File size: 5,420 Bytes
4a56cd6 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 | import Mathlib
/-!
Elementary transfer estimates used in `sections/transfer.tex`.
These results concern actual amplitude bounds and convergence of the dyadic
series; they do not assume a local smoothing estimate.
-/
namespace OAI.Analysis.Fourier.WavePackets
open _root_.MeasureTheory
/-- The pointwise estimate on an amplitude shell used to pass from the mixed
`LΒ²`/`Lβ` estimate to an `Lp` estimate. -/
theorem amplitude_shell_bound {a x p : β} (ha : 0 β€ a) (hax : a β€ x)
(hp : 2 β€ p) : a ^ (p - 2) * x ^ 2 β€ x ^ p := by
have hx : 0 β€ x := ha.trans hax
calc
a ^ (p - 2) * x ^ 2 β€ x ^ (p - 2) * x ^ 2 :=
mul_le_mul_of_nonneg_right (Real.rpow_le_rpow ha hax (sub_nonneg.mpr hp))
(sq_nonneg x)
_ = x ^ p := by
by_cases hx0 : x = 0
Β· subst x
simp [Real.zero_rpow (by linarith : p β 0)]
Β· rw [β Real.rpow_natCast, β Real.rpow_add (lt_of_le_of_ne hx (Ne.symm hx0))]
congr 1
norm_num
/-- Integrated amplitude-shell estimate. The integration measure can already
be restricted to a shell, so the hypothesis is only almost everywhere. -/
theorem integral_amplitude_shell_bound {Ξ± : Type*} [MeasurableSpace Ξ±]
{ΞΌ : Measure Ξ±} {f : Ξ± β β} {a p : β} (ha : 0 β€ a) (hp : 2 β€ p)
(hf : βα΅ x βΞΌ, a β€ f x)
(hβ : Integrable (fun x => f x ^ 2) ΞΌ)
(hβ : Integrable (fun x => f x ^ p) ΞΌ) :
a ^ (p - 2) * (β« x, f x ^ 2 βΞΌ) β€ β« x, f x ^ p βΞΌ := by
rw [β integral_const_mul]
exact integral_mono_ae (hβ.const_mul _) hβ
(hf.mono fun x hx => amplitude_shell_bound ha hx hp)
/-- The logarithmic Sobolev series in the last paragraph of the transfer
section converges precisely above the stated loss of one logarithm. -/
theorem logarithmic_dyadic_summable_iff (A B : β) :
Summable (fun j : β => (1 + (j : β)) ^ (A - B)) β A + 1 < B := by
have h := (summable_nat_add_iff 1 (f := fun j : β => (j : β) ^ (A - B)))
simp only [Nat.cast_add, Nat.cast_one] at h
rw [show (fun j : β => (1 + (j : β)) ^ (A - B)) =
(fun j : β => ((j : β) + 1) ^ (A - B)) by
funext j
rw [add_comm]]
rw [h, Real.summable_nat_rpow]
constructor <;> intro h <;> linarith
/-- Any positive Sobolev loss absorbs the polynomial loss in the dyadic
frequency index. This is the series used for the positive-loss conclusion. -/
theorem positive_loss_dyadic_summable (k : β) {Ξ΅ : β} (hΞ΅ : 0 < Ξ΅) :
Summable (fun j : β => (2 : β) ^ (-(j : β) * Ξ΅) * (1 + (j : β)) ^ k) := by
let r : β := Ξ΅ * Real.log 2
have hr : 0 < r := mul_pos hΞ΅ (Real.log_pos (by norm_num))
have hs := (summable_nat_add_iff 1).mpr
(Real.summable_pow_mul_exp_neg_nat_mul k hr)
convert hs.mul_left (Real.exp r) using 1
ext j
simp only [Nat.cast_add, Nat.cast_one, Real.rpow_def_of_pos (by norm_num : (0 : β) < 2)]
have he : Real.log 2 * (-(j : β) * Ξ΅) = r + -r * ((j : β) + 1) := by
dsimp [r]
ring
rw [he, Real.exp_add]
ring
/-- The summation step for a logarithmically weighted dyadic decomposition,
in any complete normed target space (in particular, spacetime `Lp`). -/
theorem logarithmic_dyadic_sum_bound {E : Type*} [NormedAddCommGroup E]
[CompleteSpace E] (u : β β E) {A B C : β} (hB : A + 1 < B)
(hu : β j : β, βu jβ β€ C * (1 + (j : β)) ^ (A - B)) :
Summable u β§ ββ' j, u jβ β€ C * β' j : β, (1 + (j : β)) ^ (A - B) := by
have hs := ((logarithmic_dyadic_summable_iff A B).mpr hB).mul_left C
have hs' := hs.of_norm_bounded hu
refine β¨hs', ?_β©
simpa only [tsum_mul_left] using hs'.hasSum.norm_le_of_bounded hs.hasSum hu
/-- The positive-loss summation step for a dyadic decomposition. -/
theorem positive_loss_dyadic_sum_bound {E : Type*} [NormedAddCommGroup E]
[CompleteSpace E] (u : β β E) (k : β) {Ξ΅ C : β} (hΞ΅ : 0 < Ξ΅)
(hu : β j : β, βu jβ β€ C * ((2 : β) ^ (-(j : β) * Ξ΅) * (1 + (j : β)) ^ k)) :
Summable u β§ ββ' j, u jβ β€
C * β' j : β, (2 : β) ^ (-(j : β) * Ξ΅) * (1 + (j : β)) ^ k := by
have hs := (positive_loss_dyadic_summable k hΞ΅).mul_left C
have hs' := hs.of_norm_bounded hu
refine β¨hs', ?_β©
simpa only [tsum_mul_left] using hs'.hasSum.norm_le_of_bounded hs.hasSum hu
/-- The quantitative absence of stationary points away from the cone:
the phase gradient is `X - T β’ v`, where `v` is a unit frequency vector.
This is the lower bound used by the off-cone integration-by-parts argument. -/
theorem off_cone_phase_gradient_lower_bound {E : Type*} [NormedAddCommGroup E]
[NormedSpace β E] (X v : E) {T : β} (hv : βvβ = 1)
(hTβ : 0 β€ T) (hTβ : T β€ 2) (hX : 3 β€ βXβ) :
βXβ / 3 β€ βX - T β’ vβ := by
have h := norm_sub_norm_le X (T β’ v)
rw [norm_smul, Real.norm_eq_abs, abs_of_nonneg hTβ, hv, mul_one] at h
linarith
/-- Exact exponent arithmetic in the conversion from the sector estimate to
the periodic critical estimate. -/
theorem sector_logarithmic_exponent (n : β) :
3 * (n + 3) / 2 + (n - 1) / 2 + 1 = 2 * n + 5 := by ring
/-- The interpolation power appearing between the critical and `Lβ` bounds. -/
theorem interpolated_exponent {n p : β} (hn : n β 1) (hp : p β 0) :
(n - 1) / 2 * (1 - (2 * n / (n - 1)) / p) = (n - 1) / 2 - n / p := by
field_simp
end OAI.Analysis.Fourier.WavePackets
|