Joint estimation of noise and nonlinearity in Kerr systems

TL;DR

This paper investigates the fundamental limits of simultaneously estimating loss, dephasing, and nonlinearity in Kerr quantum channels using quantum Fisher information, revealing how nonlinearity affects estimation precision and measurement strategies.

Contribution

It provides a detailed analysis of the quantum limits for joint parameter estimation in Kerr channels, highlighting the impact of nonlinearity and loss on measurement optimality.

Findings

Loss estimation improves with Kerr nonlinearity at low energy

Nonlinearity estimation is degraded by loss

Homodyne detection is nearly optimal in low-energy regimes

Abstract

We address characterization of lossy and dephasing channels in the presence of self-Kerr interaction using coherent probes. In particular, we investigate the ultimate bounds to precision in the joint estimation of loss and nonlinearity and of dephasing and nonlinearity. To this aim, we evaluate the quantum Fisher information matrix (QFIM), and compare the symmetric quantum Cram\'er-Rao bound (QCR) to the bound obtained with Fisher information matrix (FIM) of feasible quantum measurements, i.e., homodyne and double-homodyne detection. For lossy Kerr channels, our results show the loss characterization is enhanced in the presence of Kerr nonlinearity, especially in the relevant limit of small losses and low input energy, whereas the estimation of nonlinearity itself is unavoidably degraded by the presence of loss. In the low energy regime, homodyne detection of a suitably optimized quadrature represents a nearly optimal measurement. The Uhlmann curvature does not vanish, therefore loss and nonlinearity can be jointly estimated only with the addition of intrinsic quantum noise. For dephasing Kerr channels, the QFIs of the two parameters are independent of the nonlinearity, and therefore no enhancement is observed. Homodyne and double-homodyne detection are suboptimal for the estimation of dephasing and nearly optimal for nonlinearity. Also in this case, the Uhlmann curvature is nonzero, proving that the parameters cannot be jointly estimated with maximum precision.