arXiv:2610.17047 · Physical Review Letters
Enhanced Quantum Metrology via Saddle-Point Scrambling in Phase Space
Lei Shao, Hai-Jun Xing, Libin Fu
Abstract
Nonlinear effects are widely utilized in quantum metrology to enhance measurement precision by leveraging complex dynamical behaviors. Here, we propose a quantum scrambling-integrated scheme that optimizes the selection of initial states and parameter encoding by exploiting the nonlinear dynamical trajectories and their geometric properties in phase space. We consider a two-mode Bose-Einstein condensate as a typical example. Within a certain parameter range, using uncorrelated spin coherent states located at saddle points and separatrices enables the measurement precision to reach the Heisenberg scaling. An analytical expression for the quantum Fisher information under long-time evolution is provided, and our analysis further reveals that the enhancement mechanism is governed by the sensitivity of the energy spectrum to parameter variations. Compared to the criticality-enhanced protocol that requires parameters to approach the critical point, our scheme offers a distinct advantage in its applicability over a broader range of parameters. Moreover, this scheme can also be applied to models that include one- and two-axis countertwisting interactions and chaotic dynamics. This Letter offers a new perspective for advancing the application of nonlinear dynamics to enhance the parameter estimation precision in quantum systems.
AI Summary
The paper asks how classical phase-space geometry can guide the choice of initially uncorrelated states for metrology when nonlinear dynamics simultaneously generate correlations and encode an unknown parameter. Its quantum scrambling-integrated scheme selects spin-coherent initial states near saddle points or on their separatrices, rather than separating entangled-state preparation from parameter encoding. The main example is a two-mode Bose–Einstein condensate with $H=\Omega S_x-\chi S_z^2$, estimating $\Omega$. Mean-field trajectories are compared with numerically calculated quantum Fisher information (QFI) across initial-state phase space. For $\omega=2\Omega/(N\chi)<2$, a saddle exists; large QFI concentrates along its separatrices, particularly for $\omega\leq1$. The paper reports $\mathcal F_q\sim N^2$ for separatrix initial states, which can outperform the saddle itself. At representative optimized points with $\omega=1$ and $0.5$, long-time QFI is reported to exceed the critical-point benchmark at $\omega=2$ by factors of approximately $3$ and $10$, respectively. A long-time approximation, $\mathcal F_q\simeq4t^2\operatorname{Var}_{p_n}(\partial_\Omega E_n)$, relates the enhancement to the distribution of energy-level sensitivities populated by the initial state. Numerical phase-space comparisons extend the observation to two-axis countertwisting and periodically driven chaotic dynamics. Classical Fisher information and calculations including decoherence are also reported to retain a separatrix advantage. The analysis distinguishes a low-excitation, short-time QFI approximation from the long-time spectral approximation; neither supplies a uniform description of all evolution times. The proposed initial-state selection requires a phase-space model of the parameter-dependent dynamics.
Error found: The supplied paper asserts Heisenberg scaling without displaying an N-scaling calculation or specifying the fixed-resource convention needed to interpret that assertion. Equation (6) prints an inverse power in its denominator that produces Delta^6 suppression at fixed time near criticality, rather than the behavior of the corresponding expression with denominator 16 Delta^3. The printed TACT equations omit factors of N multiplying chi', so the claimed saddle positions do not solve those equations, although they do solve the equations derived from the stated Hamiltonian. The assignments of P0 and P2 are reversed between the initial point definitions and the subsequent noncritical-to-critical QFI ratios.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Journal reference
- Phys. Rev. Lett. 136, 180203 (2026)
- DOI
- 10.1103/4sn5-ngdg
- Comments
- 7 pages, 5 figures