Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Definitions for Semidefinite programs for completely bounded norms

Definition
QAlgorithms_Papers_Independent_Intakenext100_P98

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

prove2quantum

The objects the formalized statements of Semidefinite programs for completely bounded norms (arXiv:0901.4709) are stated in. Each was checked against the paper by a blind read-back and a moderator.

  • PrimalIndex: PDF p.7: index set for X direct-sum (Y tensor Z).
  • DualIndex: PDF p.7: index set for C direct-sum Z; the scalar block survives r=0.
  • realTracePairing: PDF p.3: real Hilbert--Schmidt pairing, conjugate-linear first.
  • unitTensor: PDF p.4: I_Y tensor Z, with Y first.
  • watrousChoiMatrix: PDF p.4: unnormalized, output-first Choi representation.
  • stinespringPairRepresents: PDF p.4, equation (1): equality on ALL operators, with arbitrary complex-linear Phi.
  • superAdjoint: PDF p.3: Hilbert--Schmidt adjoint in matrix-unit coordinates.
  • stabilizedTraceNorm: PDF p.4: stabilized induced trace norm, with no factor 1/2. The frozen idTensorSuperop places the n-dimensional reference first, an equivalent tensor ordering. The trace-norm ball is nonempty and compact and its linear image is bounded.
  • cbSpectral: PDF p.4: completely bounded spectral norm, using its adjoint identity.
  • primalInputCorner: PDF p.7: the primal input block.
  • primalOutputCorner: PDF p.7: the primal output/environment block.
  • environmentBlock: PDF p.7: the environment block of a dual operator.
  • stinespringSdpMap: PDF p.7: the ACTUAL block map Psi; off-diagonal input blocks are ignored. This is a concrete complex-linear function, rather than arbitrary SDP data.
  • blockAdjoint: PDF p.7: the explicit Hilbert--Schmidt adjoint of Psi.
  • stinespringSdpObjective: PDF p.7: fixed objective C = diag(0, BB*).
  • stinespringSdpBound: PDF p.7: fixed bound D = diag(1,0).
  • BlockPrimalFeasible: PDF pp.7--8: full-block positivity and D - Psi(P) positive semidefinite.
  • BlockDualFeasible: PDF pp.7--8: full-block positivity and Psi*(S) - C positive semidefinite.
  • blockPrimalValue: PDF pp.5,7: primal supremum in EReal; empty feasible set has value minus infinity.
  • blockDualValue: PDF pp.5,7: dual infimum in EReal; empty feasible set has value plus infinity.
  • channelPrimalValues: PDF p.11: actual channel-distance primal objectives, 0 <= W <= I_Y tensor rho.
  • channelDualValues: PDF p.11: ||Tr_Y Z||_infinity with Z >= 0 and Z >= J(Phi).
  • stinespringProducts: PDF p.12, Theorem 5: products over ALL Stinespring pairs and all positive-dimensional finite environments; no minimality condition.
  • fidelityPrimalValues: PDF pp.13--14: fidelity primal values for arbitrary unnormalized u,v.
  • fidelityDualValues: PDF p.13: the PSD dual of the fidelity primal.
  • fidelityPDValues: PDF p.14, last display: constrained PD dual values; no PD minimum is asserted.
  • fidelityProductValues: PDF p.15: Alberti's product objective for the two actual marginals.
Definition code
import Mathlib
import Definitions.Def_QAlgorithms_Defs_MatrixDistances
import Definitions.Def_QAlgorithms_Defs_QuantumInformation
import Definitions.Def_QAlgorithms_Defs_Channel
import Definitions.Def_QAlgorithms_Defs_Measurement

/-!
Watrous, Semidefinite programs for completely bounded norms, arXiv:0901.4709v2.
Page references are PDF pages. All new definitions belong to this paper.
Matrix and complex-linear-map types are Mathlib types, without local aliases.
Distinct real-pairing, two-operator Stinespring, and SDP-corner names avoid
confusion with unrelated accepted declarations; see ASSESSMENT.json.
Only the minimal Stinespring environment in Section 3 may have dimension zero.
-/

noncomputable section

open scoped Matrix ComplexConjugate ComplexOrder

namespace QAlgorithms.Papers.Independent.Intakenext100.P98

/-- PDF p.7: index set for X direct-sum (Y tensor Z). -/
abbrev PrimalIndex (n m r : ℕ) := Fin n ⊕ (Fin m × Fin r)

/-- PDF p.7: index set for C direct-sum Z; the scalar block survives r=0. -/
abbrev DualIndex (r : ℕ) := Fin 1 ⊕ Fin r

/-- PDF p.3: real Hilbert--Schmidt pairing, conjugate-linear first. -/
def realTracePairing {I : Type*} [Fintype I] (A B : (Matrix I I ℂ)) : ℝ :=
  (Aᴴ * B).trace.re

/-- PDF p.4: I_Y tensor Z, with Y first. -/
def unitTensor (m : ℕ) {r : ℕ} (Z : (Matrix (Fin r) (Fin r) ℂ)) : (Matrix (Fin m × Fin r) (Fin m × Fin r) ℂ) :=
  Matrix.kronecker (1 : (Matrix (Fin m) (Fin m) ℂ)) Z

/-- PDF p.4: unnormalized, output-first Choi representation. -/
def watrousChoiMatrix {n m : ℕ} (Phi : (Matrix (Fin n) (Fin n) ℂ →ₗ[ℂ] Matrix (Fin m) (Fin m) ℂ)) : (Matrix (Fin m × Fin n) (Fin m × Fin n) ℂ) :=
  fun a b => Phi (Matrix.single a.2 b.2 1) a.1 b.1

/-- PDF p.4, equation (1): equality on ALL operators, with arbitrary complex-linear Phi. -/
def stinespringPairRepresents {n m r : ℕ} (Phi : (Matrix (Fin n) (Fin n) ℂ →ₗ[ℂ] Matrix (Fin m) (Fin m) ℂ)) (A B : (Matrix (Fin m × Fin r) (Fin n) ℂ)) : Prop :=
  ∀ X : (Matrix (Fin n) (Fin n) ℂ), Phi X = traceRight (A * X * Bᴴ)

/-- PDF p.3: Hilbert--Schmidt adjoint in matrix-unit coordinates. -/
def superAdjoint {n m : ℕ} (Phi : (Matrix (Fin n) (Fin n) ℂ →ₗ[ℂ] Matrix (Fin m) (Fin m) ℂ)) : (Matrix (Fin m) (Fin m) ℂ →ₗ[ℂ] Matrix (Fin n) (Fin n) ℂ) where
  toFun Y i j := ∑ a : Fin m, ∑ b : Fin m,
    star (Phi (Matrix.single i j 1) a b) * Y a b
  map_add' Y Z := by
    ext i j
    change (∑ a : Fin m, ∑ b : Fin m,
      star (Phi (Matrix.single i j 1) a b) * (Y a b + Z a b)) =
        (∑ a : Fin m, ∑ b : Fin m, star (Phi (Matrix.single i j 1) a b) * Y a b) +
        (∑ a : Fin m, ∑ b : Fin m, star (Phi (Matrix.single i j 1) a b) * Z a b)
    simp only [mul_add, Finset.sum_add_distrib]
  map_smul' c Y := by
    ext i j
    change (∑ a : Fin m, ∑ b : Fin m,
      star (Phi (Matrix.single i j 1) a b) * (c * Y a b)) =
        c * (∑ a : Fin m, ∑ b : Fin m, star (Phi (Matrix.single i j 1) a b) * Y a b)
    simp only [Finset.mul_sum, mul_left_comm, mul_assoc]

/-- PDF p.4: stabilized induced trace norm, with no factor 1/2. The frozen
idTensorSuperop places the n-dimensional reference first, an equivalent tensor
ordering. The trace-norm ball is nonempty and compact and its linear image is bounded. -/
def stabilizedTraceNorm {n m : ℕ} (Phi : (Matrix (Fin n) (Fin n) ℂ →ₗ[ℂ] Matrix (Fin m) (Fin m) ℂ)) : ℝ :=
  sSup {v : ℝ | ∃ X : (Matrix (Fin n × Fin n) (Fin n × Fin n) ℂ),
    traceNorm X ≤ 1 ∧ v = traceNorm (idTensorSuperop (R := Fin n) Phi X)}

/-- PDF p.4: completely bounded spectral norm, using its adjoint identity. -/
def cbSpectral {n m : ℕ} (Phi : (Matrix (Fin n) (Fin n) ℂ →ₗ[ℂ] Matrix (Fin m) (Fin m) ℂ)) : ℝ :=
  stabilizedTraceNorm (superAdjoint Phi)

/-- PDF p.7: the primal input block. -/
def primalInputCorner {n m r : ℕ} (P : (Matrix (PrimalIndex n m r) (PrimalIndex n m r) ℂ)) : (Matrix (Fin n) (Fin n) ℂ) :=
  fun i j => P (Sum.inl i) (Sum.inl j)

/-- PDF p.7: the primal output/environment block. -/
def primalOutputCorner {n m r : ℕ} (P : (Matrix (PrimalIndex n m r) (PrimalIndex n m r) ℂ)) : (Matrix (Fin m × Fin r) (Fin m × Fin r) ℂ) :=
  fun i j => P (Sum.inr i) (Sum.inr j)

/-- PDF p.7: the environment block of a dual operator. -/
def environmentBlock {r : ℕ} (S : (Matrix (DualIndex r) (DualIndex r) ℂ)) : (Matrix (Fin r) (Fin r) ℂ) :=
  fun i j => S (Sum.inr i) (Sum.inr j)

/-- PDF p.7: the ACTUAL block map Psi; off-diagonal input blocks are ignored.
This is a concrete complex-linear function, rather than arbitrary SDP data. -/
def stinespringSdpMap {n m r : ℕ} (A : (Matrix (Fin m × Fin r) (Fin n) ℂ))
    (P : (Matrix (PrimalIndex n m r) (PrimalIndex n m r) ℂ)) : (Matrix (DualIndex r) (DualIndex r) ℂ) :=
  Matrix.fromBlocks (fun _ _ : Fin 1 => (primalInputCorner P).trace) 0 0
    (traceLeft (primalOutputCorner P - A * primalInputCorner P * Aᴴ))

/-- PDF p.7: the explicit Hilbert--Schmidt adjoint of Psi. -/
def blockAdjoint {n m r : ℕ} (A : (Matrix (Fin m × Fin r) (Fin n) ℂ))
    (S : (Matrix (DualIndex r) (DualIndex r) ℂ)) : (Matrix (PrimalIndex n m r) (PrimalIndex n m r) ℂ) :=
  Matrix.fromBlocks
    (S (Sum.inl 0) (Sum.inl 0) • (1 : (Matrix (Fin n) (Fin n) ℂ)) -
      Aᴴ * unitTensor m (environmentBlock S) * A)
    0 0 (unitTensor m (environmentBlock S))

/-- PDF p.7: fixed objective C = diag(0, BB*). -/
def stinespringSdpObjective {n m r : ℕ} (B : (Matrix (Fin m × Fin r) (Fin n) ℂ)) : (Matrix (PrimalIndex n m r) (PrimalIndex n m r) ℂ) :=
  Matrix.fromBlocks 0 0 0 (B * Bᴴ)

/-- PDF p.7: fixed bound D = diag(1,0). -/
def stinespringSdpBound (r : ℕ) : (Matrix (DualIndex r) (DualIndex r) ℂ) :=
  Matrix.fromBlocks (1 : (Matrix (Fin 1) (Fin 1) ℂ)) 0 0 0

/-- PDF pp.7--8: full-block positivity and D - Psi(P) positive semidefinite. -/
def BlockPrimalFeasible {n m r : ℕ} (A : (Matrix (Fin m × Fin r) (Fin n) ℂ))
    (P : (Matrix (PrimalIndex n m r) (PrimalIndex n m r) ℂ)) : Prop :=
  P.PosSemidef ∧ (stinespringSdpBound r - stinespringSdpMap A P).PosSemidef

/-- PDF pp.7--8: full-block positivity and Psi*(S) - C positive semidefinite. -/
def BlockDualFeasible {n m r : ℕ} (A B : (Matrix (Fin m × Fin r) (Fin n) ℂ))
    (S : (Matrix (DualIndex r) (DualIndex r) ℂ)) : Prop :=
  S.PosSemidef ∧ (blockAdjoint A S - stinespringSdpObjective B).PosSemidef

/-- PDF pp.5,7: primal supremum in EReal; empty feasible set has value minus infinity. -/
def blockPrimalValue {n m r : ℕ} (A B : (Matrix (Fin m × Fin r) (Fin n) ℂ)) : EReal :=
  sSup {v : EReal | ∃ P : (Matrix (PrimalIndex n m r) (PrimalIndex n m r) ℂ),
    BlockPrimalFeasible A P ∧ v = (realTracePairing (stinespringSdpObjective B) P : EReal)}

/-- PDF pp.5,7: dual infimum in EReal; empty feasible set has value plus infinity. -/
def blockDualValue {n m r : ℕ} (A B : (Matrix (Fin m × Fin r) (Fin n) ℂ)) : EReal :=
  sInf {v : EReal | ∃ S : (Matrix (DualIndex r) (DualIndex r) ℂ),
    BlockDualFeasible A B S ∧ v = (realTracePairing (stinespringSdpBound r) S : EReal)}

/-- PDF p.11: actual channel-distance primal objectives, 0 <= W <= I_Y tensor rho. -/
def channelPrimalValues {n m : ℕ} (Phi : (Matrix (Fin n) (Fin n) ℂ →ₗ[ℂ] Matrix (Fin m) (Fin m) ℂ)) : Set ℝ :=
  {v | ∃ (rho : (Matrix (Fin n) (Fin n) ℂ)) (W : (Matrix (Fin m × Fin n) (Fin m × Fin n) ℂ)),
    IsDensityMatrix rho ∧ W.PosSemidef ∧ (unitTensor m rho - W).PosSemidef ∧
    v = realTracePairing (watrousChoiMatrix Phi) W}

/-- PDF p.11: ||Tr_Y Z||_infinity with Z >= 0 and Z >= J(Phi). -/
def channelDualValues {n m : ℕ} (Phi : (Matrix (Fin n) (Fin n) ℂ →ₗ[ℂ] Matrix (Fin m) (Fin m) ℂ)) : Set ℝ :=
  {v | ∃ Z : (Matrix (Fin m × Fin n) (Fin m × Fin n) ℂ),
    Z.PosSemidef ∧ (Z - watrousChoiMatrix Phi).PosSemidef ∧ v = specNorm (traceLeft Z)}

/-- PDF p.12, Theorem 5: products over ALL Stinespring pairs and all
positive-dimensional finite environments; no minimality condition. -/
def stinespringProducts {n m : ℕ} (Phi : (Matrix (Fin n) (Fin n) ℂ →ₗ[ℂ] Matrix (Fin m) (Fin m) ℂ)) : Set ℝ :=
  {v | ∃ (r : ℕ), 0 < r ∧ ∃ A B : (Matrix (Fin m × Fin r) (Fin n) ℂ),
    stinespringPairRepresents Phi A B ∧ v = specNorm A * specNorm B}

/-- PDF pp.13--14: fidelity primal values for arbitrary unnormalized u,v. -/
def fidelityPrimalValues {m r : ℕ} (u v : EuclideanSpace ℂ (Fin m × Fin r)) : Set ℝ :=
  {a | ∃ W : (Matrix (Fin m × Fin r) (Fin m × Fin r) ℂ), W.PosSemidef ∧
    (traceLeft (pureDensity u) - traceLeft W).PosSemidef ∧
    a = realTracePairing (pureDensity v) W}

/-- PDF p.13: the PSD dual of the fidelity primal. -/
def fidelityDualValues {m r : ℕ} (u v : EuclideanSpace ℂ (Fin m × Fin r)) : Set ℝ :=
  {a | ∃ Z : (Matrix (Fin r) (Fin r) ℂ), Z.PosSemidef ∧
    (unitTensor m Z - pureDensity v).PosSemidef ∧
    a = realTracePairing (traceLeft (pureDensity u)) Z}

/-- PDF p.14, last display: constrained PD dual values; no PD minimum is asserted. -/
def fidelityPDValues {m r : ℕ} (u v : EuclideanSpace ℂ (Fin m × Fin r)) : Set ℝ :=
  {a | ∃ Z : (Matrix (Fin r) (Fin r) ℂ), Z.PosDef ∧
    realTracePairing (traceLeft (pureDensity v)) Z⁻¹ ≤ 1 ∧
    a = realTracePairing (traceLeft (pureDensity u)) Z}

/-- PDF p.15: Alberti's product objective for the two actual marginals. -/
def fidelityProductValues {m r : ℕ} (u v : EuclideanSpace ℂ (Fin m × Fin r)) : Set ℝ :=
  {a | ∃ Z : (Matrix (Fin r) (Fin r) ℂ), Z.PosDef ∧
    a = realTracePairing (traceLeft (pureDensity u)) Z * realTracePairing (traceLeft (pureDensity v)) Z⁻¹}

/-- Sanity: zero Hilbert--Schmidt pairing, PDF p.3. -/
example {I : Type*} [Fintype I] (A : (Matrix I I ℂ)) : realTracePairing 0 A = 0 := by
  simp [realTracePairing]

/-- Sanity: I_C tensor Z has no extra scalar, PDF p.4. -/
example {r : ℕ} (Z : (Matrix (Fin r) (Fin r) ℂ)) (i j : Fin r) :
    unitTensor 1 Z (0, i) (0, j) = Z i j := by
  simp [unitTensor, Matrix.kronecker, Matrix.one_apply]

/-- Sanity: the zero map has zero Choi matrix, PDF p.4. -/
example {n m : ℕ} : watrousChoiMatrix (0 : (Matrix (Fin n) (Fin n) ℂ →ₗ[ℂ] Matrix (Fin m) (Fin m) ℂ)) = 0 := by
  ext a b
  rfl

/-- Sanity: the zero pair represents zero even in a zero-dimensional environment. -/
example {n m r : ℕ} : stinespringPairRepresents (0 : (Matrix (Fin n) (Fin n) ℂ →ₗ[ℂ] Matrix (Fin m) (Fin m) ℂ))
    (0 : (Matrix (Fin m × Fin r) (Fin n) ℂ)) (0 : (Matrix (Fin m × Fin r) (Fin n) ℂ)) := by
  intro X
  ext i j
  simp [traceRight, Matrix.zero_apply]

/-- Sanity: the coordinate adjoint maps zero to zero, PDF p.3. -/
example {n m : ℕ} : superAdjoint (0 : (Matrix (Fin n) (Fin n) ℂ →ₗ[ℂ] Matrix (Fin m) (Fin m) ℂ)) = 0 := by
  ext Y i j
  change (∑ a : Fin m, ∑ b : Fin m, star (0 : ℂ) * Y a b) = 0
  simp

/-- Sanity: Psi's scalar entry is Tr X, PDF p.7. -/
example {n m r : ℕ} (A : (Matrix (Fin m × Fin r) (Fin n) ℂ)) (P : (Matrix (PrimalIndex n m r) (PrimalIndex n m r) ℂ)) :
    stinespringSdpMap A P (Sum.inl 0) (Sum.inl 0) = (primalInputCorner P).trace := by
  rfl

/-- Sanity: Psi ignores off-diagonal blocks, PDF p.7. -/
example {n m r : ℕ} (A : (Matrix (Fin m × Fin r) (Fin n) ℂ)) (P : (Matrix (PrimalIndex n m r) (PrimalIndex n m r) ℂ)) (i : Fin r) :
    stinespringSdpMap A P (Sum.inl 0) (Sum.inr i) = 0 := by
  rfl

/-- Sanity: the environment block of Psi*, PDF p.7. -/
example {n m r : ℕ} (A : (Matrix (Fin m × Fin r) (Fin n) ℂ)) (S : (Matrix (DualIndex r) (DualIndex r) ℂ))
    (i j : Fin m × Fin r) :
    blockAdjoint A S (Sum.inr i) (Sum.inr j) = unitTensor m (environmentBlock S) i j := by
  rfl

/-- Sanity: the actual objective's lower block is BB*, PDF p.7. -/
example {n m r : ℕ} (B : (Matrix (Fin m × Fin r) (Fin n) ℂ)) (i j : Fin m × Fin r) :
    stinespringSdpObjective B (Sum.inr i) (Sum.inr j) = (B * Bᴴ) i j := by
  rfl

/-- Sanity: the scalar bound is still one with an empty environment, PDF p.7. -/
example : stinespringSdpBound 0 (Sum.inl 0) (Sum.inl 0) = 1 := by
  rfl

end QAlgorithms.Papers.Independent.Intakenext100.P98
Source
Semidefinite programs for completely bounded norms, arXiv:0901.4709v2, https://arxiv.org/abs/0901.4709v2
Read-back

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

QAlgorithms.Papers.Independent.Intakenext100.P98.BlockDualFeasible

For any natural numbers n,m,rn,m,rn,m,r, any complex mr×nmr\times nmr×n matrices A,BA,BA,B, and any complex square matrix SSS indexed by Fin(1)⊔Fin(r)\mathrm{Fin}(1)\sqcup\mathrm{Fin}(r)Fin(1)⊔Fin(r), let sss be its upper-left scalar entry and ZZZ its lower-right r×rr\times rr×r block. This predicate means precisely

S⪰0and(sIn−A†(Im⊗Z)A00Im⊗Z−BB†)⪰0,S\succeq0 \quad\text{and}\quad \begin{pmatrix} sI_n-A^\dagger(I_m\otimes Z)A&0\\ 0&I_m\otimes Z-BB^\dagger \end{pmatrix}\succeq0,S⪰0and(sIn​−A†(Im​⊗Z)A0​0Im​⊗Z−BB†​)⪰0,

where ⪰0\succeq0⪰0 means Hermitian positive semidefinite and (Im⊗Z)(a,i),(b,j)=δabZij(I_m\otimes Z)_{(a,i),(b,j)}=\delta_{ab}Z_{ij}(Im​⊗Z)(a,i),(b,j)​=δab​Zij​. The off-diagonal blocks of SSS enter its positivity requirement but not the second matrix. Zero dimensions are allowed, using empty blocks and sums.

QAlgorithms.Papers.Independent.Intakenext100.P98.BlockPrimalFeasible

For any natural numbers n,m,rn,m,rn,m,r, any complex mr×nmr\times nmr×n matrix AAA, and any complex square matrix PPP indexed by Fin(n)⊔(Fin(m)×Fin(r))\mathrm{Fin}(n)\sqcup(\mathrm{Fin}(m)\times\mathrm{Fin}(r))Fin(n)⊔(Fin(m)×Fin(r)), write XXX and YYY for its upper-left and lower-right diagonal blocks. This predicate means precisely

P⪰0and(1−tr⁡X00Tr⁡m(AXA†−Y))⪰0,(Tr⁡mT)ij=∑a=0m−1T(a,i),(a,j),P\succeq0 \quad\text{and}\quad \begin{pmatrix} 1-\operatorname{tr}X&0\\ 0&\operatorname{Tr}_{m}(AXA^\dagger-Y) \end{pmatrix}\succeq0, \qquad (\operatorname{Tr}_{m}T)_{ij} =\sum_{a=0}^{m-1}T_{(a,i),(a,j)},P⪰0and(1−trX0​0Trm​(AXA†−Y)​)⪰0,(Trm​T)ij​=a=0∑m−1​T(a,i),(a,j)​,

where ⪰0\succeq0⪰0 means Hermitian positive semidefinite. The off-diagonal blocks of PPP enter its positivity requirement but not the second matrix. All dimensions may be zero, with empty sums and traces equal to zero.

QAlgorithms.Papers.Independent.Intakenext100.P98.DualIndex

For any natural number rrr, this defines the index set {0}⊔{0,…,r−1}\{0\}\sqcup\{0,\ldots,r-1\}{0}⊔{0,…,r−1} as a tagged disjoint union, of cardinality 1+r1+r1+r. The two summands remain distinct even when their numerical labels coincide; when r=0r=0r=0, only the singleton first summand remains.

QAlgorithms.Papers.Independent.Intakenext100.P98.PrimalIndex

For any natural numbers n,m,rn,m,rn,m,r, this defines the index set {0,…,n−1}⊔({0,…,m−1}×{0,…,r−1})\{0,\ldots,n-1\}\sqcup(\{0,\ldots,m-1\}\times\{0,\ldots,r-1\}){0,…,n−1}⊔({0,…,m−1}×{0,…,r−1}), where ⊔\sqcup⊔ is a tagged disjoint union. Its cardinality is n+mrn+mrn+mr. Zero dimensions are allowed: the first summand is empty when n=0n=0n=0, and the second is empty when m=0m=0m=0 or r=0r=0r=0.

QAlgorithms.Papers.Independent.Intakenext100.P98.blockAdjoint

For any natural numbers n,m,rn,m,rn,m,r, any complex mr×nmr\times nmr×n matrix AAA, and any complex square matrix SSS indexed by Fin(1)⊔Fin(r)\mathrm{Fin}(1)\sqcup\mathrm{Fin}(r)Fin(1)⊔Fin(r), let s∈Cs\in\mathbb Cs∈C be the entry of SSS in its one-dimensional upper-left block and let ZZZ be its r×rr\times rr×r lower-right block. This definition returns the matrix indexed by Fin(n)⊔(Fin(m)×Fin(r))\mathrm{Fin}(n)\sqcup(\mathrm{Fin}(m)\times\mathrm{Fin}(r))Fin(n)⊔(Fin(m)×Fin(r)) given by

(sIn−A†(Im⊗Z)A00Im⊗Z),\begin{pmatrix} sI_n-A^\dagger(I_m\otimes Z)A&0\\ 0&I_m\otimes Z \end{pmatrix},(sIn​−A†(Im​⊗Z)A0​0Im​⊗Z​),

where (Im⊗Z)(a,i),(b,j)=δabZij(I_m\otimes Z)_{(a,i),(b,j)}=\delta_{ab}Z_{ij}(Im​⊗Z)(a,i),(b,j)​=δab​Zij​. The off-diagonal blocks of SSS are unused. Neither AAA nor SSS is assumed Hermitian or positive, and zero dimensions are allowed with empty blocks and sums interpreted accordingly.

QAlgorithms.Papers.Independent.Intakenext100.P98.blockDualValue

For arbitrary natural numbers n,m,rn,m,rn,m,r, including zero, and complex mr×nmr\times nmr×n matrices A,BA,BA,B, this definition is the infimum in the extended real numbers R∪{−∞,+∞}\mathbb R\cup\{-\infty,+\infty\}R∪{−∞,+∞} of Re⁡s\operatorname{Re}sRes over all complex matrices SSS on Fin(1)⊔Fin(r)\mathrm{Fin}(1)\sqcup\mathrm{Fin}(r)Fin(1)⊔Fin(r), where sss is the upper-left scalar entry and ZZZ the lower-right block, satisfying

S⪰0,(sIn−A†(Im⊗Z)A00Im⊗Z−BB†)⪰0.S\succeq0, \qquad \begin{pmatrix} sI_n-A^\dagger(I_m\otimes Z)A&0\\ 0&I_m\otimes Z-BB^\dagger \end{pmatrix}\succeq0.S⪰0,(sIn​−A†(Im​⊗Z)A0​0Im​⊗Z−BB†​)⪰0.

Here positivity means Hermitian positive semidefiniteness, Im⊗ZI_m\otimes ZIm​⊗Z has entries δabZij\delta_{ab}Z_{ij}δab​Zij​, and the objective is the expanded real trace pairing with diag⁡(1,0)\operatorname{diag}(1,0)diag(1,0). Real objective values are embedded into the extended reals; an empty feasible set has infimum +∞+\infty+∞. Empty blocks and sums are permitted, and the definition asserts no attainment.

QAlgorithms.Papers.Independent.Intakenext100.P98.blockPrimalValue

For arbitrary natural numbers n,m,rn,m,rn,m,r, including zero, and complex mr×nmr\times nmr×n matrices A,BA,BA,B, this definition is the supremum in the extended real numbers R∪{−∞,+∞}\mathbb R\cup\{-\infty,+\infty\}R∪{−∞,+∞} of the real numbers

Re⁡tr⁡(BB†Y)\operatorname{Re}\operatorname{tr}(BB^\dagger Y)Retr(BB†Y)

over all complex matrices PPP on Fin(n)⊔(Fin(m)×Fin(r))\mathrm{Fin}(n)\sqcup(\mathrm{Fin}(m)\times\mathrm{Fin}(r))Fin(n)⊔(Fin(m)×Fin(r)), with diagonal blocks X,YX,YX,Y, satisfying

P⪰0,(1−tr⁡X00Tr⁡m(AXA†−Y))⪰0,(Tr⁡mT)ij=∑a=0m−1T(a,i),(a,j).P\succeq0, \qquad \begin{pmatrix} 1-\operatorname{tr}X&0\\ 0&\operatorname{Tr}_{m}(AXA^\dagger-Y) \end{pmatrix}\succeq0, \qquad (\operatorname{Tr}_{m}T)_{ij} =\sum_{a=0}^{m-1}T_{(a,i),(a,j)}.P⪰0,(1−trX0​0Trm​(AXA†−Y)​)⪰0,(Trm​T)ij​=a=0∑m−1​T(a,i),(a,j)​.

Each real objective value is embedded into the extended reals before taking the supremum. Here positivity means Hermitian positive semidefiniteness; the displayed objective expands the pairing Re⁡tr⁡(C†P)\operatorname{Re}\operatorname{tr}(C^\dagger P)Retr(C†P) with C=diag⁡(0,BB†)C=\operatorname{diag}(0,BB^\dagger)C=diag(0,BB†). Empty blocks and sums are allowed. The feasible set contains P=0P=0P=0; no attainment assertion is part of this definition.

QAlgorithms.Papers.Independent.Intakenext100.P98.cbSpectral

For any natural numbers n,mn,mn,m and any complex-linear map Φ\PhiΦ from complex n×nn\times nn×n matrices to complex m×mm\times mm×m matrices, define the reverse complex-linear map Ψ\PsiΨ by (Ψ(Y))ij=∑a,b=0m−1(Φ(Eij))ab‾ Yab(\Psi(Y))_{ij}=\sum_{a,b=0}^{m-1}\overline{(\Phi(E_{ij}))_{ab}}\,Y_{ab}(Ψ(Y))ij​=∑a,b=0m−1​(Φ(Eij​))ab​​Yab​, where EijE_{ij}Eij​ has entry 111 at (i,j)(i,j)(i,j) and 000 elsewhere. This declaration defines the real number sup⁡{∥(id⁡m⊗Ψ)(X)∥tr:∥X∥tr≤1}\sup\{\|(\operatorname{id}_m\otimes\Psi)(X)\|_{\mathrm{tr}}:\|X\|_{\mathrm{tr}}\le1\}sup{∥(idm​⊗Ψ)(X)∥tr​:∥X∥tr​≤1}, where XXX ranges over all complex matrices indexed by {0,…,m−1}×{0,…,m−1}\{0,\ldots,m-1\}\times\{0,\ldots,m-1\}{0,…,m−1}×{0,…,m−1} and ∥M∥tr=Re⁡tr⁡M†M\|M\|_{\mathrm{tr}}=\operatorname{Re}\operatorname{tr}\sqrt{M^\dagger M}∥M∥tr​=RetrM†M​, using continuous functional calculus for the square root. The amplified map applies Ψ\PsiΨ to each block: its entry at ((a,i),(b,j))((a,i),(b,j))((a,i),(b,j)) is (Ψ(Xab))ij(\Psi(X^{ab}))_{ij}(Ψ(Xab))ij​, where Xuvab=X(a,u),(b,v)X^{ab}_{uv}=X_{(a,u),(b,v)}Xuvab​=X(a,u),(b,v)​, a,b,u,v<ma,b,u,v<ma,b,u,v<m, and i,j<ni,j<ni,j<n. Either dimension may be zero, giving value 000.

QAlgorithms.Papers.Independent.Intakenext100.P98.channelDualValues

For any natural numbers n,mn,mn,m and any complex-linear map Φ:Mn(C)→Mm(C)\Phi:M_n(\mathbb C)\to M_m(\mathbb C)Φ:Mn​(C)→Mm​(C), this definition gives the set of all real numbers

∥Tr⁡mZ∥op\bigl\|\operatorname{Tr}_{m}Z\bigr\|_{\mathrm{op}}​Trm​Z​op​

for which there exists a complex matrix ZZZ indexed by Fin(m)×Fin(n)\mathrm{Fin}(m)\times\mathrm{Fin}(n)Fin(m)×Fin(n) satisfying Z⪰0Z\succeq0Z⪰0 and Z−J(Φ)⪰0Z-J(\Phi)\succeq0Z−J(Φ)⪰0. Here positivity means Hermitian positive semidefiniteness, the norm is the operator norm for Euclidean norms, and

(Tr⁡mZ)ij=∑a=0m−1Z(a,i),(a,j),J(Φ)(a,i),(b,j)=Φ(Eij)ab,(\operatorname{Tr}_{m}Z)_{ij}=\sum_{a=0}^{m-1}Z_{(a,i),(a,j)}, \qquad J(\Phi)_{(a,i),(b,j)}=\Phi(E_{ij})_{ab},(Trm​Z)ij​=a=0∑m−1​Z(a,i),(a,j)​,J(Φ)(a,i),(b,j)​=Φ(Eij​)ab​,

where EijE_{ij}Eij​ is the matrix unit. No channel or positivity assumption is imposed on Φ\PhiΦ. Both dimensions may be zero; if either is zero, the set is {0}\{0\}{0}, using empty matrices and sums.

QAlgorithms.Papers.Independent.Intakenext100.P98.channelPrimalValues

For any natural numbers n,mn,mn,m and any complex-linear map Φ:Mn(C)→Mm(C)\Phi:M_n(\mathbb C)\to M_m(\mathbb C)Φ:Mn​(C)→Mm​(C), this definition gives the set of all real numbers

Re⁡tr⁡(J(Φ)†W)\operatorname{Re}\operatorname{tr}\bigl(J(\Phi)^\dagger W\bigr)Retr(J(Φ)†W)

for which there exist ρ∈Mn(C)\rho\in M_n(\mathbb C)ρ∈Mn​(C) and W∈Mmn(C)W\in M_{mn}(\mathbb C)W∈Mmn​(C) satisfying

ρ⪰0,tr⁡ρ=1,W⪰0,Im⊗ρ−W⪰0.\rho\succeq0,\qquad \operatorname{tr}\rho=1,\qquad W\succeq0,\qquad I_m\otimes\rho-W\succeq0.ρ⪰0,trρ=1,W⪰0,Im​⊗ρ−W⪰0.

Here J(Φ)J(\Phi)J(Φ) is indexed by Fin(m)×Fin(n)\mathrm{Fin}(m)\times\mathrm{Fin}(n)Fin(m)×Fin(n) and has entries J(Φ)(a,i),(b,j)=Φ(Eij)abJ(\Phi)_{(a,i),(b,j)}=\Phi(E_{ij})_{ab}J(Φ)(a,i),(b,j)​=Φ(Eij​)ab​, where EijE_{ij}Eij​ is the matrix unit; (Im⊗ρ)(a,i),(b,j)=δabρij(I_m\otimes\rho)_{(a,i),(b,j)}=\delta_{ab}\rho_{ij}(Im​⊗ρ)(a,i),(b,j)​=δab​ρij​; and positivity means Hermitian positive semidefiniteness. No channel or positivity assumption is imposed on Φ\PhiΦ. Zero dimensions are included: if n=0n=0n=0, the trace-one requirement is impossible and the set is empty; if n>0n>0n>0 and m=0m=0m=0, the set is {0}\{0\}{0}.

QAlgorithms.Papers.Independent.Intakenext100.P98.environmentBlock

For any natural number rrr and any complex matrix SSS whose rows and columns are indexed by the tagged disjoint union {0}⊔{0,…,r−1}\{0\}\sqcup\{0,\ldots,r-1\}{0}⊔{0,…,r−1}, this defines its r×rr\times rr×r principal block on the second summand, with entry Sinr⁡(i),inr⁡(j)S_{\operatorname{inr}(i),\operatorname{inr}(j)}Sinr(i),inr(j)​ for 0≤i,j<r0\le i,j<r0≤i,j<r. When r=0r=0r=0, the extracted matrix is empty.

QAlgorithms.Papers.Independent.Intakenext100.P98.fidelityDualValues

For arbitrary natural numbers m,rm,rm,r and arbitrary vectors u,v∈Cmru,v\in\mathbb C^{mr}u,v∈Cmr, let Rv=vv†R_v=vv^\daggerRv​=vv† and define A∈Mr(C)A\in M_r(\mathbb C)A∈Mr​(C) by Ats=∑a=0m−1u(a,t)u(a,s)‾A_{ts}=\sum_{a=0}^{m-1}u_{(a,t)}\overline{u_{(a,s)}}Ats​=∑a=0m−1​u(a,t)​u(a,s)​​. This declaration defines the real value set

{Re⁡tr⁡(A†Z):Z∈Mr(C),Z⪰0,Im⊗Z−Rv⪰0},\left\{\operatorname{Re}\operatorname{tr}(A^\dagger Z): Z\in M_r(\mathbb C),\quad Z\succeq0,\quad I_m\otimes Z-R_v\succeq0\right\},{Retr(A†Z):Z∈Mr​(C),Z⪰0,Im​⊗Z−Rv​⪰0},

where †\dagger† denotes conjugate transpose and ⪰0\succeq0⪰0 means positive semidefinite. No normalization or nonzero assumption is imposed on either vector; dimensions may be zero, and empty sums are zero.

QAlgorithms.Papers.Independent.Intakenext100.P98.fidelityPDValues

For arbitrary natural numbers m,rm,rm,r and arbitrary vectors u,v∈Cmru,v\in\mathbb C^{mr}u,v∈Cmr, define matrices A,B∈Mr(C)A,B\in M_r(\mathbb C)A,B∈Mr​(C) by Ats=∑a=0m−1u(a,t)u(a,s)‾A_{ts}=\sum_{a=0}^{m-1}u_{(a,t)}\overline{u_{(a,s)}}Ats​=∑a=0m−1​u(a,t)​u(a,s)​​ and Bts=∑a=0m−1v(a,t)v(a,s)‾B_{ts}=\sum_{a=0}^{m-1}v_{(a,t)}\overline{v_{(a,s)}}Bts​=∑a=0m−1​v(a,t)​v(a,s)​​. This declaration defines

{Re⁡tr⁡(A†Z):Z∈Mr(C),Z is positive definite,Re⁡tr⁡(B†Z−1)≤1}.\left\{\operatorname{Re}\operatorname{tr}(A^\dagger Z): Z\in M_r(\mathbb C),\quad Z\text{ is positive definite},\quad \operatorname{Re}\operatorname{tr}(B^\dagger Z^{-1})\le1\right\}.{Retr(A†Z):Z∈Mr​(C),Z is positive definite,Retr(B†Z−1)≤1}.

Here †\dagger† denotes conjugate transpose and Z−1Z^{-1}Z−1 is the matrix inverse. The vectors need not be normalized or nonzero. Zero dimensions are permitted, with empty sums zero; positive definiteness requires strict positivity on every nonzero vector, a condition that is vacuous in dimension zero.

QAlgorithms.Papers.Independent.Intakenext100.P98.fidelityPrimalValues

For arbitrary natural numbers m,rm,rm,r and arbitrary vectors u,v∈Cmru,v\in\mathbb C^{mr}u,v∈Cmr, with coordinates indexed by (a,t)(a,t)(a,t), 0≤a<m0\le a<m0≤a<m, 0≤t<r0\le t<r0≤t<r, define Ru=uu†R_u=uu^\daggerRu​=uu†, Rv=vv†R_v=vv^\daggerRv​=vv†, and (Tr⁡mY)ts=∑a=0m−1Y(a,t),(a,s)(\operatorname{Tr}_m Y)_{ts}=\sum_{a=0}^{m-1}Y_{(a,t),(a,s)}(Trm​Y)ts​=∑a=0m−1​Y(a,t),(a,s)​. This declaration defines the real value set

{Re⁡tr⁡(Rv†W):W∈Mmr(C),W⪰0,Tr⁡mRu−Tr⁡mW⪰0},\left\{\operatorname{Re}\operatorname{tr}(R_v^\dagger W): W\in M_{mr}(\mathbb C),\quad W\succeq0,\quad \operatorname{Tr}_m R_u-\operatorname{Tr}_m W\succeq0\right\},{Retr(Rv†​W):W∈Mmr​(C),W⪰0,Trm​Ru​−Trm​W⪰0},

where †\dagger† denotes conjugate transpose and ⪰0\succeq0⪰0 means positive semidefinite. Neither vector is required to have unit norm, and zero vectors and zero dimensions are allowed, with empty sums equal to zero.

QAlgorithms.Papers.Independent.Intakenext100.P98.fidelityProductValues

For arbitrary natural numbers m,rm,rm,r and arbitrary vectors u,v∈Cmru,v\in\mathbb C^{mr}u,v∈Cmr, define A,B∈Mr(C)A,B\in M_r(\mathbb C)A,B∈Mr​(C) by Ats=∑a=0m−1u(a,t)u(a,s)‾A_{ts}=\sum_{a=0}^{m-1}u_{(a,t)}\overline{u_{(a,s)}}Ats​=∑a=0m−1​u(a,t)​u(a,s)​​ and Bts=∑a=0m−1v(a,t)v(a,s)‾B_{ts}=\sum_{a=0}^{m-1}v_{(a,t)}\overline{v_{(a,s)}}Bts​=∑a=0m−1​v(a,t)​v(a,s)​​. This declaration defines the real value set

{Re⁡tr⁡(A†Z) Re⁡tr⁡(B†Z−1):Z∈Mr(C),Z is positive definite}.\left\{ \operatorname{Re}\operatorname{tr}(A^\dagger Z)\, \operatorname{Re}\operatorname{tr}(B^\dagger Z^{-1}): Z\in M_r(\mathbb C),\quad Z\text{ is positive definite} \right\}.{Retr(A†Z)Retr(B†Z−1):Z∈Mr​(C),Z is positive definite}.

Here †\dagger† denotes conjugate transpose and Z−1Z^{-1}Z−1 is the matrix inverse. There is no additional constraint on either trace pairing. The vectors may be zero or unnormalized, and dimensions may be zero, with empty sums zero and positive definiteness vacuous on a zero-dimensional space.

QAlgorithms.Papers.Independent.Intakenext100.P98.primalInputCorner

For any natural numbers n,m,rn,m,rn,m,r and any complex matrix PPP whose rows and columns are indexed by the tagged disjoint union {0,…,n−1}⊔({0,…,m−1}×{0,…,r−1})\{0,\ldots,n-1\}\sqcup(\{0,\ldots,m-1\}\times\{0,\ldots,r-1\}){0,…,n−1}⊔({0,…,m−1}×{0,…,r−1}), this defines its n×nn\times nn×n principal block on the first summand: the entry at (i,j)(i,j)(i,j) is Pinl⁡(i),inl⁡(j)P_{\operatorname{inl}(i),\operatorname{inl}(j)}Pinl(i),inl(j)​. All dimensions may be zero; when n=0n=0n=0, the extracted matrix is empty.

QAlgorithms.Papers.Independent.Intakenext100.P98.primalOutputCorner

For any natural numbers n,m,rn,m,rn,m,r and any complex matrix PPP whose rows and columns are indexed by the tagged disjoint union {0,…,n−1}⊔({0,…,m−1}×{0,…,r−1})\{0,\ldots,n-1\}\sqcup(\{0,\ldots,m-1\}\times\{0,\ldots,r-1\}){0,…,n−1}⊔({0,…,m−1}×{0,…,r−1}), this defines its principal block on the second summand: for pairs i,j∈{0,…,m−1}×{0,…,r−1}i,j\in\{0,\ldots,m-1\}\times\{0,\ldots,r-1\}i,j∈{0,…,m−1}×{0,…,r−1}, the extracted entry is Pinr⁡(i),inr⁡(j)P_{\operatorname{inr}(i),\operatorname{inr}(j)}Pinr(i),inr(j)​. This is an mr×mrmr\times mrmr×mr matrix, and it is empty when m=0m=0m=0 or r=0r=0r=0; n=0n=0n=0 is also allowed.

QAlgorithms.Papers.Independent.Intakenext100.P98.realTracePairing

For any finite type III equipped with a finite enumeration, and any complex matrices A,BA,BA,B whose rows and columns are indexed by III, this defines their real trace pairing to be Re⁡tr⁡(A†B)\operatorname{Re}\operatorname{tr}(A^\dagger B)Retr(A†B), where A†A^\daggerA† is the conjugate transpose. Equivalently, its value is Re⁡∑i,j∈IAij‾Bij\operatorname{Re}\sum_{i,j\in I}\overline{A_{ij}}B_{ij}Re∑i,j∈I​Aij​​Bij​. The type III may be empty, in which case the value is 000.

QAlgorithms.Papers.Independent.Intakenext100.P98.stabilizedTraceNorm

For any natural numbers n,mn,mn,m and any complex-linear map Φ\PhiΦ from complex n×nn\times nn×n matrices to complex m×mm\times mm×m matrices, this defines the real supremum sup⁡{∥(id⁡n⊗Φ)(X)∥tr:∥X∥tr≤1}\sup\{\|(\operatorname{id}_n\otimes\Phi)(X)\|_{\mathrm{tr}}:\|X\|_{\mathrm{tr}}\le1\}sup{∥(idn​⊗Φ)(X)∥tr​:∥X∥tr​≤1}, where XXX ranges over all complex matrices indexed by {0,…,n−1}×{0,…,n−1}\{0,\ldots,n-1\}\times\{0,\ldots,n-1\}{0,…,n−1}×{0,…,n−1}, without any positivity or Hermitian requirement. Here ∥M∥tr=Re⁡tr⁡M†M\|M\|_{\mathrm{tr}}=\operatorname{Re}\operatorname{tr}\sqrt{M^\dagger M}∥M∥tr​=RetrM†M​, with the square root defined by continuous functional calculus, and the amplified map is defined blockwise by ((id⁡n⊗Φ)(X))(a,u),(b,v)=(Φ(Xab))uv((\operatorname{id}_n\otimes\Phi)(X))_{(a,u),(b,v)}=(\Phi(X^{ab}))_{uv}((idn​⊗Φ)(X))(a,u),(b,v)​=(Φ(Xab))uv​, where Xijab=X(a,i),(b,j)X^{ab}_{ij}=X_{(a,i),(b,j)}Xijab​=X(a,i),(b,j)​, a,b,i,j<na,b,i,j<na,b,i,j<n, and u,v<mu,v<mu,v<m. Zero dimensions are included; if n=0n=0n=0 or m=0m=0m=0, the value is 000.

QAlgorithms.Papers.Independent.Intakenext100.P98.stinespringPairRepresents

For any natural numbers n,m,rn,m,rn,m,r, any complex-linear map Φ\PhiΦ from complex n×nn\times nn×n matrices to complex m×mm\times mm×m matrices, and any complex matrices A,BA,BA,B with rows indexed by pairs (i,e)(i,e)(i,e) satisfying 0≤i<m0\le i<m0≤i<m, 0≤e<r0\le e<r0≤e<r, and columns indexed by {0,…,n−1}\{0,\ldots,n-1\}{0,…,n−1}, this defines the assertion that, for every complex n×nn\times nn×n matrix XXX, (Φ(X))ij=∑e=0r−1(AXB†)(i,e),(j,e)(\Phi(X))_{ij}=\sum_{e=0}^{r-1}(AXB^\dagger)_{(i,e),(j,e)}(Φ(X))ij​=∑e=0r−1​(AXB†)(i,e),(j,e)​ for all 0≤i,j<m0\le i,j<m0≤i,j<m. Thus the operation on the right sums the diagonal blocks in the second, rrr-dimensional factor. All dimensions may be zero; when r=0r=0r=0, the sum is empty and the assertion requires Φ(X)=0\Phi(X)=0Φ(X)=0 for every XXX.

QAlgorithms.Papers.Independent.Intakenext100.P98.stinespringProducts

For any natural numbers n,mn,mn,m and any complex-linear map Φ:Mn(C)→Mm(C)\Phi:M_n(\mathbb C)\to M_m(\mathbb C)Φ:Mn​(C)→Mm​(C), this defines the set of real numbers

{∥A∥op∥B∥op:r∈N, r>0,A,B∈C(mr)×n,Φ(X)=Tr⁡r(AXB†) for every X∈Mn(C)}.\left\{\|A\|_{\mathrm{op}}\|B\|_{\mathrm{op}}: r\in\mathbb N,\ r>0,\quad A,B\in\mathbb C^{(m r)\times n},\quad \Phi(X)=\operatorname{Tr}_r(AXB^\dagger)\ \text{for every }X\in M_n(\mathbb C)\right\}.{∥A∥op​∥B∥op​:r∈N, r>0,A,B∈C(mr)×n,Φ(X)=Trr​(AXB†) for every X∈Mn​(C)}.

The rows of A,BA,BA,B are indexed by pairs (a,t)(a,t)(a,t), with 0≤a<m0\le a<m0≤a<m, 0≤t<r0\le t<r0≤t<r; †\dagger† denotes conjugate transpose, ∥⋅∥op\|\cdot\|_{\mathrm{op}}∥⋅∥op​ is the Euclidean operator norm, and (Tr⁡rY)ab=∑t=0r−1Y(a,t),(b,t)(\operatorname{Tr}_r Y)_{ab}=\sum_{t=0}^{r-1}Y_{(a,t),(b,t)}(Trr​Y)ab​=∑t=0r−1​Y(a,t),(b,t)​. No channel or positivity assumption is imposed on Φ\PhiΦ, and A,BA,BA,B need not coincide. The dimensions n,mn,mn,m may be zero, while rrr must be strictly positive.

QAlgorithms.Papers.Independent.Intakenext100.P98.stinespringSdpBound

For every natural number rrr, this definition returns the square matrix indexed by Fin(1)⊔Fin(r)\mathrm{Fin}(1)\sqcup\mathrm{Fin}(r)Fin(1)⊔Fin(r)

(1000r×r).\begin{pmatrix} 1&0\\ 0&0_{r\times r} \end{pmatrix}.(10​00r×r​​).

The value r=0r=0r=0 is included, in which case this is the 1×11\times11×1 matrix with entry 111.

QAlgorithms.Papers.Independent.Intakenext100.P98.stinespringSdpMap

For any natural numbers n,m,rn,m,rn,m,r, any complex matrix AAA with rows indexed by Fin(m)×Fin(r)\mathrm{Fin}(m)\times\mathrm{Fin}(r)Fin(m)×Fin(r) and columns indexed by Fin(n)\mathrm{Fin}(n)Fin(n), and any complex square matrix PPP indexed by the disjoint union Fin(n)⊔(Fin(m)×Fin(r))\mathrm{Fin}(n)\sqcup(\mathrm{Fin}(m)\times\mathrm{Fin}(r))Fin(n)⊔(Fin(m)×Fin(r)), let XXX and YYY be the two diagonal blocks of PPP, of sizes nnn and mrmrmr, respectively. This definition returns the matrix on Fin(1)⊔Fin(r)\mathrm{Fin}(1)\sqcup\mathrm{Fin}(r)Fin(1)⊔Fin(r) given by

(tr⁡X00Tr⁡m(Y−AXA†)),(Tr⁡mT)ij=∑a=0m−1T(a,i),(a,j),\begin{pmatrix} \operatorname{tr}X&0\\ 0&\operatorname{Tr}_{m}(Y-AXA^\dagger) \end{pmatrix}, \qquad (\operatorname{Tr}_{m}T)_{ij} =\sum_{a=0}^{m-1}T_{(a,i),(a,j)},(trX0​0Trm​(Y−AXA†)​),(Trm​T)ij​=a=0∑m−1​T(a,i),(a,j)​,

where A†A^\daggerA† is the conjugate transpose. The off-diagonal blocks of PPP are unused, and no Hermitian or positivity assumption is imposed. All three dimensions may be zero; the corresponding blocks are empty and empty sums and traces are zero.

QAlgorithms.Papers.Independent.Intakenext100.P98.stinespringSdpObjective

For any natural numbers n,m,rn,m,rn,m,r, including zero, and any complex matrix BBB with rows indexed by Fin(m)×Fin(r)\mathrm{Fin}(m)\times\mathrm{Fin}(r)Fin(m)×Fin(r) and columns indexed by Fin(n)\mathrm{Fin}(n)Fin(n), this definition returns the square matrix on Fin(n)⊔(Fin(m)×Fin(r))\mathrm{Fin}(n)\sqcup(\mathrm{Fin}(m)\times\mathrm{Fin}(r))Fin(n)⊔(Fin(m)×Fin(r))

(000BB†).\begin{pmatrix} 0&0\\ 0&BB^\dagger \end{pmatrix}.(00​0BB†​).

There are no further assumptions on BBB; when an indexing set is empty, its block is empty and matrix products use empty sums.

QAlgorithms.Papers.Independent.Intakenext100.P98.superAdjoint

For any natural numbers n,mn,mn,m and any complex-linear map Φ\PhiΦ from complex n×nn\times nn×n matrices to complex m×mm\times mm×m matrices, this defines a complex-linear map Ψ\PsiΨ in the reverse direction by (Ψ(Y))ij=∑a=0m−1∑b=0m−1(Φ(Eij))ab‾ Yab(\Psi(Y))_{ij}=\sum_{a=0}^{m-1}\sum_{b=0}^{m-1}\overline{(\Phi(E_{ij}))_{ab}}\,Y_{ab}(Ψ(Y))ij​=∑a=0m−1​∑b=0m−1​(Φ(Eij​))ab​​Yab​ for every complex m×mm\times mm×m matrix YYY and all 0≤i,j<n0\le i,j<n0≤i,j<n, where EijE_{ij}Eij​ has its sole nonzero entry, equal to 111, at (i,j)(i,j)(i,j). Zero dimensions are allowed: when m=0m=0m=0, every displayed sum is 000, and when n=0n=0n=0, the output matrix has no entries.

QAlgorithms.Papers.Independent.Intakenext100.P98.unitTensor

For any natural numbers m,rm,rm,r and any complex r×rr\times rr×r matrix ZZZ, this defines the matrix Im⊗ZI_m\otimes ZIm​⊗Z, indexed by pairs (a,i)(a,i)(a,i) with 0≤a<m0\le a<m0≤a<m and 0≤i<r0\le i<r0≤i<r. Its entries are (Im⊗Z)(a,i),(b,j)=δabZij(I_m\otimes Z)_{(a,i),(b,j)}=\delta_{ab}Z_{ij}(Im​⊗Z)(a,i),(b,j)​=δab​Zij​. Either dimension may be zero, in which case the resulting matrix has an empty index set.

QAlgorithms.Papers.Independent.Intakenext100.P98.watrousChoiMatrix

For any natural numbers n,mn,mn,m and any complex-linear map Φ\PhiΦ from complex n×nn\times nn×n matrices to complex m×mm\times mm×m matrices, this defines a matrix JΦJ_\PhiJΦ​ indexed by {0,…,m−1}×{0,…,n−1}\{0,\ldots,m-1\}\times\{0,\ldots,n-1\}{0,…,m−1}×{0,…,n−1} through (JΦ)(a,i),(b,j)=(Φ(Eij))ab(J_\Phi)_{(a,i),(b,j)}=(\Phi(E_{ij}))_{ab}(JΦ​)(a,i),(b,j)​=(Φ(Eij​))ab​, where EijE_{ij}Eij​ is the n×nn\times nn×n matrix with entry 111 at (i,j)(i,j)(i,j) and 000 elsewhere. Zero dimensions are allowed; if n=0n=0n=0 or m=0m=0m=0, this matrix has an empty index set.

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