Quantum Sensing

Quantum sensing uses quantum systems to estimate fields, forces, phases, frequencies, temperatures, accelerations, rotations, material noise, or model parameters. From the open-system point of view, a sensor is not only a state and Hamiltonian. It is a driven, noisy, measured system whose environment may be both the signal and the limitation.

The examples below connect decoherence and measurement records to attainable precision. For complete magnetic-field protocols, see Quantum Magnetometry; for instruments and calibration, see Magnetometry.

Parameters Encoded in Noisy Quantum States

A sensing calculation begins by specifying which parameter is unknown and how it changes the measurement probabilities. Quantum Measurement as Estimation develops the statistical framework. An open-system model supplies

$$p(x\mid\theta) = \operatorname{Tr} \left[ E_x\rho_\theta \right],$$

where $\theta$ is the unknown parameter, $\rho_\theta$ is the state after preparation, evolution, noise, and control, and $\{E_x\}$ is the measurement POVM. The classical Fisher information of this model controls local statistical sensitivity:

$$\operatorname{Var}(\hat\theta) \ge \frac{1}{N\mathcal I(\theta)}$$

for $N$ independent repetitions and an unbiased estimator under regularity assumptions.

Quantum Fisher information optimizes over measurements, but the laboratory problem is usually more concrete: which state can be prepared, which control sequence is stable, which measurement is available, and which noise model is credible?

Phase Accumulation

Many sensors reduce to phase estimation. If

$$U_\theta = e^{-i\theta G},$$

then the generator $G$ determines how strongly the state moves in Hilbert space as $\theta$ changes. For a pure noiseless probe, the quantum Fisher information is

$$F_Q = 4(\Delta G)^2 .$$

This formula explains why superpositions of well-separated generator eigenvalues are sensitive. It also explains why they are fragile: the same separation that gives large phase response often makes the state vulnerable to dephasing by environmental fluctuations coupled to $G$ or a nearby observable.

For a spin or two-level sensor in a field $B$, a Ramsey-style phase is

$$\phi = \gamma B T,$$

where $\gamma$ is the relevant coupling and $T$ is interrogation time. The phase signal grows with $T$, but coherence usually decays with $T$.

Decoherence-Limited Sensitivity

A simple Ramsey signal has the form

$$\langle M\rangle = C(T) \cos(\gamma B T+\phi_0),$$

where $C(T)$ is contrast after decoherence, control error, and readout loss. Near the steepest fringe, a common scaling estimate is

$$\delta B \sim \frac{1}{ \gamma T C(T)\sqrt{N} }.$$

This expression is not a universal formula; it is a useful warning. Longer interrogation time increases phase accumulation but often decreases contrast. If

$$C(T) = e^{-T/T_2},$$

then the optimum depends on what is held fixed. At fixed number of shots $N$, this model gives $T=T_2$. At fixed total averaging time $\mathcal T$ with negligible overhead, $N=\mathcal T/T$ and the optimum is $T=T_2/2$. More generally $N\simeq\mathcal T/(T+t_d)$ for dead time $t_d$. Non-exponential decay and sequence constraints can move the optimum further.

The practical sensitivity of a sensor is therefore not determined by $T_2$ alone. It also depends on initialization fidelity, duty cycle, contrast, detector efficiency, calibration, environmental drift, dead time, and whether the signal waveform matches the control sequence.

Measurement Backaction

Every useful sensor must turn a quantum state into a classical record. That record carries information, and the interaction that creates it can disturb the sensor.

In a weak continuous measurement, the record may be idealized as

$$dI_t = 2\sqrt{\eta k}\, \langle A\rangle_t dt + dW_t,$$

where $A$ is Hermitian and dimensionless, $k$ has units of inverse time, $\eta$ is efficiency, and $dW_t^2=dt$. This record is paired with $k\mathcal D[A]\rho\,dt+\sqrt{\eta k}\mathcal H[A]\rho\,dW_t$ in the conditional equation. Keeping the record gives a conditional stochastic evolution. Discarding the record leaves dephasing in the $A$ basis.

Backaction is not synonymous with imperfection. An ideal measurement still changes the conditional state because information has become correlated with the detector. The design question is whether the backaction damages the parameter of interest, can be evaded, or can itself be used as the signal.

Optomechanical force sensing makes the tradeoff especially explicit: optical shot noise causes imprecision, while radiation-pressure fluctuations cause force backaction. The platform version is discussed in Optomechanics.

Standard Quantum Limit and Beyond

For independent probes under ordinary conditions, uncertainty often scales as

$$\delta\theta \propto \frac{1}{\sqrt{N}},$$

where the resource $N$ must be specified. Independent repetitions improve an estimator by $1/\sqrt N$ under regular conditions; an entangled-probe comparison instead varies the number of coherently encoded probes per interrogation. This is often called a standard quantum limit, but the phrase has several meanings across fields.

Ideal entangled or squeezed resources can improve a specific task. For a noiseless phase shift with a suitable generator, one sometimes obtains a scaling closer to

$$\delta\theta \propto \frac{1}{N}.$$

That ideal scaling is not a promise of practical advantage. Loss, dephasing, imperfect state preparation, detector inefficiency, and calibration overhead often erase the asymptotic gain. A trustworthy sensing claim states the resource count, noise model, estimator, bandwidth, and comparison baseline.

Squeezed States

Squeezing redistributes quantum uncertainty. For a bosonic field quadrature, a squeezed state can reduce noise in the measured quadrature while increasing noise in the conjugate quadrature. For a collective spin, spin squeezing reduces noise perpendicular to the mean spin direction at the cost of increased noise elsewhere.

For $N$ spin-1/2 particles and dimensionless collective spin $\mathbf J$, a common spin-squeezing metrology parameter is

$$\xi_R^2 = \frac{ N(\Delta J_\perp)^2 }{ |\langle\mathbf J\rangle|^2 }.$$

Values $\xi_R^2\lt1$ indicate phase sensitivity beyond the corresponding uncorrelated coherent-spin-state benchmark, assuming the contrast loss and readout model are included consistently.

Squeezing is useful only when the squeezed quadrature is the one that limits the measurement. Anti-squeezing, phase noise, optical loss, detector loss, and technical noise can rotate or leak the increased uncertainty into the measured quadrature.

For oscillator squeezing, see Squeezed States: First Encounter. For optical and two-mode resources, see Squeezed States as Entangled Modes.

Noise Spectroscopy

Sometimes the environment is the target. A controlled sensor can measure a noise spectrum by converting environmental fluctuations into phase, population change, or relaxation.

For a spin sensor under a modulation function $y(t)$, a classical magnetic fluctuation $B(t)$ produces phase

$$\phi = \gamma \int_0^T y(t)B(t)\,dt .$$

For centered stationary classical Gaussian noise, define the two-sided spectrum $S_B(\omega)=\int dt\,e^{i\omega t}\langle B(t)B(0)\rangle$ and $Y(\omega,T)=\int_0^Tdt\,y(t)e^{i\omega t}$. The Gaussian phase average gives

$$\chi(T) = \frac{\gamma^2}{2\pi} \int_0^\infty d\omega\, S_B(\omega) |Y(\omega,T)|^2,$$

The factor $1/2$ comes from $\chi=\langle\phi^2\rangle/2$. For non-Gaussian noise, higher phase cumulants can contribute. The noise-induced contrast under the stated assumptions is

$$C(T) = e^{-\chi(T)}.$$

This is the sensing version of noise spectroscopy. Ramsey, echo, Carr–Purcell, Uhrig, and other sequences select different frequency windows. The same idea appears in NV-center sensing, spin-qubit noise spectroscopy, superconducting qubits, trapped ions, and optomechanics.

Relaxometry

Relaxation can also be a signal. If environmental transverse noise has spectral weight near a transition frequency $\omega_0$, it changes the longitudinal relaxation rate:

$$\frac{1}{T_1} \propto S_\perp(\omega_0),$$

with convention-dependent constants and matrix elements. This is relaxometry. It is useful when the target is noise near a known transition, such as magnetic fluctuations from surfaces, spin baths, conductors, or materials.

Relaxometry and phase sensing probe different spectral regions. Ramsey-style sensing is sensitive to low-frequency or controlled-band noise. $T_1$ relaxometry is sensitive near transition frequencies. A sensor paper should say which one is being used.

Platform Map

Open-system sensing ideas appear in many platforms:

Platform	Typical signal	Main open-system limitation
atomic clocks	frequency shift	oscillator noise, decoherence, dead time
trapped ions	force, field, heating rate	motional heating, photon scattering
NV centers	magnetic field, temperature, material noise	surface noise, photon collection, spin bath
spin qubits	local electric, magnetic, charge noise	dephasing, relaxation, readout backaction
optomechanics	force, displacement, acceleration	imprecision-backaction tradeoff, heating
quantum optics	phase, absorption, displacement	loss, detector inefficiency, mode mismatch
mesoscopic devices	charge, current, noise	tunneling backaction, amplifier noise

The same words can hide different resources. “One photon,” “one atom,” “one trial,” and “one second of averaging” are not interchangeable resource counts.

Exercises

Ramsey Optimization with Exponential Contrast

For the rough sensitivity model

$$\delta B(T) = \frac{e^{T/T_2}}{\gamma T\sqrt{N}},$$

first hold $N$ fixed and find the optimum $T$. Then hold total averaging time $\mathcal T=NT$ fixed with zero overhead and repeat.

Solution

Minimize the logarithm:

$$\log \delta B = \frac{T}{T_2} - \log T + \text{constant}.$$

Differentiate:

$$\frac{d}{dT} \log\delta B = \frac{1}{T_2} - \frac{1}{T}.$$

The optimum satisfies

$$T = T_2.$$

At fixed total time, substitute $N=\mathcal T/T$. Then $\log\delta B=T/T_2-\tfrac12\log T+\text{constant}$, giving $T=T_2/2$. The two answers optimize different resources. Dead time, non-exponential contrast, and pulse-sequence constraints change these results.

Fisher Information for a Fringe

Consider a binary measurement with probability

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

Compute the Fisher information for $\phi$ away from points where the probability model is singular.

Solution

The probabilities are

$$p_+ = \frac{1+\cos\phi}{2}, \qquad p_- = \frac{1-\cos\phi}{2}.$$

Their derivatives are

$$\partial_\phi p_+ = - \frac{\sin\phi}{2}, \qquad \partial_\phi p_- = \frac{\sin\phi}{2}.$$

Thus

$$\mathcal I(\phi) = \frac{(\partial_\phi p_+)^2}{p_+} + \frac{(\partial_\phi p_-)^2}{p_-} = 1$$

where both probabilities are nonzero. The ideal two-outcome fringe carries unit Fisher information about phase per shot in this simple model.

Squeezing Parameter

A collective spin state has $N=10^4$, $|\langle\mathbf J\rangle|=4800$, and $(\Delta J_\perp)^2=1200$. Compute $\xi_R^2$ and say whether the state beats the coherent-spin benchmark according to this criterion.

Solution

Use

$$\xi_R^2 = \frac{ N(\Delta J_\perp)^2 }{ |\langle\mathbf J\rangle|^2 }.$$

Substitution gives

$$\xi_R^2 = \frac{10^4\times1200}{4800^2} \simeq 0.52 .$$

Since $\xi_R^2\lt1$, the state beats the corresponding coherent-spin-state phase benchmark, assuming the measurement and contrast assumptions behind this parameter are appropriate.

Relaxometry Versus Ramsey

Why does $T_1$ relaxometry probe a different part of the environmental spectrum than a Ramsey measurement?

Solution

Ramsey sensing converts low-frequency or controlled-band fluctuations into phase during an interrogation time. Its sensitivity is set by the modulation function and the accumulated phase. $T_1$ relaxometry measures transitions between energy eigenstates, so it is driven by transverse environmental noise near the transition frequency. The two protocols filter the same environment differently.

References

• C. M. Caves, “Quantum-mechanical noise in an interferometer,” Physical Review D 23, 1693-1708 (1981).
• A. A. Clerk, M. H. Devoret, S. M. Girvin, F. Marquardt, and R. J. Schoelkopf, “Introduction to quantum noise, measurement, and amplification,” Reviews of Modern Physics 82, 1155-1208 (2010).
• C. L. Degen, F. Reinhard, and P. Cappellaro, “Quantum sensing,” Reviews of Modern Physics 89, 035002 (2017). arXiv:1611.02427.
• V. Giovannetti, S. Lloyd, and L. Maccone, “Quantum metrology,” Physical Review Letters 96, 010401 (2006).
• V. Giovannetti, S. Lloyd, and L. Maccone, “Quantum-enhanced measurements: beating the standard quantum limit,” Science 306, 1330-1336 (2004).
• D. J. Wineland, J. J. Bollinger, W. M. Itano, and D. J. Heinzen, “Squeezed atomic states and projection noise in spectroscopy,” Physical Review A 50, 67-88 (1994).