Source-linked AI summary
Optimal and Secure Measurement Protocols for Quantum Sensor Networks
Zachary Eldredge, Michael Foss-Feig, Jonathan A. Gross, Steven L. Rolston, Alexey V. Gorshkov
TL;DR
The paper asks how entanglement improves estimation of a linear combination of parameters distributed across a quantum sensor network. It derives a multi-parameter quantum Fisher-information bound and introduces partial time evolution using GHZ or spin-squeezed inputs. The GHZ protocol saturates the bound, while its advantage is limited when decoherence dominates.
Problem
The paper addresses how to quantify entanglement's metrological advantage when separate qubits couple to separate parameters and the desired quantity is their weighted sum.
Method
The paper derives a multi-parameter quantum Fisher-information bound and proposes partial time evolution with GHZ or spin-squeezed input states.
Results
The GHZ partial-time-evolution protocol saturates the quantum bound for estimating α · θ.
Takeaways & Limitations
The protocol extends entanglement-enhanced Ramsey spectroscopy to spatially varying quantities and supports secure network-wide measurement of the linear combination.
Takeaways & Limitations
GHZ states are extremely sensitive to noise, and decoherence can nullify their entanglement advantage when interrogation time is limited by coherence time.
Abstract
from arXiv · showhide
Studies of quantum metrology have shown that the use of many-body entangled states can lead to an enhancement in sensitivity when compared to product states. In this paper, we quantify the metrological advantage of entanglement in a setting where the quantity to be measured is a linear function of parameters coupled to each qubit individually. We first generalize the Heisenberg limit to the measurement of non-local observables in a quantum network, deriving a bound based on the multi-parameter quantum Fisher information. We then propose a protocol that can make use of GHZ states or spin-squeezed states, and show that in the case of GHZ states the procedure is optimal, i.e., it saturates our bound.
I. INTRODUCTION
The paper studies estimating a weighted sum of parameters distributed across spatially separated qubits, where entanglement can improve precision beyond local measurements. It introduces partial time evolution to achieve the quantum-limited variance and protect the result's secrecy.
- I. INTRODUCTION: Unlike conventional entangled metrology, the network couples N parameters to N potentially separated systems and measures one linear combination of them.Potential applications include geodesy, geophysics, medical imaging, and nanoscale NMR.
- I. INTRODUCTION: The task is estimating q = α · θ, a weighted sum of unknown parameters θ_i coupled individually to N sensor qubits.The weights α_i are known, with |α_i| ≤ 1 and at least one weight equal to 1.
- I. INTRODUCTION: The proposed partial-time-evolution protocol turns each qubit's parameter coupling on for a duration proportional to its weight in q.The protocol uses entangled probes and controls when individual qubits evolve under their local parameters.
- I. INTRODUCTION: The protocol reaches the minimum variance permitted by quantum mechanics and extends the Heisenberg limit to linear combinations.The same network can perform the measurement without eavesdroppers learning details of α · θ.
- I. INTRODUCTION: Local measurements establish an entanglement-free benchmark by estimating each θ_i separately and combining the results classically.For identical local variance Var Θ0, the combined variance scales with the squared norm of α.
A. Using Fisher Information Matrix
The paper derives a multi-parameter quantum Fisher-information bound for estimating α · θ and shows that numerical optimization confirms the bound across relative parameter weights. The bound is controlled by the largest weighted contribution from an individual site.
- A. Using Fisher Information Matrix: The multi-parameter Cramér–Rao framework bounds the variance of estimating the scalar α · θ through α^T F^-1 α.The quantum version replaces classical Fisher information with quantum Fisher information, with a projection used when the matrix is singular.
- A. Using Fisher Information Matrix: The resulting bound remains valid under arbitrary time-dependent control Hamiltonians, including controls using ancilla qubits.The proof applies a single-parameter generator bound across all parameter components.
- A. Using Fisher Information Matrix: For the qubit generators in Eq. (2), the bound selects the largest contribution from one site and generalizes the usual Heisenberg bound to linear combinations.The average θ̄ recovers Heisenberg scaling, while estimating only one θ_i provides no entanglement benefit.
- A. Using Fisher Information Matrix: Numerical optimization for two qubits confirms the predicted lower bound as the relative weight α_2 changes.When α_2^2 exceeds 1/4, the second-qubit bound dominates; when α_2^2 is below 1/4, the lower bound remains 1/4.
B. Using Single-Parameter Bounds
Naive single-parameter bounds can be misleading when multiple unknown parameters influence the evolution. The paper tightens the bound by accounting for field structure and optimizing over re-parameterizations, while showing when prior knowledge can enable stronger results.
- The single-parameter Cramér–Rao approach is looser because it implicitly constrains the other components of the field.
- A basis choice can make the inferred Fisher-information bound diverge, yielding arbitrarily small lower bounds on estimator variance.
- Optimizing over all basis choices produces the tightest valid Fisher-information bound rather than relying on one potentially pathological re-parameterization.
- The optimized seminorm equals 1, giving Var(α · θ) ≥ 1/t^2, matching the paper’s tighter bound.
- If the field structure is known, a genuinely one-parameter problem can permit strategies that saturate the single-parameter bound and outperform the general result.
- Without such knowledge, learning components perpendicular to α is generally as difficult as learning the parallel component, so measuring α · θ remains optimal from ignorance.
IV. PROTOCOLS
The paper presents two optimal protocols that saturate the bound: one uses GHZ or spin-squeezed states with time-dependent control, and the other uses a more complex initial state without during-accumulation control.
- Two protocols saturate the bound: controlled phase accumulation from GHZ or spin-squeezed states, and an uncontrolled alternative using a more complicated initial state.
1. Using GHZ Input State
The GHZ protocol applies weighted phase evolution through partial-time control and reads out the resulting global parity. It achieves the quantum variance bound, while its optimality depends on estimating the full function without separately knowing individual parameters.
- Protocol: The GHZ protocol uses partial time evolution so qubit i accumulates phase proportional to α_i, producing a total phase α · θt.Control pulses implement effective evolution for time α_it while respecting α_i ∈ [-1,1].
- Protocol: The final measurement is the overall parity of locally measured σx outcomes.The parity expectation is measured over time to estimate Q.
- Optimality: The protocol's measurement saturates the variance bound for the linear combination being estimated.The paper states that the time-dependent parity measurement saturates the bound in Eq. (16).
- Optimality: GHZ evolution achieves the Fisher-information bound for the basis vector associated with the largest coefficient, α_b = 1.The projected quantum Fisher information is analyzed on the image of FQ.
- Optimality: The protocol is optimal because information about one parameter cannot improve the target estimate without also knowing the other parameters.With prior knowledge, the same bound applies asymptotically for M ≫ 1 and the scheme also saturates it.
- Security: The GHZ protocol can hide the network result from an eavesdropper who receives only a subset of nodes’ measurement results.Tracing out the central qubit removes phase information, even when that qubit is unweighted.
2. Using Spin Squeezed States
Spin-squeezed states provide an alternative input for partial-time-evolution measurements of linear functions. They can be preferable when decoherence limits interrogation time, although their scaling approaches the bound only up to numerical prefactors.
- State properties: Spin squeezing reduces variance along one collective-spin axis at the cost of increased variance along an orthogonal axis.The squeezing parameter ξ quantifies this trade-off.
- Noise regime: Squeezed states can be useful when decoherence dominates because they are more robust to noise than GHZ states.The paper motivates them for sensor networks with decoherence-limited interrogation times.
- Protocol: Spin-squeezed states can serve as inputs to partial time evolution for measuring linear functions.The protocol combines partial evolution with a final rotation and total-spin measurement.
- Sensitivity: Partial time evolution with spin-squeezed input beats the standard quantum limit when ξ ≤ ∥α∥/√N.The sensitivity is evaluated at q = 0 for small signals.
- Sensitivity: Squeezed states can achieve squeezing proportional to N^-1/2, approaching the bound up to numerical prefactors that do not scale with N.This gives favorable scaling without requiring the exact GHZ protocol.
- Extensions: Other highly entangled states, including Dicke states, may offer favorable scaling under noise with partial time evolution.The paper identifies them as additional candidate input states.
B. Time-Independent Protocols
The time-independent schemes prepare a GHZ-like or spin-squeezed state and then allow free evolution during phase accumulation, rather than using pulses for effective weighted evolution.
- B. Time-Independent Protocols: Time-independent protocols begin with either a GHZ-like state or a spin-squeezed state and use free evolution during phase accumulation.They differ from the pulse-based protocols that implement effective time α_it on qubit i.
1. Using GHZ-like Input State
The GHZ-like protocol generalizes the GHZ state by allowing spins to be aligned, flipped, or disentangled, then randomizes this choice to recover sensitivity to α · θ.
- State construction: The state |τ⟩ uses τ_j ∈ {-1, 0, 1} to specify the configuration of each spin.The construction includes the standard GHZ state as the all-τ_j = 1 case.
- State construction: τ_j = -1 flips spin j relative to GHZ, while τ_j = 0 leaves that spin disentangled.Thus the state can include only a subset of entangled spins.
- Measurement: The protocol evolves |ψ(τ)⟩ and multiplies σx outcomes for the qubits that remain entangled.The resulting observable has outcomes determined by cos^2(θ · τt/2) and sin^2(θ · τt/2).
- Randomization: Randomizing τ selects both the initial state and final measurement, with τ_j = 0 occurring according to a probability set by α_j.This randomized construction produces sensitivity to the desired weighted combination.
- Sensitivity: The resulting measurement has variance 1/t^2 to lowest order, matching the sensitivity of the time-independent protocol.The expansion is taken to lowest order in t for the target α · θ.
2. Using Spin-Squeezed States
The protocol adapts spin-squeezed states to measure linear combinations using qubit-dependent rotations, evolution, and collective measurements. A two-step implementation preserves sensitivity while requiring only single-qubit operations after state preparation.
- Protocol: Qubit-dependent z rotations precede full evolution and collective-spin measurement, replacing partial time evolution in the spin-squeezed protocol.The rotation operator adds a phase to the evolution, with angles chosen according to the target coefficients.
- Protocol: The two-step protocol measures separate quantities using η_i = φ_i and η_i = −φ_i, then combines them.The first step uses cos φ_i = α_i; the second repeats the sequence with the opposite angles.
- Sensitivity: The analysis assumes the detected phases θ_i t are small enough for the small-angle approximation.This assumption is used when evaluating the measurement sensitivity near zero signal.
- Sensitivity: 16Nξ^2/t^2 is the resulting sensitivity bound when both steps share the same total time as one time-dependent round.Using t for each step gives a tighter intermediate bound, but replacing t with t/2 makes the comparison fair.
- Implementation: After the initial squeezed state is created, the protocol requires only single-qubit operations, supporting experimental tractability.This implementation is presented as a potentially more practical route to quantum-enhanced measurements of linear combinations.
V. ENTANGLEMENT-ENHANCED MOLECULAR NMR
The paper identifies nanoscale NMR as an application for entanglement-enhanced sensor networks using spatially separated NV centers. The protocol targets linear combinations and subtraction of spatially separated field signals, with broader relevance to distributed sensing.
- Application setting: Nanoscale NMR is proposed as an application because NV-center spins accumulate phases dependent on local magnetic fields and support demonstrated entangling protocols.NV centers provide two-level quantum sensors and established platforms for quantum information processing.
- Application setting: Linear combinations of spatially separated field values could improve nanoscale NMR imaging and molecular microscopy when measured with entangled NV sensors.The cited applications combine multiple Fourier spatial modes or signals.
- Application setting: The entanglement scheme also performs simple subtraction between two qubits, enabling common-mode noise rejection.The passage introduces subtraction as an additional capability of the network.
- Outlook: Entanglement-enhanced imaging beyond single molecules is identified as a future direction, following near-standard-quantum-limit detection of a single animal neuron.The paper presents surpassing that limit as a natural next experimental step.
- Outlook: The protocol applies beyond NMR to spatially varying quantities in gravimetry, spectroscopy, rotation sensing, and mixed electric-magnetic-field measurements.The paper emphasizes that the parameters need not originate from the same physical source.