Quantum error correction for multiparameter metrology

TL;DR

This work addresses the loss of quantum advantage in multiparameter metrology with GHZ probes by introducing a quantum-error-correction framework that renders the problem effectively single-parameter. An ancilla-assisted protocol with a single GHZ probe uses an $(N{+}1)$-qubit bit-flip code to completely suppress one error channel, enabling fixed-separable measurements while preserving information via the stabilizer outcomes; dual- and triple-probe constructions restore Heisenberg scaling and saturate the QFIM bound in most cases. The authors provide detailed scaling analyses, 3D generalizations, and Bayesian simulations that demonstrate practical, few-shot Heisenberg performance. Together, these results offer a robust path to robust, high-precision multiparameter sensing with GHZ-like probes using minimal measurement complexity.

Abstract

For single-parameter sensing, Greenberger-Horne-Zeilinger (GHZ) probes achieve optimal quantum-enhanced precision across the unknown parameter range, solely relying on parameter-independent separable measurement strategies for all values of the unknown parameter. However, in the multiparameter setting, a single GHZ probe not only fails to achieve quantum advantage but also the corresponding optimal measurement becomes complex and dependent on the unknown parameters. Here, we provide a recipe for multiparameter sensing with GHZ probes using quantum error correction techniques by treating all but one unknown parameters as noise, whose effects can be corrected. This strategy restores the core advantage of single parameter GHZ-based quantum sensing, namely reaching optimally quantum-enhanced precision for all unknown parameter values while keeping the measurements separable and fixed. Specifically, given one shielded ancilla qubit per GHZ probe, our protocol extracts optimal possible precision for any probe size. While this optimal precision is shot-noise limited for a single GHZ probe, we recover the Heisenberg scaling through use of multiple complementary GHZ probes. We demonstrate the effectiveness of the protocol with Bayesian estimation.

Figures (6)

• Figure 1: Schematic of the protocol (ancilla-assisted single-probe $|\textrm{GHZ}_Z \rangle$ variant): The probe is prepared in the state $|\textrm{GHZ}_Z \rangle \propto |0 \rangle ^{\otimes (N+1)} + |1 \rangle ^{\otimes (N+1)}$ via an initial Hadamard gate and chained CNOTs. After all but one of the probe's qubits interact with the magnetic field, QEC is performed by measuring the stabilizers of the $(N+1)$-qubit bit-flip code. After applying the appropriate correction and returning the state to the codespace, the $X$-component of the magnetic field is completely suppressed and the final state is proportional to $|0 \rangle ^{\otimes (N+1)} + e^{i\phi_k} |1 \rangle ^{\otimes (N+1)}$. The acquired relative phase angle will depend on the number of detected $X$ errors, $k$, as well as $B_x$ and $B_z$ (since $\phi_k = 2 B_\textrm{eff} (N-k)$ and $B_\textrm{eff} = \arctan[B_z \tan(Bt)/B]$). Estimating $\phi_k$ corresponds to a single-parameter quantum sensing problem for which the optimal measurement basis is known to be $X$. We can then combine this protocol with an analogous one employing the phase-flip code to achieve Heisenberg scaling in the joint estimation of the two unknown parameters, $B_x$ and $B_z$.
• Figure 2: Ancilla-assisted single GHZ probe : (a) Heatmap of sensing precision quantified by $\mathrm{Tr}[F^{-1}]$ with respect to $(B_x,B_z)$. (b) Scaling with number of qubits $N$ of precision obtained via CFIM, i.e., $\mathrm{Tr}[F^{-1}]$ (hollow points) and via QFIM, i.e., $\mathrm{Tr}[Q^{-1}]$ (solid lines) for three different $(B_x,B_z)$ values ($t{=}1$ throughout).
• Figure 3: Ancilla-assisted dual GHZ probe : (a) Heatmap of sensing precision quantified by $\mathrm{Tr}[F^{-1}]$ with respect to $(B_x,B_z)$. (b) Scaling with number of qubits $N$ of precision obtained via CFIM, i.e., $\mathrm{Tr}[F^{-1}]$ (hollow points) and via QFIM, i.e., $\mathrm{Tr}[Q^{-1}]$ (solid lines) for three different $(B_x,B_z)$ values ($t{=}1$ throughout).
• Figure 4: Ancilla-free single GHZ probe : Precision Scaling with number of qubits $N$ of precision obtained via CFIM, i.e., $\mathrm{Tr}[F^{-1}]$ (hollow points) and via QFIM, i.e., $\mathrm{Tr}[Q^{-1}]$ (solid lines) for two different $(B_x,B_z)$ values ($t{=}1$ throughout).
• Figure 5: Heisenberg limit with Bayesian analysis: The left column shows the trace of the rescaled covariance matrix $M\mathrm{Cov}[\mathfrak{B}]$ alongside the inverse trace of the CFIM $\mathrm{Tr}[F^{-1}]$ as functions of $N$. Panel (a) corresponds to a single probe state $|\text{GHZ}_X\rangle$ that corrects $Z$ errors, panel (b) to a single probe $|\text{GHZ}_Z\rangle$ correcting $X$ errors, and panel (c) to the combined use of two probes, $|\text{GHZ}_X\rangle$ and $|\text{GHZ}_Z\rangle$, correcting $X$ and $Z$ errors, respectively. The right column shows the estimated values $(B_x^{\mathrm{est}}, B_z^{\mathrm{est}})$ for $N = 50$. Each point represents an estimate obtained from $M = 4000$ measurements, using either the single probe $|\textrm{GHZ}_X\rangle$, $|\textrm{GHZ}_Z\rangle$, or both.