Documentation

TauCeti.Analysis.PDE.Regularity.Inhomogeneous

Hölder continuity of weak solutions with a right-hand side #

Let a be measurable and uniformly elliptic on an open set Ω ⊆ ℝⁿ, n ≥ 1, with constants 0 < λ ≤ Λ, and let f ∈ Lʳ with r > 1 and r > n/2. This file extends De Giorgi's theorem TauCeti.PDE.exists_holderOnWith_preciseRepresentative from -∂ᵢ(aⁱʲ ∂ⱼu) = 0 to weak solutions u ∈ H¹(Ω) of

-∂ᵢ(aⁱʲ ∂ⱼu) = f in Ω, that is a(u, v) = ∫_Ω f v for every v ∈ H¹₀(Ω).

Such a solution is locally bounded and has a locally Hölder continuous representative, with an exponent depending only on λ, Λ, n, r and the normalization of the Haar measure, and with constants controlled by ‖u‖_{L²} and ‖f‖_{Lʳ}/λ. No regularity of the coefficients beyond measurability is assumed.

The right-hand side is split off one ball at a time. On B = B(x₀, ρ) ⊆ Ω, the Dirichlet problem -∂ᵢ(aⁱʲ ∂ⱼw) = f in B, w ∈ H¹₀(B), has a solution (TauCeti.PDE.UniformlyEllipticOn.existsUnique_forall_energyFormH1_eq_setIntegral_of_memLp), and Stampacchia's bound TauCeti.PDE.exists_ae_abs_value_le_mul_div_mul_rpow gives |w| ≤ C ρ^(2 - n/r) ‖f‖_{Lʳ(B)} / λ. The difference h = u - w solves the homogeneous equation on B, so De Giorgi's local boundedness and oscillation decay apply to it. The error w is small on small balls exactly because 2 - n/r > 0. Comparing the essential oscillation of u on concentric balls with that of h gives

osc_{B(x₀, s)} u ≤ C₀ (s/ρ)^α₀ osc_{B(x₀, ρ)} u + C ρ^(2 - n/r) ‖f‖_{Lʳ} / λ,

and the iteration lemma TauCeti.exists_le_mul_div_rpow_mul_add_mul_rpow_of_le_mul_add turns this into a power law with exponent min (2 - n/r) (α₀/2). The precise representative then converts the power law into a Hölder bound, as for the homogeneous equation.

Applied on small balls, this gives continuity of weak solutions whose right-hand side is only locally in Lʳ, tested against test functions; this is the form in which semilinear equations, whose right-hand side depends on the solution, are met. On a general finite-dimensional inner product space the Laplacian case follows by orthonormal coordinates.

Main declarations #

All in the namespace TauCeti.PDE.

References #

Splitting off the right-hand side on a ball #

theorem TauCeti.PDE.exists_forall_energyFormH1_eq_zero_ae_abs_sub_le {ι : Type u_1} [Fintype ι] [DecidableEq ι] {mu : MeasureTheory.Measure (EuclideanSpace ℝ ι)} [mu.IsAddHaarMeasure] [Nonempty ι] {r : NNReal} (hr : 1 < r) (hrn : ↑(Fintype.card ι) < 2 * ↑r) :
∃ (K : ℝ), 0 < K ∧ ∀ {Omega : TopologicalSpace.Opens (EuclideanSpace ℝ ι)} {a : EuclideanSpace ℝ ι → Matrix ι ι ℝ} {lam Lam : ℝ} {f : EuclideanSpace ℝ ι → ℝ} {u : ↥(W1p mu Omega 2)} {x₀ : EuclideanSpace ℝ ι} {ρ : ℝ}, UniformlyEllipticOn (↑Omega) a lam Lam → MeasureTheory.AEStronglyMeasurable a (mu.restrict ↑Omega) → MeasureTheory.MemLp f (↑r) (mu.restrict (Metric.ball x₀ ρ)) → (∀ (v : ↥(W1p0 mu Omega 2)), energyFormH1 a 0 0 u ↑v = ∫ (x : EuclideanSpace ℝ ι) in ↑Omega, f x * ↑↑(W1p.value ↑v) x ∂mu) → 0 < ρ → Metric.ball x₀ ρ ⊆ ↑Omega → ∃ (h : ↥(W1p mu { carrier := Metric.ball x₀ ρ, is_open' := ⋯ } 2)), (∀ (v : ↥(W1p0 mu { carrier := Metric.ball x₀ ρ, is_open' := ⋯ } 2)), energyFormH1 a 0 0 h ↑v = 0) ∧ ∀ᵐ (x : EuclideanSpace ℝ ι) ∂mu.restrict (Metric.ball x₀ ρ), |↑↑(W1p.value u) x - ↑↑(W1p.value h) x| ≤ K * ρ ^ (2 - ↑(Fintype.card ι) / ↑r) * (MeasureTheory.eLpNorm f (↑r) (mu.restrict (Metric.ball x₀ ρ))).toReal / lam

Splitting off the right-hand side on a ball. Let n ≥ 1 be the dimension and let r > 1 with r > n/2. There is K > 0, depending only on n, r and the normalization of mu, such that the following holds. Let a be measurable and uniformly elliptic on Ω with constants λ, Λ, and let u ∈ H¹(Ω) be a weak solution of -∂ᵢ(aⁱʲ ∂ⱼu) = f, that is a(u, v) = ∫_Ω f v for every v ∈ H¹₀(Ω), with f ∈ Lʳ(B(x₀, ρ)) and B(x₀, ρ) ⊆ Ω. Then on B = B(x₀, ρ) there is a weak solution h ∈ H¹(B) of the homogeneous equation -∂ᵢ(aⁱʲ ∂ⱼh) = 0 with

|u - h| ≤ K ρ^(2 - n/r) ‖f‖_{Lʳ(B)} / λ almost everywhere on B.

This transfers estimates for the homogeneous equation to equations with a right-hand side.

Local boundedness #

theorem TauCeti.PDE.exists_ae_abs_value_le_mul_add_of_memLp {ι : Type u_1} [Fintype ι] [DecidableEq ι] {mu : MeasureTheory.Measure (EuclideanSpace ℝ ι)} [mu.IsAddHaarMeasure] [Nonempty ι] {lam Lam : ℝ} {r : NNReal} (hr : 1 < r) (hrn : ↑(Fintype.card ι) < 2 * ↑r) :
∃ (D : ℝ), 0 < D ∧ ∀ {Omega : TopologicalSpace.Opens (EuclideanSpace ℝ ι)} {a : EuclideanSpace ℝ ι → Matrix ι ι ℝ} {f : EuclideanSpace ℝ ι → ℝ} {u : ↥(W1p mu Omega 2)} {x₀ : EuclideanSpace ℝ ι} {R : ℝ}, UniformlyEllipticOn (↑Omega) a lam Lam → MeasureTheory.AEStronglyMeasurable a (mu.restrict ↑Omega) → MeasureTheory.MemLp f (↑r) (mu.restrict (Metric.ball x₀ R)) → (∀ (v : ↥(W1p0 mu Omega 2)), energyFormH1 a 0 0 u ↑v = ∫ (x : EuclideanSpace ℝ ι) in ↑Omega, f x * ↑↑(W1p.value ↑v) x ∂mu) → 0 < R → Metric.ball x₀ R ⊆ ↑Omega → ∀ᵐ (x : EuclideanSpace ℝ ι) ∂mu.restrict (Metric.ball x₀ (R / 2)), |↑↑(W1p.value u) x| ≤ D * (R ^ (-↑(Fintype.card ι) / 2) * √(∫ (x : EuclideanSpace ℝ ι) in Metric.ball x₀ R, ↑↑(W1p.value u) x ^ 2 ∂mu) + R ^ (2 - ↑(Fintype.card ι) / ↑r) * (MeasureTheory.eLpNorm f (↑r) (mu.restrict (Metric.ball x₀ R))).toReal / lam)

Local boundedness of weak solutions with a right-hand side. Let n ≥ 1 be the dimension and let r > 1 with r > n/2. There is D > 0, depending only on λ, Λ, n, r and the normalization of mu, such that the following holds. Let a be measurable and uniformly elliptic on Ω with constants λ, Λ, and let u ∈ H¹(Ω) be a weak solution of -∂ᵢ(aⁱʲ ∂ⱼu) = f with f ∈ Lʳ(B(x₀, R)). If B(x₀, R) ⊆ Ω, then almost everywhere on B(x₀, R/2)

|u| ≤ D (R^(-n/2) ‖u‖_{L²(B(x₀, R))} + R^(2 - n/r) ‖f‖_{Lʳ(B(x₀, R))} / λ).

For f = 0 this is TauCeti.PDE.exists_ae_abs_value_le_mul_rpow_mul_sqrt_setIntegral.

Oscillation decay #

theorem TauCeti.PDE.exists_essSup_sub_essInf_le_mul_rpow_mul_add_of_memLp {ι : Type u_1} [Fintype ι] [DecidableEq ι] {mu : MeasureTheory.Measure (EuclideanSpace ℝ ι)} [mu.IsAddHaarMeasure] [Nonempty ι] {lam Lam : ℝ} {r : NNReal} (hr : 1 < r) (hrn : ↑(Fintype.card ι) < 2 * ↑r) :
∃ (α : ℝ) (C : ℝ) (K : ℝ), 0 < α ∧ α ≤ 1 ∧ 0 < C ∧ 0 < K ∧ ∀ {Omega : TopologicalSpace.Opens (EuclideanSpace ℝ ι)} {a : EuclideanSpace ℝ ι → Matrix ι ι ℝ} {f : EuclideanSpace ℝ ι → ℝ} {u : ↥(W1p mu Omega 2)} {x₀ : EuclideanSpace ℝ ι} {ρ s M : ℝ}, UniformlyEllipticOn (↑Omega) a lam Lam → MeasureTheory.AEStronglyMeasurable a (mu.restrict ↑Omega) → MeasureTheory.MemLp f (↑r) (mu.restrict (Metric.ball x₀ ρ)) → (∀ (v : ↥(W1p0 mu Omega 2)), energyFormH1 a 0 0 u ↑v = ∫ (x : EuclideanSpace ℝ ι) in ↑Omega, f x * ↑↑(W1p.value ↑v) x ∂mu) → 0 < s → s ≤ ρ → Metric.ball x₀ ρ ⊆ ↑Omega → (∀ᵐ (x : EuclideanSpace ℝ ι) ∂mu.restrict (Metric.ball x₀ ρ), |↑↑(W1p.value u) x| ≤ M) → essSup (↑↑(W1p.value u)) (mu.restrict (Metric.ball x₀ s)) - essInf (↑↑(W1p.value u)) (mu.restrict (Metric.ball x₀ s)) ≤ C * (s / ρ) ^ α * (essSup (↑↑(W1p.value u)) (mu.restrict (Metric.ball x₀ ρ)) - essInf (↑↑(W1p.value u)) (mu.restrict (Metric.ball x₀ ρ))) + K * ρ ^ (2 - ↑(Fintype.card ι) / ↑r) * (MeasureTheory.eLpNorm f (↑r) (mu.restrict (Metric.ball x₀ ρ))).toReal / lam

One step of oscillation decay with a right-hand side. Let n ≥ 1 be the dimension and let r > 1 with r > n/2. There are α ∈ (0, 1] and C, K > 0, depending only on λ, Λ, n, r and the normalization of mu, such that the following holds. Let a be measurable and uniformly elliptic on Ω with constants λ, Λ, and let u ∈ H¹(Ω) be a weak solution of -∂ᵢ(aⁱʲ ∂ⱼu) = f with f ∈ Lʳ(B(x₀, ρ)). If B(x₀, ρ) ⊆ Ω, u is essentially bounded on B(x₀, ρ) and 0 < s ≤ ρ, then the essential oscillations osc_s and osc_ρ of u on B(x₀, s) and B(x₀, ρ) satisfy

osc_s ≤ C (s/ρ)^α osc_ρ + K ρ^(2 - n/r) ‖f‖_{Lʳ(B(x₀, ρ))} / λ.

For f = 0 this is the oscillation decay TauCeti.PDE.exists_ae_value_mem_Icc_add_mul_rpow_mul_sub of the homogeneous equation.

theorem TauCeti.PDE.exists_ae_value_mem_Icc_add_mul_rpow_mul_add_of_memLp {ι : Type u_1} [Fintype ι] [DecidableEq ι] {mu : MeasureTheory.Measure (EuclideanSpace ℝ ι)} [mu.IsAddHaarMeasure] [Nonempty ι] {lam Lam : ℝ} {r : NNReal} (hr : 1 < r) (hrn : ↑(Fintype.card ι) < 2 * ↑r) :
∃ (α : ℝ) (C : ℝ), 0 < α ∧ α ≤ 1 ∧ 0 < C ∧ ∀ {Omega : TopologicalSpace.Opens (EuclideanSpace ℝ ι)} {a : EuclideanSpace ℝ ι → Matrix ι ι ℝ} {f : EuclideanSpace ℝ ι → ℝ} {u : ↥(W1p mu Omega 2)} {x₀ : EuclideanSpace ℝ ι} {R s : ℝ}, UniformlyEllipticOn (↑Omega) a lam Lam → MeasureTheory.AEStronglyMeasurable a (mu.restrict ↑Omega) → MeasureTheory.MemLp f (↑r) (mu.restrict (Metric.ball x₀ R)) → (∀ (v : ↥(W1p0 mu Omega 2)), energyFormH1 a 0 0 u ↑v = ∫ (x : EuclideanSpace ℝ ι) in ↑Omega, f x * ↑↑(W1p.value ↑v) x ∂mu) → 0 < s → s ≤ R / 2 → Metric.ball x₀ R ⊆ ↑Omega → ∃ (m' : ℝ), ∀ᵐ (x : EuclideanSpace ℝ ι) ∂mu.restrict (Metric.ball x₀ s), ↑↑(W1p.value u) x ∈ Set.Icc m' (m' + C * (s / R) ^ α * (R ^ (-↑(Fintype.card ι) / 2) * √(∫ (x : EuclideanSpace ℝ ι) in Metric.ball x₀ R, ↑↑(W1p.value u) x ^ 2 ∂mu) + R ^ (2 - ↑(Fintype.card ι) / ↑r) * (MeasureTheory.eLpNorm f (↑r) (mu.restrict (Metric.ball x₀ R))).toReal / lam))

De Giorgi's interior oscillation estimate with a right-hand side. Let n ≥ 1 be the dimension and let r > 1 with r > n/2. There are α ∈ (0, 1] and C > 0, depending only on λ, Λ, n, r and the normalization of mu, such that the following holds. Let a be measurable and uniformly elliptic on Ω with constants λ, Λ, and let u ∈ H¹(Ω) be a weak solution of -∂ᵢ(aⁱʲ ∂ⱼu) = f with f ∈ Lʳ(B(x₀, R)). If B(x₀, R) ⊆ Ω and 0 < s ≤ R/2, then almost everywhere on B(x₀, s) the function u takes values in an interval of length

C (s / R)^α (R^(-n/2) ‖u‖_{L²(B(x₀, R))} + R^(2 - n/r) ‖f‖_{Lʳ(B(x₀, R))} / λ).

For f = 0 this is the form of TauCeti.PDE.exists_ae_value_mem_Icc_add_mul_rpow_mul_rpow_mul_sqrt_setIntegral.

Hölder continuity #

theorem TauCeti.PDE.exists_holderOnWith_preciseRepresentative_of_memLp {ι : Type u_1} [Fintype ι] [DecidableEq ι] {mu : MeasureTheory.Measure (EuclideanSpace ℝ ι)} [mu.IsAddHaarMeasure] [Nonempty ι] {lam Lam : ℝ} {r : NNReal} (hr : 1 < r) (hrn : ↑(Fintype.card ι) < 2 * ↑r) :
∃ (α : NNReal) (C : ℝ), 0 < α ∧ α ≤ 1 ∧ 0 < C ∧ ∀ {Omega : TopologicalSpace.Opens (EuclideanSpace ℝ ι)} {a : EuclideanSpace ℝ ι → Matrix ι ι ℝ} {f : EuclideanSpace ℝ ι → ℝ} {u : ↥(W1p mu Omega 2)} {K : Set (EuclideanSpace ℝ ι)} {R : ℝ}, UniformlyEllipticOn (↑Omega) a lam Lam → MeasureTheory.AEStronglyMeasurable a (mu.restrict ↑Omega) → MeasureTheory.MemLp f (↑r) (mu.restrict ↑Omega) → (∀ (v : ↥(W1p0 mu Omega 2)), energyFormH1 a 0 0 u ↑v = ∫ (x : EuclideanSpace ℝ ι) in ↑Omega, f x * ↑↑(W1p.value ↑v) x ∂mu) → 0 < R → Metric.cthickening R K ⊆ ↑Omega → HolderOnWith (C * R ^ (-↑α) * (R ^ (-↑(Fintype.card ι) / 2) * √(∫ (y : EuclideanSpace ℝ ι) in ↑Omega, ↑↑(W1p.value u) y ^ 2 ∂mu) + R ^ (2 - ↑(Fintype.card ι) / ↑r) * (MeasureTheory.eLpNorm f (↑r) (mu.restrict ↑Omega)).toReal / lam)).toNNReal α (MeasureTheory.preciseRepresentative mu ↑↑(W1p.value u)) K

De Giorgi's theorem with a right-hand side. Let n ≥ 1 be the dimension and let r > 1 with r > n/2. There are α ∈ (0, 1] and C > 0, depending only on λ, Λ, n, r and the normalization of the additive Haar measure mu, such that the following holds. Let a be measurable and uniformly elliptic on Ω with constants λ, Λ, let f ∈ Lʳ(Ω), and let u ∈ H¹(Ω) be a weak solution of -∂ᵢ(aⁱʲ ∂ⱼu) = f, that is a(u, v) = ∫_Ω f v for every v ∈ H¹₀(Ω). If the closed R-thickening of a set K lies in Ω, then the precise representative of u is Hölder continuous on K with exponent α and constant

C R^(-α) (R^(-n/2) ‖u‖_{L²(Ω)} + R^(2 - n/r) ‖f‖_{Lʳ(Ω)} / λ).

The precise representative agrees with u almost everywhere on Ω (TauCeti.W1p.ae_eq_preciseRepresentative). No regularity of the coefficients beyond measurability is assumed.

Continuity for a locally integrable right-hand side #

theorem TauCeti.PDE.continuousOn_preciseRepresentative_of_forall_memLp {ι : Type u_1} [Fintype ι] [DecidableEq ι] {mu : MeasureTheory.Measure (EuclideanSpace ℝ ι)} [mu.IsAddHaarMeasure] [Nonempty ι] {lam Lam : ℝ} {r : NNReal} (hr : 2 ≤ r) (hrn : ↑(Fintype.card ι) < 2 * ↑r) {Omega : TopologicalSpace.Opens (EuclideanSpace ℝ ι)} {a : EuclideanSpace ℝ ι → Matrix ι ι ℝ} {f : EuclideanSpace ℝ ι → ℝ} {u : ↥(W1p mu Omega 2)} (h : UniformlyEllipticOn (↑Omega) a lam Lam) (ha : MeasureTheory.AEStronglyMeasurable a (mu.restrict ↑Omega)) (hf : ∀ (K : Set (EuclideanSpace ℝ ι)), IsCompact K → K ⊆ ↑Omega → MeasureTheory.MemLp f (↑r) (mu.restrict K)) (hu : ∀ (φ : TestFunction Omega ℝ ⊤), energyFormH1 a 0 0 u ((W1p.ofTestFunctionₗ mu Omega 2) φ) = ∫ (x : EuclideanSpace ℝ ι) in ↑Omega, f x * φ x ∂mu) :

Continuity of weak solutions with a locally integrable right-hand side. Let n ≥ 1 be the dimension and let r ≥ 2 with r > n/2. Let a be measurable and uniformly elliptic on Ω, and let u ∈ H¹(Ω) solve -∂ᵢ(aⁱʲ ∂ⱼu) = f against test functions, a(u, φ) = ∫_Ω f φ, where f lies in Lʳ on every compact subset of Ω. Then the precise representative of u, which agrees with u almost everywhere on Ω (TauCeti.W1p.ae_eq_preciseRepresentative), is continuous on Ω.

No integrability of f on all of Ω is assumed. On each ball B whose closure lies in Ω, the equation holds against all of H¹₀(B) (TauCeti.PDE.forall_energyFormH1_restrictL_eq_setIntegral, which uses f ∈ L²(B)), and De Giorgi's theorem with a right-hand side, TauCeti.PDE.exists_holderOnWith_preciseRepresentative_of_memLp, makes the restriction of u Hölder continuous on the concentric ball of half the radius.

theorem TauCeti.PDE.exists_continuousOn_ae_eq_of_forall_integral_inner_gradient_eq {E : Type u_2} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] [MeasurableSpace E] [BorelSpace E] [Nontrivial E] {r : NNReal} (hr : 2 ≤ r) (hrn : ↑(Module.finrank ℝ E) < 2 * ↑r) {Omega : TopologicalSpace.Opens E} {f : E → ℝ} {u : ↥(W1p MeasureTheory.volume Omega 2)} (hf : ∀ (K : Set E), IsCompact K → K ⊆ ↑Omega → MeasureTheory.MemLp f (↑r) (MeasureTheory.volume.restrict K)) (hu : ∀ (φ : TestFunction Omega ℝ ⊤), ∫ (x : E) in ↑Omega, inner ℝ (↑↑(W1p.gradient u) x) (gradient (⇑φ) x) = ∫ (x : E) in ↑Omega, f x * φ x) :
∃ (g : E → ℝ), ContinuousOn g ↑Omega ∧ ↑↑(W1p.value u) =ᵐ[MeasureTheory.volume.restrict ↑Omega] g

Continuity of weak solutions of Poisson's equation. Let E be a real inner product space of finite dimension n ≥ 1 with its volume, and let r ≥ 2 with r > n/2. If u ∈ H¹(Ω) solves -Δu = f weakly, ∫_Ω ⟪∇u, ∇φ⟫ = ∫_Ω f φ for every test function φ on Ω, where f lies in Lʳ on every compact subset of Ω, then u agrees almost everywhere on Ω with a function that is continuous on Ω.

This is the case a = 1 of TauCeti.PDE.continuousOn_preciseRepresentative_of_forall_memLp, transported to E through the coordinates of an orthonormal basis.