Skip to article frontmatterSkip to article content

Tree representations of Σ21\mathbf{\Sigma}^1_2 sets

Analytic sets are projections of closed sets. Closed sets are in NN\Baire are infinite paths through trees on N\Nat.

We call a set ANNA \subseteq \Baire YY-Souslin if AA is the projection YN[T]\exists^{Y^{\Nat}}[T] of some [T][T], where TT is a tree on N×Y\Nat \times Y, that is

A=YN[T]={α ⁣:yYN(α,y)[T]}.A = \exists^{Y^\Nat}[T] = \{\alpha \colon \exists y \in Y^\Nat \: (\alpha,y) \in [T] \}.
Proof

Assume first AA is Π11\Pi^1_1. There is a recursive tree TT on N×N\Nat \times \Nat (and hence, in LL, since “being recursive” is definable) such that

αA    T(α) is well-founded.\alpha \in A \iff T(\alpha) \text{ is well-founded}.

Hence, αA\alpha \in A if and only if there exists an order preserving map (with respect to the inverse prefix ordering) π:T(α)ω1\pi: T(\alpha) \to \omega_1. We recast this now in terms of getting an infinite branch through a tree.

Let {σi ⁣:iN}\{\sigma_i \colon i \in \Nat\} be a recursive enumeration of N<N\Nstr. We may assume for this enumeration that σii|\sigma_i| \leq i. We define a tree T~\widetilde{T} on N×ω1\Nat \times \omega_1 by

T~={(σ,τ):i,j<σ[σiσj(σσi,σi)Tτ(i)<τ(j)]}.\widetilde{T} = \{ (\sigma,\tau) : \: \forall i,j < |\sigma| \: [\sigma_i \supset \sigma_j \: \wedge \: (\sigma\Rest{|\sigma_i|}, \sigma_i) \in T \: \to \: \tau(i) < \tau(j)] \}.

The tree T~\widetilde{T} is in LL, since it is definable from TT and ω1\omega_1. Furthermore, if αA\alpha \in A, then the existence of an order-preserving map π:T(α)ω1\pi: T(\alpha) \to \omega_1 implies that there is an infinite path (α,η)(\alpha,\eta) through T~\widetilde{T}.

Conversely, if such a path (α,η)(\alpha,\eta) exists, then there is an order preserving map π:T(α)ω1\pi: T(\alpha) \to \omega_1. Hence we have

αAη(ω1)N(α,η)[T~]α(ω1)N[T~],\alpha \in A \: \leftrightarrow \: \exists \eta \in (\omega_1)^{\Nat} \: (\alpha,\eta) \in [\widetilde{T}] \: \leftrightarrow \: \alpha \in \exists^{(\omega_1)^\Nat}[\widetilde{T}],

so AA is of the desired form.

Next, we extend the representation to Σ21\Sigma^1_2.

If AA is Σ21\Sigma^1_2, then there is a Π11\Pi^1_1 set BNN×NNB \subseteq \Baire\times\Baire such that A=NNBA = \exists^{\Baire} B. Since BΠ11B \in \Pi^1_1, we can employ the tree representation of Π11\Pi^1_1 to obtain a tree TT over N×N×ω1\Nat \times \Nat \times \omega_1 such that B=(ω1)N[T]B = \exists^{(\omega_1)^\Nat} [T].

Now we recast TT as a tree TT' over N×ω1\Nat \times \omega_1 such that (ω1)N[T]=(ω1)NB\exists^{(\omega_1)^\Nat}[T'] = \exists^{(\omega_1)^\Nat} B. This is done by using a bijection between N×ω1\Nat \times \omega_1 and ω1\omega_1.

This way we can cast the N×ω1\Nat \times \omega_1 component of TT into a single ω1\omega_1 component, and thus transform the tree TT into a tree TT' over N×ω1\Nat \times \omega_1 such that (ω1)N[T]=(ω1)N[B]\exists^{(\omega_1)^\Nat}[T'] = \exists^{(\omega_1)^\Nat}[B].

Σ21\mathbf{\Sigma}^1_2 sets as unions of Borel sets

We can use Shoenfield’s tree representation to extend Corollary 1 to Σ21\Sigma^1_2 sets.

Sierpinski’s original proof used AC\AC. The following proof does not make use of choice.

Proof

Let ANNA \subseteq \Baire be Σ21\Sigma^1_2. By Theorem 1 there exists a tree TT on N×ω1\Nat \times \omega_1 such that A=(ω1)N[T]A = \exists^{(\omega_1)^\Nat}[T]. For any ξ<ω1\xi < \omega_1 let

Tξ={(σ,η)T ⁣:iηη(i)<ξ}.T^\xi = \{ (\sigma,\eta) \in T\colon \forall i \leq |\eta|\: \eta(i) < \xi \}.

Since the cofinality of ω1\omega_1 is greater than ω\omega (this can be proved without using AC\AC), every d:ωω1d: \omega \to \omega_1 has its range included in some ξ<ω1\xi < \omega_1. Thus we have

A=ξ<ω1(ω1)N[Tξ].A = \bigcup_{\xi < \omega_1} \exists^{(\omega_1)^\Nat}[T^\xi].

For all ξ<ω1\xi < \omega_1, the set (ω1)N[Tξ]\exists^{(\omega_1)^\Nat}[T^\xi] is Σ11\PS{1}, because the tree TξT^\xi is a tree on a product of countable sets and hence is isomorphic to a tree on N×N\Nat \times \Nat. By Corollary 2, each Σ11\PS{1} set is the union of 1\aleph_1 many Borel sets, from which the result follows.

As for co-analytic sets, an immediate consequence of this theorem is (using the perfect set property of Borel sets):

Absoluteness of Σ21\Sigma^1_2 relations

Shoenfield used the tree representation of Σ21\mathbf{\Sigma}^1_2 sets to establish an important absoluteness result for Σ21\Sigma^1_2 sets of reals.

Suppose ANNA \subseteq \Baire is Σ21\Sigma^1_2. Then, by the Kleene Normal Form there exists a bounded formula of second order arithmetic φ(v0,v1,v2)\varphi(v_0,v_1,v_2) such that

αA    β0β1m  φ(αm,β0m,β1m).\alpha \in A \iff \exists \beta_0 \, \forall \beta_1 \, \exists m \; \varphi(\alpha\Rest{m},\beta_0\Rest{m},\beta_1\Rest{m}).

Let MM be an inner model of ZF\ZF. Arithmetical formulas can be interpreted in ZF\ZF and we can also relativize them. This allows us to introduce a relativized version of AA by identifying, as usual, a set with the predicate that defines it:

AM(α):    (β0MNN)(β1MNN)(m)φ(αm,β0m,β1m)A^M(\alpha) :\iff (\exists \beta_0 \in M\cap \Baire) \, (\forall \beta_1\in M \cap \Baire) \, (\exists m) \: \varphi(\alpha\Rest{m},\beta_0\Rest{m},\beta_1\Rest{m})

Note that we do not have to relativize the inner natural number quantifier, since N\Nat is absolute for inner models, and also not the formula φ\varphi, since a bounded arithmetic formula translates into a bounded set-theoretic formula (with only natural number quantifiers) and is therefore absolute for MM.

We can then say that AA is absolute for MM if for any αM\alpha\in M,

AM(α)    A(α).A^M(\alpha) \iff A(\alpha).

Absoluteness can be extended and relativized in a straightforward manner with respect to some γNNM\gamma \in \Baire \cap M.

All arithmetic predicates are absolute, since all quantifiers are natural number quantifiers. Shoenfield absoluteness extends this absoluteness to Σ21\Sigma^1_2 and Π21\Pi^1_2 predicates.

Proof

We show the theorem for Σ21\Sigma^1_2 predicates. For the relativized version, one uses the relative constructible universe L[γ]L[\gamma], see Jech (2003) or Kanamori (2003).

Let AA be a Σ21\Sigma^1_2 relation. For simplicity, we assume that AA is unary. Fix a tree representation of AA as a projection of a Π11\Pi^1_1 set. So, let TT be a recursive tree on N×N×N\Nat \times \Nat \times \Nat such that

A(α)    β  T(α,β) is well-founded.A(\alpha) \iff \exists \beta \; T(\alpha,\beta) \text{ is well-founded}.

Note that TT is in MM (since it is recursive and hence definable).

Now assume αM\alpha \in M and AM(α)A^M(\alpha). Hence there is a βM\beta \in M such that T(α,β)T(\alpha,\beta) is well-founded in MM. This is equivalent to the fact that in MM there exists an order preserving mapping π:T(α,β)Ord\pi: T(\alpha,\beta) \to \mathbf{Ord}.

Since MM is an inner model and TT is absolute, the mapping exists also in VV. Hence T(α,β)T(\alpha,\beta) is well-founded in VV and thus A(α)A(\alpha).

For the converse assume that αM\alpha \in M and A(α)A(\alpha). Now we use the alternative tree representation of AA given by Theorem 1. Let ULMU \in L \subseteq M be a tree on N×ω1\Nat \times \omega_1 such that A=(ω1)NUA = \exists^{(\omega_1)^\Nat} U.

As before, let

U(α)={(αn,τ)U ⁣:nN,τ(ω1)n,}U(\alpha) = \{ (\alpha\Rest{n}, \tau)\in U \colon n \in \Nat, \tau \in (\omega_1)^n, \}

Then, for any αNN\alpha \in \Baire,

A(α)    λ(ω1)N(α,λ) infinite path through U.    U(α) not well-founded.\begin{align*} A(\alpha) & \iff \exists \lambda \in (\omega_1)^\Nat \: (\alpha,\lambda) \text{ infinite path through $U$}. \\ & \iff U(\alpha) \text{ not well-founded}. \\ \end{align*}

This means that there exists no order preserving map U(α)ω1U(\alpha) \to \omega_1. But then such a map cannot exist in MM either. Thus, U(α)U(\alpha) is a tree in MM which is ill-founded in the sense of MM. Thus, by Shoenfield’s Representation Theorem relativized to MM, AM(α)A^M(\alpha).

Absoluteness for Π21\Pi^1_2 follows by employing the same reasoning, using that the complement is Σ21\Sigma^1_2.

By analyzing the proof one sees that it actually suffices that MM is a transitive \in-model of a certain finite collection of axioms ZF\ZF such that ω1M\omega_1 \subseteq M.

The result is the best possible with respect to the analytical hierarchy, since the statement

α  [α∉L]\exists \alpha \; [\alpha \not\in L]

is Σ31\Sigma^1_3, but cannot be absolute for M=LM = L.

Shoenfield’s absoluteness theorem also holds for sentences rather than predicates, with a similar proof. This means a Σ21\Sigma^1_2 statement is true in LL if and only if it holds in VV. Many results of classical analysis are Σ21\Sigma^1_2 statements. The Shoenfield absoluteness theorem says that if they can be established under V=L\VL, they can be established in ZF\ZF alone.

On the negative side, as we will soon see, Shoenfield absoluteness also puts strong limits on the use of forcing to establish independence results in analysis.

References
  1. Jech, T. (2003). Set Theory. Springer-Verlag.
  2. Kanamori, A. (2003). The Higher Infinite (Second). Springer-Verlag.