Distributions and Convolution #
This file defines distributions on totally disconnected locally compact Hausdorff spaces, compactly supported distributions on td groups, and convolution with delta and Haar distributions.
A distribution on X is a ℂ-linear functional on the space of compactly supported locally constant functions.
- toFun : CptSupportLocConst_td X → ℂ
- map_smul' (m : ℂ) (x : CptSupportLocConst_td X) : self.toFun (m • x) = (RingHom.id ℂ) m • self.toFun x
Instances For
Equations
- Distribution_td.instFunLikeCptSupportLocConst_tdComplex = { coe := fun (F : Distribution_td X) => ⇑F.toLinearMap, coe_injective' := ⋯ }
Distributions are additive.
Distributions are ℂ-linear.
Extend a locally constant function on a subset Z to all of X by setting it to zero outside Z.
Equations
- Distribution_td.extendByZero Z f _hf_lc x = if h : x ∈ Z then f ⟨x, h⟩ else 0
Instances For
The extension by zero of a locally constant function is locally constant.
The extension by zero has compact support if Z is closed and compact.
Restriction of a distribution F on X to a subset Z. The restriction sends a function f on Z to F applied to the extension of f by zero.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The support of a distribution F is the smallest closed set S such that the restriction of F to X \ S is zero.
Equations
Instances For
The support of a distribution is closed.
A distribution vanishes on test functions whose support is disjoint from its support.
Functions and Distributions on a td group #
Translations of functions on a td group
Right translation of a function by g: (R_g φ)(x) = φ(xg).
Equations
- rightTranslate φ g x = φ (x * g)
Instances For
Left translation of a function by g: (L_g φ)(x) = φ(g⁻¹x).
Equations
- leftTranslate φ g x = φ (g⁻¹ * x)
Instances For
Right translation preserves local constancy.
Left translation preserves local constancy.
Right translation preserves compact support.
Left translation preserves compact support.
Right translation of a compactly supported locally constant function.
Equations
- φ.rightTranslate g = { toFun := rightTranslate (⇑φ) g, locallyConstant' := ⋯, hasCompactSupport' := ⋯ }
Instances For
Left translation of a compactly supported locally constant function.
Equations
- φ.leftTranslate g = { toFun := leftTranslate (⇑φ) g, locallyConstant' := ⋯, hasCompactSupport' := ⋯ }
Instances For
Distributions on a td group
A compactly supported distribution on G.
- toDistribution_td : Distribution_td G
- hasCompactSupport' : IsCompact self.toDistribution_td.support
Instances For
Equations
- CptSupportDistribution_td.instFunLikeCptSupportLocConst_tdComplex = { coe := fun (F : CptSupportDistribution_td G) => F.toDistribution_td.toFun, coe_injective' := ⋯ }
Zero compactly supported distribution.
Equations
- One or more equations did not get rendered due to their size.
For a test function φ and distribution D, the function x ↦ D(R_x φ) where R_x is right translation. This function is used to define convolution.
Equations
- D.convolveTestFun φ x = D (φ.rightTranslate x)
Instances For
The function x ↦ D(R_x φ) is locally constant.
The function x ↦ D(R_x φ) has compact support when D has compact support.
The function x ↦ D(R_x φ) as a compactly supported locally constant function.
Equations
- F.convolveTestFun' φ = { toFun := F.convolveTestFun φ, locallyConstant' := ⋯, hasCompactSupport' := ⋯ }
Instances For
Convolution of two compactly supported distributions. (F * D)(φ) = F(x ↦ D(R_x φ)) where R_x is right translation by x.
Equations
- F.conv D = { toDistribution_td := { toFun := fun (φ : CptSupportLocConst_td G) => F (D.convolveTestFun' φ), map_add' := ⋯, map_smul' := ⋯ }, hasCompactSupport' := ⋯ }
Instances For
Convolution is associative.
The delta distribution at g ∈ G: δ_g(φ) = φ(g).
Equations
- CptSupportDistribution_td.deltaDistribution g = { toDistribution_td := { toFun := fun (φ : CptSupportLocConst_td G) => φ g, map_add' := ⋯, map_smul' := ⋯ }, hasCompactSupport' := ⋯ }
Instances For
The characteristic function of a clopen compact set as a compactly supported locally constant function.
Equations
- CptSupportDistribution_td.indicatorFunction S hS_clopen hS_compact = { toFun := S.indicator fun (x : G) => 1, locallyConstant' := ⋯, hasCompactSupport' := ⋯ }
Instances For
The normalized Haar measure e_H for a compact open subgroup H. For a test function φ, e_H(φ) is the average value of φ over H with respect to the normalized Haar measure on the compact group H.
Equations
- One or more equations did not get rendered due to their size.
Instances For
e_H applied to a test function.
Properties of Haar distribution and delta functions #
The normalized Haar distribution e_H is idempotent under convolution: e_H * e_H = e_H.
Convolution of delta function with normalized Haar distribution: δ_g * e_H(φ) = e_H(R_g φ) for any g ∈ G.
Convolution of normalized Haar distribution with delta function: e_H * δ_g(φ) = e_H(L_{g⁻¹} φ) for any g ∈ G.
The proof uses that (e_H * δ_g)(φ) = e_H(x ↦ δ_g(R_x φ)) = e_H(x ↦ φ(g * x)) and (L_{g⁻¹} φ)(h) = φ(g * h), so the integrands match.
For g ∈ H, δ_g * e_H = e_H * δ_g. This uses the fact that e_H is bi-H-invariant.
Translation actions on distributions #
Left translation of a distribution by g: (g · F)(φ) = F(L_{g⁻¹} φ). This gives a left action: g · (h · F) = (gh) · F.
Equations
- F.leftTranslate g = { toDistribution_td := have __LinearMap := CptSupportDistribution_td.leftTranslate_linearMap✝ F g; { toLinearMap := __LinearMap }, hasCompactSupport' := ⋯ }
Instances For
Right translation of a distribution by g: (F · g)(φ) = F(R_g φ). This gives a right action: (F · g) · h = F · (gh).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Left translation is a left action: g · (h · F) = (gh) · F.
Right translation is a right action: (F · g) · h = F · (gh).
A distribution is left H-invariant if h · F = F for all h ∈ H.
Equations
- F.IsLeftInvariant H = ∀ h ∈ H, F.leftTranslate h = F
Instances For
A distribution is right H-invariant if F · h = F for all h ∈ H.
Equations
- F.IsRightInvariant H = ∀ h ∈ H, F.rightTranslate h = F
Instances For
A distribution is bi-H-invariant if it is both left and right H-invariant.
Equations
- F.IsBiInvariant H = (F.IsLeftInvariant H ∧ F.IsRightInvariant H)
Instances For
A compactly supported distribution is locally constant if there exists a compact open subgroup H such that F is bi-H-invariant.
Equations
- F.IsLocallyConstant = ∃ (H : OpenSubgroup G), IsCompact ↑H ∧ F.IsBiInvariant H
Instances For
A compactly supported distribution is left locally constant if there exists a compact open subgroup H such that F is left-H-invariant.
Equations
- F.IsLeftLocallyConstant = ∃ (H : OpenSubgroup G), IsCompact ↑H ∧ F.IsLeftInvariant H
Instances For
A compactly supported distribution is left locally constant if there exists a compact open subgroup H such that F is right-H-invariant.
Equations
- F.IsRightLocallyConstant = ∃ (H : OpenSubgroup G), IsCompact ↑H ∧ F.IsRightInvariant H
Instances For
The normalized Haar distribution e_H is left H-invariant.
The normalized Haar distribution e_H is right H-invariant.
The normalized Haar distribution e_H is bi-H-invariant.
Right cosets of an open subgroup are clopen.
Right cosets of a compact subgroup are compact.
A compact set is covered by finitely many right cosets of an open subgroup.
The indicator function of a right coset.
Equations
- CptSupportDistribution_td.rightCosetIndicator H hH_compact g = CptSupportDistribution_td.indicatorFunction ((fun (h : G) => h * g) '' ↑H) ⋯ ⋯
Instances For
The average value of a test function on a right coset Hg, computed using the normalized Haar measure on H.
Equations
- CptSupportDistribution_td.cosetAverage H hH_compact g φ = ∫ (h : ↥H), φ (↑h * g) ∂MeasureTheory.Measure.haarMeasure ⊤
Instances For
Left H-invariance implies F(L_{h⁻¹} φ) = F(φ) for h ∈ H.
For a left H-invariant distribution, the integral of F(L_h φ) over h ∈ H equals F(φ).
For a left H-invariant distribution F and φ supported on Hg with zero average, F(φ) = 0. This is the key technical lemma for the coset decomposition.
For a bi-H-invariant distribution F and the indicator function of a right coset Hg, the value F(χ_{Hg}) determines F's action on functions supported on that coset.
Structure theorem: Any H-invariant compactly supported distribution can be written as a finite sum of e_H * δ_{g_i} for some g_i ∈ G.
If F is bi-H-invariant and compactly supported, then there exist g₁, ..., gₖ ∈ G and a₁, ..., aₖ ∈ ℂ such that F = ∑ᵢ aᵢ (e_H * δ_{gᵢ}).
Left locally constant implies right locally constant.
Right locally constant implies left locally constant.
This is symmetric to IsLeftLocallyConstant.isRightLocallyConstant.