Silver's Theorem for Borel Equivalence Relations #
This file provides:
gandy_harrington_for_relation: Silver's theorem for Borel equivalence relations on Polish spaces — a Borel equivalence relation with uncountably many classes contains a perfect set of pairwise-inequivalent points.silver_core_polish: the dichotomy form (countable quotient or perfect set of inequivalent points), derived as a thin wrapper.silverBurgessDichotomy: the full dichotomy for standard Borel spaces.
Status #
PROVED (2026-06-10). gandy_harrington_for_relation is now sorry-free: it is the
endpoint of Miller's classical category route, fully formalized in this project
(see docs/silver-phase2-route.md for the development record):
G0Fusion.exists_gsGraph_hom— the classical Kechris–Solecki–TodorcevicG₀-dichotomy construction (positivity ideals, Lusin separation, fusion);gSGraphHomHypothesis_holds— the homomorphism input, inSilverCategoryRoute.lean;isMeagre_pullback_class_of_gSGraph_hom(Miller Prop. 6),isMeagre_of_isMeagre_sections(Kuratowski–Ulam), andmycielski_cantor(Mycielski) — the category glue;gandy_harrington_of_gSGraphHom— the assembly used below.
Consequently silver_core_polish, silverBurgessDichotomy, and the instantiation of
morley_counting are unconditional, with axioms exactly
[propext, Classical.choice, Quot.sound].
The equality case gandy_harrington_for_eq (via Cantor–Bendixson) is kept as a simple
direct proof of the smooth end of the dichotomy.
Why there was no easy reduction to a "closed relation" case. A tempting plan is to refine
the Polish topology so the Borel relation becomes closed and then run a Cantor scheme. This
is invalid in general: a closed equivalence relation on a Polish space is smooth (the class
map x ↦ [x] into the Effros–Borel space is Borel), so "potentially closed ⟹ smooth". But
E₀ (eventual equality on 2^ℕ) is a Borel equivalence relation with continuum-many classes
— so Silver applies — that is not smooth (Glimm–Effros), hence not potentially closed. So
the hard core of Silver is exactly the non-smooth relations — the G₀-dichotomy content of
the category route.
Status #
Audited invariant (axioms confirmed via #print axioms): the project's descriptive-set-theory
chain is sorry-free. gandy_harrington_for_relation, silver_core_polish,
silverBurgessDichotomy, and the unconditional instantiation of morley_counting all report
exactly [propext, Classical.choice, Quot.sound].
Silver's theorem for Borel equivalence relations. A Borel equivalence
relation on a Polish space with uncountably many classes contains a perfect set
of pairwise-inequivalent points: there is a continuous injection
f : (ℕ → Bool) → α such that distinct inputs produce r-inequivalent outputs.
Proved via Miller's classical category route: the G₀-dichotomy homomorphism
(gSGraphHomHypothesis_holds, by the fusion construction G0Fusion.exists_gsGraph_hom),
Miller's independence lemma, Kuratowski–Ulam, and Mycielski's theorem, assembled by
gandy_harrington_of_gSGraphHom.
[Equality case, unconditional] Silver's theorem when the relation is equality
(r = ⊥): an uncountable Polish space already contains a continuous injection
(ℕ → Bool) → α, so distinct inputs give distinct (hence ⊥-inequivalent) outputs.
This is exactly gandy_harrington_for_relation instantiated at r = ⊥ — but, unlike the
general Borel case, it needs no descriptive-set-theory machinery beyond mathlib's
Cantor–Bendixson theorem (IsClosed.exists_nat_bool_injection_of_not_countable applied to
Set.univ). It is the smooth end of the dichotomy; the hard content of Silver lives
entirely in the non-smooth relations (see the E₀ note on gandy_harrington_for_relation).
Silver's theorem (dichotomy form) for Borel equivalence relations on Polish spaces. Either the quotient is countable, or there exists a perfect set of pairwise-inequivalent points.
Silver-Burgess dichotomy #
The Silver-Burgess dichotomy for Borel equivalence relations on standard Borel
spaces, derived from silver_core_polish.