Non-Hermitian-enhanced quantum sensing in an optical interferometer

Technical Summary: Non-Hermitian-enhanced quantum sensing in an optical interferometer

Problem Statement
The precision of quantum parameter estimation is traditionally bounded by the quantum Cramér-Rao bound (QCRB), derived within a Hermitian measurement framework where observables yield real-valued outcomes. While recent theoretical developments in non-Hermitian physics suggest that non-Hermitian observables could offer enhanced sensitivity for parameter estimation, experimental realization has been challenging. Standard quantum measurements exclude non-Hermitian operators, and previous experimental approaches often relied on non-Hermitian evolution to encode parameters rather than directly implementing a non-Hermitian observable readout. The core challenge addressed is how to experimentally measure the complex expectation value of a non-Hermitian observable and determine if this approach yields a reduced error-propagation variance compared to optimal Hermitian strategies, particularly in the presence of noise.

Methodology
The authors experimentally realized quantum parameter estimation using a single-photon linear optical platform. The methodology involves four primary stages:

1. State Preparation: Single photons were generated via spontaneous parametric down-conversion (SPDC). The probe state was prepared as a mixed state $\rho_0 = p |\Psi_0\rangle\langle\Psi_0| + (1-p) |\Psi_1\rangle\langle\Psi_1|$ using an unbalanced Mach-Zehnder interferometer, where the parameter $p$ controls the purity.
2. Parameter Encoding: The parameter $\theta$ to be estimated was encoded into the probe state via a unitary operation $M_\theta = e^{i\theta\sigma_z/2}$.
3. Non-Hermitian Observable Readout: The team constructed a Sagnac interferometer to measure a specific non-Hermitian observable $A$. Since direct measurement of non-Hermitian operators is not standard, they decomposed the observable as $A = UR$, where $U$ is unitary and $R$ is a positive Hermitian operator.
   • In one arm of the interferometer (path-1), the photon underwent the unitary evolution $U^\dagger$.
   • In the other arm (path-2), the photon underwent a normalized positive operation $R'$. This was implemented using a Mach-Zehnder interferometer with beam displacers and wave plates to introduce controllable, polarization-dependent photon loss, effectively simulating the non-unitary operation.
   • The two paths were recombined to form interference fringes.
4. Data Reconstruction: By measuring the maximum ($I_{max}$) and minimum ($I_{min}$) photon counts and the phase shift of the interference pattern, the authors reconstructed the complex expectation value $\langle A \rangle$. The variance of the parameter estimate $(\Delta\theta)^2$ was then calculated using the error-propagation formula based on this reconstructed expectation value.

The experiment was conducted under two conditions: a noiseless environment and an environment with amplitude-damping noise (simulated by an additional Sagnac interferometer).

Key Contributions and Results

• Experimental Realization: The work provides the first experimental demonstration of quantum parameter estimation using a directly implemented non-Hermitian observable on a linear optical platform, reconstructing its complex expectation value from interference fringes.
• Reduced Variance: The authors observed that for the same probe-state model, the error-propagation variance inferred from the optimal non-Hermitian observable was smaller than that obtained from the optimal Hermitian observable. This advantage was demonstrated across a broad range of the state parameter $p$.
• Noise Robustness: Under amplitude-damping noise, the non-Hermitian readout maintained a smaller variance compared to the Hermitian benchmark. Furthermore, the noise-induced variance change rate was significantly lower for the non-Hermitian observable, indicating greater robustness against decoherence.
• Fisher Information Analysis: The authors analyzed the complete optical measurement as a physical Positive-Operator-Valued Measure (POVM), including all output ports (bright ports and the loss port). They calculated the classical Fisher information (CFI) for this complete POVM and found that it does not exceed the standard Hermitian optimum.

Significance and Claims
The paper claims that the observed enhancement is a "structured non-Hermitian readout advantage" rather than a fundamental violation of the standard quantum metrological limits. The significance of the work lies in clarifying the operational meaning of non-Hermitian observable measurements:

• Non-Hermitian readout strategies can reshape the distribution of information among different measurement channels, providing practical advantages in selected readout configurations (specifically, when focusing on the reconstructed expectation value of the non-Hermitian operator).
• This advantage remains consistent with the standard quantum limits when the complete POVM (including loss channels) is considered.
• The results offer an experimental route to non-Hermitian observable readout, demonstrating that such strategies can improve sensing precision in specific operational contexts without contradicting fundamental quantum bounds.