🌀 nonlinear sciences

First-order integrability-breaking phase transitions in dynamical systems

This paper demonstrates that both confined stochastic random walks and deterministic stadium-like billiards exhibit a first-order integrability-breaking phase transition characterized by a discontinuous jump in the diffusive observable and critical slowing down, sharing a universal set of scaling exponents that arise from a common coarse-grained diffusion mechanism.

Original authors: Anne Ketri P. da Fonseca, Marcelo de Almeida Presotto, Diego F. M. Oliveira, Edson D. Leonel

Published 2026-09-15✓ Author reviewed ⓘ

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Anne Ketri P. da Fonseca, Marcelo de Almeida Presotto, Diego F. M. Oliveira, Edson D. Leonel

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the study of how things move and change, scientists have long been fascinated by the boundary between order and chaos. For centuries, the concept of a "phase transition" has helped explain how matter shifts from one state to another, like water freezing into ice or a magnet losing its pull when heated. In these classic examples, the change is often marked by a sudden jump in a measurable property, such as density or magnetism, while the time it takes for the system to settle into its new state can become incredibly long. This idea of a sudden shift has been extended to the world of complex systems, where researchers look for similar jumps when a system moves from predictable, orderly motion to wild, unpredictable behavior. A key question in this field is whether the shift from perfect order to total chaos must always happen gradually, with the system's behavior changing smoothly as a control knob is turned, or if it can happen with a sudden, jarring leap.

A team of researchers has now found that the answer is yes: the transition can be sudden. By studying two very different systems—one involving a random walk of particles in a confined space and the other a ball bouncing inside a specific type of curved container—they discovered a new kind of transition. In both cases, they observed that as they introduced a tiny amount of disturbance to a perfectly orderly system, the system's ability to spread out and explore its space did not grow slowly. Instead, it jumped instantly from zero to a fixed, finite value. This means that even the smallest nudge away from perfect order causes the system to behave as if it were fully chaotic, exploring its entire available space immediately. Yet, despite this sudden jump in behavior, the time it takes for the system to actually reach that chaotic state becomes infinitely long as the disturbance gets smaller. This combination of a sudden jump in the final state and a slowing down of the approach to that state defines a new class of "first-order" transitions in dynamical systems, a phenomenon that challenges the previous assumption that such shifts must be gradual.

The researchers began by looking at a simple, random process. Imagine a large group of people walking back and forth inside a long, narrow hallway with walls at both ends. In a perfectly still hallway, if a person takes no steps, they stay exactly where they started. But if they take even the tiniest, random steps, they will eventually wander to every corner of the hallway. The scientists measured how far, on average, these walkers spread out over time. They found that if the step size was exactly zero, the walkers never moved, and the average distance from the start remained zero. However, the moment the step size became any number larger than zero, no matter how small, the walkers eventually explored the entire hallway, and the average distance settled at a specific, fixed value determined only by the length of the hallway. This was a sudden jump: the system went from staying put to filling the entire space instantly as soon as any movement was allowed.

To see if this strange behavior was unique to random walkers, the team turned to a completely different system: a deterministic billiard. This is a model of a particle bouncing around inside a container with curved walls, similar to a pool table but with a specific shape known as a stadium. In this system, the particle follows strict laws of physics, bouncing off the walls without any randomness. The researchers could change the shape of the walls slightly, turning a perfectly rectangular box into a stadium with curved ends. When the walls were straight, the particle's motion was predictable and orderly, and it never explored all possible angles of movement. But the moment the walls were curved, even by a microscopic amount, the particle's path became chaotic. It began to bounce in a way that eventually covered every possible angle and position within the container. Just like the random walkers, the measure of how much the particle spread out jumped from zero to a fixed, finite value the instant the walls were curved.

What makes this discovery particularly surprising is what happens as the researchers make the disturbance smaller and smaller. In many physical systems, when a change is small, the time it takes to see the full effect is also small. Here, the opposite is true. As the step size in the random walk or the curvature of the billiard walls gets closer and closer to zero, the time it takes for the system to reach that fully explored, chaotic state grows longer and longer, theoretically becoming infinite right at the point where the disturbance disappears. The system is stuck in a state of "critical slowing down," where it takes forever to settle, even though the final state it is trying to reach is completely different from the starting point. This creates a unique signature: a sudden, discontinuous change in the final behavior, paired with a diverging time scale to get there.

The researchers confirmed that both the random walkers and the bouncing ball followed the exact same mathematical rules for how they spread out and how long it took to settle. They found that the speed of the spread was normal, and the time it took to reach the final state grew in a predictable way as the disturbance shrank. This agreement between a system driven by pure chance and one driven by strict, deterministic laws suggests that this type of sudden transition is a fundamental feature of how complex systems behave, not just a quirk of one specific model. It reveals that nature can switch from order to chaos in a single, sharp step, defying the expectation that such changes must always be smooth and gradual.

This work extends the language of phase transitions, traditionally used to describe things like melting ice or magnetism, into the realm of motion and dynamics. It shows that the transition from a predictable world to a chaotic one can be just as abrupt as water turning to ice. By identifying this new class of transitions, the study provides a clearer picture of how systems respond to small changes. It suggests that in many complex environments, from the movement of particles to the behavior of large networks, a tiny push might not just nudge the system slightly, but could trigger an immediate and complete reorganization of its behavior, even if the system takes a very long time to actually get there. The findings offer a new way to understand the delicate balance between order and chaos, highlighting that the path from one to the other can be as sudden as a switch being flipped.

Technical Summary: First-Order Integrability-Breaking Phase Transitions in Dynamical Systems

Problem Statement
The transition from integrability to chaos in nonlinear dynamical systems is a fundamental problem in statistical mechanics and nonlinear dynamics. While previous studies on systems such as the standard map and time-dependent collision models have established that integrability-breaking transitions are typically continuous (second-order), characterized by a stationary order parameter that vanishes continuously as the perturbation approaches zero, a critical question remains: must this transition always be continuous? This paper investigates the existence of discontinuous (first-order) routes from integrability to chaos, specifically examining whether a finite jump in the stationary diffusive observable can coexist with critical slowing down (diverging relaxation timescales).

Methodology
The authors employ two distinct dynamical models to isolate the diffusive mechanisms underlying the transition, comparing a stochastic system with a deterministic one:

Confined Stochastic Random Walk: A one-dimensional model where independent random walkers are confined to a finite interval $[-L, L]$ with reflecting boundaries. The dynamics are governed by $x_{n+1} = x_n + \epsilon Z(n)$ , where $\epsilon$ is the perturbation parameter controlling the step amplitude and $Z(n)$ is a uniform random variable. The system is analyzed in the limit $\epsilon \to 0$ .
Deterministic Stadium-like Billiard: A two-dimensional billiard system with parabolic boundaries replacing the semicircles of a Bunimovich stadium. The geometry is controlled by a parameter $b$ , where $b=0$ represents an integrable rectangular billiard. The study focuses on the "dispersing" regime ( $b < 0$ ), where arbitrarily small curvature induces global chaos, contrasting it with the "focusing" regime ( $b > 0$ ) which exhibits mixed phase spaces.

In both systems, the authors analyze the scaling behavior of the root-mean-squared displacement (or angular roughness $\omega$ for the billiard) as a function of iteration number $n$ and the control parameter ( $\epsilon$ or $b$ ). They utilize scaling hypotheses of the form $O(n, \lambda) = \lambda^\alpha f(n/\lambda^z)$ to extract critical exponents:

$\alpha$ : Governs the dependence of the stationary observable on the perturbation.
$\beta$ : Governs the initial diffusive growth regime.
$z$ : Governs the scaling of the crossover iteration $n_x$ (relaxation timescale).

Key Results
The study reveals that both models, despite their different microscopic origins, exhibit identical scaling properties characteristic of a first-order phase transition:

Discontinuous Order Parameter ( $\alpha = 0$ ): In both systems, the stationary observable vanishes exactly at the unperturbed limit ( $\epsilon=0$ $ϵ = 0$ or $b=0$ $b = 0$ ) but jumps to a finite, geometry-controlled value for any arbitrarily small nonzero perturbation.
- For the random walk: $x_{sat} = 0$ if $\epsilon=0$ , but $x_{sat} = L/\sqrt{3}$ for any $\epsilon \neq 0$ .
- For the dispersing billiard: $\omega_{sat} = 0$ if $b=0$ , but $\omega_{sat} = \pi/\sqrt{12}$ for any $b < 0$ .
- This finite jump defines the order parameter signature of a first-order transition.
Normal Diffusion ( $\beta = 1/2$ ): In the initial growth regime ( $n \ll n_x$ ), both systems exhibit normal diffusion with an exponent $\beta \approx 1/2$ .
Critical Slowing Down ( $z = -2$ ): The crossover iteration $n_x$ $n_{x}$ , representing the time required to reach the stationary state, diverges as a power law $n_x \propto \lambda^{-2}$ $n_{x} \propto λ^{- 2}$ as the critical point is approached.
- Analytical derivation shows the diffusion coefficient $D$ vanishes quadratically with the perturbation ( $D \propto \epsilon^2$ or $D \propto b^2$ ).
- Since the accessible domain is finite, the relaxation time scales as $\tau \sim L^2/D \propto \lambda^{-2}$ , leading to the exponent $z = -2$ .

Significance and Claims
The paper claims to provide the first dynamical realization of a first-order integrability-breaking phase transition. The primary significance lies in demonstrating that a discontinuous stationary response (a finite jump in the order parameter) can coexist with a diverging dynamical timescale (critical slowing down). This combination distinguishes these transitions from the continuous (second-order) transitions previously documented in nonlinear dynamics, where the order parameter vanishes continuously.

The authors argue that the agreement between the stochastic random walk and the deterministic chaotic billiard suggests a common dynamical universality class. This universality arises not from microscopic equivalence, but from a shared coarse-grained mechanism: normal diffusion within a finite accessible domain where the diffusion coefficient vanishes quadratically at the transition.

Furthermore, the stadium-like billiard model serves as a unified setting to compare continuous and discontinuous routes from the same integrable limit. The focusing side ( $b>0$ ) exhibits a continuous transition (second-order) with $\alpha \approx 0.28$ , while the dispersing side ( $b<0$ ) exhibits the discontinuous transition (first-order) with $\alpha = 0$ .

Conclusion
The study extends the statistical-mechanics description of phase transitions to integrability-breaking dynamics. It establishes that discontinuous transitions are possible in dynamical systems, characterized by the specific exponent set $(\alpha, \beta, z) = (0, 1/2, -2)$ . The results suggest that this mechanism represents a broader class of discontinuous transitions in nonlinear dynamical systems, independent of whether the underlying dynamics are stochastic or deterministic.

Technical Summary: First-Order Integrability-Breaking Phase Transitions in Dynamical Systems

More like this