Nonreciprocal SQPT in Cavity & Magnonic Systems

• Nonreciprocal SQPT is a quantum phase transition where directional light–matter interactions yield asymmetric critical thresholds in cavity and magnonic platforms.
• It employs mechanisms like cavity rotation, directional squeezing, and sign-selective Kerr nonlinearities to trigger first- and second-order transitions, bistability, and hysteresis.
• Generalized Dicke models and rigorous microscopic treatments inform these designs, ensuring compatibility with TRK constraints and A² terms while enabling directional control.

Nonreciprocal superradiant quantum phase transition denotes a superradiant transition in which the onset of macroscopic light–matter order depends on direction, propagation channel, or an effectively directional control configuration. In the recent literature, this notion appears in driven–dissipative cavity QED and cavity magnonic systems where cavity rotation, directional squeezing, or sign-selective Kerr nonlinearities produce distinct critical thresholds, distinct steady-state phase diagrams, or even a transition in one direction but not the other (Zhu et al., 2024). Closely related precursor physics arises already in generalized Dicke models with asymmetric co- and counter-rotating couplings, where first-order transitions, phase coexistence, hysteresis, switching, and oscillatory superradiant phases emerge from interaction asymmetry plus dissipation, even when the microscopic model remains Hermitian and parity-symmetric (Stitely et al., 2020).

1. Conceptual definition and symmetry structure

A superradiant quantum phase transition (SQPT) is a transition from a normal phase with no coherent radiation to a superradiant phase with a macroscopic cavity photon field and collective matter polarization. In Dicke-type settings it is typically associated with spontaneous breaking of a discrete parity symmetry, while in Tavis–Cummings-type settings it is associated with a continuous $U(1)$ symmetry (Larson et al., 2016).

The nonreciprocal variant is characterized by direction-dependent criticality. In a spinning whispering-gallery-mode microcavity with a unidirectionally squeezed mode, the combination of cavity rotation and directional squeezing produces different critical pump strengths for forward and backward propagation, yielding nonreciprocal first- and second-order superradiant phase transitions (Zhu et al., 2024). In a spinning microwave magnonic system, the Sagnac–Fizeau shift modifies the critical driving strengths differently for clockwise and counterclockwise modes, so that the quantum phase transition can occur when the system is driven in one direction but not the other (Xu et al., 2024). In a cavity magnonic system with yttrium iron garnet, the sign of the magnon Kerr coefficient $K$, controlled by the crystallographic orientation of the bias magnetic field, produces distinct critical thresholds and steady-state phase diagrams for $K>0$ and $K<0$, which is identified as a nonreciprocal SQPT (Zhang et al., 30 Sep 2025).

A generalized open Dicke model provides an important conceptual bridge. Its Hamiltonian,

$$\hat{H} = \omega \hat{a}^{\dagger}\hat{a} + \omega_0 \hat{J}_{z} + \frac{\lambda_{-}}{\sqrt{N}}\left( \hat{a}\hat{J}_{+} + \hat{a}^{\dagger}\hat{J}_{-} \right) + \frac{\lambda_{+}}{\sqrt{N}}\left( \hat{a}\hat{J}_{-} + \hat{a}^{\dagger}\hat{J}_{+} \right),$$

retains a global $\mathbb{Z}_2$ parity symmetry but allows $\lambda_-\neq\lambda_+$, thereby asymmetrically weighting co- and counter-rotating channels (Stitely et al., 2020). Although this model does not realize spatial nonreciprocity, it establishes that asymmetric light–matter interaction channels combined with dissipation can generate history dependence, first-order superradiant onset, multistability, and hysteresis. This suggests a direct connection between interaction asymmetry and explicitly nonreciprocal SQPT architectures.

2. Generalized Dicke models as the immediate precursor of nonreciprocal SQPT physics

In the unbalanced Dicke model, the semiclassical variables

$$\alpha=\langle \hat a\rangle/\sqrt N,\qquad \beta=\langle \hat J_-\rangle/N,\qquad \gamma=\langle \hat J_z\rangle/N$$

obey nonlinear mean-field equations whose fixed points and limit cycles determine the phase diagram (Stitely et al., 2020). The relevant phases are the normal phase, a static superradiant phase, a coexistence region where normal and superradiant solutions are both locally stable, and a superradiant oscillatory phase generated by a Hopf bifurcation (Stitely et al., 2020).

The phase structure depends qualitatively on asymmetry. For balanced or weakly unbalanced couplings, the normal phase loses stability in a pitchfork bifurcation and the superradiant order parameter grows continuously, reproducing the standard second-order Dicke transition. For strongly unbalanced couplings, the onset becomes two-stage: stable and unstable superradiant equilibria first appear at finite amplitude through saddle-node bifurcations while the normal phase remains locally stable, and only at a second bifurcation does the normal state become unstable. The resulting coexistence region is the locus of first-order superradiant behavior, switching dynamics, and quantum hysteresis (Stitely et al., 2020).

The critical point separating single-stage and two-stage onset occurs at

$$\lambda_{\pm}^* =\sqrt{\frac{\omega_0}{2\omega}\,\big(\zeta \mp \omega\sqrt{\zeta}\big)}, \qquad \zeta=\kappa^2+\omega^2,$$

where saddle-node and pitchfork bifurcations coincide (Stitely et al., 2020). In this regime, the model exhibits three central nonreciprocal precursors: discontinuous superradiant onset, coexistence of parity-symmetric and parity-broken attractors, and direction-dependent history in control-parameter space.

The same paper identifies fluctuation-induced switching between normal and superradiant phases through stochastic Schrödinger equations, and a quantum hysteresis loop under linear ramps of $\lambda_+$ with ramp time $K$0 for $K$1 atoms (Stitely et al., 2020). Quantum trajectories display jumps among attractors in the coexistence region, while the oscillatory superradiant phase appears as a broadened spectral peak around the semiclassical limit-cycle frequency. These features are not yet spatial nonreciprocity, but they supply the dynamical grammar—multistability, hysteresis, metastability, and bifurcation structure—of later nonreciprocal SQPT proposals.

3. Microscopic constraints, no-go theorems, and routes to superradiant criticality

Any encyclopedia treatment of nonreciprocal SQPT must also account for the longstanding microscopic constraints on superradiant criticality. In circuit QED, a fully microscopic derivation based on BCS pseudo-spin variables and minimal coupling yields a Dicke-type Hamiltonian with both the linear dipole term and the circuit analogue of the $K$2 term,

$K$3

originating from the $K$4 contribution to the charging Hamiltonian (Xu et al., 2012). The resulting coupled-oscillator spectrum,

$K$5

supports a superradiant instability when

$K$6

which reduces to $K$7 in the thermodynamic limit (Xu et al., 2012). This establishes that a Dicke-type SQPT can occur in circuit QED for suitable microscopic parameters even when the correct quadratic term is retained.

In natural-atom cavity QED, by contrast, the Thomas–Reiche–Kuhn sum rule and the $K$8 term enforce the standard no-go theorem for continuous superradiant transitions. A three-level $K$9-system interacting with two bosonic modes shows that this restriction applies specifically to second-order onset from the normal phase: within the physically allowed TRK region, the normal state remains locally stable, yet a first-order superradiant transition can still occur when a disconnected superradiant minimum becomes globally favorable (Hayn et al., 2012). This distinction between continuous and discontinuous superradiant criticality is central for realistic nonreciprocal designs, because direction-dependent transitions often exploit multistability and first-order structure rather than a single continuous Dicke instability.

A further interacting route is provided by coupled two-level atoms with an all-to-all XY interaction,

$K>0$0

added to a Dicke Hamiltonian with an $K>0$1 term (Liu et al., 2021). The critical coupling becomes

$K>0$2

while the TRK bound remains $K>0$3 (Liu et al., 2021). For attractive interaction $K>0$4, a parameter window $K>0$5 opens in which an SQPT is compatible with the sum rule (Liu et al., 2021). This result is directly relevant to nonreciprocal SQPT because it shows that structured matter-sector interactions can lower the effective superradiant threshold without abandoning microscopic consistency.

A different gauge-invariant route appears in Landau polaritons with Rashba and Zeeman couplings. There, the order parameter is a spatially structured magnetic field rather than a uniform vector potential, and the SQPT is formulated as a zero-frequency polariton softening in Maxwell’s equation coupled to the current response kernel $K>0$6 (Manzanares et al., 2022). The in-plane Zeeman mechanism yields the criterion

$K>0$7

while Rashba- or Zeeman-induced Landau-level crossings enhance the susceptibility through nearly gapless denominators $K>0$8 (Manzanares et al., 2022). Although this work does not analyze transport nonreciprocity, it furnishes a magnetized and spin–orbit-coupled setting in which broken time-reversal symmetry, finite-$K>0$9 mode structure, and superradiant criticality coexist.

4. Explicit nonreciprocal realizations

The first explicit cavity-QED realization employs a rotating whispering-gallery-mode microcavity containing a two-level atom and a unidirectionally squeezed $K<0$0 mode. In the rotating frame at $K<0$1, the forward-pump Hamiltonian is

$K<0$2

$K<0$3

where the Sagnac–Fizeau shift $K<0$4 changes sign between forward and backward pumping (Zhu et al., 2024). In the symmetric-coupling case, the first-order critical pump strength is

$K<0$5

while the second-order critical line is given by an explicit $K<0$6 expression obtained from the NP stability matrix (Zhu et al., 2024). Forward and backward pumping yield different $K<0$7 and $K<0$8, so there are parameter windows where the system is normal in one direction and superradiant in the other (Zhu et al., 2024). The phase diagram also contains a tricritical point at $K<0$9 and a direction-dependent splitting or disappearance of multicritical points (Zhu et al., 2024).

A closely related magnonic implementation uses a spinning microwave resonator coupled to a YIG sphere with Kerr nonlinearity $$\hat{H} = \omega \hat{a}^{\dagger}\hat{a} + \omega_0 \hat{J}_{z} + \frac{\lambda_{-}}{\sqrt{N}}\left( \hat{a}\hat{J}_{+} + \hat{a}^{\dagger}\hat{J}_{-} \right) + \frac{\lambda_{+}}{\sqrt{N}}\left( \hat{a}\hat{J}_{-} + \hat{a}^{\dagger}\hat{J}_{+} \right),$$0. In the rotating frame, the effective non-Hermitian Hamiltonian is

$$\hat{H} = \omega \hat{a}^{\dagger}\hat{a} + \omega_0 \hat{J}_{z} + \frac{\lambda_{-}}{\sqrt{N}}\left( \hat{a}\hat{J}_{+} + \hat{a}^{\dagger}\hat{J}_{-} \right) + \frac{\lambda_{+}}{\sqrt{N}}\left( \hat{a}\hat{J}_{-} + \hat{a}^{\dagger}\hat{J}_{+} \right),$$1

with $$\hat{H} = \omega \hat{a}^{\dagger}\hat{a} + \omega_0 \hat{J}_{z} + \frac{\lambda_{-}}{\sqrt{N}}\left( \hat{a}\hat{J}_{+} + \hat{a}^{\dagger}\hat{J}_{-} \right) + \frac{\lambda_{+}}{\sqrt{N}}\left( \hat{a}\hat{J}_{-} + \hat{a}^{\dagger}\hat{J}_{+} \right),$$2 (Xu et al., 2024). The steady-state magnon amplitudes satisfy

$$\hat{H} = \omega \hat{a}^{\dagger}\hat{a} + \omega_0 \hat{J}_{z} + \frac{\lambda_{-}}{\sqrt{N}}\left( \hat{a}\hat{J}_{+} + \hat{a}^{\dagger}\hat{J}_{-} \right) + \frac{\lambda_{+}}{\sqrt{N}}\left( \hat{a}\hat{J}_{-} + \hat{a}^{\dagger}\hat{J}_{+} \right),$$3

and the critical drive strengths $$\hat{H} = \omega \hat{a}^{\dagger}\hat{a} + \omega_0 \hat{J}_{z} + \frac{\lambda_{-}}{\sqrt{N}}\left( \hat{a}\hat{J}_{+} + \hat{a}^{\dagger}\hat{J}_{-} \right) + \frac{\lambda_{+}}{\sqrt{N}}\left( \hat{a}\hat{J}_{-} + \hat{a}^{\dagger}\hat{J}_{+} \right),$$4 and $$\hat{H} = \omega \hat{a}^{\dagger}\hat{a} + \omega_0 \hat{J}_{z} + \frac{\lambda_{-}}{\sqrt{N}}\left( \hat{a}\hat{J}_{+} + \hat{a}^{\dagger}\hat{J}_{-} \right) + \frac{\lambda_{+}}{\sqrt{N}}\left( \hat{a}\hat{J}_{-} + \hat{a}^{\dagger}\hat{J}_{+} \right),$$5 depend explicitly on $$\hat{H} = \omega \hat{a}^{\dagger}\hat{a} + \omega_0 \hat{J}_{z} + \frac{\lambda_{-}}{\sqrt{N}}\left( \hat{a}\hat{J}_{+} + \hat{a}^{\dagger}\hat{J}_{-} \right) + \frac{\lambda_{+}}{\sqrt{N}}\left( \hat{a}\hat{J}_{-} + \hat{a}^{\dagger}\hat{J}_{+} \right),$$6, hence on the sign of $$\hat{H} = \omega \hat{a}^{\dagger}\hat{a} + \omega_0 \hat{J}_{z} + \frac{\lambda_{-}}{\sqrt{N}}\left( \hat{a}\hat{J}_{+} + \hat{a}^{\dagger}\hat{J}_{-} \right) + \frac{\lambda_{+}}{\sqrt{N}}\left( \hat{a}\hat{J}_{-} + \hat{a}^{\dagger}\hat{J}_{+} \right),$$7 (Xu et al., 2024). The result is a nonreciprocal quantum phase transition between a parity-symmetric phase, a parity-symmetry-broken phase, and a bistable phase, with a direction-dependent isolation parameter $$\hat{H} = \omega \hat{a}^{\dagger}\hat{a} + \omega_0 \hat{J}_{z} + \frac{\lambda_{-}}{\sqrt{N}}\left( \hat{a}\hat{J}_{+} + \hat{a}^{\dagger}\hat{J}_{-} \right) + \frac{\lambda_{+}}{\sqrt{N}}\left( \hat{a}\hat{J}_{-} + \hat{a}^{\dagger}\hat{J}_{+} \right),$$8 defined from $$\hat{H} = \omega \hat{a}^{\dagger}\hat{a} + \omega_0 \hat{J}_{z} + \frac{\lambda_{-}}{\sqrt{N}}\left( \hat{a}\hat{J}_{+} + \hat{a}^{\dagger}\hat{J}_{-} \right) + \frac{\lambda_{+}}{\sqrt{N}}\left( \hat{a}\hat{J}_{-} + \hat{a}^{\dagger}\hat{J}_{+} \right),$$9 and $\mathbb{Z}_2$0 (Xu et al., 2024).

A second magnonic realization eliminates literal rotation and instead uses the sign of the magnon Kerr coefficient as the nonreciprocal control variable. In a cavity magnonic system with a parametrically driven cavity and a YIG sphere, the Hamiltonian is

$\mathbb{Z}_2$1

and the sign of $\mathbb{Z}_2$2 is set by the orientation of the bias magnetic field relative to the crystallographic axes: $\mathbb{Z}_2$3 along $\mathbb{Z}_2$4, $\mathbb{Z}_2$5 along $\mathbb{Z}_2$6 (Zhang et al., 30 Sep 2025). The steady-state magnon number,

$\mathbb{Z}_2$7

selects different physical branches for $\mathbb{Z}_2$8 and $\mathbb{Z}_2$9, producing distinct phase boundaries, distinct critical thresholds, and a bidirectional contrast ratio

$\lambda_-\neq\lambda_+$0

whenever the two steady-state magnon occupations differ (Zhang et al., 30 Sep 2025). In the red-lobe regions of the contrast-ratio map, one configuration is in the superradiant phase while the other remains normal, giving ideal nonreciprocity with $\lambda_-\neq\lambda_+$1 (Zhang et al., 30 Sep 2025).

5. Order of the transition, metastability, hysteresis, and oscillatory phases

Nonreciprocal SQPTs are not limited to a single continuous instability. The recent literature explicitly shows coexistence of first- and second-order onset, bistability, and multicriticality. In the rotating microcavity QED system, the second-order line is obtained from NP instability, while the first-order line follows from the existence of real superradiant solutions at finite amplitude; the two meet at a tricritical point (Zhu et al., 2024). In the spinning magnonic system, continuous and discontinuous transitions respectively separate the parity-symmetric phase from the parity-symmetry-broken phase and from the bistable phase (Xu et al., 2024). In the Kerr-magnon proposal, the phase diagram contains normal, superradiant, bistable, and unstable regions, with the assignment of boundaries interchanged between $\lambda_-\neq\lambda_+$2 and $\lambda_-\neq\lambda_+$3 (Zhang et al., 30 Sep 2025).

The generalized open Dicke model gives the clearest microscopic account of why such structures arise. Between saddle-node and pitchfork bifurcations, the normal state and two parity-related superradiant states are simultaneously stable, so finite-$\lambda_-\neq\lambda_+$4 quantum fluctuations induce switching among them (Stitely et al., 2020). Under slow parameter ramps, this coexistence produces quantum hysteresis loops that differ from the semiclassical limit because tunnelling populates superradiant states before the pitchfork and phase-space ghosts survive briefly beyond the saddle-node (Stitely et al., 2020). These same ingredients—coexistence, metastability, and noise-activated switching—underlie the bistable regions of explicitly nonreciprocal SQPT platforms.

Beyond static order, asymmetric Dicke physics also supports a superradiant oscillatory phase generated by a Hopf bifurcation of a superradiant fixed point (Stitely et al., 2020). The resulting persistent oscillations of $\lambda_-\neq\lambda_+$5, $\lambda_-\neq\lambda_+$6, and $\lambda_-\neq\lambda_+$7 indicate that superradiant criticality in driven–dissipative systems naturally extends to limit-cycle order parameters. A plausible implication is that nonreciprocal SQPT platforms with directional gain, directional squeezing, or sign-sensitive Kerr terms can inherit similar dynamical phases whenever their effective Jacobians acquire complex-conjugate instabilities.

6. Related architectures, controversies, and theoretical status

The broader theoretical landscape shows that nonreciprocal SQPT is part of a larger family of generalized superradiant transitions in which asymmetry, multimode structure, frustration, or engineered drives replace the simplest balanced Dicke paradigm. A three-level $\lambda_-\neq\lambda_+$8 Dicke model coupled to two bosonic modes exhibits both first- and second-order superradiant transitions, but within the physically allowed TRK region the accessible transition is first-order (Hayn et al., 2012). A four-site quantum Rabi square with nearest- and next-nearest-neighbor hopping realizes a second-order transition between normal and superradiant phases and a first-order transition between antiferromagnetic and frustrated superradiant phases; the next-nearest-neighbor hopping is analytically mapped to the effect of an artificial gauge phase in the corresponding Rabi ring (Xu et al., 2024). Although that model remains reciprocal, it shows that frustrated superradiant phases and first-order superradiant-to-superradiant switching need not rely on literal nonreciprocity.

A technical controversy concerns whether Dicke- and Rabi-type superradiant transitions are “quantum” in the fluctuation-driven sense. In the Dicke, Tavis–Cummings, quantum Rabi, and Jaynes–Cummings models, the transition is mean-field in the thermodynamic or classical-oscillator limit, and quantum fluctuations either vanish asymptotically or are strictly absent because different excitation sectors do not mix (Larson et al., 2016). At the same time, these models display a non-analyticity in the ground-state energy at $\lambda_-\neq\lambda_+$9, and in the open Dicke and Rabi cases the steady state remains critical under photon loss (Larson et al., 2016). Nonreciprocal SQPT proposals inherit this ambiguity: they are clearly critical in the sense of non-analytic steady-state or ground-state structure, but their universality is generally mean-field and often driven by semiclassical bifurcation theory.

From the present body of work, several robust conclusions follow. First, explicit nonreciprocity can be implemented through cavity rotation and directional squeezing, or through sign-selective Kerr nonlinearities in cavity magnonics (Zhu et al., 2024). Second, first-order structure, bistability, and multicriticality are not peripheral effects but central organizing principles of realistic nonreciprocal SQPTs (Xu et al., 2024). Third, microscopic consistency requires careful treatment of $$\alpha=\langle \hat a\rangle/\sqrt N,\qquad \beta=\langle \hat J_-\rangle/N,\qquad \gamma=\langle \hat J_z\rangle/N$$0 terms, TRK constraints, and gauge invariance, but multilevel structure, collective interactions, finite-wavevector order, and driven–dissipative engineering all provide documented routes around the simplest no-go scenarios (Liu et al., 2021). Finally, asymmetric generalized Dicke models remain the canonical reduced description of how interaction imbalance plus loss generates the qualitative phenomena—phase coexistence, switching, hysteresis, and oscillatory superradiance—that explicit nonreciprocal platforms later realize in a genuinely directional form (Stitely et al., 2020).