Sensing and Metrology

Quantum sensing uses a controlled quantum response to infer a physical quantity. Quantum metrology asks how accurately that inference can be made with specified states, controls, measurements, time, and other resources. This chapter connects those questions: a precision bound is useful only when its parameter, resource count, noise model, and measurement can be related to the experiment that produces the data.

Helpful background. Density operators for QI supplies parameter-dependent states, measurement in circuits supplies outcome probabilities, and common noise models supplies elementary noisy channels. Familiarity with variances and conditional probabilities is enough to begin the estimation route.

Enter this chapter

Start with quantum measurement as estimation. A sensor produces observations; an estimator turns those observations into an answer. Read classical and quantum Fisher information to distinguish information in the chosen measurement from information available in the state, then Cramér–Rao bounds to identify the regularity, bias, and attainability assumptions.

For a first worked physical route, continue to Ramsey interferometry, the standard quantum limit, spin squeezing, and atomic clocks. This route connects a phase-sensitive likelihood to a laboratory estimator, a resource comparison, and a device with a continuously operating feedback loop.

For an optical route, follow Mach–Zehnder interferometry, squeezing, and Heisenberg scaling. Return to noise and decoherence in metrology before interpreting ideal scaling as an achievable experimental gain.

A likelihood, a bound, and a resource count

For a scalar parameter $\theta$, a prepared state $\rho_\theta$, and a parameter-independent POVM $\{E_x\}$, the experimental likelihood is

$$p(x\mid\theta)=\operatorname{Tr}(E_x\rho_\theta).$$

The measurement and state both matter. For $\nu$ independent identical trials, under the regularity and local unbiasedness assumptions of the scalar Cramér–Rao theorem,

$$\operatorname{Var}_\theta(\hat\theta) \geq \frac{1}{\nu F_C(\theta)} \geq \frac{1}{\nu F_Q(\theta)}.$$

Here $F_C$ is the Fisher information of one recorded trial and $F_Q$ is the quantum Fisher information of the encoded probe state. A measurement setting may be chosen from earlier data, but its value is held fixed when taking the local parameter derivative for that trial. The rightmost quantity is a lower bound, not a guarantee that the installed measurement and estimator achieve it. The Fisher-information and bounds pages develop the definitions, multiparameter qualifications, and failure modes.

For orientation, an ideal Ramsey experiment measuring a dimensionless phase $\phi$ can have

$$p(1\mid\phi)=\frac{1+\cos\phi}{2}.$$

At a regular operating point this Bernoulli likelihood has $F_C=1$ per probe. With $N$ independent probes in each of $\nu$ independent rounds, the local phase standard deviation is bounded below by $1/\sqrt{\nu N}$. If instead $\phi=\omega t$ and the unknown parameter is a frequency $\omega$, its single-probe Fisher information is $t^2$. Reporting a phase precision as a frequency precision without the interrogation time changes the problem.

The preparation time, readout time, duty cycle, bandwidth, and phase-reference resources matter when rounds are compared at fixed total laboratory time. Degen, Reinhard, and Cappellaro (2017) give a broad experimental perspective on these connections between response, control, and sensitivity.

Choose the physical quantity before the platform

Magnetometry explains how field component, waveform, bandwidth, and spatial resolution define the target. Its NV-center sensing specialization connects spin control and optical readout to several distinct sensing modes. Rydberg electrometry uses an atomic electric-field response; a spectral splitting requires a model before it becomes a calibrated field estimate.

Gravimetry and inertial sensing uses matter-wave phases relative to controlled optical references. The acceleration, rotation, vibration, and reference-frame contributions must be separated for the claimed quantity. These platform pages explain the information-extraction task; AMO and Quantum Matter explain the underlying atomic and material mechanisms.

Two routes broaden the statistical task. Distributed quantum sensing asks about a specified function of parameters across a network, such as a weighted average or difference. Its comparator must have the same total probe and communication resources. Quantum illumination is a noisy channel-discrimination problem: error probability for deciding whether a target is present replaces variance of a continuous-parameter estimator. A Cramér–Rao formula does not automatically answer that question.

Interpret enhancement with its conditions

The standard quantum limit and Heisenberg scaling compare specified families of experiments. Neither is a context-free limit for every sensor. Entanglement may increase ideal parameter sensitivity while increasing vulnerability to loss or dephasing; squeezing may reduce measured noise while also reducing signal contrast. Conversely, a useful quantum sensor need not outperform a classical competitor on every axis.

Read sensing case studies after one protocol route to compare resolution, stability, bandwidth, dynamic range, and uncertainty with matched operating conditions. Research follows changing experimental results and open questions. The enduring question here is which assumptions justify the reported information and precision.

Review the chapter

An entangled probe has four times the Fisher information per round of a reference probe, but each round takes four times as long. Does it improve precision at fixed total time? Assume independent rounds and no other costs.

Answer

The information accumulated in total time $T$ is approximately $(T/\tau)F_C$ when enough complete rounds fit in $T$. Multiplying both $F_C$ and $\tau$ by four leaves that quantity unchanged, so the local Cramér–Rao bound is unchanged. This does not prove identical practical performance: preparation failures, estimator bias, drift, and useful bandwidth may still differ. It establishes only that the per-round information gain alone is insufficient for the fixed-time claim.

References

• Degen, Christian L., Friedemann Reinhard, and Paola Cappellaro. “Quantum sensing.” Reviews of Modern Physics 89, 035002 (2017). doi:10.1103/RevModPhys.89.035002; arXiv:1611.02427 — experimental sensing protocols, control, and sensitivity.
• Holevo, Alexander S. Probabilistic and Statistical Aspects of Quantum Theory. 2nd English edition. Edizioni della Normale, 2011 — the statistical structure of quantum estimation and measurement-dependent precision bounds.