Documentation

CsdLean4.LF4.MomentMarginal

LF4 plan B (step 1): the moment marginal as a Haar marginal #

Category: 3-Local (the moment marginal as a Haar marginal).

Plan B is to discharge the hypothesis h_uniform of fs_born_volume_ratio_qubit and qubit_born_frequency_convergence, i.e. to prove

(fun p => momentMap p 0)∗ fubiniStudyMeasure p₀ = uniform on [0,1].

Everything is finite-dimensional (CPN N = ℂℙ^{N-1}, U(N)); plan B is purely a finite-dimensional measure computation — it does not touch CSD's finiteness.

This file lands the first, committable half: the measure-level reduction. Since fubiniStudyMeasure p₀ = (orbitMap p₀)∗ unitaryHaarProb and momentMap (U • p₀) i = ‖(U·rep)ᵢ‖²/‖U·rep‖² (momentMap_orbit), the moment marginal is the Haar law of the squared-modulus ratio of U acting on a representative:

(momentMap · i)∗ fubiniStudyMeasure p₀
  = (fun U => ‖(U·rep)ᵢ‖²/‖U·rep‖²)∗ unitaryHaarProb.

So h_uniform (for i = 0, N = 2) reduces exactly to: the Haar law of U ↦ ‖(U·rep)₀‖²/‖U·rep‖² on U(2) is Uniform[0,1]. With a unit representative ‖U·rep‖ = ‖rep‖, this is "|U₀₀|² (the first squared-modulus of a Haar-random unit column) is Uniform[0,1]" — the Beta(1,1) / Archimedes fact.

Remaining (the hard core). Proving that Haar marginal is uniform is the genuine analytic content with no current Mathlib support: it routes through either (Haar-orbit → uniform ) + (uniform Uniform[0,1]), or the complex-Gaussian route (normalised Gaussian → uniform sphere; |z₀|²/‖z‖² ~ Beta(1,1)). Tools: MeasureTheory.Constructions.HaarToSphere, Probability.Distributions.Beta/Gaussian. This is a multi-session build; see specs/carve-out-plan.md Tranche M, plan B.

Plan B, step 1 (the reduction). The moment-coordinate marginal of the Fubini–Study measure equals the Haar law of the squared-modulus ratio of U acting on a representative of p₀. Reduces h_uniform to a concrete distributional statement about Haar unitaries.