Smooth radial kernels of unit integral
ProvedMeasureTheory.exists_smooth_radial_probability_kernelconvolutionharmonic-functionsmean-value-propertyregularity
For every radius and every , there is a smooth compactly supported radial function on , with profile , such that
The kernel can be chosen nonnegative; that extra property is not part of this interface.
Preamble
import Mathlib.Analysis.Calculus.BumpFunction.InnerProduct import Mathlib.MeasureTheory.Measure.Haar.InnerProductSpace import Mathlib.MeasureTheory.Integral.Bochner.ContinuousLinearMap open MeasureTheory Set Metric Function open scoped ContDiff Topology set_option autoImplicit false
Formal statement
theorem MeasureTheory.exists_smooth_radial_probability_kernel {n : ℕ} {R : ℝ} (hR : 0 < R) :
∃ (k : EuclideanSpace ℝ (Fin n) → ℝ) (κ : ℝ → ℝ),
ContDiff ℝ ∞ k ∧ HasCompactSupport k ∧
(∀ y, k y = κ ‖y‖) ∧ (∀ y, R ≤ ‖y‖ → k y = 0) ∧
(∫ y, k y) = 1 := by sorrySource
Hunter, Notes on PDEs (revised 6/18/2014), printed p. 21, proof of Theorem 2.2; https://www.math.ucdavis.edu/~hunter/pdes/ch2.pdf. Kernel construction: Mathlib Analysis/Calculus/BumpFunction/InnerProduct.lean.