From coded well-orders to models: the defect bridge #
A structure whose distinguished relation fails to be a well-order carries a countable nonempty
seed of witnesses; seeding a countable fragment-elementary substructure with it and transporting
that substructure to the carrier ℕ produces a code which is not a well-order. Contrapositively,
if every code satisfying φ is a well-order then so is every model of φ ⊓ infiniteAxiom.
Deliberately independent of López–Escobar. This file imports only the well-order class, the fragment Löwenheim–Skolem machinery and the infiniteness axiom. Two very different consumers need the bridge and neither should have to drag in the other:
Descriptive/WellOrderNonBorel.lean(#33) obtains itsφfrom López–Escobar applied to a hypothetically BorelWO;Descriptive/AnalyticWellOrderBoundedness.lean(#64) obtains itsφfrom the analytic-PC sandwich, where the reduct class is only contained inWO.
That second consumer is why isWellOrder_of_realize_of_modelsOf_subset is the primary form: a
pcSentence's reduct class sits inside an invariant envelope and never equals a prescribed set.
Main results #
exists_countable_defect_seed: a well-order failure has a countable nonempty witness set.isWellOrder_of_realize_of_modelsOf_subset: containment form of the bridge.isWellOrder_of_realize: the equality-form corollary.
The defect seed #
A structure whose relation fails to be a well-order carries a countable nonempty seed of witnesses: every subset containing the seed inherits the failure. In this Mathlib a well-order is trichotomy plus well-foundedness (transitivity is derived), so there are exactly two cases: a two-element trichotomy failure, and the range of an infinite descending sequence.
The bridge: coded definability forces every model to be well-ordered #
The bridge, equality form: if a sentence defines the well-order class on codes, then every model of it conjoined with the infiniteness axiom is well-ordered.
The ModelsOf φ = wellOrderClass lt specialization of
isWellOrder_of_realize_of_modelsOf_subset; the defect-seed argument lives there and is not
repeated.