Documentation

TauCeti.Analysis.ODE.EnergyEstimate.Sobolev

Estimates for the trajectory operator on Sobolev paths #

The trajectory operator

D_A u = u' + A u : W^{1,2}(ℝ, E) → L²(ℝ, E)

(TauCeti.WeakDerivStep.trajectoryOperator, for an L∞ class of A) is controlled on Sobolev paths by any estimate it satisfies on compactly supported C¹ paths. An estimate

a ∫ ‖u'‖² + b ∫ ‖u‖² ≤ c ∫ ‖u' + A u‖² + d ∫_{[-R, R]} ‖u‖²

on such paths extends to W^{1,2}(ℝ, E), with the last term replaced by the squared norm of the restriction to a slightly larger interval, because those paths are dense (TauCeti.WeakDerivStep.dense_setOf_exists_contDiff_hasCompactSupport) and both sides are continuous. Restriction from W^{1,2}(ℝ, E) to L²(-R, R) is compact (Rellich), so an estimate with a, b > 0 is an estimate with a compact error term, the input of Peetre's lemma. This is how the Fredholm property of D_A is proved in TauCeti.Analysis.ODE.EnergyEstimate.Lyapunov.

When A is a C¹ family of symmetric operators with a spectral gap and a small derivative on the whole line, rather than only near its limits, the coercive estimate for test paths TauCeti.integral_norm_deriv_sq_add_mul_integral_norm_sq_le has no error term, so D_A is injective. This applies in particular to a constant invertible symmetric coefficient.

Main declarations #

References #

Restriction of a path with a smooth compactly supported representative #

Extending a test-path estimate to Sobolev paths. If every compactly supported C¹ path φ satisfies a ∫ ‖φ'‖² + b ∫ ‖φ‖² ≤ c ∫ ‖φ' + A φ‖² + d ∫_{[-R, R]} ‖φ‖² with d ≥ 0, then a ‖u'‖² + b ‖u‖² ≤ c ‖D_A u‖² + d ‖u‖²_{L²(-R', R')} for every u ∈ W^{1,2}(ℝ, E) and every R' > R. The coefficient of the operator is any L∞ class AL of A.

The coercive trajectory estimate on Sobolev paths. If a C¹ family of symmetric operators satisfies α ‖v‖ ≤ ‖A s v‖ and ⟪v, A' s v⟫ ≤ β ‖v‖² everywhere, then ‖u'‖² + (α² - β) ‖u‖² ≤ ‖D_A u‖² for every u ∈ W^{1,2}(ℝ, E), with no compact error term. The coefficient of the operator is any L∞ class AL of A.

theorem TauCeti.WeakDerivStep.min_one_mul_norm_sq_le_norm_trajectoryOperator_sq {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] {A : ℝ → E →L[ℝ] E} (hA : ContDiff ℝ 1 A) (hsym : ∀ (s : ℝ), (↑(A s)).IsSymmetric) {α β : ℝ} (hα : 0 ≤ α) (hgap : ∀ (s : ℝ) (v : E), α * ‖v‖ ≤ ‖(A s) v‖) (hsmall : ∀ (s : ℝ) (v : E), inner ℝ v ((deriv A s) v) ≤ β * ‖v‖ ^ 2) (AL : ↥(MeasureTheory.Lp (E →L[ℝ] E) ⊤ (MeasureTheory.volume.restrict ↑⊤))) (hAL : ↑↑AL =ᵐ[MeasureTheory.volume.restrict ↑⊤] A) (u : ↥(WeakDerivStep MeasureTheory.volume ⊤ 2 (ContinuousLinearMap.id ℝ ↥(MeasureTheory.Lp E 2 (MeasureTheory.volume.restrict ↑⊤))))) :
min 1 (α ^ 2 - β) * ‖u‖ ^ 2 ≤ ‖(trajectoryOperator AL) u‖ ^ 2

The coercive trajectory estimate in the full Sobolev norm. If a C¹ family of symmetric operators satisfies α ‖v‖ ≤ ‖A s v‖ and ⟪v, A' s v⟫ ≤ β ‖v‖² everywhere, then min 1 (α² - β) ‖u‖² ≤ ‖D_A u‖² for every u ∈ W^{1,2}(ℝ, E), where ‖u‖² = ‖u‖²_{L²} + ‖u'‖²_{L²}. The coefficient of the operator is any L∞ class AL of A.

theorem TauCeti.WeakDerivStep.min_one_mul_norm_le_norm_trajectoryOperator {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] {A : ℝ → E →L[ℝ] E} (hA : ContDiff ℝ 1 A) (hsym : ∀ (s : ℝ), (↑(A s)).IsSymmetric) {α β : ℝ} (hα : 0 ≤ α) (hgap : ∀ (s : ℝ) (v : E), α * ‖v‖ ≤ ‖(A s) v‖) (hsmall : ∀ (s : ℝ) (v : E), inner ℝ v ((deriv A s) v) ≤ β * ‖v‖ ^ 2) (AL : ↥(MeasureTheory.Lp (E →L[ℝ] E) ⊤ (MeasureTheory.volume.restrict ↑⊤))) (hAL : ↑↑AL =ᵐ[MeasureTheory.volume.restrict ↑⊤] A) (u : ↥(WeakDerivStep MeasureTheory.volume ⊤ 2 (ContinuousLinearMap.id ℝ ↥(MeasureTheory.Lp E 2 (MeasureTheory.volume.restrict ↑⊤))))) :
min 1 (α ^ 2 - β) * ‖u‖ ≤ ‖(trajectoryOperator AL) u‖

The coercive trajectory estimate as a linear bound. If a C¹ family of symmetric operators satisfies α ‖v‖ ≤ ‖A s v‖ and ⟪v, A' s v⟫ ≤ β ‖v‖² everywhere, then min 1 (α² - β) ‖u‖ ≤ ‖D_A u‖ for every u ∈ W^{1,2}(ℝ, E). For β < α² this bounds the inverse of D_A by (min 1 (α² - β))⁻¹. The coefficient of the operator is any L∞ class AL of A.

theorem TauCeti.WeakDerivStep.ker_trajectoryOperator_eq_bot {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] {A : ℝ → E →L[ℝ] E} (hA : ContDiff ℝ 1 A) (hsym : ∀ (s : ℝ), (↑(A s)).IsSymmetric) {α β : ℝ} (hα : 0 ≤ α) (hgap : ∀ (s : ℝ) (v : E), α * ‖v‖ ≤ ‖(A s) v‖) (hsmall : ∀ (s : ℝ) (v : E), inner ℝ v ((deriv A s) v) ≤ β * ‖v‖ ^ 2) (hβ : β < α ^ 2) (AL : ↥(MeasureTheory.Lp (E →L[ℝ] E) ⊤ (MeasureTheory.volume.restrict ↑⊤))) (hAL : ↑↑AL =ᵐ[MeasureTheory.volume.restrict ↑⊤] A) :

The trajectory operator of a coercive family is injective. If a C¹ family of symmetric operators satisfies α ‖v‖ ≤ ‖A s v‖ and ⟪v, A' s v⟫ ≤ β ‖v‖² everywhere with β < α², then u' + A u = 0 has no nonzero solution in W^{1,2}(ℝ, E).