This question assumes some familiarity with Jensen's fine structure analysis of the constructible universe L (https://en.wikipedia.org/wiki/Jensen_hierarchy, http://www.math.cmu.edu/~laiken/papers/FineStructure.pdf).

Everything to follow is in L.

Given some $J_\alpha$, the n-th projectum of $J_\alpha, \rho_n(J_\alpha)$ is defined as follows: the least $\rho\leq \alpha$ such that there exists a subset of $\omega\cdot \rho$ which is $\Sigma_n(J_\alpha)$ but not in $J_\alpha$. Another equivalent characterization is that $\rho_n$ is the least $\delta\leq \alpha$ such that there exists a $\Sigma_n(J_\alpha)$ that maps $\omega\cdot \delta$ onto $J_\alpha$. Of course if $1<\rho_n<\alpha$, then $J_\alpha\models \rho_n \text{ is a cardinal}$ so $\omega\cdot \rho_n = \rho_n$.

The exercise is asking to produce any arbitrary pattern of the projectums. More concretely like the following, exhibit a $J_\alpha$ such that $\rho_k(J_\alpha)=\alpha, k=0,1,2,3, \rho_4(J_\alpha)<\alpha, \rho_5<\rho_4, \rho_j=\rho_5 \forall j\geq 6$.

What I can do now is to produce one drop (I feel if somehow I know how to produce two drops then I am done). More precisely, consider $J_{\omega_2}$. Let $\xi$ be the least ordinal in $J_{\omega_2}$ which is not $\Sigma_4$-definable from $\omega_1$. Take the $\Sigma_4$ Skolem Hull in $J_{\omega_2}$ with parameters from $\omega_1 \cup \{\xi\}$, denoted by $Hull_{\Sigma_4}^{J_{\omega_2}}(\omega_1\cup \{\xi\}) \simeq_\pi J_{\beta}=Hull_{\Sigma_4}^{J_{\beta}}(\omega_1\cup \{\pi(\xi)\})$ by condensation via $\pi$. Then it's not hard to verify that $\rho_4(J_\beta)=\omega_1$ (with standard parameter $\{\pi(\xi)\}$) and $\rho_k(J_\beta)=\beta, k<4$ by elementarily.

But it is obvious that the above construction also yields that $\rho_k(J_\alpha)=\omega_1$ for all $k\geq 4$ by cardinality considerations. My feeling is that I should probably produce those projectums starting from $\rho_5$ (i.e. backtrack). But I don't see how, so far, to get another projectum drop. Thanks in advance!

  • $\begingroup$ What is $\alpha$ in the fourth paragraph? Also, is this an actual exercise? If so, perhaps you could mention where it's from? $\endgroup$
    – tci
    Feb 7, 2016 at 2:27
  • $\begingroup$ $\alpha$ is just any ordinal. Well there was some example in Jensen's original paper that produced (roughly) the pattern I mentioned above. I heard about this from a conversation but so far I couldn't either refute or confirm this. $\endgroup$
    – Jing Zhang
    Feb 7, 2016 at 2:31
  • $\begingroup$ Okay. Judging from the rest of the paragraph I thought that $\alpha$ might have been $\omega_2$. $\endgroup$
    – tci
    Feb 7, 2016 at 2:34
  • $\begingroup$ @tci sorry I found that (I looked at the wrong paragraph) they are all $\beta$. See my edits. $\endgroup$
    – Jing Zhang
    Feb 7, 2016 at 2:38
  • 1
    $\begingroup$ Did you look at "Patterns of Projecta" By Adam Krawczyk? $\endgroup$
    – William
    Apr 30, 2016 at 23:09

1 Answer 1


Let me include an argument here. Goal: given $s\in 2^{<\omega}$ we will produce a $J$-structure such that the projecta behave exactly as indicated by the string $s$ (1 means drop and 0 means stay).

Assume $V=L$. For any $s\in 2^{<\omega}$, for all infinite cardinals $\kappa$, there exists $\kappa \leq \alpha <\kappa^+$ such that in $J_\alpha$ the projecta pattern is $s$.

Fix $\kappa$. We need to demonstrate how to produce one more drop. Proceed by induction on the length of $s$, say $n$. Let $\beta\geq \kappa^+$ be such that $J_\beta$ has projecta pattern $s$. Our goal is to produce $\kappa\leq \alpha <\kappa^+$ such that the projecta pattern is $s^\frown \langle 1\rangle$, aka one more drop of projecta. Consider $A=Hull_{n+1}^{J_\beta}(\kappa \cup S)$ where $$S=\{\kappa\}\cup \{\rho_i(J_\beta): i\leq n\} \cup \{p_i: i\leq n\}\cup \{\gamma_i: i\leq n+1\}$$ where $\gamma_i$ is the parameter in $J_\beta$ for the $\Sigma_{i}$-Skolem function for $J_\beta$. We know $A\prec_{n+1} J_\beta$. What is left to check $\rho_{n+1}(A)=\kappa$ and $\rho_i(A)=\rho_i(J_\beta)$ for all $i\leq n$. Then the transitive collapse of $A$ will work.

$\rho_{n+1}(A)\leq\kappa$ because there exists a $\Sigma_{n+1}$ definable function that maps a subset of $\kappa$ onto $A$. But $\kappa$ is a cardinal. It follows that $\rho_{n+1}(A)=\kappa$.

Fix any $i\leq n$, we need to show $\rho_i(A)=\rho_i(J_\beta)$ (this also applies to the degenerate case where $\rho_i(J_\beta)=\beta$). $\rho_i(A)\leq \rho_i(J_\beta)$, since $h_i$ maps a subset of $\rho_i(J_\beta)$ onto $A$ with parameter in $A$, we know by soundness that $\rho_i(A)\leq \rho_i(J_\beta)$. Conversely, some $\Sigma_i(A)$ $h'_i$ maps $\rho_i(A)$ onto $A$. We have $A\models \forall x \exists y<\rho_i(A) h'_i(x)=y$. Note the sentence is $\Pi_{i+1}$ with parameters in $A$. By $\Sigma_{n+1}$ elementarity we know that $J_\beta \models \forall x \exists y<\rho_i(A) h'_i(x)=y$. Hence $\rho_i(J_\beta)\leq \rho_i(A)$.


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.