Documentation

LeanPool.EllipticPDE.Regularity.LowerOrderWkInfty

W^{k,∞} regularity for the transport and zeroth-order coefficients #

EllipticPdes.Sobolev.FullEllipticOp gives b and c sup bounds and measurability and no derivatives at all, which is everything the existence theory and the interior H² estimate need. Guo, Partial Differential Equations (Course Lecture Notes), Theorem VIII.3.2 (p. 65) asks for a_{ij} ∈ W^{k+2,∞} and b_i, c ∈ W^{k+1,∞}, one order less on the lower-order coefficients than on the principal part, because the lower-order terms are differentiated once less often on the way to the same conclusion. This file supplies the missing hypothesis.

Scalar predicate and bundle #

IsWkInfty f k states the hypothesis for a single scalar function, and IsWkInftyLower bundles it over the d + 1 lower-order coefficients with a constant uniform across them. Splitting it this way lets IsWkInfty.deriv be stated once and used for b, for c, and for anything else the induction differentiates.

Main declarations #

Scalar hypothesis #

structure EllipticPdes.Regularity.IsWkInfty {d : ℕ} (f : EuclideanSpace ℝ (Fin d) → ℝ) (k : ℕ) :

f ∈ W^{k,∞}, given as data: a family of weak derivatives indexed by lists of directions, each measurable and essentially bounded, with no continuity assumed. The order-zero member is f itself, so bound 0 is a sup bound on f.

Instances For
    def EllipticPdes.Regularity.IsWkInfty.mono {d : ℕ} {f : EuclideanSpace ℝ (Fin d) → ℝ} {k l : ℕ} (hf : IsWkInfty f k) (hlk : l ≤ k) :

    An order-k bundle is an order-l bundle for every l ≤ k.

    Equations
    • hf.mono hlk = { D := hf.D, D_nil := ⋯, D_meas := ⋯, D_step := ⋯, bound := hf.bound, bound_nonneg := ⋯, ess_bdd := ⋯ }
    Instances For
      def EllipticPdes.Regularity.IsWkInfty.deriv {d : ℕ} {f : EuclideanSpace ℝ (Fin d) → ℝ} {k : ℕ} (hf : IsWkInfty f (k + 1)) (m : Fin d) :
      IsWkInfty (hf.D [m]) k

      Differentiating the hypothesis. From f ∈ W^{k+1,∞}, the first derivative D [m] is in W^{k,∞}, with family α ↦ D (α ++ [m]) and the bounds shifted by one order. Appending on the right makes the lengths line up, exactly as in HasIteratedWeakDerivOn.deriv; this step lets the induction of Guo's Theorem VIII.3.2 keep its coefficient hypothesis.

      Equations
      • hf.deriv m = { D := fun (α : List (Fin d)) => hf.D (α ++ [m]), D_nil := ⋯, D_meas := ⋯, D_step := ⋯, bound := fun (j : ℕ) => hf.bound (j + 1), bound_nonneg := ⋯, ess_bdd := ⋯ }
      Instances For
        noncomputable def EllipticPdes.Regularity.IsWkInfty.ofContDiff {d : ℕ} {f : EuclideanSpace ℝ (Fin d) → ℝ} {k : ℕ} {B : ℕ → ℝ} (hf : ContDiff ℝ (↑↑k) f) (hB : ∀ (m : ℕ), 0 ≤ B m) (hbd : ∀ m ≤ k, ∀ (x : EuclideanSpace ℝ (Fin d)), ‖iteratedFDeriv ℝ m f x‖ ≤ B m) :

        Cᵏ function with bounded derivatives in W^{k,∞}. The classical iterated partials serve as the family, through hasWeakPartial_partialD, and the pointwise iteratedFDeriv bounds transfer through abs_iterPartial_le. Order zero is included here, unlike in IsCkCoeff, where EllipticCoeff.Λ already has it.

        Equations
        Instances For
          noncomputable def EllipticPdes.Regularity.IsWkInfty.const {d : ℕ} (c : ℝ) (k : ℕ) :
          IsWkInfty (fun (x : EuclideanSpace ℝ (Fin d)) => c) k

          Constants in W^{k,∞} at every order. Its derivatives past the zeroth vanish, so one bound serves every order. The datum of the induction step has a term with no coefficient at all, namely the derivative of the datum itself, and this is what lets it be treated as a weighted term like the rest.

          Equations
          Instances For
            def EllipticPdes.Regularity.IsWkInftyCoeff.entry {d : ℕ} {A : Sobolev.EllipticCoeff d} {k : ℕ} (hA : IsWkInftyCoeff A k) (i j : Fin d) :
            IsWkInfty (fun (x : EuclideanSpace ℝ (Fin d)) => A.a x i j) k

            Single entry of the coefficient matrix as a W^{k,∞} function. The matrix bundle already has a family for each entry, so the scalar bundle is that family read at a fixed pair of indices.

            Equations
            • hA.entry i j = { D := fun (α : List (Fin d)) => hA.D α i j, D_nil := ⋯, D_meas := ⋯, D_step := ⋯, bound := hA.bound, bound_nonneg := ⋯, ess_bdd := ⋯ }
            Instances For

              Bundle over the lower-order coefficients #

              Guo's hypothesis on the lower-order coefficients. Every transport component and the zeroth-order coefficient lie in W^{k,∞}, with one constant serving all of them. The uniform constant is what the estimate of Theorem VIII.3.2 is stated against, so it is recorded here rather than reconstructed as a maximum at the point of use.

              • bReg (i : Fin d) : IsWkInfty (fun (x : EuclideanSpace ℝ (Fin d)) => Op.b x i) k

                Each transport component is in W^{k,∞}.

              • cReg : IsWkInfty Op.c k

                The zeroth-order coefficient is in W^{k,∞}.

              • bound : ℕ → ℝ

                The constant uniform across the lower-order coefficients.

              • bound_nonneg (m : ℕ) : 0 ≤ self.bound m

                Every bound is nonnegative.

              • b_le (i : Fin d) (m : ℕ) : (self.bReg i).bound m ≤ self.bound m

                The transport bounds are dominated by the uniform constant.

              • c_le (m : ℕ) : self.cReg.bound m ≤ self.bound m

                The zeroth-order bound is dominated by the uniform constant.

              Instances For

                An order-k bundle is an order-l bundle for every l ≤ k.

                Equations
                • hOp.mono hlk = { bReg := fun (i : Fin d) => (hOp.bReg i).mono hlk, cReg := hOp.cReg.mono hlk, bound := hOp.bound, bound_nonneg := ⋯, b_le := ⋯, c_le := ⋯ }
                Instances For
                  theorem EllipticPdes.Regularity.IsWkInftyLower.ae_abs_b_le {d : ℕ} {Op : Sobolev.FullEllipticOp d} {k : ℕ} (hOp : IsWkInftyLower Op k) (i : Fin d) (α : List (Fin d)) (hα : α.length ≤ k) :
                  ∀ᵐ (x : EuclideanSpace ℝ (Fin d)), |(hOp.bReg i).D α x| ≤ hOp.bound α.length

                  Every transport component is essentially bounded by the uniform constant at every order up to k.

                  theorem EllipticPdes.Regularity.IsWkInftyLower.ae_abs_c_le {d : ℕ} {Op : Sobolev.FullEllipticOp d} {k : ℕ} (hOp : IsWkInftyLower Op k) (α : List (Fin d)) (hα : α.length ≤ k) :
                  ∀ᵐ (x : EuclideanSpace ℝ (Fin d)), |hOp.cReg.D α x| ≤ hOp.bound α.length

                  The zeroth-order coefficient is essentially bounded by the uniform constant at every order up to k.