Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Definitions for Post-selection technique for quantum channels with applications to quantum cryptography

Definition
QAlgorithms_Papers_Independent_Support50_P22

by Goku · Oct 23, 2026 · Mathlib 0df444a (Lean v4.33.1)

prove2quantum

The objects the formalized statements of Post-selection technique for quantum channels with applications to quantum cryptography (arXiv:0809.3019) are stated in. Each was checked against the paper by a blind read-back and a moderator.

  • tensorId: Apply a map to the first register. The two swaps explicitly convert to and from the accepted reference-first idTensorSuperop convention (PDF p.2).
  • FinitePurification: Normalized purification, system first and reference last (PDF p.4, footnote 12). The reduced state is specified, not the target postselection.
  • permuteInput: Source tensor permutation on operators: the vector action uses pi inverse, so conjugation reads matrix coordinates at x composed with pi (footnote 10).
  • pairPermutation: Simultaneous permutation of H and K copies (PDF pp.2,4).
  • pairedSymmetricSpace: The simultaneous-pair symmetric subspace, not the space of invariant mixed operators (PDF p.2 and footnote 10).
  • subspaceProjectorMatrix: Coordinate matrix of the actual orthogonal projector onto a subspace, used for the identities on PDF p.2 and the substitution in footnote 17.
  • factor: Exact symmetric-space dimension factor in Theorem 1, PDF p.2.
  • pairedCopiesEquiv: Explicit regrouping of paired copies, PDF p.2.
  • pairVector: Identify Fin(d*d) Haar coordinates with H tensor K, K a copy of H. The lexicographic equivalence is fixed independently of the map or input.
  • pairedPower: Regrouped n-fold tensor power of a bipartite operator (PDF p.2). This is the tensor power reindexed by pairedCopiesEquiv, evaluated coordinatewise.
  • densityTensorPower: Tensor power of a single-system operator, equation (1), PDF p.1.
  • hilbertSchmidtLaw: The Hilbert--Schmidt induced density law: push forward Haar pure states on H tensor a d-dimensional copy K by partial trace (PDF p.2, citation [5]). This is a constructed law, not an arbitrary mixing measure.
  • deFinettiState: The specific de Finetti state in equation (1), PDF p.1. The integrand is a continuous polynomial on the compact support of the induced probability law.
  • tauPair: Haar extension of the specified test state, PDF p.2. This is constructed from the integral; the projector identity is a conclusion below.
  • basisPurification: Canonical basis purification of the normalized projector onto S (PDF p.2). The reference is a coordinate copy of S identified by b; the formula is r^(-1/2) sum_i b_i tensor |i>. Its use below has r=g(n,d)>0.
  • postselect: Scalar-valued CP postselection functional applied on the last register. For an effect M this is (id tensor [X ↦ Tr(XM)]); its scalar codomain is literal complex numbers, not a surviving output register (Lemma 2, PDF p.2).
  • RangeSupported: Range support in a vector subspace, as in Lemma 2 (PDF p.2).
  • linearExtension: Matrix-unit complex-linear extension, used to construct the ideal protocol on all operators from its specified classical replacement rule (PDF p.3).
  • KeyOut: Finite variable-length classical output: length branch, Alice's key, Bob's key, and the complete public transcript (PDF p.3). Zero length allowed.
  • replaceKeys: Ideal key replacement S, PDF p.3. The length branch and transcript are preserved; keys are replaced by the identical uniform pair of that length. The rule discards coherences because this is a classical output operation.
  • keyReplacement: Complex-linear ideal output map S, constructed by matrix units (PDF p.3).
  • idealProtocol: F = S composed with E, constructed from E (PDF p.3).
  • publicSymmetrize: Uniform permutation average WITH a retained classical permutation label, PDF pp.3--4, footnote 16. Its probability is 1/n!, including 0!=1.
  • relabelPublic: Relabel the publicly communicated permutation p by p composed with pi; this expression pulls output coordinates back by pi inverse (footnote 16).
  • dualPairAction: Dual action U(g inverse)^T on W and the product action on V tensor W, PDF p.4, footnote 17. W uses the dual rather than the original representation.
  • invariantSpace: The fixed VECTOR subspace (V tensor dual V)^G (footnote 17, PDF p.4).
  • groupTau: The group test state is the normalized invariant-space projector, constructed from the representation (footnote 17). No Haar measure on a possibly noncompact group is used.
Definition code
import Mathlib
import Definitions.Def_QAlgorithms_Defs_Channel
import Definitions.Def_QAlgorithms_Defs_Measurement
import Definitions.Def_QAlgorithms_Defs_TensorPermutationRepresentations
import Definitions.Def_QAlgorithms_Defs_MatrixDistances
import Definitions.Def_QAlgorithms_Defs_QuditChannelFidelities
import Definitions.Def_QAlgorithms_Papers_Independent_Gap48_P20
import Definitions.Def_QAlgorithms_Papers_Independent_Information100_P69
import Definitions.Def_QAlgorithms_Papers_Independent_Dependency50_P18

/-!
Statements from Christandl--Koenig--Renner, arXiv:0809.3019v1.
All new constructions are local to this paper. Physical registers use fixed finite
complex coordinates. Trace norms have no factor 1/2. The two underspecified
operational hashing/rate claims are recorded in HARD.md, rather than weakened.
-/

noncomputable section

open scoped Matrix ComplexConjugate ComplexOrder Matrix.Norms.L2Operator
open MeasureTheory

namespace QAlgorithms.Papers.Independent.Support50.P22

attribute [local instance] Classical.decEq Classical.propDecidable

open QAlgorithms.Papers.Independent.Gap48.P20 (Copies)
open QAlgorithms.Papers.Independent.Information100.P69 (diamond)
open QAlgorithms.Papers.Independent.Dependency50.P18 (ClassicalOutput)

/-- Borel sigma algebra of the fixed finite matrix coordinates (PDF pp.1--2).
This makes the Hilbert--Schmidt pushforward a measure on actual matrices. -/
local instance postselectionMatrixMeasurableSpace (A : Type*) : MeasurableSpace (Matrix A A ℂ) := borel _

/-- The matrix measurable structure is its Borel structure (PDF pp.1--2). -/
local instance postselectionMatrixBorelSpace (A : Type*) : BorelSpace (Matrix A A ℂ) := ⟨rfl⟩

/-- Apply a map to the first register. The two swaps explicitly convert to and
from the accepted reference-first `idTensorSuperop` convention (PDF p.2). -/
def tensorId {A B R : Type*} (D : Superop A B) : Superop (A × R) (B × R) :=
  fun X p q => idTensorSuperop (R := R) D
    (fun a b => X (a.2,a.1) (b.2,b.1)) (p.2,p.1) (q.2,q.1)

/-- Normalized purification, system first and reference last (PDF p.4,
footnote 12). The reduced state is specified, not the target postselection. -/
def FinitePurification {A R : Type*} [Fintype A] [Fintype R]
    (psi : EuclideanSpace ℂ (A × R)) (rho : Matrix A A ℂ) : Prop :=
  IsState psi ∧ traceRight (pureDensity psi) = rho

/-- Source tensor permutation on operators: the vector action uses pi inverse,
so conjugation reads matrix coordinates at x composed with pi (footnote 10). -/
def permuteInput {n d : ℕ} (pi : Equiv.Perm (Fin n)) : Superop (Copies (Fin d) n) (Copies (Fin d) n) :=
  fun X a b => X (a ∘ pi) (b ∘ pi)

/-- Simultaneous permutation of H and K copies (PDF pp.2,4). -/
def pairPermutation (n d : ℕ) (pi : Equiv.Perm (Fin n)) :
    Matrix (Copies (Fin d) n × Copies (Fin d) n) (Copies (Fin d) n × Copies (Fin d) n) ℂ :=
  Matrix.kronecker (Schur.tensorPermutation d n pi) (Schur.tensorPermutation d n pi)

/-- The simultaneous-pair symmetric subspace, not the space of invariant
mixed operators (PDF p.2 and footnote 10). -/
def pairedSymmetricSpace (n d : ℕ) : Submodule ℂ (EuclideanSpace ℂ (Copies (Fin d) n × Copies (Fin d) n)) :=
  ⨅ pi : Equiv.Perm (Fin n), LinearMap.ker
    ((Matrix.toEuclideanCLM (𝕜 := ℂ) (n := Copies (Fin d) n × Copies (Fin d) n) (pairPermutation n d pi)).toLinearMap - LinearMap.id)

/-- Coordinate matrix of the actual orthogonal projector onto a subspace,
used for the identities on PDF p.2 and the substitution in footnote 17. -/
def subspaceProjectorMatrix {A : Type*} [Fintype A] [DecidableEq A]
    (S : Submodule ℂ (EuclideanSpace ℂ A)) : Matrix A A ℂ :=
  fun a b => S.starProjection (ket b) a

/-- Exact symmetric-space dimension factor in Theorem 1, PDF p.2. -/
def factor (n d : ℕ) : ℕ := Nat.choose (n + d ^ 2 - 1) n

/-- Explicit regrouping of paired copies, PDF p.2. -/
def pairedCopiesEquiv (n d : ℕ) : (Fin n → Fin d × Fin d) ≃ (Copies (Fin d) n × Copies (Fin d) n) where
  toFun x := (fun i => (x i).1, fun i => (x i).2)
  invFun x := fun i => (x.1 i,x.2 i)
  left_inv x := by funext i; exact Prod.eta (x i)
  right_inv x := by cases x; rfl

/-- Identify Fin(d*d) Haar coordinates with H tensor K, K a copy of H.
The lexicographic equivalence is fixed independently of the map or input. -/
def pairVector (d : ℕ) (psi : EuclideanSpace ℂ (Fin (d*d))) :
    EuclideanSpace ℂ (Fin d × Fin d) :=
  WithLp.toLp 2 (fun x => psi (finProdFinEquiv x))

/-- Regrouped n-fold tensor power of a bipartite operator (PDF p.2).
This is the tensor power reindexed by `pairedCopiesEquiv`, evaluated coordinatewise. -/
def pairedPower (n d : ℕ) (rho : Matrix (Fin d × Fin d) (Fin d × Fin d) ℂ) :
    Matrix (Copies (Fin d) n × Copies (Fin d) n) (Copies (Fin d) n × Copies (Fin d) n) ℂ :=
  fun a b => ∏ i, rho (a.1 i,a.2 i) (b.1 i,b.2 i)

/-- Tensor power of a single-system operator, equation (1), PDF p.1. -/
def densityTensorPower (n d : ℕ) (rho : Matrix (Fin d) (Fin d) ℂ) :
    Matrix (Copies (Fin d) n) (Copies (Fin d) n) ℂ :=
  fun a b => ∏ i, rho (a i) (b i)

/-- The Hilbert--Schmidt induced density law: push forward Haar pure states on
H tensor a d-dimensional copy K by partial trace (PDF p.2, citation [5]).
This is a constructed law, not an arbitrary mixing measure. -/
def hilbertSchmidtLaw (d : ℕ) (hd : 0 < d) : Measure (Matrix (Fin d) (Fin d) ℂ) :=
  (QuditFidelity.haarPureStateLaw (d*d) (Nat.mul_pos hd hd)).map
    (fun psi => traceRight (pureDensity (pairVector d psi)))

/-- The specific de Finetti state in equation (1), PDF p.1. The integrand is a
continuous polynomial on the compact support of the induced probability law. -/
def deFinettiState (n d : ℕ) (hd : 0 < d) : Matrix (Copies (Fin d) n) (Copies (Fin d) n) ℂ :=
  ∫ rho, densityTensorPower n d rho ∂hilbertSchmidtLaw d hd

/-- Haar extension of the specified test state, PDF p.2. This is constructed
from the integral; the projector identity is a conclusion below. -/
def tauPair (n d : ℕ) (hd : 0 < d) :
    Matrix (Copies (Fin d) n × Copies (Fin d) n) (Copies (Fin d) n × Copies (Fin d) n) ℂ :=
  ∫ psi, pairedPower n d (pureDensity (pairVector d psi))
    ∂QuditFidelity.haarPureStateLaw (d*d) (Nat.mul_pos hd hd)

/-- Canonical basis purification of the normalized projector onto S (PDF p.2).
The reference is a coordinate copy of S identified by b; the formula is
r^(-1/2) sum_i b_i tensor |i>. Its use below has r=g(n,d)>0. -/
def basisPurification {A : Type*} [Fintype A]
    (S : Submodule ℂ (EuclideanSpace ℂ A))
    (b : OrthonormalBasis (Fin (Module.finrank ℂ S)) ℂ S) :
    EuclideanSpace ℂ (A × Fin (Module.finrank ℂ S)) :=
  WithLp.toLp 2 (fun x => ((Real.sqrt (Module.finrank ℂ S) : ℂ)⁻¹) * (b x.2).val x.1)

/-- Scalar-valued CP postselection functional applied on the last register.
For an effect M this is (id tensor [X ↦ Tr(XM)]); its scalar codomain is
literal complex numbers, not a surviving output register (Lemma 2, PDF p.2). -/
def postselect {A N : Type*} [Fintype N] (M : Matrix N N ℂ)
    (X : Matrix (A × N) (A × N) ℂ) : Matrix A A ℂ :=
  fun a b => ∑ i, ∑ j, X (a,i) (b,j) * M j i

/-- Range support in a vector subspace, as in Lemma 2 (PDF p.2). -/
def RangeSupported {A : Type*} [Fintype A] (rho : Matrix A A ℂ)
    (S : Submodule ℂ (EuclideanSpace ℂ A)) : Prop :=
  ∀ psi, act rho psi ∈ S

/-- Matrix-unit complex-linear extension, used to construct the ideal protocol
on all operators from its specified classical replacement rule (PDF p.3). -/
def linearExtension {A B : Type*} [Fintype A] [DecidableEq A]
    (f : Superop A B) : Matrix A A ℂ →ₗ[ℂ] Matrix B B ℂ where
  toFun X a b := ∑ i, ∑ j, X i j * f (Matrix.single i j 1) a b
  map_add' X Y := by
    ext a b
    change (∑ i, ∑ j, (X i j + Y i j) * f (Matrix.single i j 1) a b) =
      (∑ i, ∑ j, X i j * f (Matrix.single i j 1) a b) +
        (∑ i, ∑ j, Y i j * f (Matrix.single i j 1) a b)
    simp only [add_mul, Finset.sum_add_distrib]
  map_smul' c X := by
    ext a b
    change (∑ i, ∑ j, (c * X i j) * f (Matrix.single i j 1) a b) =
      c * (∑ i, ∑ j, X i j * f (Matrix.single i j 1) a b)
    simp only [Finset.mul_sum, mul_assoc]

/-- Finite variable-length classical output: length branch, Alice's key,
Bob's key, and the complete public transcript (PDF p.3). Zero length allowed. -/
abbrev KeyOut (L C : Type*) (len : L → ℕ) :=
  (l : L) × ((Qubits (len l) × Qubits (len l)) × C)

/-- Ideal key replacement S, PDF p.3. The length branch and transcript are
preserved; keys are replaced by the identical uniform pair of that length.
The rule discards coherences because this is a classical output operation. -/
def replaceKeys {L C : Type*} [Fintype L] [Fintype C] (len : L → ℕ) :
    Superop (KeyOut L C len) (KeyOut L C len) :=
  fun X a b => if a = b ∧ a.2.1.1 = a.2.1.2 then
    ((2 : ℂ) ^ len a.1)⁻¹ * ∑ s : Qubits (len a.1), ∑ t : Qubits (len a.1),
      X ⟨a.1,((s,t),a.2.2)⟩ ⟨a.1,((s,t),a.2.2)⟩ else 0

/-- Complex-linear ideal output map S, constructed by matrix units (PDF p.3). -/
def keyReplacement {L C : Type*} [Fintype L] [Fintype C] (len : L → ℕ) :
    Matrix (KeyOut L C len) (KeyOut L C len) ℂ →ₗ[ℂ]
      Matrix (KeyOut L C len) (KeyOut L C len) ℂ :=
  linearExtension (replaceKeys len)

/-- F = S composed with E, constructed from E (PDF p.3). -/
def idealProtocol {A L C : Type*} [Fintype A] [Fintype L] [Fintype C]
    (len : L → ℕ)
    (E : Matrix A A ℂ →ₗ[ℂ] Matrix (KeyOut L C len) (KeyOut L C len) ℂ) :=
  (keyReplacement len).comp E

/-- Uniform permutation average WITH a retained classical permutation label,
PDF pp.3--4, footnote 16. Its probability is 1/n!, including 0!=1. -/
def publicSymmetrize {n d : ℕ} {B : Type*}
    (E : Superop (Copies (Fin d) n) B) :
    Superop (Copies (Fin d) n) (Equiv.Perm (Fin n) × B) :=
  fun X a b => if a.1 = b.1 then
    ((Nat.factorial n : ℂ)⁻¹) * E (permuteInput a.1 X) a.2 b.2 else 0

/-- Relabel the publicly communicated permutation p by p composed with pi;
this expression pulls output coordinates back by pi inverse (footnote 16). -/
def relabelPublic {n : ℕ} {B : Type*} (pi : Equiv.Perm (Fin n)) :
    Superop (Equiv.Perm (Fin n) × B) (Equiv.Perm (Fin n) × B) :=
  fun X a b => X (a.1 * pi.symm,a.2) (b.1 * pi.symm,b.2)

/-- Dual action U(g inverse)^T on W and the product action on V tensor W,
PDF p.4, footnote 17. W uses the dual rather than the original representation. -/
def dualPairAction {G : Type*} [Group G] {v : ℕ}
    (U : G →* Matrix.unitaryGroup (Fin v) ℂ) (g : G) :
    Matrix (Fin v × Fin v) (Fin v × Fin v) ℂ :=
  Matrix.kronecker (U g : Matrix (Fin v) (Fin v) ℂ)
    (U g⁻¹ : Matrix (Fin v) (Fin v) ℂ).transpose

/-- The fixed VECTOR subspace (V tensor dual V)^G (footnote 17, PDF p.4). -/
def invariantSpace {G : Type*} [Group G] {v : ℕ}
    (U : G →* Matrix.unitaryGroup (Fin v) ℂ) :
    Submodule ℂ (EuclideanSpace ℂ (Fin v × Fin v)) :=
  ⨅ g : G, LinearMap.ker ((Matrix.toEuclideanCLM (𝕜 := ℂ) (n := Fin v × Fin v) (dualPairAction U g)).toLinearMap - LinearMap.id)

/-- The group test state is the normalized invariant-space projector,
constructed from the representation (footnote 17). No Haar measure on a
possibly noncompact group is used. -/
def groupTau {G : Type*} [Group G] {v : ℕ}
    (U : G →* Matrix.unitaryGroup (Fin v) ℂ) :
    Matrix (Fin v × Fin v) (Fin v × Fin v) ℂ :=
  ((Module.finrank ℂ (invariantSpace U) : ℂ)⁻¹) • subspaceProjectorMatrix (invariantSpace U)

/-- Small semantic checks. These establish conventions rather than prove any
of the paper's results. -/

example (d : ℕ) : factor 0 d = 1 := by simp [factor]

example {A R : Type*} (X : Matrix (A × R) (A × R) ℂ) :
    tensorId (fun Y : Matrix A A ℂ => Y) X = X := by ext a b; rfl

example {n d : ℕ} (X : Matrix (Copies (Fin d) n) (Copies (Fin d) n) ℂ) :
    permuteInput (1 : Equiv.Perm (Fin n)) X = X := by rfl

example (d : ℕ) (rho : Matrix (Fin d) (Fin d) ℂ) (a b : Copies (Fin d) 0) :
    densityTensorPower 0 d rho a b = 1 := by simp [densityTensorPower]

example {A N : Type*} [Fintype N] [DecidableEq N]
    (X : Matrix (A × N) (A × N) ℂ) :
    postselect (1 : Matrix N N ℂ) X = traceRight X := by
  ext a b
  simp [postselect, traceRight, Matrix.one_apply]

example {L C : Type*} [Fintype L] [Fintype C] (len : L → ℕ) :
    keyReplacement len (0 : Matrix (KeyOut L C len) (KeyOut L C len) ℂ) = 0 :=
  map_zero _

example {G : Type*} [Group G] {v : ℕ} (U : G →* Matrix.unitaryGroup (Fin v) ℂ) :
    dualPairAction U 1 = 1 := by
  simp [dualPairAction]

example (n : ℕ) : factor n 1 = 1 := by simp [factor]

end QAlgorithms.Papers.Independent.Support50.P22
Source
Post-selection technique for quantum channels with applications to quantum cryptography, arXiv:0809.3019v1, https://arxiv.org/abs/0809.3019v1
Read-back

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

QAlgorithms.Papers.Independent.Support50.P22.FinitePurification

For finite types A,RA,RA,R, a vector ψ∈CA×R\psi\in\mathbb C^{A\times R}ψ∈CA×R, and a complex matrix ρ\rhoρ indexed by A×AA\times AA×A, this predicate means that ψ\psiψ has Euclidean norm 111 and

ρa,b=∑r∈Rψ(a,r) ψ(b,r)‾(a,b∈A).\rho_{a,b}=\sum_{r\in R}\psi(a,r)\,\overline{\psi(b,r)} \qquad(a,b\in A).ρa,b​=r∈R∑​ψ(a,r)ψ(b,r)​(a,b∈A).

In other words, tracing out the RRR-index of the matrix with entries ψ(a,r)ψ(b,s)‾\psi(a,r)\overline{\psi(b,s)}ψ(a,r)ψ(b,s)​ gives ρ\rhoρ. Neither type is assumed nonempty; if either is empty, the norm requirement cannot hold.

QAlgorithms.Papers.Independent.Support50.P22.KeyOut

For arbitrary types L,CL,CL,C and a function len:L→N\mathrm{len}:L\to\mathbb Nlen:L→N, this abbreviation denotes the dependent indexing type

⨆ℓ∈L({0,1}len(ℓ)×{0,1}len(ℓ)×C).\bigsqcup_{\ell\in L} \bigl(\{0,1\}^{\mathrm{len}(\ell)} \times\{0,1\}^{\mathrm{len}(\ell)} \times C\bigr).ℓ∈L⨆​({0,1}len(ℓ)×{0,1}len(ℓ)×C).

An element consists of a label ℓ\ellℓ, two Boolean strings of the same length len(ℓ)\mathrm{len}(\ell)len(ℓ), and an element of CCC. No finiteness or nonemptiness assumptions are imposed. Length zero is allowed and gives a unique empty string in each of the two string positions.

QAlgorithms.Papers.Independent.Support50.P22.RangeSupported

For any finite type AAA, complex matrix ρ\rhoρ on AAA, and complex linear subspace S⊆CAS\subseteq\mathbb C^AS⊆CA, this predicate means that matrix multiplication by ρ\rhoρ sends every Euclidean vector into SSS:

∀ψ∈CA,(a↦∑b∈Aρa,bψb)∈S.\forall\psi\in\mathbb C^A,\qquad \left(a\mapsto\sum_{b\in A}\rho_{a,b}\psi_b\right)\in S.∀ψ∈CA,(a↦b∈A∑​ρa,b​ψb​)∈S.

No positivity or trace condition is imposed on ρ\rhoρ. In particular, the zero matrix satisfies this predicate for every SSS, and the predicate holds automatically when SSS is the whole ambient space; empty AAA is allowed.

QAlgorithms.Papers.Independent.Support50.P22.basisPurification

For any finite type AAA, any complex linear subspace S⊆CAS\subseteq\mathbb C^AS⊆CA with its Euclidean inner product, and any orthonormal basis (bj)j=0r−1(b_j)_{j=0}^{r-1}(bj​)j=0r−1​ of SSS, where r=dim⁡CSr=\dim_{\mathbb C}Sr=dimC​S, this definition gives the Euclidean vector on A×{0,…,r−1}A\times\{0,\ldots,r-1\}A×{0,…,r−1} with coordinates

Ψa,j=(r)−1(bj)a.\Psi_{a,j}=(\sqrt r)^{-1}(b_j)_a.Ψa,j​=(r​)−1(bj​)a​.

The basis vector bjb_jbj​ is viewed here as a vector in the ambient space CA\mathbb C^ACA. No assumption r>0r>0r>0 is required: when S={0}S=\{0\}S={0}, the second indexing set is empty, the resulting vector is the zero vector in a zero-dimensional space, and the scalar inverse uses the convention 0−1=00^{-1}=00−1=0. The finite type AAA may itself be empty.

QAlgorithms.Papers.Independent.Support50.P22.deFinettiState

For natural numbers n,dn,dn,d and a hypothesis d>0d>0d>0, put Hn={0,…,d−1}nH_n=\{0,\ldots,d-1\}^nHn​={0,…,d−1}n. This definition gives the complex matrix

Ωn,d=∫ρ⊗n dμd(ρ),(ρ⊗n)x,y=∏i=0n−1ρxi,yi(x,y∈Hn).\Omega_{n,d}=\int \rho^{\otimes n}\,d\mu_d(\rho), \qquad (\rho^{\otimes n})_{x,y}=\prod_{i=0}^{n-1}\rho_{x_i,y_i} \quad(x,y\in H_n).Ωn,d​=∫ρ⊗ndμd​(ρ),(ρ⊗n)x,y​=i=0∏n−1​ρxi​,yi​​(x,y∈Hn​).

Here μd\mu_dμd​ is defined by sampling a unitary UUU from normalized Haar measure on the unitary group in dimension d2d^2d2, taking ψ=Ue0\psi=Ue_0ψ=Ue0​, reshaping its coordinates into va,b=ψindex⁡(a,b)v_{a,b}=\psi_{\operatorname{index}(a,b)}va,b​=ψindex(a,b)​ using the specified standard bijection {0,…,d−1}2≅{0,…,d2−1}\{0,\ldots,d-1\}^2\cong\{0,\ldots,d^2-1\}{0,…,d−1}2≅{0,…,d2−1}, and setting ρa,a′=∑bva,bva′,b‾\rho_{a,a'}=\sum_b v_{a,b}\overline{v_{a',b}}ρa,a′​=∑b​va,b​va′,b​​. The integral is the matrix-valued Bochner integral. The definition includes n=0n=0n=0, when H0H_0H0​ is a singleton and every tensor-power entry is the empty product 111.

QAlgorithms.Papers.Independent.Support50.P22.densityTensorPower

For any natural numbers n,dn,dn,d and any complex d×dd\times dd×d matrix ρ\rhoρ, this defines a matrix indexed by length-nnn strings over {0,…,d−1}\{0,\ldots,d-1\}{0,…,d−1} with entries

(densityTensorPower⁡n,dρ)a,b=∏i=0n−1ρa(i),b(i).(\operatorname{densityTensorPower}_{n,d}\rho)_{a,b} = \prod_{i=0}^{n-1}\rho_{a(i),b(i)}.(densityTensorPowern,d​ρ)a,b​=i=0∏n−1​ρa(i),b(i)​.

No density-matrix assumption is imposed on ρ\rhoρ. For n=0n=0n=0, the empty product gives the one-dimensional identity matrix, including when d=0d=0d=0; for d=0<nd=0<nd=0<n, the matrix index set is empty.

QAlgorithms.Papers.Independent.Support50.P22.dualPairAction

For any group GGG, natural number vvv, group homomorphism U:G→U(v)U:G\to\mathrm U(v)U:G→U(v), and element g∈Gg\in Gg∈G, the definition gives the matrix

DU(g)=U(g)⊗U(g−1)TD_U(g)=U(g)\otimes U(g^{-1})^{\mathsf T}DU​(g)=U(g)⊗U(g−1)T

on {0,…,v−1}×{0,…,v−1}\{0,\ldots,v-1\}\times\{0,\ldots,v-1\}{0,…,v−1}×{0,…,v−1}. Thus its entry at (i,j),(k,l)(i,j),(k,l)(i,j),(k,l) is U(g)i,kU(g−1)l,jU(g)_{i,k}U(g^{-1})_{l,j}U(g)i,k​U(g−1)l,j​. The transpose in this expression is the ordinary transpose. No topology or finiteness of GGG is required, and v=0v=0v=0 is allowed.

QAlgorithms.Papers.Independent.Support50.P22.factor

For any natural numbers n,dn,dn,d, this defines

factor⁡(n,d)=(max⁡(n+d2−1, 0)n),\operatorname{factor}(n,d)=\binom{\max(n+d^2-1,\,0)}{n},factor(n,d)=(nmax(n+d2−1,0)​),

where the subtraction in the upper argument is natural-number subtraction, truncated at zero. In particular, the value is 111 whenever n=0n=0n=0, including n=d=0n=d=0n=d=0, and is 000 whenever d=0<nd=0<nd=0<n.

QAlgorithms.Papers.Independent.Support50.P22.groupTau

For any group GGG, natural number vvv, and group homomorphism U:G→U(v)U:G\to\mathrm U(v)U:G→U(v), let

SU={ξ:(U(g)⊗U(g−1)T)ξ=ξ for every g∈G},m=dim⁡CSU.S_U=\{\xi:(U(g)\otimes U(g^{-1})^{\mathsf T})\xi=\xi \text{ for every }g\in G\}, \qquad m=\dim_{\mathbb C}S_U.SU​={ξ:(U(g)⊗U(g−1)T)ξ=ξ for every g∈G},m=dimC​SU​.

The definition gives the matrix τU=m−1PSU\tau_U=m^{-1}P_{S_U}τU​=m−1PSU​​, where PSUP_{S_U}PSU​​ is the orthogonal projection onto SUS_USU​: its (a,b)(a,b)(a,b)-entry is the aaa-coordinate of the orthogonal projection of the standard basis vector ebe_beb​. The reciprocal is taken in C\mathbb CC, with 0−1=00^{-1}=00−1=0; consequently the definition gives the zero matrix if m=0m=0m=0, including when v=0v=0v=0. No topology or finiteness of GGG is assumed.

QAlgorithms.Papers.Independent.Support50.P22.hilbertSchmidtLaw

For a natural number ddd with d>0d>0d>0, this defines a measure on complex d×dd\times dd×d matrices as follows: take normalized Haar measure on the unitary group of dimension d2d^2d2, send a unitary UUU to the vector Ue0Ue_0Ue0​, reindex that vector by pairs using the standard bijection ed:{0,…,d−1}2→{0,…,d2−1}e_d:\{0,\ldots,d-1\}^2\to\{0,\ldots,d^2-1\}ed​:{0,…,d−1}2→{0,…,d2−1}, form its rank-one matrix, and trace out the second pair coordinate. Equivalently, it is the pushforward of that Haar measure under U↦σUU\mapsto\sigma_UU↦σU​, where

(σU)i,j=∑k=0d−1Ued(i,k),0 Ued(j,k),0‾.(\sigma_U)_{i,j} = \sum_{k=0}^{d-1} U_{e_d(i,k),0}\, \overline{U_{e_d(j,k),0}}.(σU​)i,j​=k=0∑d−1​Ued​(i,k),0​Ued​(j,k),0​​.

The positivity hypothesis on ddd ensures that the coordinate 000 used in e0e_0e0​ exists.

QAlgorithms.Papers.Independent.Support50.P22.idealProtocol

For arbitrary finite types A,L,CA,L,CA,L,C, a function ℓ:L→N\ell:L\to\mathbb Nℓ:L→N, and a complex linear map EEE from matrices on AAA to matrices indexed by

Z={(l,((s,t),c)):l∈L, s,t∈{0,1}ℓ(l), c∈C},Z=\{(l,((s,t),c)):l\in L,\ s,t\in\{0,1\}^{\ell(l)},\ c\in C\},Z={(l,((s,t),c)):l∈L, s,t∈{0,1}ℓ(l), c∈C},

the definition gives the complex linear map Eideal=Rℓ∘EE_{\mathrm{ideal}}=\mathcal R_\ell\circ EEideal​=Rℓ​∘E, where Rℓ\mathcal R_\ellRℓ​ replaces the two key registers by equal uniformly distributed strings and discards all off-diagonal entries. Explicitly, for every input matrix XXX and z=(l,((s,t),c)),z′∈Zz=(l,((s,t),c)),z'\in Zz=(l,((s,t),c)),z′∈Z,

Eideal(X)z,z′={2−ℓ(l)∑u,v∈{0,1}ℓ(l)E(X)(l,((u,v),c)),(l,((u,v),c)),z=z′ and s=t,0,otherwise.E_{\mathrm{ideal}}(X)_{z,z'}= \begin{cases} 2^{-\ell(l)} \displaystyle\sum_{u,v\in\{0,1\}^{\ell(l)}} E(X)_{(l,((u,v),c)),(l,((u,v),c))}, &z=z'\text{ and }s=t,\\ 0,&\text{otherwise}. \end{cases}Eideal​(X)z,z′​=⎩⎨⎧​2−ℓ(l)u,v∈{0,1}ℓ(l)∑​E(X)(l,((u,v),c)),(l,((u,v),c))​,0,​z=z′ and s=t,otherwise.​

There is no positivity, trace-preservation, or nonemptiness assumption here. Length zero is allowed and has one empty key string.

QAlgorithms.Papers.Independent.Support50.P22.invariantSpace

For any group GGG, natural number vvv, and group homomorphism U:G→U(v)U:G\to\mathrm U(v)U:G→U(v), the definition gives the complex linear subspace

SU={ξ∈C{0,…,v−1}2:(U(g)⊗U(g−1)T)ξ=ξ for every g∈G}.S_U= \left\{\xi\in\mathbb C^{\{0,\ldots,v-1\}^2}: \bigl(U(g)\otimes U(g^{-1})^{\mathsf T}\bigr)\xi=\xi \text{ for every }g\in G\right\}.SU​={ξ∈C{0,…,v−1}2:(U(g)⊗U(g−1)T)ξ=ξ for every g∈G}.

Equivalently, it is the intersection, over all ggg, of the kernels of the linear maps U(g)⊗U(g−1)T−IU(g)\otimes U(g^{-1})^{\mathsf T}-IU(g)⊗U(g−1)T−I, acting by matrix-vector multiplication. No topology or finiteness of GGG is assumed; v=0v=0v=0 gives the zero-dimensional subspace.

QAlgorithms.Papers.Independent.Support50.P22.keyReplacement

For finite types L,CL,CL,C and a function len:L→N\mathrm{len}:L\to\mathbb Nlen:L→N, put K=⨆ℓ∈L({0,1}len(ℓ)×{0,1}len(ℓ)×C)K=\bigsqcup_{\ell\in L}(\{0,1\}^{\mathrm{len}(\ell)}\times\{0,1\}^{\mathrm{len}(\ell)}\times C)K=⨆ℓ∈L​({0,1}len(ℓ)×{0,1}len(ℓ)×C). This definition constructs the complex-linear map R\mathcal RR on matrices on KKK by the matrix-unit extension R(X)a,b=∑i,j∈KXi,jreplaceKeys⁡(Ei,j)a,b\mathcal R(X)_{a,b}=\sum_{i,j\in K}X_{i,j}\operatorname{replaceKeys}(E_{i,j})_{a,b}R(X)a,b​=∑i,j∈K​Xi,j​replaceKeys(Ei,j​)a,b​, where Ei,jE_{i,j}Ei,j​ is the matrix unit. Explicitly, for a=(ℓ,(u,v),c)a=(\ell,(u,v),c)a=(ℓ,(u,v),c),

R(X)a,b={2−len(ℓ)∑s,t∈{0,1}len(ℓ)X(ℓ,(s,t),c),(ℓ,(s,t),c),a=b and u=v,0,otherwise.\mathcal R(X)_{a,b} = \begin{cases} 2^{-\mathrm{len}(\ell)} \displaystyle\sum_{s,t\in\{0,1\}^{\mathrm{len}(\ell)}} X_{(\ell,(s,t),c),(\ell,(s,t),c)}, & a=b\text{ and }u=v,\\ 0,&\text{otherwise}. \end{cases}R(X)a,b​=⎩⎨⎧​2−len(ℓ)s,t∈{0,1}len(ℓ)∑​X(ℓ,(s,t),c),(ℓ,(s,t),c)​,0,​a=b and u=v,otherwise.​

This applies to arbitrary complex matrices, includes zero string lengths, and allows empty LLL or CCC.

QAlgorithms.Papers.Independent.Support50.P22.linearExtension

For arbitrary types A,BA,BA,B, with AAA finite and equipped with decidable equality, and any function fff from complex matrices on AAA to complex matrices on BBB, this definition constructs the complex-linear map f~\widetilde ff​ given by

f~(X)a,b=∑i∈A∑j∈AXi,j f(Ei,j)a,b,\widetilde f(X)_{a,b} = \sum_{i\in A}\sum_{j\in A}X_{i,j}\,f(E_{i,j})_{a,b},f​(X)a,b​=i∈A∑​j∈A∑​Xi,j​f(Ei,j​)a,b​,

where Ei,jE_{i,j}Ei,j​ has entry 111 at (i,j)(i,j)(i,j) and 000 elsewhere. No linearity assumption is imposed on fff; the construction uses its values on these matrix units. The type BBB need not be finite. If AAA is empty, this map is identically zero.

QAlgorithms.Papers.Independent.Support50.P22.pairPermutation

For natural numbers n,dn,dn,d, including zero, and a permutation π\piπ of {0,…,n−1}\{0,\ldots,n-1\}{0,…,n−1}, this defines a complex matrix indexed by pairs of length-nnn strings over {0,…,d−1}\{0,\ldots,d-1\}{0,…,d−1}. For such strings a,b,c,ea,b,c,ea,b,c,e, its entries are

P(a,b),(c,e)={1,a=c∘π−1 and b=e∘π−1,0,otherwise.P_{(a,b),(c,e)} = \begin{cases} 1,&a=c\circ\pi^{-1}\ \text{and}\ b=e\circ\pi^{-1},\\ 0,&\text{otherwise}. \end{cases}P(a,b),(c,e)​={1,0,​a=c∘π−1 and b=e∘π−1,otherwise.​

This is the Kronecker product of the two matrices that perform the indicated coordinate permutation on each string separately. For n=0n=0n=0, it is the one-dimensional identity matrix.

QAlgorithms.Papers.Independent.Support50.P22.pairVector

For any natural number ddd and any Euclidean vector ψ∈C{0,…,d2−1}\psi\in\mathbb C^{\{0,\ldots,d^2-1\}}ψ∈C{0,…,d2−1}, this defines a vector indexed by pairs i,j∈{0,…,d−1}i,j\in\{0,\ldots,d-1\}i,j∈{0,…,d−1} by reindexing its coordinates:

(pairVector⁡dψ)(i,j)=ψ(ed(i,j)),(\operatorname{pairVector}_d\psi)(i,j)=\psi(e_d(i,j)),(pairVectord​ψ)(i,j)=ψ(ed​(i,j)),

where ed:{0,…,d−1}×{0,…,d−1}→{0,…,d2−1}e_d:\{0,\ldots,d-1\}\times\{0,\ldots,d-1\}\to\{0,\ldots,d^2-1\}ed​:{0,…,d−1}×{0,…,d−1}→{0,…,d2−1} is the standard finite-product indexing bijection. There is no normalization requirement on ψ\psiψ. For d=0d=0d=0, both vector spaces are zero-dimensional.

QAlgorithms.Papers.Independent.Support50.P22.pairedCopiesEquiv

For any natural numbers n,dn,dn,d, including zero, this defines a bijection between length-nnn strings of pairs from {0,…,d−1}\{0,\ldots,d-1\}{0,…,d−1} and pairs of length-nnn strings from that set. It sends

x⟼(i↦(x(i))1, i↦(x(i))2),x\longmapsto\bigl(i\mapsto(x(i))_1,\ i\mapsto(x(i))_2\bigr),x⟼(i↦(x(i))1​, i↦(x(i))2​),

and its inverse sends (a,b)(a,b)(a,b) to i↦(a(i),b(i))i\mapsto(a(i),b(i))i↦(a(i),b(i)). These maps are mutually inverse. For n=0n=0n=0, both sides are singletons; for d=0<nd=0<nd=0<n, both sides are empty.

QAlgorithms.Papers.Independent.Support50.P22.pairedPower

For any natural numbers n,dn,dn,d and any complex matrix ρ\rhoρ indexed by pairs from {0,…,d−1}\{0,\ldots,d-1\}{0,…,d−1}, this defines a matrix indexed by pairs of length-nnn strings through

(pairedPower⁡n,dρ)(a,b),(c,e)=∏i=0n−1ρ(a(i),b(i)),(c(i),e(i)).(\operatorname{pairedPower}_{n,d}\rho)_{(a,b),(c,e)} = \prod_{i=0}^{n-1}\rho_{(a(i),b(i)),(c(i),e(i))}.(pairedPowern,d​ρ)(a,b),(c,e)​=i=0∏n−1​ρ(a(i),b(i)),(c(i),e(i))​.

The matrix ρ\rhoρ is arbitrary, with no positivity or trace requirement. For n=0n=0n=0, the empty product is 111, so the resulting matrix is the one-dimensional identity; for d=0<nd=0<nd=0<n, its index set is empty.

QAlgorithms.Papers.Independent.Support50.P22.pairedSymmetricSpace

For natural numbers n,dn,dn,d, including zero, this defines the complex vector subspace of Euclidean vectors indexed by pairs of length-nnn strings over {0,…,d−1}\{0,\ldots,d-1\}{0,…,d−1} that are fixed by every simultaneous coordinate permutation of the two strings. Explicitly, it consists of all vectors ψ\psiψ satisfying

ψ(a∘π,b∘π)=ψ(a,b)\psi(a\circ\pi,b\circ\pi)=\psi(a,b)ψ(a∘π,b∘π)=ψ(a,b)

for every permutation π\piπ of the nnn coordinates and every pair of strings a,ba,ba,b. It is the intersection of the kernels of the corresponding permutation operators minus the identity. For n=0n=0n=0, this is the whole one-dimensional space; for d=0<nd=0<nd=0<n, the ambient space is zero-dimensional.

QAlgorithms.Papers.Independent.Support50.P22.permuteInput

For natural numbers n,dn,dn,d, including zero, and a permutation π\piπ of {0,…,n−1}\{0,\ldots,n-1\}{0,…,n−1}, this defines a function on complex matrices indexed by strings a:{0,…,n−1}→{0,…,d−1}a:\{0,\ldots,n-1\}\to\{0,\ldots,d-1\}a:{0,…,n−1}→{0,…,d−1}. Its action is

(permuteInput⁡πX)a,b=Xa∘π,  b∘π,(a∘π)(i)=a(π(i)).(\operatorname{permuteInput}_{\pi}X)_{a,b} = X_{a\circ\pi,\;b\circ\pi}, \qquad (a\circ\pi)(i)=a(\pi(i)).(permuteInputπ​X)a,b​=Xa∘π,b∘π​,(a∘π)(i)=a(π(i)).

When n=0n=0n=0, there is one empty string and this function leaves its single matrix entry unchanged; when d=0<nd=0<nd=0<n, the matrix index set is empty.

QAlgorithms.Papers.Independent.Support50.P22.postselect

For arbitrary types A,NA,NA,N with NNN finite, any complex matrix MMM indexed by N×NN\times NN×N, and any complex matrix XXX indexed by (A×N)×(A×N)(A\times N)\times(A\times N)(A×N)×(A×N), this definition gives the matrix on AAA with entries

postselect⁡M(X)a,b=∑i∈N∑j∈NX(a,i),(b,j)Mj,i.\operatorname{postselect}_M(X)_{a,b} = \sum_{i\in N}\sum_{j\in N}X_{(a,i),(b,j)}M_{j,i}.postselectM​(X)a,b​=i∈N∑​j∈N∑​X(a,i),(b,j)​Mj,i​.

There are no positivity, normalization, or effect assumptions on either matrix, and AAA need not be finite. If NNN is empty, the result is the zero matrix.

QAlgorithms.Papers.Independent.Support50.P22.publicSymmetrize

For any natural numbers n,dn,dn,d, arbitrary type BBB, and arbitrary function EEE from complex matrices on H={0,…,d−1}{0,…,n−1}H=\{0,\ldots,d-1\}^{\{0,\ldots,n-1\}}H={0,…,d−1}{0,…,n−1} to complex matrices on BBB, the definition produces a function from matrices on HHH to matrices on Sym⁡(n)×B\operatorname{Sym}(n)\times BSym(n)×B. For every input matrix XXX, permutations σ,τ\sigma,\tauσ,τ, and u,v∈Bu,v\in Bu,v∈B, it is

SE(X)(σ,u),(τ,v)={1n! E(PσX)u,v,σ=τ,0,σ≠τ,(PσX)x,y=Xx∘σ,y∘σ.\mathcal S_E(X)_{(\sigma,u),(\tau,v)} = \begin{cases} \dfrac1{n!}\,E(P_\sigma X)_{u,v},&\sigma=\tau,\\ 0,&\sigma\ne\tau, \end{cases} \qquad (P_\sigma X)_{x,y}=X_{x\circ\sigma,y\circ\sigma}.SE​(X)(σ,u),(τ,v)​=⎩⎨⎧​n!1​E(Pσ​X)u,v​,0,​σ=τ,σ=τ,​(Pσ​X)x,y​=Xx∘σ,y∘σ​.

No linearity, channel property, or finiteness of BBB is assumed. Both n=0n=0n=0 and d=0d=0d=0 are allowed; for n=0n=0n=0 there is one position permutation and 0!=10!=10!=1.

QAlgorithms.Papers.Independent.Support50.P22.relabelPublic

For any natural number nnn, arbitrary type BBB, and permutation π\piπ of the nnn positions, the definition gives a function on complex matrices indexed by Sym⁡(n)×B\operatorname{Sym}(n)\times BSym(n)×B through

Kπ(X)(σ,u),(τ,v)=X(σπ−1,u),(τπ−1,v).\mathcal K_\pi(X)_{(\sigma,u),(\tau,v)} = X_{(\sigma\pi^{-1},u),(\tau\pi^{-1},v)}.Kπ​(X)(σ,u),(τ,v)​=X(σπ−1,u),(τπ−1,v)​.

The products are composition in the permutation group, so both public labels are multiplied on the right by π−1\pi^{-1}π−1, while u,vu,vu,v are unchanged. There is no finiteness assumption on BBB, and n=0n=0n=0 is included.

QAlgorithms.Papers.Independent.Support50.P22.replaceKeys

For finite types L,CL,CL,C and a function len:L→N\mathrm{len}:L\to\mathbb Nlen:L→N, let K=⨆ℓ∈L({0,1}len(ℓ)×{0,1}len(ℓ)×C)K=\bigsqcup_{\ell\in L}(\{0,1\}^{\mathrm{len}(\ell)}\times\{0,1\}^{\mathrm{len}(\ell)}\times C)K=⨆ℓ∈L​({0,1}len(ℓ)×{0,1}len(ℓ)×C). This definition gives a function on arbitrary complex matrices on KKK. For indices a=(ℓ,(u,v),c)a=(\ell,(u,v),c)a=(ℓ,(u,v),c) and b∈Kb\in Kb∈K, its output is

replaceKeys⁡(X)a,b={2−len(ℓ)∑s,t∈{0,1}len(ℓ)X(ℓ,(s,t),c),(ℓ,(s,t),c),a=b and u=v,0,otherwise.\operatorname{replaceKeys}(X)_{a,b} = \begin{cases} 2^{-\mathrm{len}(\ell)} \displaystyle\sum_{s,t\in\{0,1\}^{\mathrm{len}(\ell)}} X_{(\ell,(s,t),c),(\ell,(s,t),c)}, & a=b\text{ and }u=v,\\ 0,&\text{otherwise}. \end{cases}replaceKeys(X)a,b​=⎩⎨⎧​2−len(ℓ)s,t∈{0,1}len(ℓ)∑​X(ℓ,(s,t),c),(ℓ,(s,t),c)​,0,​a=b and u=v,otherwise.​

No positivity or normalization is required of XXX. Length zero is included, with singleton string sets and factor 111; if LLL or CCC is empty, KKK is empty.

QAlgorithms.Papers.Independent.Support50.P22.subspaceProjectorMatrix

For a finite type AAA equipped with decidable equality and a complex linear subspace S⊆CAS\subseteq\mathbb C^AS⊆CA, this defines the matrix of the orthogonal projection onto SSS in the coordinate basis:

Pa,b=(proj⁡Seb)(a),P_{a,b}=(\operatorname{proj}_S e_b)(a),Pa,b​=(projS​eb​)(a),

where eb(a)=1e_b(a)=1eb​(a)=1 when a=ba=ba=b and 000 otherwise. No nonemptiness assumption is made on AAA; for empty AAA, this is the empty matrix.

QAlgorithms.Papers.Independent.Support50.P22.tauPair

For natural numbers n,dn,dn,d with d>0d>0d>0, put Hn={0,…,d−1}nH_n=\{0,\ldots,d-1\}^nHn​={0,…,d−1}n. Sample UUU from normalized Haar measure on the unitary group in dimension d2d^2d2, let ψ=Ue0\psi=Ue_0ψ=Ue0​, and reshape it into va,b=ψindex⁡(a,b)v_{a,b}=\psi_{\operatorname{index}(a,b)}va,b​=ψindex(a,b)​ using the specified standard bijection between pairs of ddd-valued indices and indices below d2d^2d2. This definition gives the complex matrix on Hn×HnH_n\times H_nHn​×Hn​ whose entries are

(τn,d)(x,y),(x′,y′)=∫∏i=0n−1vxi,yi vxi′,yi′‾ dU.(\tau_{n,d})_{(x,y),(x',y')} = \int \prod_{i=0}^{n-1} v_{x_i,y_i}\,\overline{v_{x'_i,y'_i}} \,dU.(τn,d​)(x,y),(x′,y′)​=∫i=0∏n−1​vxi​,yi​​vxi′​,yi′​​​dU.

Thus the integrand is the grouped tensor power of the matrix vv∗vv^*vv∗. The integral is a matrix-valued Bochner integral. For n=0n=0n=0, its indexing set is a singleton and its integrand is the scalar matrix 111, by the empty-product convention.

QAlgorithms.Papers.Independent.Support50.P22.tensorId

For arbitrary types A,B,RA,B,RA,B,R and an arbitrary function DDD from complex matrices indexed by A×AA\times AA×A to complex matrices indexed by B×BB\times BB×B, this defines a function on matrices with an additional RRR-register. For a matrix XXX indexed by A×RA\times RA×R, its output is given by

(tensorId⁡D)(X)(b,r),(b′,s)=D ⁣((X(a,r),(a′,s))a,a′∈A)b,b′.(\operatorname{tensorId}D)(X)_{(b,r),(b',s)} = D\!\left((X_{(a,r),(a',s)})_{a,a'\in A}\right)_{b,b'}.(tensorIdD)(X)(b,r),(b′,s)​=D((X(a,r),(a′,s)​)a,a′∈A​)b,b′​.

Thus DDD is applied separately to each matrix block specified by the two reference indices r,sr,sr,s. The types may be empty or infinite, and no linearity or channel assumption is imposed on DDD.

Human review
  • Endorsed by Shuze Chen · Oct 24, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Goku · Oct 24, 2026

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

  • Endorsed by Gohan · Oct 29, 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