When a quantum state leaves its starting point before perfect transfer

Can a quantum state lose all overlap with its starting configuration before it arrives perfectly at the other end of a chain? Jasmin G. Enriquez and Anastasiia Minenkova address that question in a preprint, Early State Exclusion for 5 × 5 Jacobi Matrices. For a specified class of five-qubit models, they report a complete mathematical criterion separating chains that show this behavior from those that do not.

The phenomenon is called early state exclusion. Despite language about a state “vanishing,” it does not mean that quantum information is destroyed. It means that the evolving state has zero overlap with the initial state: mathematically, the two states are orthogonal. That condition is different from perfect arrival at the destination.

Two events, two different tests

The researchers study a one-dimensional chain with interactions between neighboring qubits. A qubit is a quantum system used to encode quantum information. The transfer dynamics considered here are represented by a five-by-five Jacobi matrix: a symmetric array whose nonzero entries lie on the main diagonal and immediately beside it. Those neighboring entries describe the connections along the chain.

Perfect state transfer occurs when evolution takes a state initially localized at one end to the other end, apart from an overall phase. Early state exclusion asks a separate question: does the overlap with the starting state become exactly zero at some positive time before the earliest perfect transfer?

That timing restriction matters. Testing against a later transfer event could count behavior that happened only after perfect transfer had already occurred. The paper consistently defines exclusion relative to the first arrival.

A criterion written in the spectrum

The central result concerns chains with mirror-symmetric matrix entries and five eigenvalues arranged around a common center. Eigenvalues are the spectral values that determine the oscillation frequencies entering the evolution. Here they have the form c − x, c − y, c, c + y and c + x, with x greater than y greater than zero.

Among these matrices that already realize perfect end-to-end transfer, the authors prove that early state exclusion is absent if and only if x is an integer multiple of y. Otherwise, it occurs. Both possibilities describe infinitely many matrices.

For example, the paper presents a spectrum centered at zero with x = 6 and y = 1 that has no early exclusion. A different example, x = 6 and y = 5, does exhibit it. These are examples within the perfect-transfer setting—not a claim that any arbitrary choice of two spacings guarantees perfect transfer.

From numerical clues to a proof

Numerical exploration first suggested the distinction. To establish it analytically, the authors express the return overlap using cosine functions whose frequencies depend on x and y. This converts a question about matrix evolution into a question about when an oscillating real-valued function crosses zero.

In the integer-multiple case, they compare two trigonometric expressions and show that the return overlap cannot vanish before the first perfect-transfer time. In the other case, they identify times at which the relevant difference changes sign. Continuity then guarantees a zero between them.

The paper also constructs chains with any prescribed positive even number of early-exclusion events. In its chosen normalization, setting y = 2k + 1 and x = 2k + 2 produces 2k such events before perfect transfer.

Finally, shifting every eigenvalue by the same amount c only adds an overall phase to the evolving state. That does not change the magnitude of its overlap with the initial state, so the classification extends from spectra centered at zero to the stated centered family.

What remains unresolved

This is a mathematical preprint, not an experimental demonstration or a tested communication device. Its exact-zero and perfect-transfer results concern the specified model; they do not establish how a physical implementation would tolerate noise or imperfectly controlled interactions.

Nor does the theorem settle the question for every odd chain length. The authors present numerical examples for seven, nine and eleven qubits, but leave a general treatment beyond five open. The advance is a precise classification for one tractable family—not a universal rule for quantum networks.