Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Eqs.(11),(13): exact coherent Chernoff/Bhattacharyya bounds

Open
QAlgorithms.Papers.Independent.Intakenext100.P85.coherent_exact_bounds

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

prove2quantum

Let NS,NB≥0N_S,N_B\ge0NS​,NB​≥0, 0<κ<10<\kappa<10<κ<1, and M≥1M\ge1M≥1. Transmit MMM coherent modes each of real amplitude NS\sqrt{N_S}NS​​ through independent copies of the target channel. The hypotheses have priors 1/21/21/2, and reception permits every joint quantum measurement. The one-mode returned states ρ0,ρ1\rho_0,\rho_1ρ0​,ρ1​ are Gaussian, with means (0,0)(0,0)(0,0) and (κNS,0)(\sqrt{\kappa N_S},0)(κNS​​,0) and common covariance (2NB+1)I/4(2N_B+1)I/4(2NB​+1)I/4. Set

c=κNS(NB+1−NB)2.c=\kappa N_S(\sqrt{N_B+1}-\sqrt{N_B})^2.c=κNS​(NB​+1​−NB​​)2.

Their actual operator overlap satisfies Tr⁡(ρ01/2ρ11/2)=e−c\operatorname{Tr}(\rho_0^{1/2}\rho_1^{1/2})=e^{-c}Tr(ρ01/2​ρ11/2​)=e−c, and the optimized quantum Chernoff bound equals the Bhattacharyya bound e−Mc/2e^{-Mc}/2e−Mc/2. The minimum equal-prior error obeys

1−1−e−2Mc2≤Pe,CS(M)≤e−Mc2.\frac{1-\sqrt{1-e^{-2Mc}}}{2}\le P_{e,\mathrm{CS}}^{(M)}\le\frac{e^{-Mc}}2.21−1−e−2Mc​​≤Pe,CS(M)​≤2e−Mc​.

All states are constructed from the coherent input and physical channel; Gaussian moments and bounded realizations are conclusions. These finite-parameter bounds remain valid at zero signal energy or zero background.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Intakenext100_P85
Formal statement
noncomputable section

open MeasureTheory Filter

open scoped BigOperators ENNReal Topology

attribute [local instance] Classical.propDecidable Classical.decEq

namespace QAlgorithms.Papers.Independent.Intakenext100.P85

open QAlgorithms.Papers.Independent.Claimed50.P19 (Cov)

open QAlgorithms.Papers.Independent.Renew46.P41 (coherentProductCoefficient)

/-- Source p.3, Eqs.(11),(13): exact coherent-state overlap, Chernoff
minimum and finite-copy equal-prior error sandwich. All joint effects are
allowed. Real source means are (sqrt(k NS),0), not (sqrt(k)*NS,0). -/
theorem coherent_exact_bounds (k NS NB : ℝ) (hk : 0 < k) (hk1 : k < 1)
    (hNS : 0 ≤ NS) (hNB : 0 ≤ NB) (M : ℕ) (hM : 0 < M)
    (W : G.FockWeylSystem 1) :
    let alpha := constantAmplitude 1 NS
    let rho0 := coherentReturn false k NB alpha
    let rho1 := coherentReturn true k NB alpha
    (∀ present : Bool,
      BlockRealized M present k NB (coherentArray (constantAmplitude M NS)) ∧
      T.IsDensity (coherentReturn present k NB (constantAmplitude M NS)) ∧
      HasArray (realizeOccupationArray (coherentReturn present k NB (constantAmplitude M NS)))
        (coherentReturn present k NB (constantAmplitude M NS))) ∧
    SourceGaussian W rho0 0 (Matrix.diagonal (fun _ => bathB NB / 4)) ∧
    SourceGaussian W rho1
      (fun i => if i.2 = 0 then Real.sqrt (k * NS) else 0)
      (Matrix.diagonal (fun _ => bathB NB / 4)) ∧
    G.chernoffTrace (realizeOccupationArray rho0) (realizeOccupationArray rho1) (1 / 2) =
      Real.exp (-coherentExponent k NS NB) ∧
    G.quantumChernoff (realizeOccupationArray rho0) (realizeOccupationArray rho1) M =
      ENNReal.ofReal (Real.exp (-(M : ℝ) * coherentExponent k NS NB) / 2) ∧
    ENNReal.ofReal (coherentLower k ((M : ℝ) * NS) NB) ≤
      optimumError (coherentReturn false k NB (constantAmplitude M NS))
        (coherentReturn true k NB (constantAmplitude M NS)) ∧
    optimumError (coherentReturn false k NB (constantAmplitude M NS))
        (coherentReturn true k NB (constantAmplitude M NS)) ≤
      ENNReal.ofReal (Real.exp (-(M : ℝ) * coherentExponent k NS NB) / 2) := by sorry

end QAlgorithms.Papers.Independent.Intakenext100.P85

end
Source
Quantum Illumination with Gaussian States, arXiv:0810.0534v1, PDF p. 3, Eqs.(11),(13): exact coherent Chernoff/Bhattacharyya bounds
Read-back

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

For all real 0<k<10<k<10<k<1, NS,NB≥0N_S,N_B\ge0NS​,NB​≥0, every positive integer MMM, and every one-mode Weyl system WWW specified below, the following realization, Gaussian, and error statements hold. For mmm modes and amplitude α∈Cm\alpha\in\mathbb C^mα∈Cm, put

cα(j)=∏r=0m−1e−∣αr∣2/2αrjrjr!,C(α)ij=cα(i)cα(j)‾.c_\alpha(j)=\prod_{r=0}^{m-1}e^{-|\alpha_r|^2/2}\frac{\alpha_r^{j_r}}{\sqrt{j_r!}}, \qquad C(\alpha)_{ij}=c_\alpha(i)\overline{c_\alpha(j)}.cα​(j)=r=0∏m−1​e−∣αr​∣2/2jr​!​αrjr​​​,C(α)ij​=cα​(i)cα​(j)​.

For any occupation kernel AAA, let R(A)R(A)R(A) be a chosen bounded operator on the corresponding ℓ2\ell^2ℓ2-space with matrix AAA, or zero if none exists. Define wt(h)=(t/(t+1))h/(t+1)w_t(h)=(t/(t+1))^h/(t+1)wt​(h)=(t/(t+1))h/(t+1) and

βη(m,ℓ;p,q)=1m+ℓ=p+qm!ℓ!p!q!∑0≤r≤pr≤m, m−r≤q(pr)(qm−r)(η)r+q−(m−r)(−1−η)p−r(1−η)m−r.\beta_\eta(m,\ell;p,q)= \mathbf1_{m+\ell=p+q}\sqrt{\frac{m!\ell!}{p!q!}} \sum_{\substack{0\le r\le p\\r\le m,\ m-r\le q}} \binom pr\binom q{m-r} (\sqrt\eta)^{r+q-(m-r)} (-\sqrt{1-\eta})^{p-r}(\sqrt{1-\eta})^{m-r}.βη​(m,ℓ;p,q)=1m+ℓ=p+q​p!q!m!ℓ!​​0≤r≤pr≤m, m−r≤q​∑​(rp​)(m−rq​)(η​)r+q−(m−r)(−1−η​)p−r(1−η​)m−r.

The single-coordinate maps used are

L0(A)mn=1m=nwNB(m)∑aAaa,L1(A)mn=∑(h,ℓ)∈N2h≤m+ℓ, h≤n+ℓwNB/(1−k)(h) βk(m,ℓ;m+ℓ−h,h) βk(n,ℓ;n+ℓ−h,h) R(A)m+ℓ−h,n+ℓ−h.L_0(A)_{mn}=\mathbf1_{m=n}w_{N_B}(m)\sum_aA_{aa},\qquad L_1(A)_{mn}= \sum_{\substack{(h,\ell)\in\mathbb N^2\\h\le m+\ell,\ h\le n+\ell}} w_{N_B/(1-k)}(h)\, \beta_k(m,\ell;m+\ell-h,h)\, \beta_k(n,\ell;n+\ell-h,h)\, R(A)_{m+\ell-h,n+\ell-h}.L0​(A)mn​=1m=n​wNB​​(m)a∑​Aaa​,L1​(A)mn​=(h,ℓ)∈N2h≤m+ℓ, h≤n+ℓ​∑​wNB​/(1−k)​(h)βk​(m,ℓ;m+ℓ−h,h)βk​(n,ℓ;n+ℓ−h,h)R(A)m+ℓ−h,n+ℓ−h​.

To apply LpL_pLp​ to coordinate rrr of an mmm-mode kernel XXX, its (i,j)(i,j)(i,j)-entry is obtained by applying LpL_pLp​ to the slice Aab=Xi[r:=a],j[r:=b]A_{ab}=X_{i[r:=a],j[r:=b]}Aab​=Xi[r:=a],j[r:=b]​ and taking entry (ir,jr)(i_r,j_r)(ir​,jr​); apply these coordinate maps in order r=0,…,m−1r=0,\ldots,m-1r=0,…,m−1, and denote the result on C(α)C(\alpha)C(α) by Yp(α)Y_p(\alpha)Yp​(α). All sums use the total infinite-sum definitions. Set αr(m)=NS\alpha^{(m)}_r=\sqrt{N_S}αr(m)​=NS​​, ρp=Yp(α(1))\rho_p=Y_p(\alpha^{(1)})ρp​=Yp​(α(1)), and

e=kNS(NB+1−NB)2.e=kN_S\bigl(\sqrt{N_B+1}-\sqrt{N_B}\bigr)^2.e=kNS​(NB​+1​−NB​​)2.

For both p=0,1p=0,1p=0,1, every coordinate slice of every intermediate kernel, including the initial and final stages, in the MMM-coordinate computation from C(α(M))C(\alpha^{(M)})C(α(M)) is realized with exactly its entries by its chosen bounded operator; Yp(α(M))Y_p(\alpha^{(M)})Yp​(α(M)) is a density kernel; and R(Yp(α(M)))R(Y_p(\alpha^{(M)}))R(Yp​(α(M))) has precisely that matrix. Here a density kernel has an expansion ∑rur(i)vr(j)‾\sum_r u_r(i)\overline{v_r(j)}∑r​ur​(i)vr​(j)​ with summable products of ℓ2\ell^2ℓ2-norms, real nonnegative finite quadratic forms, and summable real diagonal of sum one. The one-mode Weyl operators W(ξ)W(\xi)W(ξ) are assumed unitary, with entries dmn(−ξ1+iξ0)d_{mn}(-\xi_1+i\xi_0)dmn​(−ξ1​+iξ0​), where

dmn(z)=e−∣z∣2/2m! n!∑h=0min⁡(m,n)zm−h(−z‾)n−hh!(m−h)!(n−h)!.d_{mn}(z)=e^{-|z|^2/2}\sqrt{m!\,n!} \sum_{h=0}^{\min(m,n)} \frac{z^{m-h}(-\overline z)^{n-h}}{h!(m-h)!(n-h)!}.dmn​(z)=e−∣z∣2/2m!n!​h=0∑min(m,n)​h!(m−h)!(n−h)!zm−h(−z)n−h​.

For p=0,1p=0,1p=0,1, R(ρp)R(\rho_p)R(ρp​) has matrix ρp\rho_pρp​, is positive and self-adjoint, admits a pointwise nuclear representation Tx=∑r⟨vr,x⟩urTx=\sum_r\langle v_r,x\rangle u_rTx=∑r​⟨vr​,x⟩ur​ with summable products of norms, and has summable diagonal of trace one. With V=2NB+14I2V=\frac{2N_B+1}{4}I_2V=42NB​+1​I2​, its covariance assertions require 4V4V4V symmetric positive definite and 4V+iJ4V+iJ4V+iJ positive semidefinite, J=(01−10)J=\left(\begin{smallmatrix}0&1\\-1&0\end{smallmatrix}\right)J=(0−1​10​); every trace diagonal below is summable, and

tr⁡(R(ρ0)W(ξ))=e−2ξTVξ,tr⁡(R(ρ1)W(ξ))=e−2ξTVξ+2ikNS ξ0.\operatorname{tr}(R(\rho_0)W(\xi))=e^{-2\xi^{\mathsf T}V\xi}, \qquad \operatorname{tr}(R(\rho_1)W(\xi)) =e^{-2\xi^{\mathsf T}V\xi+2i\sqrt{kN_S}\,\xi_0}.tr(R(ρ0​)W(ξ))=e−2ξTVξ,tr(R(ρ1​)W(ξ))=e−2ξTVξ+2ikNS​​ξ0​.

Writing TsT^sTs for the real power defined by continuous functional calculus and tr⁡T=∑j⟨ej,Tej⟩\operatorname{tr}T=\sum_j\langle e_j,Te_j\rangletrT=∑j​⟨ej​,Tej​⟩, the exact trace and infimum identities are

Re⁡tr⁡ ⁣(R(ρ0)1/2R(ρ1)1/2)=e−e,12(inf⁡0<s<1[Re⁡tr⁡ ⁣(R(ρ0)sR(ρ1)1−s)]+)M=[e−Me2]+,\operatorname{Re}\operatorname{tr}\!\left(R(\rho_0)^{1/2}R(\rho_1)^{1/2}\right)=e^{-e}, \qquad \frac12\left( \inf_{0<s<1} \left[\operatorname{Re}\operatorname{tr}\!\left(R(\rho_0)^sR(\rho_1)^{1-s}\right)\right]_+ \right)^M =\left[\frac{e^{-Me}}2\right]_+,Retr(R(ρ0​)1/2R(ρ1​)1/2)=e−e,21​(0<s<1inf​[Retr(R(ρ0​)sR(ρ1​)1−s)]+​)M=[2e−Me​]+​,

where [x]+=max⁡(x,0)[x]_+=\max(x,0)[x]+​=max(x,0), viewed in [0,∞][0,\infty][0,∞] where appropriate. Finally, for kernels A0,A1A_0,A_1A0​,A1​ on NM\mathbb N^MNM, define

Perr(A0,A1)=inf⁡E boundedE=E∗≥0, I−E=(I−E)∗≥0[Re⁡tr⁡(ER(A0))+Re⁡tr⁡((I−E)R(A1))2]+.P_{\mathrm{err}}(A_0,A_1) =\inf_{\substack{E\text{ bounded}\\E=E^*\ge0,\ I-E=(I-E)^*\ge0}} \left[ \frac{\operatorname{Re}\operatorname{tr}(ER(A_0))+ \operatorname{Re}\operatorname{tr}((I-E)R(A_1))}{2} \right]_+.Perr​(A0​,A1​)=E boundedE=E∗≥0, I−E=(I−E)∗≥0​inf​[2Retr(ER(A0​))+Retr((I−E)R(A1​))​]+​.

Then

[1−1−exp⁡ ⁣(−2kMNS(NB+1−NB)2)2]+≤Perr ⁣(Y0(α(M)),Y1(α(M)))≤[e−Me2]+.\left[ \frac{1-\sqrt{1-\exp\!\left(-2kMN_S (\sqrt{N_B+1}-\sqrt{N_B})^2\right)}}2 \right]_+ \le P_{\mathrm{err}}\!\left(Y_0(\alpha^{(M)}),Y_1(\alpha^{(M)})\right) \le\left[\frac{e^{-Me}}2\right]_+.​21−1−exp(−2kMNS​(NB​+1​−NB​​)2)​​​+​≤Perr​(Y0​(α(M)),Y1​(α(M)))≤[2e−Me​]+​.

The assumptions allow M=1M=1M=1, NS=0N_S=0NS​=0, and NB=0N_B=0NB​=0.

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