Determinacy From Woodins II
Posted on February 8, 2017
This is a continuation of my last post on determinacy, where we began the proof of projective determinacy. We’ve reduced the statement to showing that every projective set is homogeneously Suslin, which will be shown here, modulo a key lemma from Martin and Steel (1989).
This post is part of a series on determinacy:
- An Overview of Determinacy Axioms
- Determinacy From Woodins I
- Determinacy From Woodins II
- Determinacy From Woodins III
- From Determinacy to a Woodin I
- From Determinacy to a Woodin II
- The Structure of Games
- Borel Determinacy
- HOD Models of Determinacy
- Limitations of ZFC Determinacy
- Mice and Long Games
As previously mentioned we will inductively show that every $\bf\Pi^1_1$ set is homogeneously Suslin, assuming a measurable above a limit of Woodins. The start of the induction is the following theorem, derived from Martin’s proof of $\bf\Pi^1_1$-determinacy from a measurable.
Theorem (Martin, '70). Suppose $\kappa$ is a measurable cardinal. Then every $\bf\Pi^1_1$ set is $\kappa$-homogeneously Suslin.
I’ll here define the tree and measures that witness this fact, but without showing that it works. For the interested reader, this is 32.1 and 31.1 in Kanamori. Fix some $\bf\Pi^1_1$ set $A\subseteq{^\omega\omega}$. By the characterisation of $\bf\Pi^1_1$ sets (see [Kanamori, 13.1]) we can find a tree $T$ on $\omega\times\omega$ such that, given any $x\in{^\omega\omega}$,
$$ x\in A\text{ iff }T_x\text{ is wellfounded.} (1) $$
Fix some recursive enumeration ${^{<\omega}\omega}=\{s_i\mid i<\omega\}$ and define for each $x\in{^\omega\omega}$ a strict linear ordering $<_x$ on $\omega$ as $i<_x j$ iff
$$ s_i\notin T_x\land s_j\notin T_x\land i < j\text{ or }s_i\notin T_x\land s_j\in T_x\text{ or }s_i\in T_x\land s_j\in T_x\land s_i<_{\text{KB}}s_j, $$
where $<_{\text{KB}}$ is the Kleene-Brouwer ordering of ${^{<\omega}\omega}$. We then see by (1) that
$$ x\in A\text{ iff }<_x\text{ is a wellordering on }\omega. (2) $$
Given any $s\in{^{<\omega}\omega}$ define $T_{\langle s\rangle}:=\{t\in{^{<\omega}\kappa}\mid \exists n<|s|:\langle s\upharpoonright n,t\rangle\in T\}$ and furthermore set $<_s$ to be the ordering on $|s|$ defined exactly as $<_x$ but replacing $T_x$ with $T_{\langle s\rangle}$. Note that $s\subseteq t\Rightarrow <_s\subseteq <_t$ and $<_x=\bigcup_{n<\omega}<_{x\upharpoonright n}$ for any $x\in{^\omega\omega}$. We can now define the tree we’re interested in as
$$ T^*:=\{\langle s,t\rangle\mid\forall i,j<|s|:t_i < t_j\leftrightarrow i<_s j\}. $$
Note here that the second coordinates are providing witnesses for the wellfoundedness of the ordering associated to the first coordinate. By (2) we then see that indeed, $p[T^*]=A$. So far so good. As for the measures, letting $\mu$ be a normal measure on $\kappa$, define for each $s\in{^{<\omega}\omega}$
$$ \mu_s:=\{x\subseteq T^*_s\mid\exists H\in\mu:[H]^{|s|}\subseteq\text{ran}"x\}. $$
To see that this works one uses that $\kappa\rightarrow(\mu)^{<\omega}$ holds by Rowbottom’s theorem – see details in 32.1 and 32.2 in Kanamori.
Having covered the base case we then want to go from assuming that every $\bf\Pi^1_n$ set is homogeneously Suslin to the corresponding fact about $\bf\Pi^1_{n+1}$ sets. Firstly, we need to weaken the notion of a homogeneously Suslin set.
Definition(s). A tree $T$ over $^k\omega\times X$ for any set $X$ is $\kappa$-weakly homogeneous if there exists a countable set $\sigma$ of $\kappa$-complete measures such that for any $x\in p[T]$ there is a countably complete towert $\langle\mu_n\mid n<\omega\rangle\in{^\omega\sigma}$ with $T_{x\upharpoonright n}\in\mu_n$ for every $n<\omega$.
A set $A\subseteq{^k({^{<\omega}\omega})}$ is then $\kappa$-weakly homogeneously Suslin if $A=p[T]$ for a $\kappa$-weakly homogeneous tree, and weakly homogeneously Suslin if it’s $\kappa$-weakly homogeneously Suslin for some $\kappa$.
An alternative characterisation of the weakly homogeneously Suslin sets is that $A\subseteq{^k({^{<\omega}\omega})}$ is $\kappa$-weakly homogeneously Suslin iff $A=pB$ for some $\kappa$-homogeneously Suslin set $B\subseteq{^{k+1}({^{<\omega}\omega})}$ (see 32.3 in Kanamori). The following fact is then the crucial connection between the homogeneously Suslin sets and their weak variants.
Key Lemma (Martin-Steel, '89). Let $\delta$ be a Woodin cardinal and fix some $A\subseteq{^\omega\omega}$. Then if $A$ is $\delta^+$-weakly homogeneously Suslin, $\lnot A$ is $\alpha$-homogeneously Suslin for every $\alpha<\delta$.
This is the main theorem in Martin and Steel (1989), and for the purposes of this post we’ll just black box the result. Using this Key Lemma we can now prove projective determinacy.
Theorem (Martin-Steel, '89). Assume there exist $n$ Woodins and a measurable above. Then $\bf\Pi^1_{n+1}$ determinacy holds.
Proof. Let $\langle \delta_{i+1}\mid i < n\rangle$ be an increasing enumeration of the Woodins, set $\delta_0:=\omega$ and let $\kappa>\delta_n$ be the measurable. For each $i < n$ fix some ordinal $\alpha_i\in(\delta_i,\delta_{i+1})$ and set $\alpha_n:=\kappa$. We will show by induction on $i$ that every $\bf\Pi^1_{i+1}$ set is $\alpha_{n-i}$-homogeneously Suslin for every $i\in[0,n]$.
For $i=0$ we get the result by the above theorem due to Martin. Assume now that every $\bf\Pi^1_{i+1}$ set is $\alpha_{n-i}$-homogeneously Suslin for some $i < n$ and fix some $\bf\Pi^1_{i+2}$ set $A\subseteq{^k({^\omega\omega})}$. By definition of $\bf\Pi^1_{i+2}$ we can fix some $\bf\Pi^1_{i+1}$ set $B\subseteq{^{k+1}({^\omega\omega})}$ such that $A=\lnot pB$.
By our inductive hypothesis $B$ is $\alpha_{n-i}$-homogeneously Suslin, so that $\lnot A$ is $\alpha_{n-i}$-weakly homogeneously Suslin by the above alternative characterisation of the weakly homogeneously Suslin sets. Since $\delta_{n-i}\in(\alpha_{n-(i+1)},\alpha_{n-i})$, the Key Lemma implies that $A$ is $\alpha_{n-(i+1)}$-homogeneously Suslin. As homogeneously Suslin sets are determined, we’re done. QED
So, modulo the Key Lemma, a result I might or might not write up a proof of at some point, we’ve shown projective determinacy from infinitely many Woodins and a measurable above. It turns out that from this hypothesis it’s even possible to prove that all sets of reals in $L(\mathbb R)$ are determined, a tremendous result due to Woodin, which will be covered next time.