🔢 mathematics

Macroscopic Multistability and Bifurcations in Theta-Neuron Networks with Distributed Delays

This paper derives a single delay differential equation for the order parameter of an all-to-all coupled theta-neuron network with distributed delays, analyzing the stability and bifurcations of two distinct equilibrium families to demonstrate how delay kernels influence multistability, stability switching, and the emergence of periodic dynamics.

Original authors: Lavinia Bîrdac, Alexandru Fikl, Eva Kaslik, Raluca Mureşan

Published 2026-07-21

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Lavinia Bîrdac, Alexandru Fikl, Eva Kaslik, Raluca Mureşan

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 or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a bustling city where millions of people are trying to decide when to clap their hands. Some clap in perfect unison, creating a thunderous roar; others clap at random, making a chaotic mess. This is the world of collective dynamics, a branch of science that studies how huge groups of individual units—like neurons in a brain or fireflies in a forest—coordinate their actions. At the heart of this coordination is a tricky problem: time delays. In the real world, nothing happens instantly. It takes time for a signal to travel down a nerve, for a chemical to cross a gap between cells, or for a message to reach the next person in line. These tiny lags can turn a calm rhythm into a wild, unpredictable dance, causing the group to suddenly switch from silence to a frenzy, or to get stuck in a loop of oscillating behavior. Scientists call this "stability switching," and understanding it is crucial for figuring out how brains work, how seizures start, and how to keep complex systems from crashing.

Now, picture a specific type of neuron called a theta neuron. Think of it as a tiny, perfect metronome that ticks around in a circle. When you have just one, it's easy to predict. But when you have millions of them all talking to each other, the math gets so messy it feels like trying to track every single raindrop in a storm. This is where the paper by Lavinia Bîrdac and her team comes in. They didn't just simulate a few neurons; they used a clever mathematical shortcut (a "magic trick" known as the Watanabe–Strogatz reduction) to shrink the entire infinite crowd down to a single, manageable equation. This equation acts like a "group mind," describing the average behavior of the whole network.

The researchers discovered that this "group mind" has two distinct personalities, or families of states. The first family, Type 1, lives on the edge of a circle and represents the neurons all firing together in perfect sync. The second family, Type 2, lives on a straight line inside the circle and represents a more scattered, unsynchronized state. The big question was: what happens when you introduce a delay? Does the group stay calm, or does it start wobbling?

The paper finds that the answer depends entirely on who is in charge and what kind of delay you use. If the neurons are in a specific "negative" mood (mathematically, when a parameter called $\kappa$ is negative), the group is stubborn. No matter how long the delay is, their stability doesn't change; they stay exactly as they were. However, if they are in a "positive" mood ( $\kappa > 0$ ), the delay becomes a powerful switch.

Here is where it gets wild. The team found that for the synchronized group (Type 1), a delay can suddenly make them lose their cool and start oscillating in a new, rhythmic pattern. This happens through a specific mechanism called a Hopf bifurcation, which is like a tightrope walker suddenly deciding to start juggling. The paper proves that for certain delays, this juggling act is stable and safe (a "supercritical" bifurcation), leading to a new, steady rhythm.

But the story is even more dramatic for the unsynchronized group (Type 2). Here, the delay acts like a mood ring that flips back and forth. As the delay gets longer, the group can go from unstable to stable, then back to unstable, then stable again. It's a game of musical chairs where the delay determines who is sitting and who is standing. The researchers showed that this "stability switching" only happens with a specific kind of delay called a Dirac kernel (which is like a single, sharp tap of a drum). If the delay is spread out smoothly (like a Gamma kernel), the group stays put and doesn't switch moods at all.

Through computer simulations, the authors visualized these changes. They showed that depending on the delay, the network can end up with two different stable states existing at the same time (multistability), or it can settle into a brand-new, repeating cycle of activity. They even found regions where the math predicted no stable state, yet the simulation showed a stable, dancing rhythm on the edge of the circle, suggesting there are hidden patterns the simple equations missed.

In short, this paper maps out the "mood swings" of a giant neural network. It tells us that time delays aren't just annoying lags; they are the conductors of the orchestra, capable of turning a chaotic noise into a symphony, or a steady beat into a frantic jig, depending on the exact timing and the nature of the delay. The findings are backed by rigorous math and supported by detailed computer simulations, offering a clear, albeit complex, picture of how time shapes the collective behavior of the brain's most fundamental units.

Technical Summary: Macroscopic Multistability and Bifurcations in Theta-Neuron Networks with Distributed Delays

Problem Statement
Large populations of interacting oscillatory units, such as neuronal networks, exhibit complex collective dynamics including synchronization, multistability, and transitions between quiescent and firing regimes. While microscopic descriptions involve high-dimensional nonlinear systems, macroscopic observables (e.g., mean firing rates) are often tractable via phase oscillator models. The theta neuron serves as a canonical model for type-I neuronal excitability. However, the stability and bifurcation structure of globally coupled (all-to-all) networks of identical theta neurons, specifically when synaptic interactions are mediated by distributed time delays, has not been systematically analyzed. Previous work addressed instantaneous interactions or discrete delays, but a unified framework treating both discrete and distributed delays within the thermodynamic limit was lacking.

Methodology
The authors employ the Watanabe–Strogatz reduction to analyze a network of $N$ identical theta neurons coupled all-to-all. By assuming uniformly distributed constants of motion and taking the thermodynamic limit ( $N \to \infty$ ), the high-dimensional system is reduced to a single delay differential equation (DDE) for a complex order parameter $z(t)$ . This reduction is equivalent to the Ott–Antonsen ansatz.

Key methodological steps include:

Model Formulation: The synaptic current is modeled as a convolution of the network activity with a delay kernel $l_\tau$ . The authors define a general class of "admissible" delay kernels that includes discrete Dirac delays and continuous distributions (e.g., Gamma kernels) via a reference probability measure.
Dimensional Reduction: The complex DDE for $z(t)$ is converted into a two-dimensional real system in polar coordinates $(\rho, \phi)$ .
Equilibrium Classification: The equilibria of the reduced system are classified into two geometrically distinct families:
- Type 1: Equilibria located on the unit circle ( $|z|=1$ ), corresponding to fully synchronized states.
- Type 2: Equilibria located on the real axis ( $z \in (-1, 1)$ ), corresponding to nonsynchronized stationary phase distributions.
Stability Analysis: The local stability of these equilibria is determined by deriving scalar characteristic equations involving the Laplace–Stieltjes transform of the delay kernel. The authors analyze the roots of these equations to determine stability criteria and conditions for Hopf bifurcations.
Kernel Specifics: The analysis is applied to three specific kernel families: the Dirac kernel (discrete delay), the weak Gamma kernel, and the strong Gamma kernel.
Numerical Validation: Numerical simulations of the reduced system with discrete delays are performed to verify analytical predictions, illustrating phase portraits, basins of attraction, and coexistence of attractors.

Key Contributions and Results

Unified Framework: The paper provides a unified analytical framework for treating both discrete and distributed delays in theta-neuron networks using the Laplace–Stieltjes transform of the delay kernel.
Equilibrium Geometry: The study rigorously classifies equilibria into Type 1 (unit circle) and Type 2 (real axis), deriving explicit polynomial conditions for their existence in the $(\kappa, \eta)$ parameter plane.
Type 1 Stability (Unit Circle):
- Equilibria on the upper semicircle are always unstable.
- Equilibria on the lower semicircle exhibit delay-independent stability when the coupling strength $\kappa < 0$ .
- Destabilization of Type 1 equilibria occurs only when $\kappa > 0$ and $\eta < 0$ . In the discrete-delay case, this leads to a Hopf bifurcation at a critical delay, which the authors prove to be supercritical (yielding stable limit cycles).
Type 2 Stability (Real Axis):
- The delay-free saddle point remains unstable for all admissible delay kernels.
- The delay-free center exhibits delay-dependent stability: it becomes asymptotically stable for small delays if $\kappa > 0$ and unstable if $\kappa < 0$ .
- Kernel Dependence: For weak and strong Gamma kernels, the Type 2 center does not undergo Hopf bifurcations in the analyzed parameter range. However, for the Dirac kernel, an infinite sequence of Hopf thresholds exists, leading to stability switching (alternating between stable and unstable states) as the delay increases.
Multistability and Coexistence: The analysis reveals regimes where multiple stable attractors coexist. Specifically, the numerical simulations demonstrate the coexistence of stable equilibria and stable periodic orbits, as well as the coexistence of multiple stable cycles, depending on the delay magnitude and parameter regime.
Stability Switching: The paper explicitly maps the regions in the parameter space where stability switching occurs for Type 2 equilibria under discrete delays, identifying boundaries where the system regains stability after initial destabilization.

Significance and Claims
The authors claim that their work provides the first systematic local stability and bifurcation analysis for all-to-all coupled identical theta-neuron networks with distributed synaptic delays. The primary significance lies in:

Macroscopic Insight: The results offer macroscopic stability predictions for infinite networks, linking microscopic parameters to collective behaviors like synchronization (Type 1) and desynchronization (Type 2).
Delay Effects: The study demonstrates that delays are not merely perturbations but can fundamentally alter the system's dynamics by inducing stability switching, creating multistability, and generating delay-induced periodic dynamics.
Kernel Sensitivity: The findings highlight that the qualitative behavior of the network is highly sensitive to the choice of delay kernel (e.g., discrete vs. distributed), with discrete delays capable of inducing complex switching behaviors not present in the Gamma kernel cases studied.

The paper concludes that while the analysis is local and restricted to the invariant family selected by the uniform distribution of constants of motion, it successfully captures the emergence of collective oscillations and the complex interplay between equilibrium branches and delay-induced instabilities. Future work is suggested to extend these results to nonuniform constants of motion, heterogeneous populations, and finite-size effects.

More like this