arXiv:2609.36004
Digital Quantum Algorithms for Generating and Utilizing Spin Squeezed States
Mingru Yang, Ruby Wei, Chao Yin
Abstract
Spin squeezed states (SSSs) are conventionally viewed as analog resources for quantum metrology. Here we develop algorithms to efficiently generate and exploit SSSs on a digital quantum computer. We introduce an adaptive local-circuit protocol that prepares SSSs with squeezing parameter $ξ$ in depth $\mathrm{O}\left(\log{1/ξ}\right)$, yielding an exponential improvement over non-adaptive approaches. We prove that this depth is optimal for permutation symmetric SSSs by deriving a matching lower bound on entanglement entropy. We further leverage the squeezing paradigm to deterministically prepare the Dicke states using asymptotically fewer resources than prior methods. Remarkably, classical simulations of our preparation protocols under realistic noise models predicts a metrological gain of 4.2 dB beyond the standard quantum limit with 55 superconducting qubits in total including ancillae. Our work paves the way for performing quantum metrology on near-term digital quantum devices.
AI Summary
The paper studies the digital-circuit resources needed to prepare spin-squeezed states (SSSs), their entanglement structure, and their use in Dicke-state preparation. Squeezing is defined by extensive transverse polarization and collective variance $\operatorname{Var}(Z)\leq N\xi^2$. A one-dimensional adaptive protocol prepares permutation-symmetric SSSs in depth $O(\log(1/\xi))$, using $2N+O(\log(1/\xi))$ qubits, for $\xi\geq cN^{-1/2}\log^6N$. Its success probability is $1-O(1/\log(1/\xi))$. Gaussian-amplitude ancillas, measurement-assisted global fanout, and an inverse QFT implement a coarse-grained Gaussian measurement of collective magnetization. This contrasts with the $\Omega(\xi^{-2/d})$ depth required by spatially local unitary circuits in $d$ dimensions; the adaptive model permits nonlocal classical feedforward. A matching one-dimensional depth lower bound follows from an $\Omega(\log(1/\xi))$ smooth log-rank entanglement bound across an equal bipartition. The proof truncates the Dicke expansion and bounds reduced-state eigenvalues using Dicke-state Schmidt coefficients. Recursive squeezing and adaptive rotations also prepare the equatorial Dicke state in $O(\log N)$ adaptive depth with constant success probability and linear width. Stochastic-noise simulations predict 4.2 dB Wineland metrological gain with 15 data qubits, 55 total qubits, and 0.25% acceptance. Noise mitigation uses repeated measurements, postselection, and two-dimensional decoding. Simulations omit coherent errors and idealize several circuit stages. Preparation remains sensitive to fanout faults; the squeezing construction excludes a polylogarithmic neighborhood of the Heisenberg limit. Entanglement optimality assumes permutation symmetry and one-dimensional geometry; extensions to general Dicke states and shorter higher-dimensional protocols remain open.
Central claim may not hold: The abstract claims deterministic Dicke-state preparation, but the proved O(log N)-depth protocol succeeds only with constant probability and requires repetition for eventual success. The abstract's squeezing-preparation statement omits both the theorem's restriction ξ ≥ c₁N^(-1/2)log⁶N and its nonunit success probability. The claimed exponential improvement over non-adaptive approaches is established only against local unitary preparation, not general non-adaptive circuits allowing measurement and postselection. The claimed optimal local depth is proved only in one dimension for permutation-symmetric targets, while shorter-depth preparation in higher dimensions remains open. The predicted 4.2 dB gain is conditional on 0.25% postselection acceptance and a noise simulation that treats several circuit components as ideal.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 6+29 pages