+)^{M_\xi}\neq ((2^{\kappa})^+)^{M_{\xi'}}$, so $M_\xi \neq M_{\xi'}$, and certainly $j_\xi \neq j_{\xi'}$.

The answer. Assume that for any distinct weak generators $\xi,\xi'$ of $j$, $j_\xi\neq j_{\xi'}$. Suppose $X$ is a set, suppose $x,x'\in j(X)$, and assume the derived ultrafilter $U_x = \{A\subseteq X : x\in j(A)\}$ is equal to $U_{x'} = \{A\subseteq X : x'\in j(A)\}$. We will show $x = x'$. Let $\xi$ and $\xi'$ be the weak generators associated to $x$ and $x'$ respectively. Since $U_x = U_{x'}$, the associated ultrapower embeddings $j_x$ and $j_{x'}$ are also equal. It follows that $j_\xi = j_x = j_{x'} = j_{\xi'}$. Thus by our assumption about weak generators, $\xi = \xi'$. In particular, $k_x = k_{\xi} = k_{\xi'} = k_{x'}$. Therefore $$x = k_x([\text{id}]_{U_x}) = k_{x'}([\text{id}]_{U_{x'}}) = x'$$ Hence $x = x'$, as desired.

  • $\begingroup$ Thank you! I think along the same lines the following about iterations of ultrapowers is also true: if $g_0: [\kappa]^2\to \kappa$ and $g_1: [\kappa]^2\to \kappa$ are representing two different ordinals in $\mathrm{Ult}(V, E^2)$, then the ultrafilters derived from the map $V\mapsto \mathrm{Ult}(V, E^2)$ (with $[g_0], [g_1]$) will be different. $\endgroup$
    – Otto
    Apr 5, 2019 at 20:21
  • $\begingroup$ There are distinct ordinals in the ultrapower by $E^2$ whose derived ultrafilters are equal, namely $\kappa$ and $j_E(\kappa)$ where $\kappa = \text{crt}(j_E)$. $\endgroup$ Apr 5, 2019 at 20:33
  • $\begingroup$ right of course. I meant to say if they represent ordinals in the same interval $[\kappa_k, \kappa_{k+1})$ $\endgroup$
    – Otto
    Apr 5, 2019 at 20:39
  • $\begingroup$ This seems more subtle to me $\endgroup$ Apr 5, 2019 at 21:07

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.