Driven-Dissipative Quantum Rabi Network

• Driven-dissipative quantum Rabi networks are systems of coupled quantum Rabi units where local light-matter interactions experience external driving and various dissipation channels.
• They showcase complex nonequilibrium phenomena including shifted phase transition thresholds, metastability, and the breakdown of standard Lindblad approximations in ultrastrong regimes.
• Experimental platforms like circuit QED validate these models, emphasizing the impact of dressed-basis dissipation and inter-site photon hopping on steady-state behavior.

A driven-dissipative quantum Rabi network is a coupled assembly of quantum Rabi units—cavity or oscillator modes locally hybridized with two-level systems through the full Rabi interaction, including counter-rotating terms—subject to loss, thermalization, or other open-system channels, and often to coherent or parametric driving. In current research, the topic is addressed through both explicit networks, such as dissipative Rabi-Hubbard lattices and Rabi dimers, and a larger body of single-site open-quantum-Rabi studies that determine the local nonequilibrium physics inherited by any array. The central issues are the validity of dressed-basis dissipation in ultrastrong and deep-strong coupling, the structure of dissipative phase transitions, metastability and symmetry restoration, and the way local drive and dissipation compete with inter-site coupling to generate collective steady states or long-lived transient phases (Ye et al., 2021, Huang et al., 2019, Mercurio et al., 13 Mar 2026).

1. Local quantum Rabi element and the meaning of “driven-dissipative”

The local building block is the quantum Rabi model

$$\hat{H}_\mathrm{QRM} = \omega_{\rm c}\hat{a}^\dagger\hat{a} + \frac{\omega_\mathrm{q}}{2}\hat{\sigma}_z + g(\hat{a}+\hat{a}^\dagger)\hat{\sigma}_x,$$

where the interaction contains both rotating terms $\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger$ and counter-rotating terms $\hat{\sigma}_+\hat a^\dagger+\hat{\sigma}_-\hat a$. This is the defining distinction from the Jaynes-Cummings approximation, obtained for $g\ll \omega_{\rm c},\omega_{\rm q}$ by neglecting the counter-rotating contributions, so that

$$\hat{H}_\mathrm{JCM} = \omega_{\rm c}\hat{a}^\dagger\hat{a} + \frac{\omega_\mathrm{q}}{2}\hat{\sigma}_z + g(\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger).$$

In driven open settings, the driven Jaynes-Cummings model has the well-known critical drive $F_c=g/2$, or $\tilde F_c=1$ in the rescaled variable $\tilde F=2F/g$, whereas the full driven quantum Rabi model shifts and eventually destroys that transition as ultrastrong and deep-strong coupling are approached (Mercurio et al., 13 Mar 2026).

“Driven” is realized in several non-equivalent ways across the literature. One formulation applies a periodic drive directly in the dressed basis of the QRM,

$$\hat{H}_\mathrm{QRM}^\mathrm{tot}(t) = \hat{H}_\mathrm{QRM} + F\left(\hat X^+ e^{i\omega_{\rm d} t}+\hat X^- e^{-i\omega_{\rm d} t}\right),$$

with period $T=2\pi/\omega_{\rm d}$ (Mercurio et al., 13 Mar 2026). Another engineers an effective open QRM in a rotating frame,

$\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger$0

using longitudinal modulation and transverse microwave driving in circuit QED (Ning et al., 19 Jun 2025). A further extension adds a static transverse term to realize a generalized or driven Rabi Hamiltonian in cold atoms, while preserving access to ultrastrong, deep-strong, and dispersive deep-strong coupling (Schneeweiss et al., 2017).

“Dissipative” likewise spans several channels. The simplest is single-photon loss,

$\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger$1

or, equivalently, $\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger$2 (Mercurio et al., 13 Mar 2026, Ning et al., 19 Jun 2025). Other works include cavity loss together with spin relaxation, thermal baths, sub-Ohmic environments, or nonlinear two-photon decay. The common feature is that the network nodes are not isolated Rabi systems but local open quantum systems whose steady states and transients are determined by a Liouvillian or Floquet-Liouville dynamics rather than by unitary spectroscopy alone.

2. From single nodes to dimers and lattices

The most explicit network realization in the supplied literature is the dissipative Rabi-Hubbard lattice,

$\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger$3

with on-site Hamiltonian

$\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger$4

After single-site mean-field decoupling, this becomes

$\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger$5

with order parameter $\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger$6. Here the network is a lattice of hybrid light-matter sites coupled by photon hopping, and the delocalization transition is interpreted as a photonic localization-delocalization transition or, equivalently, a normal-to-coherent ordering transition (Ye et al., 2021).

The minimal network element is the Rabi dimer,

$\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger$7

with local driven Rabi subsystems

$\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger$8

This system already exhibits the essential network ingredients: local Rabi hybridization, inter-resonator hopping, site-selective driving, and a structured environment coupled through the qubits rather than directly through the photons (Huang et al., 2019).

A broader, explicitly identified network extension is the natural lattice model

$\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger$9

with local loss

$\hat{\sigma}_+\hat a^\dagger+\hat{\sigma}_-\hat a$0

This expression is presented as the natural continuation of the experimentally realized single-site open QRM in circuit QED, where each site carries tunable $\hat{\sigma}_+\hat a^\dagger+\hat{\sigma}_-\hat a$1, $\hat{\sigma}_+\hat a^\dagger+\hat{\sigma}_-\hat a$2, and $\hat{\sigma}_+\hat a^\dagger+\hat{\sigma}_-\hat a$3, and inter-cavity hopping $\hat{\sigma}_+\hat a^\dagger+\hat{\sigma}_-\hat a$4 would supply the genuinely many-body component (Ning et al., 19 Jun 2025).

Not all related models are spatial networks in the literal sense. The generalized Jaynes-Cummings–Rabi model with $\hat{\sigma}_+\hat a^\dagger+\hat{\sigma}_-\hat a$5 two-state systems collectively coupled to one cavity mode,

$\hat{\sigma}_+\hat a^\dagger+\hat{\sigma}_-\hat a$6

is collective rather than lattice-resolved, but it remains directly relevant because it determines the local driven-open Rabi physics that would enter mean-field or mode-based treatments of a network (Gutierrez-Jauregui et al., 2018). Similarly, the cold-atom implementation based on an optical lattice produces an array of independent on-site QRMs; controlled tunneling is not developed there, but the platform already supplies the site-resolved QRM architecture from which a network could plausibly be built (Schneeweiss et al., 2017).

3. Open-system structure, dressed-basis dissipation, and observables

A central technical distinction in driven-dissipative quantum Rabi networks is whether the local light-matter coupling remains weak enough for bare-basis Lindblad operators to be reliable. In ultrastrong and deep-strong coupling, several works emphasize that they are not. For periodic open systems, the density matrix evolves as

$\hat{\sigma}_+\hat a^\dagger+\hat{\sigma}_-\hat a$7

with $\hat{\sigma}_+\hat a^\dagger+\hat{\sigma}_-\hat a$8. In the driven QRM, the appropriate dissipator is formulated in the dressed basis, with an Ohmic environment $\hat{\sigma}_+\hat a^\dagger+\hat{\sigma}_-\hat a$9, so that the weak-coupling Lindblad form with $g\ll \omega_{\rm c},\omega_{\rm q}$0 is recovered only as a limit (Mercurio et al., 13 Mar 2026).

The same principle appears in the dissipative Rabi-Hubbard lattice. There, each mean-field site couples to qubit and cavity baths through $g\ll \omega_{\rm c},\omega_{\rm q}$1 and $g\ll \omega_{\rm c},\omega_{\rm q}$2, but the long-time dynamics is reduced to a dressed master equation in the eigenbasis $g\ll \omega_{\rm c},\omega_{\rm q}$3 of $g\ll \omega_{\rm c},\omega_{\rm q}$4: $g\ll \omega_{\rm c},\omega_{\rm q}$5 with rates determined by dressed matrix elements of $g\ll \omega_{\rm c},\omega_{\rm q}$6 and $g\ll \omega_{\rm c},\omega_{\rm q}$7. This dressed-basis treatment is explicitly contrasted with the standard local Lindblad equation, which overestimates dissipation at strong coupling and can even alter the phase boundary qualitatively (Ye et al., 2021).

A related refinement concerns observables. In ultrastrong coupling, the physically meaningful output is not the bare photon number but a dressed positive-frequency field operator. For the driven QRM,

$g\ll \omega_{\rm c},\omega_{\rm q}$8

is used instead of $g\ll \omega_{\rm c},\omega_{\rm q}$9, because the latter may differ significantly from the output field intensity once counter-rotating hybridization becomes important (Mercurio et al., 13 Mar 2026). This lesson is echoed by open-anisotropic and parametrically amplified Rabi models, where Wigner functions, photon-number fluctuations, and branch-resolved observables become the appropriate steady-state diagnostics in place of simple bare occupations (Wang et al., 14 Jan 2026, Zhu et al., 18 Mar 2026).

A realistic driven-dissipative quantum Rabi network therefore requires two simultaneous renormalizations: the local dissipation must be expressed in the dressed basis of each strongly hybridized node, and the local observables must be recast in terms of dressed output operators or dressed transitions rather than naive subsystem operators. A plausible implication is that network-level transport, ordering, and criticality depend sensitively on the local definition of both loss and measurement.

4. Dissipative phase transitions and critical structure

The most general formulation of dissipative phase transitions in time-periodic open quantum systems is given in terms of the one-period Floquet propagator,

$$\hat{H}_\mathrm{JCM} = \omega_{\rm c}\hat{a}^\dagger\hat{a} + \frac{\omega_\mathrm{q}}{2}\hat{\sigma}_z + g(\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger).$$0

rather than an assumed time-independent Floquet Lindbladian. Its eigenproblem,

$$\hat{H}_\mathrm{JCM} = \omega_{\rm c}\hat{a}^\dagger\hat{a} + \frac{\omega_\mathrm{q}}{2}\hat{\sigma}_z + g(\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger).$$1

defines the periodic steady state and the leading relaxation modes. A Floquet dissipative phase transition is then identified by nonanalyticity of the period-averaged steady-state observable and spectrally by the leading nontrivial Floquet eigenvalue approaching the unit circle. This provides the appropriate diagnostic framework for periodically driven Rabi arrays that cannot be reduced exactly to static frames (Mercurio et al., 13 Mar 2026).

Within that framework, the driven Jaynes-Cummings model shows the expected second-order transition at $$\hat{H}_\mathrm{JCM} = \omega_{\rm c}\hat{a}^\dagger\hat{a} + \frac{\omega_\mathrm{q}}{2}\hat{\sigma}_z + g(\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger).$$2, while the driven QRM deviates from it in two ways. First, as $$\hat{H}_\mathrm{JCM} = \omega_{\rm c}\hat{a}^\dagger\hat{a} + \frac{\omega_\mathrm{q}}{2}\hat{\sigma}_z + g(\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger).$$3 approaches ultrastrong coupling, the critical line is shifted toward smaller $$\hat{H}_\mathrm{JCM} = \omega_{\rm c}\hat{a}^\dagger\hat{a} + \frac{\omega_\mathrm{q}}{2}\hat{\sigma}_z + g(\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger).$$4; the counter-rotating light-matter terms therefore lower the effective critical drive relative to the Jaynes-Cummings prediction $$\hat{H}_\mathrm{JCM} = \omega_{\rm c}\hat{a}^\dagger\hat{a} + \frac{\omega_\mathrm{q}}{2}\hat{\sigma}_z + g(\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger).$$5. Second, in deep-strong coupling the dissipative phase transition disappears altogether because of light-matter decoupling, and the response reduces to that of an effectively uncoupled driven harmonic oscillator, with output $$\hat{H}_\mathrm{JCM} = \omega_{\rm c}\hat{a}^\dagger\hat{a} + \frac{\omega_\mathrm{q}}{2}\hat{\sigma}_z + g(\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger).$$6 at fixed $$\hat{H}_\mathrm{JCM} = \omega_{\rm c}\hat{a}^\dagger\hat{a} + \frac{\omega_\mathrm{q}}{2}\hat{\sigma}_z + g(\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger).$$7 (Mercurio et al., 13 Mar 2026). For network theory, this establishes that replacing local Jaynes-Cummings nodes by true Rabi nodes is not a perturbative correction: the underlying local nonequilibrium mechanism can be reshaped or extinguished.

In the dissipative Rabi-Hubbard lattice, the key critical variable is the hopping $$\hat{H}_\mathrm{JCM} = \omega_{\rm c}\hat{a}^\dagger\hat{a} + \frac{\omega_\mathrm{q}}{2}\hat{\sigma}_z + g(\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger).$$8. At zero temperature, in the deep-strong coupling regime, the dressed-master-equation analysis gives

$$\hat{H}_\mathrm{JCM} = \omega_{\rm c}\hat{a}^\dagger\hat{a} + \frac{\omega_\mathrm{q}}{2}\hat{\sigma}_z + g(\hat{\sigma}_+\hat a+\hat{\sigma}_-\hat a^\dagger).$$9

with $F_c=g/2$0. Hence $F_c=g/2$1 as $F_c=g/2$2, even in the presence of dissipation. This explicitly contradicts the earlier finite minimum critical hopping obtained from the standard Lindblad treatment and identifies dressed-basis dissipation as the decisive ingredient in the strong-coupling phase diagram (Ye et al., 2021).

Two further single-node generalizations show how rich the local critical structure can become before any lattice coupling is added. In the parametrically amplified open QRM with both one- and two-photon decay,

$F_c=g/2$3

the classical-oscillator limit yields four composite phases, an inverted regime in which superradiance occurs only for sufficiently small $F_c=g/2$4, and first- and second-order dissipative phase transitions separated by a tricritical point; two-photon dissipation stabilizes the superradiant phase and controls the tricritical structure (Zhu et al., 18 Mar 2026). In the anisotropic open QRM with explicit $F_c=g/2$5 term,

$F_c=g/2$6

anisotropy generically overcomes the no-go theorem that forbids the superradiant instability on the isotropic line. The resulting steady-state phase diagram contains normal, superradiant, and bistable phases, tricritical points, and isolated bistable regions, with anomalous photon-fluctuation scaling $F_c=g/2$7 appearing near the intersection of the two critical-line branches rather than at the tricritical points (Wang et al., 14 Jan 2026).

Taken together, these results imply that a driven-dissipative quantum Rabi network is not controlled by a single “superradiant threshold.” Depending on local drive, anisotropy, nonlinear decay, and coupling regime, the appropriate local ingredients may be second-order criticality, first-order bistability, tricriticality, or complete loss of the local transition mechanism. A plausible implication is that many-body phase diagrams of true Rabi lattices should be expected to inherit this heterogeneity rather than resemble a simple Dicke or Jaynes-Cummings network.

5. Metastability, symmetry restoration, and long-lived nonequilibrium states

Metastability is a recurrent feature of the driven-dissipative QRM and is likely to be central in any network generalization. In the periodically driven dissipative Rabi model, a Floquet-Liouville analysis shows that once $F_c=g/2$8 becomes comparable to or larger than $F_c=g/2$9, a new timescale emerges that can exceed the natural relaxation time by orders of magnitude. The density matrix assumes the long-time form

$\tilde F_c=1$0

where $\tilde F_c=1$1 is the periodic steady state and $\tilde F_c=1$2 the dominant slow mode. The associated metastable states can differ strongly from the final steady state: one extremal state displays enhanced antibunching and another strong bunching, while the true periodic steady state lies between them (Boité et al., 2016).

The physical mechanism there is a parity-structured bottleneck in dressed-state relaxation. After the parity shift in the Rabi spectrum, the driven transition $\tilde F_c=1$3 can decay either directly or through a cascaded channel. In ultrastrong coupling, the first and third steps of the cascaded path become very slow, generating a tiny Floquet-Liouville gap and hence a long-lived metastable manifold. For network-oriented thinking, this is important because it ties slow relaxation not to weak bare losses but to the dressed spectral fine structure of the local node.

A complementary mechanism appears in the dissipative QRM with cavity loss and weak spin relaxation. There, the nominal superradiant phase becomes a superradiant metastable phase: symmetry-breaking states are stable only for a finite time because each spin-jump event acts as a strong perturbation that can drive the system from a symmetry-breaking state to the parity-preserving saddle point before it relaxes back to a broken-symmetry state. The exact steady state therefore restores parity, even though mean-field predicts two symmetry-breaking branches. The metastable lifetime is obtained from the Liouvillian gap,

$\tilde F_c=1$4

and remains finite in the thermodynamic limit whenever $\tilde F_c=1$5 (Xiao et al., 26 Nov 2025).

These two mechanisms—Floquet bottlenecks and jump-induced restoration—establish a general warning for driven-dissipative quantum Rabi networks. Apparent ordered or symmetry-broken states can be long-lived but not truly stationary. A plausible implication is that network-level phases inferred from mean-field order parameters or finite-time simulations must be checked against Liouvillian spectral gaps, switching pathways, and the possibility of symmetry-restored exact steady states.

6. Experimental platforms, measurements, and computational methods

The clearest experimental realization of an open QRM in the supplied literature is the circuit-QED implementation based on an Xmon qubit coupled to a lossy resonator. The effective rotating-frame Hamiltonian

$\tilde F_c=1$6

is engineered through longitudinal modulation and transverse driving, while the open-system dynamics is modeled as

$\tilde F_c=1$7

The experiment reconstructs photon-number distributions $\tilde F_c=1$8, tracks the transient build-up of the average photon number, and shows that stronger coupling populates higher photon-number states more strongly even in the presence of substantial loss. For network modeling, this establishes the experimentally benchmarked single-site Liouvillian from which a driven-dissipative Rabi array would be constructed (Ning et al., 19 Jun 2025).

Another directly relevant diagnostic is qubit transmission spectroscopy in the driven, dissipative ultrastrong-coupling Rabi model. There the local Hamiltonian includes a strong drive and a weak probe, both subsystems are lossy, and the measured observable is the qubit response rather than the cavity field. The resulting spectra exhibit avoided crossings, oscillator sidebands, and strong-drive Floquet structures governed by Laguerre and Bessel dressing factors. This makes clear that, in a network, probing qubits and probing resonators need not access the same sector of the hybridized spectrum (Magazzù et al., 2021).

Critical sensing constitutes a further application of the open QRM. In quantum thermometry with a dissipative quantum Rabi system, the cavity becomes a transducer of the spin-bath temperature, and the optimal precision occurs at the normal-to-superradiant critical point rather than at the exceptional point of the effective anti-$\tilde F_c=1$9-symmetric cavity description. Near the critical point, direct photon detection becomes an excellent proxy for the optimal quantum measurement. A plausible network implication is that thermometry or parameter estimation in a Rabi lattice should be organized around the soft collective mode rather than around isolated non-Hermitian degeneracies (Xie et al., 2021).

Theoretical and numerical methods for this subject are correspondingly diverse. Exact or near-exact single-site dynamics beyond weak coupling have been treated using an influence-functional stochastic Schrödinger equation, which incorporates coherent drive, non-Markovian Ohmic spin dissipation, and cavity leakage within a unified local-node description (Henriet et al., 2014). In the dispersive regime of the full dissipative Rabi model, a Keldysh-contour hybrid perturbation theory yields compact analytic expressions for Bloch-Siegert shifts, Purcell rates, dressed dephasing, and photon-assisted dephasing without invoking the rotating-wave approximation (Müller, 2020). For periodically driven open systems, practical Floquet-propagator diagonalization has been implemented with QuantumToolbox.jl and the Arnoldi-Lindblad method (Mercurio et al., 13 Mar 2026). At the machine-learning level, driven neural quantum propagators learn Liouville-space propagators rather than state trajectories, but their current demonstrations remain confined to low-dimensional systems and the approach inherits exponential scaling with system size, so applicability to a realistic multi-site quantum Rabi network is only approximate in principle (Zhang et al., 2024).

The present state of the subject is therefore structurally asymmetric. Explicit driven-dissipative quantum Rabi networks are represented by dimers and mean-field lattices, while a larger literature determines the local open-QRM physics—dressed dissipation, criticality, metastability, nonlinear decay, anisotropy, and parametric driving—that any full network theory must incorporate. A plausible implication is that the next decisive advances will require combining those local ingredients with genuine many-site methods capable of resolving collective Liouvillian spectra, spatial correlations, and finite-frequency Floquet criticality in arrays of strongly hybridized Rabi nodes.