Theorem 2 — in two-party Bell tests with M_A and M_B inputs, η*_NS ≤ (M_A + M_B − 2)/(M_A M_B − 1)
ProvedBellDetEff.TwoParty.theorem_2Consider two-party Bell tests in which A has inputs and B has inputs, with and . Let be the efficiency bound of eq. (A3): the infimum, over all no-signaling behaviours with these inputs and arbitrary finite output alphabets, of the largest detector efficiency at which the behaviour can still be simulated by a local hidden variable model. Then
Together with the matching lower bound of Theorem 1 (obtained in the paper by citing a local model of Massar and Pironio), this shows that is the exact detection-efficiency threshold for demonstrating the nonlocality of the maximally nonlocal no-signaling theory with two parties. In particular it tends to as the number of inputs grows.
Formalization Note The hypotheses and are implicit in the paper. At the right-hand side is , which Lean evaluates to , and the claim would be false: with one input each, every behaviour is local and . The bound is a real number, with no natural-number subtraction. is the infimum over all output alphabets and all no-signaling behaviours, and local hidden variable models range over arbitrary probability spaces with measurable response functions.
import Mathlib import Definitions.Def_BellDetEff_TwoParty_Efficiency
namespace BellDetEff.TwoParty
/-- Theorem 2 (Cao–Peng, arXiv:1607.04350v1, p. 2, eq. (5)): in two-party Bell tests with `MA`
and `MB` inputs, `η*_NS ≤ (MA + MB − 2)/(MA MB − 1)`. -/
theorem theorem_2 (MA MB : ℕ) (hA : 1 ≤ MA) (hB : 1 ≤ MB) (hAB : 2 ≤ MA * MB) :
etaStarNS MA MB ≤ ((MA : ℝ) + MB - 2) / ((MA : ℝ) * MB - 1) := by sorry
end BellDetEff.TwoParty
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.