Smoothness and derivative bound for x-derivative slices
ProvedcontDiff_iteratedDeriv_slice_and_norm_iteratedFDeriv_le_norm_iteratedFDeriv_addLet be a natural number and let be a function which is over (smoothness of order in ); fix and . Consider the slice function sending to , that is, the -th derivative in the first variable of , taken at the point , with the second variable frozen at . The conclusion is a conjunction: first, is over on ; second, for every and every , the norm of the -th iterated Fréchet derivative of at , as a continuous multilinear map, is at most the norm of the -th iterated Fréchet derivative of at the point . Thus the bound holds with constant , uniformly in and .
This is the quantitative slicing estimate for a smooth function of one real variable and further real variables: partial differentiation times in the first variable and restriction to the slice costs nothing in operator norm beyond shifting the order of the total derivative by . It is used in the construction of smooth bounds for derivatives of oscillatory integrals, feeding MeasureTheory.exists_forall_contDiff_norm_iteratedDeriv_integral_cexp_mul_le_prod_of_contDiff.
import Mathlib set_option maxHeartbeats 4000000 set_option synthInstance.maxHeartbeats 400000 set_option backward.isDefEq.respectTransparency.types false set_option autoImplicit false open MeasureTheory
theorem contDiff_iteratedDeriv_slice_and_norm_iteratedFDeriv_le_norm_iteratedFDeriv_add
{n : ℕ} (Φ : ℝ × (Fin n → ℝ) → ℂ) (hΦ : ContDiff ℝ (⊤ : ℕ∞) Φ) (j : ℕ) (x : ℝ) :
ContDiff ℝ (⊤ : ℕ∞) (fun y' : Fin n → ℝ => iteratedDeriv j (fun t : ℝ => Φ (t, y')) x) ∧
∀ (N : ℕ) (y : Fin n → ℝ),
‖iteratedFDeriv ℝ N (fun y' : Fin n → ℝ => iteratedDeriv j (fun t : ℝ => Φ (t, y')) x) y‖ ≤
‖iteratedFDeriv ℝ (N + j) Φ (x, y)‖ := by sorry