Interaction-Based Readout in Quantum Metrology

• Interaction-based readout is a measurement strategy where engineered nonlinear interactions post-encoding convert fragile phase information into an accessible observable, enabling Heisenberg-limited precision.
• It employs techniques like one-axis twisting and spin-cat state preparation to transform the detection problem from requiring fine resolution to simple population-difference measurements.
• The method shows enhanced detection-noise robustness and is applicable in various platforms including Bose–Einstein condensates, trapped ions, and superconducting circuits.

Searching arXiv for relevant papers on interaction-based readout in quantum metrology and related architectures.

Interaction-based readout is a readout strategy in which controlled interactions are applied after parameter encoding and before measurement, so that phase information is processed into an experimentally accessible observable. In quantum metrology, the term denotes protocols where interparticle interactions are used during detection to amplify phase information into a signal that is robust to detection noise and compatible with coarse-grained measurements. In the spin-cat metrology setting, the canonical example is a composite readout built from one-axis twisting, two $\pi/2$ pulses, and a population measurement, which allows non-Gaussian spin cat states to saturate their ultimate precision bounds with simple $\hat J_z$ detection rather than parity measurement or full number-resolved detection (Huang et al., 2018).

1. Definition and conceptual structure

In the standard interferometric decomposition, one prepares an input state $|\psi\rangle_{\mathrm{in}}$, encodes the unknown parameter through a unitary $\hat U(\phi)=e^{-i\phi \hat G}$, and then measures the output. Standard linear readout uses only a linear rotation before measuring a collective spin component, typically a $\pi/2$-pulse followed by measurement of $\hat J_z$. Interaction-based readout replaces that final linear stage by a controlled many-body evolution (Huang et al., 2018).

For the spin-cat protocol, the readout is

$$\hat U_{\mathrm{IBR}}(\chi t) = \hat R_x^\dagger\!\left(\frac{\pi}{2}\right)\, \hat U_{\mathrm{non}}(\chi t)\, \hat R_x\!\left(\frac{\pi}{2}\right),$$

with

$$\hat U_{\mathrm{non}}(\chi t)=e^{-i\hat H_{\mathrm{OAT}} t}=e^{i\chi t \hat J_z^2}, \qquad \hat R_x\!\left(\frac{\pi}{2}\right)=e^{i\frac{\pi}{2}\hat J_x}.$$

The final observable is the population difference $\hat J_z$. The defining feature is that nonlinear many-body dynamics are inserted into the readout itself, rather than being used only for state preparation (Huang et al., 2018).

Two distinctions are central. First, interaction-based readout differs from linear readout because it can make simple population measurements optimal for states whose quantum limit would otherwise require parity or single-particle-resolved detection. Second, it is broader than twisting echo: the readout unitary need not be the perfect time-reversal of state preparation, so $\hat U_R \neq \hat U^\dagger$ in general (Huang et al., 2018).

2. Spin cat states and the metrological setting

The spin-cat construction is formulated for a two-mode bosonic system with fixed particle number $\hat J_z$0, collective spin length $\hat J_z$1, and Dicke basis $\hat J_z$2. The collective operators are

$\hat J_z$3

A spin coherent state is

$\hat J_z$4

and the relevant macroscopic superposition is

$\hat J_z$5

With $\hat J_z$6, and for $\hat J_z$7 with $\hat J_z$8, this becomes the spin cat state

$\hat J_z$9

which satisfies $|\psi\rangle_{\mathrm{in}}$0 (Huang et al., 2018).

The metrological task is frequency estimation under

$|\psi\rangle_{\mathrm{in}}$1

For a pure state, the quantum Fisher information is $|\psi\rangle_{\mathrm{in}}$2; here $|\psi\rangle_{\mathrm{in}}$3, so

$|\psi\rangle_{\mathrm{in}}$4

For spin cat states with two well-separated peaks at $|\psi\rangle_{\mathrm{in}}$5, one has

$|\psi\rangle_{\mathrm{in}}$6

and therefore

$|\psi\rangle_{\mathrm{in}}$7

The corresponding quantum Cramér–Rao bound is

$|\psi\rangle_{\mathrm{in}}$8

This is Heisenberg-like scaling, $|\psi\rangle_{\mathrm{in}}$9, for all spin cat states satisfying the cat condition; at $\hat U(\phi)=e^{-i\phi \hat G}$0, the state is GHZ-like and $\hat U(\phi)=e^{-i\phi \hat G}$1 (Huang et al., 2018).

3. Readout sequence and saturation of the quantum limit

The full protocol is preparation, phase encoding, interaction-based readout, and population measurement: $\hat U(\phi)=e^{-i\phi \hat G}$2 The measured observable is $\hat U(\phi)=e^{-i\phi \hat G}$3, and phase sensitivity can be estimated by error propagation,

$\hat U(\phi)=e^{-i\phi \hat G}$4

For a general symmetric input state

$\hat U(\phi)=e^{-i\phi \hat G}$5

the case $\hat U(\phi)=e^{-i\phi \hat G}$6 is analytically tractable (Huang et al., 2018).

In that case, after the full sequence the conditional probability of outcome $\hat U(\phi)=e^{-i\phi \hat G}$7 in a $\hat U(\phi)=e^{-i\phi \hat G}$8 measurement is

$\hat U(\phi)=e^{-i\phi \hat G}$9

The associated classical Fisher information is

$\pi/2$0

At the special operating point $\pi/2$1,

$\pi/2$2

Since $\pi/2$3, the sum is $\pi/2$4, and therefore

$\pi/2$5

Thus, at $\pi/2$6, interaction-based readout followed by population measurement saturates the quantum Cramér–Rao bound. For spin cat states,

$\pi/2$7

so the simple $\pi/2$8 measurement becomes optimal (Huang et al., 2018).

The same conclusion appears in the error-propagation picture. At $\pi/2$9,

$\hat J_z$0

so full reconstruction of the measurement distribution is not required in order to attain the optimal precision bound. For $\hat J_z$1, the protocol still yields Heisenberg scaling $\hat J_z$2 with optimized $\hat J_z$3, although not tight to the quantum-Cramér–Rao prefactor (Huang et al., 2018).

4. Detection-noise robustness and relation to echo protocols

Detection noise is modeled by Gaussian smearing of the ideal $\hat J_z$4 distribution: $\hat J_z$5 with normalization coefficients $\hat J_z$6. The measured expectation value becomes

$\hat J_z$7

For $\hat J_z$8, the relative precision stays near unity so long as

$\hat J_z$9

which yields the critical noise scale

$$\hat U_{\mathrm{IBR}}(\chi t) = \hat R_x^\dagger\!\left(\frac{\pi}{2}\right)\, \hat U_{\mathrm{non}}(\chi t)\, \hat R_x\!\left(\frac{\pi}{2}\right),$$0

The robustness is therefore linear in $$\hat U_{\mathrm{IBR}}(\chi t) = \hat R_x^\dagger\!\left(\frac{\pi}{2}\right)\, \hat U_{\mathrm{non}}(\chi t)\, \hat R_x\!\left(\frac{\pi}{2}\right),$$1, and states with smaller $$\hat U_{\mathrm{IBR}}(\chi t) = \hat R_x^\dagger\!\left(\frac{\pi}{2}\right)\, \hat U_{\mathrm{non}}(\chi t)\, \hat R_x\!\left(\frac{\pi}{2}\right),$$2 are more robust because $$\hat U_{\mathrm{IBR}}(\chi t) = \hat R_x^\dagger\!\left(\frac{\pi}{2}\right)\, \hat U_{\mathrm{non}}(\chi t)\, \hat R_x\!\left(\frac{\pi}{2}\right),$$3 is smaller (Huang et al., 2018).

A common misconception is that optimal interaction-based readout must be exact time reversal. That is not generally the case. In the twist-and-turn setting, the optimum interaction-based readout is “not the obvious case of perfect time reversal” (Mirkhalaf et al., 2018). More generally, one can construct an optimal protocol with substantial flexibility, allowing it to remain robust to detection noise while still saturating the quantum limit (Nolan et al., 2017).

The comparison to twisting echo on spin-squeezed states is especially sharp. Twisting echo schemes start from Gaussian squeezed states and use an inverse one-axis-twisting stage as the readout. By contrast, the spin-cat protocol uses non-Gaussian inputs and a readout that does not require exact reversal of the preparation dynamics. The resulting detection-noise tolerance reaches $$\hat U_{\mathrm{IBR}}(\chi t) = \hat R_x^\dagger\!\left(\frac{\pi}{2}\right)\, \hat U_{\mathrm{non}}(\chi t)\, \hat R_x\!\left(\frac{\pi}{2}\right),$$4, whereas twisting echo on squeezed states is described as robust only up to $$\hat U_{\mathrm{IBR}}(\chi t) = \hat R_x^\dagger\!\left(\frac{\pi}{2}\right)\, \hat U_{\mathrm{non}}(\chi t)\, \hat R_x\!\left(\frac{\pi}{2}\right),$$5 (Huang et al., 2018).

5. Experimental realization and limitations

The readout sequence was designed to be compatible with current experimental techniques. The required ingredients are: preparation of spin cat states, phase encoding under $$\hat U_{\mathrm{IBR}}(\chi t) = \hat R_x^\dagger\!\left(\frac{\pi}{2}\right)\, \hat U_{\mathrm{non}}(\chi t)\, \hat R_x\!\left(\frac{\pi}{2}\right),$$6, one-axis twisting generated by

$$\hat U_{\mathrm{IBR}}(\chi t) = \hat R_x^\dagger\!\left(\frac{\pi}{2}\right)\, \hat U_{\mathrm{non}}(\chi t)\, \hat R_x\!\left(\frac{\pi}{2}\right),$$7

two $$\hat U_{\mathrm{IBR}}(\chi t) = \hat R_x^\dagger\!\left(\frac{\pi}{2}\right)\, \hat U_{\mathrm{non}}(\chi t)\, \hat R_x\!\left(\frac{\pi}{2}\right),$$8 pulses about the $$\hat U_{\mathrm{IBR}}(\chi t) = \hat R_x^\dagger\!\left(\frac{\pi}{2}\right)\, \hat U_{\mathrm{non}}(\chi t)\, \hat R_x\!\left(\frac{\pi}{2}\right),$$9-axis, and a population-difference measurement $$\hat U_{\mathrm{non}}(\chi t)=e^{-i\hat H_{\mathrm{OAT}} t}=e^{i\chi t \hat J_z^2}, \qquad \hat R_x\!\left(\frac{\pi}{2}\right)=e^{i\frac{\pi}{2}\hat J_x}.$$0. Candidate platforms include two-mode Bose–Einstein condensates and trapped-ion systems, where collective spin variables and Ising-type interactions can be realized (Huang et al., 2018).

Spin cat states can be generated via nonlinear dynamics in two-mode Bose–Einstein condensates or via adiabatic ground-state preparation by tuning atom–atom interactions. The parameter $$\hat U_{\mathrm{non}}(\chi t)=e^{-i\hat H_{\mathrm{OAT}} t}=e^{i\chi t \hat J_z^2}, \qquad \hat R_x\!\left(\frac{\pi}{2}\right)=e^{i\frac{\pi}{2}\hat J_x}.$$1 is set by interaction strength and preparation time. The one-axis-twisting Hamiltonian arises naturally in collisional Bose–Einstein condensates and has already been used to generate spin squeezing. Population measurement requires only coarse detection, provided the resolution satisfies the linear-in-$$\hat U_{\mathrm{non}}(\chi t)=e^{-i\hat H_{\mathrm{OAT}} t}=e^{i\chi t \hat J_z^2}, \qquad \hat R_x\!\left(\frac{\pi}{2}\right)=e^{i\frac{\pi}{2}\hat J_x}.$$2 noise threshold above (Huang et al., 2018).

The main decoherence model treated during nonlinear evolution is correlated dephasing,

$$\hat U_{\mathrm{non}}(\chi t)=e^{-i\hat H_{\mathrm{OAT}} t}=e^{i\chi t \hat J_z^2}, \qquad \hat R_x\!\left(\frac{\pi}{2}\right)=e^{i\frac{\pi}{2}\hat J_x}.$$3

Under this model, spin cat states remain sub-shot-noise even for sizable $$\hat U_{\mathrm{non}}(\chi t)=e^{-i\hat H_{\mathrm{OAT}} t}=e^{i\chi t \hat J_z^2}, \qquad \hat R_x\!\left(\frac{\pi}{2}\right)=e^{i\frac{\pi}{2}\hat J_x}.$$4, especially for moderate $$\hat U_{\mathrm{non}}(\chi t)=e^{-i\hat H_{\mathrm{OAT}} t}=e^{i\chi t \hat J_z^2}, \qquad \hat R_x\!\left(\frac{\pi}{2}\right)=e^{i\frac{\pi}{2}\hat J_x}.$$5, where the optimal evolution times are shorter. Strong dephasing still degrades performance, and particle loss and other decoherence channels were not treated in detail. The analytical saturation results also rely on the symmetry condition $$\hat U_{\mathrm{non}}(\chi t)=e^{-i\hat H_{\mathrm{OAT}} t}=e^{i\chi t \hat J_z^2}, \qquad \hat R_x\!\left(\frac{\pi}{2}\right)=e^{i\frac{\pi}{2}\hat J_x}.$$6 for the input coefficients (Huang et al., 2018).

6. Broader usage of the term

Outside spin-cat metrology, the same label is used for indirect readout architectures in which the measured subsystem is inferred through an engineered interaction rather than through direct projective access. Representative examples appear across cavity QED, topological qubits, and nanoelectromechanical devices.

Platform	Engineered interaction	Readout channel
Superconducting qubit–oscillator	Modulated longitudinal qubit–oscillator interaction	Qubit-state-dependent cavity drive (Didier et al., 2015)
Majorana box qubit	Parametric modulation of a longitudinal qubit–resonator interaction	Resonator displacement and homodyne signal (Grimsmo et al., 2018)
Carbon-nanotube double quantum dot	Curvature-induced spin-phonon coupling	Pauli spin blockade leakage current (Ohm et al., 2011)
NV ensemble in a microwave cavity	Off-resonant dispersive spin–cavity interaction	Cavity phase or quadrature shift (Wang et al., 24 May 2026)

In these settings, the measured degree of freedom is not observed directly. Instead, an interaction converts the target information into a more accessible observable: a cavity field, a leakage current, or a collective phase response. This suggests that interaction-based readout is best understood as a family of indirect measurement protocols in which a deliberately engineered interaction maps fragile or inaccessible information onto a robust readout channel. In quantum metrology, the spin-cat protocol is a particularly explicit realization of that idea because it shows, analytically, that nonlinear readout can make a simple population measurement both optimal and robust (Huang et al., 2018).