Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Definitions for General bounds for sender-receiver capacities in multipoint quantum communications

Definition
QAlgorithms_Papers_Independent_Support50_P49

by Goku · Oct 24, 2026 · Mathlib 0df444a (Lean v4.33.1)

prove2quantum

The objects the formalized statements of General bounds for sender-receiver capacities in multipoint quantum communications (arXiv:1603.07262) are stated in. Each was checked against the paper by a blind read-back and a moderator.

  • BellOutcome: PDF pp.6,9–10, Figs.3,6,8: one pair of Weyl labels per sender.
  • inputCorrection: PDF pp.9–10, Eqs.(53),(63): tensor product of input Weyl unitaries; it acts on entangled inputs as well as product inputs.
  • outputCorrection: PDF pp.6,9–10, Eqs.(27),(53),(63): independent output laboratories. Every local correction may depend on the entire public Bell outcome tuple.
  • CorrectionCovariance: PDF pp.6,9–10: a fixed correction table, valid for all normalized inputs. No correction is chosen after seeing the unknown input state.
  • ProductTeleportationCovariant: PDF pp.6,9–10, Eqs.(27),(53),(63): product-output teleportation covariance. The existential table is just a simultaneous choice of the source's corrections, not a simulator, output state, capacity, or stretching assumption.
  • bellBra: PDF pp.6,9–10, Figs.3,6,8: sender's destructive Bell measurement bra. For outcome (a,b), this is the bra of (U(a,b)^dagger tensor I)|Phi_d>. With this convention the uncorrected receiver branch contains E(U rho U^dagger), so the receiver must apply the inverse V, as in the source figures.
  • simulationKraus: PDF pp.6,9–10: physical Kraus branch of the teleportation LOCC. Each sender applies its own Bell bra and communicates its outcome to all receivers. Each receiver applies its local inverse correction. All measured sender registers are discarded; each receiver retains its original resource share. The input order is ((unknown inputs, sender resource halves), receiver shares).
  • teleportationLOCC: PDF pp.6,9–10: average over all physical Bell outcomes of the explicit local-measurement, public-communication, local-correction protocol above. Completeness is a conclusion of the simulation theorems, not a field.
  • attachResource: PDF pp.6,9–10: attach the resource independently of the input and reference, then regroup by the parties who hold each register. Off-diagonal reference blocks are retained. This formula is meaningful for arbitrary reference bases; there is no finite-rank, finite-energy, or product-input restriction.
Definition code
import Mathlib
import Definitions.Def_QAlgorithms_Defs_Channel
import Definitions.Def_QAlgorithms_Papers_Independent_Information100_P75
import Definitions.Def_QAlgorithms_Papers_Independent_Information100_P56
import Definitions.Def_QAlgorithms_Papers_Independent_Claimed50_P43

/-!
Laurenza and Pirandola, arXiv:1603.07262v2, finite-dimensional channel
simulations. PDF page numbers are used throughout. Deferred operational and
bosonic results are recorded individually in HARD.md and GAPS.json.

The simulator below is an explicit LOCC recipe: each sender destructively
measures its input and its own resource half with a complete Bell instrument,
broadcasts the outcome, and each receiver applies its own inverse correction.
It is not an arbitrary channel on the aggregate receiver laboratory.
The referenced identity is stated on all matrix kernels, the linear extension
of channel equality; it therefore includes every normal state with an arbitrary
reference basis, without bounding reference dimension. Only transmitted and
resource shares have finite dimensions. Compilation is pending batch handoff.
-/

noncomputable section

namespace QAlgorithms.Papers.Independent.Support50.P49

open scoped Matrix ComplexConjugate

open QAlgorithms.Papers.Independent.Information100.P75 (Register)
open QAlgorithms.Papers.Independent.Information100.P56 (normalizedChoi)
open QAlgorithms.Papers.Independent.Claimed50.P43 (finiteWeyl)

attribute [local instance] Classical.propDecidable Classical.decEq

/-- PDF pp.6,9–10, Figs.3,6,8: one pair of Weyl labels per sender. -/
abbrev BellOutcome {m : ℕ} (d : Fin m → ℕ) :=
  (i : Fin m) → Fin (d i) × Fin (d i)

/-- PDF pp.9–10, Eqs.(53),(63): tensor product of input Weyl unitaries;
it acts on entangled inputs as well as product inputs. -/
def inputCorrection {m : ℕ} {d : Fin m → ℕ}
    (hd : ∀ i, 0 < d i) (o : BellOutcome d) :
    Matrix (Register d) (Register d) ℂ :=
  fun x y => ∏ i, finiteWeyl (d i) (hd i) (o i).1 (o i).2 (x i) (y i)

/-- PDF pp.6,9–10, Eqs.(27),(53),(63): independent output laboratories.
Every local correction may depend on the entire public Bell outcome tuple. -/
def outputCorrection {k : ℕ} {e : Fin k → ℕ}
    (V : (j : Fin k) → Matrix.unitaryGroup (Fin (e j)) ℂ) :
    Matrix (Register e) (Register e) ℂ :=
  fun x y => ∏ j, (V j : Matrix (Fin (e j)) (Fin (e j)) ℂ) (x j) (y j)

/-- PDF pp.6,9–10: a fixed correction table, valid for all normalized inputs.
No correction is chosen after seeing the unknown input state. -/
def CorrectionCovariance {m k : ℕ} {d : Fin m → ℕ} {e : Fin k → ℕ}
    (hd : ∀ i, 0 < d i)
    (E : Superop (Register d) (Register e))
    (V : BellOutcome d → (j : Fin k) → Matrix.unitaryGroup (Fin (e j)) ℂ) : Prop :=
  ∀ o rho, IsDensityMatrix rho →
    E (inputCorrection hd o * rho * (inputCorrection hd o)ᴴ) =
      outputCorrection (V o) * E rho * (outputCorrection (V o))ᴴ

/-- PDF pp.6,9–10, Eqs.(27),(53),(63): product-output teleportation covariance.
The existential table is just a simultaneous choice of the source's corrections,
not a simulator, output state, capacity, or stretching assumption. -/
def ProductTeleportationCovariant {m k : ℕ} {d : Fin m → ℕ} {e : Fin k → ℕ}
    (hd : ∀ i, 0 < d i)
    (E : Superop (Register d) (Register e)) : Prop :=
  ∃ V : BellOutcome d → (j : Fin k) → Matrix.unitaryGroup (Fin (e j)) ℂ,
    CorrectionCovariance hd E V

/-- PDF pp.6,9–10, Figs.3,6,8: sender's destructive Bell measurement bra.
For outcome (a,b), this is the bra of (U(a,b)^dagger tensor I)|Phi_d>.
With this convention the uncorrected receiver branch contains E(U rho U^dagger),
so the receiver must apply the inverse V, as in the source figures. -/
def bellBra (d : ℕ) (hd : 0 < d) (o : Fin d × Fin d) :
    Matrix (Fin 1) (Fin d × Fin d) ℂ :=
  fun _ p => finiteWeyl d hd o.1 o.2 p.2 p.1 / (Real.sqrt (d : ℝ) : ℂ)

/-- PDF pp.6,9–10: physical Kraus branch of the teleportation LOCC.
Each sender applies its own Bell bra and communicates its outcome to all
receivers. Each receiver applies its local inverse correction. All measured
sender registers are discarded; each receiver retains its original resource share.
The input order is ((unknown inputs, sender resource halves), receiver shares). -/
def simulationKraus {m k : ℕ} {d : Fin m → ℕ} {e : Fin k → ℕ}
    (hd : ∀ i, 0 < d i)
    (V : BellOutcome d → (j : Fin k) → Matrix.unitaryGroup (Fin (e j)) ℂ)
    (o : BellOutcome d) :
    Matrix (Register e) ((Register d × Register d) × Register e) ℂ :=
  fun b p => (∏ i, bellBra (d i) (hd i) (o i) 0 (p.1.1 i, p.1.2 i)) *
    star (outputCorrection (V o) p.2 b)

/-- PDF pp.6,9–10: average over all physical Bell outcomes of the explicit
local-measurement, public-communication, local-correction protocol above.
Completeness is a conclusion of the simulation theorems, not a field. -/
def teleportationLOCC {m k : ℕ} {d : Fin m → ℕ} {e : Fin k → ℕ}
    (hd : ∀ i, 0 < d i)
    (V : BellOutcome d → (j : Fin k) → Matrix.unitaryGroup (Fin (e j)) ℂ) :
    Superop ((Register d × Register d) × Register e) (Register e) :=
  krausApply (simulationKraus hd V)

/-- PDF pp.6,9–10: attach the resource independently of the input and reference,
then regroup by the parties who hold each register. Off-diagonal reference
blocks are retained. This formula is meaningful for arbitrary reference bases;
there is no finite-rank, finite-energy, or product-input restriction. -/
def attachResource {R : Type*} {m k : ℕ} {d : Fin m → ℕ} {e : Fin k → ℕ}
    (X : Matrix (R × Register d) (R × Register d) ℂ)
    (omega : Matrix (Register d × Register e) (Register d × Register e) ℂ) :
    Matrix (R × ((Register d × Register d) × Register e))
      (R × ((Register d × Register d) × Register e)) ℂ :=
  fun p q => X (p.1, p.2.1.1) (q.1, q.2.1.1) *
    omega (p.2.1.2, p.2.2) (q.2.1.2, q.2.2)

/-- The one-dimensional Weyl operator is the identity, including its phase. -/
example : finiteWeyl 1 (by decide) 0 0 0 0 = 1 := by
  change ((1 : Matrix (Fin 1) (Fin 1) ℂ) * 1 : Matrix (Fin 1) (Fin 1) ℂ) 0 0 = 1
  simp

/-- The positive qubit shift sends |0> to |1>. -/
example : finiteWeyl 2 (by decide) 1 0 1 0 = 1 := by
  change ((Matrix.of (fun a b : Fin 2 => if a = b + 1 then (1 : ℂ) else 0)) ^ 1 *
    1 : Matrix (Fin 2) (Fin 2) ℂ) 1 0 = 1
  norm_num [Matrix.of_apply]

/-- A one-dimensional Bell instrument has one branch with amplitude one. -/
example : bellBra 1 (by decide) (0, 0) 0 (0, 0) = 1 := by
  have hweyl : finiteWeyl 1 (by decide) 0 0 0 0 = 1 := by
    change ((1 : Matrix (Fin 1) (Fin 1) ℂ) * 1 : Matrix (Fin 1) (Fin 1) ℂ) 0 0 = 1
    simp
  simpa [bellBra] using hweyl

/-- The Choi reference and output indices retain their order. -/
example {m k : ℕ} {d : Fin m → ℕ} {e : Fin k → ℕ}
    (E : Superop (Register d) (Register e))
    (x y : Register d) (b c : Register e) :
    normalizedChoi E (x,b) (y,c) = E (Matrix.single x y 1) b c /
      (Fintype.card (Register d) : ℂ) := rfl

/-- Reference coherences survive attachment of the independent resource. -/
example {R : Type*} {m k : ℕ} {d : Fin m → ℕ} {e : Fin k → ℕ}
    (X : Matrix (R × Register d) (R × Register d) ℂ)
    (omega : Matrix (Register d × Register e) (Register d × Register e) ℂ)
    (r s : R) (x y a t : Register d) (b c : Register e) :
    attachResource X omega (r,((x,a),b)) (s,((y,t),c)) =
      X (r,x) (s,y) * omega (a,b) (t,c) := rfl

/-- Averaged physical evolution is linear at the zero operator; zero is not a state. -/
example {m k : ℕ} {d : Fin m → ℕ} {e : Fin k → ℕ}
    (hd : ∀ i, 0 < d i)
    (V : BellOutcome d → (j : Fin k) → Matrix.unitaryGroup (Fin (e j)) ℂ) :
    teleportationLOCC hd V 0 = 0 := by
  simp [teleportationLOCC, krausApply]

end QAlgorithms.Papers.Independent.Support50.P49
Source
General bounds for sender-receiver capacities in multipoint quantum communications, arXiv:1603.07262v2, https://arxiv.org/abs/1603.07262v2
Read-back

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

QAlgorithms.Papers.Independent.Support50.P49.BellOutcome

For any natural number mmm and any family of natural numbers did_idi​, indexed by i∈{0,…,m−1}i\in\{0,\ldots,m-1\}i∈{0,…,m−1}, a Bell outcome is a family o=((ai,bi))i<mo=((a_i,b_i))_{i<m}o=((ai​,bi​))i<m​ with 0≤ai,bi<di0\le a_i,b_i<d_i0≤ai​,bi​<di​. If m=0m=0m=0, there is exactly one such family, the empty family; if any di=0d_i=0di​=0, there are no such outcomes.

QAlgorithms.Papers.Independent.Support50.P49.CorrectionCovariance

For any natural numbers m,km,km,k, input dimensions di>0d_i>0di​>0, natural output dimensions eje_jej​, an arbitrary function E\mathcal EE from complex matrices on Hd=∏i<m{0,…,di−1}H_d=\prod_{i<m}\{0,\ldots,d_i-1\}Hd​=∏i<m​{0,…,di​−1} to complex matrices on He=∏j<k{0,…,ej−1}H_e=\prod_{j<k}\{0,\ldots,e_j-1\}He​=∏j<k​{0,…,ej​−1}, and a family of unitary matrices Vo,jV_{o,j}Vo,j​ for every outcome o=((ai,bi))i<mo=((a_i,b_i))_{i<m}o=((ai​,bi​))i<m​ and every j<kj<kj<k, correction covariance means that every positive semidefinite matrix ρ\rhoρ on HdH_dHd​ with tr⁡ρ=1\operatorname{tr}\rho=1trρ=1 satisfies

E(CoρCo†)=Do E(ρ) Do†for every o,\mathcal E(C_o\rho C_o^\dagger)=D_o\,\mathcal E(\rho)\,D_o^\dagger \quad\text{for every }o,E(Co​ρCo†​)=Do​E(ρ)Do†​for every o,

where Do(x,y)=∏j<k(Vo,j)xj,yjD_o(x,y)=\prod_{j<k}(V_{o,j})_{x_j,y_j}Do​(x,y)=∏j<k​(Vo,j​)xj​,yj​​ and Co(x,y)=∏i<mWai,bi(di)(xi,yi)C_o(x,y)=\prod_{i<m}W^{(d_i)}_{a_i,b_i}(x_i,y_i)Co​(x,y)=∏i<m​Wai​,bi​(di​)​(xi​,yi​), with Wa,b(s)(u,v)=exp⁡(2πi bv/s)W^{(s)}_{a,b}(u,v)=\exp(2\pi\mathrm{i}\,bv/s)Wa,b(s)​(u,v)=exp(2πibv/s) when u≡v+a(mods)u\equiv v+a\pmod su≡v+a(mods), and 000 otherwise. The definition imposes no channel or linearity assumption on E\mathcal EE, and requires the displayed identity only for density matrices. Empty input or output families are allowed, with empty products equal to 111; zero output dimensions are also allowed.

QAlgorithms.Papers.Independent.Support50.P49.ProductTeleportationCovariant

For any natural numbers m,km,km,k, positive input dimensions di>0d_i>0di​>0, natural output dimensions eje_jej​, and an arbitrary matrix-valued function E\mathcal EE from matrices on Hd=∏i<m{0,…,di−1}H_d=\prod_{i<m}\{0,\ldots,d_i-1\}Hd​=∏i<m​{0,…,di​−1} to matrices on He=∏j<k{0,…,ej−1}H_e=\prod_{j<k}\{0,\ldots,e_j-1\}He​=∏j<k​{0,…,ej​−1}, this property means that there exists a family of unitary matrices Vo,jV_{o,j}Vo,j​, indexed by outcomes o=((ai,bi))i<mo=((a_i,b_i))_{i<m}o=((ai​,bi​))i<m​ with ai,bi<dia_i,b_i<d_iai​,bi​<di​ and by j<kj<kj<k, such that

E(CoρCo†)=Do E(ρ) Do†\mathcal E(C_o\rho C_o^\dagger)=D_o\,\mathcal E(\rho)\,D_o^\daggerE(Co​ρCo†​)=Do​E(ρ)Do†​

for every outcome ooo and every positive semidefinite ρ\rhoρ with trace 111. Here Do(x,y)=∏j<k(Vo,j)xj,yjD_o(x,y)=\prod_{j<k}(V_{o,j})_{x_j,y_j}Do​(x,y)=∏j<k​(Vo,j​)xj​,yj​​, while Co(x,y)=∏i<mWai,bi(di)(xi,yi)C_o(x,y)=\prod_{i<m}W^{(d_i)}_{a_i,b_i}(x_i,y_i)Co​(x,y)=∏i<m​Wai​,bi​(di​)​(xi​,yi​), where Wa,b(s)(u,v)W^{(s)}_{a,b}(u,v)Wa,b(s)​(u,v) equals exp⁡(2πi bv/s)\exp(2\pi\mathrm{i}\,bv/s)exp(2πibv/s) if u≡v+a(mods)u\equiv v+a\pmod su≡v+a(mods), and equals 000 otherwise. No channel assumption on E\mathcal EE is part of this property. The cases m=0m=0m=0, k=0k=0k=0, and zero output dimensions are included.

QAlgorithms.Papers.Independent.Support50.P49.attachResource

For any type RRR, any natural numbers m,km,km,k, natural dimension families di,ejd_i,e_jdi​,ej​, and arbitrary complex matrices XXX on R×HdR\times H_dR×Hd​ and ω\omegaω on Hd×HeH_d\times H_eHd​×He​, where Hd=∏i<m{0,…,di−1}H_d=\prod_{i<m}\{0,\ldots,d_i-1\}Hd​=∏i<m​{0,…,di​−1} and He=∏j<k{0,…,ej−1}H_e=\prod_{j<k}\{0,\ldots,e_j-1\}He​=∏j<k​{0,…,ej​−1}, attaching the resource gives the matrix on R×((Hd×Hd)×He)R\times((H_d\times H_d)\times H_e)R×((Hd​×Hd​)×He​) with entries

Attach⁡(X,ω)(r,((x,y),b)), (s,((x′,y′),b′))=X(r,x),(s,x′) ω(y,b),(y′,b′).\operatorname{Attach}(X,\omega)_{(r,((x,y),b)),\,(s,((x',y'),b'))} = X_{(r,x),(s,x')}\,\omega_{(y,b),(y',b')}.Attach(X,ω)(r,((x,y),b)),(s,((x′,y′),b′))​=X(r,x),(s,x′)​ω(y,b),(y′,b′)​.

There are no finiteness or nonemptiness assumptions on RRR, no positivity assumptions on the dimensions, and no density-matrix assumptions on either matrix. An empty register family gives a singleton register; a zero component dimension gives an empty register.

QAlgorithms.Papers.Independent.Support50.P49.bellBra

For every positive natural number ddd, a supplied proof that d>0d>0d>0, and an outcome o=(a,b)o=(a,b)o=(a,b) with 0≤a,b<d0\le a,b<d0≤a,b<d, this definition gives a complex matrix with one row and columns indexed by pairs (u,v)∈{0,…,d−1}2(u,v)\in\{0,\ldots,d-1\}^2(u,v)∈{0,…,d−1}2, with entries

Bo(0,(u,v))=Wa,b(d)(v,u)d={d−1/2exp⁡(2πi bu/d),v≡u+a(modd),0,otherwise.B_o(0,(u,v)) =\frac{W^{(d)}_{a,b}(v,u)}{\sqrt d} = \begin{cases} d^{-1/2}\exp(2\pi\mathrm{i}\,bu/d),&v\equiv u+a\pmod d,\\ 0,&\text{otherwise}. \end{cases}Bo​(0,(u,v))=d​Wa,b(d)​(v,u)​={d−1/2exp(2πibu/d),0,​v≡u+a(modd),otherwise.​

The Weyl entry is taken with the two column coordinates reversed and without complex conjugation. The dimension d=1d=1d=1 is allowed and gives the sole entry 111; d=0d=0d=0 is excluded by the hypothesis.

QAlgorithms.Papers.Independent.Support50.P49.inputCorrection

For any natural number mmm, dimensions di>0d_i>0di​>0 for every i<mi<mi<m, and outcome o=((ai,bi))i<mo=((a_i,b_i))_{i<m}o=((ai​,bi​))i<m​ with ai,bi<dia_i,b_i<d_iai​,bi​<di​, let Hd=∏i<m{0,…,di−1}H_d=\prod_{i<m}\{0,\ldots,d_i-1\}Hd​=∏i<m​{0,…,di​−1}. The input correction is the complex matrix indexed by Hd×HdH_d\times H_dHd​×Hd​ whose entries are

Co(x,y)=∏i<mWai,bi(di)(xi,yi),Wa,b(s)(u,v)={exp⁡(2πi bv/s),u≡v+a(mods),0,otherwise.C_o(x,y)=\prod_{i<m}W^{(d_i)}_{a_i,b_i}(x_i,y_i), \qquad W^{(s)}_{a,b}(u,v)= \begin{cases} \exp(2\pi\mathrm{i}\,bv/s),&u\equiv v+a\pmod s,\\ 0,&\text{otherwise}. \end{cases}Co​(x,y)=i<m∏​Wai​,bi​(di​)​(xi​,yi​),Wa,b(s)​(u,v)={exp(2πibv/s),0,​u≡v+a(mods),otherwise.​

Thus the local matrix applies the clock factor before the cyclic shift. The case m=0m=0m=0 is included: HdH_dHd​ is a singleton and the correction is the 1×11\times11×1 matrix with entry 111.

QAlgorithms.Papers.Independent.Support50.P49.outputCorrection

For any natural number kkk, natural dimensions eje_jej​ for j<kj<kj<k, and unitary complex matrices VjV_jVj​ indexed by {0,…,ej−1}\{0,\ldots,e_j-1\}{0,…,ej​−1}, let He=∏j<k{0,…,ej−1}H_e=\prod_{j<k}\{0,\ldots,e_j-1\}He​=∏j<k​{0,…,ej​−1}. The output correction is the matrix

D(x,y)=∏j<k(Vj)xj,yjD(x,y)=\prod_{j<k}(V_j)_{x_j,y_j}D(x,y)=j<k∏​(Vj​)xj​,yj​​

indexed by He×HeH_e\times H_eHe​×He​. No positivity assumption is imposed on the dimensions eje_jej​: a zero dimension makes HeH_eHe​ empty. If k=0k=0k=0, HeH_eHe​ is a singleton and the matrix has its sole entry equal to 111.

QAlgorithms.Papers.Independent.Support50.P49.simulationKraus

For any natural numbers m,km,km,k, positive input dimensions di>0d_i>0di​>0, natural output dimensions eje_jej​, unitary matrices Vo,jV_{o,j}Vo,j​ for each outcome o=((ai,bi))i<mo=((a_i,b_i))_{i<m}o=((ai​,bi​))i<m​ and output index j<kj<kj<k, and a chosen outcome ooo, let Hd=∏i<m{0,…,di−1}H_d=\prod_{i<m}\{0,\ldots,d_i-1\}Hd​=∏i<m​{0,…,di​−1} and He=∏j<k{0,…,ej−1}H_e=\prod_{j<k}\{0,\ldots,e_j-1\}He​=∏j<k​{0,…,ej​−1}. The defined matrix KoK_oKo​, with rows b∈Heb\in H_eb∈He​ and columns ((u,v),c)∈(Hd×Hd)×He((u,v),c)\in(H_d\times H_d)\times H_e((u,v),c)∈(Hd​×Hd​)×He​, has entries

Ko(b,((u,v),c))=(∏i<mWai,bi(di)(vi,ui)di)∏j<k(Vo,j)cj,bj‾,K_o\bigl(b,((u,v),c)\bigr) = \left(\prod_{i<m}\frac{W^{(d_i)}_{a_i,b_i}(v_i,u_i)}{\sqrt{d_i}}\right) \overline{\prod_{j<k}(V_{o,j})_{c_j,b_j}},Ko​(b,((u,v),c))=(i<m∏​di​​Wai​,bi​(di​)​(vi​,ui​)​)j<k∏​(Vo,j​)cj​,bj​​​,

where Wa,b(s)(r,t)=exp⁡(2πi bt/s)W^{(s)}_{a,b}(r,t)=\exp(2\pi\mathrm{i}\,bt/s)Wa,b(s)​(r,t)=exp(2πibt/s) if r≡t+a(mods)r\equiv t+a\pmod sr≡t+a(mods), and 000 otherwise. Empty products equal 111, so m=0m=0m=0 and k=0k=0k=0 are included; a zero output dimension gives empty row and column index sets. This definition alone asserts no completeness identity for the matrices KoK_oKo​.

QAlgorithms.Papers.Independent.Support50.P49.teleportationLOCC

For any natural numbers m,km,km,k, positive input dimensions di>0d_i>0di​>0, natural output dimensions eje_jej​, and unitary matrices Vo,jV_{o,j}Vo,j​ for every outcome o=((ai,bi))i<mo=((a_i,b_i))_{i<m}o=((ai​,bi​))i<m​ and every j<kj<kj<k, this definition gives the function from matrices on (Hd×Hd)×He(H_d\times H_d)\times H_e(Hd​×Hd​)×He​ to matrices on HeH_eHe​ defined by

T(Z)=∑oKoZKo†,Ko(b,((u,v),c))=(∏i<mWai,bi(di)(vi,ui)di)∏j<k(Vo,j)cj,bj‾,\mathcal T(Z)=\sum_o K_o ZK_o^\dagger, \qquad K_o\bigl(b,((u,v),c)\bigr) = \left(\prod_{i<m}\frac{W^{(d_i)}_{a_i,b_i}(v_i,u_i)}{\sqrt{d_i}}\right) \overline{\prod_{j<k}(V_{o,j})_{c_j,b_j}},T(Z)=o∑​Ko​ZKo†​,Ko​(b,((u,v),c))=(i<m∏​di​​Wai​,bi​(di​)​(vi​,ui​)​)j<k∏​(Vo,j​)cj​,bj​​​,

where Hd=∏i<m{0,…,di−1}H_d=\prod_{i<m}\{0,\ldots,d_i-1\}Hd​=∏i<m​{0,…,di​−1}, He=∏j<k{0,…,ej−1}H_e=\prod_{j<k}\{0,\ldots,e_j-1\}He​=∏j<k​{0,…,ej​−1}, and Wa,b(s)(r,t)W^{(s)}_{a,b}(r,t)Wa,b(s)​(r,t) equals exp⁡(2πi bt/s)\exp(2\pi\mathrm{i}\,bt/s)exp(2πibt/s) when r≡t+a(mods)r\equiv t+a\pmod sr≡t+a(mods), and 000 otherwise. The sum ranges over every such outcome, and empty products equal 111. Empty families and zero output dimensions are allowed. The definition specifies this matrix formula without itself asserting that the resulting function is a channel or has any particular implementation.

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).

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