Documentation

CsdLean4.Mathlib.LinearAlgebra.Projectivization.MeasureSpace

Measurable structure on projectivization #

Category: 1-Mathlib (CSD-free Mathlib upstream candidates).

Under [RCLike K] and finite-dimensional normed V, the projectivization ℙ K V carries a canonical Borel MeasurableSpace structure derived from its quotient topology (Topology.lean). This file installs:

Provenance #

Staged as upstream Mathlib material. All declarations live under namespace Projectivization with no CsdLean4-namespace prefix; the file is intended to land in Mathlib/LinearAlgebra/Projectivization/MeasureSpace.lean once usage stabilises. Discharges items 4.1–4.6 of specs/projectivization-plan.md (the full MeasureSpace.lean mathematical scope).

Hypothesis pattern #

[RCLike K] [NormedAddCommGroup V] [NormedSpace K V] [FiniteDimensional K V], matching the Topology.lean NormedFiniteDim section. Under these hypotheses, ℙ K V is a compact Hausdorff space; the Borel σ-algebra is the natural measurable structure. lift_measurable and measurable_iff_measurable_comp_mk' additionally take [MeasurableSpace V] [BorelSpace V] so the source subtype {v : V // v ≠ 0} inherits a Borel structure via Subtype.borelSpace that agrees with the Mathlib-canonical borel _ (callers typically borelize V).

Tags #

projectivization, projective space, Borel measurable space, quotient measurable space, scale-invariant measurable function

Second-countability of Projectivization K V. Free consequence of the open-quotient-map structure (Topology.lean's isOpenMap_mk' + isQuotientMap_mk') and second-countability of the source: V is second-countable (finite-dim normed over RCLike is proper via FiniteDimensional.proper_rclike, and proper metric spaces are second-countable via secondCountable_of_proper), so the open subtype {v : V // v ≠ 0} is second-countable (Subtype.secondCountableTopology), and the open quotient map carries that to ℙ K V (Topology.IsQuotientMap.secondCountableTopology).

@[instance_reducible]

The Borel σ-algebra on Projectivization K V, derived from its quotient topology (Topology.lean).

Gated on [RCLike K] and finite-dim normed V so it does not shadow the generic Quotient.instMeasurableSpace in algebraic-geometry contexts (where K is an abstract field and the analytic structure is not relevant).

Equations

Projectivization K V is a BorelSpace: the installed measurable space agrees with borel _ by definition.

Singletons in ℙ K V are measurable. Follows from T2 (closed singletons; established in Topology.lean's Projectivization.instT2Space) plus the Borel structure (closed sets are measurable).

The canonical surjection {v : V // v ≠ 0} → ℙ K V is measurable. Follows from continuity (Topology.lean's continuous_mk') via Continuous.measurable.

Stated under additional [MeasurableSpace V] [BorelSpace V] hypotheses so the source subtype {v : V // v ≠ 0} inherits a Borel MeasurableSpace via Subtype.borelSpace. Most callers will supply these (typically via borelize V); the resulting MeasurableSpace V agrees with borel V.

Coincidence of Borel and coinduced σ-algebras #

The key fact underwriting lift_measurable: under [RCLike K] + finite-dim normed V, the Borel σ-algebra on ℙ K V (the one installed as instMeasurableSpace) coincides with the σ-algebra obtained by pushing the Borel σ-algebra on {v : V // v ≠ 0} along mk'. Mathlib's Continuous.map_borel_eq (Polish.Basic) discharges this given PolishSpace on the source.

PolishSpace V is automatic for finite-dim normed V over [RCLike K] (separable + completely metrizable). The subtype {v : V // v ≠ 0} is open in V (complement of the closed singleton {0}), hence Polish via IsOpen.polishSpace.

The Borel σ-algebra on ℙ K V coincides with the σ-algebra coinduced from the Borel σ-algebra on {v : V // v ≠ 0} via mk'.

This is the coincidence lemma: it lets lift_measurable (below) reduce measurability of Projectivization.lift f hf (against the Borel σ-algebra) to measurability of f (against the Borel σ-algebra on the nonzero subtype). Without this lemma, the two σ-algebras might differ and measurable_from_quotient (which works with the coinduced version) would not suffice.

V is Polish under our hypotheses (ProperSpace V via FiniteDimensional.proper_rclike gives second-countable + complete + metrizable, hence Polish via the Mathlib.Topology.MetricSpace.Polish PolishSpace instance for separable + completely metrizable). The nonzero subtype is open in V, hence also Polish via IsOpen.polishSpace.

lift_measurable and the measurability characterisation #

theorem Projectivization.lift_measurable {K : Type u_1} {V : Type u_2} [RCLike K] [NormedAddCommGroup V] [NormedSpace K V] [FiniteDimensional K V] [MeasurableSpace V] [BorelSpace V] {α : Type u_3} [MeasurableSpace α] (f : { v : V // v 0 }α) (hf : ∀ (a b : { v : V // v 0 }) (t : K), a = t bf a = f b) (hf_meas : Measurable f) :

A scale-invariant measurable function on the nonzero subtype descends to a measurable function on ℙ K V. Routes the measurability of Projectivization.lift f hf through the coincidence lemma borel_eq_map_mk'.

Hypotheses: f : {v : V // v ≠ 0} → α is K-scale-invariant (hf, the same hypothesis required by Projectivization.lift); f is measurable for the Borel σ-algebra on the nonzero subtype (callers typically supply [MeasurableSpace V] [BorelSpace V] so this is the natural σ-algebra).

This is the load-bearing user-facing lemma for LF4 §3 + §8 — it lets callers build measurable functions on ℙ K V from measurable scale-invariant functions on V \ {0}, which is how preparations encode rep-maps in the LF2 / LF3 chain.

A function out of ℙ K V is measurable iff its precomposition with mk' is measurable. Companion to lift_measurable for the case where the function is already defined on ℙ K V (rather than constructed via Projectivization.lift).