OctoInformBeta

OpenAI preprint, September 2026 · 1. Introduction · Section 1

Theorem 1.1: The scalar optimum gives operational capacity and optimal product phase signals

Amplitude damping has additive Holevo information.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

Fix a damping parameter γ\gamma and a stationary excited-state population ν\nu in [0,1][0,1]. The communication problem concerns repeated uses of the generalized amplitude-damping channel 𝒩γ,ν\mathcal N_{\gamma,\nu}.

The computational basis ℬn\mathcal B_n of nn qubits is the set Fin⁡(n)→Fin⁡(2)\operatorname{Fin}(n)\to\operatorname{Fin}(2) of bit strings, where Fin⁡(m)={0,…,m−1}\operatorname{Fin}(m)=\{0,\ldots,m-1\}.✦ An nn-qubit operator AA is a complex matrix indexed by ℬn\mathcal B_n in both rows and columns.✦ A density operator PP on nn qubits is a positive semidefinite nn-qubit operator with trace one.✦

The single-qubit Kraus operators of 𝒩γ,ν\mathcal N_{\gamma,\nu} are K0(1)=1−ν(1001−γ),K1(1)=γ(1−ν)(0100),K2(1)=ν(1−γ001),K3(1)=γν(0010).\begin{aligned} K^{(1)}_0&=\sqrt{1-\nu} \begin{pmatrix}1&0\\0&\sqrt{1-\gamma}\end{pmatrix},& K^{(1)}_1&=\sqrt{\gamma(1-\nu)} \begin{pmatrix}0&1\\0&0\end{pmatrix},\\ K^{(1)}_2&=\sqrt{\nu} \begin{pmatrix}\sqrt{1-\gamma}&0\\0&1\end{pmatrix},& K^{(1)}_3&=\sqrt{\gamma\nu} \begin{pmatrix}0&0\\1&0\end{pmatrix}. \end{aligned}(1)✦ The symbol ** denotes the adjoint. The tensor product Kraus operators KrK_r, with r:Fin⁡(n)→Fin⁡(4)r:\operatorname{Fin}(n)\to\operatorname{Fin}(4), and the memoryless channel are formed from (1) by Kr(i,j)=∏k=0n−1Krk(1)(ik,jk),i,j∈ℬn,𝒩γ,ν⊗n(P)=∑rKrPKr*.\begin{aligned} K_r(i,j)&=\prod_{k=0}^{n-1}K^{(1)}_{r_k}(i_k,j_k), \qquad i,j\in\mathcal B_n,\\ \mathcal N_{\gamma,\nu}^{\otimes n}(P)&=\sum_r K_rPK_r^*. \end{aligned}(2)

Ensembles and Communication Rates

For an operator PP, the von Neumann entropy S(P)S(P) is the real trace of the spectral function x↦−xln⁡xx\mapsto-x\ln x applied to PP.✦ For density operators, its formula and the binary entropy hh are S(P)=−Tr⁡(Pln⁡P),h(q)=−qln⁡q−(1−q)ln⁡(1−q),0ln⁡0=0.S(P)=-\operatorname{Tr}(P\ln P),\qquad h(q)=-q\ln q-(1-q)\ln(1-q),\qquad 0\ln0=0.(3) Both entropies in (3) are in nats.

A finite quantum ensemble E={(wa,ρa)}a=0m−1E=\{(w_a,\rho_a)\}_{a=0}^{m-1} on nn qubits consists of density operators ρa\rho_a and nonnegative weights waw_a summing to one.✦ The ensemble Holevo information χ(E;γ,ν)\chi(E;\gamma,\nu) of a finite quantum ensemble EE on nn qubits is its output entropy difference in bits; the nn-use Holevo information χn(γ,ν)\chi_n(\gamma,\nu) of 𝒩γ,ν\mathcal N_{\gamma,\nu} is the supremum of this difference over all such ensembles: χ(E;γ,ν)=S(∑awa𝒩γ,ν⊗n(ρa))−∑awaS(𝒩γ,ν⊗n(ρa))ln⁡2,χn(γ,ν)=supEχ(E;γ,ν).\begin{aligned} \chi(E;\gamma,\nu) &=\frac{S\!\left(\sum_a w_a \mathcal N_{\gamma,\nu}^{\otimes n}(\rho_a)\right) -\sum_a w_a S\!\left( \mathcal N_{\gamma,\nu}^{\otimes n}(\rho_a)\right)}{\ln2},\\ \chi_n(\gamma,\nu)&=\sup_E\chi(E;\gamma,\nu). \end{aligned}(4)

An nn-use classical code CnC_n consists of a positive number MnM_n of messages, encoding density operators ρa\rho_a, and positive semidefinite decoding operators DaD_a on nn qubits with ∑aDa=I\sum_aD_a=I, where II is the identity operator on nn qubits.✦ For an nn-use classical code CnC_n with MnM_n equally likely messages, the average decoding error en(Cn)e_n(C_n) is en(Cn)=1−1Mn∑a=0Mn−1Re⁡Tr⁡(Da𝒩γ,ν⊗n(ρa)).e_n(C_n)=1-\frac{1}{M_n}\sum_{a=0}^{M_n-1} \operatorname{Re}\operatorname{Tr}\!\left( D_a\mathcal N_{\gamma,\nu}^{\otimes n}(\rho_a)\right).(5) An achievable rate RR for 𝒩γ,ν\mathcal N_{\gamma,\nu} is a real number in bits per channel use for which there is a sequence of codes CnC_n with en(Cn)→0e_n(C_n)\to0 in (5) and, for every ε>0\varepsilon>0, ln⁡Mn/(nln⁡2)≥R−ε\ln M_n/(n\ln2)\ge R-\varepsilon for all sufficiently large nn. The unassisted classical capacity C(γ,ν)C(\gamma,\nu) of 𝒩γ,ν\mathcal N_{\gamma,\nu} is the supremum of its achievable rates: C(γ,ν)=sup⁡{R∈ℝ:R is achievable}.C(\gamma,\nu)=\sup\{R\in\mathbb R:R\text{ is achievable}\}.(6)✦

Phase Signals and the Scalar Objective

For an input excitation population p∈[0,1]p\in[0,1], the binary phase signal amplitudes ap(s,b)a_p(s,b), with s,b∈Fin⁡(2)s,b\in\operatorname{Fin}(2), are ap(s,b)={1−p,b=0,(−1)sp,b=1.a_p(s,b)= \begin{cases} \sqrt{1-p},&b=0,\\ (-1)^s\sqrt p,&b=1. \end{cases}(7)✦ The single-qubit phase vectors ϕ+\phi_+ and ϕ−\phi_- have computational-basis amplitudes ap(0,b)a_p(0,b) and ap(1,b)a_p(1,b), respectively, for b∈Fin⁡(2)b\in\operatorname{Fin}(2).

The product signal amplitude Ap,s(i)A_{p,s}(i) and product signal operator ρp,s\rho_{p,s}, for s,i,j∈ℬns,i,j\in\mathcal B_n, are defined from (7) by Ap,s(i)=∏k=0n−1ap(sk,ik),ρp,s(i,j)=Ap,s(i)Ap,s(j)¯.A_{p,s}(i)=\prod_{k=0}^{n-1}a_p(s_k,i_k),\qquad \rho_{p,s}(i,j)=A_{p,s}(i)\overline{A_{p,s}(j)}.(8) The phase value Φn(γ,ν,p)\Phi_n(\gamma,\nu,p) is the entropy difference with equal weights on the operators in (8): Φn(γ,ν,p)=S(2−n∑s∈ℬn𝒩γ,ν⊗n(ρp,s))−2−n∑s∈ℬnS(𝒩γ,ν⊗n(ρp,s))ln⁡2.\Phi_n(\gamma,\nu,p)=\frac{ S\!\left(2^{-n}\sum_{s\in\mathcal B_n} \mathcal N_{\gamma,\nu}^{\otimes n}(\rho_{p,s})\right) -2^{-n}\sum_{s\in\mathcal B_n} S\!\left(\mathcal N_{\gamma,\nu}^{\otimes n}(\rho_{p,s})\right)}{\ln2}.(9)✦ The phase value in (9) is measured in bits.

The qubit entropy gg, the phase output determinant vv, and the single-qubit phase pair objective ff are g(u)=h(1+1−4u2),v(γ,ν,p)=γν(1−ν)+γ(1−γ)(p−ν)2,f(γ,ν,p)=h((1−γ)p+γν)−g(v(γ,ν,p)).\begin{aligned} g(u)&=h\!\left(\frac{1+\sqrt{1-4u}}2\right),\\ v(\gamma,\nu,p)&=\gamma\nu(1-\nu) +\gamma(1-\gamma)(p-\nu)^2,\\ f(\gamma,\nu,p)&=h((1-\gamma)p+\gamma\nu)-g(v(\gamma,\nu,p)). \end{aligned}(10) An optimal population pp for (γ,ν)(\gamma,\nu) is a point of [0,1][0,1] such that f(γ,ν,q)≤f(γ,ν,p)f(\gamma,\nu,q)\le f(\gamma,\nu,p) for every q∈[0,1]q\in[0,1], with ff as in (10).✦

The Claim

Theorem 1. Let γ,ν∈[0,1]\gamma,\nu\in[0,1]. There is an optimal population p*p_* such that, for every integer n≥1n\ge1, χn(γ,ν)=nχ1(γ,ν)=nln⁡2f(γ,ν,p*),C(γ,ν)=χ1(γ,ν),χn(γ,ν)=Φn(γ,ν,p)for every optimal population p.\begin{gathered} \chi_n(\gamma,\nu)=n\chi_1(\gamma,\nu) =\frac{n}{\ln2}f(\gamma,\nu,p_*),\qquad C(\gamma,\nu)=\chi_1(\gamma,\nu),\\ \chi_n(\gamma,\nu)=\Phi_n(\gamma,\nu,p) \quad\text{for every optimal population }p. \end{gathered}(11)✦ For every optimal population pp and every integer n≥1n\ge1, each operator ρp,s\rho_{p,s} with s∈ℬns\in\mathcal B_n is a density operator.✦

In the paper, writing χ(𝒩γ,ν⊗n)\chi(\mathcal N_{\gamma,\nu}^{\otimes n}) for the nn-use Holevo information and C(𝒩γ,ν)C(\mathcal N_{\gamma,\nu}) for the unassisted classical capacity, the result reads: for all γ,ν∈[0,1]\gamma,\nu\in[0,1] and every integer n≥1n\ge1, χ(𝒩γ,ν⊗n)=nχ(𝒩γ,ν)=nln⁡2max⁡0≤p≤1f(γ,ν,p)\chi(\mathcal N_{\gamma,\nu}^{\otimes n}) =n\chi(\mathcal N_{\gamma,\nu}) =\frac{n}{\ln2}\max_{0\le p\le1}f(\gamma,\nu,p) and C(𝒩γ,ν)=χ(𝒩γ,ν)C(\mathcal N_{\gamma,\nu})=\chi(\mathcal N_{\gamma,\nu}), with independent products of the two equiprobable signals ϕ+\phi_+ and ϕ−\phi_- attaining the Holevo information at every block length for every maximizing pp.

What Is Checked

The claim (11) is stated over finite ensembles of nn-qubit density operators, the tensor powers (2), and the vanishing-error code sequences defining the capacity (6), with γ,ν∈[0,1]\gamma,\nu\in[0,1].