Nonreciprocal Optomechanical Entanglement

• Nonreciprocal optomechanical entanglement is a quantum phenomenon where the entanglement in hybrid bosonic systems varies based on the input direction or sign of a control parameter.
• The topic utilizes Gaussian dynamics, including linearized Hamiltonians and Lyapunov equations, to model directional entanglement influenced by Sagnac shifts, asymmetric coupling, and synthetic gauge phases.
• Applications focus on directional quantum-state engineering and quantum information processing, with implementations ranging from spinning resonators and Fabry–Perot cavities to hybrid magnonic platforms.

Nonreciprocal optomechanical entanglement denotes a regime in which the amount, distribution, or even existence of quantum entanglement in an optomechanical or hybrid bosonic system depends on direction. In the most direct implementations, “direction” means driving a device from opposite sides so that left and right inputs generate different steady-state entanglement. In other implementations, the same asymmetry is encoded by a sign-reversible control parameter—such as a Sagnac–Fizeau shift, a Barnett-induced frequency shift, or a Kerr-induced magnon shift—that plays the role of forward versus backward configuration. Across recent work, the phenomenon has been realized or proposed in spinning whispering-gallery-mode resonators, asymmetric Fabry–Perot cavities, cascaded optomechanical networks, cavity magnomechanics, exciton optomechanics, and closed-loop optomechanical systems with directional squeezing (Berinyuy et al., 23 Jan 2025, Wu et al., 5 Jun 2026, Jiao et al., 2022, Lin et al., 2 Mar 2026, Jiao et al., 2024).

1. Definition and conceptual scope

In current usage, nonreciprocal entanglement is not limited to a single bipartition or a single physical platform. It includes direction-dependent photon–phonon entanglement, photon–vibration entanglement, vibration–vibration entanglement, microwave–optical entanglement, magnon–phonon entanglement, and tripartite continuous-variable entanglement, provided that the corresponding entanglement measure changes when the direction of driving or the sign of a directional control parameter is reversed (Berinyuy et al., 23 Jan 2025, Liu et al., 2024, Jiao et al., 2024). A central point established across the literature is that nonreciprocity here concerns quantum correlations themselves, not merely transport asymmetry.

The modern literature separates several logically distinct notions. Xu and Li’s three-mode system established optical nonreciprocal response and a three-port optomechanical circulator through a gauge-invariant phase difference between linearized couplings, but did not explicitly compute entanglement (Xu et al., 2015). A silicon optomechanical circuit later demonstrated synthetic magnetism and reservoir engineering for nonreciprocal transport and directional amplification, thereby supplying a microchip architecture in which nonreciprocal quantum-state engineering becomes plausible (Fang et al., 2016). Explicit nonreciprocal optomechanical entanglement was then formulated in spinning-resonator cavity optomechanics, where photons and phonons can be strongly entangled in one direction but fully uncorrelated in the other, even in regimes with no classical nonreciprocity (Jiao et al., 2020).

The term also requires some care because “direction” is platform-dependent. In spinning WGM and asymmetric Fabry–Perot systems, direction refers to forward versus backward optical driving (Berinyuy et al., 23 Jan 2025, Wu et al., 5 Jun 2026). In Barnett- and Kerr-based magnomechanical systems, direction is encoded by the sign of the Barnett shift or by the sign of the Kerr coefficient under magnetic-field reversal (Chabar et al., 29 Sep 2025, Chen et al., 2023). In directional squeezing schemes, the asymmetry is tied to whether the coherent drive is applied from the squeezing-injection side (Jiao et al., 2024). This broader usage suggests that nonreciprocal optomechanical entanglement is best understood as direction-sensitive Gaussian entanglement generated by explicitly broken time-reversal symmetry or by direction-dependent effective couplings.

2. Gaussian dynamical framework and entanglement quantifiers

The prevailing theoretical framework is a linearized Gaussian treatment. One starts from a driven nonlinear Hamiltonian, adds dissipation and input noise through quantum Langevin equations, and expands each mode operator as a steady-state mean plus fluctuation under strong driving. The resulting fluctuation dynamics takes the form

$\dot{\chi}(t)=A\chi(t)+n(t),$

or equivalently $$\dot{\mathbf u}(t)=\mathbf A \mathbf u(t)+\mathbf n(t)$$, where the drift matrix contains the effective detunings, linearized couplings, damping rates, and any directional asymmetry induced by rotation, structural imbalance, synthetic phases, or parametric processes (Berinyuy et al., 23 Jan 2025, Wu et al., 5 Jun 2026).

Because the noise is Gaussian, the steady state is completely characterized by the covariance matrix $V$, with entries $$V_{ij}=\frac{1}{2}\langle \chi_i\chi_j+\chi_j\chi_i\rangle$$. In steady state, $V$ is obtained from the Lyapunov equation

$A V + V A^T = -D,$

with diffusion matrix $D$ determined by optical vacuum noise and thermal occupations of the mechanical, vibrational, magnonic, or excitonic baths (Berinyuy et al., 23 Jan 2025, Wu et al., 5 Jun 2026). Stability is imposed by requiring that all eigenvalues of the drift matrix have negative real parts; several works explicitly invoke the Routh–Hurwitz criterion or numerical eigenvalue checks (Berinyuy et al., 23 Jan 2025, Wu et al., 5 Jun 2026, Xu et al., 2018).

For bipartite entanglement, the standard measure is the logarithmic negativity. Given a $4\times 4$ reduced covariance matrix $V_{\rm sub}$, one computes the smallest symplectic eigenvalue $\zeta$ or $$\dot{\mathbf u}(t)=\mathbf A \mathbf u(t)+\mathbf n(t)$$0 of the partially transposed covariance matrix and defines

$$\dot{\mathbf u}(t)=\mathbf A \mathbf u(t)+\mathbf n(t)$$1

or equivalently $$\dot{\mathbf u}(t)=\mathbf A \mathbf u(t)+\mathbf n(t)$$2 in the Fabry–Perot formulation (Berinyuy et al., 23 Jan 2025, Wu et al., 5 Jun 2026). For tripartite continuous-variable states, the minimum residual contangle is frequently adopted:

$$\dot{\mathbf u}(t)=\mathbf A \mathbf u(t)+\mathbf n(t)$$3

with $$\dot{\mathbf u}(t)=\mathbf A \mathbf u(t)+\mathbf n(t)$$4 and contangle defined as the square of the logarithmic negativity (Lin et al., 2 Mar 2026, Jiao et al., 2024).

Directionality is quantified by contrast-type measures. In the molecular WGM system, the bidirectional contrast ratio is

$$\dot{\mathbf u}(t)=\mathbf A \mathbf u(t)+\mathbf n(t)$$5

with $$\dot{\mathbf u}(t)=\mathbf A \mathbf u(t)+\mathbf n(t)$$6 for reciprocal entanglement and $$\dot{\mathbf u}(t)=\mathbf A \mathbf u(t)+\mathbf n(t)$$7 for ideal nonreciprocity (Berinyuy et al., 23 Jan 2025). In the asymmetric Fabry–Perot cavity, Wu et al. use

$$\dot{\mathbf u}(t)=\mathbf A \mathbf u(t)+\mathbf n(t)$$8

in direct analogy with the classical transmission asymmetry $$\dot{\mathbf u}(t)=\mathbf A \mathbf u(t)+\mathbf n(t)$$9 (Wu et al., 5 Jun 2026). Barnett- and Kerr-based magnomechanical papers use the same absolute-difference-over-sum structure for bipartite and tripartite measures (Chabar et al., 29 Sep 2025, Chen et al., 2023).

3. Microscopic sources of nonreciprocity

The microscopic origin of nonreciprocal optomechanical entanglement varies by platform, but the common mechanism is a direction-dependent modification of effective detunings, effective damping, or phase-sensitive couplings.

A first major class relies on rotation-induced Sagnac–Fizeau shifts. In spinning WGM systems, the cavity resonance acquires a shift $V$0 whose sign depends on propagation direction. The same laser frequency can therefore appear effectively red-detuned in one direction and blue-detuned in the opposite direction. Since blue detuning enhances two-mode squeezing and red detuning favors cooling, the entanglement becomes direction-dependent (Berinyuy et al., 23 Jan 2025, Jiao et al., 2020). In cascaded spinning resonators, this same mechanism permits the individual optimization of the local light–motion interaction in each resonator, enabling nonreciprocal remote entanglement between distant mechanical oscillators (Jiao et al., 2022).

A second class uses structural asymmetry. In asymmetric Fabry–Perot cavities, $V$1 implies $V$2, and the resulting directional differences in intracavity field and radiation pressure make forward and backward steady states inequivalent (Wu et al., 5 Jun 2026). Here the asymmetry is geometric rather than rotational, but the effect is again a direction-dependent linearized coupling $V$3 and an associated direction-dependent covariance matrix.

A third class uses synthetic gauge phases and dissipative interference. In the three-mode model of Xu and Li, the crucial parameter is the gauge-invariant phase $V$4, and time-reversal symmetry is broken when $V$5 (Xu et al., 2015). In integrated optomechanical circuits, synthetic flux and mechanically mediated dissipative coupling can be arranged so that coherent and dissipative pathways interfere destructively in one direction and not the other, yielding nonreciprocal transport and directional amplification (Fang et al., 2016). Although those works focus on response rather than entanglement, they supply the phase-engineering and reservoir-engineering templates for nonreciprocal Gaussian correlation control.

A fourth class exploits active or internal nonlinear resources. In spinning exciton optomechanics, an intracavity optical parametric amplifier modifies effective detunings and damping rates through $V$6 and $V$7, and thereby regulates the Sagnac-induced nonreciprocal entanglement (Lin et al., 2 Mar 2026). In magnomechanics, the Barnett effect adds a sign-reversible magnon frequency shift $V$8, while Kerr nonlinearity produces both a frequency shift and a pair-magnon term whose sign depends on magnetic-field direction (Chabar et al., 29 Sep 2025, Chen et al., 2023, Liu et al., 2024). In closed-loop WGM optomechanics, directional two-photon driving plus a phase-matched squeezed reservoir creates a squeezed optical mode only for one input direction, so that the enhanced effective couplings $V$9 exist only on that side (Jiao et al., 2024).

Platform class	Nonreciprocal ingredient	Representative entanglement outputs
Spinning WGM optomechanics	Sagnac–Fizeau shift $$V_{ij}=\frac{1}{2}\langle \chi_i\chi_j+\chi_j\chi_i\rangle$$0	photon–phonon, photon–vibration, vibration–vibration
Asymmetric Fabry–Perot cavity	$$V_{ij}=\frac{1}{2}\langle \chi_i\chi_j+\chi_j\chi_i\rangle$$1, $$V_{ij}=\frac{1}{2}\langle \chi_i\chi_j+\chi_j\chi_i\rangle$$2	forward/backward optomechanical entanglement
Phase-engineered three-mode circuits	gauge phase $$V_{ij}=\frac{1}{2}\langle \chi_i\chi_j+\chi_j\chi_i\rangle$$3, synthetic flux	directional optical or optomechanical Gaussian correlations
Barnett/Kerr/OPA hybrids	$$V_{ij}=\frac{1}{2}\langle \chi_i\chi_j+\chi_j\chi_i\rangle$$4, Kerr shift, parametric gain	magnon–phonon, microwave–optical, photon–exciton
Directional squeezing loops	one-sided squeezed optical mode	bipartite and tripartite nonreciprocal entanglement

4. Representative physical realizations

A particularly explicit realization is the molecular cavity optomechanical system with $$V_{ij}=\frac{1}{2}\langle \chi_i\chi_j+\chi_j\chi_i\rangle$$5 molecules inside a spinning WGM resonator coupled to an auxiliary optical cavity. The molecular vibrations are grouped into two collective modes $$V_{ij}=\frac{1}{2}\langle \chi_i\chi_j+\chi_j\chi_i\rangle$$6 and $$V_{ij}=\frac{1}{2}\langle \chi_i\chi_j+\chi_j\chi_i\rangle$$7, with collective couplings $$V_{ij}=\frac{1}{2}\langle \chi_i\chi_j+\chi_j\chi_i\rangle$$8 and $$V_{ij}=\frac{1}{2}\langle \chi_i\chi_j+\chi_j\chi_i\rangle$$9. In this setting, nonreciprocal photon–vibration entanglement $V$0 and nonreciprocal vibration–vibration entanglement $V$1 arise because CCW driving with $V$2 pushes the system toward the blue sideband, whereas CW driving with $V$3 behaves more like a red-detuned cooling configuration. The paper identifies $V$4 and $V$5 as optimal for enhanced nonreciprocity in the CCW case, uses experimentally feasible parameters such as $V$6 THz, $V$7 GHz, $V$8, $V$9, $A V + V A^T = -D,$0, and $A V + V A^T = -D,$1, and reports that vibration–vibration entanglement is greatly enhanced as $A V + V A^T = -D,$2 grows while remaining nonzero at high temperatures around $A V + V A^T = -D,$3 K (Berinyuy et al., 23 Jan 2025).

Wu et al. analyze a single-mode asymmetric Fabry–Perot cavity formed by a movable metasurface mirror and a fixed mirror, with forward and backward driving defined by the incident side. For $A V + V A^T = -D,$4, $A V + V A^T = -D,$5, $A V + V A^T = -D,$6 mW, $A V + V A^T = -D,$7 mK, and $A V + V A^T = -D,$8 MHz, they find $A V + V A^T = -D,$9 and $D$0 near optimal detuning. In the asymmetric cavity, forward entanglement persists up to $D$1 K in the 2D detuning–temperature plot and up to about $D$2 K at fixed $D$3, whereas backward entanglement disappears around $D$4 K. For a symmetric cavity with the same total optical dissipation, the maximal entanglement is $D$5, the 2D temperature limit is about $D$6 K, and the fixed-detuning limit is about $D$7 K (Wu et al., 5 Jun 2026).

In spinning exciton optomechanics with an embedded optical parametric amplifier, the Sagnac shift makes the photon–exciton, photon–phonon, exciton–phonon, and tripartite entanglements direction-dependent, while the OPA selectively reshapes them. The OPA significantly enhances photon–exciton entanglement and tripartite entanglement but weakens photon–phonon and exciton–phonon entanglement. With $D$8 and $D$9, photon–exciton entanglement survives up to about $4\times 4$0 K for $4\times 4$1 and about $4\times 4$2 K for $4\times 4$3, whereas without OPA it vanishes around $4\times 4$4 K (Lin et al., 2 Mar 2026).

Hybrid magnonic platforms supply several additional realizations. In Barnett-effect magnomechanics with an OPA, the YIG rotation frequency enters as a Barnett-induced shift $4\times 4$5 in the magnon detuning, and ideal nonreciprocity can be reached for cavity–magnon, magnon–phonon, and cavity–phonon entanglements by tuning detuning, coupling regime, OPA gain, and phase (Chabar et al., 29 Sep 2025). In cavity-magnon optomechanics with magnon Kerr nonlinearity, the sign of the Kerr coefficient is controlled by the magnetic-field direction, so the same device can switch from reciprocal to ideally nonreciprocal bipartite and tripartite entanglement (Chen et al., 2023). In Kerr-modified cavity optomagnomechanics, the same mechanism is extended to microwave–optical photon–photon entanglement, optical photon–magnon entanglement, and magnon–phonon entanglement, with ideal nonreciprocity achievable through the sign of the Kerr-induced shift $4\times 4$6 (Liu et al., 2024).

Nonreciprocal entanglement can also be distributed over distance. In a cascaded configuration of two telecommunication-fiber-coupled spinning WGM optomechanical resonators, remote entanglement between two spatially separated mechanical oscillators can be produced only through driving from one specific input direction. For frequency-mismatched oscillators, spinning one resonator can enhance the maximum entanglement by factors up to about $4\times 4$7 for $4\times 4$8 and $4\times 4$9 for $V_{\rm sub}$0 compared with the static mismatched case, and the revival coefficient $V_{\rm sub}$1 can exceed $V_{\rm sub}$2 or even $V_{\rm sub}$3 in suitable parameter regions (Jiao et al., 2022).

5. Tunability, robustness, and common misunderstandings

A defining feature of the subject is that nonreciprocity is often tunable rather than fixed. In the molecular WGM system, the auxiliary-cavity detuning $V_{\rm sub}$4 switches the bidirectional contrast from nearly zero to nearly one; the regions $V_{\rm sub}$5 and $V_{\rm sub}$6 support maximal nonreciprocity for both photon–vibration and vibration–vibration entanglement, while around $V_{\rm sub}$7 the nonreciprocity vanishes (Berinyuy et al., 23 Jan 2025). In Barnett- and Kerr-controlled magnomechanical systems, contrast ratios can likewise be tuned continuously from reciprocal to ideal nonreciprocal behavior by detuning, nonlinear gain, or magnetic-field orientation (Chabar et al., 29 Sep 2025, Chen et al., 2023).

Robustness is equally platform-specific. In the spinning WGM cavity of Jiao and collaborators, the Sagnac effect can revive optomechanical entanglement against backscattering losses, with the revival factor $V_{\rm sub}$8 reaching up to $V_{\rm sub}$9 for suitable $\zeta$0, and the nonreciprocal entanglement remains significant up to temperatures around $\zeta$1 mK (Jiao et al., 2020). In the Fabry–Perot platform, asymmetry improves both strength and thermal robustness relative to a symmetric cavity, even though the same total optical dissipation is maintained (Wu et al., 5 Jun 2026). In the feedback-assisted hybrid magnomechanical–optomechanical scheme, increasing the beam-splitter reflectivity enhances stationary entanglement and steering and can effectively counter thermal noise (Chabar et al., 13 Jul 2025). In exciton optomechanics with OPA, photon–exciton entanglement becomes robust against cavity dissipation, whereas phonon-involving and tripartite entanglements can become less thermally robust (Lin et al., 2 Mar 2026).

Several recurring misconceptions are corrected explicitly in the literature. First, classical and quantum nonreciprocities are not equivalent. Wu et al. show that $\zeta$2 and $\zeta$3 do not exhibit a simple positive correlation: the cavity can be highly nonreciprocal classically while the entanglement is small and nearly symmetric, or the entanglement can be strongly nonreciprocal while classical transmission is almost reciprocal (Wu et al., 5 Jun 2026). Second, quantum nonreciprocity can exist even in the absence of classical nonreciprocity: in the spinning-resonator scheme, there are detunings for which the intracavity photon numbers are reciprocal while the entanglement is strongly direction-dependent (Jiao et al., 2020). Third, nonlinearity alone does not guarantee nonreciprocal behavior. In the three-mode system with one mechanically coupled optical mode, the optomechanically induced Kerr nonlinearity is necessary but not sufficient; an additional impedance-matching condition must be broken, otherwise the effective nonlinearities in the two directions become equal and the response remains reciprocal (Xu et al., 2018).

A further conceptual clarification is that “forward” and “backward” are not always literal propagation directions. In Kerr-modified cavity-magnon optomechanics and in cavity-magnon optomechanics with magnon Kerr effect, the direction parameter is the sign of the Kerr coefficient under magnetic-field reversal (Liu et al., 2024, Chen et al., 2023). In Barnett-effect systems it is the sign of $\zeta$4 (Chabar et al., 29 Sep 2025). In directional quantum squeezing, the distinction is whether the coherent drive enters from the squeezing injection direction (Jiao et al., 2024). This suggests that nonreciprocal optomechanical entanglement is fundamentally a question of asymmetric quantum-state generation under a reversed control configuration, not only of unidirectional wave propagation.

6. Applications and emerging directions

The applications stated across these works concentrate on directional quantum information processing. The molecular WGM proposal identifies quantum information transmission and the development of nonreciprocal quantum devices as direct motivations (Berinyuy et al., 23 Jan 2025). The asymmetric Fabry–Perot study positions asymmetric cavities as a platform to explore the connection between classical and quantum nonreciprocities and as a route toward magnet-free nonreciprocal quantum photonic and phononic circuits (Wu et al., 5 Jun 2026). Directional squeezing in closed-loop optomechanics is proposed as a source of bipartite and tripartite entanglement and of one-way or two-way EPR steering, with explicit relevance to quantum secure direct communication and one-way quantum computing (Jiao et al., 2024).

Hybrid bosonic systems extend the scope from optomechanical entanglement proper to directional transduction and network functionality. The exciton-optomechanical OPA scheme points toward room-temperature nonreciprocal quantum technologies and directional quantum interfaces between optical fields and solid-state excitons (Lin et al., 2 Mar 2026). Barnett- and feedback-assisted magnomechanical systems identify nonreciprocal single-phonon devices, asymmetric steering, multipartite entanglement, and thermally robust quantum links as plausible targets (Chabar et al., 29 Sep 2025, Chabar et al., 13 Jul 2025). The integrated optomechanical circuit with synthetic magnetism and reservoir engineering indicates the feasibility of more general nonreciprocal optical devices and, by extension, topological phases for both light and sound on a microchip (Fang et al., 2016).

Several research directions recur across the literature. One is the expansion from bipartite to multipartite nonreciprocal resources, including genuine tripartite entanglement and asymmetric EPR steering in closed loops and hybrid chains (Jiao et al., 2024, Chabar et al., 13 Jul 2025). Another is the use of active resources—OPA gain, coherent feedback, synthetic flux, or directional squeezing—to decouple entanglement robustness from classical transmission performance (Lin et al., 2 Mar 2026, Chabar et al., 13 Jul 2025). A third is the extension from single devices to networks: cascaded WGM resonators already demonstrate that directional entanglement can be distributed between remote mechanical oscillators, and the same design logic suggests larger chiral optomechanical networks (Jiao et al., 2022). A plausible implication is that the field is converging on a unified control paradigm in which direction-dependent detunings, phase-engineered couplings, and engineered reservoirs are combined to make entanglement itself a directional resource rather than a by-product of directional transport.