ConceptComplete

Paradoxical Consequences of the Axiom of Choice

The axiom of choice implies several results that are counterintuitive or "paradoxical," demonstrating the power and subtlety of non-constructive existence proofs.


Non-Measurable Sets

Theorem6.3Vitali's theorem

Assuming AC, there exists a subset of [0,1][0,1] that is not Lebesgue measurable.

ExampleConstruction of a Vitali set

Define an equivalence relation on [0,1][0,1]: xyx \sim y iff xyQx - y \in \mathbb{Q}. By AC, choose one representative from each equivalence class to form a set V[0,1]V \subseteq [0,1].

If VV were measurable with measure mm, then [0,1]qQ[1,1](V+q)[1,2][0,1] \subseteq \bigcup_{q \in \mathbb{Q} \cap [-1,1]} (V + q) \subseteq [-1,2]. By countable additivity and translation invariance: 1qm(V+q)=qm31 \leq \sum_{q} m(V+q) = \sum_q m \leq 3. But if m=0m = 0, the sum is 0<10 < 1; if m>0m > 0, the sum is >3\infty > 3. Contradiction.


The Banach-Tarski Paradox

Theorem6.4Banach-Tarski paradox

Assuming AC, the closed unit ball B3R3B^3 \subset \mathbb{R}^3 can be partitioned into finitely many pieces which, after rotations and translations, reassemble into two copies of B3B^3. More precisely, there exist disjoint sets A1,,Ak,B1,,BlA_1, \ldots, A_k, B_1, \ldots, B_l with B3=A1AkB1BlB^3 = A_1 \sqcup \cdots \sqcup A_k \sqcup B_1 \sqcup \cdots \sqcup B_l, and rigid motions σi,τj\sigma_i, \tau_j such that σ1(A1)σk(Ak)=B3\sigma_1(A_1) \sqcup \cdots \sqcup \sigma_k(A_k) = B^3 and τ1(B1)τl(Bl)=B3\tau_1(B_1) \sqcup \cdots \sqcup \tau_l(B_l) = B^3.

RemarkKey ingredients

The Banach-Tarski paradox relies on:

  1. The axiom of choice (to select representatives from orbits).
  2. The non-amenability of the free group F2F_2 (which can be embedded in SO(3)SO(3) via rotations).
  3. Dimension 3\geq 3: In dimensions 1 and 2, finitely additive translation-invariant measures exist on all bounded sets (Banach), so Banach-Tarski fails. The paradox is specific to n3n \geq 3 because SO(n)SO(n) contains free subgroups for n3n \geq 3.

Other Counterintuitive Consequences

ExampleFurther consequences of AC
  1. Every infinite set has a countably infinite subset (using DC, which follows from AC).

  2. Every surjection has a right inverse: given f:ABf: A \twoheadrightarrow B, AC provides g:BAg: B \to A with f(g(b))=bf(g(b)) = b.

  3. The existence of well-orderings of R\mathbb{R}: No explicit well-ordering of R\mathbb{R} is known or definable, yet AC guarantees one exists.

  4. Non-principal ultrafilters on N\mathbb{N} exist: These are used throughout logic and combinatorics but cannot be explicitly constructed.

RemarkConsistency and independence

Godel (1938) proved that AC is consistent with ZF (if ZF is consistent): he showed AC holds in the constructible universe LL. Cohen (1963) proved that ¬\negAC is also consistent with ZF using the method of forcing. Thus AC is independent of ZF, and mathematicians may choose whether to adopt it. Most mainstream mathematics assumes ZFC (ZF + AC).