Bright Two-Mode Squeezed States (bTMSS)

• Bright two-mode squeezed states (bTMSS) are Gaussian entangled states that combine coherent displacement with two-mode squeezing to yield high brightness and nonclassical intermode correlations.
• They are generated using techniques such as optical parametric amplification, four-wave mixing, and hybrid optomechanical systems, ensuring the preservation of entanglement under large mean photon numbers.
• Practical implementations highlight bTMSS applications in quantum metrology, atomic quantum memory interfaces, and distributed sensing, demonstrating enhanced sensitivity and compatibility with atomic platforms.

Bright two-mode squeezed states (bTMSS) are Gaussian entangled states in which two-mode squeezing is combined with coherent displacement or seeding, so that Einstein–Podolsky–Rosen-type intermode correlations coexist with high optical brightness. In the literature, bTMSS appear as seeded outputs of optical parametric amplifiers, bright twin beams generated by non-degenerate four-wave mixing in hot alkali vapors, pulsed type-II parametric downconversion states in waveguides, and effective two-mode squeezed states in hybrid optomechanical systems. Their defining practical feature is that large mean fields do not remove the nonclassical correlations responsible for reduced intensity-difference noise, squeezed joint quadratures, and continuous-variable entanglement, which in turn makes bTMSS relevant to atom-resonant interfaces, quantum memories, precision metrology, and distributed sensing (Kamble et al., 2024, Kim et al., 2018).

1. State definition and Gaussian structure

A general pure bTMSS can be written as

$$|\psi\rangle = D_a(\alpha)\,D_b(\beta)\,S_2(\xi)\,|0,0\rangle,$$

with

$$S_2(\xi)=\exp[\xi \hat a \hat b-\xi^\ast \hat a^\dagger \hat b^\dagger],$$

where $\xi=r e^{i\phi}$ is the complex squeezing parameter. A commonly used special case is the bright-seeded configuration

$$|\psi\rangle_{\rm bTMSS}=S_2(r)\,|u,0\rangle = S_2(r)\,D_a(u)\,|0,0\rangle,$$

in which one mode is seeded and the other is left in vacuum (Kamble et al., 2024).

In this representation, the state remains Gaussian and is therefore fully characterized by first and second moments. The bright component is encoded in the displacement vector, while the nonclassical resource resides in the covariance matrix. For the two-mode squeezed vacuum, the mean photon number in each mode is $\bar n_{\rm sq}=\sinh^2(r)$; when mode $j$ is displaced by a coherent amplitude $\alpha_j$, the total mean photon number becomes

$$\langle \hat n_j\rangle = |\alpha_j|^2 + \sinh^2(r),$$

and the total two-mode brightness is $\sum_{j=1}^2 \langle \hat n_j\rangle$ (Jensen et al., 2010).

The Heisenberg-picture form of the bright two-mode squeezing transformation makes the coexistence of seeding and pair creation explicit: $$\hat a_s=\hat a_0\cosh r+\hat b_0^\dagger \sinh r,\qquad \hat b_s=\hat b_0\cosh r+\hat a_0^\dagger \sinh r.$$ This form is frequently used in continuous-variable sensing analyses because it directly exposes how gain $$S_2(\xi)=\exp[\xi \hat a \hat b-\xi^\ast \hat a^\dagger \hat b^\dagger],$$0 amplifies a seed and how $$S_2(\xi)=\exp[\xi \hat a \hat b-\xi^\ast \hat a^\dagger \hat b^\dagger],$$1 injects correlated excitations into the conjugate mode (Kannath et al., 2 Aug 2025).

2. Correlations, squeezing observables, and brightness

The two canonical observables used to characterize bTMSS are joint quadratures and intensity-difference noise. With quadratures defined as

$$S_2(\xi)=\exp[\xi \hat a \hat b-\xi^\ast \hat a^\dagger \hat b^\dagger],$$2

two-mode squeezing generates correlations for which the joint quadratures $$S_2(\xi)=\exp[\xi \hat a \hat b-\xi^\ast \hat a^\dagger \hat b^\dagger],$$3 have variances $$S_2(\xi)=\exp[\xi \hat a \hat b-\xi^\ast \hat a^\dagger \hat b^\dagger],$$4. In separability tests, one may use the Duan criterion in the form

$$S_2(\xi)=\exp[\xi \hat a \hat b-\xi^\ast \hat a^\dagger \hat b^\dagger],$$5

which holds for any $$S_2(\xi)=\exp[\xi \hat a \hat b-\xi^\ast \hat a^\dagger \hat b^\dagger],$$6 in the treatment of seeded bTMSS used for absorption and gain metrology (Kamble et al., 2024).

A common operational signature of bright twin beams is reduced noise in the photon-number difference. For a two-mode squeezed vacuum, or equally bright coherent seed plus four-wave mixing, the variance of the intensity-difference observable $$S_2(\xi)=\exp[\xi \hat a \hat b-\xi^\ast \hat a^\dagger \hat b^\dagger],$$7 is

$$S_2(\xi)=\exp[\xi \hat a \hat b-\xi^\ast \hat a^\dagger \hat b^\dagger],$$8

and the squeezing level in decibels is

$$S_2(\xi)=\exp[\xi \hat a \hat b-\xi^\ast \hat a^\dagger \hat b^\dagger],$$9

In loss models relevant to hot-vapor four-wave mixing, the normalized intensity-difference noise is written as

$\xi=r e^{i\phi}$0

with parametric gain $\xi=r e^{i\phi}$1 (Jain et al., 4 Jul 2025, Monsa et al., 20 Jan 2026).

The bright seed changes the mean field but does not remove the entanglement structure. The literature explicitly contrasts bTMSS with both limiting cases: compared to two-mode squeezed vacuum, bTMSS carries a much larger mean photon number while preserving two-mode entanglement; compared to a coherent state, it retains quadrature squeezing and intermode correlations. The seed merely shifts means and does not destroy the entanglement inherent in $\xi=r e^{i\phi}$2 (Kamble et al., 2024).

3. Generation mechanisms and experimental realizations

The canonical $\xi=r e^{i\phi}$3 realization uses an optical parametric amplifier (OPA) pumped by a strong classical field at twice the signal frequency. In the undepleted-pump approximation, the interaction Hamiltonian is

$\xi=r e^{i\phi}$4

with squeezing parameter $\xi=r e^{i\phi}$5; injecting a bright coherent beam in one mode while leaving the other in vacuum yields a state of the form $\xi=r e^{i\phi}$6 (Kamble et al., 2024).

Representative realizations span waveguide PDC and several atomic-vapor four-wave-mixing architectures:

Platform	Mechanism	Reported figures
PP-KTP waveguide	Type II parametric downconversion	Mean photon number up to 2.5 per pulse; equivalent to 11 dB of two-mode squeezing (Eckstein et al., 2010)
Hot $\xi=r e^{i\phi}$7Rb, double-$\xi=r e^{i\phi}$8	Non-degenerate FWM near the D1 lines of $\xi=r e^{i\phi}$9Rb and $$|\psi\rangle_{\rm bTMSS}=S_2(r)\,|u,0\rangle = S_2(r)\,D_a(u)\,|0,0\rangle,$$0Rb	$$|\psi\rangle_{\rm bTMSS}=S_2(r)\,|u,0\rangle = S_2(r)\,D_a(u)\,|0,0\rangle,$$1 dB with both modes resonant with $$|\psi\rangle_{\rm bTMSS}=S_2(r)\,|u,0\rangle = S_2(r)\,D_a(u)\,|0,0\rangle,$$2Rb; $$|\psi\rangle_{\rm bTMSS}=S_2(r)\,|u,0\rangle = S_2(r)\,D_a(u)\,|0,0\rangle,$$3 dB and $$|\psi\rangle_{\rm bTMSS}=S_2(r)\,|u,0\rangle = S_2(r)\,D_a(u)\,|0,0\rangle,$$4 dB when only one mode is resonant (Kim et al., 2018)
Compact fiber-coupled hot $$|\psi\rangle_{\rm bTMSS}=S_2(r)\,|u,0\rangle = S_2(r)\,D_a(u)\,|0,0\rangle,$$5Rb	FWM at 795 nm with fiber-coupled inputs and outputs	4.4 dB after output fibers at 1 MHz with 135 mW pump power (Jain et al., 4 Jul 2025)
Ultra compact hot $$|\psi\rangle_{\rm bTMSS}=S_2(r)\,|u,0\rangle = S_2(r)\,D_a(u)\,|0,0\rangle,$$6Rb	FWM at 795 nm with a single fiber-coupled input, EOPM, and one Fabry–Pérot etalon	Up to $$|\psi\rangle_{\rm bTMSS}=S_2(r)\,|u,0\rangle = S_2(r)\,D_a(u)\,|0,0\rangle,$$7 dB at 0.8 MHz with 300 mW pump power (Monsa et al., 20 Jan 2026)

The pulsed PP-KTP waveguide implementation is notable for demonstrating a genuinely single-mode source in the telecom regime, with the single-mode character established through $$|\psi\rangle_{\rm bTMSS}=S_2(r)\,|u,0\rangle = S_2(r)\,D_a(u)\,|0,0\rangle,$$8 measurements and a mean photon number up to 2.5 per pulse, corresponding to 11 dB of two-mode squeezing (Eckstein et al., 2010). By contrast, the hot-vapor FWM sources are narrowband and spectrally compatible with alkali atomic transitions, which shifts their relevance toward atom–light interfaces and deployable continuous-variable systems (Kim et al., 2018, Jain et al., 4 Jul 2025).

Beyond purely optical nonlinear media, a hybrid three-mode cavity-optomechanical system has been proposed in which modulation of strong drives on two photon modes yields an effective two-photon squeezing Hamiltonian,

$$|\psi\rangle_{\rm bTMSS}=S_2(r)\,|u,0\rangle = S_2(r)\,D_a(u)\,|0,0\rangle,$$9

or equivalently

$\bar n_{\rm sq}=\sinh^2(r)$0

In that treatment, the intracavity classical amplitudes can be $\bar n_{\rm sq}=\sinh^2(r)$1–$\bar n_{\rm sq}=\sinh^2(r)$2, so the photon populations are explicitly bright while the effective interaction remains Gaussian (Qi, 4 Apr 2025).

4. Atom-resonant implementations and quantum memory interfaces

A major line of development has been the production of narrowband bTMSS on or near alkali resonances. In hot $\bar n_{\rm sq}=\sinh^2(r)$3Rb, non-degenerate FWM in a double-$\bar n_{\rm sq}=\sinh^2(r)$4 configuration can be operated so that both generated modes are simultaneously on resonance with transitions in the D1 line of $\bar n_{\rm sq}=\sinh^2(r)$5Rb, specifically one mode with the $\bar n_{\rm sq}=\sinh^2(r)$6 to $\bar n_{\rm sq}=\sinh^2(r)$7 transition and the other with the $\bar n_{\rm sq}=\sinh^2(r)$8 to $\bar n_{\rm sq}=\sinh^2(r)$9 transition. For this fully resonant configuration the reported intensity-difference squeezing is $j$0 dB, increasing to $j$1 dB and $j$2 dB when only one mode is resonant with the D1 $j$3 to $j$4 or $j$5 to $j$6 transition, respectively. The stated motivation is not only enhanced sensitivity in atomic-based sensors but also the deterministic entanglement of two distant atomic ensembles (Kim et al., 2018).

The memory interface for entangled bright two-mode squeezed inputs has been demonstrated in room-temperature cesium vapor. The reported memory consists of two cells, one for each mode, filled with cesium atoms at room temperature, with a memory time of about 1 msec. The input states are two-mode squeezed by 6.0 dB, have a variable orientation of squeezing, and are displaced by a few vacuum units. Preservation of quantum coherence is established by a measured memory fidelity of $j$7, which exceeds the benchmark of $j$8 for the best possible classical memory for the specified displacement range (Jensen et al., 2010).

Compact FWM sources at 795 nm sharpen the interface between bTMSS generation and alkali-based hardware. The fiber-coupled source reported in 2025 is narrowband and was explicitly described as ideal for atomic-based quantum sensing and quantum networking configurations that rely on atomic quantum memories (Jain et al., 4 Jul 2025). The 2026 ultra-compact implementation likewise emphasizes direct coupling to EIT- or Raman-based atomic quantum memories in $j$9Rb or $\alpha_j$0Rb, with twin beams centered on the Rb D1 line, MHz linewidths, and squeezing bandwidths of a few MHz (Monsa et al., 20 Jan 2026).

5. Metrological and networked sensing roles

In single-sample parameter estimation, bTMSS have been analyzed as probes for weak absorption and small optical gain. Two measurement schemes are considered: balanced photodetection and time-reversed metrology, both using two-mode bright squeezed light. The reported maximum quantum advantage over coherent-state probing is 3.7 times for the absorption parameter $\alpha_j$1 and 8.4 times for $\alpha_j$2. For gain estimation, the reported maximum quantum advantage is around 2.81 times for $\alpha_j$3 and around 6.28 times for $\alpha_j$4. These results are compared with the Cramér–Rao bound for a two-mode bright squeezed state, and the discussion distinguishes operating regimes in which balanced detection is simpler from regimes in which full SU(1,1) anti-squeezing becomes superior (Kamble et al., 2024).

The same basic resource has been extended to distributed sensing. In a network of optical gyroscopes, the task is to estimate the global phase shift corresponding to the average angular rotation across multiple spatially separated Sagnac loops. Two architectures are compared: an $\alpha_j$5 mode-entangled bTMSS configuration produced from a single bright pair and symmetric beam-splitter networks, and a separable $\alpha_j$6-bTMSS configuration using $\alpha_j$7 independent parametric amplifiers. Under 5% photon loss in every channel in the system, the proposed scheme shows a sensitivity enhancement of $\alpha_j$8 dB beyond the shot-noise limit, with an initial squeezing of $\alpha_j$9 dB. In the detailed analysis for $$\langle \hat n_j\rangle = |\alpha_j|^2 + \sinh^2(r),$$0, the entangled-versus-separable enhancement is reported as $$\langle \hat n_j\rangle = |\alpha_j|^2 + \sinh^2(r),$$1 $$\langle \hat n_j\rangle = |\alpha_j|^2 + \sinh^2(r),$$2 dB), and the mode-entangled architecture outperforms separable bTMSS throughout the practical parameter range considered (Kannath et al., 2 Aug 2025).

These metrological results are consistent with the broader role of bright twin beams in atom-based sensing. The narrowband FWM sources at 795 nm are repeatedly positioned as suited to atomic-based quantum sensing, while the atom-resonant $$\langle \hat n_j\rangle = |\alpha_j|^2 + \sinh^2(r),$$3Rb realization explicitly frames on-resonance squeezing as a route to enhanced sensitivity in metrology experiments based on atomic ensembles (Kim et al., 2018).

6. Limitations, stability, and development directions

The practical performance of bTMSS is strongly conditioned by loss, mode matching, and dynamical stability. In the compact fiber-coupled FWM source, residual optical losses of $$\langle \hat n_j\rangle = |\alpha_j|^2 + \sinh^2(r),$$4 per arm at the fiber facets and connectors reduce squeezing from $$\langle \hat n_j\rangle = |\alpha_j|^2 + \sinh^2(r),$$5 dB before fiber coupling to the measured $$\langle \hat n_j\rangle = |\alpha_j|^2 + \sinh^2(r),$$6 dB after the output fibers. The same study identifies imperfect overlap of correlated coherence areas, arising from the multimode character of FWM, as a source of admixture of uncorrelated thermal modes and excess noise (Jain et al., 4 Jul 2025).

The ultra-compact 2026 source makes the loss budget more explicit. It reports internal atomic absorption loss of $$\langle \hat n_j\rangle = |\alpha_j|^2 + \sinh^2(r),$$7, etalon/fiber/EOPM insertion loss of $$\langle \hat n_j\rangle = |\alpha_j|^2 + \sinh^2(r),$$8 dB, total PBS and AR-coated transmission $$\langle \hat n_j\rangle = |\alpha_j|^2 + \sinh^2(r),$$9, detector quantum efficiency $\sum_{j=1}^2 \langle \hat n_j\rangle$0, an electronic noise floor $\sum_{j=1}^2 \langle \hat n_j\rangle$1 dB below the measured squeezing, and an overall detection efficiency of approximately 80–85%. Under these conditions, with parametric gain $\sum_{j=1}^2 \langle \hat n_j\rangle$2, the measured slope ratio of approximately 0.135 corresponds to $\sum_{j=1}^2 \langle \hat n_j\rangle$3 dB of intensity-difference squeezing (Monsa et al., 20 Jan 2026).

A different limitation appears in open-system optomechanical generation. There, static covariance-matrix stability requires $\sum_{j=1}^2 \langle \hat n_j\rangle$4, which also limits steady-state squeezing to $\sum_{j=1}^2 \langle \hat n_j\rangle$5 dB. The optomechanical analysis shows, however, that by optimizing the squeezing quadrature operator one can obtain stable two-mode squeezing even in unstable system dynamics, and that the squeezing level can surpass the maximum achievable under stable system conditions. In that framework, the optimized quadrature remains bounded even when individual covariance-matrix elements diverge, so the usual steady-state stability condition is not the only meaningful operating criterion (Qi, 4 Apr 2025).

Current engineering directions follow directly from these bottlenecks. In hot-vapor FWM sources, the stated improvement paths include AR-coating fiber ends to reduce reflections below 1%, optimizing the pump-beam spatial profile to enlarge the coherence area and improve mode matching, and increasing pump power or cell length while keeping technical noise low (Jain et al., 4 Jul 2025). More generally, the recent narrowband sources indicate a movement toward low-SWaP, modular architectures that preserve the defining bTMSS combination of high flux, intermode quantum correlations, and direct compatibility with atomic platforms (Jain et al., 4 Jul 2025, Monsa et al., 20 Jan 2026).