Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Recoverable PBT for an odd number of qubit ports

Open
QAlgorithms.Papers.Independent.Intakenext100.P53.recoverable_pbt_odd

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

prove2quantum

Let N≥1N\ge1N≥1 be odd and set J=(N+1)/2J=(N+1)/2J=(N+1)/2. Use the qubit resource OprobO_{\mathrm{prob}}Oprob​ from Eq.(15), the success effects Si=ρ−1/2σiρ−1/2S_i=\rho^{-1/2}\sigma_i\rho^{-1/2}Si​=ρ−1/2σi​ρ−1/2, and the failure projector P0=I−supp⁡ρP_0=I-\operatorname{supp}\rhoP0​=I−suppρ. The inverse square root vanishes on the kernel. In the maximal-spin failure sector Alice measures

e0=∣J,0⟩,em±=(∣J,m⟩±∣J,−m⟩)/2,m=1,…,J.e_0=|J,0\rangle,\qquad e_m^\pm=(|J,m\rangle\pm|J,-m\rangle)/\sqrt2,\quad m=1,\ldots,J.e0​=∣J,0⟩,em±​=(∣J,m⟩±∣J,−m⟩)/2​,m=1,…,J.

These vectors form an orthonormal basis of the failure sector, so the success and refined failure effects form a POVM. Each successful port has probability 1/(N+3)1/(N+3)1/(N+3) and teleports the input exactly. Writing cN=3/((N+2)(N+3))c_N=3/((N+2)(N+3))cN​=3/((N+2)(N+3)), the unnormalized failure state on reference DDD and all outputs BBB is

ω0=cN∑m=−JJ∣J,m⟩⟨J,m∣,tr⁡ω0=3N+3.\omega_0=c_N\sum_{m=-J}^{J}|J,m\rangle\langle J,m|,\qquad\operatorname{tr}\omega_0=\frac3{N+3}.ω0​=cN​m=−J∑J​∣J,m⟩⟨J,m∣,trω0​=N+33​.

Each refined failure outcome has probability cNc_NcN​ and a normalized pure conditional state whose reference marginal is ID/2I_D/2ID​/2. There are unitaries UzU_zUz​ on all NNN output qubits, chosen solely from the classical refinement outcome zzz, such that for every finite reference RRR and density operator TTT on R⊗CR\otimes CR⊗C, recovery followed by retaining a fixed output qubit gives cNTc_NTcN​T in branch zzz. Successful port iii gives T/(N+3)T/(N+3)T/(N+3) without correction. With the same Alice measurement, ignoring the refinement and choosing a uniformly random port on failure instead gives

F=NN+3.F=\frac{N}{N+3}.F=N+3N​.

At N=1N=1N=1 the resource is a singlet and Alice's success singlet plus the three refinement vectors constitute the usual Bell measurement. This is the constructed hybrid of port selection and faithful teleportation using all outputs.

Formalization Note The printed Eq.(16) scalar is corrected as forced by its preceding normalized sandwich; even NNN is excluded.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Intakenext100_P53
Formal statement
noncomputable section

namespace QAlgorithms.Papers.Independent.Intakenext100.P53

open scoped Classical ComplexConjugate ComplexOrder Matrix Matrix.Norms.L2Operator

-- Accepted generic matrix and positive-operator interfaces. The support is the
-- orthogonal range projection; all uses below are on positive physical rho.

open QAlgorithms.Papers.Independent.Information100.P21 (Op)

open QAlgorithms.Papers.Independent.Information100.P07 (support)

open QAlgorithms.Papers.Independent.Next46.P39 (inverseSqrt)

/-- PDF pp.3-4, Eqs.(16)-(18), summary p.5: explicit odd-N recoverable PBT.
The printed Eq.(16) coefficient 6*2^N/((N+2)(N+3)) is corrected to
3/((N+2)(N+3)), forced by its preceding sandwich and the normalized Eq.(15).
Unitaries depend only on the classical refinement outcome, precede every input,
and act on all B qubits. The alternative branch is correction-free port selection. -/
theorem recoverable_pbt_odd (N : ℕ) (hN : 1 ≤ N) (hodd : Odd N) :
    let O := probabilisticResource N
    let p0 : ℝ := 3 / (((N : ℝ) + 2) * ((N : ℝ) + 3))
    ExactProbabilistic O (probabilisticSRM 2 N) ∧
    (∀ i : Fin N, pbtBranchProbability O (srm 2 N i) = 1 / ((N : ℝ) + 3)) ∧
    (∀ z : Refinement N, QAlgorithms.IsState (refinementVector N z)) ∧
    (∀ z w : Refinement N, inner ℂ (refinementVector N z) (refinementVector N w) =
      if z = w then 1 else 0) ∧
    (∑ z : Refinement N, refinementEffect N z) = probabilisticSRM 2 N none ∧
    (∀ z : Fin N ⊕ Refinement N, (recoveryMeasurement N z).PosSemidef) ∧
    (∑ z : Fin N ⊕ Refinement N, recoveryMeasurement N z) = 1 ∧
    pbtBranch O (probabilisticSRM 2 N none) =
      (p0 : ℂ) • ∑ k : Fin (N + 2), QAlgorithms.pureDensity (bdDicke N k.val) ∧
    pbtBranchProbability O (probabilisticSRM 2 N none) = 3 / ((N : ℝ) + 3) ∧
    (∀ z : Refinement N, pbtBranchProbability O (refinementEffect N z) = p0 ∧
      ∃ v : EuclideanSpace ℂ (Fin 2 × Ports 2 N), QAlgorithms.IsState v ∧
        pbtBranch O (refinementEffect N z) = (p0 : ℂ) • QAlgorithms.pureDensity v ∧
        QAlgorithms.traceRight (QAlgorithms.pureDensity v) = (1 / 2 : ℂ) • (1 : Op (Fin 2))) ∧
    (∃ U : Refinement N → Op (Ports 2 N),
      (∀ z, U z ∈ Matrix.unitaryGroup (Ports 2 N) ℂ) ∧
      ∀ (R : Type) [Fintype R] [DecidableEq R] (T : Op (R × Fin 2)),
        QAlgorithms.IsDensityMatrix T →
        (∀ i : Fin N, selectPort i (pbtChannelBranch O (srm 2 N i) T) =
          ((1 / ((N : ℝ) + 3) : ℝ) : ℂ) • T) ∧
        (∀ z : Refinement N,
          selectPort (⟨0, by omega⟩ : Fin N)
            (recoverBranch (U z) (pbtChannelBranch O (refinementEffect N z) T)) = (p0 : ℂ) • T)) ∧
    pbtFidelity O (completedSRM 2 N) = (N : ℝ) / ((N : ℝ) + 3) ∧
    probabilisticResource 1 = 1 ∧ srm 2 1 (0 : Fin 1) = portProjector (d := 2) (0 : Fin 1) ∧
    refinementVector 1 (Sum.inl ()) = onePortBell false false ∧
    refinementVector 1 (Sum.inr (0, false)) = onePortBell false true ∧
    refinementVector 1 (Sum.inr (0, true)) = onePortBell true true := by sorry

end QAlgorithms.Papers.Independent.Intakenext100.P53

end
Source
Some remarks on port-based teleportation, arXiv:1506.01555v2, PDF p. 5, IV, Eqs.(16)-(18): recoverable PBT for qubits and odd port count
Read-back

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

For every natural number N≥1N\ge1N≥1 that is odd, the following assertions hold. To specify all the quantities involved, let BN={0,1}{0,…,N−1}\mathcal B_N=\{0,1\}^{\{0,\ldots,N-1\}}BN​={0,1}{0,…,N−1} be the port configurations, let AN=BN×{0,1}\mathcal A_N=\mathcal B_N\times\{0,1\}AN​=BN​×{0,1}, let III denote the identity on the space indicated by its context, let M†M^\daggerM† denote conjugate transpose, and write ∣v⟩⟨v∣|v\rangle\langle v|∣v⟩⟨v∣ for the matrix with entries vxvy‾v_x\overline{v_y}vx​vy​​. A density matrix means a positive semidefinite matrix of trace 111, and a state vector means a vector of Euclidean norm 111. Define the single-qubit matrices

X=(0110),Y=(0−ii0),Z=(100−1),X=\begin{pmatrix}0&1\\1&0\end{pmatrix},\qquad Y=\begin{pmatrix}0&-\mathrm i\\ \mathrm i&0\end{pmatrix},\qquad Z=\begin{pmatrix}1&0\\0&-1\end{pmatrix},X=(01​10​),Y=(0i​−i0​),Z=(10​0−1​),

and, on the NNN ports, define

KN=(∑j=0N−1Xj2)2+(∑j=0N−1Yj2)2+(∑j=0N−1Zj2)2,K_N=\left(\sum_{j=0}^{N-1}\frac{X_j}{2}\right)^2+ \left(\sum_{j=0}^{N-1}\frac{Y_j}{2}\right)^2+ \left(\sum_{j=0}^{N-1}\frac{Z_j}{2}\right)^2,KN​=(j=0∑N−1​2Xj​​)2+(j=0∑N−1​2Yj​​)2+(j=0∑N−1​2Zj​​)2,

where a subscript jjj means that the indicated matrix acts on port jjj and the identity acts on all other ports. Let CtC_tCt​ be the matrix obtained by applying the scalar function that is 111 at t2(t2+1)\frac t2(\frac t2+1)2t​(2t​+1) and 000 elsewhere to KNK_NKN​ by continuous functional calculus, put Dt=Re⁡tr⁡CtD_t=\operatorname{Re}\operatorname{tr}C_tDt​=RetrCt​, and let TN={t∈{0,…,N}:t mod 2=N mod 2}\mathcal T_N=\{t\in\{0,\ldots,N\}:t\bmod2=N\bmod2\}TN​={t∈{0,…,N}:tmod2=Nmod2}. The resource operator in the statement is

O=2N∑t∈TN(t+1)2 ∑t∈TN(t+1)2Dt Ct,p0=3(N+2)(N+3).O=\sqrt{\frac{2^N}{\sum_{t\in\mathcal T_N}(t+1)^2}}\, \sum_{t\in\mathcal T_N}\sqrt{\frac{(t+1)^2}{D_t}}\,C_t, \qquad p_0=\frac{3}{(N+2)(N+3)}.O=∑t∈TN​​(t+1)22N​​t∈TN​∑​Dt​(t+1)2​​Ct​,p0​=(N+2)(N+3)3​.

Define sac=(−1)a/2s_{ac}=(-1)^a/\sqrt2sac​=(−1)a/2​ if c=1−ac=1-ac=1−a, and sac=0s_{ac}=0sac​=0 otherwise, and let S=∣s⟩⟨s∣S=|s\rangle\langle s|S=∣s⟩⟨s∣. For each port iii, let B≠i\mathcal B_{\ne i}B=i​ consist of configurations on all ports other than iii, and define the matrix

(Vi)(a,c),r=saic 1{a∣≠i=r},Pi=ViVi†,σi=2−(N−1)Pi,F=∑iσi.(V_i)_{(a,c),r} =s_{a_i c}\,\mathbf1_{\{a|_{\ne i}=r\}},\qquad P_i=V_iV_i^\dagger,\qquad \sigma_i=2^{-(N-1)}P_i,\qquad F=\sum_i\sigma_i.(Vi​)(a,c),r​=sai​c​1{a∣=i​=r}​,Pi​=Vi​Vi†​,σi​=2−(N−1)Pi​,F=i∑​σi​.

Let LLL be the continuous functional calculus of FFF for x↦(x)−1x\mapsto(\sqrt x)^{-1}x↦(x​)−1, let PFP_FPF​ be the orthogonal projector onto the range of FFF, and put

Ei=LσiL,E∅=I−PF,E^i=Ei+1NE∅.E_i=L\sigma_iL,\qquad E_\varnothing=I-P_F,\qquad \widehat E_i=E_i+\frac1N E_\varnothing.Ei​=Lσi​L,E∅​=I−PF​,Ei​=Ei​+N1​E∅​.

Thus EiE_iEi​, E∅E_\varnothingE∅​, and E^i\widehat E_iEi​ respectively expand the successful square-root effects, their failure effect, and their completed effects used in the statement; scalar inverses here and below use 0−1=00^{-1}=00−1=0. Define resource coefficients

Ωa,u=∑b∈BNOa,b∏j=0N−1sbjuj.\Omega_{a,u}=\sum_{b\in\mathcal B_N}O_{a,b}\prod_{j=0}^{N-1}s_{b_j u_j}.Ωa,u​=b∈BN​∑​Oa,b​j=0∏N−1​sbj​uj​​.

For a finite reference index set R\mathcal RR, a matrix TTT on R×{0,1}\mathcal R\times\{0,1\}R×{0,1}, and an effect EEE on AN\mathcal A_NAN​, define the branch matrix on R×BN\mathcal R\times\mathcal B_NR×BN​ by

(BER(T))(α,u),(β,v)=∑a,b∈BN∑c,e∈{0,1}E(b,e),(a,c) Ωa,uΩb,v‾ T(α,c),(β,e).\bigl(\mathcal B_E^{\mathcal R}(T)\bigr)_{(\alpha,u),(\beta,v)} =\sum_{a,b\in\mathcal B_N}\sum_{c,e\in\{0,1\}} E_{(b,e),(a,c)}\, \Omega_{a,u}\overline{\Omega_{b,v}}\, T_{(\alpha,c),(\beta,e)}.(BER​(T))(α,u),(β,v)​=a,b∈BN​∑​c,e∈{0,1}∑​E(b,e),(a,c)​Ωa,u​Ωb,v​​T(α,c),(β,e)​.

Write BE=BE{0,1}(S)\mathcal B_E=\mathcal B_E^{\{0,1\}}(S)BE​=BE{0,1}​(S) and q(E)=Re⁡tr⁡BEq(E)=\operatorname{Re}\operatorname{tr}\mathcal B_Eq(E)=RetrBE​. If ins⁡i(c,r)\operatorname{ins}_i(c,r)insi​(c,r) is the port configuration with value ccc at iii and values rrr elsewhere, define port selection by

(ΔiH)(α,c),(β,e)=∑r∈B≠iH(α,ins⁡i(c,r)),(β,ins⁡i(e,r)).\bigl(\Delta_i H\bigr)_{(\alpha,c),(\beta,e)} =\sum_{r\in\mathcal B_{\ne i}} H_{(\alpha,\operatorname{ins}_i(c,r)),(\beta,\operatorname{ins}_i(e,r))}.(Δi​H)(α,c),(β,e)​=r∈B=i​∑​H(α,insi​(c,r)),(β,insi​(e,r))​.

For 0≤k≤N+10\le k\le N+10≤k≤N+1, let DkD_kDk​ be the vector on AN\mathcal A_NAN​ whose coefficient at (a,c)(a,c)(a,c) is (N+1k)−1/2\binom{N+1}{k}^{-1/2}(kN+1​)−1/2 when ∑jaj+c=k\sum_j a_j+c=k∑j​aj​+c=k, and zero otherwise, and let D~k\widetilde D_kDk​ be the vector on {0,1}×BN\{0,1\}\times\mathcal B_N{0,1}×BN​ with D~k(c,a)=Dk(a,c)\widetilde D_k(c,a)=D_k(a,c)Dk​(c,a)=Dk​(a,c). Set h=(N+1)/2h=(N+1)/2h=(N+1)/2, using natural-number division, and let the refinement indices be

ZN={∗} ⊔ ({0,…,h−1}×{false,true}).\mathcal Z_N=\{*\}\ \sqcup\ \bigl(\{0,\ldots,h-1\}\times\{\mathrm{false},\mathrm{true}\}\bigr).ZN​={∗} ⊔ ({0,…,h−1}×{false,true}).

Their vectors and effects, and the combined measurement effects, are

v∗=Dh,v(m,b)=Dh−(m+1)+(if b then −1 else 1)Dh+(m+1)2,Az=E∅∣vz⟩⟨vz∣E∅,Mi=Ei,Mz=Az,v_*=D_h,\qquad v_{(m,b)}=\frac{D_{h-(m+1)}+ \bigl(\text{if }b\text{ then }-1\text{ else }1\bigr)D_{h+(m+1)}}{\sqrt2}, \qquad A_z=E_\varnothing|v_z\rangle\langle v_z|E_\varnothing,\qquad M_i=E_i,\quad M_z=A_z,v∗​=Dh​,v(m,b)​=2​Dh−(m+1)​+(if b then −1 else 1)Dh+(m+1)​​,Az​=E∅​∣vz​⟩⟨vz​∣E∅​,Mi​=Ei​,Mz​=Az​,

where MMM is indexed by the disjoint union of ports and refinement indices; the natural subtraction in h−(m+1)h-(m+1)h−(m+1) has no underflow for these indices. With this notation, the theorem asserts that tr⁡(O†O)=2N\operatorname{tr}(O^\dagger O)=2^Ntr(O†O)=2N, that the effects E∅,E0,…,EN−1E_\varnothing,E_0,\ldots,E_{N-1}E∅​,E0​,…,EN−1​ are all positive semidefinite and sum to III, and that

ΔiBEi=q(Ei) S,q(Ei)=1N+3for every port i,∥vz∥=1,⟨vz,vw⟩={1z=w,0z≠w,for every z,w∈ZN,∑z∈ZNAz=E∅,Mξ⪰0for every combined index ξ,∑ξMξ=I,BE∅=p0∑k=0N+1∣D~k⟩⟨D~k∣,q(E∅)=3N+3.\begin{gathered} \Delta_i\mathcal B_{E_i}=q(E_i)\,S,\qquad q(E_i)=\frac1{N+3}\quad\text{for every port }i,\\ \|v_z\|=1,\qquad \langle v_z,v_w\rangle= \begin{cases}1&z=w,\\0&z\ne w,\end{cases} \quad\text{for every }z,w\in\mathcal Z_N,\\ \sum_{z\in\mathcal Z_N}A_z=E_\varnothing,\qquad M_\xi\succeq0\quad\text{for every combined index }\xi,\qquad \sum_\xi M_\xi=I,\\ \mathcal B_{E_\varnothing} =p_0\sum_{k=0}^{N+1}|\widetilde D_k\rangle\langle\widetilde D_k|, \qquad q(E_\varnothing)=\frac3{N+3}. \end{gathered}Δi​BEi​​=q(Ei​)S,q(Ei​)=N+31​for every port i,∥vz​∥=1,⟨vz​,vw​⟩={10​z=w,z=w,​for every z,w∈ZN​,z∈ZN​∑​Az​=E∅​,Mξ​⪰0for every combined index ξ,ξ∑​Mξ​=I,BE∅​​=p0​k=0∑N+1​∣Dk​⟩⟨Dk​∣,q(E∅​)=N+33​.​

For every refinement index zzz, it further asserts q(Az)=p0q(A_z)=p_0q(Az​)=p0​ and the existence of a unit vector wzw_zwz​ on {0,1}×BN\{0,1\}\times\mathcal B_N{0,1}×BN​ such that

BAz=p0∣wz⟩⟨wz∣,∑a∈BNwz(c,a)wz(e,a)‾=12 1{c=e}(c,e∈{0,1}).\mathcal B_{A_z}=p_0|w_z\rangle\langle w_z|, \qquad \sum_{a\in\mathcal B_N}w_z(c,a)\overline{w_z(e,a)} =\frac12\,\mathbf1_{\{c=e\}}\quad(c,e\in\{0,1\}).BAz​​=p0​∣wz​⟩⟨wz​∣,a∈BN​∑​wz​(c,a)wz​(e,a)​=21​1{c=e}​(c,e∈{0,1}).

There exists a single family of unitary matrices UzU_zUz​ on the NNN ports, indexed by all z∈ZNz\in\mathcal Z_Nz∈ZN​, such that, for every reference type R\mathcal RR equipped with a finite enumeration and decidable equality and every density matrix TTT on R×{0,1}\mathcal R\times\{0,1\}R×{0,1},

ΔiBEiR(T)=1N+3Tfor every port i,Δ0 ⁣[(IR⊗Uz)BAzR(T)(IR⊗Uz)†]=p0Tfor every refinement z.\begin{aligned} \Delta_i\mathcal B_{E_i}^{\mathcal R}(T) &=\frac1{N+3}T &&\text{for every port }i,\\ \Delta_0\!\left[(I_{\mathcal R}\otimes U_z) \mathcal B_{A_z}^{\mathcal R}(T) (I_{\mathcal R}\otimes U_z)^\dagger\right] &=p_0T &&\text{for every refinement }z. \end{aligned}Δi​BEi​R​(T)Δ0​[(IR​⊗Uz​)BAz​R​(T)(IR​⊗Uz​)†]​=N+31​T=p0​T​​for every port i,for every refinement z.​

The reference type is allowed to be empty; in that case there is no density matrix TTT, so this requirement on TTT is vacuous. The theorem also asserts the completed-effect fidelity identity

∑i=0N−1Re⁡tr⁡(S ΔiBE^i)=NN+3.\sum_{i=0}^{N-1}\operatorname{Re}\operatorname{tr} \bigl(S\,\Delta_i\mathcal B_{\widehat E_i}\bigr) =\frac{N}{N+3}.i=0∑N−1​Retr(SΔi​BEi​​)=N+3N​.

Finally, including the permitted boundary case N=1N=1N=1, it asserts the following identities for the same constructions evaluated with one port:

O1=I,E0(1)=P0(1),v∗(1)=∣01⟩+∣10⟩2,v(0,false)(1)=∣00⟩+∣11⟩2,v(0,true)(1)=∣00⟩−∣11⟩2.O_1=I,\qquad E_0^{(1)}=P_0^{(1)},\qquad v_*^{(1)}=\frac{|01\rangle+|10\rangle}{\sqrt2},\qquad v_{(0,\mathrm{false})}^{(1)}=\frac{|00\rangle+|11\rangle}{\sqrt2},\qquad v_{(0,\mathrm{true})}^{(1)}=\frac{|00\rangle-|11\rangle}{\sqrt2}.O1​=I,E0(1)​=P0(1)​,v∗(1)​=2​∣01⟩+∣10⟩​,v(0,false)(1)​=2​∣00⟩+∣11⟩​,v(0,true)(1)​=2​∣00⟩−∣11⟩​.
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