import OAI.Analysis.Fourier.SquareFunction.KernelTail open OAI.Analysis.Fourier.Wave /-! Regularity of the actual convolutions entering the finite-vector sharp argument. In particular its `L¹` hypothesis follows from kernel decay. -/ namespace OAI.Analysis.Fourier.SquareFunction open _root_.MeasureTheory Set open scoped ENNReal Convolution noncomputable def convolveKernel {n : ℕ} {H : Type*} [NormedAddCommGroup H] [NormedSpace ℂ H] (K : Space n → H) (f : Space n → ℂ) (x : Space n) : H := ∫ y, f y • K (x - y) theorem integrable_convolveKernel {n : ℕ} {H : Type*} [NormedAddCommGroup H] [NormedSpace ℂ H] [CompleteSpace H] {K : Space n → H} {f : Space n → ℂ} (hf : Integrable f) (hK : Integrable K) : Integrable (convolveKernel K f) := hf.integrable_convolution (L := ContinuousLinearMap.lsmul ℂ ℂ) hK theorem continuous_convolveKernel {n : ℕ} {H : Type*} [NormedAddCommGroup H] [NormedSpace ℂ H] [CompleteSpace H] {K : Space n → H} {f : Space n → ℂ} (hf : Integrable f) (hK : Continuous K) (hB : BddAbove (range fun x => ‖K x‖)) : Continuous (convolveKernel K f) := hB.continuous_convolution_right_of_integrable (L := ContinuousLinearMap.lsmul ℂ ℂ) hf hK theorem integrable_convolveKernel_integrand {n : ℕ} {H : Type*} [NormedAddCommGroup H] [NormedSpace ℂ H] [CompleteSpace H] {K : Space n → H} {f : Space n → ℂ} (hf : Integrable f) (hK : Continuous K) (hB : BddAbove (range fun x => ‖K x‖)) (x : Space n) : Integrable (fun y => f y • K (x - y)) := by have hb : BddAbove ((fun i => ‖K i‖) '' ((fun t => x - t) ⁻¹' univ)) := by simpa only [preimage_univ, image_univ] using hB exact hb.convolutionExistsAt (L := ContinuousLinearMap.lsmul ℂ ℂ) MeasurableSet.univ (subset_univ _) hf.integrableOn hK.aestronglyMeasurable theorem norm_convolveKernel_le {n : ℕ} {H : Type*} [NormedAddCommGroup H] [NormedSpace ℂ H] [CompleteSpace H] {K : Space n → H} {f : Space n → ℂ} (hf : Integrable f) (hK : Continuous K) {B : ℝ} (hB : ∀ x, ‖K x‖ ≤ B) (x : Space n) : ‖convolveKernel K f x‖ ≤ (∫ y, ‖f y‖) * B := by have hbi : BddAbove (range fun x => ‖K x‖) := ⟨B, by rintro _ ⟨y, rfl⟩; exact hB y⟩ calc _ ≤ ∫ y, ‖f y • K (x - y)‖ := norm_integral_le_integral_norm _ _ ≤ ∫ y, ‖f y‖ * B := by apply integral_mono (integrable_convolveKernel_integrand hf hK hbi x).norm (hf.norm.mul_const B) intro y dsimp only rw [norm_smul] exact mul_le_mul_of_nonneg_left (hB _) (norm_nonneg _) _ = _ := integral_mul_const _ _ /-- Polynomial kernel decay supplies the global first moment required to remove the root-average error in the sharp-function argument. -/ theorem integrable_of_polynomial_decay {n : ℕ} {H : Type*} [NormedAddCommGroup H] {a q C : ℝ} (ha : 0 < a) (hq : (n : ℝ) < q) {K : Space n → H} (hKm : AEStronglyMeasurable K) (hK : ∀ u, ‖K u‖ ≤ C * a ^ n * (a * ‖u‖ + 1) ^ (-q)) : Integrable K := by refine ⟨hKm, ?_⟩ change (∫⁻ u, ‖K u‖ₑ) < ∞ have hmass : (∫⁻ u : Space n, ENNReal.ofReal ((a * ‖u‖ + 1) ^ (-q))) < ∞ := by have h := polynomial_kernel_mass_scaling ha q (0 : Space n) simp only [dist_zero_right] at h rw [h] exact ENNReal.mul_lt_top ENNReal.ofReal_lt_top (polynomialKernelMass_lt_top hq) calc _ ≤ ∫⁻ u, ENNReal.ofReal (C * a ^ n) * ENNReal.ofReal ((a * ‖u‖ + 1) ^ (-q)) := by apply lintegral_mono intro u have h := ENNReal.ofReal_le_ofReal (hK u) rw [ofReal_norm, ENNReal.ofReal_mul' (Real.rpow_nonneg (by positivity) _)] at h exact h _ = ENNReal.ofReal (C * a ^ n) * ∫⁻ u : Space n, ENNReal.ofReal ((a * ‖u‖ + 1) ^ (-q)) := lintegral_const_mul' _ _ ENNReal.ofReal_ne_top _ < ∞ := ENNReal.mul_lt_top ENNReal.ofReal_lt_top hmass theorem bounded_of_polynomial_decay {n : ℕ} {H : Type*} [NormedAddCommGroup H] {a q C : ℝ} (ha : 0 ≤ a) (hq : 0 ≤ q) (hC : 0 ≤ C) {K : Space n → H} (hK : ∀ u, ‖K u‖ ≤ C * a ^ n * (a * ‖u‖ + 1) ^ (-q)) : ∀ u, ‖K u‖ ≤ C * a ^ n := by intro u calc _ ≤ C * a ^ n * (a * ‖u‖ + 1) ^ (-q) := hK u _ ≤ C * a ^ n * 1 := mul_le_mul_of_nonneg_left (Real.rpow_le_one_of_one_le_of_nonpos (by nlinarith [mul_nonneg ha (norm_nonneg u)]) (neg_nonpos.mpr hq)) (mul_nonneg hC (pow_nonneg ha _)) _ = _ := mul_one _ /-- Exact local/tail splitting of the actual convolution, justified by integrability of both summands. -/ theorem convolveKernel_indicator_add_compl {n : ℕ} {H : Type*} [NormedAddCommGroup H] [NormedSpace ℂ H] [CompleteSpace H] {K : Space n → H} {f : Space n → ℂ} (hf : Integrable f) (hK : Continuous K) (hB : BddAbove (range fun x => ‖K x‖)) {P : Set (Space n)} (hP : MeasurableSet P) (x : Space n) : convolveKernel K f x = convolveKernel K (P.indicator f) x + convolveKernel K (Pᶜ.indicator f) x := by rw [convolveKernel, convolveKernel, convolveKernel, ← integral_add (integrable_convolveKernel_integrand (hf.indicator hP) hK hB x) (integrable_convolveKernel_integrand (hf.indicator hP.compl) hK hB x)] apply integral_congr_ae filter_upwards [] with y by_cases hy : y ∈ P <;> simp [hy] end OAI.Analysis.Fourier.SquareFunction