Documentation

TauCeti.Analysis.PDE.Regularity.Interior

Interior H² regularity for a constant principal coefficient #

Let A be a constant, uniformly elliptic coefficient matrix and let u ∈ H¹(Ω) be a weak solution of the divergence-form equation

-div(A ∇u) + ⟨b, ∇u⟩ + c u = f in Ω, with f ∈ L²(Ω) and b, c ∈ L∞(Ω),

meaning ∫_Ω (⟨∇v, A ∇u⟩ + ⟨b, ∇u⟩ v + c u v) = ∫_Ω f v for every v ∈ H¹₀(Ω). No boundary condition is imposed on u and nothing is assumed about ∂Ω. This file proves that u ∈ H²_loc(Ω): on every open V whose closure is compact and contained in Ω, the restriction of u is the first-order part of an element of W^{2,2}(V).

Localization #

The lower-order terms belong to L², so moving them to the right-hand side gives -div(A ∇u) = g with g = f - ⟨b, ∇u⟩ - c u ∈ L²(Ω). No sign or smallness condition on b or c is needed. The whole-space theorem TauCeti.PDE.UniformlyEllipticOn.exists_lowerOrder_eq does the analytic work; what is proved here is that a cutoff of a local solution is a global one. For ψ smooth and compactly supported in Ω, the product ψ u, extended by zero, lies in H¹(ℝⁿ) and solves

-div(A ∇(ψ u)) = ψ g - ⟨∇u, A ∇ψ⟩ - ⟨∇ψ, A ∇u⟩ - u div(A ∇ψ) on ℝⁿ,

whose right-hand side is again in L², because every term carrying u or ∇u also carries a bounded, compactly supported factor built from ψ. The identity is checked against test functions φ on ℝⁿ, which suffices by density (TauCeti.PDE.isWeakSolutionDirichlet_iff_forall_testFunction). The term ψ ∇u is handled by testing the equation for u against ψ φ ∈ H¹₀(Ω), and the term u ∇ψ by the definition of the weak derivative of u, tested against the components of φ A ∇ψ, which are test functions on Ω. Choosing ψ = 1 near closure V makes ψ u agree with u on V.

Main declarations #

References #

noncomputable def TauCeti.PDE.divMatrixGradient {ι : Type u_1} [Fintype ι] (A : Matrix ι ι ℝ) (ψ : EuclideanSpace ℝ ι → ℝ) (x : EuclideanSpace ℝ ι) :

The divergence div(A ∇ψ) = ∑ᵢ ∂ᵢ (A ∇ψ)ᵢ of the conormal field of ψ, computed in the standard basis. It is the zeroth-order coefficient that commuting the operator -div(A ∇ ·) past a cutoff ψ produces.

Equations
  • One or more equations did not get rendered due to their size.
Instances For
    @[simp]
    theorem TauCeti.PDE.divMatrixGradient_eq_zero_of_notMem_tsupport {ι : Type u_1} [Fintype ι] {A : Matrix ι ι ℝ} {ψ : EuclideanSpace ℝ ι → ℝ} (hψ : ContDiff ℝ (↑⊤) ψ) {x : EuclideanSpace ℝ ι} (hx : x ∉ tsupport ψ) :

    div(A ∇ψ) vanishes off the support of ψ.

    theorem TauCeti.PDE.continuous_divMatrixGradient {ι : Type u_1} [Fintype ι] {A : Matrix ι ι ℝ} {ψ : EuclideanSpace ℝ ι → ℝ} (hψ : ContDiff ℝ (↑⊤) ψ) :

    div(A ∇ψ) is continuous for smooth ψ.

    noncomputable def TauCeti.PDE.localizedForcing {ι : Type u_1} [Fintype ι] {mu : MeasureTheory.Measure (EuclideanSpace ℝ ι)} [mu.IsAddHaarMeasure] {Omega : TopologicalSpace.Opens (EuclideanSpace ℝ ι)} (A : Matrix ι ι ℝ) (ψ : EuclideanSpace ℝ ι → ℝ) (f : ↥(MeasureTheory.Lp ℝ 2 (mu.restrict ↑Omega))) (u : ↥(W1p mu Omega 2)) (x : EuclideanSpace ℝ ι) :

    The right-hand side of the equation satisfied by a cutoff ψ u of a weak solution u of -div(A ∇u) = f:

    -div(A ∇(ψ u)) = ψ f - ⟨∇u, A ∇ψ⟩ - ⟨∇ψ, A ∇u⟩ - u div(A ∇ψ).

    Equations
    • One or more equations did not get rendered due to their size.
    Instances For
      theorem TauCeti.PDE.memLp_localizedForcing {ι : Type u_1} [Fintype ι] {mu : MeasureTheory.Measure (EuclideanSpace ℝ ι)} [mu.IsAddHaarMeasure] {Omega : TopologicalSpace.Opens (EuclideanSpace ℝ ι)} {A : Matrix ι ι ℝ} {ψ : EuclideanSpace ℝ ι → ℝ} (hψ : ContDiff ℝ (↑⊤) ψ) (hψc : HasCompactSupport ψ) (f : ↥(MeasureTheory.Lp ℝ 2 (mu.restrict ↑Omega))) (u : ↥(W1p mu Omega 2)) :
      MeasureTheory.MemLp (localizedForcing A ψ f u) 2 (mu.restrict ↑Omega)

      The localized forcing term is square integrable, since each of its terms is an L²(Ω) function times a bounded one.

      theorem TauCeti.PDE.exists_isWeakSolutionDirichlet_extendByZeroL_contDiffSMul {ι : Type u_1} [Fintype ι] {mu : MeasureTheory.Measure (EuclideanSpace ℝ ι)} [mu.IsAddHaarMeasure] {Omega : TopologicalSpace.Opens (EuclideanSpace ℝ ι)} {A : Matrix ι ι ℝ} {f : ↥(MeasureTheory.Lp ℝ 2 (mu.restrict ↑Omega))} {u : ↥(W1p mu Omega 2)} (hu : ∀ (v : ↥(W1p0 mu Omega 2)), energyFormH1 (fun (x : EuclideanSpace ℝ ι) => A) 0 0 u ↑v = ∫ (x : EuclideanSpace ℝ ι) in ↑Omega, ↑↑f x * ↑↑(W1p.value ↑v) x ∂mu) {ψ : EuclideanSpace ℝ ι → ℝ} (hψ : ContDiff ℝ (↑⊤) ψ) (hψc : HasCompactSupport ψ) (hts : tsupport ψ ⊆ ↑Omega) :
      ∃ (w : ↥(W1p0 mu ⊤ 2)) (g : ↥(MeasureTheory.Lp ℝ 2 (mu.restrict ↑⊤))), IsWeakSolutionDirichlet (fun (x : EuclideanSpace ℝ ι) => A) 0 0 g w ∧ (∀ᵐ (x : EuclideanSpace ℝ ι) ∂mu, ↑↑g x = (↑Omega).indicator (localizedForcing A ψ f u) x) ∧ (∀ᵐ (x : EuclideanSpace ℝ ι) ∂mu, ↑↑(W1p.value ↑w) x = (↑Omega).indicator (fun (y : EuclideanSpace ℝ ι) => ψ y * ↑↑(W1p.value u) y) x) ∧ ∀ᵐ (x : EuclideanSpace ℝ ι) ∂mu, ↑↑(W1p.gradient ↑w) x = (↑Omega).indicator (fun (y : EuclideanSpace ℝ ι) => ψ y • ↑↑(W1p.gradient u) y + ↑↑(W1p.value u) y • gradient ψ y) x

      A cutoff of a weak solution solves an equation on the whole space. Let u ∈ H¹(Ω) be a weak solution of -div(A ∇u) = f in Ω for a constant matrix A, with f ∈ L²(Ω) and no boundary condition, and let ψ be smooth and compactly supported in Ω. Then ψ u, extended by zero, is a weak solution of -div(A ∇w) = g on the whole space for g ∈ L²(ℝⁿ) given almost everywhere by

      g = ψ f - ⟨∇u, A ∇ψ⟩ - ⟨∇ψ, A ∇u⟩ - u div(A ∇ψ), extended by zero.

      No ellipticity and no regularity of ∂Ω is needed.

      theorem TauCeti.PDE.exists_isWeakSolutionDirichlet_top_ae_eq_on_of_isCompact {ι : Type u_1} [Fintype ι] {mu : MeasureTheory.Measure (EuclideanSpace ℝ ι)} [mu.IsAddHaarMeasure] {Omega : TopologicalSpace.Opens (EuclideanSpace ℝ ι)} {A : Matrix ι ι ℝ} {f : ↥(MeasureTheory.Lp ℝ 2 (mu.restrict ↑Omega))} {u : ↥(W1p mu Omega 2)} (hu : ∀ (v : ↥(W1p0 mu Omega 2)), energyFormH1 (fun (x : EuclideanSpace ℝ ι) => A) 0 0 u ↑v = ∫ (x : EuclideanSpace ℝ ι) in ↑Omega, ↑↑f x * ↑↑(W1p.value ↑v) x ∂mu) {S : Set (EuclideanSpace ℝ ι)} (hS : IsCompact S) (hSΩ : S ⊆ ↑Omega) :
      ∃ (w : ↥(W1p0 mu ⊤ 2)) (g : ↥(MeasureTheory.Lp ℝ 2 (mu.restrict ↑⊤))), IsWeakSolutionDirichlet (fun (x : EuclideanSpace ℝ ι) => A) 0 0 g w ∧ (∀ᵐ (x : EuclideanSpace ℝ ι) ∂mu, x ∈ S → ↑↑(W1p.value ↑w) x = ↑↑(W1p.value u) x) ∧ (∀ᵐ (x : EuclideanSpace ℝ ι) ∂mu, x ∈ S → ↑↑(W1p.gradient ↑w) x = ↑↑(W1p.gradient u) x) ∧ ∀ᵐ (x : EuclideanSpace ℝ ι) ∂mu, x ∈ S → ↑↑g x = ↑↑f x

      Localizing a weak solution to the whole space. Let u ∈ H¹(Ω) be a weak solution of -div(A ∇u) = f in Ω for a constant matrix A, with f ∈ L²(Ω) and no boundary condition. Near any compact S ⊆ Ω, u agrees, in value and in gradient, with a weak solution w ∈ H¹(ℝⁿ) of an equation -div(A ∇w) = g on the whole space, with g ∈ L²(ℝⁿ) and g = f almost everywhere on S.

      One may take for w the product of u with a smooth cutoff equal to one near S and compactly supported in Ω, extended by zero (TauCeti.PDE.exists_isWeakSolutionDirichlet_extendByZeroL_contDiffSMul). This is the device that reduces interior regularity to regularity on the whole space.

      theorem TauCeti.PDE.UniformlyEllipticOn.exists_lowerOrder_eq_restrictL {ι : Type u_1} [Fintype ι] {mu : MeasureTheory.Measure (EuclideanSpace ℝ ι)} [mu.IsAddHaarMeasure] {Omega : TopologicalSpace.Opens (EuclideanSpace ℝ ι)} {A : Matrix ι ι ℝ} {lam : ℝ} (hlam : 0 < lam) (hA : ∀ (ξ : EuclideanSpace ℝ ι), lam * ‖ξ‖ ^ 2 ≤ ξ.ofLp ⬝ᵥ A.mulVec ξ.ofLp) {b : EuclideanSpace ℝ ι → EuclideanSpace ℝ ι} {c : EuclideanSpace ℝ ι → ℝ} (hb : MeasureTheory.MemLp b ⊤ (mu.restrict ↑Omega)) (hc : MeasureTheory.MemLp c ⊤ (mu.restrict ↑Omega)) {f : ↥(MeasureTheory.Lp ℝ 2 (mu.restrict ↑Omega))} {u : ↥(W1p mu Omega 2)} (hu : ∀ (v : ↥(W1p0 mu Omega 2)), energyFormH1 (fun (x : EuclideanSpace ℝ ι) => A) b c u ↑v = ∫ (x : EuclideanSpace ℝ ι) in ↑Omega, ↑↑f x * ↑↑(W1p.value ↑v) x ∂mu) {V : TopologicalSpace.Opens (EuclideanSpace ℝ ι)} (hV : IsCompact (closure ↑V)) (hVΩ : closure ↑V ⊆ ↑Omega) :
      ∃ (U : Wkp mu V 2 2), Wkp.lowerOrder 1 U = (W1p.restrictL ⋯) u

      Interior H² regularity for a constant principal coefficient. Let A be a constant, uniformly elliptic matrix, let b, c ∈ L∞(Ω), and let u ∈ H¹(Ω) be a weak solution of

      -∂ⱼ(Aⁱʲ ∂ᵢu) + ⟨b, ∇u⟩ + c u = f in Ω, with f ∈ L²(Ω),

      in the sense that ∫_Ω (⟨∇v, A ∇u⟩ + ⟨b, ∇u⟩ v + c u v) = ∫_Ω f v for every v ∈ H¹₀(Ω), with no boundary condition on u. Then u ∈ H²_loc(Ω): on every open V whose closure is compact and contained in Ω, the restriction of u is the first-order part of an element of W^{2,2}(V).

      No regularity of ∂Ω, sign of c, or smallness of the lower-order terms is assumed.