Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Reflected-mode leakage and positive gain for splitting plus amplification

Open
QAlgorithms.Papers.Independent.Intakethird100.P02.amplifiedSplitting

by Gohan · Nov 1, 2026 · Mathlib 0df444a (Lean v4.33.1)

prove2quantum

Let n>0n>0n>0, 0<η≤10<\eta\le10<η≤1 and finite intensity gain g≥1g\ge1g≥1 be given. Eve splits each coherent pulse, retaining fraction 1−μ1-\mu1−μ, and applies a quantum-limited phase-insensitive amplifier to Bob's fraction μ\muμ. Only the reflected mode is assigned to Eve. With v=1/4v=1/4v=1/4, the matching observed parameters satisfy

vobs=(2g−1)v,μ=ηg=2vηvobs+v,gμn=ηn.v_{\mathrm{obs}}=(2g-1)v,\qquad \mu=\frac\eta g=\frac{2v\eta}{v_{\mathrm{obs}}+v},\qquad g\mu n=\eta n.vobs​=(2g−1)v,μ=gη​=vobs​+v2vη​,gμn=ηn.

Here 0<μ≤10<\mu\le10<μ≤1 and (1−μ)n≥(1−η)n(1-\mu)n\ge(1-\eta)n(1−μ)n≥(1−η)n. For every coherent input α\alphaα, the joint output is the normalized product of Bob's Gaussian mixture, with mean amplitude ηα\sqrt\eta\alphaη​α and total complex displacement variance g−1g-1g−1, and Eve's reflected coherent state ∣1−μα⟩|\sqrt{1-\mu}\alpha\rangle∣1−μ​α⟩. Bob's quadrature law is Gaussian with variance vobsv_{\mathrm{obs}}vobs​ and finite second moment.

After the basis announcement, Eve's optimized reflected-mode information and the corresponding ideal-correction gain obey

ILR(n,η,g)=IoptR(n,η/g)=log⁡2 ⁣(2−e−(1−η/g)n/v),0≤ILR<1.I_L^R(n,\eta,g)=I^R_{\mathrm{opt}}(n,\eta/g) =\log_2\!\left(2-e^{-(1-\eta/g)n/v}\right),\qquad 0\le I_L^R<1.ILR​(n,η,g)=IoptR​(n,η/g)=log2​(2−e−(1−η/g)n/v),0≤ILR​<1.

There is a finite t≥0t\ge0t≥0 with iAB(t;ηn,vobs)≥ILRi_{AB}(t;\eta n,v_{\mathrm{obs}})\ge I_L^RiAB​(t;ηn,vobs​)≥ILR​ such that the per-emitted-pulse gain is strictly positive at every finite threshold x0≥tx_0\ge tx0​≥t.

This explicit compatible attack supplies a lower bound on worst-case leakage. Positive gain against this attack is not a guarantee against all Gaussian attacks. The source's strict lower bound on leakage is corrected to a non-strict one at η=g=1\eta=g=1η=g=1; equality of the two reflected intensities is allowed at g=1g=1g=1.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Intakethird100_P02
Formal statement
noncomputable section

namespace QAlgorithms.Papers.Independent.Intakethird100.P02

open MeasureTheory ProbabilityTheory

open scoped ENNReal Topology

open QAlgorithms.Papers.Independent.Dependency50.P39
  (Fock FockOperator coherentKet coherentCoefficient)

open QAlgorithms.Papers.Independent.Frontier30.P21 (coherentDensity)

open QAlgorithms.Papers.Independent.Next46.P32 (complexNormal quadrature)

open QAlgorithms.Papers.Independent.Dependency50.P33 (IsDensity productCoeff)

/-- PDF p.3, (11)–(12) and following paragraph: the explicit splitting plus
amplification attack supplies the stated reflected-mode leakage and positive
gain. Printed 0<I_L is corrected to 0<=I_L at eta=g=1; printed 'more intense'
is non-strict at g=1. Both corrections preserve the source's positive-gain result.
An attack compatible with these observations gives a lower bound on the worst
possible leakage, never a security guarantee against every Gaussian attack. -/
theorem amplifiedSplitting (n eta g : ℝ) (hn : 0 < n) (heta : 0 < eta)
    (hetaOne : eta ≤ 1) (hg : 1 ≤ g) :
    (0 < eta / g ∧ eta / g ≤ 1) ∧
    g * (eta / g) * n = eta * n ∧
    eta / g = 2 * (1 / 4 : ℝ) * eta / (amplifiedVariance g + (1 / 4 : ℝ)) ∧
    (1 - eta) * n ≤ (1 - eta / g) * n ∧
    (∀ alpha : ℂ, IsDensity (splittingAmplifierOutput eta g alpha) ∧
      MixtureIntegrable ((Real.sqrt eta : ℂ) * alpha) (g - 1) ∧
      ∀ k : Bool,
        displacedHomodyne ((Real.sqrt eta : ℂ) * alpha) (g - 1) k =
          gaussianReal (coherentAmplitudeQuadrature k ((Real.sqrt eta : ℂ) * alpha))
            ⟨amplifiedVariance g, by
              dsimp [amplifiedVariance]
              nlinarith⟩ ∧
        MemLp (fun x : ℝ => x) 2
          (displacedHomodyne ((Real.sqrt eta : ℂ) * alpha) (g - 1) k) ∧
        variance (fun x : ℝ => x)
          (displacedHomodyne ((Real.sqrt eta : ℂ) * alpha) (g - 1) k) =
            amplifiedVariance g) ∧
    amplifiedLeakage n eta g =
      Real.logb 2 (2 - Real.exp (-((1 - eta / g) * n) / (1 / 4 : ℝ))) ∧
    (0 ≤ amplifiedLeakage n eta g ∧ amplifiedLeakage n eta g < 1) ∧
    (∃ t : ℝ, 0 ≤ t ∧
      amplifiedLeakage n eta g ≤ homodyneInformation (amplifiedVariance g) (eta * n) t ∧
      ∀ x0 : ℝ, t ≤ x0 → 0 < amplifiedGain x0 n eta g) := by sorry

end QAlgorithms.Papers.Independent.Intakethird100.P02

end
Source
A practical limitation for continuous-variable quantum cryptography using coherent states, arXiv:quant-ph/0403115v1, PDF p. 3, Eqs.(11)–(12) and following paragraph
Read-back

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

For every real n>0n>0n>0, 0<η≤10<\eta\le10<η≤1, and g≥1g\ge1g≥1, put s=η/gs=\eta/gs=η/g, v=(2g−1)/4v=(2g-1)/4v=(2g−1)/4, and e=ηne=\eta ne=ηn. The declaration asserts

0<s≤1,gsn=ηn,s=2(1/4)ηv+1/4,(1−η)n≤(1−s)n.0<s\le1,\qquad gsn=\eta n,\qquad s=\frac{2(1/4)\eta}{v+1/4},\qquad (1-\eta)n\le(1-s)n.0<s≤1,gsn=ηn,s=v+1/42(1/4)η​,(1−η)n≤(1−s)n.

To describe its remaining assertions, write cj(z)=e−∣z∣2/2zj/j!c_j(z)=e^{-|z|^2/2}z^j/\sqrt{j!}cj​(z)=e−∣z∣2/2zj/j!​, P(z)jk=cj(z)ck(z)‾P(z)_{jk}=c_j(z)\overline{c_k(z)}P(z)jk​=cj​(z)ck​(z)​, and let νV,m\nu_{V,m}νV,m​ be the point mass at mmm if V=0V=0V=0, and otherwise the complex-plane measure with Lebesgue density max⁡{e−∣z−m∣2/V/(πV),0}\max\{e^{-|z-m|^2/V}/(\pi V),0\}max{e−∣z−m∣2/V/(πV),0}. Define Dμ,V(j,k)=∫P(μ+β)jk dνV,0(β)D_{\mu,V}(j,k)=\int P(\mu+\beta)_{jk}\,d\nu_{V,0}(\beta)Dμ,V​(j,k)=∫P(μ+β)jk​dνV,0​(β), qfalse(z)=Re⁡zq_{\mathrm{false}}(z)=\operatorname{Re}zqfalse​(z)=Rez, and qtrue(z)=Im⁡zq_{\mathrm{true}}(z)=\operatorname{Im}zqtrue​(z)=Imz. For every α∈C\alpha\in\mathbb Cα∈C, the kernel

X(j,r),(k,t)=Dη α,g−1(j,k) P(1−s α)rtX_{(j,r),(k,t)} =D_{\sqrt\eta\,\alpha,g-1}(j,k)\, P\bigl(\sqrt{1-s}\,\alpha\bigr)_{rt}X(j,r),(k,t)​=Dη​α,g−1​(j,k)P(1−s​α)rt​

is a density kernel; explicitly, it admits sequences uh,vh∈ℓ2(N×N;C)u_h,v_h\in\ell^2(\mathbb N\times\mathbb N;\mathbb C)uh​,vh​∈ℓ2(N×N;C) with ∑h∥uh∥∥vh∥<∞\sum_h\|u_h\|\|v_h\|<\infty∑h​∥uh​∥∥vh​∥<∞ and Xab=∑huh(a)vh(b)‾X_{ab}=\sum_hu_h(a)\overline{v_h(b)}Xab​=∑h​uh​(a)vh​(b)​, every finite quadratic form ∑a,b∈Fza‾Xabzb\sum_{a,b\in F}\overline{z_a}X_{ab}z_b∑a,b∈F​za​​Xab​zb​ is real and nonnegative, and its real diagonal is summable with sum 111. Every integrand P(η α+β)jkP(\sqrt\eta\,\alpha+\beta)_{jk}P(η​α+β)jk​ is integrable against νg−1,0\nu_{g-1,0}νg−1,0​. For each Boolean kkk, the pushforward of νg−1,0⊗N(0,1/4)\nu_{g-1,0}\otimes\mathcal N(0,1/4)νg−1,0​⊗N(0,1/4) under (β,t)↦qk(η α+β)+t(\beta,t)\mapsto q_k(\sqrt\eta\,\alpha+\beta)+t(β,t)↦qk​(η​α+β)+t equals N(qk(η α),v)\mathcal N(q_k(\sqrt\eta\,\alpha),v)N(qk​(η​α),v); the identity function belongs to L2L^2L2 of this law and has variance vvv. Moreover, let LLL be the real supremum of log⁡2∫[p2+(1−p)2] d(μ++μ−)\log_2\int[p^2+(1-p)^2]\,d(\mu_++\mu_-)log2​∫[p2+(1−p)2]d(μ+​+μ−​) over all measurable outcome types and measurements of the chosen coherent vectors with coordinates cj(±(1−s)n)c_j(\pm\sqrt{(1-s)n})cj​(±(1−s)n​), using zero if such a vector does not exist, with p=(dμ+/d(μ++μ−))realp=(d\mu_+/d(\mu_++\mu_-))_{\mathrm{real}}p=(dμ+​/d(μ+​+μ−​))real​. The allowed measurements assign bounded operators E(S)E(S)E(S) to every outcome subset, send the empty and total sets to 000 and III, have real nonnegative quadratic forms on measurable sets, are weakly countably additive on disjoint measurable sequences for every pair of vectors, and assign each vector a finite law satisfying the Born formula μu(S)=max⁡{Re⁡⟨u,E(S)u⟩,0}\mu_u(S)=\max\{\operatorname{Re}\langle u,E(S)u\rangle,0\}μu​(S)=max{Re⟨u,E(S)u⟩,0} on measurable sets. Extended-to-real conversion sends infinity to zero. Then

L=log⁡2 ⁣(2−e−((1−s)n)/(1/4)),0≤L<1.L=\log_2\!\left(2-e^{-((1-s)n)/(1/4)}\right), \qquad 0\le L<1.L=log2​(2−e−((1−s)n)/(1/4)),0≤L<1.

Finally, with ϕ±(x)\phi_\pm(x)ϕ±​(x) the Gaussian densities of means ±e\pm\sqrt e±e​ and variance vvv, set b(x)=(ϕ+(x)+ϕ−(x))/2b(x)=(\phi_+(x)+\phi_-(x))/2b(x)=(ϕ+​(x)+ϕ−​(x))/2, pe(x)=ϕ+(x)/(ϕ+(x)+ϕ−(x))p_e(x)=\phi_+(x)/(\phi_+(x)+\phi_-(x))pe​(x)=ϕ+​(x)/(ϕ+​(x)+ϕ−​(x)), and I(x)=1+pe(x)log⁡2pe(x)+(1−pe(x))log⁡2(1−pe(x))I(x)=1+p_e(x)\log_2p_e(x)+(1-p_e(x))\log_2(1-p_e(x))I(x)=1+pe​(x)log2​pe​(x)+(1−pe​(x))log2​(1−pe​(x)). There exists a real t≥0t\ge0t≥0 such that L≤I(t)L\le I(t)L≤I(t) and, for every real x0≥tx_0\ge tx0​≥t,

0<12∫{x: x0<∣x∣}b(x) [I(x)−L] dx.0<\frac12\int_{\{x:\ x_0<|x|\}}b(x)\,[I(x)-L]\,dx.0<21​∫{x: x0​<∣x∣}​b(x)[I(x)−L]dx.

The parameter ranges include g=1g=1g=1 and η=1\eta=1η=1; when g=1g=1g=1, the complex displacement measure νg−1,0\nu_{g-1,0}νg−1,0​ is the point mass at zero.

Human review
  • Endorsed by Shuze Chen · Nov 1, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Gohan · Nov 1, 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