Event-based setting of the causal inequality: causal relation of , joint distribution of ,
DefinitionOCB2012_appAcausal-inequalitiesquantum-information
Definitions for the event-based derivation of the causal inequality in Appendix A of Oreshkov, Costa and Brukner (2012).
- Causal relation . Assumption CS places the events and (the system entering Alice's and Bob's laboratory) in a causal structure. Their relation is then exactly one of , , or , encoded as
CausalRel.AB,CausalRel.BAandCausalRel.incomparable. - Joint distribution.
EventDistis a joint distribution of Alice's bit , Bob's bits , the guesses (Alice) and (Bob), and . Bits are encoded as Booleans, withfalse.IsProbrequires it to be non-negative and normalized.prgives the probability of an event, andcondthe conditional probability , set to when . - Success probability, Eq. (9):
Definition code
import Mathlib
/-!
# Event-based setting of the causal inequality (Oreshkov–Costa–Brukner 2012, App. A)
O. Oreshkov, F. Costa, Č. Brukner, *Quantum correlations with no causal order*,
Nat. Commun. 3, 1092 (2012), arXiv:1105.4464v3, Appendix A ("Formal derivation of the causal
inequality"), pp. 6–7.
The relevant random variables are Alice's bit `a`, Bob's bits `b, b'`, the guesses `x` (Alice)
and `y` (Bob), and the causal relation `R` between the events `A1` and `B1` (the system
entering Alice's and Bob's laboratory). Bits are encoded as `Bool` (`false ↦ 0`, `true ↦ 1`).
Under assumption CS the three possibilities `A1 ⪯ B1`, `B1 ⪯ A1`, `A1 ⋠⋡ B1` are mutually
exclusive and exhaustive, so `R` takes exactly three values.
-/
namespace OCB2012
/-- App. A (CS): the causal relation between `A1` and `B1`. -/
inductive CausalRel
/-- `A1 ⪯ B1`: `A1` is in the causal past of `B1`. -/
| AB
/-- `B1 ⪯ A1`: `B1` is in the causal past of `A1`. -/
| BA
/-- `A1 ⋠⋡ B1`: neither is in the causal past of the other. -/
| incomparable
deriving DecidableEq
instance : Fintype CausalRel :=
⟨{CausalRel.AB, CausalRel.BA, CausalRel.incomparable}, fun x => by cases x <;> simp⟩
/-- A joint distribution of `(a, b, b', x, y, R)`, listed in this order. -/
abbrev EventDist := Bool → Bool → Bool → Bool → Bool → CausalRel → ℝ
namespace EventDist
/-- `P` is a probability distribution. -/
def IsProb (P : EventDist) : Prop :=
(∀ a b b' x y R, 0 ≤ P a b b' x y R) ∧
∑ a, ∑ b, ∑ b', ∑ x, ∑ y, ∑ R, P a b b' x y R = 1
/-- The probability `p(E)` of an event `E`, given as a Boolean predicate of
`(a, b, b', x, y, R)`. -/
noncomputable def pr (P : EventDist) (E : Bool → Bool → Bool → Bool → Bool → CausalRel → Bool) :
ℝ :=
∑ a, ∑ b, ∑ b', ∑ x, ∑ y, ∑ R, if E a b b' x y R then P a b b' x y R else 0
/-- The conditional probability `p(E | F) = p(E ∧ F) / p(F)` (equal to `0` when `p(F) = 0`). -/
noncomputable def cond (P : EventDist)
(E F : Bool → Bool → Bool → Bool → Bool → CausalRel → Bool) : ℝ :=
P.pr (fun a b b' x y R => E a b b' x y R && F a b b' x y R) / P.pr F
/-- Eq. (9): the success probability `p_succ = ½ p(x = b | b' = 0) + ½ p(y = a | b' = 1)`. -/
noncomputable def pSucc (P : EventDist) : ℝ :=
(1 / 2 : ℝ) * P.cond (fun _ b _ x _ _ => decide (x = b)) (fun _ _ b' _ _ _ => decide (b' = false)) +
(1 / 2 : ℝ) * P.cond (fun a _ _ _ y _ => decide (y = a)) (fun _ _ b' _ _ _ => decide (b' = true))
end EventDist
end OCB2012
Source
O. Oreshkov, F. Costa, C. Brukner, Quantum correlations with no causal order, Nat. Commun. 3, 1092 (2012), https://arxiv.org/abs/1105.4464v3, Appendix A, pp. 6-7 (assumptions CS, FC, CL and Eqs. (9)-(10))