Droplet-Induced Swap Phase Dynamics

• The droplet-induced swap phase is a non-equilibrium state where metastable droplets initiate oscillatory reversals between two competing spin species.
• Mean-field analysis shows a Hopf bifurcation leading to limit cycle oscillations, marking the transition from static ferromagnetic order to a dynamic swap regime.
• Dimensional studies reveal that while 2D systems destabilize the swap phase via spiral defects, 3D systems maintain robust oscillations with droplet-capture mechanisms affecting static order.

Searching arXiv for the cited papers and closely related context. arXiv_search("(Avni et al., 2024) nonreciprocal Ising model swap phase droplet-induced swap; (Yao et al., 31 Dec 2025) active phase separation necking rupture cavitation", 10, "search")

In the nonreciprocal Ising model, the droplet-induced swap phase denotes a non-equilibrium oscillatory ordered regime in which the two spin species do not remain in a static ferromagnetic configuration, but instead alternate their magnetization through a sequence initiated by nucleation and growth of metastable droplets. Within the broader swap phenomenology, this droplet-controlled regime is central to understanding why static order can fail, why oscillatory order can emerge, and why the fate of the system depends sharply on spatial dimension, symmetry, and the structure of the inter-species couplings (Avni et al., 2024).

1. Model, nonreciprocity, and broken detailed balance

The underlying model places two Ising variables on every lattice site,

$\sigma_i^A,\sigma_i^B \in \{+1,-1\},$

with state vector

$$\vec{\sigma}=\{\sigma_1^A,\dots,\sigma_{L^d}^A,\sigma_1^B,\dots,\sigma_{L^d}^B\}.$$

Same-species spins couple symmetrically to nearest neighbors with strength $J$. The distinctive ingredient is the on-site inter-species coupling matrix

$$K_{\alpha\beta}=\begin{pmatrix} 0 & K_+ + K_-\ K_+ - K_- & 0 \end{pmatrix},$$

so spin flips are governed not by a single global energy but by a species-dependent “selfish energy”

$$E_i^\alpha=-J\sum_{j\,\text{nn of }i}\sigma_i^\alpha\sigma_j^\alpha - K_{\alpha\beta}\sigma_i^\alpha\sigma_i^\beta.$$

The dynamics is Glauber-like, with rates

$$w(F_i^\alpha\vec{\sigma}\mid \vec{\sigma}) =\frac{1}{2\tau}\left[1-\tanh\!\left(\frac{\Delta E_i^\alpha}{2k_BT}\right)\right].$$

Because the couplings are generally not reciprocal, the dynamics does not satisfy detailed balance. A nontrivial Kolmogorov cycle with ratio $e^{8\tilde K_-}$ shows that the system is genuinely out of equilibrium whenever $K_-\neq 0$. This broken reciprocity is the prerequisite for the swap state: one species tends to align with the other while the other pushes in the opposite direction, so static alignment and dynamical alternation compete rather than being reducible to equilibrium free-energy minimization (Avni et al., 2024).

A particularly important limit is the fully anti-symmetric case $K_+=0$, so $K_-=K>0$. Then the coarse-grained dynamics becomes

$$\vec{\sigma}=\{\sigma_1^A,\dots,\sigma_{L^d}^A,\sigma_1^B,\dots,\sigma_{L^d}^B\}.$$0

with $$\vec{\sigma}=\{\sigma_1^A,\dots,\sigma_{L^d}^A,\sigma_1^B,\dots,\sigma_{L^d}^B\}.$$1. In this case the four local states $$\vec{\sigma}=\{\sigma_1^A,\dots,\sigma_{L^d}^A,\sigma_1^B,\dots,\sigma_{L^d}^B\}.$$2 are related by rotations in the $$\vec{\sigma}=\{\sigma_1^A,\dots,\sigma_{L^d}^A,\sigma_1^B,\dots,\sigma_{L^d}^B\}.$$3 plane, giving an enhanced $$\vec{\sigma}=\{\sigma_1^A,\dots,\sigma_{L^d}^A,\sigma_1^B,\dots,\sigma_{L^d}^B\}.$$4 symmetry. That symmetry becomes decisive for both the oscillatory phase portrait and the droplet mechanism.

2. Mean-field swap dynamics and its bifurcation structure

At mean-field level, the local magnetizations

$$\vec{\sigma}=\{\sigma_1^A,\dots,\sigma_{L^d}^A,\sigma_1^B,\dots,\sigma_{L^d}^B\}.$$5

obey

$$\vec{\sigma}=\{\sigma_1^A,\dots,\sigma_{L^d}^A,\sigma_1^B,\dots,\sigma_{L^d}^B\}.$$6

The mean-field phase portrait contains three phases: disorder, static order, and swap. For small $$\vec{\sigma}=\{\sigma_1^A,\dots,\sigma_{L^d}^A,\sigma_1^B,\dots,\sigma_{L^d}^B\}.$$7, the only stable fixed point is $$\vec{\sigma}=\{\sigma_1^A,\dots,\sigma_{L^d}^A,\sigma_1^B,\dots,\sigma_{L^d}^B\}.$$8. For larger $$\vec{\sigma}=\{\sigma_1^A,\dots,\sigma_{L^d}^A,\sigma_1^B,\dots,\sigma_{L^d}^B\}.$$9 and vanishing nonreciprocity, the usual pitchfork bifurcation yields static ferromagnetic order. When $J$0, however, the origin becomes a Hopf point with eigenvalues

$J$1

so that at $J$2 a stable limit cycle appears. This limit cycle is the swap phase (Avni et al., 2024).

In the $J$3 plane, the swap trajectory circles around the origin, with $J$4 and $J$5 oscillating out of phase. Close to the Hopf point the oscillation is nearly circular and harmonic. Farther from the bifurcation it becomes square-like because of the underlying $J$6 anisotropy. The mean-field boundary between swap and static order is a $J$7-symmetric SNIC bifurcation, where the limit cycle collides with four pairs of stable and saddle points and the period diverges.

This mean-field structure is important because it identifies the swap state as a genuine dynamical ordered phase rather than a transient relaxation pattern. At the same time, the later finite-dimensional analysis shows that mean-field coexistence of static order and swap does not directly determine the large-scale fate of the system once droplets and defects are admitted.

3. Droplet-induced alternation and the loss of static order

The droplet-induced mechanism explains why static order is unstable in the fully anti-symmetric model and how alternation between species can emerge from local nucleation events. In a static ordered state, both species are aligned on average, but because the inter-species couplings have opposite signs, one species is effectively in a metastable state. The paper gives an intuitive mapping to two equilibrium Ising systems in opposite effective fields: one species sits in a local minimum and can lower its selfish free energy by nucleating an opposite-magnetization droplet (Avni et al., 2024).

Once such a droplet exceeds a critical radius $J$8, the bulk gain from the field-like bias exceeds the surface-tension cost, and the droplet expands rather than shrinking. This reversal then makes the other species metastable, which nucleates a droplet of its own. The result is a dynamical sequence of magnetization flips. In finite systems, the system can linger in static order, then a droplet flips one species, then the other, and so on; this is the droplet-induced swap regime.

In the fully anti-symmetric case, the paper argues that static order is unstable in any finite dimension. The critical droplet size is finite whenever $J$9 and $$K_{\alpha\beta}=\begin{pmatrix} 0 & K_+ + K_-\ K_+ - K_- & 0 \end{pmatrix},$$0 are finite, so fluctuations will eventually nucleate expanding droplets. Because $$K_{\alpha\beta}=\begin{pmatrix} 0 & K_+ + K_-\ K_+ - K_- & 0 \end{pmatrix},$$1- and $$K_{\alpha\beta}=\begin{pmatrix} 0 & K_+ + K_-\ K_+ - K_- & 0 \end{pmatrix},$$2-droplets are symmetric in that limit, the growing droplets of the two species expand at the same speed, and a nested “child” droplet cannot catch and extinguish its parent. The flip cascade therefore does not self-arrest. This is the specific sense in which the swap dynamics is “droplet-induced”: the oscillatory exchange of magnetization is not merely accompanied by droplets but is generated by their irreversible growth.

4. Dimensional dependence: 2D defect instability and 3D stability

The finite-dimensional realization of swap dynamics depends strongly on spatial dimension. In 2D, simulations show transient or finite-size oscillations, but the swap phase is destabilized by spiral defects. These defects are topological structures in a local angle variable $$K_{\alpha\beta}=\begin{pmatrix} 0 & K_+ + K_-\ K_+ - K_- & 0 \end{pmatrix},$$3 that encodes the pair $$K_{\alpha\beta}=\begin{pmatrix} 0 & K_+ + K_-\ K_+ - K_- & 0 \end{pmatrix},$$4 as four discrete angles, and in the fully anti-symmetric case they have four arms because of the $$K_{\alpha\beta}=\begin{pmatrix} 0 & K_+ + K_-\ K_+ - K_- & 0 \end{pmatrix},$$5 symmetry. As system size increases, the oscillatory state yields to defect proliferation and disorder. The order parameters

$$K_{\alpha\beta}=\begin{pmatrix} 0 & K_+ + K_-\ K_+ - K_- & 0 \end{pmatrix},$$6

both shrink with $$K_{\alpha\beta}=\begin{pmatrix} 0 & K_+ + K_-\ K_+ - K_- & 0 \end{pmatrix},$$7, indicating that the apparent 2D swap state is not thermodynamically stable (Avni et al., 2024).

In 3D, by contrast, the swap phase is stable. Simulations show spatially homogeneous noisy oscillations with a well-defined global period and a phase shift between $$K_{\alpha\beta}=\begin{pmatrix} 0 & K_+ + K_-\ K_+ - K_- & 0 \end{pmatrix},$$8 and $$K_{\alpha\beta}=\begin{pmatrix} 0 & K_+ + K_-\ K_+ - K_- & 0 \end{pmatrix},$$9. Here $$E_i^\alpha=-J\sum_{j\,\text{nn of }i}\sigma_i^\alpha\sigma_j^\alpha - K_{\alpha\beta}\sigma_i^\alpha\sigma_i^\beta.$$0 and $$E_i^\alpha=-J\sum_{j\,\text{nn of }i}\sigma_i^\alpha\sigma_j^\alpha - K_{\alpha\beta}\sigma_i^\alpha\sigma_i^\beta.$$1 remain finite above a critical $$E_i^\alpha=-J\sum_{j\,\text{nn of }i}\sigma_i^\alpha\sigma_j^\alpha - K_{\alpha\beta}\sigma_i^\alpha\sigma_i^\beta.$$2, and the disorder-to-swap transition is continuous in the low-$$E_i^\alpha=-J\sum_{j\,\text{nn of }i}\sigma_i^\alpha\sigma_j^\alpha - K_{\alpha\beta}\sigma_i^\alpha\sigma_i^\beta.$$3, low-$$E_i^\alpha=-J\sum_{j\,\text{nn of }i}\sigma_i^\alpha\sigma_j^\alpha - K_{\alpha\beta}\sigma_i^\alpha\sigma_i^\beta.$$4 regime. Finite-size scaling gives

$$E_i^\alpha=-J\sum_{j\,\text{nn of }i}\sigma_i^\alpha\sigma_j^\alpha - K_{\alpha\beta}\sigma_i^\alpha\sigma_i^\beta.$$5

in excellent agreement with the 3D XY values rather than 3D Ising. The physical interpretation offered is symmetry-based: the swap phase breaks a continuous time-translation symmetry, and near the Hopf point the complex order parameter behaves like an $$E_i^\alpha=-J\sum_{j\,\text{nn of }i}\sigma_i^\alpha\sigma_j^\alpha - K_{\alpha\beta}\sigma_i^\alpha\sigma_i^\beta.$$6 field. On that basis the oscillatory state is described as a classical continuous time crystal, with coherence time scaling as

$$E_i^\alpha=-J\sum_{j\,\text{nn of }i}\sigma_i^\alpha\sigma_j^\alpha - K_{\alpha\beta}\sigma_i^\alpha\sigma_i^\beta.$$7

The droplet-induced regime persists within this 3D setting but becomes more subtle. At intermediate couplings, droplets govern the alternation of magnetization while the system can still synchronize globally. As $$E_i^\alpha=-J\sum_{j\,\text{nn of }i}\sigma_i^\alpha\sigma_j^\alpha - K_{\alpha\beta}\sigma_i^\alpha\sigma_i^\beta.$$8 increases or $$E_i^\alpha=-J\sum_{j\,\text{nn of }i}\sigma_i^\alpha\sigma_j^\alpha - K_{\alpha\beta}\sigma_i^\alpha\sigma_i^\beta.$$9 decreases, the critical droplet grows and droplets become sparser. The system then crosses from a noisy homogeneous swap regime to one where individual droplets drive magnetization flips. In a finite system, the oscillation period can even scale like $$w(F_i^\alpha\vec{\sigma}\mid \vec{\sigma}) =\frac{1}{2\tau}\left[1-\tanh\!\left(\frac{\Delta E_i^\alpha}{2k_BT}\right)\right].$$0 in the single-droplet limit. The authors note, however, that global synchronization $$w(F_i^\alpha\vec{\sigma}\mid \vec{\sigma}) =\frac{1}{2\tau}\left[1-\tanh\!\left(\frac{\Delta E_i^\alpha}{2k_BT}\right)\right].$$1 can decrease in this droplet regime because different regions fall out of phase, and the ultimate thermodynamic fate of that high-$$w(F_i^\alpha\vec{\sigma}\mid \vec{\sigma}) =\frac{1}{2\tau}\left[1-\tanh\!\left(\frac{\Delta E_i^\alpha}{2k_BT}\right)\right].$$2 part of the swap region is left open.

5. Asymmetric couplings and the droplet-capture mechanism

The general asymmetric case $$w(F_i^\alpha\vec{\sigma}\mid \vec{\sigma}) =\frac{1}{2\tau}\left[1-\tanh\!\left(\frac{\Delta E_i^\alpha}{2k_BT}\right)\right].$$3 changes the droplet narrative by breaking the exact symmetry between the two species. The swap phase itself remains robust in the sense reported numerically: in 2D it remains unstable, in 3D it remains stable, and the disorder-to-swap transition retains the same XY-like continuous criticality in the appropriate regime. What changes is the fate of static order (Avni et al., 2024).

Reciprocal coupling reduces the $$w(F_i^\alpha\vec{\sigma}\mid \vec{\sigma}) =\frac{1}{2\tau}\left[1-\tanh\!\left(\frac{\Delta E_i^\alpha}{2k_BT}\right)\right].$$4 symmetry to the diagonal $$w(F_i^\alpha\vec{\sigma}\mid \vec{\sigma}) =\frac{1}{2\tau}\left[1-\tanh\!\left(\frac{\Delta E_i^\alpha}{2k_BT}\right)\right].$$5. Then $$w(F_i^\alpha\vec{\sigma}\mid \vec{\sigma}) =\frac{1}{2\tau}\left[1-\tanh\!\left(\frac{\Delta E_i^\alpha}{2k_BT}\right)\right].$$6- and $$w(F_i^\alpha\vec{\sigma}\mid \vec{\sigma}) =\frac{1}{2\tau}\left[1-\tanh\!\left(\frac{\Delta E_i^\alpha}{2k_BT}\right)\right].$$7-droplets do not grow identically. The species that couples more strongly to the other can form a smaller critical droplet and expand faster. This asymmetry enables the droplet-capture mechanism: a droplet from one species can be overtaken and destroyed by a nested droplet of the other species before it reaches system size. In the simulations, this can re-stabilize static order when $$w(F_i^\alpha\vec{\sigma}\mid \vec{\sigma}) =\frac{1}{2\tau}\left[1-\tanh\!\left(\frac{\Delta E_i^\alpha}{2k_BT}\right)\right].$$8, because the faster-growing nested droplet “eats” the larger one and stops the flip cascade.

This mechanism clarifies an important misconception. The destabilization of static order is not an inevitable consequence of nonreciprocity alone; rather, it depends on whether droplet growth can run away unchecked. In the fully anti-symmetric case it can, because the two species’ droplets expand at the same speed. In the general asymmetric case it need not, because differential growth rates permit capture. The phase diagrams therefore show that static order and swap are often separated not by a direct transition but by an intermediate disordered regime, especially in 3D. Static order is thus restored not by suppressing nucleation altogether, but by altering the competitive kinetics of nested droplets.

6. Relation to other non-equilibrium droplet pathways

A related but distinct non-equilibrium droplet mechanism appears in the active lattice gas studied in “Non-equilibrium pathways between cluster morphologies in active phase separation: necking, rupture and cavitation” (Yao et al., 31 Dec 2025). There the system is a 2D active lattice gas undergoing motility-induced phase separation, with a geometric phase transition between two metastable liquid-cluster morphologies: a system-spanning slab and a compact droplet. The dynamics breaks global equilibrium because of self-propulsion and orientational dynamics, and the relevant control parameter is the Péclet number $$w(F_i^\alpha\vec{\sigma}\mid \vec{\sigma}) =\frac{1}{2\tau}\left[1-\tanh\!\left(\frac{\Delta E_i^\alpha}{2k_BT}\right)\right].$$9, which measures distance from equilibrium.

That work uses two coordinates: a neck or channel coordinate $e^{8\tilde K_-}$0, which tracks the narrowest connection and changes sign between slab-like and droplet-like states, and a roundness coordinate

$e^{8\tilde K_-}$1

which is small for slabs and close to $e^{8\tilde K_-}$2 for circular droplets. Forward flux sampling shows a pronounced directional asymmetry. The droplet-to-slab transition remains equilibrium-like: the droplet slowly elongates, a narrow vapor channel forms, a rapid topological connection occurs across the periodic box, and the cluster relaxes into a slab. By contrast, slab-to-droplet depends on rare non-equilibrium fluctuations. At low $e^{8\tilde K_-}$3, equilibrium-like and non-equilibrium routes compete; at high $e^{8\tilde K_-}$4, the equilibrium-like route is entirely suppressed, and all observed reactive trajectories begin with a large vapor bubble inside the slab. The topology then changes through rupture or cavitation, with a near-vertical jump in $e^{8\tilde K_-}$5 at low $e^{8\tilde K_-}$6.

This comparison does not identify the two models, which address different observables and different ordered states. It does, however, suggest a broader non-equilibrium principle: once detailed balance is broken, droplet or bubble fluctuations can determine the mechanism of state switching rather than merely altering a transition rate. In the nonreciprocal Ising model, droplets trigger alternation between species; in the active lattice gas, vapor bubbles trigger morphology change. In both cases, the reverse pathway is not generally the time reverse of the forward one, and rare localized fluctuations become the decisive objects for understanding the macroscopic dynamics.