Helstrom bound

In plain terms

Someone prepares one of two known states and hands it to you, and you have to guess which. If the states were classical probability distributions the best strategy would be obvious, and the success rate would be a simple distance between the distributions. Quantum states can overlap in a way distributions can't, so the question is whether quantum mechanics allows a cleverer test than the classical analogy suggests, or a worse one.

Helstrom answered it exactly. Among all possible measurements, including ones that entangle the state with an apparatus and process the result, the best chance of guessing right is one half plus half the trace distance between the two states, where the trace distance is the natural quantum generalisation of the distance between two probability distributions. No test beats that, and one particular test achieves it, namely the measurement that asks which state is more likely at each point of the spectrum of their difference.

The two extremes are the ones worth remembering. Identical states force you to guess at random, and perfectly distinguishable states allow a test that never errs. Everything in between is measured by one number, which is why trace distance is the right notion of how different two quantum states are.

In CSD

The corpus proves the bound and its attainment together, so the trace distance gets its operational meaning rather than remaining a definition. The optimality half says no test does better than the projector onto the positive part of the weighted difference of the two states, and the attainment half computes that projector's success probability exactly. Both rest on the Jordan decomposition machinery already in the trace-distance module, since the variational bound and the identity that makes it tight are statements about positive and negative parts.

The result is stated at equal priors and in general, in success form and in error form, with the two extremes proved separately: indistinguishable states force a coin flip, perfectly distinguishable ones permit an error-free test. It's the converse companion to the data-processing inequality, as channels can't increase distinguishability and this says distinguishability is exactly what a measurement can extract. The complementary problem, zero error at the cost of an inconclusive outcome, is unambiguous discrimination, proved elsewhere in the corpus.

Mathematically

A test is an effect E with 0 <= E <= 1 (IsTest), guessing the first state on outcome E. With A = p0 rho0 - p1 rho1 the Helstrom operator and helstromTest the projector onto its positive eigenspace, re_trace_posPart_eq gives Re Tr(A+) = (||A||_1 + Re Tr A)/2, the Jordan-decomposition identity; re_trace_mul_le_helstrom is optimality, from the variational bound Re Tr(A P) <= Re Tr(A+) for 0 <= P <= 1; re_trace_mul_helstrom is attainment, from A times the positive projector being A+.

successProb_le and successProb_helstromTest give P_success <= (1 + D(rho0, rho1))/2 with equality at helstromTest, errorProb_ge and errorProb_helstromTest the error form, helstrom_indistinguishable and helstrom_perfect the two extremes, and successProbPrior_le with successProbPrior_helstromTest the general-prior statement P_success <= (1 + ||p0 rho0 - p1 rho1||_1)/2. Finite-dimensional matrices over the complex numbers throughout.

The name

Carl Helstrom worked out the bound in the late 1960s at the University of California, San Diego, as part of a programme applying statistical decision theory to optical communication, and his 1976 book Quantum Detection and Estimation Theory founded the subject now called quantum estimation. Alexander Holevo proved closely related results independently in Moscow at the same time.

The bound's practical face is in quantum key distribution and in optical receivers, where the question of how well an eavesdropper or a detector can distinguish two signals is precisely Helstrom's.

Related entries

• Holevo boundA quantum system can be prepared in infinitely many distinct ways, yet no reader can extract more classical information from it than from a classical system of the same size, and the measurement channel of CSD is proved to pass exactly one bit per qubit.
• Quantum channelThe most general thing that can happen to a quantum state is captured in CSD as a concrete structure, a family of operators with one completeness law, and every certificate of its physicality is proved rather than assumed.
• Luders ruleThe textbook's second rule, which says how a state changes when it's measured, isn't a second rule in CSD, as it's what the record layer does.
• Naimark dilationEvery noisy, imperfect measurement made in a real laboratory is an ideal textbook measurement carried out on a larger system, and that single fact is what lets CSD's derivation of the Born weights cover actual experiments rather than idealised ones.
• von Neumann entropyA quantum whole can be perfectly known while its parts are as uncertain as they could possibly be, and entropy is where that stops being a metaphor and becomes arithmetic.