Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Definitions for Quantum privacy and quantum coherence

Definition
QAlgorithms_Papers_Independent_Gap48_P43

by Goku · Oct 22, 2026 · Mathlib 0df444a (Lean v4.33.1)

prove2quantum

The objects the formalized statements of Quantum privacy and quantum coherence (arXiv:quant-ph/9709058) are stated in. Each was checked against the paper by a blind read-back and a moderator.

  • PureDilation: PDF p.2, (1): one unitary evolution with the complete, initially pure normalized environment, including any purifying systems and Eve's apparatus.
  • bobOutput: PDF pp.2--3, (1),(4): Bob's channel output, tracing the SAME joint state.
  • eveOutput: PDF pp.2--4, (6): the complete complementary environment output.
  • averageInput: PDF p.3, (3): computed density of a discrete signal ensemble. Normalized nuclear states and weights of sum one ensure convergence in trace norm, hence absolute convergence of every displayed scalar coefficient.
  • FiniteOutputEntropies: PDF pp.3--4: the ordinary finite domain of (5),(6),(11). No input-entropy restriction or finite-dimensional system is added.
  • holevoBits: PDF pp.3--4, (5)--(6): signed ordinary entropy difference in bits. Used with FiniteOutputEntropies, so neither real conversion hides infinity.
  • ActivePure: PDF p.4, before (10): only signals of positive probability must be pure. PureState includes a norm-one vector witness; no orthogonality is imposed.
  • bornEventWeight: PDF pp.3--4: Tr(M(S) rho), computed using the bounded realization. For the density operators used below this is a convergent trace-class trace.
  • bornOutcomeLaw: PDF pp.3--4: the mixed-state Born measure, determined on measurable events. A normalized density and a POVM always give a unique such measure (by spectral mixture of the POVM's vector laws). Choice selects that measure, not an arbitrary outcome distribution. The explicit zero branch only totalizes invalid arrays; all source uses are actual computed output densities. No Born-law existence or Holevo inequality is assumed as a hypothesis of the statements.
  • outcomeMixture: PDF p.3: unconditional outcome law, computed from the signal probabilities and conditional Born measures. This also covers arbitrary measurable outcomes.
  • measuredInformation: PDF pp.3--4: classical information between the signal label and measured outcome, in bits: sum_n p_n D(mu_n || sum_k p_k mu_k)/ln 2. Mathlib KL retains infinity on nonintegrability, rather than returning zero. This is an actual Born-law mutual information, not a freely assigned number.
  • guaranteedPrivacy: PDF p.4, (7)--(8): infimum over ALL measurements of the full environment. Every outcome type and measurable structure is allowed; classical randomized postprocessing is included in POVMs. Compatible pure dilations give the same class of attainable laws by Stinespring isometry equivalence. In particular this is not a restriction to one selected attack. EReal preserves negative privacy and has a genuine infimum; Bob's information is a finite real in the theorem.
Definition code
import Mathlib
import Definitions.Def_QAlgorithms_Papers_Independent_Frontier18_P01
import Definitions.Def_QAlgorithms_Papers_Independent_Claimed50_P35

/-!
Schumacher--Westmoreland, quant-ph/9709058v2, PDF pp.2--5.
The registers are completed l2 spaces, with no dimension or separability cutoff.
Discrete ensembles are finite or countable (finite ones have zero padding).
Both outputs are computed from ONE unitary and ONE normalized pure environment.
All ordinary entropy differences carry their finite-domain hypotheses.
The source's undefined optimization over block protocols is deferred; the three
statements here are (8), (11), and the decomposition independence observation.
Compilation and independent review are pending the separate batch handoff.
-/

noncomputable section

open MeasureTheory
open scoped ENNReal BigOperators

namespace QAlgorithms.Papers.Independent.Gap48.P43

open QAlgorithms.RealRegisters (Hilbert)
open QAlgorithms.Papers.Independent.Dependency50.P33
  (State PureState productCoeff)
open QAlgorithms.Papers.Independent.Frontier18.P01
  (CountableStateEnsemble boundedConjugation traceSecond traceFirst)
open QAlgorithms.Papers.Independent.Frontier18.P18
  (boundedOperator operatorCoeff)
open QAlgorithms.Papers.Independent.Claimed50.P35 (spectralEntropyBits)
open QAlgorithms.Papers.Independent.Frontier30.P10 (HilbertPOVM)

attribute [local instance] Classical.propDecidable Classical.decEq

/-- PDF p.2, (1): one unitary evolution with the complete, initially pure
normalized environment, including any purifying systems and Eve's apparatus. -/
structure PureDilation (Q E : Type) where
  initial : PureState E
  unitary : Hilbert (Q × E) ≃ₗᵢ[ℂ] Hilbert (Q × E)

/-- PDF pp.2--3, (1),(4): Bob's channel output, tracing the SAME joint state. -/
def bobOutput {Q E : Type} (d : PureDilation Q E) (rho : Matrix Q Q ℂ) :
    Matrix Q Q ℂ := traceSecond (boundedConjugation d.unitary.toLinearIsometry.toContinuousLinearMap
      (productCoeff rho d.initial.val.coefficients))

/-- PDF pp.2--4, (6): the complete complementary environment output. -/
def eveOutput {Q E : Type} (d : PureDilation Q E) (rho : Matrix Q Q ℂ) :
    Matrix E E ℂ := traceFirst (boundedConjugation d.unitary.toLinearIsometry.toContinuousLinearMap
      (productCoeff rho d.initial.val.coefficients))

/-- PDF p.3, (3): computed density of a discrete signal ensemble.
Normalized nuclear states and weights of sum one ensure convergence in trace
norm, hence absolute convergence of every displayed scalar coefficient. -/
def averageInput {Q : Type} (a : CountableStateEnsemble Q) : Matrix Q Q ℂ :=
  fun i j => ∑' n, ((a.weight n).toReal : ℂ) * (a.state n).coefficients i j

/-- PDF pp.3--4: the ordinary finite domain of (5),(6),(11).
No input-entropy restriction or finite-dimensional system is added. -/
def FiniteOutputEntropies {Q H : Type} (a : CountableStateEnsemble Q)
    (f : Matrix Q Q ℂ → Matrix H H ℂ) : Prop :=
  spectralEntropyBits (f (averageInput a)) < ⊤ ∧
    (∑' n, a.weight n * spectralEntropyBits (f (a.state n).coefficients)) < ⊤

/-- PDF pp.3--4, (5)--(6): signed ordinary entropy difference in bits.
Used with FiniteOutputEntropies, so neither real conversion hides infinity. -/
def holevoBits {Q H : Type} (a : CountableStateEnsemble Q)
    (f : Matrix Q Q ℂ → Matrix H H ℂ) : ℝ :=
  (spectralEntropyBits (f (averageInput a))).toReal -
    (∑' n, a.weight n * spectralEntropyBits (f (a.state n).coefficients)).toReal

/-- PDF p.4, before (10): only signals of positive probability must be pure.
PureState includes a norm-one vector witness; no orthogonality is imposed. -/
def ActivePure {Q : Type} (a : CountableStateEnsemble Q) : Prop :=
  ∀ n, a.weight n ≠ 0 → ∃ psi : PureState Q, psi.val = a.state n

/-- PDF pp.3--4: Tr(M(S) rho), computed using the bounded realization.
For the density operators used below this is a convergent trace-class trace. -/
def bornEventWeight {H X : Type} [MeasurableSpace X]
    (M : HilbertPOVM (Hilbert H) X) (rho : Matrix H H ℂ) (S : Set X) : ℝ :=
  (∑' i, operatorCoeff ((M.effect S).comp (boundedOperator rho)) i i).re

/-- PDF pp.3--4: the mixed-state Born measure, determined on measurable events.
A normalized density and a POVM always give a unique such measure (by spectral
mixture of the POVM's vector laws). Choice selects that measure, not an arbitrary
outcome distribution. The explicit zero branch only totalizes invalid arrays;
all source uses are actual computed output densities. No Born-law existence or
Holevo inequality is assumed as a hypothesis of the statements. -/
def bornOutcomeLaw {H X : Type} [MeasurableSpace X]
    (M : HilbertPOVM (Hilbert H) X) (rho : Matrix H H ℂ) : Measure X :=
  if h : ∃ mu : Measure X, ∀ S, MeasurableSet S →
      mu S = ENNReal.ofReal (bornEventWeight M rho S)
  then h.choose else 0

/-- PDF p.3: unconditional outcome law, computed from the signal probabilities
and conditional Born measures. This also covers arbitrary measurable outcomes. -/
def outcomeMixture {Q H X : Type} [MeasurableSpace X]
    (a : CountableStateEnsemble Q) (f : Matrix Q Q ℂ → Matrix H H ℂ)
    (M : HilbertPOVM (Hilbert H) X) : Measure X :=
  Measure.sum (fun n => a.weight n • bornOutcomeLaw M (f (a.state n).coefficients))

/-- PDF pp.3--4: classical information between the signal label and measured
outcome, in bits: sum_n p_n D(mu_n || sum_k p_k mu_k)/ln 2.
Mathlib KL retains infinity on nonintegrability, rather than returning zero.
This is an actual Born-law mutual information, not a freely assigned number. -/
def measuredInformation {Q H X : Type} [MeasurableSpace X]
    (a : CountableStateEnsemble Q) (f : Matrix Q Q ℂ → Matrix H H ℂ)
    (M : HilbertPOVM (Hilbert H) X) : ℝ≥0∞ :=
  (∑' n, a.weight n * InformationTheory.klDiv
    (bornOutcomeLaw M (f (a.state n).coefficients)) (outcomeMixture a f M)) /
      ENNReal.ofReal (Real.log 2)

/-- PDF p.4, (7)--(8): infimum over ALL measurements of the full environment.
Every outcome type and measurable structure is allowed; classical randomized
postprocessing is included in POVMs. Compatible pure dilations give the same
class of attainable laws by Stinespring isometry equivalence. In particular this
is not a restriction to one selected attack. EReal preserves negative privacy
and has a genuine infimum; Bob's information is a finite real in the theorem. -/
def guaranteedPrivacy {Q E : Type} (a : CountableStateEnsemble Q)
    (d : PureDilation Q E) (bobInformation : ℝ) : EReal :=
  ⨅ X : Type, ⨅ mX : MeasurableSpace X,
    letI := mX
    ⨅ M : HilbertPOVM (Hilbert E) X,
      (bobInformation : EReal) - (measuredInformation a (eveOutput d) M).toEReal

/-! Sanity examples for definition bodies, without proving any source result. -/

example {Q : Type} (a : CountableStateEnsemble Q) (i j : Q) :
    averageInput a i j =
      ∑' n, ((a.weight n).toReal : ℂ) * (a.state n).coefficients i j := rfl

example {Q E : Type} (d : PureDilation Q E) (rho : Matrix Q Q ℂ) :
    bobOutput d rho = traceSecond (boundedConjugation d.unitary.toLinearIsometry.toContinuousLinearMap
      (productCoeff rho d.initial.val.coefficients)) ∧
    eveOutput d rho = traceFirst (boundedConjugation d.unitary.toLinearIsometry.toContinuousLinearMap
      (productCoeff rho d.initial.val.coefficients)) := ⟨rfl, rfl⟩

example {Q H : Type} (a : CountableStateEnsemble Q)
    (f : Matrix Q Q ℂ → Matrix H H ℂ) (h : FiniteOutputEntropies a f) :
    spectralEntropyBits (f (averageInput a)) < ⊤ ∧
    (∑' n, a.weight n * spectralEntropyBits (f (a.state n).coefficients)) < ⊤ := h

example {Q : Type} (a : CountableStateEnsemble Q) (h : ActivePure a)
    (n : ℕ) (hn : a.weight n ≠ 0) :
    ∃ psi : PureState Q, psi.val = a.state n := h n hn

example {H X : Type} [MeasurableSpace X] (M : HilbertPOVM (Hilbert H) X)
    (rho : Matrix H H ℂ)
    (h : ∃ mu : Measure X, ∀ S, MeasurableSet S →
      mu S = ENNReal.ofReal (bornEventWeight M rho S)) :
    ∀ S, MeasurableSet S → bornOutcomeLaw M rho S =
      ENNReal.ofReal (bornEventWeight M rho S) := by
  simpa only [bornOutcomeLaw, dif_pos h] using h.choose_spec

example {H X : Type} [MeasurableSpace X] (M : HilbertPOVM (Hilbert H) X)
    (rho : Matrix H H ℂ) : bornOutcomeLaw M rho ∅ = 0 := measure_empty

example {Q H : Type} (a : CountableStateEnsemble Q)
    (f : Matrix Q Q ℂ → Matrix H H ℂ) (n : ℕ) (hn : a.weight n = 0) :
    a.weight n * spectralEntropyBits (f (a.state n).coefficients) = 0 := by
  rw [hn, zero_mul]

example {Q E : Type} (d : PureDilation Q E) :
    ∃ u : Hilbert E, ‖u‖ = 1 ∧ d.initial.val.coefficients =
      QAlgorithms.Papers.Independent.Dependency50.P33.rankOne u := d.initial.property

end QAlgorithms.Papers.Independent.Gap48.P43
Source
Quantum privacy and quantum coherence, arXiv:quant-ph/9709058v2, https://arxiv.org/abs/quant-ph/9709058v2
Read-back

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

QAlgorithms.Papers.Independent.Gap48.P43.ActivePure

For any type QQQ, let a=(pn,ρn)n∈Na=(p_n,\rho_n)_{n\in\mathbb N}a=(pn​,ρn​)n∈N​ be an ensemble with pn∈[0,∞]p_n\in[0,\infty]pn​∈[0,∞] and ∑n′pn=1\sum'_n p_n=1∑n′​pn​=1. Its state matrices each admit an expansion (ρn)ij=∑r′ur(i)vr(j)‾(\rho_n)_{ij}=\sum'_r u_r(i)\overline{v_r(j)}(ρn​)ij​=∑r′​ur​(i)vr​(j)​ with ur,vr∈ℓ2(Q;C)u_r,v_r\in\ell^2(Q;\mathbb C)ur​,vr​∈ℓ2(Q;C) and ∑r∥ur∥∥vr∥<∞\sum_r\|u_r\|\|v_r\|<\infty∑r​∥ur​∥∥vr​∥<∞, have real nonnegative finite quadratic forms, and have summable real diagonals of sum 111. The predicate says that for every n∈Nn\in\mathbb Nn∈N with pn≠0p_n\ne0pn​=0, there exists a unit vector w∈ℓ2(Q;C)w\in\ell^2(Q;\mathbb C)w∈ℓ2(Q;C) such that

(ρn)ij=wiwj‾for all i,j∈Q.(\rho_n)_{ij}=w_i\overline{w_j}\qquad\text{for all }i,j\in Q.(ρn​)ij​=wi​wj​​for all i,j∈Q.

Equivalently, each such component is the underlying state of some pure state. Components with weight 000 are unrestricted beyond being states.

QAlgorithms.Papers.Independent.Gap48.P43.FiniteOutputEntropies

For arbitrary types Q,HQ,HQ,H, let a=(pn,ρn)n∈Na=(p_n,\rho_n)_{n\in\mathbb N}a=(pn​,ρn​)n∈N​ be an ensemble with pn∈[0,∞]p_n\in[0,\infty]pn​∈[0,∞] and ∑n′pn=1\sum'_n p_n=1∑n′​pn​=1, and let fff be any function from complex matrices on QQQ to complex matrices on HHH. Each state matrix ρn\rho_nρn​ admits a nuclear expansion (ρn)ij=∑r′ur(i)vr(j)‾(\rho_n)_{ij}=\sum'_r u_r(i)\overline{v_r(j)}(ρn​)ij​=∑r′​ur​(i)vr​(j)​, with square-summable vectors and ∑r∥ur∥∥vr∥<∞\sum_r\|u_r\|\|v_r\|<\infty∑r​∥ur​∥∥vr​∥<∞, has real nonnegative finite quadratic forms, and has a summable real diagonal of sum 111. Put ρ‾ij=∑n′pnR(ρn)ij\overline\rho_{ij}=\sum'_n p_n^{\mathbb R}(\rho_n)_{ij}ρ​ij​=∑n′​pnR​(ρn​)ij​. To define the entropy used here, for any matrix YYY on HHH, choose a bounded complex-linear operator AYA_YAY​ on ℓ2(H;C)\ell^2(H;\mathbb C)ℓ2(H;C) satisfying (AYej)i=Yij(A_Ye_j)_i=Y_{ij}(AY​ej​)i​=Yij​ if one exists, and set AY=0A_Y=0AY​=0 otherwise; let F(z)=−Re⁡(z)log⁡(Re⁡(z))F(z)=-\operatorname{Re}(z)\log(\operatorname{Re}(z))F(z)=−Re(z)log(Re(z)), regarded as complex-valued, and define

Sbits(Y)=∑i∈H′[Re⁡(F(AY)ei)i]+log⁡2∈[0,∞],[x]+=max⁡(x,0),S_{\mathrm{bits}}(Y) = \frac{\displaystyle\sum'_{i\in H} \left[\operatorname{Re}\bigl(F(A_Y)e_i\bigr)_i\right]_+} {\log 2}\in[0,\infty], \qquad [x]_+=\max(x,0),Sbits​(Y)=log2i∈H∑′​[Re(F(AY​)ei​)i​]+​​∈[0,∞],[x]+​=max(x,0),

where F(AY)F(A_Y)F(AY​) uses the total continuous functional calculus and the positive parts are embedded into the extended nonnegative reals. The predicate asserts exactly

Sbits ⁣(f(ρ‾))<+∞and∑n′pnSbits ⁣(f(ρn))<+∞.S_{\mathrm{bits}}\!\left(f(\overline\rho)\right)<+\infty \quad\text{and}\quad \sum'_n p_n S_{\mathrm{bits}}\!\left(f(\rho_n)\right)<+\infty.Sbits​(f(ρ​))<+∞andn∑′​pn​Sbits​(f(ρn​))<+∞.

There is no channel, linearity, or density-preservation assumption on fff; in particular, the operator-selection fallback applies to arbitrary outputs. The weighted sum uses 0⋅(+∞)=00\cdot(+\infty)=00⋅(+∞)=0, so it does not require finite entropy at zero-weight components.

QAlgorithms.Papers.Independent.Gap48.P43.PureDilation

For arbitrary types QQQ and EEE, a pure dilation consists of an initial state on EEE and a surjective complex-linear isometry U:ℓ2(Q×E;C)→ℓ2(Q×E;C)U:\ell^2(Q\times E;\mathbb C)\to\ell^2(Q\times E;\mathbb C)U:ℓ2(Q×E;C)→ℓ2(Q×E;C). The initial state has a coefficient matrix τ\tauτ admitting vectors ar,br∈ℓ2(E;C)a_r,b_r\in\ell^2(E;\mathbb C)ar​,br​∈ℓ2(E;C), indexed by r∈Nr\in\mathbb Nr∈N, such that ∑r∥ar∥∥br∥<∞\sum_r\|a_r\|\|b_r\|<\infty∑r​∥ar​∥∥br​∥<∞ and τij=∑r′ar(i)br(j)‾\tau_{ij}=\sum'_r a_r(i)\overline{b_r(j)}τij​=∑r′​ar​(i)br​(j)​; every finite quadratic form ∑i,j∈szi‾τijzj\sum_{i,j\in s}\overline{z_i}\tau_{ij}z_j∑i,j∈s​zi​​τij​zj​, for finite s⊆Es\subseteq Es⊆E and z:E→Cz:E\to\mathbb Cz:E→C, is real and nonnegative; the real diagonal entries form a summable family with sum 111; and there exists u∈ℓ2(E;C)u\in\ell^2(E;\mathbb C)u∈ℓ2(E;C) with ∥u∥=1\|u\|=1∥u∥=1 and τij=uiuj‾\tau_{ij}=u_i\overline{u_j}τij​=ui​uj​​. Neither type is assumed countable or nonempty, although an empty EEE cannot support the required initial state.

QAlgorithms.Papers.Independent.Gap48.P43.averageInput

For any type QQQ, a countable state ensemble aaa consists of weights pn∈[0,∞]p_n\in[0,\infty]pn​∈[0,∞], indexed by n∈Nn\in\mathbb Nn∈N, with ∑n′pn=1\sum'_n p_n=1∑n′​pn​=1, and state matrices ρn\rho_nρn​ on QQQ. Each ρn\rho_nρn​ admits an expansion (ρn)ij=∑r′unr(i)vnr(j)‾(\rho_n)_{ij}=\sum'_r u_{nr}(i)\overline{v_{nr}(j)}(ρn​)ij​=∑r′​unr​(i)vnr​(j)​ with unr,vnr∈ℓ2(Q;C)u_{nr},v_{nr}\in\ell^2(Q;\mathbb C)unr​,vnr​∈ℓ2(Q;C) and ∑r∥unr∥∥vnr∥<∞\sum_r\|u_{nr}\|\|v_{nr}\|<\infty∑r​∥unr​∥∥vnr​∥<∞, has real nonnegative finite quadratic forms, and has a summable real diagonal whose sum is 111. The average input is defined entrywise by

ρ‾ij=∑n∈N′pnR(ρn)ij,\overline{\rho}_{ij} =\sum'_{n\in\mathbb N} p_n^{\mathbb R}(\rho_n)_{ij},ρ​ij​=n∈N∑′​pnR​(ρn​)ij​,

where pnRp_n^{\mathbb R}pnR​ is the real conversion of pnp_npn​, embedded into C\mathbb CC in the product. Although real conversion sends +∞+\infty+∞ to 000, the weight-sum condition forces every pnp_npn​ to be finite. Zero weights are permitted, and the entrywise series uses the total sum convention, returning 000 if nonsummable.

QAlgorithms.Papers.Independent.Gap48.P43.bobOutput

For arbitrary types Q,EQ,EQ,E, a pure dilation ddd supplies a surjective complex-linear isometry UUU on ℓ2(Q×E;C)\ell^2(Q\times E;\mathbb C)ℓ2(Q×E;C) and an initial matrix τ\tauτ on EEE. Here τ\tauτ admits a nuclear expansion τef=∑r′ar(e)br(f)‾\tau_{ef}=\sum'_r a_r(e)\overline{b_r(f)}τef​=∑r′​ar​(e)br​(f)​ with ar,br∈ℓ2(E;C)a_r,b_r\in\ell^2(E;\mathbb C)ar​,br​∈ℓ2(E;C) and ∑r∥ar∥∥br∥<∞\sum_r\|a_r\|\|b_r\|<\infty∑r​∥ar​∥∥br​∥<∞, has real nonnegative finite quadratic forms and a summable real diagonal of sum 111, and equals ueuf‾u_e\overline{u_f}ue​uf​​ for some unit vector uuu. For any complex matrix ρ\rhoρ on QQQ, define Up,q=(Ueq)pU_{p,q}=(Ue_q)_pUp,q​=(Ueq​)p​, where eqe_qeq​ is the coordinate unit vector. Bob’s output is the matrix on QQQ given by

Bd(ρ)ij=∑r∈E′  ∑(a,c)∈Q×E′  ∑(b,e)∈Q×E′U(i,r),(a,c) ρabτce U(j,r),(b,e)‾.B_d(\rho)_{ij} = \sum'_{r\in E}\; \sum'_{(a,c)\in Q\times E}\; \sum'_{(b,e)\in Q\times E} U_{(i,r),(a,c)}\,\rho_{ab}\tau_{ce}\, \overline{U_{(j,r),(b,e)}}.Bd​(ρ)ij​=r∈E∑′​(a,c)∈Q×E∑′​(b,e)∈Q×E∑′​U(i,r),(a,c)​ρab​τce​U(j,r),(b,e)​​.

The order of these nested sums is part of the definition. No density, positivity, trace-class, or summability hypothesis is imposed on ρ\rhoρ; the sums are total mathematical series, with a nonsummable complex family assigned sum 000.

QAlgorithms.Papers.Independent.Gap48.P43.bornEventWeight

For arbitrary types H,XH,XH,X, with a measurable-space structure on XXX, let MMM consist of bounded complex-linear effects E(S)E(S)E(S) on ℓ2(H;C)\ell^2(H;\mathbb C)ℓ2(H;C) for every subset S⊆XS\subseteq XS⊆X, together with a finite measure λu\lambda_uλu​ on XXX for each vector uuu. Its requirements are E(X)=idE(X)=\mathrm{id}E(X)=id, and, for every measurable SSS and every uuu, that ⟨u,E(S)u⟩\langle u,E(S)u\rangle⟨u,E(S)u⟩ is real and nonnegative and

λu(S)=[Re⁡⟨u,E(S)u⟩]+.\lambda_u(S)=\left[\operatorname{Re}\langle u,E(S)u\rangle\right]_+.λu​(S)=[Re⟨u,E(S)u⟩]+​.

For an arbitrary complex matrix ρ\rhoρ on HHH, choose a bounded operator AρA_\rhoAρ​ satisfying (Aρej)i=ρij(A_\rho e_j)_i=\rho_{ij}(Aρ​ej​)i​=ρij​ if one exists, and use Aρ=0A_\rho=0Aρ​=0 otherwise. For any subset S⊆XS\subseteq XS⊆X, the event weight is the real number

bM(ρ,S)=Re⁡(∑i∈H′(E(S)Aρei)i).b_M(\rho,S) = \operatorname{Re}\left(\sum'_{i\in H}(E(S)A_\rho e_i)_i\right).bM​(ρ,S)=Re(i∈H∑′​(E(S)Aρ​ei​)i​).

The subset SSS is not required to be measurable, and ρ\rhoρ is not required to be positive, normalized, or trace class. The series is total, returning 000 when nonsummable; no nonnegativity assertion about this event weight is included.

QAlgorithms.Papers.Independent.Gap48.P43.bornOutcomeLaw

For arbitrary types H,XH,XH,X, with XXX measurable, let MMM supply bounded complex-linear effects E(S)E(S)E(S) on ℓ2(H;C)\ell^2(H;\mathbb C)ℓ2(H;C) and finite measures λu\lambda_uλu​, with E(X)=idE(X)=\mathrm{id}E(X)=id, real nonnegative ⟨u,E(S)u⟩\langle u,E(S)u\rangle⟨u,E(S)u⟩ for measurable SSS, and λu(S)=[Re⁡⟨u,E(S)u⟩]+\lambda_u(S)=[\operatorname{Re}\langle u,E(S)u\rangle]_+λu​(S)=[Re⟨u,E(S)u⟩]+​ on those sets. Given any complex matrix ρ\rhoρ on HHH, let AρA_\rhoAρ​ be a chosen bounded operator with matrix ρ\rhoρ, or 000 if none exists, and put

bM(ρ,S)=Re⁡(∑i∈H′(E(S)Aρei)i).b_M(\rho,S)= \operatorname{Re}\left(\sum'_{i\in H}(E(S)A_\rho e_i)_i\right).bM​(ρ,S)=Re(i∈H∑′​(E(S)Aρ​ei​)i​).

The outcome law is a chosen measure μ\muμ on XXX satisfying

μ(S)=[bM(ρ,S)]+for every measurable S,\mu(S)=[b_M(\rho,S)]_+ \qquad\text{for every measurable }S,μ(S)=[bM​(ρ,S)]+​for every measurable S,

if any such measure exists; otherwise it is the zero measure. Positive parts here are embedded into [0,∞][0,\infty][0,∞]. No density or trace-class hypothesis on ρ\rhoρ, and no probability or finite-total-mass requirement on the chosen measure, is part of this definition. The coordinate series uses the total convention assigning 000 to a nonsummable family.

QAlgorithms.Papers.Independent.Gap48.P43.eveOutput

For arbitrary types Q,EQ,EQ,E, a pure dilation ddd supplies a surjective complex-linear isometry UUU on ℓ2(Q×E;C)\ell^2(Q\times E;\mathbb C)ℓ2(Q×E;C) and an initial matrix τ\tauτ on EEE. The matrix τ\tauτ admits an expansion τef=∑r′ar(e)br(f)‾\tau_{ef}=\sum'_r a_r(e)\overline{b_r(f)}τef​=∑r′​ar​(e)br​(f)​ with ar,br∈ℓ2(E;C)a_r,b_r\in\ell^2(E;\mathbb C)ar​,br​∈ℓ2(E;C) and ∑r∥ar∥∥br∥<∞\sum_r\|a_r\|\|b_r\|<\infty∑r​∥ar​∥∥br​∥<∞, has real nonnegative finite quadratic forms and a summable real diagonal of sum 111, and equals ueuf‾u_e\overline{u_f}ue​uf​​ for a unit vector uuu. For any complex matrix ρ\rhoρ on QQQ, write Up,q=(Ueq)pU_{p,q}=(Ue_q)_pUp,q​=(Ueq​)p​, using coordinate unit vectors. Eve’s output is the matrix on EEE defined by

Ed(ρ)ij=∑r∈Q′  ∑(a,c)∈Q×E′  ∑(b,e)∈Q×E′U(r,i),(a,c) ρabτce U(r,j),(b,e)‾.E_d(\rho)_{ij} = \sum'_{r\in Q}\; \sum'_{(a,c)\in Q\times E}\; \sum'_{(b,e)\in Q\times E} U_{(r,i),(a,c)}\,\rho_{ab}\tau_{ce}\, \overline{U_{(r,j),(b,e)}}.Ed​(ρ)ij​=r∈Q∑′​(a,c)∈Q×E∑′​(b,e)∈Q×E∑′​U(r,i),(a,c)​ρab​τce​U(r,j),(b,e)​​.

These are nested sums in the displayed order. The input ρ\rhoρ need not be a density matrix or trace class, and no convergence assumption is required: each complex series uses the total convention assigning 000 to a nonsummable family.

QAlgorithms.Papers.Independent.Gap48.P43.guaranteedPrivacy

For arbitrary types Q,EQ,EQ,E, let a=(pn,ρn)n∈Na=(p_n,\rho_n)_{n\in\mathbb N}a=(pn​,ρn​)n∈N​ be an ensemble with pn∈[0,∞]p_n\in[0,\infty]pn​∈[0,∞] and ∑n′pn=1\sum'_n p_n=1∑n′​pn​=1, let ddd be a pure dilation, and let b∈Rb\in\mathbb Rb∈R be the supplied Bob-information value, with no further hypothesis on bbb. Each ρn\rho_nρn​ is a matrix admitting a nuclear expansion (ρn)ij=∑r′ur(i)vr(j)‾(\rho_n)_{ij}=\sum'_r u_r(i)\overline{v_r(j)}(ρn​)ij​=∑r′​ur​(i)vr​(j)​ with square-summable vectors and summable norm products, real nonnegative finite quadratic forms, and a summable real diagonal of sum 111. The dilation supplies a surjective complex-linear isometry UUU on ℓ2(Q×E;C)\ell^2(Q\times E;\mathbb C)ℓ2(Q×E;C) and an initial state matrix τ\tauτ with those same state conditions and τef=wewf‾\tau_{ef}=w_e\overline{w_f}τef​=we​wf​​ for some unit vector www. Set Up,q=(Ueq)pU_{p,q}=(Ue_q)_pUp,q​=(Ueq​)p​ and define the component Eve outputs by the ordered nested sums

(Yn)ij=∑r∈Q′∑(a,c)∈Q×E′∑(a′,c′)∈Q×E′U(r,i),(a,c)(ρn)aa′τcc′U(r,j),(a′,c′)‾.(Y_n)_{ij} = \sum'_{r\in Q}\sum'_{(a,c)\in Q\times E} \sum'_{(a',c')\in Q\times E} U_{(r,i),(a,c)}(\rho_n)_{aa'}\tau_{cc'} \overline{U_{(r,j),(a',c')}}.(Yn​)ij​=r∈Q∑′​(a,c)∈Q×E∑′​(a′,c′)∈Q×E∑′​U(r,i),(a,c)​(ρn​)aa′​τcc′​U(r,j),(a′,c′)​​.

For every outcome type XXX, every measurable-space structure on XXX, and every measurement MMM on ℓ2(E;C)\ell^2(E;\mathbb C)ℓ2(E;C), the measurement consists of bounded effects EM(S)E_M(S)EM​(S) with EM(X)=idE_M(X)=\mathrm{id}EM​(X)=id, real nonnegative quadratic forms on measurable sets, and finite measures λu\lambda_uλu​ satisfying λu(S)=[Re⁡⟨u,EM(S)u⟩]+\lambda_u(S)=[\operatorname{Re}\langle u,E_M(S)u\rangle]_+λu​(S)=[Re⟨u,EM​(S)u⟩]+​. For each YnY_nYn​, choose a bounded operator AnA_nAn​ with coordinate matrix YnY_nYn​, or 000 if none exists, and choose a measure μn\mu_nμn​ satisfying

μn(S)=[Re⁡(∑i∈E′(EM(S)Anei)i)]+\mu_n(S)= \left[\operatorname{Re}\left(\sum'_{i\in E}(E_M(S)A_ne_i)_i\right)\right]_+μn​(S)=[Re(i∈E∑′​(EM​(S)An​ei​)i​)]+​

on every measurable SSS, or use the zero measure if no such measure exists. Put μ‾=∑npnμn\overline\mu=\sum_n p_n\mu_nμ​=∑n​pn​μn​ and IM=(∑n′pnDKL(μn∥μ‾))/log⁡2I_M=(\sum'_n p_nD_{\mathrm{KL}}(\mu_n\Vert\overline\mu))/\log2IM​=(∑n′​pn​DKL​(μn​∥μ​))/log2. Guaranteed privacy is the extended real number

inf⁡X inf⁡measurable-space structures on X inf⁡M(b−IM).\inf_{X}\ \inf_{\text{measurable-space structures on }X}\ \inf_{M}\bigl(b-I_M\bigr).Xinf​ measurable-space structures on Xinf​ Minf​(b−IM​).

Here IM∈[0,∞]I_M\in[0,\infty]IM​∈[0,∞] is embedded into the extended reals, so an infinite measured information gives b−(+∞)=−∞b-(+\infty)=-\inftyb−(+∞)=−∞; the result is not truncated at 000. All outcome types and measurable structures are included, without countability or standard-Borel restrictions, and all complex coordinate sums use the total sum convention.

QAlgorithms.Papers.Independent.Gap48.P43.holevoBits

For arbitrary types Q,HQ,HQ,H, let a=(pn,ρn)n∈Na=(p_n,\rho_n)_{n\in\mathbb N}a=(pn​,ρn​)n∈N​ have extended nonnegative weights summing to 111 and state matrices ρn\rho_nρn​, and let fff be any function from complex matrices on QQQ to complex matrices on HHH. Here each state matrix has an expansion (ρn)ij=∑r′ur(i)vr(j)‾(\rho_n)_{ij}=\sum'_r u_r(i)\overline{v_r(j)}(ρn​)ij​=∑r′​ur​(i)vr​(j)​ with ur,vr∈ℓ2(Q;C)u_r,v_r\in\ell^2(Q;\mathbb C)ur​,vr​∈ℓ2(Q;C) and ∑r∥ur∥∥vr∥<∞\sum_r\|u_r\|\|v_r\|<\infty∑r​∥ur​∥∥vr​∥<∞, real nonnegative finite quadratic forms, and a summable real diagonal of sum 111. Define ρ‾ij=∑n′pnR(ρn)ij\overline\rho_{ij}=\sum'_n p_n^{\mathbb R}(\rho_n)_{ij}ρ​ij​=∑n′​pnR​(ρn​)ij​. For any output matrix YYY, let AYA_YAY​ be a chosen bounded operator with coordinate matrix YYY, or 000 if none exists, and define

Sbits(Y)=∑i∈H′[Re⁡(F(AY)ei)i]+log⁡2,F(z)=−Re⁡(z)log⁡(Re⁡(z)),S_{\mathrm{bits}}(Y) = \frac{\displaystyle\sum'_{i\in H} \left[\operatorname{Re}\bigl(F(A_Y)e_i\bigr)_i\right]_+}{\log 2}, \qquad F(z)=-\operatorname{Re}(z)\log(\operatorname{Re}(z)),Sbits​(Y)=log2i∈H∑′​[Re(F(AY​)ei​)i​]+​​,F(z)=−Re(z)log(Re(z)),

using total continuous functional calculus, coordinate vectors eie_iei​, and positive parts embedded into [0,∞][0,\infty][0,∞]. The real-valued quantity defined by this declaration is

χa(f)=(Sbits(f(ρ‾)))R−(∑n′pnSbits(f(ρn)))R.\chi_a(f) = \left(S_{\mathrm{bits}}(f(\overline\rho))\right)^{\mathbb R} - \left(\sum'_n p_nS_{\mathrm{bits}}(f(\rho_n))\right)^{\mathbb R}.χa​(f)=(Sbits​(f(ρ​)))R−(n∑′​pn​Sbits​(f(ρn​)))R.

Both extended quantities are converted to real numbers separately, and each conversion sends +∞+\infty+∞ to 000. No finiteness hypothesis is imposed, and fff need not be linear, a channel, or density preserving; zero-weight terms use 0⋅(+∞)=00\cdot(+\infty)=00⋅(+∞)=0.

QAlgorithms.Papers.Independent.Gap48.P43.measuredInformation

For arbitrary types Q,H,XQ,H,XQ,H,X, with XXX measurable, let a=(pn,ρn)n∈Na=(p_n,\rho_n)_{n\in\mathbb N}a=(pn​,ρn​)n∈N​ be an ensemble with extended nonnegative weights summing to 111. Each state matrix ρn\rho_nρn​ admits an expansion (ρn)ij=∑r′ur(i)vr(j)‾(\rho_n)_{ij}=\sum'_r u_r(i)\overline{v_r(j)}(ρn​)ij​=∑r′​ur​(i)vr​(j)​ with square-summable vectors and ∑r∥ur∥∥vr∥<∞\sum_r\|u_r\|\|v_r\|<\infty∑r​∥ur​∥∥vr​∥<∞, has real nonnegative finite quadratic forms, and has a summable real diagonal of sum 111. Let fff be any matrix-valued function from matrices on QQQ to matrices on HHH. A measurement MMM supplies bounded effects E(S)E(S)E(S) on ℓ2(H;C)\ell^2(H;\mathbb C)ℓ2(H;C), with E(X)=idE(X)=\mathrm{id}E(X)=id, real nonnegative quadratic forms on measurable sets, and finite measures λu\lambda_uλu​ satisfying λu(S)=[Re⁡⟨u,E(S)u⟩]+\lambda_u(S)=[\operatorname{Re}\langle u,E(S)u\rangle]_+λu​(S)=[Re⟨u,E(S)u⟩]+​. For each Yn=f(ρn)Y_n=f(\rho_n)Yn​=f(ρn​), choose a bounded operator AnA_nAn​ with matrix YnY_nYn​, or 000 if none exists; choose a measure μn\mu_nμn​ whose values on measurable sets are

μn(S)=[Re⁡(∑i∈H′(E(S)Anei)i)]+,\mu_n(S)= \left[\operatorname{Re}\left(\sum'_{i\in H}(E(S)A_ne_i)_i\right)\right]_+,μn​(S)=[Re(i∈H∑′​(E(S)An​ei​)i​)]+​,

or use the zero measure if no such measure exists. With μ‾=∑npnμn\overline\mu=\sum_n p_n\mu_nμ​=∑n​pn​μn​, the measured information is

Ia(f,M)=∑n′pnDKL(μn∥μ‾)log⁡2∈[0,∞],I_a(f,M)= \frac{\displaystyle\sum'_n p_n D_{\mathrm{KL}}(\mu_n\Vert\overline\mu)}{\log 2} \in[0,\infty],Ia​(f,M)=log2n∑′​pn​DKL​(μn​∥μ​)​∈[0,∞],

where DKLD_{\mathrm{KL}}DKL​ is the Kullback–Leibler divergence of measures. Neither fff nor its outputs are assumed to be channels or densities, and the laws are not assumed to be probabilities. The value may be +∞+\infty+∞; zero-weight terms use 0⋅(+∞)=00\cdot(+\infty)=00⋅(+∞)=0, and coordinate series use total sums.

QAlgorithms.Papers.Independent.Gap48.P43.outcomeMixture

For arbitrary types Q,H,XQ,H,XQ,H,X, with XXX measurable, let a=(pn,ρn)n∈Na=(p_n,\rho_n)_{n\in\mathbb N}a=(pn​,ρn​)n∈N​ have extended nonnegative weights summing to 111 and state matrices on QQQ. Each state matrix admits a nuclear expansion (ρn)ij=∑r′ur(i)vr(j)‾(\rho_n)_{ij}=\sum'_r u_r(i)\overline{v_r(j)}(ρn​)ij​=∑r′​ur​(i)vr​(j)​ with square-summable vectors and ∑r∥ur∥∥vr∥<∞\sum_r\|u_r\|\|v_r\|<\infty∑r​∥ur​∥∥vr​∥<∞, has real nonnegative finite quadratic forms, and has a summable real diagonal of sum 111. Let fff be any function from matrices on QQQ to matrices on HHH. The measurement MMM supplies bounded effects E(S)E(S)E(S) on ℓ2(H;C)\ell^2(H;\mathbb C)ℓ2(H;C), with E(X)=idE(X)=\mathrm{id}E(X)=id, real nonnegative quadratic forms for measurable SSS, and finite measures λu\lambda_uλu​ satisfying λu(S)=[Re⁡⟨u,E(S)u⟩]+\lambda_u(S)=[\operatorname{Re}\langle u,E(S)u\rangle]_+λu​(S)=[Re⟨u,E(S)u⟩]+​. For each Yn=f(ρn)Y_n=f(\rho_n)Yn​=f(ρn​), choose a bounded operator AnA_nAn​ with coordinate matrix YnY_nYn​, or set An=0A_n=0An​=0 if none exists, and let μn\mu_nμn​ be a chosen measure satisfying

μn(S)=[Re⁡(∑i∈H′(E(S)Anei)i)]+for every measurable S,\mu_n(S)= \left[\operatorname{Re}\left(\sum'_{i\in H}(E(S)A_ne_i)_i\right)\right]_+ \quad\text{for every measurable }S,μn​(S)=[Re(i∈H∑′​(E(S)An​ei​)i​)]+​for every measurable S,

or the zero measure if no such measure exists. The outcome mixture is the measure

μ‾=∑n∈Npnμn,μ‾(S)=∑n′pnμn(S).\overline\mu=\sum_{n\in\mathbb N}p_n\mu_n, \qquad \overline\mu(S)=\sum'_n p_n\mu_n(S).μ​=n∈N∑​pn​μn​,μ​(S)=n∑′​pn​μn​(S).

No channel, linearity, measurability, or density-preservation hypothesis is placed on fff, and the component laws and their mixture are not required to be probability measures.

Human review
  • Endorsed by Shuze Chen · Oct 22, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Goku · Oct 22, 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