An exact theory explains the speed limits of measurement-driven quantum preparation

Preparing symmetry through local checks

In an arXiv preprint, T. Dörstel and colleagues develop an exact mathematical description of a measurement-and-feedback protocol for preparing Dicke states. These states are equal superpositions of all arrangements with a specified number of spin-up qubits. They are unchanged when any two qubits exchange positions.

The protocol checks pairs of qubits for a singlet, an antisymmetric two-qubit state that violates this exchange symmetry. When a measurement detects a singlet, a conditional single-qubit rotation converts it into a symmetric state. On a connected measurement network, satisfying all the local symmetry constraints places the system in the fully symmetric Dicke subspace.

This is “frustration-free” control: the target satisfies every constraint simultaneously. Once the ideal protocol reaches that subspace, subsequent checks and corrections leave it unchanged. With feedback that preserves magnetization—the total spin along a chosen axis—an initial state confined to one magnetization sector converges to that sector’s unique Dicke state.

A smaller, exactly closed description

Tracking a general quantum state becomes exponentially more demanding as qubits are added. The authors instead track singlet projectors: operators whose expectation values give the probability of detecting a singlet on each possible pair.

Their evolution forms a closed system. Each projector evolves into a linear combination of other pair projectors, without requiring additional many-qubit observables. For N qubits, there are only N(N−1)/2 such quantities.

The equations are identical to those of a classical random walk with absorption. A configuration specifies two distinct endpoints. Measurements move those endpoints along measured bonds; feedback removes weight when the endpoints form a bond where correction operates. “First passage” refers to reaching that removal set.

This is an exact description of the pair observables, not a claim that the entire quantum state is classical. The authors also derive bounds connecting total singlet weight to the probability of being outside the target subspace, translating pair relaxation into preparation guarantees.

Transport sets the pace—until removal does

For a periodic one-dimensional chain with nearest-neighbor measurements and maximal feedback, the slowest pair decay rate is exactly 2[1−cos(π/N)], in units of the measurement rate per bond. For large N, it approaches π²/N²: the associated relaxation time grows quadratically with system size.

That rate is an exact eigenvalue of the full quantum evolution generator. However, the authors prove only that the full generator’s slowest nonzero decay rate lies between half this value and the value itself. Equality at all system sizes remains unproven.

The study also establishes a transport-independent limit within this class of symmetric, closed pair dynamics. Feedback cannot remove singlet weight faster than its removal rate averaged over all pair configurations. Local feedback reaches only order N configurations among order N² possibilities, imposing a relaxation time at least linear in N at fixed local rates.

Random partial-SWAP operations can accelerate endpoint motion while preserving the target. Sufficiently fast motion changes the chain’s relaxation from quadratic to linear scaling, but further acceleration cannot overcome the removal limit. This improvement requires additional operations whose rate increases with system size.

Monitoring every pair reaches the same linear scaling without those transport operations, at the same total measurement-attempt rate as the chain. The tradeoff is access to nonlocal pair measurements. This is not a universal speed limit for all methods of preparing Dicke states.

Final states, fluctuations and errors

Pair relaxation alone does not specify the final state when the input spans several magnetization sectors. A second closed set of conserved collective-spin quantities determines the final populations and coherences between Dicke states. A third closure, involving products of singlet and triplet projectors, determines ensemble covariances rather than individual measurement histories.

For a specified model of independent, isotropic single-qubit Pauli errors, the authors obtain exact stationary pair profiles and recovery rates. Their weak-noise upper bound on stationary leakage grows as the error rate per site times N³ for the bare chain, or N² in the removal-limited regime. These are bounds, not exact leakage values or a size-independent error threshold. Because these errors change magnetization, the guarantee concerns the symmetric subspace, not preservation of one chosen Dicke state.

Scope of the result

The work combines analytic derivations with numerical comparisons against full quantum evolution at accessible sizes; it does not report an experimental demonstration. The supplied source is a preprint, not evidence of peer review. Its exact results depend on the specified measurements, feedback and error channels. Coherent exchange at finite singlet density and extensions of the feedback closure beyond qubits remain open questions.