Collective Spin Cat States

• Collective spin cat states are macroscopic superpositions of spin coherent states in symmetric many-body systems, forming highly entangled quantum resources.
• They are generated through deterministic nonlinear evolutions, ancilla-mediated exchanges, or measurement-induced protocols to reach Heisenberg-limited metrology.
• Practical implementations require precise control of interaction strengths and mitigation of decoherence to preserve nonclassical coherence and enhance metrological performance.

Collective spin cat states are macroscopic superpositions in the collective-spin degrees of freedom of many-body systems, typically formulated in the totally symmetric spin-$J=N/2$ manifold of $N$ spin-$1/2$ particles or, equivalently, in two-mode bosonic realizations of $su(2)$. A standard form is the equal superposition of two spin coherent states (SCSs), while broader formulations include even/odd parity superpositions and “generalized cat states” identified by coherence between eigen-subspaces of an additive observable whose eigenvalues differ by $O(N)$ (Huang et al., 2014, Tatsuta et al., 2024). Across the literature, these states appear as non-Gaussian entangled resources for Heisenberg-limited metrology, as targets of nonlinear collective dynamics such as one-axis twisting and Jaynes–Cummings-type evolution, and as logical codewords or syndrome sectors in collective-spin quantum error correction (Opatrný et al., 2012, Groiseau et al., 2021, Franke et al., 16 Mar 2026).

1. Definitions, conventions, and canonical forms

The basic kinematic setting is a collective spin built from two bosonic modes $a,b$ or from $N$ physical spins. The collective operators are written, depending on convention, as

$$\hat J_x=\tfrac12(a^\dagger b+b^\dagger a),\quad \hat J_y=\tfrac{1}{2i}(a^\dagger b-b^\dagger a),$$

with $\hat J_z$ appearing either as $\tfrac12(a^\dagger a-b^\dagger b)$ or $N$0; the sign difference reflects mode labeling rather than a physical distinction (Opatrný et al., 2012, Huang et al., 2014). In the spin language, $N$1 or $N$2 specifies the collective representation.

A spin coherent state pointing in Bloch-sphere direction $N$3 is commonly written as

$N$4

or, equivalently, in the Dicke basis $N$5,

$N$6

A standard two-component spin cat is then

$N$7

with

$N$8

so that $N$9 yields the GHZ/NOON state

$1/2$0

whereas $1/2$1 makes the two SCSs coincide, reducing the state to an unentangled coherent spin state (Huang et al., 2014).

A second widely used parametrization emphasizes parity. In that form,

$1/2$2

In the Holstein–Primakoff regime one introduces a bosonic amplitude $1/2$3, so small $1/2$4 corresponds to $1/2$5, and the even/odd spin cats reduce to the familiar bosonic even/odd cat structure (Mok et al., 2024).

A broader notion appears in the generalized-cat literature. For an arbitrary $1/2$6-spin state $1/2$7, an additive observable $1/2$8, and the trace norm, the “coherence index” $1/2$9 is defined by

$su(2)$0

States with $su(2)$1 are called generalized cat states because they carry coherence between macroscopically distinct sectors of $su(2)$2 (Tatsuta et al., 2024). This formulation makes explicit that a collective spin cat need not be restricted to a pure superposition of exactly two CSSs.

2. Nonlinear Hamiltonians and deterministic generation

A central route to collective spin cat states is deterministic nonlinear evolution. In Rydberg-blockaded atomic ensembles, one introduces two ground states $su(2)$3, two Rydberg levels $su(2)$4, and the “perfect blockade” condition permitting at most one Rydberg excitation. The resulting time-dependent Hamiltonian combines an effective Jaynes–Cummings coupling with a transverse Raman term,

$su(2)$5

with adiabatic STIRAP-like sequencing used to transform the initial $su(2)$6 coherent state into an extremal eigenstate of the pure JC Hamiltonian. The same framework also supports a direct half-pulse cat protocol: using only $su(2)$7 for

$su(2)$8

the equatorial spin-coherent superposition evolves approximately into

$su(2)$9

The paper identifies an optimal regime $O(N)$0, $O(N)$1, $O(N)$2, and $O(N)$3 up to a few $O(N)$4 (Opatrný et al., 2012).

A second major mechanism is one-axis twisting generated from an engineered Dicke model. In the dispersive limit of a generalized Dicke Hamiltonian, adiabatic elimination of the bosonic mode gives an effective spin-only master equation with

$O(N)$5

Starting from $O(N)$6, ideal no-decay dynamics produce a cat at

$O(N)$7

namely a superposition of the collective-spin extremes. The same work also emphasizes the conditional character of the preparation: when there are no quantum jumps, long-time evolution approaches $O(N)$8, while an early jump probabilistically projects the system into one of finitely many entangled-state cycles distinguished by jump frequency $O(N)$9 (Groiseau et al., 2021).

Ancilla-mediated exchange couplings provide a third deterministic route. For two spin ensembles $a,b$0 and $a,b$1, the resonant interaction

$a,b$2

maps an initial SCS of $a,b$3 and a GHZ-like ancillary state of $a,b$4 into a cat of $a,b$5. For $a,b$6 the quarter-revival time is $a,b$7; for general $a,b$8, preparing $a,b$9 in $N$0 yields

$N$1

so the effective coupling is $N$2 and cat formation is $N$3 times faster. In the one-axis-twisting regime this same setting also generates equally weighted $N$4-component superpositions at rational fractions of the revival time (Dooley et al., 2014).

Rydberg–Kerr protocols extend the deterministic landscape beyond two-component cats. For a pure Kerr Hamiltonian $N$5, an equatorial CSS evolves at

$N$6

into an $N$7-component equatorial superposition $N$8. The same analysis gives $N$9. In the resonant regime, the collective Hamiltonian becomes $$\hat J_x=\tfrac12(a^\dagger b+b^\dagger a),\quad \hat J_y=\tfrac{1}{2i}(a^\dagger b-b^\dagger a),$$0, the effective quadratic nonlinearity scales as $$\hat J_x=\tfrac12(a^\dagger b+b^\dagger a),\quad \hat J_y=\tfrac{1}{2i}(a^\dagger b-b^\dagger a),$$1, and the absence of the adiabaticity constraint permits fast switching of the Rydberg drive (Khazali, 2023).

3. Heralded and measurement-induced generation

A distinct class of protocols uses conditioning rather than purely deterministic unitary evolution. In the single-photon “painting” scheme, a cavity photon imparts a collective rotation through the dispersive interaction

$$\hat J_x=\tfrac12(a^\dagger b+b^\dagger a),\quad \hat J_y=\tfrac{1}{2i}(a^\dagger b-b^\dagger a),$$2

If the incident wavepacket is a superposition of two short pulses separated by $$\hat J_x=\tfrac12(a^\dagger b+b^\dagger a),\quad \hat J_y=\tfrac{1}{2i}(a^\dagger b-b^\dagger a),$$3, then a photodetection event at late time heralds the atomic state

$$\hat J_x=\tfrac12(a^\dagger b+b^\dagger a),\quad \hat J_y=\tfrac{1}{2i}(a^\dagger b-b^\dagger a),$$4

Because success is heralded, the protocol can be implemented with a weak coherent field rather than a deterministic single-photon source, and the same formalism extends to arbitrary Dicke-basis superpositions by suitable multi-tone envelope design (Davis et al., 2018).

Repeated ancilla readout can also create metrologically useful cats from a thermal initial state without direct control over the spin ensemble. In that approach, an ancillary qubit couples collectively through

$$\hat J_x=\tfrac12(a^\dagger b+b^\dagger a),\quad \hat J_y=\tfrac{1}{2i}(a^\dagger b-b^\dagger a),$$5

and a Ramsey-type measurement cycle yields Kraus operators

$$\hat J_x=\tfrac12(a^\dagger b+b^\dagger a),\quad \hat J_y=\tfrac{1}{2i}(a^\dagger b-b^\dagger a),$$6

After $$\hat J_x=\tfrac12(a^\dagger b+b^\dagger a),\quad \hat J_y=\tfrac{1}{2i}(a^\dagger b-b^\dagger a),$$7 readouts with $$\hat J_x=\tfrac12(a^\dagger b+b^\dagger a),\quad \hat J_y=\tfrac{1}{2i}(a^\dagger b-b^\dagger a),$$8 “+” outcomes, the dominant asymptotic state is

$$\hat J_x=\tfrac12(a^\dagger b+b^\dagger a),\quad \hat J_y=\tfrac{1}{2i}(a^\dagger b-b^\dagger a),$$9

Typical trajectories have $\hat J_z$0, so the ensemble is driven toward $\hat J_z$1. The paper shows that $\hat J_z$2 has $\hat J_z$3 and hence $\hat J_z$4, establishing the output as a generalized cat state (Tatsuta et al., 2024).

Near the Dicke critical point, photon-number measurement on the ground state provides another heralded mechanism. For the Dicke Hamiltonian

$\hat J_z$5

the normal-phase ground state near $\hat J_z$6 is strongly light–matter entangled and anti-squeezed in the photon quadrature. A photon-number–resolving measurement with outcome $\hat J_z$7 projects the atomic sector to

$\hat J_z$8

and the resulting spin state has fidelity $\hat J_z$9 with a cat-like superposition of two SCSs pointing at $\tfrac12(a^\dagger a-b^\dagger b)$0 in the $\tfrac12(a^\dagger a-b^\dagger b)$1–$\tfrac12(a^\dagger a-b^\dagger b)$2 plane. The cat-size proxy

$\tfrac12(a^\dagger a-b^\dagger b)$3

grows monotonically with $\tfrac12(a^\dagger a-b^\dagger b)$4 and as $\tfrac12(a^\dagger a-b^\dagger b)$5, while the success probability decreases toward the thermodynamic limit as $\tfrac12(a^\dagger a-b^\dagger b)$6 for fixed $\tfrac12(a^\dagger a-b^\dagger b)$7 and fixed $\tfrac12(a^\dagger a-b^\dagger b)$8 (Nakamoto et al., 15 May 2026).

4. Diagnostics, witnesses, and metrological quantifiers

The most direct figure of merit is fidelity to an ideal target cat. In the Rydberg-blockade setting this is

$\tfrac12(a^\dagger a-b^\dagger b)$9

and analogous expressions appear for SCS-cat targets, multicomponent cats, and magnon-cat analogues (Opatrný et al., 2012, Dooley et al., 2014, Sun et al., 2021). Fidelity is operationally useful but does not by itself capture the macroscopic character of the superposition.

Several nonclassicality witnesses are standard. The Wineland spin-squeezing parameter is

$N$00

but for an ideal cat $N$01, so the cited work explicitly notes that one must instead use a generalized criterion of “macroscopic superposition size” (Opatrný et al., 2012). Spin or bosonic Wigner functions are correspondingly central. In the spherical-harmonics expansion,

$N$02

and a negative dip is a witness of nonclassicality. In the Holstein–Primakoff approximation one also records $N$03 and parity oscillations $N$04 (Opatrný et al., 2012, Dooley et al., 2014).

Quantum Fisher information provides both a metrological benchmark and, in one line of work, a decoherence predictor. Under one-body loss, the quantum Cramér–Rao bound is

$N$05

with lossless GHZ input giving $N$06 and separable SCS input giving $N$07. Moderate-$N$08 spin cats retain sub-SQL performance under particle loss and interpolate between these limits (Huang et al., 2014). In atom–light state preparation, the same resource becomes a fragility measure: for an auxiliary light field with QFI $N$09, the reduced atomic QFI of a maximal cat obeys

$N$10

showing that the decoherence rate scales with the product $N$11 (Nolan et al., 2016).

Restricted-subspace Bell diagnostics reveal a more specialized structural feature. For spin-$N$12 Bell cat states, a universal Bell-type inequality $N$13 can be formulated when measurement outcomes are restricted to the spin-coherent-state subspace. Its violation occurs only for half-integer $N$14, not integer $N$15, and is attributed to the non-trivial Berry phase between south- and north-pole gauges (Gu et al., 2019). This does not state that all collective spin cats display Bell nonlocality in the full Hilbert space; the same paper explicitly states that violation disappears there except for spin-$N$16.

5. Decoherence, dissipation, and stability

Collective spin cats are fragile to dissipation, but the dominant mechanisms depend strongly on platform. In Rydberg-blockaded ensembles, the cited error channels are Rydberg spontaneous decay with rate $N$17, laser phase noise or differential dephasing between $N$18 and $N$19, and blockade imperfections allowing double Rydberg excitation with probability $N$20. The leading fidelity estimates are

$N$21

so one optimizes $N$22, $N$23, and pulse duration jointly (Opatrný et al., 2012).

In the engineered Dicke model, the effective collective decay rate is

$N$24

The fidelity at the cat time is suppressed by factors $N$25, and numerical studies reported in the paper require

$N$26

to maintain $N$27 (Groiseau et al., 2021).

A particularly detailed stability analysis concerns inhomogeneous dephasing. For free evolution under $N$28 with $N$29, the averaged short-time fidelities of even and odd cats satisfy

$N$30

In the small-amplitude limit,

$N$31

so the odd cat dephases on the bare timescale $N$32, whereas the even cat is $N$33 slower. For $N$34, parity sensitivity disappears and both parities behave like a spin coherent state. Reintroducing the two-body dissipative stabilization $N$35, the mean-field analysis yields a synchronization threshold $N$36, beyond which full dephasing occurs (Mok et al., 2024).

The preparation field can itself be the source of decoherence. For dispersive atom–light coupling with a coherent auxiliary field of mean occupation $N$37, keeping light-induced decoherence negligible for a fixed rotation $N$38 requires

$N$39

equivalently $N$40 for a maximal cat. This formalizes a recurring constraint: the more metrologically powerful the target cat, the more classical the auxiliary control field must be if one wants to suppress back-action-induced entanglement with the controller (Nolan et al., 2016).

6. Metrology, fault-tolerant encoding, and related collective-spin platforms

In quantum metrology, spin cats interpolate between the ideal Heisenberg-scaling but loss-fragile GHZ limit and the robust but SQL-limited coherent-state limit. Under one-body loss, moderate-$N$41 cats remain robust and can still beat the SQL for realistic $N$42. After phase accumulation and a final $N$43 pulse about $N$44, parity measurement in mode $N$45 stays close to the ultimate quantum Cramér–Rao bound and outperforms simple population-difference readout over a broad dissipative regime (Huang et al., 2014). This is one of the clearest demonstrations that non-Gaussianity, rather than maximal entanglement alone, is the relevant resource in realistic lossy interferometry.

Collective spin cats also serve as codewords. Spin-$N$46-Cat codes encode a logical qubit in superpositions of equatorial coherent states separated by azimuthal angle $N$47. In the totally symmetric $N$48 sector,

$N$49

The code exploits a modular decomposition of the symmetric Hilbert space into $N$50 sectors labeled by $N$51, corrects collective and individual dephasing, excitation, and decay errors, and admits a universal, fault-tolerant, and bias-preserving gate set implemented with first-order interactions in central-spin systems. For $N$52, $N$53–10, and $N$54–1000, the reported improvement factor over a bare Dicke qubit is of order $N$55 (Franke et al., 16 Mar 2026).

Related platforms broaden the concept beyond atomic ensembles. Magnon cat states represent macroscopic quantum superpositions of collective magnetic excitations and can be prepared remotely through pulsed optomagnonic entanglement followed by photon subtraction or subtraction–addition and homodyne postselection. For $N$56, $N$57, $N$58, and $N$59, the odd cat has $N$60, $N$61, $N$62, $N$63, while the even cat has $N$64, $N$65, $N$66, $N$67; the Wigner-negativity lifetimes are $N$68 and $N$69, respectively (Sun et al., 2021). In a separate hybrid architecture, a flux-tunable transmon directly coupled to a YIG Kittel mode realizes a radiation-pressure interaction that heralds even or odd magnon cats after qubit rotation and projective measurement, with typical cat-formation time $N$70–$N$71 (Kounalakis et al., 2022). A further extension uses modular-variable projection to map cat-state-like magnon entanglement to effective spin states and detect it through CHSH violation (Zhang et al., 3 Sep 2025).

Taken together, these results establish collective spin cat states as a family rather than a single construction: they include two-component and multicomponent CSS superpositions, parity-resolved cats, generalized $N$72 macroscopic superpositions, dissipation-conditioned cats, and modular collective-spin codewords. This suggests that the most useful unifying criterion is not a unique wavefunction ansatz but the coexistence of collective encoding, macroscopically distinct branches, and measurable nonclassical coherence across those branches.