Chapter NsCutoffUniformity
DefinitionChapterNsCutoffUniformitydefinitionsformalizationtimepiece
Formal definitions for the timepiece Lean 4 formalization (source chapter BookProof/ChapterNsCutoffUniformity.lean): generated def bundle for ChapterNsCutoffUniformity. See BookProof/ChapterNsCutoffUniformity.lean for full context.
Definition code
import Definitions.Def_ChapterNsFourierElimination
import Definitions.Def_ChapterNavierStokesFullEulerianFock
import Mathlib
/-!
# Uniformity of the reduced forms under the energy cutoff
The first of the two residual obligations of item 4 of the Navier–Stokes plan items of
`CONSOLIDATED_PLAN.md` (§5.4(1) of `DESIGN_COMPARISON_N_20260915.md`): after the Fourier
elimination the reduced constraint forms carry the mode coefficients `i k_j` and `−|k|²`, and
under the energy cutoff `|k_j| ≤ Λ` they must be bounded by `O(Λ)` **uniformly in the parcel
number `n`** — that uniformity is what makes the `ℓ²`-lift of the sector Hamiltonians possible,
and it is why the cutoff is load-bearing rather than cosmetic.
What is proved here about the reduced family `redFormPoly` of
`BookProof.ChapterNsFourierElimination`:
* `norm_coeff_redVisc_le`, `norm_coeff_redAdvectPoly_le`, `norm_coeff_redMomentumPoly_le` and the
headline `norm_coeff_redFormPoly_le` — **every coefficient of every reduced form of every parcel
is bounded by `cutoffBound nu Λ = 1 + 3Λ + 3|ν|Λ²`** as soon as `|k_j| ≤ Λ`. The bound does not
depend on the parcel number `n`, on the parcel `p`, on which of the seven forms is taken, or on
the monomial: it is `O(Λ)` in exactly the sense the obligation asks for (and `O(Λ²)` through the
viscous coefficient `ν|k|²`);
* `totalDegree_redFormPoly_le` — the reduced forms are of degree at most two, again uniformly;
* `vars_redFormPoly_subset` — **parcel locality**: a reduced form involves only the six
coordinates of its own parcel. With the fixed number `7` of forms per parcel this is the
sparsity that makes the row/column (Schur) data of the sector Hamiltonians independent of `n`;
* `norm_coeff_redFormPoly_unbounded_of_no_cutoff` — the cutoff is **necessary**: without it the
coefficients of the reduced family are unbounded, since the advection coefficient is `k_j`
itself.
**What this does and does not settle.** It settles the coefficient half of the uniformity
obligation — the data entering the Schur bounds are `O(Λ)` and parcel local, uniformly in `n`.
The operator-level relative bound `‖H_n x‖ ≤ K ‖(N_n + 1) x‖` with `K` independent of `n` is not
proved here; on the landed route it is not needed either, because the comparison actually used is
the lifted Friedrichs extension of the reduced Hamiltonian itself
(`BookProof.NsFullEuler.nsRedFullOuterN_esa`), for which the Faris–Lavine commutator constant is
`c = 0`.
Everything is `sorry`-free and `axiom`-free.
-/
namespace BookProof.NsCutoffUniformity
open MvPolynomial BookProof.NsFullEuler
noncomputable section
variable {n : ℕ}
/-! ## 1. Coefficient bounds for one monomial -/
variable {ι : Type*}
theorem norm_coeff_X_le (a : ι) (m : ι →₀ ℕ) : ‖coeff m (X a : MvPolynomial ι ℂ)‖ ≤ 1 := by
classical
rw [coeff_X']
split <;> simp
theorem norm_coeff_C_mul_X_le (c : ℂ) (a : ι) (m : ι →₀ ℕ) :
‖coeff m (C c * X a : MvPolynomial ι ℂ)‖ ≤ ‖c‖ := by
classical
rw [coeff_C_mul, norm_mul, coeff_X']
split <;> simp
theorem norm_coeff_C_mul_X_mul_X_le (c : ℂ) (a b : ι) (m : ι →₀ ℕ) :
‖coeff m (C c * (X a * X b) : MvPolynomial ι ℂ)‖ ≤ ‖c‖ := by
classical
have hx : (X a * X b : MvPolynomial ι ℂ)
= monomial (Finsupp.single a 1 + Finsupp.single b 1) 1 := by
rw [X, X, monomial_mul, one_mul]
rw [hx, coeff_C_mul, norm_mul, coeff_monomial]
split <;> simp
/-! ## 2. The cutoff bound -/
/-- The uniform bound on the coefficients of the reduced forms under the cutoff `|k_j| ≤ Λ`. -/
def cutoffBound (nu Λ : ℝ) : ℝ := 1 + 3 * Λ + 3 * |nu| * Λ ^ 2
theorem cutoffBound_nonneg (nu : ℝ) {Λ : ℝ} (hΛ : 0 ≤ Λ) : 0 ≤ cutoffBound nu Λ := by
unfold cutoffBound
positivity
/-- Under the cutoff the viscous coefficient `ν |k|²` is bounded by `3 |ν| Λ²`. -/
theorem abs_visc_coeff_le {nu Λ : ℝ} {k : Fin 3 → ℝ} (hΛ : 0 ≤ Λ) (hk : ∀ j, |k j| ≤ Λ) :
|nu * ∑ j : Fin 3, (k j) ^ 2| ≤ 3 * |nu| * Λ ^ 2 := by
have hsum : ∑ j : Fin 3, (k j) ^ 2 ≤ 3 * Λ ^ 2 := by
have hbound : ∀ j : Fin 3, (k j) ^ 2 ≤ Λ ^ 2 := by
intro j
have := hk j
nlinarith [abs_nonneg (k j), sq_abs (k j)]
calc ∑ j : Fin 3, (k j) ^ 2 ≤ ∑ _j : Fin 3, Λ ^ 2 := Finset.sum_le_sum fun j _ => hbound j
_ = 3 * Λ ^ 2 := by simp [Finset.sum_const]
have hnonneg : 0 ≤ ∑ j : Fin 3, (k j) ^ 2 := Finset.sum_nonneg fun j _ => sq_nonneg _
rw [abs_mul, abs_of_nonneg hnonneg]
calc |nu| * ∑ j : Fin 3, (k j) ^ 2 ≤ |nu| * (3 * Λ ^ 2) := by
exact mul_le_mul_of_nonneg_left hsum (abs_nonneg nu)
_ = 3 * |nu| * Λ ^ 2 := by ring
/-! ## 3. The coefficient bounds of the reduced forms -/
/-- The real residual form has coefficients bounded by `1 + 3|ν|Λ²`. -/
theorem norm_coeff_redVisc_le (nu : ℝ) {Λ : ℝ} {k : Fin 3 → ℝ} (hΛ : 0 ≤ Λ)
(hk : ∀ j, |k j| ≤ Λ) (p : Fin n) (i : Fin 3) (m : Fin (n * 6) →₀ ℕ) :
‖coeff m (redVisc nu k n p i)‖ ≤ 1 + 3 * |nu| * Λ ^ 2 := by
rw [redVisc, coeff_add]
refine (norm_add_le _ _).trans ?_
gcongr
· exact norm_coeff_X_le _ _
· refine (norm_coeff_C_mul_X_le _ _ _).trans ?_
rw [Complex.norm_real, Real.norm_eq_abs]
exact abs_visc_coeff_le hΛ hk
/-! ## 4. Degree and parcel locality -/
section Degree
variable (nu : ℝ) (k : Fin 3 → ℝ)
end Degree
end
end BookProof.NsCutoffUniformity
Source