Documentation

TauCeti.Analysis.BoundedVariation.LebesgueDecomposition

Lebesgue decomposition of the BV derivative #

For a locally integrable function of locally finite variation, write Du = σ |Du| using polarDensity. Relative to a σ-finite reference measure ν on the open domain, the absolutely continuous part of Du has density acDerivativeDensity μ Ω u ν = (d|Du|/dν) σ. The remaining part is σ (|Du|.singularPart ν). We prove the distributional identity for their sum and the splitting of the variation measure, including local integrability of the density. Taking ν = μ.comap Subtype.val gives the usual absolutely continuous/singular decomposition on the domain. No global finiteness of |Du| or integrability of u is required.

The variation measure satisfies |Du| = ν.withDensity (fun x => ‖acDerivativeDensity μ Ω u ν x‖ₑ) + |Du|.singularPart ν (withDensity_enorm_acDerivativeDensity_add_singularPart). The second measure is mutually singular to ν by Mathlib's Measure.mutuallySingular_singularPart. The distributional identity describes the derivative through its action on test fields; no globally finite vector measure is imposed on an unbounded domain. The density is unique ν-almost everywhere among locally integrable densities in such a decomposition (ae_eq_acDerivativeDensity), including when the competing singular reference measure is not σ-finite.

References #

The density of the absolutely continuous part of Du relative to ν on Ω. It is the Radon–Nikodym density of |Du| multiplied by its polar field.

Instances For

    The defining formula for the absolutely continuous derivative density.

    The absolutely continuous derivative density is measurable.

    theorem TauCeti.ae_enorm_acDerivativeDensity_eq_rnDeriv {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] [MeasurableSpace E] [BorelSpace E] {μ : MeasureTheory.Measure E} {Ω : TopologicalSpace.Opens E} {u : E → ℝ} (hu : MeasureTheory.LocallyIntegrableOn u (↑Ω) μ) (hfin : ∀ (K : Set E), IsCompact K → K ⊆ ↑Ω → ∃ (U : TopologicalSpace.Opens E), K ⊆ ↑U ∧ totalVariationOn μ U u ≠ ⊤) (ν : MeasureTheory.Measure ↥Ω) :
    (fun (x : ↥Ω) => ‖acDerivativeDensity μ Ω u ν x‖ₑ) =ᵐ[ν] (variationMeasure μ Ω u).rnDeriv ν

    The norm of the absolutely continuous derivative density is exactly d|Du|/dν, including where this Radon–Nikodym derivative vanishes.

    The variation measure splits into the norm of the absolutely continuous derivative and the singular variation measure. The latter is mutually singular to ν.

    theorem TauCeti.lintegral_enorm_acDerivativeDensity_add_singularPart {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] [MeasurableSpace E] [BorelSpace E] {μ : MeasureTheory.Measure E} {Ω : TopologicalSpace.Opens E} {u : E → ℝ} (hu : MeasureTheory.LocallyIntegrableOn u (↑Ω) μ) (hfin : ∀ (K : Set E), IsCompact K → K ⊆ ↑Ω → ∃ (U : TopologicalSpace.Opens E), K ⊆ ↑U ∧ totalVariationOn μ U u ≠ ⊤) (ν : MeasureTheory.Measure ↥Ω) [MeasureTheory.SigmaFinite ν] (s : Set ↥Ω) :
    ∫⁻ (x : ↥Ω) in s, ‖acDerivativeDensity μ Ω u ν x‖ₑ ∂ν + ((variationMeasure μ Ω u).singularPart ν) s = (variationMeasure μ Ω u) s

    On every subset, variation is the integral of the norm of the absolutely continuous derivative plus the singular variation, as an extended-valued identity.

    The absolutely continuous derivative density is locally integrable even when the whole derivative has infinite variation.

    Finite total variation implies integrability of the absolutely continuous derivative density. Global integrability of the original function is not required.

    theorem TauCeti.integral_divergenceCLM_mul_eq_neg_integral_inner_acDerivativeDensity_sub_singularPart {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] [MeasurableSpace E] [BorelSpace E] {μ : MeasureTheory.Measure E} {Ω : TopologicalSpace.Opens E} {u : E → ℝ} (hu : MeasureTheory.LocallyIntegrableOn u (↑Ω) μ) (hfin : ∀ (K : Set E), IsCompact K → K ⊆ ↑Ω → ∃ (U : TopologicalSpace.Opens E), K ⊆ ↑U ∧ totalVariationOn μ U u ≠ ⊤) (ν : MeasureTheory.Measure ↥Ω) [MeasureTheory.SigmaFinite ν] (φ : TestFunction Ω E ⊤) :
    ∫ (x : E), (TestFunction.divergenceCLM φ) x * u x ∂μ = -∫ (x : ↥Ω), inner ℝ (φ ↑x) (acDerivativeDensity μ Ω u ν x) ∂ν - ∫ (x : ↥Ω), inner ℝ (φ ↑x) (polarDensity μ Ω u x) ∂(variationMeasure μ Ω u).singularPart ν

    Lebesgue decomposition of the BV derivative. The distributional derivative is the sum of the density acDerivativeDensity μ Ω u ν against ν and the polar field against the singular part of |Du|. Both pairings are integrable because the test field has compact support.

    theorem TauCeti.ae_eq_acDerivativeDensity {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] [MeasurableSpace E] [BorelSpace E] {μ : MeasureTheory.Measure E} {Ω : TopologicalSpace.Opens E} {u : E → ℝ} (hu : MeasureTheory.LocallyIntegrableOn u (↑Ω) μ) (hfin : ∀ (K : Set E), IsCompact K → K ⊆ ↑Ω → ∃ (U : TopologicalSpace.Opens E), K ⊆ ↑U ∧ totalVariationOn μ U u ≠ ⊤) (ν : MeasureTheory.Measure ↥Ω) [MeasureTheory.SigmaFinite ν] {g τ : ↥Ω → E} {ξ : MeasureTheory.Measure ↥Ω} (hg : MeasureTheory.LocallyIntegrable g ν) (hτ : MeasureTheory.LocallyIntegrable τ ξ) (hξ : ξ.MutuallySingular ν) (h : ∀ (φ : TestFunction Ω E ⊤), ∫ (x : E), (TestFunction.divergenceCLM φ) x * u x ∂μ = -∫ (x : ↥Ω), inner ℝ (φ ↑x) (g x) ∂ν - ∫ (x : ↥Ω), inner ℝ (φ ↑x) (τ x) ∂ξ) :
    g =ᵐ[ν] acDerivativeDensity μ Ω u ν

    Uniqueness of the absolutely continuous derivative density. Any locally integrable g representing Du together with a locally integrable density against a measure singular to ν agrees with acDerivativeDensity at ν-almost every point.

    theorem TauCeti.acDerivativeDensity_congr_ae {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] [MeasurableSpace E] [BorelSpace E] {μ : MeasureTheory.Measure E} {Ω : TopologicalSpace.Opens E} {u : E → ℝ} (ν : MeasureTheory.Measure ↥Ω) {v : E → ℝ} (h : u =ᵐ[μ.restrict ↑Ω] v) :
    acDerivativeDensity μ Ω u ν =ᵐ[ν] acDerivativeDensity μ Ω v ν

    The absolutely continuous derivative density is unchanged almost everywhere if the original function is changed on a null subset of its domain.