arXiv:2610.17790
Multi-parameter estimation enhanced by level crossing Hamiltonian structure with Bayesian feedback
Ryo Tatsumi, Le Bin Ho, Hiroaki Matsueda
Abstract
Level crossings, where two quantum energy levels intersect, are ubiquitous in quantum systems and play a central role in controlling quantum dynamics. In this study, we establish a framework that connects these level crossing structures to the ultimate precision limits of quantum metrology. By analyzing a Hamiltonian with the level crossings structure, we identify the conditions under which multiple parameters can be estimated simultaneously with maximal sensitivity near the level crossing point. We then construct a resource efficient sensing protocol based on unitary quench dynamics and Bayesian estimation, enabling high precision, low cost parameter estimation without requiring entanglement or complex control. We present a protocol whose estimation precision saturates the Holevo Cramer Rao bound, the tightest achievable precision bound in multi parameter estimation. Using the two level model and the two spin Ising model as illustrative examples, we demonstrate how the protocol can be applied to a system with level crossing Hamiltonian structure. Our findings reveal that level crossings serve as universal operating points for precision quantum metrology, offering a simple and scalable path toward practical multi parameter sensing across diverse quantum platforms.
AI Summary
The paper asks whether level-crossing Hamiltonian structure can support simultaneous estimation of two parameters without ground-state preparation. It studies $H=\delta(s)\sigma_z+q(a)\sigma_x$, with a crossing at $\delta=q=0$, using analytic Fisher-information calculations and numerical Bayesian estimation. The ground-state quantum Fisher information matrix (QFIM) is singular: enhanced single-parameter sensitivities near the crossing do not identify both parameters jointly. For a parameter-independent probe evolved for time $t$, the state $(|0\rangle+i|1\rangle)/\sqrt{2}$ instead gives a diagonal QFIM at the crossing. In the minimal model $H=s\sigma_z+a\sigma_x$, both diagonal entries equal $4t^2$. Randomly alternating measurements of $\sigma_x$ and $\sigma_z$ accounts for measurement incompatibility and attains the locally derived Holevo bound, $\operatorname{Tr}C^{-1}=(|\partial_s\delta|^{-1}+|\partial_aq|^{-1})^2/(4t^2)$. For the minimal model, this yields total variance $1/(nt^2)$ asymptotically for $n$ measurements, twice the SLD bound. An adaptive protocol offsets the Hamiltonian parameters using posterior means to steer operation toward the crossing. Simulations with $10^4$ updates show posterior concentration and variances approaching the local classical bound. A two-spin Ising example maps approximately to the same model near $(B_z,B_x)=(J,0)$, predicting total variance $(3+2\sqrt{2})/(8nt^2)$. The stated qualifications are that nonlinear parameter derivatives can move sensitivity maxima away from the crossing, and additional levels limit the two-level approximation. Larger exchange coupling improves that approximation, although additional-level contributions can broaden the sensitive parameter region. Holevo optimality is established at the crossing; global estimation is demonstrated numerically.
Central claim may not hold: The paper establishes local Holevo-bound saturation for a particular two-parameter qubit model, not the abstract's universal optimality claim for level crossings. The Bayesian-feedback simulations demonstrate selected examples but do not establish a general global-convergence or bound-attainment guarantee. The prescribed two-spin probe is entangled, with concurrence 1/2, so that implementation does not support the advertised absence of entangled-probe requirements. The full two-spin model has Q_BxBx=8t²+sin²(2Jt)/J² at the crossing, contradicting the asserted exact effective-qubit value and leaving full-model Holevo saturation unestablished. The eight-element Ising POVM is not normalized as written: its effects sum to [(1+p)/2]I rather than I.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 13 pages, 8 figures