Weighted Cauchy–Schwarz with coordinate duplication
ProvedGreenTao.coordinate_cauchy_schwarzadditive-combinatoricsgowers-uniformitygreen-tao
Let be finite sets, let and , and suppose and . With uniform averages,
This is the weighted coordinate-duplication step of the repeated Cauchy–Schwarz argument: collects the other variables and are independent copies of the coordinate being eliminated. It is valid for signed and nonnegative weights that can vanish.
Preamble
import Definitions.Def_GreenTao_Pseudorandom import Mathlib open scoped BigOperators Topology open Filter GreenTao
Formal statement
theorem GreenTao.coordinate_cauchy_schwarz {α β : Type} [Fintype α] [Fintype β]
(G w : α → ℝ) (H : α → β → ℝ)
(hw : ∀ a, 0 ≤ w a) (hG : ∀ a, |G a| ≤ w a) :
(avg fun a => G a * avg (H a)) ^ 2 ≤
avg w * avg (fun a => avg fun b => avg fun c => w a * H a b * H a c) := by sorrySource
Green and Tao, The primes contain arbitrarily long arithmetic progressions, https://arxiv.org/html/math/0404188v6, §5 Lemma 5.4, equation (5.19), and its proof by separating a coordinate and applying weighted Cauchy–Schwarz. This is its coordinate-invariant finite averaging step, not the full indexed J_d/P_d specialization.