Triadic percolation on multilayer networks

This paper introduces the multilayer triadic percolation (MTP) model, which generalizes triadic interactions to multilayer networks and reveals a significantly richer dynamical landscape—including Neimark-Sacker bifurcations, pseudo-periodic oscillations, and period-two cycles—compared to the chaotic behavior observed in single-layer systems.

The Gist

Imagine a vast, bustling city made of two different neighborhoods, Neighborhood A and Neighborhood B. In this city, the "roads" (connections) between buildings don't just exist or disappear on their own; they are controlled by "traffic wardens" (regulator nodes).

This paper introduces a new way of studying how these cities function when the wardens can do two things:

1. Watch their own neighborhood: A warden in Neighborhood A can decide to open or close a road between two other buildings in Neighborhood A.
2. Watch the other neighborhood: A warden in Neighborhood A can also decide to open or close a road between two buildings in Neighborhood B.

The researchers call this system Multilayer Triadic Percolation (MTP). They wanted to see what happens when these two neighborhoods talk to each other through these wardens, compared to a city where wardens only watch their own block.

Here is the breakdown of their findings using simple analogies:

1. The Old Way: A Single Neighborhood

In previous studies (the "single-layer" model), researchers looked at just one neighborhood. They found that if the wardens are too strict or too chaotic, the city's "giant connected area" (the part of the city where everyone can reach everyone else) doesn't just settle down. Instead, it starts to oscillate.

• The Analogy: Imagine the size of the connected city growing and shrinking like a breathing lung. Sometimes it doubles its rhythm, then quadruples, until it becomes completely unpredictable and chaotic. This is like a drumbeat that speeds up until it's a mess of noise.
• The Catch: In this single neighborhood, this chaotic breathing only happened if the wardens had a "negative" attitude (they liked to close roads). If they were only "positive" (only opening roads), the city would just suddenly collapse or stay steady.

2. The New Discovery: Two Neighborhoods Talking

The authors added a second neighborhood and let the wardens from both sides influence the roads in both places. This created a much more complex dance.

The Big Surprise: The "Spiral Dance" (Neimark–Sacker Bifurcation)
When the two neighborhoods interact, the city doesn't just speed up its breathing rhythm. It starts to spin.

• The Analogy: Imagine a figure skater. In the single neighborhood, the skater just spins faster and faster until they fall (chaos). But in the two-neighborhood system, the skater starts to wobble in a beautiful, spiraling circle. The size of the connected city doesn't just go up and down; it traces a complex, looping path that can last for a very long time or look like a never-ending, slightly irregular pattern.
• Why it matters: This "spiral" behavior is impossible in a single neighborhood. It only happens because the two layers are influencing each other.

The Second Surprise: Chaos Without Negativity
In the old model, you needed "negative" wardens (those who close roads) to get the system to oscillate wildly.

• The Analogy: In the two-neighborhood city, the researchers found that even if all the wardens are "positive" (they only want to open roads and help), the city can still start to oscillate between being fully active and completely silent.
• The Result: The city can flip-flop between "All lights on" and "Total blackout" in a regular rhythm, even without anyone trying to shut things down. This is something that never happened in the single-neighborhood model.

3. The "Traffic Light" Effect

The researchers mapped out exactly when these changes happen. They found that the rules are tricky:

• Sometimes, making the wardens more strict (adding more negative regulation) actually stabilizes the city, stopping the chaos. This is counter-intuitive; usually, we think more rules mean more chaos, but here, more rules can calm the system down.
• The system has three "tipping points":
  1. The Upper Limit: Where the stable city first starts to wobble.
  2. The Middle Ground: Where the city might settle back down after a period of chaos.
  3. The Lower Limit: Where the city finally collapses into silence.

Summary

Think of this paper as discovering a new type of traffic system.

• Old System: One layer of traffic lights. If they get confused, the traffic jams and clears in a simple, predictable, or chaotic rhythm.
• New System: Two layers of traffic lights talking to each other. This creates a spiral dance of traffic flow that is richer and more complex. It allows for wild swings between "gridlock" and "free-flow" even if all the lights are programmed to be helpful.

The authors conclude that real-world systems—like the human brain (where glial cells regulate neurons) or ecosystems—are likely more like this two-layer system than the simple one-layer model. Understanding this "spiral dance" helps us see why these complex systems behave so dynamically and unpredictably.

Technical Summary

Technical Summary: Triadic Percolation on Multilayer Networks

Problem Statement
Real-world systems, such as brain networks, climate systems, and ecological networks, are increasingly recognized as being better represented by higher-order interactions rather than simple dyadic networks. While triadic percolation models have previously been established to describe how regulatory nodes modulate interactions between other nodes (turning percolation into a dynamical system with period-doubling and routes to chaos), these models have been restricted to single-layer networks and hypergraphs. In many real scenarios, such as the interaction between glia and neurons, the system comprises multiple layers with distinct functional roles. The existing single-layer framework fails to capture the complex interplay between intra-layer and inter-layer regulatory interactions, which may lead to dynamical behaviors not observed in simpler topologies.

Methodology
The authors propose the Multilayer Triadic Percolation (MTP) model, a generalization of triadic percolation to networks with two layers (Layer A and Layer B). The model incorporates:

1. Structural Networks: Random networks in each layer with defined degree distributions (assumed Poisson for analytical tractability).
2. Triadic Regulatory Interactions: Directed signed interactions where nodes regulate structural links. These are categorized as:
   • Intralayer: Nodes within a layer regulate links within the same layer.
   • Interlayer: Nodes in one layer regulate links in the other layer.
3. Dynamical Algorithm: The evolution of the Giant Component (GC) is defined by an iterative two-step process:
   • Step 1 (Percolation): Determine the active nodes (those belonging to the GC) based on the current state of structural links.
   • Step 2 (Regulation): Update the probability of link retention ($p(t)$) based on a Boolean rule involving the activity of positive and negative regulators from both layers, plus stochastic noise.

The dynamics are mathematically encoded as a two-dimensional map relating the fraction of active nodes in each layer ($R_A, R_B$) to the link retention probabilities of the previous time step. This contrasts with single-layer triadic percolation, which is captured by a one-dimensional map. The authors analyze the stability of fixed points using the Jacobian matrix of this map to identify bifurcation thresholds and characterize the route to chaos.

Key Contributions and Results
The study reveals that the simultaneous presence of intra- and inter-layer regulations induces novel dynamical states absent in single-layer counterparts:

1. Neimark–Sacker Bifurcation: Unlike single-layer systems, the MTP model exhibits a Neimark–Sacker bifurcation. This occurs when a pair of complex-conjugate eigenvalues of the Jacobian cross the unit circle, leading to the emergence of invariant cycles. The system displays oscillations of arbitrarily large periods or pseudo-periodic (quasi-periodic) oscillations. This phenomenon requires both positive and negative regulatory interactions and the coupling of intra- and inter-layer dynamics.
2. Period-Two Oscillations without Negative Regulation: In single-layer networks, period-doubling oscillations typically require negative regulatory interactions. The MTP model demonstrates that period-two oscillations can emerge even when triadic interactions are exclusively positive, provided they are exclusively interlayer. This occurs due to the decoupling of even and odd time steps in the iterative map, allowing for convergence to different fixed points based on initial conditions.
3. Complex Phase Diagrams: The authors characterize three critical thresholds ($p^u_c, p^s_c, p^l_c$) governing the stability of the non-trivial fixed point. They find that the nature of the bifurcation (discontinuous, period-doubling, or Neimark–Sacker) depends sensitively on the balance between intra- and inter-layer regulation strengths. Notably, increasing the strength of negative regulation can sometimes stabilize the steady state, a non-monotonic behavior not observed in single-layer models.
4. Generalization: While the primary analysis focuses on $M=2$ layers, the framework is shown to generalize to $M$-layer networks via $M$-dimensional maps, suggesting even richer dynamical behaviors for higher dimensions.

Significance
The paper claims that the MTP model offers a more realistic framework for understanding critical phenomena in interconnected systems where regulatory mechanisms span multiple functional layers. By integrating percolation theory with nonlinear dynamical systems, the work demonstrates that multilayer structural and regulatory interactions can generate a rich variety of dynamical states, including quasi-periodic oscillations and complex bifurcation scenarios. These findings provide new insights into the dynamics of real-world systems such as brain networks (glia-neuron interactions) and ecological networks, suggesting that the multilayer nature of these systems is crucial for their time-dependent critical behavior. Furthermore, the authors posit that this framework could serve as a tool for adaptively controlling network activity due to its rich dynamical repertoire.