Lean, informalized

Nielsen & Chuang · Chapter 12, Quantum information theory · p. 563

Exercise 12.13 - A mixture of two operations has at least the average entropy exchange of the two

Entropy exchange is concave in the quantum operation.

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 Exercise

A quantum system with a finite dimensional state space ℋ\mathcal{H} is prepared in a state ρ\rho and passed through a noisy device. With probability pp the device acts as a quantum operation ℰ1\mathcal{E}_1, and otherwise as a second operation ℰ2\mathcal{E}_2. Exercise 12.13 asks to show that the entropy exchange is concave in the operation: the entropy exchange of ρ\rho under the device is at least the average, with weights pp and 1−p1-p, of its entropy exchanges under ℰ1\mathcal{E}_1 and under ℰ2\mathcal{E}_2.✦

A Quantum System

A density operator ρ\rho on ℋ\mathcal{H} is a positive operator on ℋ\mathcal{H} of trace one; the density operators are the states of the system.✦

A quantum operation ℰ\mathcal{E} on ℋ\mathcal{H} is given by its operation elements, operators EiE_i on ℋ\mathcal{H} indexed by a finite set, and it is trace preserving when ∑iEi†Ei=I.\sum_i E_i^\dagger E_i = I .(1)

For a weight pp and two operations on ℋ\mathcal{H}, ℰ1\mathcal{E}_1 with elements EiE_i and ℰ2\mathcal{E}_2 with elements FkF_k, the mixture pℰ1+(1−p)ℰ2p\,\mathcal{E}_1 + (1-p)\,\mathcal{E}_2 is the operation whose elements are the EiE_i scaled by p\sqrt{p} together with the FkF_k scaled by 1−p\sqrt{1-p}: {pEi}i∪{1−pFk}k.\bigl\{\sqrt{p}\,E_i\bigr\}_i \,\cup\, \bigl\{\sqrt{1-p}\,F_k\bigr\}_k .(2)✦

Entropy Exchange

The Shannon entropy of a distribution λ\lambda on a finite set is H(λ)=−∑mλmlog⁡λm,H(\lambda) = -\sum_m \lambda_m \log \lambda_m ,(3)✦ measured in nats. The von Neumann entropy S(T)S(T) of a Hermitian operator TT on a finite dimensional inner product space of dimension nn is the Shannon entropy of its eigenvalues λ1,…,λn\lambda_1, \dots, \lambda_n, counted with multiplicity; for an operator that is not Hermitian it is set to zero.✦

For an operator ρ\rho on ℋ\mathcal{H} and a quantum operation ℰ\mathcal{E} with elements EiE_i, the matrix WW is the complex square matrix indexed by the operation elements, with entries Wij=tr⁡(EiρEj†).W_{ij} = \operatorname{tr}\bigl(E_i\, \rho\, E_j^\dagger\bigr) .(4)✦ The entropy exchange S(ρ,ℰ)S(\rho, \mathcal{E}) of ρ\rho under ℰ\mathcal{E} is the von Neumann entropy of WW, taken as an operator on the complex Euclidean space with one basis vector for each operation element: S(ρ,ℰ)=S(W).S(\rho, \mathcal{E}) = S(W) .(5)✦

The Claim

Theorem 1. Let ρ\rho be a density operator on ℋ\mathcal{H}, let ℰ1\mathcal{E}_1 and ℰ2\mathcal{E}_2 be trace preserving quantum operations on ℋ\mathcal{H} with elements EiE_i and FkF_k indexed by finite sets, and let 0≤p≤10 \le p \le 1. Then pS(ρ,ℰ1)+(1−p)S(ρ,ℰ2)≤S(ρ,pℰ1+(1−p)ℰ2).p\, S(\rho, \mathcal{E}_1) + (1-p)\, S(\rho, \mathcal{E}_2) \le S\bigl(\rho,\, p\,\mathcal{E}_1 + (1-p)\,\mathcal{E}_2\bigr) .(6)✦

What Is Checked

The claim is stated over a density operator ρ\rho on a finite dimensional Hilbert space, two families of operation elements each satisfying (1), a weight pp between 00 and 11, the mixture (2), and the entropy exchange (5), built from the matrix (4), the von Neumann entropy and the Shannon entropy (3).

What the library proves about these constructions is the inequality (6), for every such state, pair of trace preserving operations and weight.