Nonreciprocal Phase Transition in Many-Body Systems

• Nonreciprocal phase transition is defined by asymmetric interactions that break detailed balance and convert static order into dynamic states like oscillations or traveling waves.
• The phenomenon is analyzed through non-Hermitian dynamics, exceptional points, and bifurcation theory, revealing transitions controlled by eigenvalue softening and eigenvector coalescence.
• Applications span classical statistical mechanics, active matter, and open quantum systems, impacting universality classes and spectral transitions with boundary-sensitive effects.

Nonreciprocal phase transition denotes a qualitative change in collective behavior driven by asymmetric interactions, typically encoded by couplings with $G_{12}\neq G_{21}$ or, equivalently, by a nonvanishing antisymmetric component $G^{(A)}=(G-G^\top)/2$. Because such dynamics generally cannot be reduced to a single potential, they commonly violate detailed balance, generate non-Hermitian or non-normal linearized operators, and replace purely static ordering transitions by transitions toward limit cycles, traveling waves, time crystals, oscillatory amorphous states, or direction-dependent spectral phases (Fruchart et al., 2020, Lorenzana et al., 22 Sep 2025). Across classical statistical mechanics, active matter, non-Hermitian lattices, and open quantum systems, the central question is whether the antisymmetric part of the coupling merely shifts thresholds or instead reorganizes the attractor structure, the symmetry-breaking pattern, or the universality class itself (Avni et al., 2023, Chiacchio et al., 2023).

1. Definition and conceptual scope

In coupled dynamical systems, nonreciprocity means that the influence of one subsystem on another differs from the reverse influence. A convenient decomposition writes

$$G = G^{(S)} + G^{(A)},\qquad G^{(S)}_{ij}=\tfrac12(G_{ij}+G_{ji}),\qquad G^{(A)}_{ij}=\tfrac12(G_{ij}-G_{ji}),$$

and for two fields the scalar parameter

$\varepsilon=\tfrac12(G_{12}-G_{21})$

measures the antisymmetric part directly. Within the perturbative framework of coupled Langevin-type equations, the nonreciprocal perturbation is represented in the MSRJD action by

$$\delta S=\varepsilon\int dx\,dt\,(\hat\phi_1\phi_2-\hat\phi_2\phi_1),$$

which identifies nonreciprocity as a specific operator rather than a vague departure from equilibrium (Lorenzana et al., 22 Sep 2025).

This asymmetry has several nonequilibrium consequences. In the two-species nonreciprocal Ising model, each spin flips according to its own “selfish energy,” so the dynamics cannot be derived from a single global energy; detailed balance is broken, and the broken reciprocity appears as non-Hermitian linearized dynamics and finite entropy production in oscillatory states (Avni et al., 2023). In reservoir-engineered fermionic chains, nonreciprocity is operationally defined by the asymmetry of response or steady-state observables under spatial inversion, and it manifests simultaneously in deterministic drift and dissipative noise (Soares et al., 21 May 2025).

Accordingly, “nonreciprocal phase transition” has a broader meaning than an ordinary equilibrium transition with asymmetric couplings. It can refer to a transition between static and time-dependent many-body phases, a transition between reciprocal and nonreciprocal dissipative transport regimes, or a direction-dependent non-Hermitian spectral transition. What unifies these cases is that the antisymmetric coupling changes the long-time phase or the phase boundary itself.

2. Spectral and bifurcation mechanisms

A central theoretical theme is the appearance of exceptional points, Hopf bifurcations, and related non-Hermitian singularities. In the general framework for nonreciprocal matter, spontaneous breaking of a continuous symmetry provides a Goldstone mode at zero eigenvalue; nonreciprocity can then force the Goldstone mode to coalesce with a damped mode at an exceptional point, producing a defective Jordan block and converting a static ordered phase into a time-dependent one (Fruchart et al., 2020). In this sense, the phase transition is controlled not only by eigenvalue softening but also by eigenvector coalescence.

The multipopulation $O(2)$-symmetric theory extends this picture beyond two coupled sectors. There, nonreciprocal transitions are classified by the topology of attractors on tori and by the structure of the non-Hermitian Jacobian. The catalog includes critical exceptional points at chiral onset, Hopf bifurcations, limit-cycle saddle-node bifurcations, and symmetric homoclinic orbit bifurcations that dynamically restore $\mathbb{Z}_2$ symmetry; for $N\geq 4$, the homoclinic route can generate Shilnikov chaos (Weis et al., 22 Jul 2025).

A canonical mean-field example is supplied by the nonreciprocal Ising model. For homogeneous magnetizations,

$$\partial_t m = J_{\mathrm{MF}}m,\qquad J_{\mathrm{MF}}=(\tilde J-1)I+\tilde K\,\epsilon,$$

with eigenvalues

$\lambda_\pm=(\tilde J-1)\pm i\tilde K.$

The disordered fixed point undergoes a Hopf bifurcation at $G^{(A)}=(G-G^\top)/2$0, with oscillation frequency $G^{(A)}=(G-G^\top)/2$1 at onset. Beyond the linear regime, the homogeneous oscillatory “swap” phase is bounded at larger $G^{(A)}=(G-G^\top)/2$2 by a saddle-node-on-invariant-circle bifurcation, so the mean-field transition sequence already mixes Hopf and global bifurcation mechanisms (Avni et al., 2023).

Exceptional points are nevertheless not universal. In the non-reciprocal Dicke model, adiabatic elimination of the cavity produces EPs at the normal-to-dynamical boundary, but the full light-matter dynamics retains the nonreciprocal phase transition even when the dynamical mediator smooths away the EP singularities. The transition then appears as spontaneous $G^{(A)}=(G-G^\top)/2$3-symmetry breaking without requiring exceptional points in the full spectrum (Chiacchio et al., 2023). A nonreciprocal phase transition should therefore not be identified too narrowly with EP physics, even though EPs organize a large class of examples.

3. Classical statistical and continuum realizations

The two-species nonreciprocal Ising model provides the clearest statistical-mechanical realization. Mean field predicts disordered, static ordered, and oscillatory swap phases, but fluctuations drastically revise this picture. In two dimensions, the swap phase is destroyed by the proliferation of spiral defects, and static order is destabilized by rare droplet growth. In three dimensions, by contrast, the swap phase is stable in a finite part of parameter space, the coherence time scales as $G^{(A)}=(G-G^\top)/2$4, and finite-size scaling at $G^{(A)}=(G-G^\top)/2$5 gives $G^{(A)}=(G-G^\top)/2$6, $G^{(A)}=(G-G^\top)/2$7, and $G^{(A)}=(G-G^\top)/2$8, in agreement with 3D XY rather than 3D Ising criticality. Static order is absent in the thermodynamic limit for fully antisymmetric interspecies couplings because droplets of the metastable species eventually grow, but asymmetric couplings can restore static order through a droplet-capture mechanism (Avni et al., 2023).

A distinct route appears in the single-species Ising model with state-dependent “vision-cone” couplings. There, nonreciprocity can break spin-flip symmetry directly. In the $G^{(A)}=(G-G^\top)/2$9 case, it shifts the continuous PM–FM transition temperature; for $$G = G^{(S)} + G^{(A)},\qquad G^{(S)}_{ij}=\tfrac12(G_{ij}+G_{ji}),\qquad G^{(A)}_{ij}=\tfrac12(G_{ij}-G_{ji}),$$0, it generates a cubic term in the Landau-like free-energy expansion and adds a discontinuous transition with hysteresis and metastability. Static and dynamic scaling in two dimensions give exponents compatible with the Ising values for $$G = G^{(S)} + G^{(A)},\qquad G^{(S)}_{ij}=\tfrac12(G_{ij}+G_{ji}),\qquad G^{(A)}_{ij}=\tfrac12(G_{ij}-G_{ji}),$$1, $$G = G^{(S)} + G^{(A)},\qquad G^{(S)}_{ij}=\tfrac12(G_{ij}+G_{ji}),\qquad G^{(A)}_{ij}=\tfrac12(G_{ij}-G_{ji}),$$2, and $$G = G^{(S)} + G^{(A)},\qquad G^{(S)}_{ij}=\tfrac12(G_{ij}+G_{ji}),\qquad G^{(A)}_{ij}=\tfrac12(G_{ij}-G_{ji}),$$3, but the order-parameter exponent $$G = G^{(S)} + G^{(A)},\qquad G^{(S)}_{ij}=\tfrac12(G_{ij}+G_{ji}),\qquad G^{(A)}_{ij}=\tfrac12(G_{ij}-G_{ji}),$$4 increases with nonreciprocity, which suggests that, within numerical precision, the model does not belong to the 2D Ising universality class (Garcés et al., 2024).

Driven-dissipative bosonic condensates reveal a complementary boundary-sensitive scenario. In a one-dimensional lattice with coherent hopping, incoherent pump, two-particle loss, and correlated loss $$G = G^{(S)} + G^{(A)},\qquad G^{(S)}_{ij}=\tfrac12(G_{ij}+G_{ji}),\qquad G^{(A)}_{ij}=\tfrac12(G_{ij}-G_{ji}),$$5, periodic boundaries always select a finite-momentum traveling-wave condensate. Open boundaries instead generate a much richer phase diagram: vacuum windows, static edge-kink condensates, dynamical phases with spontaneous particle-hole symmetry breaking, and a critical exceptional point where a second zero mode coalesces with the $$G = G^{(S)} + G^{(A)},\qquad G^{(S)}_{ij}=\tfrac12(G_{ij}+G_{ji}),\qquad G^{(A)}_{ij}=\tfrac12(G_{ij}-G_{ji}),$$6 Goldstone mode (Belyansky et al., 7 Feb 2025).

Nonreciprocity also reshapes glass transitions. In the bipartite spherical Sherrington–Kirkpatrick model with antisymmetric inter-population coupling $$G = G^{(S)} + G^{(A)},\qquad G^{(S)}_{ij}=\tfrac12(G_{ij}+G_{ji}),\qquad G^{(A)}_{ij}=\tfrac12(G_{ij}-G_{ji}),$$7, dynamical mean-field theory yields an exceptional-point-mediated transition from a static disorder phase to an oscillating amorphous phase. The transition remains at $$G = G^{(S)} + G^{(A)},\qquad G^{(S)}_{ij}=\tfrac12(G_{ij}+G_{ji}),\qquad G^{(A)}_{ij}=\tfrac12(G_{ij}-G_{ji}),$$8, the non-reciprocal Edwards–Anderson parameter is

$$G = G^{(S)} + G^{(A)},\qquad G^{(S)}_{ij}=\tfrac12(G_{ij}+G_{ji}),\qquad G^{(A)}_{ij}=\tfrac12(G_{ij}-G_{ji}),$$9

and the aging correlator acquires oscillations,

$\varepsilon=\tfrac12(G_{12}-G_{21})$0

showing that glassiness and aging can survive nonreciprocity in an oscillatory form rather than being destroyed outright (Lorenzana et al., 2024).

4. Open quantum and non-Hermitian realizations

In open quantum many-body systems, nonreciprocal phase transitions arise naturally from chiral reservoirs, cavity-mediated couplings, and engineered dissipation. For active quantum spins coupled through chiral waveguides, the directional couplings

$\varepsilon=\tfrac12(G_{12}-G_{21})$1

become antagonistic when $\varepsilon=\tfrac12(G_{12}-G_{21})$2. In the thermodynamic limit this produces a Hopf bifurcation to a traveling-wave time-crystalline state with spontaneous $\varepsilon=\tfrac12(G_{12}-G_{21})$3-symmetry breaking; the onset frequency is

$\varepsilon=\tfrac12(G_{12}-G_{21})$4

and continuous monitoring can reveal symmetry-broken quantum trajectories even at finite $\varepsilon=\tfrac12(G_{12}-G_{21})$5 (Nadolny et al., 2024).

The non-reciprocal Dicke model supplies another archetype. A lossy cavity mediates asymmetric interactions between two collective spin species, giving an effective interaction matrix with $\varepsilon=\tfrac12(G_{12}-G_{21})$6. The model has a discrete $\varepsilon=\tfrac12(G_{12}-G_{21})$7 symmetry, and its transition from a stationary normal phase to a dynamical phase is interpreted as spontaneous $\varepsilon=\tfrac12(G_{12}-G_{21})$8-symmetry breaking. Remarkably, this nonreciprocal phase transition survives in the full light-matter dynamics without requiring exceptional points, even though EPs appear in the adiabatically eliminated description (Chiacchio et al., 2023).

Reservoir engineering can also produce dissipative nonreciprocal transitions in fermionic chains. With jump operators

$\varepsilon=\tfrac12(G_{12}-G_{21})$9

the noninteracting system is gapless on the line $$\delta S=\varepsilon\int dx\,dt\,(\hat\phi_1\phi_2-\hat\phi_2\phi_1),$$0, where density and current relax as $$\delta S=\varepsilon\int dx\,dt\,(\hat\phi_1\phi_2-\hat\phi_2\phi_1),$$1. Weak interactions preserve these signatures and support volume-law trajectory entanglement, while stronger interactions open a many-body dissipative gap, restore reciprocity dynamically, and produce exponential relaxation; for $$\delta S=\varepsilon\int dx\,dt\,(\hat\phi_1\phi_2-\hat\phi_2\phi_1),$$2, the crossover occurs at $$\delta S=\varepsilon\int dx\,dt\,(\hat\phi_1\phi_2-\hat\phi_2\phi_1),$$3 (Soares et al., 21 May 2025).

Magnonic platforms realize direction-dependent quantum phase boundaries in two distinct ways. In a spinning microwave magnonic system, the Sagnac–Fizeau shift

$$\delta S=\varepsilon\int dx\,dt\,(\hat\phi_1\phi_2-\hat\phi_2\phi_1),$$4

changes the effective detuning differently for clockwise and counterclockwise driving, so the first- and second-order thresholds $$\delta S=\varepsilon\int dx\,dt\,(\hat\phi_1\phi_2-\hat\phi_2\phi_1),$$5 and $$\delta S=\varepsilon\int dx\,dt\,(\hat\phi_1\phi_2-\hat\phi_2\phi_1),$$6 split and a quantum phase transition can occur in one direction but not the other (Xu et al., 2024). In cavity magnonics with magnon Kerr effect, the sign of the Kerr coefficient $$\delta S=\varepsilon\int dx\,dt\,(\hat\phi_1\phi_2-\hat\phi_2\phi_1),$$7 depends on the crystallographic orientation of the YIG sphere, and the steady-state phase diagrams for $$\delta S=\varepsilon\int dx\,dt\,(\hat\phi_1\phi_2-\hat\phi_2\phi_1),$$8 and $$\delta S=\varepsilon\int dx\,dt\,(\hat\phi_1\phi_2-\hat\phi_2\phi_1),$$9 differ markedly; the nonreciprocity is quantified by the bidirectional contrast ratio $O(2)$0 (Zhang et al., 30 Sep 2025).

A more explicitly non-Hermitian version appears in chiral cavity QED. There, directional dissipation gives different atomic decay rates for the CW and CCW sectors, so the two effective $O(2)$1 non-Hermitian Hamiltonians have different exceptional-point thresholds

$O(2)$2

The same device can therefore be $O(2)$3-unbroken in one direction and $O(2)$4-broken in the other, and the nonreciprocal phase region also exhibits nonreciprocal photon blockade (Cai et al., 2024). In non-Hermitian lattice models, the term can refer more narrowly to spectral and skin-effect transitions: under open boundaries, spectra may pass through real–complex–imaginary or real–complex regimes, and the non-Hermitian skin effect can reverse direction when point gaps close and reopen (Zeng et al., 2022).

5. Universality and relevance of nonreciprocity

A major theoretical question is when a small antisymmetric coupling changes the universality class. The perturbative criterion developed for pairs of asymmetrically coupled fields states that, for identical uncoupled subsystems, the first-order correction scales as

$O(2)$5

Uniform nonreciprocity is therefore relevant whenever $O(2)$6, equivalently when the tree-level scaling exponent

$O(2)$7

is positive. For random antisymmetric perturbations of identical uncoupled fields, relevance is controlled by

$O(2)$8

whereas for randomness around an already nonreciprocal swap transition the standard Harris condition $O(2)$9 is recovered (Lorenzana et al., 22 Sep 2025).

These criteria match explicit many-body examples. In the 3D nonreciprocal Ising model with fully antisymmetric interspecies couplings, the disorder-to-swap transition is not Ising-like: it is continuous at moderate $\mathbb{Z}_2$0 and falls in the 3D XY class, reflecting the emergence of a phase degree of freedom associated with oscillatory order (Avni et al., 2023). By contrast, the same relevance analysis shows that if a reciprocal coupling $\mathbb{Z}_2$1 already mixes the fields in the unperturbed problem, the antisymmetric perturbation can be irrelevant and the asymptotic Ising universality class can survive (Lorenzana et al., 22 Sep 2025).

Not all departures from reciprocity lead to a complete universality-class replacement. In the 2D single-species vision-cone Ising model, $\mathbb{Z}_2$2, $\mathbb{Z}_2$3, and $\mathbb{Z}_2$4 remain close to standard Ising/Model-A values while $\mathbb{Z}_2$5 drifts upward with increasing nonreciprocity, indicating a partial but systematic deviation rather than a clean crossover to a well-established new class (Garcés et al., 2024). In dissipative interacting fermions, the existence of a continuous transition with gap opening is supported numerically, but critical exponents were not extracted, so the universality class remains open (Soares et al., 21 May 2025).

6. Diagnostics, boundary sensitivity, and unresolved issues

Because nonreciprocal phases are often dynamical, their identification requires observables beyond static order parameters. In the nonreciprocal Ising model, the synchronization measure

$\mathbb{Z}_2$6

and the phase-space angular momentum

$\mathbb{Z}_2$7

distinguish disorder, static order, and swap phases. In continuum active matter, irreversibility across a nonreciprocal $\mathbb{Z}_2$8-breaking transition is quantified by

$\mathbb{Z}_2$9

and the motion-induced contribution is controlled by the global polar order parameter $N\geq 4$0 (Avni et al., 2023, Alston et al., 2023). In fermionic and magnonic platforms, the relevant diagnostics are steady-state current $N\geq 4$1, momentum-space asymmetry $N\geq 4$2, edge accumulation $N\geq 4$3, bidirectional contrast ratio $N\geq 4$4, and isolation parameter $N\geq 4$5 (Soares et al., 21 May 2025, Zhang et al., 30 Sep 2025, Xu et al., 2024).

Boundary conditions are often decisive rather than incidental. In driven-dissipative condensates, periodic boundaries always yield a traveling-wave condensate, while open boundaries can stabilize vacuum phases, static edge-kink condensates, and critical-exceptional-point-controlled dynamical phases with distinct bulk and edge behavior (Belyansky et al., 7 Feb 2025). In non-Hermitian lattice problems, the open-boundary spectrum can become entirely real or imaginary, and skin-effect reversal occurs precisely at analytic boundaries where point gaps close and reopen (Zeng et al., 2022). This boundary sensitivity is one reason the phrase “nonreciprocal phase transition” spans both bulk critical phenomena and spectral transitions tied to open geometry.

Several common simplifications are therefore misleading. Nonreciprocity does not generically destroy order: it can stabilize 3D time-crystalline swap phases and oscillatory glass phases with aging (Lorenzana et al., 2024). Nor are exceptional points invariably required: they organize many classical and non-Hermitian transitions, but the full non-reciprocal Dicke model shows that spontaneous $N\geq 4$6-breaking transitions can persist without EPs in the full spectrum (Chiacchio et al., 2023). The main unresolved directions concern higher-dimensional and spatially extended generalizations, stronger disorder, non-Markovian reservoirs, long-range couplings, and the systematic classification of chaotic and homoclinic nonreciprocal transitions in multipopulation systems (Weis et al., 22 Jul 2025).

Nonreciprocal phase transition has thus become an umbrella concept for symmetry-changing and attractor-changing phenomena driven by antisymmetric couplings. Its modern usage covers Hopf and homoclinic transitions in classical many-body systems, universality-class changes such as Ising-to-XY crossover, dissipative transitions with dynamical reciprocity restoration, $N\geq 4$7- and particle-hole-symmetry breaking, and spectral/skin-effect transitions under open boundaries. The common thread is that the antisymmetric part of the interaction is not a small perturbation to equilibrium ordering, but a structural ingredient that can reorganize both the phase diagram and the meaning of order itself.