Hille-Yosida generation from the resolvent estimates
ProvedEthierKurtz.oblique_hille_yosida_generationmarkov-processesprobability
Hille--Yosida generation from the resolvent estimates.
Assume the dissipative estimate and the dense-range statement for the closed reflected graph of Theorem 1.5. Then there exists a strongly continuous contraction semigroup on that is positive () and conservative (), whose generator is exactly :
This is the generation step itself: the two resolvent estimates produce the semigroup by the Hille--Yosida theorem, while positivity and conservativeness follow from the maximum principle for the interior operator together with the oblique boundary condition.
Formalization Note Lean takes the two estimates as explicit hypotheses and quantifies over , with convergence in the filter.
Preamble
import Definitions.Def_EthierKurtz_BoundaryCTwiceHolder import Definitions.Def_EthierKurtz_BoundaryCOnceHolder import Definitions.Def_EthierKurtz_IsOutwardUnitNormal import Definitions.Def_EthierKurtz_obliqueDiffusionGraph import Definitions.Def_EthierKurtz_IsStronglyContinuousContractionSemigroup open Filter open scoped Topology BoundedContinuousFunction namespace EthierKurtz
Formal statement
/-- Hille-Yosida generation from the resolvent estimates: dissipativity plus
dense range for the closed reflected graph yields the conservative positive
strongly continuous contraction semigroup whose generator is exactly the
graph. Positivity and conservativeness come from the maximum principle for
the interior operator with the oblique boundary condition. -/
theorem oblique_hille_yosida_generation (n : ℕ) (hd : 2 ≤ n + 1)
(Ω : Set (EuclideanSpace ℝ (Fin (n + 1))))
(hbounded : Bornology.IsBounded Ω) (hconnected : IsConnected Ω)
(hopen : IsOpen Ω) (μ : ℝ) (hμ : 0 < μ ∧ μ ≤ 1)
(hboundary : BoundaryCTwiceHolder Ω μ)
(a : EuclideanSpace ℝ (Fin (n + 1)) → Matrix (Fin (n + 1)) (Fin (n + 1)) ℝ)
(b c normal : EuclideanSpace ℝ (Fin (n + 1)) → EuclideanSpace ℝ (Fin (n + 1)))
(ha : ∀ x ∈ Ω, (a x).PosSemidef)
(haHolder : ∀ i j, ComponentHolder Ω μ (fun x => a x i j))
(hbHolder : ∀ i, ComponentHolder Ω μ (fun x => b x i))
(helliptic : ∃ ε : ℝ, 0 < ε ∧ ∀ x ∈ Ω,
∀ θ : EuclideanSpace ℝ (Fin (n + 1)), ‖θ‖ = 1 →
ε ≤ ∑ i, ∑ j, θ i * a x i j * θ j)
(hc : ∀ i, BoundaryCOnceHolder Ω μ (fun x => c x i))
(hnormal : ∀ x ∈ frontier Ω, IsOutwardUnitNormal Ω x (normal x))
(hoblique : ∃ ε : ℝ, 0 < ε ∧ ∀ x ∈ frontier Ω,
ε ≤ ∑ i, c x i * normal x i)
(hest : ∀ fg ∈ closure (obliqueDiffusionGraph Ω μ a b c),
∀ lam : ℝ, 0 < lam → lam * ‖fg.1‖ ≤ ‖lam • fg.1 - fg.2‖)
(hrange : ∃ lam₀ : ℝ, 0 < lam₀ ∧ ∀ h : (closure Ω) →ᵇ ℝ, ∀ delta : ℝ,
0 < delta → ∃ fg ∈ closure (obliqueDiffusionGraph Ω μ a b c),
‖(lam₀ • fg.1 - fg.2) - h‖ < delta) :
let A := closure (obliqueDiffusionGraph Ω μ a b c)
∃ T : ℝ → ((closure Ω) →ᵇ ℝ) →L[ℝ] ((closure Ω) →ᵇ ℝ),
IsStronglyContinuousContractionSemigroup T ∧
(∀ t : ℝ, 0 ≤ t → ∀ f, (∀ x, 0 ≤ f x) → ∀ x, 0 ≤ T t f x) ∧
(∀ t : ℝ, 0 ≤ t → T t 1 = 1) ∧
(∀ f g, Tendsto (fun t : ℝ => t⁻¹ • (T t f - f))
(𝓝[>] (0 : ℝ)) (𝓝 g) ↔ (f, g) ∈ A) := by sorry
Source
Stewart N. Ethier and Thomas G. Kurtz, Markov Processes: Characterization and Convergence, Wiley, 1986. Chapter 8, Section 1, Theorem 1.5, printed p. 369 (PDF p. 378); Hille-Yosida generation with (1.15) and (1.19)--(1.20).