Lean, informalized

Nielsen & Chuang · Chapter 12, Quantum information theory · p. 531

Theorem 12.1 - How much one measurement outcome can tell about which state was prepared

The information a measurement yields is bounded by entropies.

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

Alice has a finite collection of states ρx\rho_x of one quantum system QQ. She draws the label xx at random with probability pxp_x, prepares ρx\rho_x and hands the system to Bob, who knows the states and the probabilities but not the label drawn. Bob performs a measurement on QQ and reads an outcome yy. With XX the label and YY the outcome, the mutual information H(X:Y)H(X{:}Y) measures what Bob has learned about Alice’s choice. Theorem 12.1 bounds it, for every measurement Bob may perform, by a quantity built from the states and the probabilities alone.

A Quantum System

A quantum system QQ is a physical system whose state space is a finite dimensional complex Hilbert space.✦ A state ρ\rho of QQ is a positive operator of unit trace on the state space of QQ.✦

The mixture of states ρx\rho_x of QQ, indexed by xx in a finite set, with weights pxp_x that are nonnegative and sum to one, is the state ρ\rho of QQ whose density operator is the convex combination ρ=∑xpxρx.\rho = \sum_x p_x \rho_x .(1)✦ It is the state Bob assigns to the system he receives, not knowing xx.

The von Neumann entropy of a state ρ\rho of QQ is S(ρ)=−tr⁡(ρlog⁡ρ),S(\rho) = -\operatorname{tr}(\rho \log \rho) ,(2)✦ the Shannon entropy of the eigenvalues of ρ\rho, measured in nats.

The Measurement

A measurement on QQ with outcomes y=1,…,ry = 1, \ldots, r is a family of measurement operators MyM_y on the state space of QQ, one for each outcome, that satisfy the completeness relation ∑yMy†My=I.\sum_y M_y^\dagger M_y = I .(3)✦ The positive operators Ey=My†MyE_y = M_y^\dagger M_y, one for each outcome yy, are the effects of the measurement; together they form its positive operator-valued measure (POVM). When the measurement is performed on a state ρ\rho of QQ, the outcome yy occurs with the probability given by the Born rule, p(y)=tr⁡(MyρMy†)=tr⁡(Eyρ).p(y) = \operatorname{tr}(M_y \rho M_y^\dagger) = \operatorname{tr}(E_y \rho) .(4)✦

The Claim

When Bob performs the measurement on the system Alice prepared, the label xx and the outcome yy occur together with probability p(x,y)=pxtr⁡(Eyρx).p(x, y) = p_x \operatorname{tr}(E_y \rho_x) .(5)

Theorem 1. Let QQ be a quantum system and let ρx\rho_x be states of QQ, indexed by xx in a finite set and prepared with probabilities pxp_x, with mixture ρ\rho as in (1). For every measurement on QQ with outcomes y=1,…,ry = 1, \ldots, r, the label XX and the outcome YY, distributed as in (5), satisfy H(X:Y)≤S(ρ)−∑xpxS(ρx).H(X{:}Y) \leq S(\rho) - \sum_x p_x S(\rho_x) .(6)✦

What Is Checked

The claim is stated over a quantum system QQ, states of QQ with their mixture (1) and their von Neumann entropy (2), and a measurement on QQ whose outcome probabilities are the Born probabilities (4) of its POVM.

About these constructions the library shows that a mixture of states with nonnegative weights summing to one is again a state, positive with unit trace, and that the effects Ey=My†MyE_y = M_y^\dagger M_y of a measurement are positive and sum to the identity by the completeness relation (3). For every state, the Born probabilities (4) therefore form a probability distribution over the outcomes.