Intermittency-induced regimes in fast-slow dynamical systems

arXiv:2609.25513 · nlin.CD · Submitted 2026-09-22 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Nonlinear Sciences. The day in nonlinear science: chaos, patterns, networks and adaptation.

Kai: I'm Kai.

Mira: And I'm Mira, here with Lev.

Kai: We're looking at the paper: "Intermittency-induced regimes in fast-slow dynamical systems".

Mira: Here's the gist of the paper. Intermittent dynamics are ubiquitous in the Earth system and often arise from the interaction of processes evolving on different time scales.

Kai: Intermittent bursts in a fast forcing system propagate to and reshape the dynamics of a slower response system that would otherwise settle onto a quasi-stationary or weakly oscillatory regime.

Mira: Across both systems investigated, intermittency in the forcing progressively reshapes the attractor of the slow response system and drives it into different regimes.

Kai: System Models Investigated The work addresses this question using two coupled models of increasing complexity: a low-dimensional Lorenz-63 system and the spatially extended Kuramoto–Sivashinsky equation.

Title and authors: Kai: We're diving deeper now into what this paper actually says about how intermittency works in these fast-slow systems. The title itself, "Intermittency-induced regimes in fast-slow dynamical systems," sets the stage for understanding those transitions.

Mira: It’s really about how the nature of that interaction dictates the resulting dynamics, and they use two very different mathematical structures to test this idea. We've got a low-dimensional Lorenz system and a spatially extended Kuramoto–Sivashinsky equation.

Lev: From an engineering standpoint, it’s interesting because we see how these abstract concepts map onto actual equations that need to be solved computationally or experimentally. The L63 system is simpler, which is good for initial understanding.

Kai: Right, the paper uses the L63 system and the KSE equation to explore this relationship across different complexities. They aren't just looking at one thing; they are testing how this mechanism holds up in different mathematical settings.

Mira: And they focus on understanding how intermittent information moves between these systems evolving at different time scales, which is a key question for multiscale dynamics. This study is also motivated by a climate analogy—think of an atmospheric burst forcing an oceanic system to see what happens under those conditions.

Lev: It’s interesting that they are testing this on two such distinct models. The KSE equation, being spatially extended, introduces a lot more complexity because the dynamics aren't just point-wise; they have spatial structure to deal with.

Kai: And the objective is to figure out under what conditions intermittency can actually be induced in that slower subsystem, not just when it happens naturally in the driver.

The paper's summary: Mira: So, summarizing their core findings, the paper shows that increasing the frequency of those intermittent bursts in the fast forcing system causes a progressive displacement of the response attractor away from its unperturbed state. This displacement doesn't go on forever; it hits a limit where it just stays put.

Kai: That saturation behavior is pretty important because it tells us there’s an upper bound to how much this intermittent forcing can reshape the slow system's long-term behavior. It’s not infinite change, which is good for stability analysis.

Lev: From a hardware perspective, if we think about a real system, this saturation limit suggests that there's a point where adding more driving noise just stops causing significant structural changes in the response dynamics. The overhead of the forcing doesn't scale forever.

Mira: They are using something called the Wasserstein distance to quantify this displacement between the induced attractor and the unperturbed one. This metric is how they measure that change in shape and location.

Kai: And they show this happens in both systems, which is a strong result because it means this isn't just an artifact of one particular mathematical model; it’s a general property of these coupled dynamics.

Lev: That generality is what makes me think about what this means for real-world applications—if this scaling law holds, we can predict the limits of how much forcing can perturb a slow system.

The paper's improvements: Kai: Moving on to how the authors suggest improving this work, they focus on using extreme value statistics to better characterize those regime transitions they see. They use the variance of sampled local maxima as a proxy for detecting structural changes in the induced attractor.

Mira: They also look at the shape parameter of a fitted Generalized Extreme Value distribution, or GEV distribution. This parameter is described as being sensitive to regime transitions in both systems. It’s a diagnostic tool they use to pinpoint where the system switches behavior.

Lev: That sounds like it would be useful for experimental work because we can measure those extreme events and see if the statistical signature changes as we tune the parameters of the forcing system. It connects the math to measurable statistics on real data.

Kai: They also explore synchronization as a specific mechanism for many of these regime transitions they observe. They found that stripe-like structures in L63, in block maxima variance, and even in time-averaged phase-locking maps all appear in the same regions of the coupling strength and time-scale separation plane.

Mira: They confirm this synchronization by replacing the intermittent driver with a purely periodic forcing. When they do that, those bands saturate at a Rayleigh parameter around one point, which strongly suggests phase locking is what’s driving those transitions.

Lev: If phase synchronization is the mechanism, it means we might be looking for specific conditions in our experiments where we need to ensure the fast and slow systems are locked in a certain way to see that transition happen.

Conclusion: Kai: So, wrapping up this paper on "Intermittency-induced regimes in fast-slow dynamical systems," they show that intermittency progressively reshapes the slow system's attractor until it saturates at a limit, and they use extreme value statistics to track the regime changes.

Mira: The main implication is that we have a better way to characterize when and how this information transfer happens by looking at phase synchronization in specific parameter spaces. It gives us concrete regions where we expect these transitions to occur across different models.

Lev: For real hardware, understanding the saturation limit is crucial for designing controllers or systems that need to operate under noisy inputs; knowing the upper bound on how much noise can disrupt your slow process is valuable overhead information.

Kai: We also see a scaling law for the response delay, which follows an exponential scaling with time scale separation, suggesting this is a general property of coupled type-I intermittent dynamics. That’s a universal prediction we can test in other systems.

Mira: It really helps us connect these abstract models to physical intuition, using that climate analogy and seeing how the mathematical structures align with those observations. It’s about understanding the mechanism of information transfer itself.

Lev: I just think it's solid work because they are connecting these different aspects—the geometry of the attractor, the extreme statistics, and the phase locking—into one coherent picture for fast-slow systems.

Kai: Yeah, it’s a good piece of work. We’ll keep an eye on how this connects to other areas of nonlinear dynamics next.

Department of Physics and Astronomy, University of Bologna · Royal Meteorological Institute of Belgium · School of Physical & Mathematical Sciences, Nanyang Technological University

nlin.CD

Submitted: 2026-09-22

Updated: 2026-11-06

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 90/100

The gist: Intermittent dynamics are ubiquitous in the Earth system and often arise from the interaction of processes evolving on different time scales.

Key concepts

Attractor Displacement
This measures how much the long-term behavior (attractor) of the slow system changes when subjected to intermittent forcing. The study shows this displacement increases with burst frequency but eventually levels off, indicating a saturation point beyond which further bursts have less impact.
Generalized Extreme Value (GEV) Distribution
This statistical tool is used to analyze extreme fluctuations in the response system by fitting a specific probability distribution to its largest observed values. The shape parameter of this distribution helps researchers distinguish between different dynamic regimes, such as those associated with synchronization or chaos.
Phase Synchronization
This is identified as the primary mechanism causing regime shifts. It occurs when the phases of the fast and slow systems become locked together. The study maps where this synchronization appears in various system parameters, confirming its role in driving transitions between different dynamic states.
Delay Response Scaling
This describes how quickly a response system reacts to an input signal. The analysis reveals that the optimal time lag required for the best response scales exponentially with the time-scale separation ($ au$). This scaling is a general feature of coupled type-I intermittent dynamics.

Terminology

Summary

Intermittent dynamics are ubiquitous in the Earth system and often arise from the interaction of processes evolving on different time scales. p. 0 Intermittent bursts in a fast forcing system propagate to and reshape the dynamics of a slower response system that would otherwise settle onto a quasi-stationary or weakly oscillatory regime. p. 0 Across both systems investigated, intermittency in the forcing progressively reshapes the attractor of the slow response system and drives it into different regimes. p. 0

System Models Investigated

The work addresses this question using two coupled models of increasing complexity: a low-dimensional Lorenz-63 system and the spatially extended Kuramoto–Sivashinsky equation. The systems considered are two idealized models of increasing complexity: a two coupled Lorenz 63 (L63) system [18], following the formulation proposed in [19], in which the feedback term is suppressed; and a coupled spatially extended system based on the Kuramoto–Sivashinsky equation (KSE) [20, 21]. p. 1 The overarching objective of this work is to understand how intermittent information can be transmitted between systems evolving on different time scales. p. 1

Quantifying Attractor Displacement

Using the Wasserstein distance, the authors show that increasing the frequency of intermittent events progressively displaces the response attractor from its unperturbed counterpart up to a limit beyond which this deviation saturates. The results for both systems demonstrate this saturation behavior. p. 3 For the L63 system, the distance grows rapidly with increasing Rayleigh parameter as it approaches a limit beyond which it saturates. p. 0 Similarly, for the KSE system, the distance rises sharply as viscosity decreases (increasing burst frequency) and then plateaus. p. 3

Regime Transitions and Extreme Statistics

Varying forcing intensity and time-scale separation drives distinct regime transitions which are characterized through the variance of ensemble maxima, the power spectra of both systems, and extreme value statistics. The variance of sampled local maxima serves as a proxy to detect regime transitions and structural changes in the induced attractor. p. 4 For the L63 system, the transition to chaos occurs rather progressively as both time-scale separation and coupling strength increase. p. 4 The shape parameter of the fitted GEV distribution provides a sensitive diagnostic of regime transitions in both systems. p. 7

Mechanism of Information Transfer

The study identifies phase synchronization as the specific mechanism responsible for many of the regime transitions observed. The stripe-like structures found in the case of L63, in the variance of block maxima, and in the time-averaged phase-locking maps all occupy the same regions of the (C, tau) plane. p. 6 This synchronization is confirmed by replacing the intermittent driver with a purely periodic forcing to show that corresponding bands saturate at R ≈ 1. p. 5

Delay Response Scaling

The mutual-information-based estimate of the optimal lag shows that the response delay follows a common exponential scaling law with tau in both systems, decreasing as the two time scales become comparable. This robust scaling across different models suggests it reflects a general property of coupled type-I intermittent dynamics. p. 0 The L63 estimate is less smooth than the KSE counterpart, displaying substantial scatter and comparatively large error bars. p. 9

System-Specific Realizations

The comparison between L63 and KSE shows that the way this general mechanism is realized is strongly shaped by the internal structure of each system. In the Lorenz case, coupling strength C consistently emerges as the dominant control parameter. p. 5 The KSE system exhibits a markedly more efficient and less noisy synchronization. p. 5 In the KSE case, extremes are generically bounded (Weibull-type) rather than Gumbel-type at strong coupling. p. 7

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Robust attractor characterization for slow systems: The AI system can predict regime transitions in a slow subsystem by calculating a metric such as the Wasserstein distance between its induced attractor and its unperturbed reference attractor, showing increasing the frequency of intermittent events progressively displaces the response attractor from its unperturbed counterpart, up to a limit beyond which this deviation saturates.

  2. Extreme value prediction via Generalized Extreme Value theory: The AI can predict extreme fluctuations in response variables by fitting a GEV distribution to block maxima, allowing it to distinguish between regimes using the shape parameter of the GEV distribution, such as identifying positive ㉯ indicates a Fréchet-type, heavy-tailed distribution.

  3. Synchronization mechanism identification: The system can identify phase synchronization as the specific mechanism for regime transitions by finding regions in the coupling strength and time-scale separation where "the stripe-like structures found in the case of L63, in the variance of the block maxima, in the positive-to-negative transitions of the GEV shape parameter (with the bounded, Weibull-type regime specifically associated with synchronized states), and in the time-averaged phase-locking maps R(C, ㏋), all occupy the same regions of the (C, ㏋) plane."

  4. Optimal delay estimation: The AI can estimate the optimal time lag between driver and response signals using mutual information based estimates to determine how the optimal lag decreases as ㏋ approaches unity, i.e., as the time scales of the driver and response become comparable, providing a universal prediction for exponential type scaling of delays.

  5. System-specific regime mapping: The AI can distinguish between system behaviors by identifying that in the L63 case, coupling strength C consistently emerges as the dominant control parameter, governing the transition to chaos, whereas in KSE, the coupling strength now plays a more prominent role.

Related papers