Documentation

TauCeti.NumberTheory.HeckeRing.FundamentalDomain

The Hecke coset representatives tile a fundamental domain #

If S is a fundamental domain for φ(Γ₂), the translates of S by the images of the representatives aᵥ = rightCosetRep D v = δ τᵥ⁻¹ tile one for φ(Γ₁) ⊓ φ(δ) φ(Γ₂) φ(δ)⁻¹.

This is generic: the ambient group G, the acting group P and the space α acted on are all arbitrary, and no step uses matrices, the upper half-plane, or a slash action. It sits here beside rightCosetRep rather than in the modular-forms layer that consumes it, so a generic consumer need not import the slash action to reach it.

The statement is made along a homomorphism φ because the double cosets frequently live in a group that does not act on the space of interest. In the motivating instantiation φ is TauCeti.ratPosToPSL2R, whose source is the positive-determinant subgroup GL(2, ℚ)⁺ — that restriction being forced for the fractional-linear action specifically, since a negative-determinant matrix carries ℍ to the lower half-plane.

Main results #

Provenance #

Adapted from the AINTLIB LeanModularForms project (LeanModularForms/HeckeRIngs/GL2/AdjointTheory/FDTransport.lean, https://github.com/CBirkbeck/AINTLIB, commit 6d87d596a5372d5b122c47b7082d4c3afa9b7c3b, Apache-2.0, Chris Birkbeck). There the transport is carried out at PSL level with the source group fixed at SL(2, ℤ): Gamma_p_α_FD_finite_index_decomp (FDTransport.lean:128) takes φ : SL(2, ℤ) →* G_outer, so it is generic in the acting group but not in the source. The statement here is generic in the source group, the acting group and the space acted on alike.

References #

theorem DoubleCoset.isFundamentalDomain_iUnion_rightCosetRep_smul {G : Type u_1} [Group G] {Δ : Submonoid G} {Γ₁ Γ₂ : Subgroup G} (D : HeckeCoset Δ Γ₁ Γ₂) {P : Type u_2} {α : Type u_3} [Group P] [MeasurableSpace α] [MulAction P α] (φ : G →* P) {H : Subgroup G} [Countable (DecompQuotient Γ₂ Γ₁ (↑(Quotient.out D))⁻¹)] {S : Set α} {μ : MeasureTheory.Measure α} (h₂ : Γ₁ ≤ H) (hconj : ∀ y ∈ Γ₂, ↑(Quotient.out D) * y * (↑(Quotient.out D))⁻¹ ∈ H) (hker : φ.ker ⊓ H ≤ Γ₁) (hS : MeasureTheory.IsFundamentalDomain (↥(Subgroup.map φ Γ₂)) S μ) (hδ : MeasureTheory.Measure.QuasiMeasurePreserving (fun (x : α) => (φ ↑(Quotient.out D))⁻¹ • x) μ μ) (hnull : ∀ (v : DecompQuotient Γ₂ Γ₁ (↑(Quotient.out D))⁻¹), MeasureTheory.NullMeasurableSet ((φ ↑(Quotient.out v))⁻¹ • S) μ) :
MeasureTheory.IsFundamentalDomain (↥(Subgroup.map φ Γ₁ ⊓ ConjAct.toConjAct (φ ↑(Quotient.out D)) • Subgroup.map φ Γ₂)) (⋃ (v : DecompQuotient Γ₂ Γ₁ (↑(Quotient.out D))⁻¹), φ (rightCosetRep D v) • S) μ

The images of the Hecke coset representatives tile a fundamental domain. If S is a fundamental domain for φ(Γ₂), the translates of S by the images of the representatives aᵥ = rightCosetRep D v = δ τᵥ⁻¹ tile one for φ(Γ₁) ⊓ φ(δ) φ(Γ₂) φ(δ)⁻¹.

Supply H rather than injectivity of φ: any subgroup containing Γ₁, receiving the conjugate δ Γ₂ δ⁻¹, and meeting ker φ inside Γ₁ — neither Γ₂ ≤ H nor conjugation-stability of H is needed. The determinant-one subgroup serves when φ collapses no more of it than {±1} and {±1} ≤ Γ₁ — both hold for ratPosToPSL2R over any Γ₀(N), but neither follows from the statement, which constrains φ only through hker.

Why injectivity is not the alternative, for the intended instantiation: P there acts faithfully and is a matrix group modulo scalars, so with -I ∈ Γ₂ an injective φ would make φ (-I) a non-identity element of φ(Γ₂) acting trivially, and hS could then hold for no set of positive measure — MeasureTheory.IsFundamentalDomain demands a.e.-disjointness over distinct group elements. Nothing in the statement itself forces this: the action is arbitrary here.

The acted-on space is an arbitrary measurable α, not ℍ: no hypothesis and no step of the proof uses upper-half-plane structure, and the tiling lemma underneath (MeasureTheory.IsFundamentalDomain.iUnion_mul_smul_of_transversal) is already stated at that generality. ℍ is simply what the modular-forms consumers instantiate it at.