Multiplex Percolation Framework

• Multiplex percolation is a framework unifying interconnected network layers, characterized by interdependencies and distinctive phase transitions.
• It employs generating functions and message-passing algorithms to analyze connectivity and predict hybrid and continuous transitions under failure.
• The framework has practical applications in modeling cascading failures, brain networks, and infrastructure robustness while revealing novel universality classes.

The multiplex percolation framework unifies and extends classical percolation theory to systems in which a fixed set of nodes engages in multiple, distinct types of interactions—each encoded as an independent “layer” or edge type—while asking that systemic connectivity or robustness be defined relative to the interplay of all these layers. Unlike monoplex percolation, where the emergence of macroscopic connectivity reduces to a single network structure, multiplex percolation incorporates strong inter-layer dependencies, non-local viability conditions, and diverse dynamical rules that yield rich and fundamentally new phase behaviors, including hybrid (first–second order) transitions, multiple thresholds, and novel universality classes.

1. Foundational Definitions and Classes of Multiplex Percolation

In multiplex percolation, the system is represented by a set of nodes present in $M$ layers, each with its own network topology $G^{[\alpha]}$ ($\alpha=1,\dots,M$). Edges within each layer represent distinct interaction types, and percolation processes explore the formation and stability of large-scale connectivity under random removal of nodes or links.

Frameworks for multiplex percolation are classified according to connectivity requirements:

• Strong (Viable/Mutually Connected) Percolation: A node belongs to the giant component if and only if it is connected via paths in every layer simultaneously to all other nodes in the component (“mutually connected”). This condition is nonlocal and enforces systemic interdependence (Baxter et al., 2016).
• Weak Percolation: Looser variants require only that each node have at least one neighbor in each layer (pruning percolation) or become (re-)activated if it gains support in all layers (bootstrap percolation) (Baxter et al., 2013).
• k-Core and Higher-Order Percolation: Stronger requirements entail that each node maintain at least $k_\alpha$ neighbors in layer $\alpha$ (vector $k$-core), or that group (hyperedge) interactions are present (Azimi-Tafreshi et al., 2014, Sun et al., 2021).
• Coupled Cascade and Dependency Schemes: Directed dependency links, history-dependent prunings, or cascading processes further modulate the percolation transition (Niu et al., 2016, Li et al., 2020).

The formal framework adopts generating functions, self-consistent message-passing equations, or recursive pruning/growth algorithms to characterize macroscopic connectedness.

2. Mathematical Formalism and Core Equations

The analytical machinery is typically built on the locally tree-like approximation, with multivariate generating functions for the degree distributions across all layers:

• For the joint degree distribution $P(q_1,\dots,q_M)$, define

$$G_0(x_1,\dots,x_M) = \sum_{q_1,\dots,q_M} P(q_1,\dots,q_M)\prod_{\alpha=1}^M x_\alpha^{q_\alpha}$$

and marginal/excess generating functions for each layer (e.g., $G_1^{[\alpha]}$).

Strong Percolation (Mutually Connected Giant Component)

Let $X_\alpha$ denote the probability that following a layer-$G^{[\alpha]}$0 edge leads to an infinite subtree whose nodes each also possess connectivity in all other layers:

$G^{[\alpha]}$1

The giant component size is then

$G^{[\alpha]}$2

The onset and nature of the transition are determined by linear stability and tangency (Jacobian) conditions (Baxter et al., 2016).

Weak Percolation (Bootstrap and Pruning)

For weak pruning percolation (WPP), the probability $G^{[\alpha]}$3 that a layer-$G^{[\alpha]}$4 edge leads to a surviving node is governed by

$G^{[\alpha]}$5

For weak bootstrap percolation (WBP), seeds become active and non-seeds require support from each layer (Baxter et al., 2013).

Directed Dependency and Cascade Models

Introducing a directed dependency (e.g., $G^{[\alpha]}$6, distribution of incoming dependencies), transitions depend on $G^{[\alpha]}$7 (fraction with zero dependencies) and $G^{[\alpha]}$8 (fraction with at most one), leading to thresholds (Niu et al., 2016):

$G^{[\alpha]}$9

Bond Percolation and Multilayer Overlap

The critical phenomena also hinge on link overlap and interlayer degree–degree correlations. Overlap modifies generating functions and can shift transitions to being continuous or hybrid discontinuous, depending on the overlap fraction and correlation structure (Cellai et al., 2013, Min et al., 2014, Cellai et al., 2016).

3. Phase Transitions and Universality Classes

Multiplex percolation displays both continuous (second-order) and hybrid (first–second order) transitions:

• Hybrid Transitions: In strong multiplex percolation ($\alpha=1,\dots,M$0), the emergence of the giant component is typically hybrid: at the critical point, the order parameter jumps discontinuously from zero to a large value, with mean-square-root scaling above threshold (Battiston et al., 5 May 2026, Baxter et al., 2016).
• Continuous Transitions: Weak percolation, and strong percolation in the presence of complete overlap or certain parameter regimes, can exhibit standard continuous percolation transitions.
• Tricritical Points and Crossover: Partial interdependence, heterogeneity in dependency parameters, or varying overlap interpolates between continuous and hybrid regimes, giving rise to tricritical points (Cellai et al., 2013, Niu et al., 2016, 2612.05163).
• Multiple Transitions and Multimodality: Anticorrelated or partially overlapping layers in multiplex networks can generate multiple percolation thresholds, each layer percolating at distinct values of the control parameter (Hackett et al., 2015).
• Universality: The upper critical dimension for multiplex percolation is $\alpha=1,\dots,M$1, distinct from ordinary percolation ($\alpha=1,\dots,M$2); exponents and fractal dimensions differ systematically from ordinary percolation even for $\alpha=1,\dots,M$3 (Grassberger, 2015).

4. Algorithmic and Computational Approaches

• Message-Passing and Belief Propagation: Self-consistent cavity equations or Belief-Propagation systems capture both average and large-deviation properties of the giant component under locally tree-like assumptions (Cellai et al., 2016, Bianconi, 2018). These methods extend to networks with overlap, generalized dependencies, and directed links.
• SOS Mapping and Surface-Growth Interpretation: On duplexes, viable clusters and cascades can be mapped to solid-on-solid (SOS) surface-growth processes, enabling ultra-fast algorithms that compute the viable landscape in $\alpha=1,\dots,M$4 time (Grassberger, 2015).
• Heuristics for Optimal Percolation: Multiplex optimal percolation—identifying minimal dismantling sets—uses extensions of simulated annealing, greedy centrality-based heuristics, and leaf-removal refinements. The size and composition of minimum sets are sensitive to overlap and interlayer correlations (Osat et al., 2017).
• Self-organized and History-dependent Methods: Algorithms for risk-perception, epidemic manageability, or brain network robustness interpret percolation as an evolving, memory-dependent process on multiplex layers (Bagnoli et al., 2014, Li et al., 2020).

5. Impact of Structural Correlations and Overlap

Multiplex percolation is highly sensitive to interlayer correlations:

• Edge Overlap: Even limited overlap can convert a hybrid transition to continuous, facilitate robustness, or fundamentally alter threshold locations (Cellai et al., 2013, Min et al., 2014, Bianconi et al., 2016).
• Degree-Degree Correlation: Positive interlayer correlations (nodes having high degree in multiple layers) generally raise robustness; negative correlations or anticorrelation can induce multiple transitions (Hackett et al., 2015, Osat et al., 2017).
• Directed or Weighted Dependencies: Directed dependencies, non-trivial support hierarchies or history dependence can localize abrupt transitions and govern the stability or instability regime boundary (Niu et al., 2016, Li et al., 2020).

6. Representative Results and Applications

Key physical insights and applications arising from the multiplex percolation framework:

• Catastrophic Cascades: Hybrid transitions generate abrupt collective failures; mean-field calculations predict order parameters, but actual cascade scaling deviates from mean-field even on ER graphs (Grassberger, 2015).
• Brain, Transportation, Infrastructure: The framework successfully models functional vulnerability in neural, inter-airline, inter-utility, and combined communication infrastructures, often outperforming single-layer analyses (Li et al., 2020, Bianconi et al., 2016).
• Weak Percolation and Contagion: Weak bootstrap percolation on multiplexes displays critical windows where repair or seeding causes abrupt (dis)appearance of macroscopic connectivity, of relevance to epidemic and infrastructure recovery (Baxter et al., 2013, Baxter et al., 2016).
• Clique Percolation and Community Structure: Multiplex clique percolation extends community detection to require simultaneous coherence in $\alpha=1,\dots,M$5 layers, raising thresholds and illuminating overlapping, multiplex-cohesive communities (Afsarmanesh et al., 2016).
• Robustness Quantification: Metrics such as “safeguard centrality,” large-deviation rate functions, and optimal dismantling sets uniquely capture multiplex network fragility and the centrality of specific nodes under realistic failure scenarios (Coghi et al., 2018, Bianconi, 2018, Osat et al., 2017).
• Higher-Order Interactions: Extension to hypergraphs and simplicial complexes relates multiplex percolation with $\alpha=1,\dots,M$6-core and higher-order percolation phenomena, unifying multiple domains under a common analytic umbrella (Sun et al., 2021).

7. Open Problems and Theoretical Advances

• Generalization to Arbitrary Layer-Number and Partial Overlap: Analytical tractability rapidly diminishes with growing $\alpha=1,\dots,M$7 and arbitrary overlap structure; unifying frameworks capable of dealing with arbitrary overlap and directed dependencies are an active area (Cellai et al., 2016, Min et al., 2014).
• Interplay with Dynamics and Non-Equilibrium Transitions: Coupling of percolation with other dynamics (diffusion, reaction, games, evolutionary processes) on multiplexes presents further nontrivial emergent phenomena (Battiston et al., 5 May 2026).
• Classification of Critical Exponents and Universality Classes: The mapping between geometric (percolation) and dynamical (cascade, surface growth) exponents, and the classification of hybrid and multicritical phenomena in multiplex settings, remains a lively field.

The multiplex percolation framework thus provides a unified, technically rigorous foundation for understanding connectivity, robustness, epidemic spreading, and failure cascades in layered, interdependent, or multimodal complex systems. It captures a spectrum of critical behaviors and practical vulnerabilities inaccessible to classical single-network percolation and is central to ongoing research in multilayer network science (Baxter et al., 2016, Battiston et al., 5 May 2026, Grassberger, 2015).