Documentation

LeanPool.EllipticPDE.Regularity.Localise.FiniteOrder

Localising coefficients of finite order on an open set #

Evans, Partial Differential Equations (2nd ed.), §6.3.1, Theorem 1 (p. 327) assumes a^{ij} ∈ C¹(U) and b^i, c ∈ L^∞(U), and Theorem 2 (p. 332) assumes a^{ij}, b^i, c ∈ C^{m+1}(U). Neither asks a bound on a^{ij} over U, and neither asks anything of a coefficient off U. Localise/LocalOp.lean blends coefficients smooth on U into a FullEllipticOp; this file does the same at finite order, and for lower-order coefficients that are only essentially bounded and almost everywhere strongly measurable on U.

Blend #

For a test function χ of U valued in [0, 1], the principal part is ã = λ I + χ (a − λ I), as in localOp. At order m the product χ (a − λ I) is C^m on the whole space with compact support, so every derivative up to order m is bounded (exists_iteratedFDeriv_bound_of_le), which is IsCkCoeff at order m. The lower-order coefficients are supplied already cut off. For L^∞(U) coefficients the cutoff multiplies a measurable modification (AEStronglyMeasurable.mk), which agrees with the coefficient almost everywhere on U; for C^m(U) coefficients it multiplies the coefficient itself.

Main declarations #

theorem EllipticPdes.Regularity.contDiff_mul_of_contDiffOn_of_le {d : ℕ} {U : Set (EuclideanSpace ℝ (Fin d))} (hU : IsOpen U) {χ g : EuclideanSpace ℝ (Fin d) → ℝ} (hχ : Sobolev.IsTestFn U χ) {n : ℕ∞} (hg : ContDiffOn ℝ (↑n) g U) :
ContDiff ℝ ↑n fun (x : EuclideanSpace ℝ (Fin d)) => χ x * g x

A function of class C^n on an open U, multiplied by a test function supported in U, is of class C^n on the whole space.

theorem EllipticPdes.Regularity.exists_iteratedFDeriv_bound_of_le {d : ℕ} {g : EuclideanSpace ℝ (Fin d) → ℝ} {m : ℕ} (hg : ContDiff ℝ (↑m) g) (hc : HasCompactSupport g) :
∃ (B : ℕ → ℝ), (∀ (j : ℕ), 0 ≤ B j) ∧ ∀ j ≤ m, ∀ (x : EuclideanSpace ℝ (Fin d)), ‖iteratedFDeriv ℝ j g x‖ ≤ B j

A compactly supported function of class C^m has every iterated derivative of order at most m bounded, with a nonnegative bound at each order.

A compactly supported function of class C^m lies in W^{m,∞}.

theorem EllipticPdes.Regularity.exists_sup_bound_continuous {d : ℕ} {g : EuclideanSpace ℝ (Fin d) → ℝ} (hg : Continuous g) (hc : HasCompactSupport g) :
∃ (C : ℝ), 0 ≤ C ∧ ∀ (x : EuclideanSpace ℝ (Fin d)), |g x| ≤ C

A continuous compactly supported function has a nonnegative uniform bound.

Principal part of finite order #

Principal part of class C^m on an open set. Entries of class C^m on U, uniformly elliptic almost everywhere on U with constant lam > 0, with no bound, no symmetry and nothing asked off U.

Instances For
    def EllipticPdes.Regularity.PrincipalOn.gA {d : ℕ} {U : Set (EuclideanSpace ℝ (Fin d))} {m : ℕ} (P : PrincipalOn d U m) (χ : EuclideanSpace ℝ (Fin d) → ℝ) (i j : Fin d) (x : EuclideanSpace ℝ (Fin d)) :

    The principal part shifted down to λ I, cut off by χ.

    Equations
    Instances For
      def EllipticPdes.Regularity.PrincipalOn.aT {d : ℕ} {U : Set (EuclideanSpace ℝ (Fin d))} {m : ℕ} (P : PrincipalOn d U m) (χ : EuclideanSpace ℝ (Fin d) → ℝ) (x : EuclideanSpace ℝ (Fin d)) (i j : Fin d) :

      The blended principal part λ I + χ (a − λ I).

      Equations
      Instances For
        theorem EllipticPdes.Regularity.PrincipalOn.gA_contDiff {d : ℕ} {U : Set (EuclideanSpace ℝ (Fin d))} {m : ℕ} (P : PrincipalOn d U m) (hU : IsOpen U) {χ : EuclideanSpace ℝ (Fin d) → ℝ} (hχ : Sobolev.IsTestFn U χ) (i j : Fin d) :
        ContDiff ℝ (↑m) (P.gA χ i j)
        theorem EllipticPdes.Regularity.PrincipalOn.gA_hasCompactSupport {d : ℕ} {U : Set (EuclideanSpace ℝ (Fin d))} {m : ℕ} (P : PrincipalOn d U m) {χ : EuclideanSpace ℝ (Fin d) → ℝ} (hχ : Sobolev.IsTestFn U χ) (i j : Fin d) :
        theorem EllipticPdes.Regularity.PrincipalOn.gA_continuous {d : ℕ} {U : Set (EuclideanSpace ℝ (Fin d))} {m : ℕ} (P : PrincipalOn d U m) (hU : IsOpen U) {χ : EuclideanSpace ℝ (Fin d) → ℝ} (hχ : Sobolev.IsTestFn U χ) (i j : Fin d) :
        Continuous (P.gA χ i j)
        theorem EllipticPdes.Regularity.PrincipalOn.quad_aT {d : ℕ} {U : Set (EuclideanSpace ℝ (Fin d))} {m : ℕ} (P : PrincipalOn d U m) (χ : EuclideanSpace ℝ (Fin d) → ℝ) (x : EuclideanSpace ℝ (Fin d)) (ξ : Fin d → ℝ) :
        ∑ i : Fin d, ∑ j : Fin d, P.aT χ x i j * ξ i * ξ j = P.lam * ∑ i : Fin d, ξ i ^ 2 + χ x * (∑ i : Fin d, ∑ j : Fin d, P.a x i j * ξ i * ξ j - P.lam * ∑ i : Fin d, ξ i ^ 2)

        The quadratic form of the blend, as a convex combination of λ |ξ|² and the quadratic form of a.

        theorem EllipticPdes.Regularity.PrincipalOn.aT_elliptic {d : ℕ} {U : Set (EuclideanSpace ℝ (Fin d))} {m : ℕ} (P : PrincipalOn d U m) {χ : EuclideanSpace ℝ (Fin d) → ℝ} (hχ : Sobolev.IsTestFn U χ) (hχ01 : ∀ (x : EuclideanSpace ℝ (Fin d)), χ x ∈ Set.Icc 0 1) (hUm : MeasurableSet U) :
        ∀ᵐ (x : EuclideanSpace ℝ (Fin d)), ∀ (ξ : Fin d → ℝ), P.lam * ∑ i : Fin d, ξ i ^ 2 ≤ ∑ i : Fin d, ∑ j : Fin d, P.aT χ x i j * ξ i * ξ j

        Uniform ellipticity of the blend on the whole space at the constant λ.

        noncomputable def EllipticPdes.Regularity.blendOp {d : ℕ} {U : Set (EuclideanSpace ℝ (Fin d))} {m : ℕ} (P : PrincipalOn d U m) (hU : IsOpen U) {χ : EuclideanSpace ℝ (Fin d) → ℝ} (hχ : Sobolev.IsTestFn U χ) (hχ01 : ∀ (x : EuclideanSpace ℝ (Fin d)), χ x ∈ Set.Icc 0 1) (b : EuclideanSpace ℝ (Fin d) → Fin d → ℝ) (c : EuclideanSpace ℝ (Fin d) → ℝ) {Bs Cs : ℝ} (hBs : 0 ≤ Bs) (hCs : 0 ≤ Cs) (hbm : ∀ (i : Fin d), Measurable fun (x : EuclideanSpace ℝ (Fin d)) => b x i) (hcm : Measurable c) (hbb : ∀ (i : Fin d), ∀ᵐ (x : EuclideanSpace ℝ (Fin d)), |b x i| ≤ Bs) (hcb : ∀ᵐ (x : EuclideanSpace ℝ (Fin d)), |c x| ≤ Cs) :

        Blended operator at finite order. Principal part λ I + χ (a − λ I), and lower-order coefficients b, c supplied measurable and essentially bounded on the whole space.

        Equations
        • One or more equations did not get rendered due to their size.
        Instances For
          theorem EllipticPdes.Regularity.blendOp_isCkCoeff {d : ℕ} {U : Set (EuclideanSpace ℝ (Fin d))} {m : ℕ} (P : PrincipalOn d U m) (hU : IsOpen U) {χ : EuclideanSpace ℝ (Fin d) → ℝ} (hχ : Sobolev.IsTestFn U χ) (hχ01 : ∀ (x : EuclideanSpace ℝ (Fin d)), χ x ∈ Set.Icc 0 1) (b : EuclideanSpace ℝ (Fin d) → Fin d → ℝ) (c : EuclideanSpace ℝ (Fin d) → ℝ) {Bs Cs : ℝ} (hBs : 0 ≤ Bs) (hCs : 0 ≤ Cs) (hbm : ∀ (i : Fin d), Measurable fun (x : EuclideanSpace ℝ (Fin d)) => b x i) (hcm : Measurable c) (hbb : ∀ (i : Fin d), ∀ᵐ (x : EuclideanSpace ℝ (Fin d)), |b x i| ≤ Bs) (hcb : ∀ᵐ (x : EuclideanSpace ℝ (Fin d)), |c x| ≤ Cs) :
          Nonempty (IsCkCoeff (blendOp P hU hχ hχ01 b c hBs hCs hbm hcm hbb hcb).toEllipticCoeff m)

          C^m bundle of the blend. Each entry is a constant plus a compactly supported function of class C^m, so its derivatives of order 1 to m are bounded.

          theorem EllipticPdes.Regularity.blendOp_a_eq {d : ℕ} {U : Set (EuclideanSpace ℝ (Fin d))} {m : ℕ} (P : PrincipalOn d U m) (hU : IsOpen U) {χ : EuclideanSpace ℝ (Fin d) → ℝ} (hχ : Sobolev.IsTestFn U χ) (hχ01 : ∀ (x : EuclideanSpace ℝ (Fin d)), χ x ∈ Set.Icc 0 1) (b : EuclideanSpace ℝ (Fin d) → Fin d → ℝ) (c : EuclideanSpace ℝ (Fin d) → ℝ) {Bs Cs : ℝ} (hBs : 0 ≤ Bs) (hCs : 0 ≤ Cs) (hbm : ∀ (i : Fin d), Measurable fun (x : EuclideanSpace ℝ (Fin d)) => b x i) (hcm : Measurable c) (hbb : ∀ (i : Fin d), ∀ᵐ (x : EuclideanSpace ℝ (Fin d)), |b x i| ≤ Bs) (hcb : ∀ᵐ (x : EuclideanSpace ℝ (Fin d)), |c x| ≤ Cs) {x : EuclideanSpace ℝ (Fin d)} (hx : χ x = 1) :
          (blendOp P hU hχ hχ01 b c hBs hCs hbm hcm hbb hcb).a x = P.a x

          The blend is the given principal part wherever the cutoff is one.

          theorem EllipticPdes.Regularity.exists_cutoff_interior_one {d : ℕ} {U K : Set (EuclideanSpace ℝ (Fin d))} (hU : IsOpen U) (hK : IsCompact K) (hKU : K ⊆ U) :
          ∃ (χ : EuclideanSpace ℝ (Fin d) → ℝ) (_ : Sobolev.IsTestFn U χ) (_ : ∀ (x : EuclideanSpace ℝ (Fin d)), χ x ∈ Set.Icc 0 1) (W : Set (EuclideanSpace ℝ (Fin d))), IsOpen W ∧ K ⊆ W ∧ W ⊆ U ∧ ∀ x ∈ W, χ x = 1

          The cutoff of exists_isTestFn_one_nhdsSet_of_isCompact and the open interior of the set where it is one, which contains the compact set and sits inside U.

          Coefficients of Theorem 1 #

          Coefficients of Evans §6.3.1 Theorem 1 on an open set. a^{ij} ∈ C¹(U), uniformly elliptic almost everywhere on U, and b^i, c ∈ L^∞(U): almost everywhere strongly measurable and essentially bounded on U. Nothing is asked off U.

          Instances For

            A cutoff of U times a measurable modification of a function essentially bounded on U is measurable and essentially bounded on the whole space.

            theorem EllipticPdes.Regularity.exists_localOp_C1 {d : ℕ} {U : Set (EuclideanSpace ℝ (Fin d))} (P : C1OpOn d U) (hU : IsOpen U) {K : Set (EuclideanSpace ℝ (Fin d))} (hK : IsCompact K) (hKU : K ⊆ U) :
            ∃ (Op : Sobolev.FullEllipticOp d) (W : Set (EuclideanSpace ℝ (Fin d))), IsOpen W ∧ K ⊆ W ∧ W ⊆ U ∧ Nonempty (IsC1Coeff Op.toEllipticCoeff) ∧ (∀ x ∈ W, Op.a x = P.a x) ∧ (∀ (i : Fin d), ∀ᵐ (x : EuclideanSpace ℝ (Fin d)) ∂MeasureTheory.volume.restrict W, Op.b x i = P.b x i) ∧ ∀ᵐ (x : EuclideanSpace ℝ (Fin d)) ∂MeasureTheory.volume.restrict W, Op.c x = P.c x

            Localisation of the coefficients of Theorem 1. For every compact K ⊆ U there are a global operator with C¹ principal part of bounded derivative and an open W with K ⊆ W ⊆ U, on which the principal part agrees with the given one everywhere and the lower-order coefficients agree with the given ones almost everywhere.

            Coefficients of Theorem 2 #

            Coefficients of Evans §6.3.1 Theorem 2 on an open set at order m. a^{ij}, b^i, c of class C^m on U, with a^{ij} uniformly elliptic almost everywhere on U. Evans's hypothesis at order m is this structure at m + 1. Nothing is asked off U.

            Instances For
              theorem EllipticPdes.Regularity.exists_localOp_Ck {d : ℕ} {U : Set (EuclideanSpace ℝ (Fin d))} {m : ℕ} (P : CkOpOn d U m) (hU : IsOpen U) {K : Set (EuclideanSpace ℝ (Fin d))} (hK : IsCompact K) (hKU : K ⊆ U) :
              ∃ (Op : Sobolev.FullEllipticOp d) (W : Set (EuclideanSpace ℝ (Fin d))), IsOpen W ∧ K ⊆ W ∧ W ⊆ U ∧ Nonempty (IsCkCoeff Op.toEllipticCoeff m) ∧ Nonempty (IsWkInftyLower Op m) ∧ (∀ x ∈ W, Op.a x = P.a x) ∧ (∀ (i : Fin d), ∀ᵐ (x : EuclideanSpace ℝ (Fin d)) ∂MeasureTheory.volume.restrict W, Op.b x i = P.b x i) ∧ ∀ᵐ (x : EuclideanSpace ℝ (Fin d)) ∂MeasureTheory.volume.restrict W, Op.c x = P.c x

              Localisation of the coefficients of Theorem 2. For every compact K ⊆ U there are a global operator with every coefficient of class C^m with bounded derivatives, packaged as IsCkCoeff and IsWkInftyLower at order m, and an open W with K ⊆ W ⊆ U on which it agrees with the given coefficients.

              Transport of the weak formulation along almost everywhere equal coefficients #

              theorem EllipticPdes.Regularity.LocalWeakSol.congr_ae {d : ℕ} {W : Set (EuclideanSpace ℝ (Fin d))} {a a' : EuclideanSpace ℝ (Fin d) → Fin d → Fin d → ℝ} {b b' : EuclideanSpace ℝ (Fin d) → Fin d → ℝ} {c c' f f' u : EuclideanSpace ℝ (Fin d) → ℝ} {G : Fin d → EuclideanSpace ℝ (Fin d) → ℝ} (ha : ∀ (i j : Fin d), ∀ᵐ (x : EuclideanSpace ℝ (Fin d)) ∂MeasureTheory.volume.restrict W, a x i j = a' x i j) (hb : ∀ (i : Fin d), ∀ᵐ (x : EuclideanSpace ℝ (Fin d)) ∂MeasureTheory.volume.restrict W, b x i = b' x i) (hc : ∀ᵐ (x : EuclideanSpace ℝ (Fin d)) ∂MeasureTheory.volume.restrict W, c x = c' x) (hf : ∀ᵐ (x : EuclideanSpace ℝ (Fin d)) ∂MeasureTheory.volume.restrict W, f x = f' x) (h : LocalWeakSol W a b c f u G) :
              LocalWeakSol W a' b' c' f' u G

              Coefficients equal on W almost everywhere, with a principal part equal everywhere on W, give the same local weak formulation on W.