Distributed Collective Squeezed Mode

• Distributed collective squeezed mode is a delocalized observable whose variance is suppressed below the vacuum threshold, enabling nonlocal quantum correlations.
• It spans diverse systems—optical, spin, and phononic—using weighted summations and covariance analysis to enhance multimode sensing precision.
• Metrological gains depend on optimal generator alignment and efficient mode extraction techniques, validated through precise noise reduction benchmarks.

A distributed collective squeezed mode is a squeezed degree of freedom delocalized over multiple physical modes or sensor nodes. In continuous-variable settings it is typically a collective bosonic mode $\hat b=\sum_{i=1}^N w_i \hat a_i$ or a collective quadrature $\hat O_{\mathbf g}=\mathbf g^T \hat R$ whose variance is reduced below vacuum or an appropriate separability benchmark. In spin networks it is a weighted collective spin observable such as $M=\sum_k w_k J_{z,k}$ or a transverse component of the total spin with variance below the coherent-spin-state or quantum-projection-noise reference. Across the cited literature, the concept appears in distributed sensing, multimode Gaussian optics, trapped-ion motion, optical-lattice bosons, atomic-vapor arrays, and even high-harmonic manifolds. Its metrological relevance depends not only on reduced noise, but on alignment between the squeezed collective mode and the generator of the parameter to be estimated, as well as on a correct resource-normalized benchmark (Gessner et al., 2017, Malia et al., 2022, McGuinness, 2023, Zhou et al., 24 Apr 2026).

1. Definition and formal scope

In multimode continuous-variable theory, the basic object is a collective observable built from the phase-space vector $$\hat R=(\hat x_1,\hat p_1,\ldots,\hat x_N,\hat p_N)^T$$. For a real weight vector $\mathbf g$, the collective quadrature is $\hat O_{\mathbf g}=\mathbf g^T \hat R$, with variance $$\mathrm{Var}(\hat O_{\mathbf g})=\mathbf g^T V \mathbf g$$. A distributed collective squeezed mode is obtained when the optimal squeezed direction has support on several modes, so that the noise reduction is intrinsically nonlocal rather than confined to a single oscillator. In the bosonic mode language, this same structure is written as $\hat b=\sum_i w_i \hat a_i$, with collective quadratures $$X_b(\theta)=(\hat b^\dagger e^{i\theta}+\hat b e^{-i\theta})/\sqrt{2}$$ and squeezing signaled by sub-vacuum variance in some quadrature (Gessner et al., 2017, Zhou et al., 24 Apr 2026).

In distributed spin systems, the corresponding object is a weighted sum of node-local collective spins. For $M$ sensor nodes with local collective spin operators $\hat O_{\mathbf g}=\mathbf g^T \hat R$0, the distributed collective spin is $\hat O_{\mathbf g}=\mathbf g^T \hat R$1, and a target squeezed mode can be defined as $\hat O_{\mathbf g}=\mathbf g^T \hat R$2. In the atomic-network literature this mode is prepared by a shared QND measurement that reduces $\hat O_{\mathbf g}=\mathbf g^T \hat R$3 below the quantum projection noise limit of an unentangled coherent spin state, creating nonlocal anticorrelations between nodes (Malia et al., 2022).

The distinction between local and distributed squeezing is central. Local squeezing reduces noise in one mode or at one node. Distributed collective squeezing instead suppresses the variance of a nonlocal linear combination. In the continuous-variable entanglement framework, this difference is operationalized by replacing one variance in a Heisenberg-type product bound with the correlation-free covariance $\hat O_{\mathbf g}=\mathbf g^T \hat R$4; the resulting multimode squeezing coefficient

$\hat O_{\mathbf g}=\mathbf g^T \hat R$5

satisfies $\hat O_{\mathbf g}=\mathbf g^T \hat R$6 for separable states, so $\hat O_{\mathbf g}=\mathbf g^T \hat R$7 certifies entanglement and identifies an optimally squeezed distributed collective mode (Gessner et al., 2017).

2. Covariance structure, mode extraction, and fundamental bounds

Because most implementations are Gaussian or approximately Gaussian, the covariance matrix is the natural description. In optical and electromechanical networks, Bloch–Messiah or Schmidt decompositions rotate the physical basis into supermodes, each with its own squeezing parameter. A distributed collective squeezed mode is then one of these supermodes, often the dominant one, while orthogonal combinations remain vacuum-like or less squeezed. This perspective underlies both spectral supermodes in SPDC sources and collective normal modes in reservoir-engineered networks (Roman-Rodriguez et al., 2023, Kouadou et al., 2022, Karnieli et al., 20 Sep 2025).

The 2026 analysis of quantum limits on squeezing recasts the problem in terms of commutator and noise budgets. For passive, reservoir-engineered two-mode networks satisfying physical realizability, stability, diagonal dissipation, no hidden loss channels, and uncorrelated inputs, the mode-optimal variances obey

$\hat O_{\mathbf g}=\mathbf g^T \hat R$8

With additional local parametric drives, the optimum bound becomes

$\hat O_{\mathbf g}=\mathbf g^T \hat R$9

which approaches $M=\sum_k w_k J_{z,k}$0 as $M=\sum_k w_k J_{z,k}$1. These results formalize a trade-off already implicit in multimode squeezing experiments: squeezing concentrated into one collective mode necessarily constrains what remains available in orthogonal modes (Zhou et al., 24 Apr 2026).

Mode extraction has itself become a control problem. A variational scheme based on self-configuring photonic networks learns the dominant supermodes sequentially by using homodyne variance as a cost function. For $M=\sum_k w_k J_{z,k}$2 modes, the first $M=\sum_k w_k J_{z,k}$3 most significant supermodes can be discovered with $M=\sum_k w_k J_{z,k}$4 physical elements and optimization steps, rather than full quadratic overhead. In the frequency domain, inverse-designed surrogate networks reduce full-network implementations to $M=\sum_k w_k J_{z,k}$5 and even $M=\sum_k w_k J_{z,k}$6 modulated cavities, depending on the frequency-bin architecture (Karnieli et al., 20 Sep 2025).

3. Distributed quantum sensing and generator alignment

The metrological use of a distributed collective squeezed mode is controlled by the generator of parameter encoding. If $M=\sum_k w_k J_{z,k}$7, then the relevant single-shot uncertainty is

$M=\sum_k w_k J_{z,k}$8

not the noise of an arbitrary readout variable. In distributed sensing, the collective mode must therefore coincide with the encoded observable $M=\sum_k w_k J_{z,k}$9 or with the corresponding weighted quadrature of an optical network. This generator–readout alignment is explicit in both atomic-network and continuous-variable proposals (McGuinness, 2023, Kannath et al., 2 Aug 2025).

In mode-entangled spin-squeezed atomic networks, a shared QND measurement couples simultaneously to several atomic ensembles and conditionally squeezes a weighted collective spin mode. The 2022 experiment reported a clock network with up to four nodes, claiming up to $$\hat R=(\hat x_1,\hat p_1,\ldots,\hat x_N,\hat p_N)^T$$0 dB better precision than a mode-separable network and $$\hat R=(\hat x_1,\hat p_1,\ldots,\hat x_N,\hat p_N)^T$$1 dB improvement relative to a CSS quantum projection noise limit, and also demonstrated an atom-interferometric differential protocol (Malia et al., 2022). A different optical realization uses bright two-mode squeezed states distributed by symmetric beam-splitter networks so that the symmetric normal mode carries the squeezing aligned to the average Sagnac phase of a gyroscope network. In that architecture, the collective estimator $$\hat R=(\hat x_1,\hat p_1,\ldots,\hat x_N,\hat p_N)^T$$2 approaches the QCRB in the bright-seed limit, and for $$\hat R=(\hat x_1,\hat p_1,\ldots,\hat x_N,\hat p_N)^T$$3 loss in every channel with $$\hat R=(\hat x_1,\hat p_1,\ldots,\hat x_N,\hat p_N)^T$$4 dB initial squeezing the reported sensitivity gain is $$\hat R=(\hat x_1,\hat p_1,\ldots,\hat x_N,\hat p_N)^T$$5 dB beyond the shot-noise limit (Kannath et al., 2 Aug 2025).

Trapped-ion work shows the same logic in a discrete motional setting. Reservoir engineering can stabilize a two-mode squeezed steady state whose commuting collective quadratures $$\hat R=(\hat x_1,\hat p_1,\ldots,\hat x_N,\hat p_N)^T$$6 and $$\hat R=(\hat x_1,\hat p_1,\ldots,\hat x_N,\hat p_N)^T$$7 are simultaneously squeezed, allowing joint estimation of two collective displacements with improvements of $$\hat R=(\hat x_1,\hat p_1,\ldots,\hat x_N,\hat p_N)^T$$8 and $$\hat R=(\hat x_1,\hat p_1,\ldots,\hat x_N,\hat p_N)^T$$9 dB beyond the classical limit. Related laser-free proposals use Coulomb-mediated nonlinear coupling and time-varying electric fields to create a phonon two-mode squeezed state or a beam-splitter transformation between collective vibrational modes, with Heisenberg scaling for appropriate beam-splitter metrology (Li et al., 2023, Nikolova et al., 2023).

4. Optical supermodes in spectral, spatial, and harmonic manifolds

In multimode photonics, the distributed collective squeezed mode is often literally a supermode. A telecom-wavelength single-pass SPDC source in a ppKTP waveguide generated significant squeezing in more than $\mathbf g$0 frequency supermodes, with maximum measured squeezing exceeding $\mathbf g$1 dB. In an 8-frexel covariance-matrix reconstruction, $\mathbf g$2 of bipartitions, namely $\mathbf g$3, violated PPT, and programmable LO shaping allowed reconfiguration into few-node cluster states whose nullifiers were measured below shot noise (Roman-Rodriguez et al., 2023). A closely related platform combining frequency and time multiplexing demonstrated $\mathbf g$4 squeezed spectral modes at $\mathbf g$5 MHz, with best single-mode squeezing of $\mathbf g$6 dB for HG$\mathbf g$7, direct pulse-by-pulse readout, and inseparability in $\mathbf g$8 of $\mathbf g$9 bipartitions of an 8-frexel covariance matrix (Kouadou et al., 2022).

Spatial and spatiotemporal implementations make the same point in a different basis. A non-collinear single-pass BBO source probed with a spatiotemporally shaped LO showed squeezing of $\hat O_{\mathbf g}=\mathbf g^T \hat R$0, $\hat O_{\mathbf g}=\mathbf g^T \hat R$1, $\hat O_{\mathbf g}=\mathbf g^T \hat R$2, and $\hat O_{\mathbf g}=\mathbf g^T \hat R$3 dB in HG$\hat O_{\mathbf g}=\mathbf g^T \hat R$4–HG$\hat O_{\mathbf g}=\mathbf g^T \hat R$5, and covariance analysis revealed multiple temporal eigenmodes together with at least two principal spatial modes per temporal mode (Volpe et al., 2020). Cascaded phase-only SLMs have also been used to map a squeezed HG$\hat O_{\mathbf g}=\mathbf g^T \hat R$6 state into higher-order or arbitrary complex transverse modes; the reported output squeezing includes $\hat O_{\mathbf g}=\mathbf g^T \hat R$7 dB for HG$\hat O_{\mathbf g}=\mathbf g^T \hat R$8 and $\hat O_{\mathbf g}=\mathbf g^T \hat R$9 dB for an arbitrary “QMC” pattern, showing that a single squeezed input can be redistributed across an engineered complex amplitude profile (Ma et al., 2020).

Warm-vapor polarization self-rotation provides a complementary spatial example. There the squeezed vacuum is not a single transverse mode but a coherent superposition of higher-order Laguerre–Gauss modes with different Gouy phases, so the local oscillator defines the collective detection mode. Experimentally, $$\mathrm{Var}(\hat O_{\mathbf g})=\mathbf g^T V \mathbf g$$0 dB baseline squeezing was observed, but masks and focus position strongly altered the measured noise, indicating a spatial multimode structure; the accompanying model suggested that the dominant $$\mathrm{Var}(\hat O_{\mathbf g})=\mathbf g^T V \mathbf g$$1 mode alone would exhibit more than $$\mathrm{Var}(\hat O_{\mathbf g})=\mathbf g^T V \mathbf g$$2 dB if isolated (Zhang et al., 2015). At the opposite spectral extreme, bichromatically driven high-harmonic generation can produce an even-harmonic manifold that is effectively a single rank-1 Gaussian collective mode, with weights set by the HHG response coefficients $$\mathrm{Var}(\hat O_{\mathbf g})=\mathbf g^T V \mathbf g$$3 and effective squeezing parameter $$\mathrm{Var}(\hat O_{\mathbf g})=\mathbf g^T V \mathbf g$$4 in the linear-response regime (Rivera-Dean et al., 21 Jun 2026).

5. Engineered distribution, synchronization, and protection

Several recent works treat distributed collective squeezing as something to be designed rather than merely observed. In a paraffin-coated $$\mathrm{Var}(\hat O_{\mathbf g})=\mathbf g^T V \mathbf g$$5Rb vapor cell, a common, motion-homogenized ground-state coherence acted as a global dissipative reservoir for a 30-beam array of polarization-squeezed channels. The experiment realized $$\mathrm{Var}(\hat O_{\mathbf g})=\mathbf g^T V \mathbf g$$6 dB of squeezing in a 30-beam array, verified synchronization behavior, and found improved purity and faster recovery from perturbations as the array size increased, indicating that the symmetric collective optical mode is being jointly regulated by the moving atoms (Wang et al., 27 May 2026).

In quantum error correction, the distributed two-mode-squeezing GKP construction uses one active two-mode squeezer and an array of beamsplitters to distribute continuous-variable correlations over many ancillae. The collective ancilla mode $$\mathrm{Var}(\hat O_{\mathbf g})=\mathbf g^T V \mathbf g$$7 becomes the partner of the data oscillator in an EPR-like pair, and the resulting analog noise suppression scales as

$$\mathrm{Var}(\hat O_{\mathbf g})=\mathbf g^T V \mathbf g$$8

The same construction yields qubit-code distances comparable to previous brute-force multimode searches while requiring only one active squeezing element (Brady et al., 2024).

Circuit-QED proposals pursue similar goals with qubit-mediated Gaussian control. A Rabi-driven qubit dispersively coupled to one or two oscillators can, under modulated Jaynes–Cummings interactions, generate conditional squeezing Hamiltonians for single-mode and two-mode collective quadratures. Simulations reported $$\mathrm{Var}(\hat O_{\mathbf g})=\mathbf g^T V \mathbf g$$9 dB intra-cavity single-mode squeezing, $\hat b=\sum_i w_i \hat a_i$0 dB “superimposed single-mode” squeezing in the collective supermode basis, and $\hat b=\sum_i w_i \hat a_i$1 dB two-mode squeezing (Blumenthal et al., 30 Jul 2025). In optical-lattice bosons, Monte-Carlo optimization of Bose–Hubbard control sequences does not target a named supermode directly, but it engineers multimode states with $\hat b=\sum_i w_i \hat a_i$2, which the paper identifies as an intermediate scaling between SQL and HL and attributes to a finite $\hat b=\sum_i w_i \hat a_i$3-measure subset of Hilbert space containing metrologically useful distributed correlations (Shao et al., 19 Nov 2025).

6. Benchmarking, controversy, and network deployment

The central benchmarking issue is that squeezing of a collective observable is not, by itself, a proof of metrological gain. The benchmark must match the estimator, operating point, contrast, and resource accounting. This point is made explicitly in the 2023 critique of the four-node atomic-clock experiment. That critique argues that the reported “$\hat b=\sum_i w_i \hat a_i$4” analysis examined fluctuations of a difference variable rather than a phase-estimator uncertainty, used an incorrect QPN reference, compared different operating points, and did not establish entanglement from the projective QND readout alone. Using the only phase-calibrated dataset, it extracted an experimental phase uncertainty of about $\hat b=\sum_i w_i \hat a_i$5 rad and compared it with a proper QPN benchmark of $\hat b=\sum_i w_i \hat a_i$6 when total duration is included, or $\hat b=\sum_i w_i \hat a_i$7 mrad when only pulse durations are counted, concluding that the demonstration was more than two to three orders of magnitude worse than the relevant QPN limit rather than $\hat b=\sum_i w_i \hat a_i$8–$\hat b=\sum_i w_i \hat a_i$9 dB better (McGuinness, 2023). The original experiment, by contrast, reported $$X_b(\theta)=(\hat b^\dagger e^{i\theta}+\hat b e^{-i\theta})/\sqrt{2}$$0 dB improvement relative to CSS QPN and $$X_b(\theta)=(\hat b^\dagger e^{i\theta}+\hat b e^{-i\theta})/\sqrt{2}$$1 dB beyond the mode-separable limit (Malia et al., 2022).

A separate line of work validates distributed collective squeezing under realistic networking conditions rather than disputing estimator definitions. Two-mode squeezing distributed through separate fibers, while coexisting with classical telecom traffic and a multiplexed LO, yielded $$X_b(\theta)=(\hat b^\dagger e^{i\theta}+\hat b e^{-i\theta})/\sqrt{2}$$2 dB coexistent two-mode squeezing after two separate 5-km spools and $$X_b(\theta)=(\hat b^\dagger e^{i\theta}+\hat b e^{-i\theta})/\sqrt{2}$$3 dB after deployed campus fibers of about $$X_b(\theta)=(\hat b^\dagger e^{i\theta}+\hat b e^{-i\theta})/\sqrt{2}$$4 m and $$X_b(\theta)=(\hat b^\dagger e^{i\theta}+\hat b e^{-i\theta})/\sqrt{2}$$5 km. The same study reported that no measurable excess noise from the classical channels was observed beyond insertion loss, and that 10 GbE throughput was unaffected (Chapman et al., 2023).

Taken together, these results sharpen the meaning of the term. A distributed collective squeezed mode is not merely any nonlocal observable with reduced variance. It is a delocalized mode whose squeezing survives the relevant loss channels, whose entanglement or nonclassicality can be certified in the correct basis, and whose utility for sensing or information processing depends on whether the distributed mode is the one actually interrogated by the protocol. Under that criterion, the concept spans supermode optics, spin networks, phononic normal modes, synchronized optical arrays, and bosonic codes, but its status as a metrological resource remains inseparable from benchmarking discipline.