Theorem 4.27 — interior H² regularity of weak solutions of −∑∂ᵢ(aᵢⱼ∂ⱼu) = f
ProvedHunterPDE.Regularity.interior_H2_regularityLet be open. Let the coefficients be symmetric, , and uniformly elliptic on with constant , and let . If is a weak solution of in , then for every , and
where the constant depends only on , , and the — not on or .
Weak solutions are a priori only in ; this theorem shows they have square-integrable second weak derivatives away from the boundary, so that the equation holds pointwise almost everywhere (a strong solution).
Formalization Note. The hypotheses , and uniform ellipticity are the standing assumptions (4.17)–(4.18) of §4.6 on the operator. is ContDiffOn ℝ 1 on , as printed (no regularity up to the boundary). The constant is ∃ C : ℝ, quantified after , the coefficients, and and before and ; the norms are ℝ≥0∞-valued and enters as ENNReal.ofReal C. is the norm of Definition 3.23 with , and weak derivatives on .
import Mathlib import Definitions.Def_HunterPDE_Regularity_WeakSolution import Definitions.Def_HunterPDE_Regularity_DiffQuotient open MeasureTheory open scoped ENNReal
namespace HunterPDE.Regularity
/-- Theorem 4.27 of Hunter, *Notes on PDEs* (revised 6/18/2014), p. 112 (interior `H²`
regularity): let `Ω` be an open set in `ℝⁿ`, `aᵢⱼ ∈ C¹(Ω)` and `f ∈ L²(Ω)`. If `u ∈ H¹(Ω)` is a
weak solution of `L u = −∑ᵢⱼ ∂ᵢ(aᵢⱼ ∂ⱼu) = f` ((4.34)–(4.37)), then `u ∈ H²(Ω′)` for every
`Ω′ ⋐ Ω`, and (4.39) `‖u‖_{H²(Ω′)} ≤ C (‖f‖_{L²(Ω)} + ‖u‖_{L²(Ω)})` where `C` depends only on
`n`, `Ω′`, `Ω` and `aᵢⱼ`.
The standing assumptions of §4.6 on `L` are hypotheses: `aᵢⱼ ∈ L^∞(Ω)`, `aᵢⱼ = aⱼᵢ` ((4.17)) and
uniform ellipticity with constant `θ` ((4.18)). `C¹(Ω)` is `ContDiffOn ℝ 1` on `Ω` (no
regularity up to the boundary, as printed). The constant `C` is chosen after `Ω`, the
coefficients (and their ellipticity constant) and `Ω′`, and before `f` and `u`. -/
theorem interior_H2_regularity {n : ℕ} (Ω : Set (EuclideanSpace ℝ (Fin n))) (hΩ : IsOpen Ω)
(a : Fin n → Fin n → EuclideanSpace ℝ (Fin n) → ℝ)
(ha_C1 : ∀ i j, ContDiffOn ℝ 1 (a i j) Ω)
(ha_bdd : ∀ i j, MemLp (a i j) ∞ (volume.restrict Ω))
(ha_symm : ∀ i j, a i j = a j i) (θ : ℝ) (hθ : UniformlyElliptic Ω a θ)
(Ω' : Set (EuclideanSpace ℝ (Fin n))) (hΩ' : CompactlyContained Ω' Ω) :
∃ C : ℝ, ∀ f u : EuclideanSpace ℝ (Fin n) → ℝ,
MemLp f 2 (volume.restrict Ω) → IsWeakSolution Ω a f u →
MemW 2 2 Ω' u ∧
sobolevNorm 2 2 Ω' u ≤ ENNReal.ofReal C *
(eLpNorm f 2 (volume.restrict Ω) + eLpNorm u 2 (volume.restrict Ω)) := by sorry
end HunterPDE.Regularity
Read-back
What the Lean code literally says, in plain math · claude-opus-5-5
Let (implicit), and write for -dimensional Euclidean space with Lebesgue measure . For a set and , write for the seminorm of with respect to restricted to . This is Lean's eLpNorm, which is defined for every function and takes the value when the integral diverges.
Data and hypotheses. The statement fixes:
- a set that is open;
- functions for , defined on all of , such that for every :
- is continuously differentiable on , in the sense of
ContDiffOnof order . It is differentiable within at each point of with a derivative continuous on . Nothing is required outside or up to . - is essentially bounded on : it is a.e.-strongly measurable for and .
- as functions on the whole of , not only on .
- is continuously differentiable on , in the sense of
- a real number , with no sign condition stated in this code, and the hypothesis .
- a set with .
, , , and are definitions from the imported bundle files Def_HunterPDE_Regularity_WeakSolution and Def_HunterPDE_Regularity_DiffQuotient. Their bodies are not part of the code given here, so their content is not expanded in this read-back. Every property below that rests on them holds exactly as those definitions say, and no more.
In particular, nothing in the statement itself requires any of the following. They hold only if one of the bundle definitions imposes them:
- is square-integrable or has weak derivatives;
- is open, nonempty or bounded;
- .
Conclusion. There exists a real number such that for all functions the following holds. If
- (a.e.-strongly measurable for with ), and
- holds,
then both of the following are true:
- satisfies the bundle predicate with parameters on . The name suggests membership in the Sobolev space , but its content is whatever the bundle definition says.
- The following inequality holds in the extended non-negative reals :
The factor is ENNReal.ofReal C: a negative is replaced by .
Order of quantifiers. is chosen after , , the , and (and the hypotheses about them), and before and . So one constant works for all and , but it may depend on everything else, including . No explicit form or dependence of is asserted. The imported difference-quotient file contributes no symbol that appears in the statement.
Degenerate cases.
- Infinite norm of . If , which the statement itself does not exclude, the right-hand side is when , and the inequality is then trivially true. When it is , which is under the convention of . In the same way, makes the right-hand side for every , forcing .
- . is a single point carrying Lebesgue measure , and there are no indices . So the coefficient hypotheses are empty, and is either empty or that point.
- . The restricted measure is zero. Then every condition holds trivially, the right-hand side equals , and the conclusion demands together with . This case is live only if can hold, which depends on that definition, for instance whether it permits .
- Unsatisfiable hypotheses. If or cannot be satisfied, for example for a nonpositive if the definition requires positivity, the theorem is vacuous for those data. If is never satisfiable, any witnesses the statement vacuously. Which of these occur cannot be determined from the code shown.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.