Non-Reciprocal Phase Transition

• Non-reciprocal phase transitions are nonequilibrium phenomena driven by asymmetric interactions that break detailed balance and lead to dynamic order.
• They are characterized by non-Hermitian spectral features, exceptional points, and direction-dependent critical thresholds across theoretical and experimental systems.
• Applications span chiral magnets, active matter, and driven-dissipative quantum systems, offering insights into dynamic symmetry restoration and unconventional phase behavior.

Searching arXiv for the cited paper and closely related work on non-reciprocal phase transitions to ground the article. {"^^^^^^^^2query2 OR \2^^^^^^^^"Non-reciprocal spin excitations across the skyrmion-paramagnetic phase transition in MnSi\"","max_results":5} {"^^^^^^^^2query2 phase transitions\" (&&&^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^&&&) OR (Garcés et al., 2024) OR (Soares et al., 21 May 2025)","max_results":^^^^2(Weber et al., 6 Feb 2026) OR \2query2^^^^} Non-reciprocal phase transition denotes a class of nonequilibrium phase transitions in which asymmetric couplings—typically written as PRESERVED_PLACEHOLDER_^^^^2query2^^^^ or PRESERVED_PLACEHOLDER_^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^—alter the existence, stability, and critical structure of macroscopic phases. In contrast to reciprocal transitions governed by a symmetric interaction matrix or an equilibrium free-energy functional, non-reciprocal dynamics generically break detailed balance, produce non-Hermitian or non-normal linearized generators, and can lead to time-dependent ordered states, parity-time-symmetry breaking, discontinuous jumps superposed on continuous transitions, absorbing-to-chaotic transitions, or the persistence of directional response across a conventional thermodynamic boundary (&&&^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^&&&).

^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^. Formal structure

A standard starting point is the decomposition

$$J = J^{(s)} + J^{(a)},\qquad J^{(s)}_{ab}=J^{(s)}_{ba},\qquad J^{(a)}_{ab}=-J^{(a)}_{ba},$$

which isolates the reciprocal and non-reciprocal sectors of the coupling matrix. In mean-field $O(2)$ systems, complex population order parameters are written as $z_a=R_a e^{i\phi_a}$, while in scalar relaxational systems one often studies asymmetrically coupled fields $\phi_1,\phi_2$ with $g_{AB}\neq g_{BA}$ (Weis et al., 22 Jul 2025, Lorenzana et al., 22 Sep 2025).

The dynamical consequence of $J^{(a)}\neq 0$ is not merely a perturbation of equilibrium criticality. In several formulations the antisymmetric sector breaks detailed balance and destroys any Hamiltonian or Boltzmann stationary measure, so that the transition must be defined through dynamical observables: fixed-point stability, the emergence of limit cycles or chaos, direction-dependent transport, or the onset of finite entropy production (Garcés et al., 2024, Alston et al., 2023).

Current usage does not single out one universal phenomenon. The same label covers, among other cases, transitions from static to oscillatory order in coupled spin models, from unique fixed points to chaotic attractors in ecosystems, from absorbing states to active turbulence in dense driven matter, from critical non-reciprocal phases to reciprocity-restored dissipative phases in open fermion chains, and from reciprocal to direction-dependent parity-time transitions in photonics (Avni et al., 2023, &&&^^^^2(Weber et al., 6 Feb 2026) OR \2query2^^^^&&&, &&&^^^^2(Weber et al., 6 Feb 2026) OR \2(Weber et al., 6 Feb 2026) OR \2^^^^&&&, Soares et al., 21 May 2025, &&&^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^3&&&).

2. Spectral organization, exceptional points, and bifurcations

A central organizing theme is the appearance of non-Hermitian spectral singularities. In the general symmetry-based theory for $O(2)$-equivariant dynamics, non-reciprocity makes the Jacobian non-Hermitian and non-normal, so eigenvectors need not be orthogonal and exceptional points can control the transition to time-dependent phases where a spontaneously broken symmetry is dynamically restored (&&&^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^&&&). In multipopulation extensions, the reduced phase dynamics on $\mathbb{T}^{N-1}$ supports chiral limit cycles, non-winding limit cycles, quasiperiodic tori, and chaos; the relevant transitions include critical exceptional points, Hopf bifurcations, saddle-node bifurcations on limit cycles, and homoclinic orbit bifurcations (Weis et al., 22 Jul 2025).

The exceptional-point scenario is especially explicit when a Goldstone mode is already present. In that case, a second mode can coalesce with the zero mode at criticality, producing defective Jacobians or defective Floquet monodromies. For PRESERVED_PLACEHOLDER_^^^^2(Weber et al., 6 Feb 2026) OR \2query2^^^^ populations, a homoclinic merger of two PRESERVED_PLACEHOLDER_^^^^2(Weber et al., 6 Feb 2026) OR \2(Weber et al., 6 Feb 2026) OR \2^^^^-broken chiral orbits can dynamically restore PRESERVED_PLACEHOLDER_^^^^2(Weber et al., 6 Feb 2026) OR \22^^^^ symmetry and, when the saddle quantity is positive, generate Shilnikov chaos (Weis et al., 22 Jul 2025).

Exceptional points, however, are not universal prerequisites. The non-reciprocal Dicke model realizes a non-stationary phase as spontaneous PRESERVED_PLACEHOLDER_^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^3-symmetry breaking of the steady state, and its adiabatically eliminated spin-only description has exceptional points at the phase boundary; yet the full model with dynamical photon mediation retains the non-reciprocal phase transition without exceptional points and without an underlying broken symmetry being required a priori (&&&^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^7&&&). A plausible implication is that exceptional points are frequent organizing centers rather than a necessary definition of the phenomenon.

3. Magnetic realizations

In chiral magnets, non-reciprocity appears already at the level of elementary excitations. In MnSi, non-reciprocity means that spin waves do not obey PRESERVED_PLACEHOLDER_^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^4; the Dzyaloshinskii–Moriya interaction allowed by the non-centrosymmetric space group PRESERVED_PLACEHOLDER_^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^5 and an applied magnetic field together shift magnon branches away from PRESERVED_PLACEHOLDER_^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^6 and redistribute spectral weight between PRESERVED_PLACEHOLDER_^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^7 (&&&^^^^2query2^^^^&&&). Near the skyrmion–paramagnetic boundary, inelastic neutron scattering at PRESERVED_PLACEHOLDER_^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^8 and PRESERVED_PLACEHOLDER_^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^9 shows a broad, structured skyrmion magnon band that evolves smoothly, within instrumental resolution, into quasi-elastic paramagnons in the fluctuation-disordered regime at $$J = J^{(s)} + J^{(a)},\qquad J^{(s)}_{ab}=J^{(s)}_{ba},\qquad J^{(a)}_{ab}=-J^{(a)}_{ba},$$^^^^2query2^^^^–J=J(s)+J(a),Jab(s)=Jba(s),Jab(a)=−Jba(a),J = J^{(s)} + J^{(a)},\qquad J^{(s)}_{ab}=J^{(s)}_{ba},\qquad J^{(a)}_{ab}=-J^{(a)}_{ba},^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^ (&&&^^^^2query2^^^^&&&).

The notable point is that the non-reciprocal character does not disappear at the skyrmion–paramagnetic transition. The maximal spectral intensity remains shifted away from $$J = J^{(s)} + J^{(a)},\qquad J^{(s)}_{ab}=J^{(s)}_{ba},\qquad J^{(a)}_{ab}=-J^{(a)}_{ba},$$2, the asymmetry reverses when $$J = J^{(s)} + J^{(a)},\qquad J^{(s)}_{ab}=J^{(s)}_{ba},\qquad J^{(a)}_{ab}=-J^{(a)}_{ba},$$3 is flipped, and the effect stays discernible up to $$J = J^{(s)} + J^{(a)},\qquad J^{(s)}_{ab}=J^{(s)}_{ba},\qquad J^{(a)}_{ab}=-J^{(a)}_{ba},$$4 under $$J = J^{(s)} + J^{(a)},\qquad J^{(s)}_{ab}=J^{(s)}_{ba},\qquad J^{(a)}_{ab}=-J^{(a)}_{ba},$$5, whereas at $$J = J^{(s)} + J^{(a)},\qquad J^{(s)}_{ab}=J^{(s)}_{ba},\qquad J^{(a)}_{ab}=-J^{(a)}_{ba},$$6 the inelastic line shapes are symmetric and non-reciprocity is absent (&&&^^^^2query2^^^^&&&). This indicates that non-reciprocity here is a symmetry property tied to chirality and broken time-reversal symmetry rather than to static skyrmion order itself.

A different magnetic route uses reservoir engineering. In magnetic metals under continuous light injection, decay through a virtually excited localized state yields a dissipative inter-spin coupling

$$J = J^{(s)} + J^{(a)},\qquad J^{(s)}_{ab}=J^{(s)}_{ba},\qquad J^{(a)}_{ab}=-J^{(a)}_{ba},$$7

which depends on the local dissipation rate and therefore becomes non-reciprocal when only part of the system is illuminated (&&&2^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^&&&). Applied to layered ferromagnets, this mechanism drives a static-to-time-dependent chiral phase transition, with the sign inversion threshold set by $$J = J^{(s)} + J^{(a)},\qquad J^{(s)}_{ab}=J^{(s)}_{ba},\qquad J^{(a)}_{ab}=-J^{(a)}_{ba},$$8 (&&&2^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^&&&).

Glassy magnetism provides yet another variant. In a bipartite spherical Sherrington–Kirkpatrick model with deterministic antisymmetric coupling $$J = J^{(s)} + J^{(a)},\qquad J^{(s)}_{ab}=J^{(s)}_{ba},\qquad J^{(a)}_{ab}=-J^{(a)}_{ba},$$9, dynamical mean-field theory finds an exceptional-point-mediated transition from a static disorder phase to an oscillating amorphous phase at $O(2)$^^^^2query2^^^^, with critical relaxation

$O(2)$^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^

at the transition and oscillatory aging below it (Lorenzana et al., 2024).

4. Lattice spin models, coarsening, and defect-controlled criticality

A minimal scalar example is the single-species Ising model with state-dependent non-reciprocal couplings and Glauber dynamics. On the fully connected graph, the magnetization obeys

$O(2)$2

and the cubic term in the effective Landau expansion appears when $O(2)$3, explicitly breaking spin-flip symmetry and producing metastability, hysteresis, and a first-order transition on top of the usual continuous paramagnetic–ferromagnetic line (Garcés et al., 2024). In two-dimensional simulations, the continuous transition retains Ising-like $O(2)$4 and $O(2)$5, with $O(2)$6 and $O(2)$7 for $O(2)$8, while $O(2)$9 increases from $z_a=R_a e^{i\phi_a}$^^^^2query2^^^^ to $z_a=R_a e^{i\phi_a}$^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^ and the coarsening exponent remains $z_a=R_a e^{i\phi_a}$2 (Garcés et al., 2024).

A two-species Ising generalization with antisymmetric on-site couplings realizes a different mechanism. Mean field predicts disordered, static ordered, and swap phases, the last being a limit cycle in the $z_a=R_a e^{i\phi_a}$3 plane born at a Hopf line $z_a=R_a e^{i\phi_a}$4 with linear frequency $z_a=R_a e^{i\phi_a}$5 (Avni et al., 2023). In finite dimensions, static order is destabilized by droplet growth when the couplings are fully antisymmetric, but an explicit asymmetry $z_a=R_a e^{i\phi_a}$6 can re-stabilize it through droplet capture (Avni et al., 2024).

The fate of the time-dependent phase is dimension-dependent. In two dimensions, the swap phase is destroyed by spiral defects; both the synchronization order parameter

$z_a=R_a e^{i\phi_a}$7

and the phase-space angular momentum vanish with increasing size (Avni et al., 2023). In three dimensions, by contrast, the disorder-to-swap transition is continuous with XY critical exponents $z_a=R_a e^{i\phi_a}$8, $z_a=R_a e^{i\phi_a}$9, and $\phi_1,\phi_2$^^^^2query2^^^^, and the coherence time scales as $\phi_1,\phi_2$^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^, consistent with robust temporal order in the thermodynamic limit (Avni et al., 2023).

5. Active matter, ecosystems, and entropy production

In high-dimensional ecology, the asymmetric MacArthur consumer-resource model introduces non-reciprocity by decoupling the consumption coefficients $\phi_1,\phi_2$2 from the impact coefficients $\phi_1,\phi_2$3, with reciprocity parameter $\phi_1,\phi_2$4 (&&&^^^^2(Weber et al., 6 Feb 2026) OR \2query2^^^^&&&). Cavity theory yields an instability condition

$\phi_1,\phi_2$5

and the transition is continuous, with dynamical susceptibility variances diverging as $\phi_1,\phi_2$6 (&&&^^^^2(Weber et al., 6 Feb 2026) OR \2query2^^^^&&&). Below the threshold the system reaches a unique uninvadable steady state; above it the phase is chaotic, with positive maximal Lyapunov exponent and finite Kaplan–Yorke dimension (&&&^^^^2(Weber et al., 6 Feb 2026) OR \2query2^^^^&&&).

In dense disordered active matter, non-reciprocal pair forces can instead produce an absorbing-to-chaotic transition. For overdamped particles interacting via

$\phi_1,\phi_2$7

the system arrests into an absorbing amorphous state for $\phi_1,\phi_2$8 and enters a chaotic active-turbulent phase for $\phi_1,\phi_2$9, with $g_{AB}\neq g_{BA}$^^^^2query2^^^^ (&&&^^^^2(Weber et al., 6 Feb 2026) OR \2(Weber et al., 6 Feb 2026) OR \2^^^^&&&). The activity grows as $g_{AB}\neq g_{BA}$^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^ with $g_{AB}\neq g_{BA}$2, the correlation length scales as $g_{AB}\neq g_{BA}$3 with $g_{AB}\neq g_{BA}$4, the persistent time scales as $g_{AB}\neq g_{BA}$5 with $g_{AB}\neq g_{BA}$6, and $g_{AB}\neq g_{BA}$7, all consistent with directed percolation in two dimensions (&&&^^^^2(Weber et al., 6 Feb 2026) OR \2(Weber et al., 6 Feb 2026) OR \2^^^^&&&).

For conserved continuum fields, the transition can be expressed directly in thermodynamic irreversibility. In a binary Cahn–Hilliard model with antisymmetric cross-diffusion $g_{AB}\neq g_{BA}$8, the informational entropy production rate is

$g_{AB}\neq g_{BA}$9

and the deterministic PT-breaking threshold is

$J^{(a)}\neq 0$^^^^2query2^^^^

in the binary specialization (Alston et al., 2023). Above $J^{(a)}\neq 0$^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^, the fields enter a traveling-wave phase with nonzero global polar order parameter $J^{(a)}\neq 0$2, and the macroscopic contribution to irreversibility obeys

$J^{(a)}\neq 0$3

In the one-mode weak-noise regime,

$J^{(a)}\neq 0$4

which makes the $J^{(a)}\neq 0$5 scaling explicit (Alston et al., 2023).

6. Quantum, photonic, and driven-dissipative many-body systems

In photonics, a direct route to non-reciprocal phase transition is dynamic gain–loss modulation. For two waveguide modes coupled by a traveling-wave imaginary-index modulation, the Floquet quasi-energies are

$J^{(a)}\neq 0$6

with an exceptional point at $J^{(a)}\neq 0$7 (&&&^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^3&&&). Forward phase matching gives $J^{(a)}\neq 0$8 and therefore a thresholdless broken-$J^{(a)}\neq 0$9 transition $O(2)$^^^^2query2^^^^, whereas backward propagation remains in the exact phase until $O(2)$^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^ exceeds the finite threshold $O(2)$2 (&&&^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^3&&&). The result is a direction-dependent $O(2)$3 transition rather than merely asymmetric transmission.

A chiral quantum-optical version uses a waveguide-coupled micro-ring and a V-type atom. Because the clockwise and counterclockwise whispering-gallery modes address transitions with different decay rates, the effective two-mode problem has direction-dependent exceptional points

$O(2)$4

which in the reported parameter set give $O(2)$5 and $O(2)$6 (Cai et al., 2024). The forward and backward configurations can therefore fall into different $O(2)$7 phases at the same global parameters, and the non-reciprocal phase region also supports non-reciprocal photon blockade (Cai et al., 2024).

Driven-dissipative many-body platforms exhibit still richer behavior. In the non-reciprocal Dicke model, phase-engineered light–matter couplings generate asymmetric effective inter-species couplings $O(2)$8 after adiabatic elimination, and the dynamical phase is identified with spontaneous $O(2)$9-symmetry breaking of the nonlinear steady state (&&&^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^7&&&). In a one-dimensional Lindbladian bosonic chain with correlated single-particle loss, periodic boundaries always select finite-momentum traveling-wave condensates, whereas open boundaries produce vacuum, static condensate, and multiple dynamical phases, including spontaneous particle–hole symmetry breaking at a critical exceptional point and edge–bulk differentiated regimes (Belyansky et al., 7 Feb 2025).

Open fermionic systems supply two further templates. In an interacting spinless-fermion chain with non-local gain and loss,

$\mathbb{T}^{N-1}$^^^^2query2^^^^

the line $\mathbb{T}^{N-1}$^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^ closes the dissipative gap and gives power-law relaxation with exponent $\mathbb{T}^{N-1}$2, while increasing the interaction $\mathbb{T}^{N-1}$3 to order $\mathbb{T}^{N-1}$4 opens a many-body dissipative gap, restores reciprocity dynamically, and converts the decay to exponential (Soares et al., 21 May 2025). In a reservoir-engineered non-reciprocal Kitaev chain, the correlation Hamiltonian undergoes a pairing-induced transition; for aligned non-reciprocal hopping and pairing, the spectrum collapses at $\mathbb{T}^{N-1}$5 into an $\mathbb{T}^{N-1}$6-fold exceptional point, separating a non-reciprocal phase with directional spreading and slow relaxation from a density-wave phase with short relaxation and boundary-induced modulation (Brighi et al., 28 Oct 2025).

7. Criticality, relevance, and unresolved problems

The present literature does not support a single universality class for non-reciprocal phase transitions. Instead, different models realize different critical structures: directed percolation in absorbing-to-chaotic active turbulence, XY exponents in the three-dimensional disorder-to-swap transition, Ising-like $\mathbb{T}^{N-1}$7 and $\mathbb{T}^{N-1}$8 but modified $\mathbb{T}^{N-1}$9 in the single-species non-reciprocal Ising model, continuous susceptibility divergence at an ecology-to-chaos threshold, and interaction-driven reciprocity restoration in open quantum matter (&&&^^^^2(Weber et al., 6 Feb 2026) OR \2(Weber et al., 6 Feb 2026) OR \2^^^^&&&, Avni et al., 2023, Garcés et al., 2024, &&&^^^^2(Weber et al., 6 Feb 2026) OR \2query2^^^^&&&, Soares et al., 21 May 2025). This suggests that non-reciprocity is a mechanism that reshapes critical behavior rather than a fixed universality class by itself.

A perturbative criterion for when non-reciprocity changes universality was derived for two asymmetrically coupled scalar fields with Model-A-type dynamics. For identical uncoupled fields, the linear correction scales as PRESERVED_PLACEHOLDER_^^^^2(Weber et al., 6 Feb 2026) OR \2query2query2^^^^, so deterministic non-reciprocity is relevant if PRESERVED_PLACEHOLDER_^^^^2(Weber et al., 6 Feb 2026) OR \2query2(Weber et al., 6 Feb 2026) OR \2^^^^ (Lorenzana et al., 22 Sep 2025). For random antisymmetric coupling, relevance at an equilibrium-type transition requires

PRESERVED_PLACEHOLDER_^^^^2(Weber et al., 6 Feb 2026) OR \2query22^^^^

whereas around a nonequilibrium swap transition the Harris-type condition reduces to

PRESERVED_PLACEHOLDER_^^^^2(Weber et al., 6 Feb 2026) OR \2query2^^^^3

with the appropriate nonequilibrium PRESERVED_PLACEHOLDER_^^^^2(Weber et al., 6 Feb 2026) OR \2query2^^^^4 (Lorenzana et al., 22 Sep 2025).

Several issues remain open across the field. In multipopulation PRESERVED_PLACEHOLDER_^^^^2(Weber et al., 6 Feb 2026) OR \2query2^^^^5 systems, finite-size fluctuations and spatial inhomogeneity can shift phase boundaries and alter torus and homoclinic bifurcations (Weis et al., 22 Jul 2025). In the two-species non-reciprocal Ising model, the large-PRESERVED_PLACEHOLDER_^^^^2(Weber et al., 6 Feb 2026) OR \2query2^^^^6 three-dimensional regime with scroll waves is not fully resolved (Avni et al., 2024). In open fermion chains, a precise universality classification and a full Liouvillian spectral theory of the interaction-driven transition are still lacking (Soares et al., 21 May 2025). In non-reciprocal driven-dissipative condensates, the critical theory beyond mean field near the critical exceptional point is also open (Belyansky et al., 7 Feb 2025).

The common conclusion is narrower than a universal taxonomy but stronger than a collection of isolated examples: asymmetric couplings can change not only transport and mode propagation but the very ontology of phases, producing time-dependent order, dynamic symmetry restoration, direction-dependent critical thresholds, or persistent non-reciprocal fluctuations even after static order disappears (&&&^^^^2(Weber et al., 6 Feb 2026) OR \2^^^^&&&, &&&^^^^2query2^^^^&&&).