math / data /lean /OAI /Analysis /Fourier /SquareFunction /GoodLambda.lean
introvoyz045's picture
Migrated from GitHub
4a56cd6 verified
Raw History Blame Contribute Delete
4.57 kB
import OAI.Analysis.Fourier.SquareFunction.Maximal
/-! The layer-cake and absorption step of the dyadic sharp-function argument. -/
namespace OAI.Analysis.Fourier.SquareFunction
open _root_.MeasureTheory Set
open scoped ENNReal
/-- Distribution-function inequalities add with exactly the same constants
after taking positive moments. -/
theorem power_moment_le_of_superlevel_add {X : Type*} [MeasurableSpace X]
(μ : Measure X) {u v w : X → ℝ}
(hu : Measurable u) (hv : Measurable v) (hw : Measurable w)
(hun : ∀ x, 0 ≤ u x) (hvn : ∀ x, 0 ≤ v x) (hwn : ∀ x, 0 ≤ w x)
{p : ℝ} (hp : 0 < p) (c : ℝ≥0∞)
(h : ∀ t > 0, μ {x | t < u x} ≤ μ {x | t < v x} + c * μ {x | t < w x}) :
(∫⁻ x, ENNReal.ofReal (u x) ^ p ∂μ) ≤
(∫⁻ x, ENNReal.ofReal (v x) ^ p ∂μ) +
c * ∫⁻ x, ENNReal.ofReal (w x) ^ p ∂μ := by
have hm (f : X → ℝ) : Measurable (fun t : ℝ => μ {x | t < f x} *
ENNReal.ofReal p * ENNReal.ofReal t ^ (p - 1)) := by
have hmono : Antitone (fun t : ℝ => μ {x | t < f x}) :=
fun _ _ hst => measure_mono (fun _ hx => lt_of_le_of_lt hst hx)
exact (hmono.measurable.mul measurable_const).mul
(measurable_id.ennreal_ofReal.pow_const _)
rw [lintegral_power_layercake μ hu hun hp,
lintegral_power_layercake μ hv hvn hp, lintegral_power_layercake μ hw hwn hp]
calc
_ ≤ ∫⁻ t in Ioi (0 : ℝ), (μ {x | t < v x} * ENNReal.ofReal p *
ENNReal.ofReal t ^ (p - 1)) + c *
(μ {x | t < w x} * ENNReal.ofReal p * ENNReal.ofReal t ^ (p - 1)) := by
apply lintegral_mono_ae
filter_upwards [ae_restrict_mem measurableSet_Ioi] with t ht
calc
_ ≤ (μ {x | t < v x} + c * μ {x | t < w x}) *
ENNReal.ofReal p * ENNReal.ofReal t ^ (p - 1) := by
gcongr
exact h t ht
_ = _ := by ring
_ = _ := by rw [lintegral_add_left (hm v), lintegral_const_mul c (hm w)]
/-- The small same-side term in a good-lambda estimate can be absorbed when
the left moment is finite. This is why the source first works with finite
families whose square function already belongs to `Lᵖ`. -/
theorem ennreal_absorb {U V : ℝ≥0∞} {θ : ℝ}
(hθ : 0 ≤ θ) (hθ1 : θ < 1) (hU : U < ∞)
(h : U ≤ V + ENNReal.ofReal θ * U) :
U ≤ ENNReal.ofReal (1 / (1 - θ)) * V := by
have hs : 0 < 1 - θ := sub_pos.mpr hθ1
have hc : 0 < (1 : ℝ) / (1 - θ) := div_pos zero_lt_one hs
by_cases hV : V = ∞
· rw [hV, ENNReal.mul_top (ENNReal.ofReal_ne_zero_iff.mpr hc)]
exact le_top
have hmul : ENNReal.ofReal θ * U ≠ ∞ :=
ENNReal.mul_ne_top ENNReal.ofReal_ne_top hU.ne
have hr := (ENNReal.toReal_le_toReal hU.ne (ENNReal.add_ne_top.mpr ⟨hV, hmul⟩)).mpr h
rw [ENNReal.toReal_add hV hmul, ENNReal.toReal_mul, ENNReal.toReal_ofReal hθ] at hr
apply (ENNReal.toReal_le_toReal hU.ne
(ENNReal.mul_ne_top ENNReal.ofReal_ne_top hV)).mp
rw [ENNReal.toReal_mul, ENNReal.toReal_ofReal hc.le]
calc
U.toReal ≤ V.toReal / (1 - θ) := (le_div_iff₀ hs).mpr (by nlinarith)
_ = (1 / (1 - θ)) * V.toReal := by ring
/-- A good-lambda distribution inequality yields the corresponding strong
positive-moment estimate, with its explicit absorption constant. -/
theorem power_moment_le_of_good_lambda {X : Type*} [MeasurableSpace X]
(μ : Measure X) {u v : X → ℝ} (hu : Measurable u) (hv : Measurable v)
(hun : ∀ x, 0 ≤ u x) (hvn : ∀ x, 0 ≤ v x)
{p a η : ℝ} (hp : 0 < p) (ha : 0 ≤ a) (hη : 0 ≤ η)
(hsmall : η * a ^ p < 1)
(hfin : (∫⁻ x, ENNReal.ofReal (u x) ^ p ∂μ) < ∞)
(hgood : ∀ t > 0, μ {x | t < u x} ≤ μ {x | t < v x} +
ENNReal.ofReal η * μ {x | t < a * u x}) :
(∫⁻ x, ENNReal.ofReal (u x) ^ p ∂μ) ≤
ENNReal.ofReal (1 / (1 - η * a ^ p)) *
∫⁻ x, ENNReal.ofReal (v x) ^ p ∂μ := by
have hscale : (∫⁻ x, ENNReal.ofReal (a * u x) ^ p ∂μ) =
ENNReal.ofReal (a ^ p) * ∫⁻ x, ENNReal.ofReal (u x) ^ p ∂μ := by
simp_rw [ENNReal.ofReal_mul ha, ENNReal.mul_rpow_of_nonneg _ _ hp.le,
ENNReal.ofReal_rpow_of_nonneg ha hp.le]
exact lintegral_const_mul _ (hu.ennreal_ofReal.pow_const p)
have h := power_moment_le_of_superlevel_add μ hu hv (hu.const_mul a)
hun hvn (fun x => mul_nonneg ha (hun x)) hp (ENNReal.ofReal η) hgood
rw [hscale, ← mul_assoc, ← ENNReal.ofReal_mul hη] at h
exact ennreal_absorb (mul_nonneg hη (Real.rpow_nonneg ha _)) hsmall hfin h
end OAI.Analysis.Fourier.SquareFunction