ProofComplete

Proof of the Perfect Set Property for Analytic Sets

Every uncountable analytic set contains a perfect subset. This classical result, due to Suslin, shows that the continuum hypothesis holds for analytic sets.


Statement

Theorem9.8Perfect set theorem for analytic sets

Let XX be a Polish space and AβŠ†XA \subseteq X an analytic (Ξ£11\boldsymbol{\Sigma}^1_1) set. If AA is uncountable, then AA contains a perfect set (a nonempty closed set with no isolated points). In particular, ∣A∣=c|A| = \mathfrak{c}.


Proof

Proof

Since AA is analytic, there exists a continuous surjection f:ωω→Af: \omega^\omega \to A (or we can write A=p[T]A = p[\mathcal{T}] for a tree T\mathcal{T}). We use the Cantor-Bendixson analysis combined with the tree representation.

Step 1: Tree representation. Write AA as the projection of a closed set FβŠ†X×ωωF \subseteq X \times \omega^\omega: A=Ο€X[F]A = \pi_X[F]. This FF can be represented by a tree T\mathcal{T} on X<ω×ω<Ο‰X^{<\omega} \times \omega^{<\omega}: F=[T]F = [\mathcal{T}] (the set of infinite branches).

Step 2: The Suslin scheme. Define a Suslin scheme (Fs)sβˆˆΟ‰<Ο‰(F_s)_{s \in \omega^{<\omega}} by setting Fs={x∈X:βˆƒyβˆˆΟ‰Ο‰, yΒ extendsΒ s, (x,y)∈F}F_s = \{x \in X : \exists y \in \omega^\omega,\, y \text{ extends } s,\, (x,y) \in F\}. Then A=⋃nβ‹‚kF(n0,…,nk)A = \bigcup_{n} \bigcap_{k} F_{(n_0, \ldots, n_k)} where the union-intersection is over branches.

Step 3: Splitting. An uncountable analytic set has a splitting tree: a perfect subtree SβŠ†Ο‰<Ο‰S \subseteq \omega^{<\omega} such that for every node s∈Ss \in S, the set Fs∩AF_s \cap A is uncountable, and ss has at least two children s⌒0,s⌒1∈Ss^\frown 0, s^\frown 1 \in S with Fs⌒0F_{s^\frown 0} and Fs⌒1F_{s^\frown 1} both uncountable.

The construction proceeds by induction: at each level, use the fact that if As=Fs∩AA_s = F_s \cap A is uncountable, it cannot be the union of countably many singletons. Since each As=⋃nAs⌒nA_s = \bigcup_n A_{s^\frown n}, at least one (in fact two, by a pigeonhole argument) must be uncountable.

Step 4: Extracting a perfect set. The splitting tree SS determines a continuous injection Ο†:2Ο‰β†’A\varphi: 2^\omega \to A: for each x∈2Ο‰x \in 2^\omega, follow the corresponding branch of SS to obtain a unique point Ο†(x)∈A\varphi(x) \in A. The image Ο†[2Ο‰]\varphi[2^\omega] is a perfect set contained in AA.

More precisely: refine the neighborhoods so that diam(Fs)β†’0\mathrm{diam}(F_s) \to 0 along each branch (possible since XX is metrizable). Then each branch of SS determines a unique point in AA, and the resulting set is homeomorphic to 2Ο‰2^\omega (the Cantor set), hence perfect. β– \blacksquare

β– 

Consequences

RemarkThe continuum hypothesis for analytic sets

Since every uncountable analytic set has cardinality c\mathfrak{c} (it contains a perfect set of size 2β„΅02^{\aleph_0}), there are no analytic sets of intermediate cardinality between β„΅0\aleph_0 and c\mathfrak{c}. This means the continuum hypothesis holds for analytic sets, regardless of whether CH holds in general.

This was one of the early triumphs of descriptive set theory: resolving CH for a natural and extensive class of sets.

ExampleFailure for coanalytic sets

The perfect set property for Ξ 11\boldsymbol{\Pi}^1_1 (coanalytic) sets is independent of ZFC:

  • If V=LV = L, there exists an uncountable coanalytic set with no perfect subset (an uncountable Ξ 11\boldsymbol{\Pi}^1_1 set of constructible reals).
  • If there exists a measurable cardinal (or 0β™―0^\sharp exists), then every uncountable Ξ 11\boldsymbol{\Pi}^1_1 set has a perfect subset.

This illustrates the boundary between ZFC-provable and independent statements in descriptive set theory.