Documentation

LeanPool.CIVAxisymmetric.Comparison.Mollifier

The spatial mollifier of lem:aniso:comparison #

The comparison argument mollifies in space with the standard family η_ε = ε^{-m} η(·/ε), where η is a fixed smooth nonnegative kernel of unit mass supported in the unit ball. The profile η is the normalized bump kernel of the ambient library, and the family is obtained from it by the explicit rescaling, so that the scaling identities below are available.

The bounds recorded here are the two that the commutator estimate of DiPerna and Lions needs: the kernel has unit mass and is supported in the ball of radius ε, and the integral of the length of its gradient is C_η/ε with C_η the corresponding integral for the profile. The ball is the one of the sup norm of Vec m, which is the norm carried by that type.

The fixed profile #

noncomputable def CIV.mollifierProfile (m : ℕ) :
CKN.Vec m → ℝ

The fixed mollifier profile η: the normalized smooth bump of radius one on Vec m.

Equations
Instances For

    The profile is supported in the unit ball.

    The rescaled family #

    noncomputable def CIV.mollifierKernel (m : ℕ) (ε : ℝ) :
    CKN.Vec m → ℝ

    The rescaled kernel η_ε(x) = ε^{-m} η(x/ε).

    Equations
    Instances For
      theorem CIV.mollifierKernel_nonneg {m : ℕ} {ε : ℝ} (hε : 0 < ε) (x : CKN.Vec m) :
      theorem CIV.mollifierKernel_eq_zero {m : ℕ} {ε : ℝ} (hε : 0 < ε) {x : CKN.Vec m} (hx : ε ≤ ‖x‖) :

      The kernel η_ε vanishes outside the ball of radius ε.

      theorem CIV.integral_mollifierKernel {m : ℕ} {ε : ℝ} (hε : 0 < ε) :
      ∫ (x : CKN.Vec m), mollifierKernel m ε x = 1

      The kernel η_ε has unit mass.

      The gradient of the kernel #

      noncomputable def CIV.mollifierGradConst (m : ℕ) :

      The gradient constant C_η = ∫ ∑_i |∂_iη| of the fixed profile.

      Equations
      Instances For
        theorem CIV.fderiv_mollifierKernel_apply {m : ℕ} (ε : ℝ) (x v : CKN.Vec m) :
        (fderiv ℝ (mollifierKernel m ε) x) v = (ε ^ m)⁻¹ * (ε⁻¹ * (fderiv ℝ (mollifierProfile m) (ε⁻¹ • x)) v)
        theorem CIV.integral_sum_abs_fderiv_mollifierKernel {m : ℕ} {ε : ℝ} (hε : 0 < ε) :
        ∫ (x : CKN.Vec m), ∑ i : Fin m, |(fderiv ℝ (mollifierKernel m ε) x) (CKN.basisVec i)| = mollifierGradConst m / ε

        The scaling of the gradient integral: ∫ |∇η_ε| = C_η/ε.

        theorem CIV.fderiv_mollifierKernel_eq_zero {m : ℕ} {ε : ℝ} (hε : 0 < ε) {w : CKN.Vec m} (hw : ε < ‖w‖) (v : CKN.Vec m) :
        (fderiv ℝ (mollifierKernel m ε) w) v = 0

        The gradient of η_ε vanishes outside the ball of radius ε.

        theorem CIV.continuous_fderiv_mollifierKernel {m : ℕ} (ε : ℝ) (v : CKN.Vec m) :
        Continuous fun (w : CKN.Vec m) => (fderiv ℝ (mollifierKernel m ε) w) v
        theorem CIV.hasCompactSupport_fderiv_mollifierKernel {m : ℕ} {ε : ℝ} (hε : 0 < ε) (v : CKN.Vec m) :
        HasCompactSupport fun (w : CKN.Vec m) => (fderiv ℝ (mollifierKernel m ε) w) v