Documentation

TauCeti.Analysis.Sobolev.W1p.CompactSupport

Compactly supported Sobolev functions have zero boundary values #

W^{1,p}_0(Ω) is the closure of the test functions C_c^∞(Ω) in W^{1,p}(Ω), the Sobolev form of the homogeneous Dirichlet condition. Deciding that a given function lies in it is, in general, a boundary question. This file settles the case in which there is no boundary to answer for: if u ∈ W^{1,p}(Ω) vanishes almost everywhere outside a compact K ⊆ Ω, then

u ∈ W^{1,p}_0(Ω).

Only the value component is assumed to vanish; the gradient then vanishes off K by itself, because a Sobolev function that vanishes on an open set has vanishing weak gradient there (TauCeti.W1p.gradient_ae_eq_zero_of_value_ae_eq_zero).

The argument #

Extending u by zero gives a genuine element of W^{1,p}(ℝⁿ) (TauCeti.HasWeakFDerivOn.indicator_of_isCompact), which is where the compact support is used: for a general u ∈ W^{1,p}(Ω) the zero-extension need not be weakly differentiable at all. On the whole space every Sobolev function is a limit of test functions (TauCeti.W1p.mem_w1p0Submodule_top), but those test functions are supported anywhere in ℝⁿ, so they do not exhibit u as a limit of test functions on Ω. Multiplying by a cutoff χ which is 1 on K and compactly supported inside Ω repairs that: it leaves the extension of u unchanged, while carrying every test function on ℝⁿ into the image of C_c^∞(Ω). Since that image is closed — extension by zero is an isometry of W^{1,p}_0(Ω) onto its image — the limit stays in it, and restricting the isometry back gives u itself.

Consequences #

TauCeti.W1p.contDiffSMul_mem_w1p0Submodule_of_hasCompactSupport is the form localization arguments use: multiplying any u ∈ W^{1,p}(Ω) by a smooth cutoff compactly supported in Ω produces an element of W^{1,p}_0(Ω). The companion TauCeti.W1p.contDiffSMul_mem_w1p0Submodule needs u to lie in W^{1,p}_0(Ω) already, which is exactly what the compact support replaces here. This is the localization step of Meyers--Serrin density and of interior estimates.

Main declarations #

References #

L. C. Evans, Partial Differential Equations, §5.3.3; H. Brezis, Functional Analysis, Sobolev Spaces and Partial Differential Equations, Lemma 9.5.

Zero boundary values #

theorem TauCeti.W1p.mem_w1p0Submodule_of_isCompact {E : Type u_1} [MeasurableSpace E] [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] [BorelSpace E] {mu : MeasureTheory.Measure E} [mu.IsAddHaarMeasure] {Omega : TopologicalSpace.Opens E} {p : ENNReal} [Fact (1 ≤ p)] {K : Set E} (hp : p ≠ ⊤) {u : ↥(W1p mu Omega p)} (hK : IsCompact K) (hKO : K ⊆ ↑Omega) (hu : ∀ᵐ (x : E) ∂mu.restrict ↑Omega, x ∉ K → ↑↑(value u) x = 0) :
u ∈ w1p0Submodule mu Omega p

A compactly supported Sobolev function lies in W^{1,p}_0(Ω). If u ∈ W^{1,p}(Ω) vanishes almost everywhere outside a compact K ⊆ Ω, then u is a W^{1,p}-limit of test functions on Ω.

No regularity of ∂Ω is assumed, and none is needed: the hypothesis keeps u away from the boundary, so a cutoff can separate it from ∂Ω.

Multiplication by a compactly supported cutoff #

theorem TauCeti.W1p.contDiffSMul_mem_w1p0Submodule_of_hasCompactSupport {E : Type u_1} [MeasurableSpace E] [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] [BorelSpace E] {mu : MeasureTheory.Measure E} [mu.IsAddHaarMeasure] {Omega : TopologicalSpace.Opens E} {p : ENNReal} [Fact (1 ≤ p)] (hp : p ≠ ⊤) {psi : E → ℝ} (hpsi : ContDiff ℝ (↑⊤) psi) {M : ℝ} (hM : 0 ≤ M) (hpsiM : ∀ x ∈ Omega, |psi x| ≤ M) (hgradM : ∀ x ∈ Omega, ‖_root_.gradient psi x‖ ≤ M) (hcpt : HasCompactSupport psi) (hts : tsupport psi ⊆ ↑Omega) (u : ↥(W1p mu Omega p)) :
contDiffSMul psi hpsi hM hpsiM hgradM u ∈ w1p0Submodule mu Omega p

Multiplying by a cutoff compactly supported in Ω lands in W^{1,p}_0(Ω). Unlike TauCeti.W1p.contDiffSMul_mem_w1p0Submodule, nothing is assumed about the boundary behaviour of u: the support of ψ keeps the product away from ∂Ω. This is the localization device that turns a statement about W^{1,p}(Ω) into one about the test-function closure.

Localisation to the whole space #

theorem TauCeti.W1p.value_extendByZeroL_contDiffSMul_ae {E : Type u_1} [MeasurableSpace E] [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] [BorelSpace E] {mu : MeasureTheory.Measure E} [mu.IsAddHaarMeasure] {Omega : TopologicalSpace.Opens E} {p : ENNReal} [Fact (1 ≤ p)] {psi : E → ℝ} (hpsi : ContDiff ℝ (↑⊤) psi) {M : ℝ} (hM : 0 ≤ M) (hpsiM : ∀ x ∈ Omega, |psi x| ≤ M) (hgradM : ∀ x ∈ Omega, ‖_root_.gradient psi x‖ ≤ M) (u : ↥(W1p mu Omega p)) (hw : contDiffSMul psi hpsi hM hpsiM hgradM u ∈ w1p0Submodule mu Omega p) :
∀ᵐ (x : E) ∂mu, ↑↑(value ↑((W1p0.extendByZeroL ⋯) ⟨contDiffSMul psi hpsi hM hpsiM hgradM u, hw⟩)) x = (↑Omega).indicator (fun (y : E) => psi y * ↑↑(value u) y) x

The value of an extended cutoff product. For ψ smooth and u ∈ W^{1,p}(Ω) with ψ u ∈ W^{1,p}_0(Ω), the extension of ψ u by zero to the whole space is ψ u on Ω and vanishes off Ω.

theorem TauCeti.W1p.gradient_extendByZeroL_contDiffSMul_ae {E : Type u_1} [MeasurableSpace E] [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] [BorelSpace E] {mu : MeasureTheory.Measure E} [mu.IsAddHaarMeasure] {Omega : TopologicalSpace.Opens E} {p : ENNReal} [Fact (1 ≤ p)] {psi : E → ℝ} (hpsi : ContDiff ℝ (↑⊤) psi) {M : ℝ} (hM : 0 ≤ M) (hpsiM : ∀ x ∈ Omega, |psi x| ≤ M) (hgradM : ∀ x ∈ Omega, ‖_root_.gradient psi x‖ ≤ M) (u : ↥(W1p mu Omega p)) (hw : contDiffSMul psi hpsi hM hpsiM hgradM u ∈ w1p0Submodule mu Omega p) :
∀ᵐ (x : E) ∂mu, ↑↑(gradient ↑((W1p0.extendByZeroL ⋯) ⟨contDiffSMul psi hpsi hM hpsiM hgradM u, hw⟩)) x = (↑Omega).indicator (fun (y : E) => psi y • ↑↑(gradient u) y + ↑↑(value u) y • _root_.gradient psi y) x

The Leibniz rule for an extended cutoff product. For ψ smooth and u ∈ W^{1,p}(Ω) with ψ u ∈ W^{1,p}_0(Ω), the extension of ψ u by zero to the whole space has weak gradient ψ ∇u + u ∇ψ on Ω and 0 off Ω.

theorem TauCeti.W1p.exists_top_value_gradient_ae_eq_on_of_isCompact {E : Type u_1} [MeasurableSpace E] [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] [BorelSpace E] {mu : MeasureTheory.Measure E} [mu.IsAddHaarMeasure] {Omega : TopologicalSpace.Opens E} {p : ENNReal} [Fact (1 ≤ p)] (hp : p ≠ ⊤) (u : ↥(W1p mu Omega p)) {S : Set E} (hS : IsCompact S) (hSO : S ⊆ ↑Omega) :
∃ (w : ↥(W1p mu ⊤ p)), (∀ᵐ (x : E) ∂mu, x ∈ S → ↑↑(value w) x = ↑↑(value u) x) ∧ ∀ᵐ (x : E) ∂mu, x ∈ S → ↑↑(gradient w) x = ↑↑(gradient u) x

Localisation to the whole space. For 1 ≤ p < ∞, a Sobolev function u ∈ W^{1,p}(Ω) agrees, in value and in gradient, almost everywhere on any compact S ⊆ Ω with some w ∈ W^{1,p}(ℝⁿ). 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.