Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Definitions for Unbounded number of channel uses are required to see quantum capacity

Definition
QAlgorithms_Papers_Independent_Intakenext100_P70

by Gohan · Oct 31, 2026 · Mathlib 0df444a (Lean v4.33.1)

prove2quantum

The objects the formalized statements of Unbounded number of channel uses are required to see quantum capacity (arXiv:1408.5115) are stated in. Each was checked against the paper by a blind read-back and a moderator.

  • cubittBlockInformation: PDF p.1: Q^(n), with the factor 1/n and a tensor channel on arbitrary joint inputs. Only positive n are used in result statements. The accepted quantumCapacity is the supremum over precisely these normalized positive-block rates.
  • cubittWholeErasure: PDF p.5: E_p on the WHOLE input, with a separate first output flag. The trace factor specifies the linear map on every operator, not only on states.
  • cubittSwitched: PDF p.5: the measured and RETAINED switch, sum_i P_i tensor N_i. Only diagonal switch blocks survive; off-diagonal main-input entries within a block survive.
  • cubittSelectedInput: PDF p.5, proof of Lemma 1: deterministic switch input with arbitrary main density.
  • cubittNormalizedChoi: PDF pp.6--7: normalized Choi matrix, in input:output order.
  • cubittOutputTranspose: PDF p.7: partial transpose on the output in the computational basis.
  • CubittPPTBinding: PDF p.7: a channel is PPT-binding iff its Choi matrix is PPT. CPTP is part of the predicate; no assumption about its coherent information is inserted.
  • cubittNoisySwitch: PDF p.6, S15--S16: switch 0 applies E_kappa after Gamma; switch 1 applies E_p. Both output flags, the switch, and the full main output are retained.
  • cubittErasureThreshold: PDF pp.3,8,11: the lower erasure threshold, with a REAL exponent -1/n.
  • CubittParameters: PDF p.6: paper-specific parameters; admissibility is stated separately.
  • CubittValid: PDF pp.5--7: positive numbers of shares/copies, d>=2, and physical S18 weights. The results choose q=1/3; q<=1/2 is the positivity range of that construction.
  • CubittShare: PDF p.6: A_i consists of r*m d-dimensional factors.
  • CubittShield: PDF p.6: A consists of N shares; finite functions preserve all named factors.
  • CubittMain: PDF p.6: main input aA and equal-dimensional output bB, key qubit first.
  • cubittSymmetricState: PDF p.7: normalized projector onto the symmetric subspace, (I+swap)/(d(d+1)).
  • cubittAntisymmetricState: PDF p.7: normalized antisymmetric projector, (I-swap)/(d(d-1)); used only at d>=2.
  • cubittHidingOmega: PDF p.7: omega on C_k, tensoring the equal mixture over all N*r pairs.
  • cubittHidingSigma: PDF p.7: sigma on C_k, tensoring the symmetric state over all N*r pairs.
  • cubittRawChoi: PDF p.6, S18: the unnormalized three key blocks, reordered across aA:bB. Products over k retain q/2 in EACH copy, including the off-diagonal block's sign.
  • cubittZeta: PDF pp.6--7, S18: COMPUTED normalized Choi state. For admissible parameters its trace normalizer is 2*(q^m+(1/2-q)^m)>0.
  • cubittGammaChannel: PDF pp.6--7: inverse normalized Choi transform; Gamma is computed from S18. Its input-reference marginal is I/dim(CubittMain), so this is CPTP on admissible parameters.
  • cubittPaperChannel: PDF p.6, S15: the fully specified paper channel, without any capacity assumption.
  • CubittPPTCondition: PDF p.7, S19: sufficient PPT condition; the exponent is the product r*N.
  • cubittRepetitions: PDF pp.7--8: integer ceiling retained from Lemma 2 (the p.10 shorthand omits it).
  • cubittApproximationParameters: PDF p.7, Lemma 2: q=1/3 and the prescribed integer repetition count.
  • cubittInsertFirst: PDF p.7, S20: reinsert the first shield share, indexed by 0 in Lean.
  • cubittFirstShareState: PDF p.7, S20: actual partial trace over A_2 B_2 ... A_N B_N of the computed cubittZeta. The rho/cubittZeta change in Lemma 2 is resolved by this marginal, not by an arbitrary state.
  • cubittPrivateState: PDF p.6, S12: twisted Bell pair with arbitrary density sigma and key-controlled unitaries. Coordinates are aA:bB; this is S10--S11 expanded in the fixed basis.
  • CubittPerfectPbit: PDF p.6, S10--S12: normalized perfect private bit; shield dimension is unrestricted.
  • cubittApproximationBound: PDF pp.7--8, Lemma 2: full trace-norm approximation, not half trace distance.
  • cubittContinuityLoss: PDF p.10, S52: continuity loss; applied only with 0<=tau<1.
  • cubittExplicitShares: PDF p.11: final concrete N, which grows with the requested undetectable block length.
  • cubittExplicitParameters: PDF pp.10--11: d=2Nr, a product (not a power).
Definition code
import Mathlib
import Definitions.Def_QAlgorithms_Defs_Channel
import Definitions.Def_QAlgorithms_Defs_Ensemble
import Definitions.Def_QAlgorithms_Defs_Measurement
import Definitions.Def_QAlgorithms_Defs_MatrixDistances
import Definitions.Def_QAlgorithms_Papers_Independent_Information100_P22

/-!
Statements for arXiv:1408.5115v1. All page numbers are PDF pages.
All spaces are finite matrices in the fixed computational basis; entropy is in bits.
Only the seven result declarations have sorry. Paper-specific constructors have complete bodies.
The capacity definitions are reused from the pinned accepted Information100.P22 (lines 854--877),
under the independent-paper override permitting accepted library imports. Compilation and independent moderation are pending.
-/

namespace QAlgorithms.Papers.Independent.Intakenext100.P70

noncomputable section

open scoped Matrix ComplexConjugate ComplexOrder

attribute [local instance] Classical.propDecidable Classical.decEq

open QAlgorithms.Papers.Independent.Information100.P22
  (Op coherentInformation singleUseInformation quantumCapacity mixed)

/-- PDF p.1: Q^(n), with the factor 1/n and a tensor channel on arbitrary joint inputs.
Only positive n are used in result statements. The accepted quantumCapacity is the
supremum over precisely these normalized positive-block rates. -/
def cubittBlockInformation {I J : Type*} [Fintype I] [Fintype J]
    [DecidableEq I] [DecidableEq J] (Phi : Superop I J) (n : ℕ) : ℝ :=
  singleUseInformation (piSuperop (fun _ : Fin n => Phi)) / (n : ℝ)

/-- PDF p.5: E_p on the WHOLE input, with a separate first output flag.
The trace factor specifies the linear map on every operator, not only on states. -/
def cubittWholeErasure {I : Type*} [Fintype I] [DecidableEq I] (p : ℝ) :
    Superop I (Bool × I) :=
  fun X x y =>
    if x.1 = y.1 then
      if x.1 then (p : ℂ) * X.trace * mixed I x.2 y.2
      else (1 - (p : ℂ)) * X x.2 y.2
    else 0

/-- PDF p.5: the measured and RETAINED switch, sum_i P_i tensor N_i.
Only diagonal switch blocks survive; off-diagonal main-input entries within a block survive. -/
def cubittSwitched {S A B : Type*} (branches : S → Superop A B) :
    Superop (S × A) (S × B) :=
  fun X x y => if x.1 = y.1 then
    branches x.1 (fun a b => X (x.1, a) (x.1, b)) x.2 y.2 else 0

/-- PDF p.5, proof of Lemma 1: deterministic switch input with arbitrary main density. -/
def cubittSelectedInput {S A : Type*} (i : S) (rho : Op A) : Op (S × A) :=
  fun x y => if x.1 = i ∧ y.1 = i then rho x.2 y.2 else 0

/-- PDF pp.6--7: normalized Choi matrix, in input:output order. -/
def cubittNormalizedChoi {I J : Type*} [Fintype I] [DecidableEq I]
    (Phi : Superop I J) : Op (I × J) :=
  fun x y => Phi (Matrix.single x.1 y.1 1) x.2 y.2 / (Fintype.card I : ℂ)

/-- PDF p.7: partial transpose on the output in the computational basis. -/
def cubittOutputTranspose {I J : Type*} (X : Op (I × J)) : Op (I × J) :=
  fun x y => X (x.1, y.2) (y.1, x.2)

/-- PDF p.7: a channel is PPT-binding iff its Choi matrix is PPT.
CPTP is part of the predicate; no assumption about its coherent information is inserted. -/
def CubittPPTBinding {I J : Type*} [Fintype I] [Fintype J] [DecidableEq I]
    (Phi : Superop I J) : Prop :=
  IsChannel Phi ∧ (cubittOutputTranspose (cubittNormalizedChoi Phi)).PosSemidef

/-- PDF p.6, S15--S16: switch 0 applies E_kappa after Gamma; switch 1 applies E_p.
Both output flags, the switch, and the full main output are retained. -/
def cubittNoisySwitch {I : Type*} [Fintype I] [DecidableEq I]
    (Gamma : Superop I I) (kappa p : ℝ) : Superop (Bool × I) (Bool × (Bool × I)) :=
  cubittSwitched fun s => if s then cubittWholeErasure p else cubittWholeErasure kappa ∘ Gamma

/-- PDF pp.3,8,11: the lower erasure threshold, with a REAL exponent -1/n. -/
def cubittErasureThreshold (n : ℕ) (kappa : ℝ) : ℝ :=
  (1 + kappa ^ n) ^ (-(1 : ℝ) / (n : ℝ))

/-- PDF p.6: paper-specific parameters; admissibility is stated separately. -/
structure CubittParameters where
  N : ℕ
  r : ℕ
  m : ℕ
  d : ℕ
  q : ℝ

/-- PDF pp.5--7: positive numbers of shares/copies, d>=2, and physical S18 weights.
The results choose q=1/3; q<=1/2 is the positivity range of that construction. -/
def CubittValid (P : CubittParameters) : Prop :=
  0 < P.N ∧ 0 < P.r ∧ 0 < P.m ∧ 2 ≤ P.d ∧ 0 < P.q ∧ P.q ≤ 1 / 2

/-- PDF p.6: A_i consists of r*m d-dimensional factors. -/
abbrev CubittShare (r m d : ℕ) := Fin r → Fin m → Fin d

/-- PDF p.6: A consists of N shares; finite functions preserve all named factors. -/
abbrev CubittShield (P : CubittParameters) := Fin P.N → CubittShare P.r P.m P.d

/-- PDF p.6: main input aA and equal-dimensional output bB, key qubit first. -/
abbrev CubittMain (P : CubittParameters) := Bool × CubittShield P

/-- PDF p.7: normalized projector onto the symmetric subspace, (I+swap)/(d(d+1)). -/
def cubittSymmetricState (d : ℕ) : Op (Fin d × Fin d) :=
  fun x y => ((if x = y then 1 else 0) +
    (if x.1 = y.2 ∧ x.2 = y.1 then 1 else 0)) / ((d : ℂ) * ((d : ℂ) + 1))

/-- PDF p.7: normalized antisymmetric projector, (I-swap)/(d(d-1)); used only at d>=2. -/
def cubittAntisymmetricState (d : ℕ) : Op (Fin d × Fin d) :=
  fun x y => ((if x = y then 1 else 0) -
    (if x.1 = y.2 ∧ x.2 = y.1 then 1 else 0)) / ((d : ℂ) * ((d : ℂ) - 1))

/-- PDF p.7: omega on C_k, tensoring the equal mixture over all N*r pairs. -/
def cubittHidingOmega (P : CubittParameters) (k : Fin P.m)
    (x y : CubittShield P × CubittShield P) : ℂ :=
  ∏ i, ∏ j, (cubittSymmetricState P.d (x.1 i j k, x.2 i j k) (y.1 i j k, y.2 i j k) +
    cubittAntisymmetricState P.d (x.1 i j k, x.2 i j k) (y.1 i j k, y.2 i j k)) / 2

/-- PDF p.7: sigma on C_k, tensoring the symmetric state over all N*r pairs. -/
def cubittHidingSigma (P : CubittParameters) (k : Fin P.m)
    (x y : CubittShield P × CubittShield P) : ℂ :=
  ∏ i, ∏ j, cubittSymmetricState P.d (x.1 i j k, x.2 i j k) (y.1 i j k, y.2 i j k)

/-- PDF p.6, S18: the unnormalized three key blocks, reordered across aA:bB.
Products over k retain q/2 in EACH copy, including the off-diagonal block's sign. -/
def cubittRawChoi (P : CubittParameters) : Op (CubittMain P × CubittMain P) :=
  fun x y =>
    let u := (x.1.2, x.2.2)
    let v := (y.1.2, y.2.2)
    if x.1.1 = x.2.1 ∧ y.1.1 = y.2.1 then
      if x.1.1 = y.1.1 then
        ∏ k, (P.q : ℂ) / 2 * (cubittHidingOmega P k u v + cubittHidingSigma P k u v)
      else
        ∏ k, (P.q : ℂ) / 2 * (cubittHidingOmega P k u v - cubittHidingSigma P k u v)
    else if x.1.1 ≠ x.2.1 ∧ x.1.1 = y.1.1 ∧ x.2.1 = y.2.1 then
      ∏ k, ((1 / 2 : ℂ) - (P.q : ℂ)) * cubittHidingSigma P k u v
    else 0

/-- PDF pp.6--7, S18: COMPUTED normalized Choi state.
For admissible parameters its trace normalizer is 2*(q^m+(1/2-q)^m)>0. -/
def cubittZeta (P : CubittParameters) : Op (CubittMain P × CubittMain P) :=
  (2 * ((P.q : ℂ) ^ P.m + ((1 / 2 : ℂ) - (P.q : ℂ)) ^ P.m))⁻¹ • cubittRawChoi P

/-- PDF pp.6--7: inverse normalized Choi transform; Gamma is computed from S18.
Its input-reference marginal is I/dim(CubittMain), so this is CPTP on admissible parameters. -/
def cubittGammaChannel (P : CubittParameters) : Superop (CubittMain P) (CubittMain P) :=
  fun X b c => (Fintype.card (CubittMain P) : ℂ) *
    ∑ a, ∑ a2, X a a2 * cubittZeta P (a, b) (a2, c)

/-- PDF p.6, S15: the fully specified paper channel, without any capacity assumption. -/
def cubittPaperChannel (P : CubittParameters) (kappa p : ℝ) := cubittNoisySwitch (cubittGammaChannel P) kappa p

/-- PDF p.7, S19: sufficient PPT condition; the exponent is the product r*N. -/
def CubittPPTCondition (P : CubittParameters) : Prop :=
  0 < P.q ∧ P.q ≤ 1 / 3 ∧
    ((P.d : ℝ) / ((P.d : ℝ) - 1)) ^ (P.r * P.N) ≤ (1 - P.q) / P.q

/-- PDF pp.7--8: integer ceiling retained from Lemma 2 (the p.10 shorthand omits it). -/
def cubittRepetitions (m : ℕ) : ℕ := 2 * m + Nat.ceil (Real.logb 2 (m : ℝ))

/-- PDF p.7, Lemma 2: q=1/3 and the prescribed integer repetition count. -/
def cubittApproximationParameters (N m d : ℕ) : CubittParameters :=
  ⟨N, cubittRepetitions m, m, d, 1 / 3⟩

/-- PDF p.7, S20: reinsert the first shield share, indexed by 0 in Lean. -/
def cubittInsertFirst (P : CubittParameters) (hN : 0 < P.N) (x : CubittShare P.r P.m P.d)
    (rest : {i : Fin P.N // i ≠ ⟨0, hN⟩} → CubittShare P.r P.m P.d) : CubittShield P :=
  fun i => if h : i = ⟨0, hN⟩ then x else rest ⟨i, h⟩

/-- PDF p.7, S20: actual partial trace over A_2 B_2 ... A_N B_N of the computed cubittZeta.
The rho/cubittZeta change in Lemma 2 is resolved by this marginal, not by an arbitrary state. -/
def cubittFirstShareState (P : CubittParameters) (hN : 0 < P.N) :
    Op ((Bool × CubittShare P.r P.m P.d) × (Bool × CubittShare P.r P.m P.d)) :=
  fun x y => ∑ aRest, ∑ bRest,
    cubittZeta P ((x.1.1, cubittInsertFirst P hN x.1.2 aRest),
      (x.2.1, cubittInsertFirst P hN x.2.2 bRest))
      ((y.1.1, cubittInsertFirst P hN y.1.2 aRest),
      (y.2.1, cubittInsertFirst P hN y.2.2 bRest))

/-- PDF p.6, S12: twisted Bell pair with arbitrary density sigma and key-controlled unitaries.
Coordinates are aA:bB; this is S10--S11 expanded in the fixed basis. -/
def cubittPrivateState {I : Type*} [Fintype I] [DecidableEq I] (sigma : Op (I × I))
    (U : Bool → Bool → Matrix.unitaryGroup (I × I) ℂ) :
    Op ((Bool × I) × (Bool × I)) :=
  fun x y => if x.1.1 = x.2.1 ∧ y.1.1 = y.2.1 then
    (1 / 2 : ℂ) * (((U x.1.1 x.1.1 : Op (I × I)) * sigma *
      (U y.1.1 y.1.1 : Op (I × I))ᴴ) (x.1.2, x.2.2) (y.1.2, y.2.2))
    else 0

/-- PDF p.6, S10--S12: normalized perfect private bit; shield dimension is unrestricted. -/
def CubittPerfectPbit {I : Type*} [Fintype I] [DecidableEq I]
    (rho : Op ((Bool × I) × (Bool × I))) : Prop :=
  IsDensityMatrix rho ∧ ∃ sigma : Op (I × I), IsDensityMatrix sigma ∧
    ∃ U : Bool → Bool → Matrix.unitaryGroup (I × I) ℂ, rho = cubittPrivateState sigma U

/-- PDF pp.7--8, Lemma 2: full trace-norm approximation, not half trace distance. -/
def cubittApproximationBound (m : ℕ) : ℝ :=
  16 * Real.sqrt (m : ℝ) * (2 : ℝ) ^ (-(m : ℝ) / 4)

/-- PDF p.10, S52: continuity loss; applied only with 0<=tau<1. -/
def cubittContinuityLoss (tau : ℝ) : ℝ := 4 * tau + 2 * binEntropy2 tau

/-- PDF p.11: final concrete N, which grows with the requested undetectable block length. -/
def cubittExplicitShares (n : ℕ) : ℕ := 2 * n * 4 ^ n

/-- PDF pp.10--11: d=2*N*r, a product (not a power). -/
def cubittExplicitParameters (n m : ℕ) : CubittParameters :=
  cubittApproximationParameters (cubittExplicitShares n) m (2 * cubittExplicitShares n * cubittRepetitions m)

/-! Small constructor sanity examples. No result theorem is proved in this stage. -/

example {I : Type*} [Fintype I] [DecidableEq I] (X : Op I) (a b : I) :
    cubittWholeErasure 0 X (false, a) (false, b) = X a b := by
  simp [cubittWholeErasure]

example {I : Type*} [Fintype I] [DecidableEq I] (X : Op I) (a b : I) :
    cubittWholeErasure 1 X (true, a) (true, b) = X.trace * mixed I a b := by
  simp [cubittWholeErasure]

example {A B : Type*} (branches : Bool → Superop A B) (X : Op (Bool × A)) (a b : B) :
    cubittSwitched branches X (false, a) (true, b) = 0 := by
  simp [cubittSwitched]

example {I J : Type*} (X : Op (I × J)) : cubittOutputTranspose (cubittOutputTranspose X) = X := rfl

example : cubittSymmetricState 2 (0, 0) (0, 0) = 1 / 3 := by
  norm_num [cubittSymmetricState]

example : cubittAntisymmetricState 2 (0, 1) (1, 0) = -(1 / 2) := by
  norm_num [cubittAntisymmetricState]

example (P : CubittParameters) (hN : 0 < P.N) (x : CubittShare P.r P.m P.d)
    (rest : {i : Fin P.N // i ≠ ⟨0, hN⟩} → CubittShare P.r P.m P.d) :
    cubittInsertFirst P hN x rest ⟨0, hN⟩ = x := by
  simp [cubittInsertFirst]

example : cubittExplicitShares 1 = 8 := rfl

example (m : ℕ) : (cubittExplicitParameters 1 m).q = 1 / 3 := rfl

end

end QAlgorithms.Papers.Independent.Intakenext100.P70
Source
Unbounded number of channel uses are required to see quantum capacity, arXiv:1408.5115v1, https://arxiv.org/abs/1408.5115v1
Read-back

What the Lean code literally says, in plain math · gpt-6.1-sol

QAlgorithms.Papers.Independent.Intakenext100.P70.CubittMain

For any parameter tuple P=(N,r,m,d,q)P=(N,r,m,d,q)P=(N,r,m,d,q), with N,r,m,d∈NN,r,m,d\in\mathbb NN,r,m,d∈N and q∈Rq\in\mathbb Rq∈R, the type CubittMain(P)\mathrm{CubittMain}(P)CubittMain(P) consists of pairs (b,s)(b,s)(b,s), where b∈{false,true}b\in\{\mathrm{false},\mathrm{true}\}b∈{false,true} and sss is an array with entries sijk∈{0,…,d−1}s_{ijk}\in\{0,\ldots,d-1\}sijk​∈{0,…,d−1}, indexed by 0≤i<N0\le i<N0≤i<N, 0≤j<r0\le j<r0≤j<r, and 0≤k<m0\le k<m0≤k<m. No restrictions on the parameters are imposed. If any array index range is empty, there are exactly two such pairs; if all three ranges are nonempty and d=0d=0d=0, there are none.

QAlgorithms.Papers.Independent.Intakenext100.P70.CubittPPTBinding

For finite types I,JI,JI,J, with decidable equality on III, and any function Φ:Mat⁡I(C)→Mat⁡J(C)\Phi:\operatorname{Mat}_I(\mathbb C)\to\operatorname{Mat}_J(\mathbb C)Φ:MatI​(C)→MatJ​(C), this predicate requires both that Φ\PhiΦ have a finite Kraus representation and that a specified matrix be positive semidefinite. Explicitly, there must exist a natural number rrr and matrices Eℓ∈Mat⁡J×I(C)E_\ell\in\operatorname{Mat}_{J\times I}(\mathbb C)Eℓ​∈MatJ×I​(C), for 0≤ℓ<r0\le\ell<r0≤ℓ<r, such that

∑ℓ=0r−1Eℓ†Eℓ=1,Φ(X)=∑ℓ=0r−1EℓXEℓ†for every complex matrix X,\sum_{\ell=0}^{r-1}E_\ell^\dagger E_\ell=1, \qquad \Phi(X)=\sum_{\ell=0}^{r-1}E_\ell X E_\ell^\dagger \quad\text{for every complex matrix }X,ℓ=0∑r−1​Eℓ†​Eℓ​=1,Φ(X)=ℓ=0∑r−1​Eℓ​XEℓ†​for every complex matrix X,

and the matrix

T(a,b),(a′,b′)=Φ(Ea,a′)b′,b∣I∣T_{(a,b),(a',b')}=\frac{\Phi(E_{a,a'})_{b',b}}{|I|}T(a,b),(a′,b′)​=∣I∣Φ(Ea,a′​)b′,b​​

must be positive semidefinite, where Ea,a′E_{a,a'}Ea,a′​ is the matrix unit on III. The types may be empty, and r=0r=0r=0 is allowed; the normalization equation still applies, and division by zero uses the total complex-field convention.

QAlgorithms.Papers.Independent.Intakenext100.P70.CubittPPTCondition

For any parameter tuple P=(N,r,m,d,q)P=(N,r,m,d,q)P=(N,r,m,d,q), with N,r,m,d∈NN,r,m,d\in\mathbb NN,r,m,d∈N and q∈Rq\in\mathbb Rq∈R, this predicate means precisely

0<q,q≤13,(dd−1)rN≤1−qq,0<q,\qquad q\le\frac13,\qquad \left(\frac{d}{d-1}\right)^{rN}\le\frac{1-q}{q},0<q,q≤31​,(d−1d​)rN≤q1−q​,

where the fractions and subtraction are real arithmetic and the exponent rNrNrN is a natural number. It imposes no additional restrictions on N,r,m,dN,r,m,dN,r,m,d, and mmm does not occur in the conditions. In particular, d=0d=0d=0 and d=1d=1d=1 are included, with real division by zero defined as zero; if rN=0rN=0rN=0, the left-hand side is 111, including when its base is zero.

QAlgorithms.Papers.Independent.Intakenext100.P70.CubittParameters

This structure consists of five fields: natural numbers N,r,m,dN,r,m,dN,r,m,d and a real number qqq. It imposes no further conditions: any of the natural-number fields may be zero, and qqq may be any real number.

QAlgorithms.Papers.Independent.Intakenext100.P70.CubittPerfectPbit

For any finite type III with decidable equality and any complex matrix ρ\rhoρ on (Bool×I)2(\mathrm{Bool}\times I)^2(Bool×I)2, this predicate means that ρ\rhoρ is positive semidefinite with trace 111, and there exist a positive semidefinite, trace-one matrix σ\sigmaσ on I×II\times II×I and a Boolean-indexed family of unitaries Uα,βU_{\alpha,\beta}Uα,β​ on I×II\times II×I such that

ρ((α,a),(β,b)),((γ,c),(δ,d))={12(Uα,ασUγ,γ†)(a,b),(c,d),α=β and γ=δ,0,otherwise.\rho_{((\alpha,a),(\beta,b)),((\gamma,c),(\delta,d))} = \begin{cases} \frac12\bigl(U_{\alpha,\alpha}\sigma U_{\gamma,\gamma}^{\dagger}\bigr)_{(a,b),(c,d)}, &\alpha=\beta\ \text{and}\ \gamma=\delta,\\ 0,&\text{otherwise}. \end{cases}ρ((α,a),(β,b)),((γ,c),(δ,d))​={21​(Uα,α​σUγ,γ†​)(a,b),(c,d)​,0,​α=β and γ=δ,otherwise.​

The off-diagonal members of the unitary family are existentially supplied but do not affect this equality. Empty III is included in the definition, but then the trace-one conditions cannot hold.

QAlgorithms.Papers.Independent.Intakenext100.P70.CubittShare

For arbitrary natural numbers r,m,dr,m,dr,m,d, this abbreviates the type of functions assigning to every j∈{0,…,r−1}j\in\{0,\ldots,r-1\}j∈{0,…,r−1} a function from {0,…,m−1}\{0,\ldots,m-1\}{0,…,m−1} to {0,…,d−1}\{0,\ldots,d-1\}{0,…,d−1}, equivalently an r×mr\times mr×m array with entries in a ddd-element set. If r=0r=0r=0 or m=0m=0m=0, this type has exactly one element, including when d=0d=0d=0; if r,m>0r,m>0r,m>0 and d=0d=0d=0, it has no elements.

QAlgorithms.Papers.Independent.Intakenext100.P70.CubittShield

For any parameter tuple P=(N,r,m,d,q)P=(N,r,m,d,q)P=(N,r,m,d,q), where N,r,m,dN,r,m,dN,r,m,d are arbitrary natural numbers and q∈Rq\in\mathbb Rq∈R, the type CubittShield(P)\mathrm{CubittShield}(P)CubittShield(P) consists of arrays

s:{0,…,N−1}×{0,…,r−1}×{0,…,m−1}⟶{0,…,d−1}.s:\{0,\ldots,N-1\}\times\{0,\ldots,r-1\}\times\{0,\ldots,m-1\}\longrightarrow\{0,\ldots,d-1\}.s:{0,…,N−1}×{0,…,r−1}×{0,…,m−1}⟶{0,…,d−1}.

There are no positivity or validity hypotheses on PPP. If any of N,r,mN,r,mN,r,m is zero, this type contains exactly one empty array; if all three are positive and d=0d=0d=0, it is empty.

QAlgorithms.Papers.Independent.Intakenext100.P70.CubittValid

For a record PPP consisting of natural numbers N,r,m,dN,r,m,dN,r,m,d and a real number qqq, this predicate holds exactly when

N>0,r>0,m>0,d≥2,0<q≤12.N>0,\qquad r>0,\qquad m>0,\qquad d\ge2,\qquad 0<q\le\frac12.N>0,r>0,m>0,d≥2,0<q≤21​.

All six inequalities are required simultaneously; there are no additional conditions.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittAntisymmetricState

For every natural number ddd, this defines a complex square matrix AdA_dAd​ indexed by pairs in {0,…,d−1}2\{0,\ldots,d-1\}^2{0,…,d−1}2. For every row x=(x1,x2)x=(x_1,x_2)x=(x1​,x2​) and column y=(y1,y2)y=(y_1,y_2)y=(y1​,y2​), its entry is

(Ad)x,y=1{x=y}−1{x1=y2,  x2=y1}d(d−1),(A_d)_{x,y} = \frac{\mathbf 1_{\{x=y\}}-\mathbf 1_{\{x_1=y_2,\;x_2=y_1\}}}{d(d-1)},(Ad​)x,y​=d(d−1)1{x=y}​−1{x1​=y2​,x2​=y1​}​​,

with all arithmetic in C\mathbb CC, so d−1d-1d−1 here is complex subtraction. No hypothesis d≥2d\ge2d≥2 is imposed. For d=0d=0d=0 the index set is empty; for d=1d=1d=1 the unique entry is zero, using the convention that division by zero equals zero.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittApproximationBound

For every natural number mmm, including m=0m=0m=0, this defines the real number

16m 2−m/4,16\sqrt m\,2^{-m/4},16m​2−m/4,

where mmm is converted to a real number before taking the square root and forming the real exponent. Its value at m=0m=0m=0 is 000. The declaration defines this number without asserting any approximation inequality.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittApproximationParameters

For arbitrary natural numbers N,m,dN,m,dN,m,d, this defines a parameter record with fields

(N,r,m,d,q)=(N,  2m+⌈log⁡2m⌉N,  m,  d,  13),(N,r,m,d,q)=\left(N,\;2m+\left\lceil\log_2 m\right\rceil_{\mathbb N},\;m,\;d,\;\frac13\right),(N,r,m,d,q)=(N,2m+⌈log2​m⌉N​,m,d,31​),

where ⌈x⌉N\lceil x\rceil_{\mathbb N}⌈x⌉N​ denotes the least natural number at least xxx, with value 000 when x≤0x\le0x≤0, and the total real logarithm has log⁡20=0\log_2 0=0log2​0=0. Thus m=0m=0m=0 gives r=0r=0r=0. No positivity or validity conditions on N,m,dN,m,dN,m,d are required or asserted.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittBlockInformation

For any finite types I,JI,JI,J with decidable equality, any function Φ:Mat⁡I(C)→Mat⁡J(C)\Phi:\operatorname{Mat}_I(\mathbb C)\to\operatorname{Mat}_J(\mathbb C)Φ:MatI​(C)→MatJ​(C), and any natural number nnn, this defines the real number S(Φ[n])/nS(\Phi^{[n]})/nS(Φ[n])/n. Here Φ[n]\Phi^{[n]}Φ[n] is the function on matrices indexed by InI^nIn whose entries, indexed by y,y′∈Jny,y'\in J^ny,y′∈Jn, are

Φ[n](X)y,y′=∑x,x′∈InXx,x′∏i=0n−1Φ(Exi,xi′)yi,yi′,\Phi^{[n]}(X)_{y,y'}=\sum_{x,x'\in I^n}X_{x,x'}\prod_{i=0}^{n-1}\Phi(E_{x_i,x'_i})_{y_i,y'_i},Φ[n](X)y,y′​=x,x′∈In∑​Xx,x′​i=0∏n−1​Φ(Exi​,xi′​​)yi​,yi′​​,

where Ea,bE_{a,b}Ea,b​ has entry 111 at (a,b)(a,b)(a,b) and zero elsewhere. For a function Ψ:Mat⁡K(C)→Mat⁡L(C)\Psi:\operatorname{Mat}_K(\mathbb C)\to\operatorname{Mat}_L(\mathbb C)Ψ:MatK​(C)→MatL​(C), S(Ψ)S(\Psi)S(Ψ) is the real supremum, over all positive semidefinite ρ\rhoρ of trace 111, of

H(Ψ(ρ))−H ⁣((id⁡K⊗Ψ)(∣vρ⟩⟨vρ∣)).H(\Psi(\rho))-H\!\left((\operatorname{id}_K\otimes\Psi)(|v_\rho\rangle\langle v_\rho|)\right).H(Ψ(ρ))−H((idK​⊗Ψ)(∣vρ​⟩⟨vρ​∣)).

The vector vρv_\rhovρ​ has coordinates (vρ)a,b=(cfc⁡(⋅,ρ))b,a(v_\rho)_{a,b}=(\operatorname{cfc}(\sqrt{\cdot},\rho))_{b,a}(vρ​)a,b​=(cfc(⋅​,ρ))b,a​, the blockwise extension is ((id⁡K⊗Ψ)(Z))(a,b),(a′,b′)=Ψ((Z(a,u),(a′,u′))u,u′)b,b′((\operatorname{id}_K\otimes\Psi)(Z))_{(a,b),(a',b')}=\Psi((Z_{(a,u),(a',u')})_{u,u'})_{b,b'}((idK​⊗Ψ)(Z))(a,b),(a′,b′)​=Ψ((Z(a,u),(a′,u′)​)u,u′​)b,b′​, and H(A)=−Re⁡tr⁡(cfc⁡(t↦tlog⁡2t,A))H(A)=-\operatorname{Re}\operatorname{tr}(\operatorname{cfc}(t\mapsto t\log_2t,A))H(A)=−Retr(cfc(t↦tlog2​t,A)), using the total continuous functional calculus. No linearity or channel assumption is imposed on Φ\PhiΦ. Empty sums and products retain their usual values; the real supremum of an empty set is 000, and n=0n=0n=0 is allowed, in which case the final division by zero gives 000.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittContinuityLoss

For every real number τ\tauτ, this defines

4τ+2ln⁡2(−τlog⁡τ−(1−τ)log⁡(1−τ)).4\tau+\frac{2}{\ln2}\bigl(-\tau\log\tau-(1-\tau)\log(1-\tau)\bigr).4τ+ln22​(−τlogτ−(1−τ)log(1−τ)).

Here log⁡\loglog is the total real logarithm, with log⁡0=0\log0=0log0=0 and log⁡x=ln⁡∣x∣\log x=\ln|x|logx=ln∣x∣ for x≠0x\ne0x=0. Thus the definition applies to all real τ\tauτ, without a hypothesis 0≤τ≤10\le\tau\le10≤τ≤1, and gives values 000 at τ=0\tau=0τ=0 and 444 at τ=1\tau=1τ=1. It does not itself assert a continuity inequality.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittErasureThreshold

For every natural number nnn and every real number κ\kappaκ, this defines

(1+κn)−1/n,(1+\kappa^n)^{-1/n},(1+κn)−1/n,

where the inner power has natural exponent nnn and the outer power is the total real-power operation. No positivity assumption on nnn, κ\kappaκ, or the base is imposed. In particular, when n=0n=0n=0, the inner power is 111, the exponent is 000 because real division by zero gives zero, and the value is 111; a zero base with a negative exponent also receives the total-operation value 000.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittExplicitParameters

For arbitrary natural numbers n,mn,mn,m, put s=2n4ns=2n4^ns=2n4n and r=2m+⌈log⁡2m⌉Nr=2m+\lceil\log_2m\rceil_{\mathbb N}r=2m+⌈log2​m⌉N​, where the natural-number ceiling is truncated below at 000 and log⁡20=0\log_2 0=0log2​0=0. This defines the parameter record

(N,r,m,d,q)=(s, r, m, 2sr, 13).(N,r,m,d,q)=\left(s,\ r,\ m,\ 2sr,\ \frac13\right).(N,r,m,d,q)=(s, r, m, 2sr, 31​).

No validity conditions are required or asserted. In particular, n=0n=0n=0 gives N=d=0N=d=0N=d=0, and m=0m=0m=0 gives r=d=0r=d=0r=d=0.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittExplicitShares

For every natural number nnn, this defines the natural number

2n 4n.2n\,4^n.2n4n.

The input n=0n=0n=0 is included and gives 000; no positivity hypothesis is required.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittFirstShareState

Let P=(N,r,m,d,q)P=(N,r,m,d,q)P=(N,r,m,d,q), where N,r,m,dN,r,m,dN,r,m,d are natural numbers and qqq is real, and assume N>0N>0N>0. Put H=(Fin r→Fin m→Fin d)H=(\mathrm{Fin}\,r\to\mathrm{Fin}\,m\to\mathrm{Fin}\,d)H=(Finr→Finm→Find), S=(Fin N→H)S=(\mathrm{Fin}\,N\to H)S=(FinN→H), and E(s,a)0=sE(s,a)_0=sE(s,a)0​=s, E(s,a)i=aiE(s,a)_i=a_iE(s,a)i​=ai​ for i≠0i\ne0i=0, where aaa assigns shares to the nonzero indices. This defines a complex matrix FFF indexed by (Bool×H)2(\mathrm{Bool}\times H)^2(Bool×H)2, whose entries, for x=((α,s),(β,t))x=((\alpha,s),(\beta,t))x=((α,s),(β,t)) and y=((α′,s′),(β′,t′))y=((\alpha',s'),(\beta',t'))y=((α′,s′),(β′,t′)), are

Fx,y=∑a,bζ((α,E(s,a)),(β,E(t,b))),((α′,E(s′,a)),(β′,E(t′,b))).F_{x,y}=\sum_{a,b} \zeta_{((\alpha,E(s,a)),(\beta,E(t,b))), ((\alpha',E(s',a)),(\beta',E(t',b)))}.Fx,y​=a,b∑​ζ((α,E(s,a)),(β,E(t,b))),((α′,E(s′,a)),(β′,E(t′,b)))​.

Here both sums range independently over all assignments of shares to the indices i≠0i\ne0i=0, and ζ\zetaζ is the following explicitly defined matrix on (Bool×S)2(\mathrm{Bool}\times S)^2(Bool×S)2. For pairs z,w∈(Fin d)2z,w\in(\mathrm{Fin}\,d)^2z,w∈(Find)2, set

Dz,w±=1z=w±1z1=w2, z2=w1d(d±1).D^\pm_{z,w}= \frac{\mathbf1_{z=w}\pm\mathbf1_{z_1=w_2,\ z_2=w_1}}{d(d\pm1)}.Dz,w±​=d(d±1)1z=w​±1z1​=w2​, z2​=w1​​​.

For u=(u1,u2),v=(v1,v2)∈S2u=(u_1,u_2),v=(v_1,v_2)\in S^2u=(u1​,u2​),v=(v1​,v2​)∈S2, define

Ωk(u,v)=∏i<N∏j<rD(u1(i,j,k),u2(i,j,k)),(v1(i,j,k),v2(i,j,k))++D(u1(i,j,k),u2(i,j,k)),(v1(i,j,k),v2(i,j,k))−2,Σk(u,v)=∏i<N∏j<rD(u1(i,j,k),u2(i,j,k)),(v1(i,j,k),v2(i,j,k))+.\Omega_k(u,v)=\prod_{i<N}\prod_{j<r} \frac{D^+_{(u_1(i,j,k),u_2(i,j,k)),(v_1(i,j,k),v_2(i,j,k))} +D^-_{(u_1(i,j,k),u_2(i,j,k)),(v_1(i,j,k),v_2(i,j,k))}}2, \quad \Sigma_k(u,v)=\prod_{i<N}\prod_{j<r} D^+_{(u_1(i,j,k),u_2(i,j,k)),(v_1(i,j,k),v_2(i,j,k))}.Ωk​(u,v)=i<N∏​j<r∏​2D(u1​(i,j,k),u2​(i,j,k)),(v1​(i,j,k),v2​(i,j,k))+​+D(u1​(i,j,k),u2​(i,j,k)),(v1​(i,j,k),v2​(i,j,k))−​​,Σk​(u,v)=i<N∏​j<r∏​D(u1​(i,j,k),u2​(i,j,k)),(v1​(i,j,k),v2​(i,j,k))+​.

For row ((α,u1),(β,u2))((\alpha,u_1),(\beta,u_2))((α,u1​),(β,u2​)) and column ((α′,v1),(β′,v2))((\alpha',v_1),(\beta',v_2))((α′,v1​),(β′,v2​)), the corresponding entry of ζ\zetaζ is 1/[2(qm+(12−q)m)]1/[2(q^m+(\frac12-q)^m)]1/[2(qm+(21​−q)m)] times

{∏k<mq2(Ωk+Σk),α=β, α′=β′, α=α′,∏k<mq2(Ωk−Σk),α=β, α′=β′, α≠α′,∏k<m(12−q)Σk,α≠β, α=α′, β=β′,0,otherwise,\begin{cases} \displaystyle\prod_{k<m}\frac q2(\Omega_k+\Sigma_k), &\alpha=\beta,\ \alpha'=\beta',\ \alpha=\alpha',\\ \displaystyle\prod_{k<m}\frac q2(\Omega_k-\Sigma_k), &\alpha=\beta,\ \alpha'=\beta',\ \alpha\ne\alpha',\\ \displaystyle\prod_{k<m}\left(\frac12-q\right)\Sigma_k, &\alpha\ne\beta,\ \alpha=\alpha',\ \beta=\beta',\\ 0,&\text{otherwise}, \end{cases}⎩⎨⎧​k<m∏​2q​(Ωk​+Σk​),k<m∏​2q​(Ωk​−Σk​),k<m∏​(21​−q)Σk​,0,​α=β, α′=β′, α=α′,α=β, α′=β′, α=α′,α=β, α=α′, β=β′,otherwise,​

with Ωk,Σk\Omega_k,\Sigma_kΩk​,Σk​ evaluated at (u,v)(u,v)(u,v). All scalars in these matrix formulas are interpreted in C\mathbb CC. No validity, positivity, or trace-one hypothesis is imposed: r,m,d,qr,m,d,qr,m,d,q are unrestricted, empty sums and products have values 000 and 111, and division by zero has value 000.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittGammaChannel

For arbitrary parameters P=(N,r,m,d,q)P=(N,r,m,d,q)P=(N,r,m,d,q), with N,r,m,d∈NN,r,m,d\in\mathbb NN,r,m,d∈N and q∈Rq\in\mathbb Rq∈R, let I={false,true}×{0,…,d−1}N×r×mI=\{\mathrm{false},\mathrm{true}\}\times\{0,\ldots,d-1\}^{N\times r\times m}I={false,true}×{0,…,d−1}N×r×m and D=∣I∣=2dNrmD=|I|=2d^{Nrm}D=∣I∣=2dNrm. This defines a function ΓP\Gamma_PΓP​ from all complex I×II\times II×I matrices to complex I×II\times II×I matrices by

(ΓP(X))b,c=D2(qm+(12−q)m)∑a∈I∑a′∈IXa,a′ H((a,b),(a′,c)).(\Gamma_P(X))_{b,c} = \frac{D}{2\bigl(q^m+(\tfrac12-q)^m\bigr)} \sum_{a\in I}\sum_{a'\in I}X_{a,a'}\,H((a,b),(a',c)).(ΓP​(X))b,c​=2(qm+(21​−q)m)D​a∈I∑​a′∈I∑​Xa,a′​H((a,b),(a′,c)).

Here HHH is specified entrywise as follows: for x=((α,s),(β,t))x=((\alpha,s),(\beta,t))x=((α,s),(β,t)) and y=((γ,s′),(η,t′))y=((\gamma,s'),(\eta,t'))y=((γ,s′),(η,t′)), put

δijk=1{sijk=sijk′,  tijk=tijk′},εijk=1{sijk=tijk′,  tijk=sijk′},\delta_{ijk}=\mathbf 1_{\{s_{ijk}=s'_{ijk},\;t_{ijk}=t'_{ijk}\}}, \quad \varepsilon_{ijk}=\mathbf 1_{\{s_{ijk}=t'_{ijk},\;t_{ijk}=s'_{ijk}\}},δijk​=1{sijk​=sijk′​,tijk​=tijk′​}​,εijk​=1{sijk​=tijk′​,tijk​=sijk′​}​, Ok=∏i<N∏j<r12(δijk+εijkd(d+1)+δijk−εijkd(d−1)),Sk=∏i<N∏j<rδijk+εijkd(d+1),O_k=\prod_{i<N}\prod_{j<r}\frac12 \left(\frac{\delta_{ijk}+\varepsilon_{ijk}}{d(d+1)} +\frac{\delta_{ijk}-\varepsilon_{ijk}}{d(d-1)}\right), \qquad S_k=\prod_{i<N}\prod_{j<r}\frac{\delta_{ijk}+\varepsilon_{ijk}}{d(d+1)},Ok​=i<N∏​j<r∏​21​(d(d+1)δijk​+εijk​​+d(d−1)δijk​−εijk​​),Sk​=i<N∏​j<r∏​d(d+1)δijk​+εijk​​,

and

H(x,y)={∏k<mq2(Ok+Sk),α=β, γ=η, α=γ,∏k<mq2(Ok−Sk),α=β, γ=η, α≠γ,∏k<m(12−q)Sk,α≠β, α=γ, β=η,0,otherwise.H(x,y)= \begin{cases} \displaystyle\prod_{k<m}\frac q2(O_k+S_k), &\alpha=\beta,\ \gamma=\eta,\ \alpha=\gamma,\\[3pt] \displaystyle\prod_{k<m}\frac q2(O_k-S_k), &\alpha=\beta,\ \gamma=\eta,\ \alpha\ne\gamma,\\[3pt] \displaystyle\prod_{k<m}(\tfrac12-q)S_k, &\alpha\ne\beta,\ \alpha=\gamma,\ \beta=\eta,\\[3pt] 0,&\text{otherwise}. \end{cases}H(x,y)=⎩⎨⎧​k<m∏​2q​(Ok​+Sk​),k<m∏​2q​(Ok​−Sk​),k<m∏​(21​−q)Sk​,0,​α=β, γ=η, α=γ,α=β, γ=η, α=γ,α=β, α=γ, β=η,otherwise.​

The formula applies to every XXX, with no density-matrix or parameter hypotheses. Entry arithmetic is complex, division by zero is zero, empty sums are zero, and empty products are 111; in particular, if III is empty, this is the unique function between the corresponding singleton matrix spaces.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittHidingOmega

For any parameters P=(N,r,m,d,q)P=(N,r,m,d,q)P=(N,r,m,d,q), with N,r,m,d∈NN,r,m,d\in\mathbb NN,r,m,d∈N and q∈Rq\in\mathbb Rq∈R, any k∈{0,…,m−1}k\in\{0,\ldots,m-1\}k∈{0,…,m−1}, and any two pairs of shield arrays x=(x(1),x(2))x=(x^{(1)},x^{(2)})x=(x(1),x(2)) and y=(y(1),y(2))y=(y^{(1)},y^{(2)})y=(y(1),y(2)), whose entries lie in {0,…,d−1}\{0,\ldots,d-1\}{0,…,d−1} and have indices i<N,j<r,ℓ<mi<N,j<r,\ell<mi<N,j<r,ℓ<m, this defines the complex number

ΩP,k(x,y)=∏i=0N−1∏j=0r−112(δij+εijd(d+1)+δij−εijd(d−1)),\Omega_{P,k}(x,y) = \prod_{i=0}^{N-1}\prod_{j=0}^{r-1} \frac12\left( \frac{\delta_{ij}+\varepsilon_{ij}}{d(d+1)} + \frac{\delta_{ij}-\varepsilon_{ij}}{d(d-1)} \right),ΩP,k​(x,y)=i=0∏N−1​j=0∏r−1​21​(d(d+1)δij​+εij​​+d(d−1)δij​−εij​​),

where

δij=1{xijk(1)=yijk(1),  xijk(2)=yijk(2)},εij=1{xijk(1)=yijk(2),  xijk(2)=yijk(1)}.\delta_{ij}=\mathbf 1_{\{x^{(1)}_{ijk}=y^{(1)}_{ijk},\;x^{(2)}_{ijk}=y^{(2)}_{ijk}\}}, \qquad \varepsilon_{ij}=\mathbf 1_{\{x^{(1)}_{ijk}=y^{(2)}_{ijk},\;x^{(2)}_{ijk}=y^{(1)}_{ijk}\}}.δij​=1{xijk(1)​=yijk(1)​,xijk(2)​=yijk(2)​}​,εij​=1{xijk(1)​=yijk(2)​,xijk(2)​=yijk(1)​}​.

The fractions are complex fractions, with division by zero equal to zero. The parameter qqq does not enter the formula. Empty products equal 111, so the value is 111 if N=0N=0N=0 or r=0r=0r=0; if m=0m=0m=0, there is no possible argument kkk.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittHidingSigma

For any parameters P=(N,r,m,d,q)P=(N,r,m,d,q)P=(N,r,m,d,q), with N,r,m,d∈NN,r,m,d\in\mathbb NN,r,m,d∈N and q∈Rq\in\mathbb Rq∈R, any k∈{0,…,m−1}k\in\{0,\ldots,m-1\}k∈{0,…,m−1}, and pairs of shield arrays x=(x(1),x(2))x=(x^{(1)},x^{(2)})x=(x(1),x(2)), y=(y(1),y(2))y=(y^{(1)},y^{(2)})y=(y(1),y(2)) with entries in {0,…,d−1}\{0,\ldots,d-1\}{0,…,d−1} and indices i<N,j<r,ℓ<mi<N,j<r,\ell<mi<N,j<r,ℓ<m, this defines

ΣP,k(x,y)=∏i=0N−1∏j=0r−11{xijk(1)=yijk(1),  xijk(2)=yijk(2)}+1{xijk(1)=yijk(2),  xijk(2)=yijk(1)}d(d+1)∈C.\Sigma_{P,k}(x,y) = \prod_{i=0}^{N-1}\prod_{j=0}^{r-1} \frac{ \mathbf 1_{\{x^{(1)}_{ijk}=y^{(1)}_{ijk},\;x^{(2)}_{ijk}=y^{(2)}_{ijk}\}} + \mathbf 1_{\{x^{(1)}_{ijk}=y^{(2)}_{ijk},\;x^{(2)}_{ijk}=y^{(1)}_{ijk}\}} }{d(d+1)} \in\mathbb C.ΣP,k​(x,y)=i=0∏N−1​j=0∏r−1​d(d+1)1{xijk(1)​=yijk(1)​,xijk(2)​=yijk(2)​}​+1{xijk(1)​=yijk(2)​,xijk(2)​=yijk(1)​}​​∈C.

No parameter validity assumptions are imposed, and qqq does not enter the formula. Complex division by zero equals zero. If N=0N=0N=0 or r=0r=0r=0, the empty product gives 111; if m=0m=0m=0, there is no argument kkk.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittInsertFirst

Let P=(N,r,m,d,q)P=(N,r,m,d,q)P=(N,r,m,d,q) be a parameter record with natural-number fields N,r,m,dN,r,m,dN,r,m,d and real field qqq, and assume N>0N>0N>0. Write H={0,…,d−1}{0,…,r−1}×{0,…,m−1}H=\{0,\ldots,d-1\}^{\{0,\ldots,r-1\}\times\{0,\ldots,m-1\}}H={0,…,d−1}{0,…,r−1}×{0,…,m−1} for the set of shares. Given x∈Hx\in Hx∈H and a function rest:{i∈{0,…,N−1}:i≠0}→H\mathrm{rest}:\{i\in\{0,\ldots,N-1\}:i\ne0\}\to Hrest:{i∈{0,…,N−1}:i=0}→H, this defines the shield configuration s:{0,…,N−1}→Hs:\{0,\ldots,N-1\}\to Hs:{0,…,N−1}→H by

si={x,i=0,rest(i),i≠0.s_i=\begin{cases}x,&i=0,\\ \mathrm{rest}(i),&i\ne0.\end{cases}si​={x,rest(i),​i=0,i=0.​

There are no restrictions on r,m,d,qr,m,d,qr,m,d,q; in particular, N=1N=1N=1 permits the unique function on the empty set as rest\mathrm{rest}rest.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittNoisySwitch

For any finite type III with decidable equality, any function Γ:Mat⁡I(C)→Mat⁡I(C)\Gamma:\operatorname{Mat}_I(\mathbb C)\to\operatorname{Mat}_I(\mathbb C)Γ:MatI​(C)→MatI​(C), and arbitrary real numbers κ,p\kappa,pκ,p, this defines a function from matrices indexed by {false,true}×I\{\mathrm{false},\mathrm{true}\}\times I{false,true}×I to matrices indexed by {false,true}×({false,true}×I)\{\mathrm{false},\mathrm{true}\}\times(\{\mathrm{false},\mathrm{true}\}\times I){false,true}×({false,true}×I). For an input XXX, set Xa,b(s)=X(s,a),(s,b)X^{(s)}_{a,b}=X_{(s,a),(s,b)}Xa,b(s)​=X(s,a),(s,b)​ and set Ytrue=X(true)Y_{\mathrm{true}}=X^{(\mathrm{true})}Ytrue​=X(true), Yfalse=Γ(X(false))Y_{\mathrm{false}}=\Gamma(X^{(\mathrm{false})})Yfalse​=Γ(X(false)), with λtrue=p\lambda_{\mathrm{true}}=pλtrue​=p and λfalse=κ\lambda_{\mathrm{false}}=\kappaλfalse​=κ. The output entries are

N(X)(s,(u,a)),(t,(v,b))={0,s≠t or u≠v,λs tr⁡(Ys) ∣I∣−1δa,b,s=t, u=v=true,(1−λs)(Ys)a,b,s=t, u=v=false.\mathcal N(X)_{(s,(u,a)),(t,(v,b))} = \begin{cases} 0,&s\ne t\text{ or }u\ne v,\\ \lambda_s\,\operatorname{tr}(Y_s)\,|I|^{-1}\delta_{a,b},&s=t,\ u=v=\mathrm{true},\\ (1-\lambda_s)(Y_s)_{a,b},&s=t,\ u=v=\mathrm{false}. \end{cases}N(X)(s,(u,a)),(t,(v,b))​=⎩⎨⎧​0,λs​tr(Ys​)∣I∣−1δa,b​,(1−λs​)(Ys​)a,b​,​s=t or u=v,s=t, u=v=true,s=t, u=v=false.​

There is no channel or linearity assumption on Γ\GammaΓ and no range restriction on κ,p\kappa,pκ,p. The type III may be empty, and inverses use 0−1=00^{-1}=00−1=0.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittNormalizedChoi

For any finite type III with decidable equality, any type JJJ, and any function Φ:Mat⁡I(C)→Mat⁡J(C)\Phi:\operatorname{Mat}_I(\mathbb C)\to\operatorname{Mat}_J(\mathbb C)Φ:MatI​(C)→MatJ​(C), this defines a complex matrix indexed by I×JI\times JI×J through

C(a,b),(a′,b′)=Φ(Ea,a′)b,b′∣I∣,C_{(a,b),(a',b')}=\frac{\Phi(E_{a,a'})_{b,b'}}{|I|},C(a,b),(a′,b′)​=∣I∣Φ(Ea,a′​)b,b′​​,

where Ea,a′E_{a,a'}Ea,a′​ has entry 111 at (a,a′)(a,a')(a,a′) and zero elsewhere, and the denominator is interpreted in C\mathbb CC. No linearity or channel assumption is imposed on Φ\PhiΦ, and JJJ need not be finite. The type III may be empty; complex division by zero is total and gives zero.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittOutputTranspose

For arbitrary types I,JI,JI,J and any complex matrix XXX indexed by I×JI\times JI×J, this defines the matrix TJ(X)T_J(X)TJ​(X) by

TJ(X)(a,b),(a′,b′)=X(a,b′),(a′,b).T_J(X)_{(a,b),(a',b')}=X_{(a,b'),(a',b)}.TJ​(X)(a,b),(a′,b′)​=X(a,b′),(a′,b)​.

Only the two JJJ indices are exchanged. No finiteness, positivity, normalization, or nonemptiness assumption is imposed.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittPaperChannel

For arbitrary P=(N,r,m,d,q)P=(N,r,m,d,q)P=(N,r,m,d,q), with natural N,r,m,dN,r,m,dN,r,m,d and real qqq, and arbitrary real numbers κ,p\kappa,pκ,p, this defines a function from complex matrices indexed by {false,true}×I\{\mathrm{false},\mathrm{true}\}\times I{false,true}×I to complex matrices indexed by {false,true}×({false,true}×I)\{\mathrm{false},\mathrm{true}\}\times(\{\mathrm{false},\mathrm{true}\}\times I){false,true}×({false,true}×I), where I={false,true}×{0,…,d−1}N×r×mI=\{\mathrm{false},\mathrm{true}\}\times\{0,\ldots,d-1\}^{N\times r\times m}I={false,true}×{0,…,d−1}N×r×m and D=∣I∣D=|I|D=∣I∣. For an input matrix XXX, write Xs(a,b)=X(s,a),(s,b)X_s(a,b)=X_{(s,a),(s,b)}Xs​(a,b)=X(s,a),(s,b)​. Set Ytrue=XtrueY_{\mathrm{true}}=X_{\mathrm{true}}Ytrue​=Xtrue​, Yfalse=ΓP(Xfalse)Y_{\mathrm{false}}=\Gamma_P(X_{\mathrm{false}})Yfalse​=ΓP​(Xfalse​), ttrue=pt_{\mathrm{true}}=pttrue​=p, and tfalse=κt_{\mathrm{false}}=\kappatfalse​=κ. The output entry is

ΦP(X)(s,(e,b)),(s′,(f,c))={(1−ts)(Ys)b,c,s=s′, e=f=false,ts tr⁡(Ys) D−11{b=c},s=s′, e=f=true,0,otherwise.\Phi_P(X)_{(s,(e,b)),(s',(f,c))} = \begin{cases} (1-t_s)(Y_s)_{b,c},&s=s',\ e=f=\mathrm{false},\\ t_s\,\operatorname{tr}(Y_s)\,D^{-1}\mathbf 1_{\{b=c\}}, &s=s',\ e=f=\mathrm{true},\\ 0,&\text{otherwise}. \end{cases}ΦP​(X)(s,(e,b)),(s′,(f,c))​=⎩⎨⎧​(1−ts​)(Ys​)b,c​,ts​tr(Ys​)D−11{b=c}​,0,​s=s′, e=f=false,s=s′, e=f=true,otherwise.​

The function ΓP\Gamma_PΓP​ used here is given by

(ΓP(Z))b,c=D2(qm+(12−q)m)∑a,a′∈IZa,a′ H((a,b),(a′,c)).(\Gamma_P(Z))_{b,c} = \frac{D}{2(q^m+(\tfrac12-q)^m)} \sum_{a,a'\in I}Z_{a,a'}\,H((a,b),(a',c)).(ΓP​(Z))b,c​=2(qm+(21​−q)m)D​a,a′∈I∑​Za,a′​H((a,b),(a′,c)).

For x=((α,s),(β,t))x=((\alpha,s),(\beta,t))x=((α,s),(β,t)), y=((γ,s′),(η,t′))y=((\gamma,s'),(\eta,t'))y=((γ,s′),(η,t′)), define δijk=1{sijk=sijk′, tijk=tijk′}\delta_{ijk}=\mathbf 1_{\{s_{ijk}=s'_{ijk},\,t_{ijk}=t'_{ijk}\}}δijk​=1{sijk​=sijk′​,tijk​=tijk′​}​, εijk=1{sijk=tijk′, tijk=sijk′}\varepsilon_{ijk}=\mathbf 1_{\{s_{ijk}=t'_{ijk},\,t_{ijk}=s'_{ijk}\}}εijk​=1{sijk​=tijk′​,tijk​=sijk′​}​, and

Ok=∏i<N,j<r12(δijk+εijkd(d+1)+δijk−εijkd(d−1)),Sk=∏i<N,j<rδijk+εijkd(d+1).O_k=\prod_{i<N,j<r}\frac12 \left(\frac{\delta_{ijk}+\varepsilon_{ijk}}{d(d+1)} +\frac{\delta_{ijk}-\varepsilon_{ijk}}{d(d-1)}\right), \qquad S_k=\prod_{i<N,j<r}\frac{\delta_{ijk}+\varepsilon_{ijk}}{d(d+1)}.Ok​=i<N,j<r∏​21​(d(d+1)δijk​+εijk​​+d(d−1)δijk​−εijk​​),Sk​=i<N,j<r∏​d(d+1)δijk​+εijk​​.

Then

H(x,y)={∏k<mq2(Ok+Sk),α=β, γ=η, α=γ,∏k<mq2(Ok−Sk),α=β, γ=η, α≠γ,∏k<m(12−q)Sk,α≠β, α=γ, β=η,0,otherwise.H(x,y)= \begin{cases} \prod_{k<m}\frac q2(O_k+S_k),& \alpha=\beta,\ \gamma=\eta,\ \alpha=\gamma,\\ \prod_{k<m}\frac q2(O_k-S_k),& \alpha=\beta,\ \gamma=\eta,\ \alpha\ne\gamma,\\ \prod_{k<m}(\tfrac12-q)S_k,& \alpha\ne\beta,\ \alpha=\gamma,\ \beta=\eta,\\ 0,&\text{otherwise}. \end{cases}H(x,y)=⎩⎨⎧​∏k<m​2q​(Ok​+Sk​),∏k<m​2q​(Ok​−Sk​),∏k<m​(21​−q)Sk​,0,​α=β, γ=η, α=γ,α=β, γ=η, α=γ,α=β, α=γ, β=η,otherwise.​

All matrix-entry arithmetic is complex. There are no restrictions requiring κ,p\kappa,pκ,p to lie in [0,1][0,1][0,1], or requiring PPP to be valid; division by zero equals zero and empty products equal 111.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittPrivateState

For any finite type III with decidable equality, any complex matrix σ\sigmaσ indexed by I×II\times II×I, and any family of unitary matrices Uα,βU_{\alpha,\beta}Uα,β​ on I×II\times II×I, indexed by two Booleans, this defines the matrix RRR on (Bool×I)2(\mathrm{Bool}\times I)^2(Bool×I)2 by

R((α,a),(β,b)),((γ,c),(δ,d))={12(Uα,ασUγ,γ†)(a,b),(c,d),α=β and γ=δ,0,otherwise.R_{((\alpha,a),(\beta,b)),((\gamma,c),(\delta,d))} = \begin{cases} \frac12\bigl(U_{\alpha,\alpha}\sigma U_{\gamma,\gamma}^{\dagger}\bigr)_{(a,b),(c,d)}, &\alpha=\beta\ \text{and}\ \gamma=\delta,\\ 0,&\text{otherwise}. \end{cases}R((α,a),(β,b)),((γ,c),(δ,d))​={21​(Uα,α​σUγ,γ†​)(a,b),(c,d)​,0,​α=β and γ=δ,otherwise.​

Only the diagonal members Uα,αU_{\alpha,\alpha}Uα,α​ enter this formula, although all four members of the supplied family must be unitary. The matrix σ\sigmaσ need not be positive semidefinite or have trace one, and III may be empty, in which case the resulting matrix is empty.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittRawChoi

For arbitrary parameters P=(N,r,m,d,q)P=(N,r,m,d,q)P=(N,r,m,d,q), where N,r,m,d∈NN,r,m,d\in\mathbb NN,r,m,d∈N and q∈Rq\in\mathbb Rq∈R, this defines a complex matrix HPH_PHP​ indexed by pairs ((a,s),(b,t))((a,s),(b,t))((a,s),(b,t)), with a,ba,ba,b Boolean and s,ts,ts,t arrays in {0,…,d−1}N×r×m\{0,\ldots,d-1\}^{N\times r\times m}{0,…,d−1}N×r×m. To specify a row x=((a,s),(b,t))x=((a,s),(b,t))x=((a,s),(b,t)) and column y=((c,s′),(e,t′))y=((c,s'),(e,t'))y=((c,s′),(e,t′)), put

δijk=1{sijk=sijk′,  tijk=tijk′},εijk=1{sijk=tijk′,  tijk=sijk′},\delta_{ijk}=\mathbf 1_{\{s_{ijk}=s'_{ijk},\;t_{ijk}=t'_{ijk}\}}, \qquad \varepsilon_{ijk}=\mathbf 1_{\{s_{ijk}=t'_{ijk},\;t_{ijk}=s'_{ijk}\}},δijk​=1{sijk​=sijk′​,tijk​=tijk′​}​,εijk​=1{sijk​=tijk′​,tijk​=sijk′​}​,

and define, for k<mk<mk<m,

Ωk=∏i<N∏j<r12(δijk+εijkd(d+1)+δijk−εijkd(d−1)),Σk=∏i<N∏j<rδijk+εijkd(d+1).\Omega_k=\prod_{i<N}\prod_{j<r} \frac12\left(\frac{\delta_{ijk}+\varepsilon_{ijk}}{d(d+1)} +\frac{\delta_{ijk}-\varepsilon_{ijk}}{d(d-1)}\right), \qquad \Sigma_k=\prod_{i<N}\prod_{j<r} \frac{\delta_{ijk}+\varepsilon_{ijk}}{d(d+1)}.Ωk​=i<N∏​j<r∏​21​(d(d+1)δijk​+εijk​​+d(d−1)δijk​−εijk​​),Σk​=i<N∏​j<r∏​d(d+1)δijk​+εijk​​.

Then its entry is

(HP)x,y={∏k<mq2(Ωk+Σk),a=b, c=e, a=c,∏k<mq2(Ωk−Σk),a=b, c=e, a≠c,∏k<m(12−q)Σk,a≠b, a=c, b=e,0,otherwise.(H_P)_{x,y}= \begin{cases} \displaystyle\prod_{k<m}\frac q2(\Omega_k+\Sigma_k), & a=b,\ c=e,\ a=c,\\[3pt] \displaystyle\prod_{k<m}\frac q2(\Omega_k-\Sigma_k), & a=b,\ c=e,\ a\ne c,\\[3pt] \displaystyle\prod_{k<m}\left(\frac12-q\right)\Sigma_k, & a\ne b,\ a=c,\ b=e,\\[3pt] 0,&\text{otherwise}. \end{cases}(HP​)x,y​=⎩⎨⎧​k<m∏​2q​(Ωk​+Σk​),k<m∏​2q​(Ωk​−Σk​),k<m∏​(21​−q)Σk​,0,​a=b, c=e, a=c,a=b, c=e, a=c,a=b, a=c, b=e,otherwise.​

All displayed entry arithmetic is complex arithmetic. There are no validity hypotheses on PPP; division by zero is zero and empty products are 111, including each product over kkk when m=0m=0m=0.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittRepetitions

For every natural number mmm, this defines the natural number

2m+⌈log⁡mlog⁡2⌉+,2m+\left\lceil\frac{\log m}{\log 2}\right\rceil_+,2m+⌈log2logm​⌉+​,

where log⁡\loglog is the real natural logarithm and ⌈x⌉+\lceil x\rceil_+⌈x⌉+​ is the least natural number greater than or equal to xxx, hence is zero for x≤0x\le0x≤0. The definition includes m=0m=0m=0: the total real logarithm has log⁡0=0\log0=0log0=0, so the resulting number is 000.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittSelectedInput

For arbitrary types S,AS,AS,A, an element i∈Si\in Si∈S, and an arbitrary complex matrix ρ\rhoρ indexed by AAA, this defines the complex matrix indexed by S×AS\times AS×A with entries

X(s,a),(t,b)={ρa,b,s=i and t=i,0,otherwise.X_{(s,a),(t,b)} = \begin{cases} \rho_{a,b},&s=i\text{ and }t=i,\\ 0,&\text{otherwise}. \end{cases}X(s,a),(t,b)​={ρa,b​,0,​s=i and t=i,otherwise.​

Neither finiteness of the types nor positivity or normalization of ρ\rhoρ is required; specifying iii requires SSS to be inhabited.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittSwitched

For arbitrary types S,A,BS,A,BS,A,B and a family of functions Φs:Mat⁡A(C)→Mat⁡B(C)\Phi_s:\operatorname{Mat}_A(\mathbb C)\to\operatorname{Mat}_B(\mathbb C)Φs​:MatA​(C)→MatB​(C) indexed by s∈Ss\in Ss∈S, this defines a function from matrices indexed by S×AS\times AS×A to matrices indexed by S×BS\times BS×B by

S(X)(s,b),(t,c)={Φs(X(s))b,c,s=t,0,s≠t,Xa,a′(s)=X(s,a),(s,a′).\mathcal S(X)_{(s,b),(t,c)} = \begin{cases} \Phi_s(X^{(s)})_{b,c},&s=t,\\ 0,&s\ne t, \end{cases} \qquad X^{(s)}_{a,a'}=X_{(s,a),(s,a')}.S(X)(s,b),(t,c)​={Φs​(X(s))b,c​,0,​s=t,s=t,​Xa,a′(s)​=X(s,a),(s,a′)​.

Thus each diagonal SSS block is passed to its corresponding function, and every off-diagonal SSS block of the output is zero. The types need not be finite or nonempty, and the branch functions need not be linear or channels.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittSymmetricState

For every natural number ddd, this defines a complex square matrix SdS_dSd​ indexed by pairs in {0,…,d−1}2\{0,\ldots,d-1\}^2{0,…,d−1}2. For every row x=(x1,x2)x=(x_1,x_2)x=(x1​,x2​) and column y=(y1,y2)y=(y_1,y_2)y=(y1​,y2​), its entry is

(Sd)x,y=1{x=y}+1{x1=y2,  x2=y1}d(d+1),(S_d)_{x,y} = \frac{\mathbf 1_{\{x=y\}}+\mathbf 1_{\{x_1=y_2,\;x_2=y_1\}}}{d(d+1)},(Sd​)x,y​=d(d+1)1{x=y}​+1{x1​=y2​,x2​=y1​}​​,

where indicators and the denominator are interpreted in C\mathbb CC. The definition includes d=0d=0d=0, when the matrix has an empty index set; complex division is total, with division by zero equal to zero.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittWholeErasure

For any finite type III with decidable equality and any real ppp, this defines a function from complex matrices indexed by III to complex matrices indexed by {false,true}×I\{\mathrm{false},\mathrm{true}\}\times I{false,true}×I. Writing d=∣I∣d=|I|d=∣I∣, its value on an arbitrary matrix XXX is

Ep(X)(s,a),(t,b)={0,s≠t,p tr⁡(X) d−1δa,b,s=t=true,(1−p)Xa,b,s=t=false.\mathcal E_p(X)_{(s,a),(t,b)} = \begin{cases} 0,&s\ne t,\\ p\,\operatorname{tr}(X)\,d^{-1}\delta_{a,b},&s=t=\mathrm{true},\\ (1-p)X_{a,b},&s=t=\mathrm{false}. \end{cases}Ep​(X)(s,a),(t,b)​=⎩⎨⎧​0,ptr(X)d−1δa,b​,(1−p)Xa,b​,​s=t,s=t=true,s=t=false.​

Scalars are regarded as complex numbers, and δa,b\delta_{a,b}δa,b​ is 111 when a=ba=ba=b and 000 otherwise. There is no restriction 0≤p≤10\le p\le10≤p≤1 and no assumption on XXX. The type III may be empty; inverses are total, with 0−1=00^{-1}=00−1=0.

QAlgorithms.Papers.Independent.Intakenext100.P70.cubittZeta

For arbitrary P=(N,r,m,d,q)P=(N,r,m,d,q)P=(N,r,m,d,q), with natural N,r,m,dN,r,m,dN,r,m,d and real qqq, this defines a complex matrix ζP\zeta_PζP​ on pairs of Boolean–shield pairs, where a shield is an array in {0,…,d−1}N×r×m\{0,\ldots,d-1\}^{N\times r\times m}{0,…,d−1}N×r×m. For row x=((a,s),(b,t))x=((a,s),(b,t))x=((a,s),(b,t)) and column y=((c,s′),(e,t′))y=((c,s'),(e,t'))y=((c,s′),(e,t′)), define

δijk=1{sijk=sijk′,  tijk=tijk′},εijk=1{sijk=tijk′,  tijk=sijk′},\delta_{ijk}=\mathbf 1_{\{s_{ijk}=s'_{ijk},\;t_{ijk}=t'_{ijk}\}}, \quad \varepsilon_{ijk}=\mathbf 1_{\{s_{ijk}=t'_{ijk},\;t_{ijk}=s'_{ijk}\}},δijk​=1{sijk​=sijk′​,tijk​=tijk′​}​,εijk​=1{sijk​=tijk′​,tijk​=sijk′​}​, Ωk=∏i<N∏j<r12(δijk+εijkd(d+1)+δijk−εijkd(d−1)),Σk=∏i<N∏j<rδijk+εijkd(d+1).\Omega_k=\prod_{i<N}\prod_{j<r}\frac12 \left(\frac{\delta_{ijk}+\varepsilon_{ijk}}{d(d+1)} +\frac{\delta_{ijk}-\varepsilon_{ijk}}{d(d-1)}\right), \qquad \Sigma_k=\prod_{i<N}\prod_{j<r}\frac{\delta_{ijk}+\varepsilon_{ijk}}{d(d+1)}.Ωk​=i<N∏​j<r∏​21​(d(d+1)δijk​+εijk​​+d(d−1)δijk​−εijk​​),Σk​=i<N∏​j<r∏​d(d+1)δijk​+εijk​​.

Its entries are

(ζP)x,y=12(qm+(12−q)m){∏k<mq2(Ωk+Σk),a=b, c=e, a=c,∏k<mq2(Ωk−Σk),a=b, c=e, a≠c,∏k<m(12−q)Σk,a≠b, a=c, b=e,0,otherwise.(\zeta_P)_{x,y} = \frac{1}{2\bigl(q^m+(\tfrac12-q)^m\bigr)} \begin{cases} \displaystyle\prod_{k<m}\frac q2(\Omega_k+\Sigma_k), &a=b,\ c=e,\ a=c,\\[3pt] \displaystyle\prod_{k<m}\frac q2(\Omega_k-\Sigma_k), &a=b,\ c=e,\ a\ne c,\\[3pt] \displaystyle\prod_{k<m}(\tfrac12-q)\Sigma_k, &a\ne b,\ a=c,\ b=e,\\[3pt] 0,&\text{otherwise}. \end{cases}(ζP​)x,y​=2(qm+(21​−q)m)1​⎩⎨⎧​k<m∏​2q​(Ωk​+Σk​),k<m∏​2q​(Ωk​−Σk​),k<m∏​(21​−q)Σk​,0,​a=b, c=e, a=c,a=b, c=e, a=c,a=b, a=c, b=e,otherwise.​

All arithmetic here is in C\mathbb CC. No nonzero-denominator or validity assumptions are imposed: zero has inverse zero, and empty products equal 111. In particular, when m=0m=0m=0, the prefactor is 1/41/41/4.

Human review
  • Endorsed by Shuze Chen · Oct 31, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Gohan · Oct 31, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me