Closed cat-code joint probabilities and multiplicities
ProvedQAlgorithms.Papers.Independent.Support50.P40.cat_jointProbabilitiesLet and let , . Encode with the ordinary cat code and decode by the CNOT network of Fig. 7. Fix one syndrome string of length with Hamming weight . Let be its joint probability with residual Bell outcome , rather than a conditional probability or the sum over all syndromes of weight .
The four masses are
There are exactly syndromes of weight , for . Consequently, writing , the exact signed yield is
All zero-log terms are zero. The definition uses the independent physical Pauli law and the actual decoder's residual label, so these formulas identify those probabilities. At the empty syndrome gives masses and the hashing yield.
import Definitions.Def_QAlgorithms_Papers_Independent_Support50_P40
namespace QAlgorithms.Papers.Independent.Support50.P40
open QAlgorithms QAlgorithms.Stabilizer
open scoped ComplexConjugate Matrix
noncomputable section
/-- PDF pp.17--18, Eqs.(36)--(39): the four joint masses for ONE specified
syndrome of weight r. Multiplicity is absent here; summing all syndromes
of that weight introduces `choose(m,r)`. Block size is p=m+1, including p=1. -/
theorem cat_jointProbabilities (m : ℕ) (f : ℝ) (hf0 : 0 ≤ f) (hf1 : f ≤ 1)
(i : Qubits m) :
let p := m + 1
let r := hammingWeight i
let g := noiseG f
catJoint m f i .X = (2 : ℝ) ^ (p - r - 1) * g ^ (p - r) * (f + g) ^ r ∧
catJoint m f i .Y = (2 : ℝ) ^ (p - r - 1) * g ^ (p - r) * (f + g) ^ r ∧
catJoint m f i .I = (if r = 0 then ((f + g) ^ p + (f - g) ^ p) / 2
else (2 : ℝ) ^ (r - 1) * g ^ r * (f + g) ^ (p - r)) ∧
catJoint m f i .Z = (if r = 0 then ((f + g) ^ p - (f - g) ^ p) / 2
else (2 : ℝ) ^ (r - 1) * g ^ r * (f + g) ^ (p - r)) ∧
(∀ t : ℕ, t ≤ m →
(Finset.univ.filter (fun s : Qubits m => hammingWeight s = t)).card = Nat.choose m t) ∧
catYield m f =
(1 + ∑ t ∈ Finset.range (m + 1), (Nat.choose m t : ℝ) *
(let a := ∑ j : Pauli, catClosedMass m f t j;
-a * Real.logb 2 a + ∑ j : Pauli,
catClosedMass m f t j * Real.logb 2 (catClosedMass m f t j))) / (m + 1) := by sorry
end
end QAlgorithms.Papers.Independent.Support50.P40
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
For every integer , every real satisfying , and every bit string , put , , and . An error label is a string , with weight equal to its number of nonidentity entries and mass . Define its amplitude bits by , , and its phase bits by , . Its syndrome is the -bit string ; its logical label is according as is . Let be the sum of the error masses over labels with and . Then all four equalities hold:
Moreover, for every integer with , exactly strings in have weight . For , let denote the corresponding right-hand side above with , and put . Writing , with zero terms interpreted using , and , the quantity defined by equals
The statement includes , when the syndrome is the unique empty string and , as well as and ; powers with exponent zero have value , including .
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.