Diffusion Dynamics on Multiplex Networks

• Multiplex networks are interconnected layers where diffusion occurs simultaneously, governed by a supra-Laplacian framework that integrates intra- and inter-layer dynamics.
• The diffusion analysis reveals regimes such as superdiffusion, optimal coupling, and directionality-induced jamming, controlled by spectral properties and network structure.
• These insights inform models of contagion, innovation spread, and complex system control, with applications in social, biological, and technological domains.

A multiplex network is a system in which the same set of nodes interact through multiple types of edges, each type organized as a distinct network layer. Diffusion dynamics on multiplex networks concerns the study of how quantities such as mass, probability, information, or influence spread within and between these layers according to their combinatorial structure and interlayer coupling. Compared to single-layer (monoplex) topologies, multiplex diffusion introduces unique spectral regimes, non-trivial forms of superdiffusion, emergent patterns, and sensitivity to higher-order structure and directionality. Foundational results show that diffusion is fundamentally controlled by the spectral properties of the multiplex’s supra-Laplacian, encoding both intra- and inter-layer connectivity. Research on multiplex diffusion spans the mathematical analysis of time scales, synchronization, epidemic thresholds, pattern formation, optimality conditions, and higher-order dynamics, and it informs applications in physical, social, biological, and technological systems.

1. Mathematical Frameworks for Multiplex Diffusion

Diffusion processes on multiplex networks are typically formalized via dynamical systems built on a supra-Laplacian operator that integrates all intra- and inter-layer flows. For a multiplex with $M$ layers and $N$ nodes per layer, denote the Laplacian of layer $\alpha$ as $L^{[\alpha]}$ (for undirected layers, $$L_{ij}^{[\alpha]}=k_i^{[\alpha]}\delta_{ij} - a_{ij}^{[\alpha]}$$). Interlayer couplings are encoded as uniform or node-specific constants, or with a general interlayer adjacency $W^I$, which may itself be directed or weighted. The combined system is governed by

$\frac{d\mathbf{x}}{dt} = -\mathcal{L}\,\mathbf{x}$

where $\mathcal{L}$ is the $NM \times NM$ block-matrix supra-Laplacian, e.g., for $M=2$:

$N$0

with $N$1 the interlayer coupling. The solution can be decomposed into normal modes, $N$2, with eigenvalues $N$3 of $N$4. The conversion rate and the existence of steady states are thus controlled by the spectral properties of the supra-Laplacian (Gomez et al., 2012, Sole-Ribalta et al., 2013, Torres-Hugas et al., 2024, Cencetti et al., 2019).

For directed and asymmetric layers or interlayer links, the supra-Laplacian becomes non-symmetric, requiring analysis in terms of the real parts of its spectrum, and non-normality effects become prominent (Tejedor et al., 2017, Wang et al., 2020, Bouchet et al., 26 Oct 2025).

Higher-order multiplexes, where interactions occur not merely on nodes but also on links or higher-dimensional simplices, require generalizations such as the multiplex Hodge Laplacian, coupling via overlaps of simplices and accommodating Dirac-type dynamics (Krishnagopal et al., 2023).

2. Spectral Regimes and Superdiffusion

A central concept is superdiffusion (“super-diffusion” Editor's term)—the regime where the global mixing time on the multiplex is shorter than that for any single layer in isolation. This is rigorously characterized by the Fiedler eigenvalue $N$5 of the supra-Laplacian:

$N$6

Superdiffusion emerges through the interplay between layers, especially in the strong coupling regime, where the effective operator is the Laplacian of the (possibly weighted) average network, $N$7, with the largest $N$8 potentially exceeding $N$9 (Gomez et al., 2012, Sole-Ribalta et al., 2013, Cencetti et al., 2019, Torres-Hugas et al., 2024).

The onset and intensity of superdiffusion are determined by several structural properties:

• Minimum node strength: A structural predictor $\alpha$0 reliably forecasts superdiffusion; positive $\alpha$1 implies the worst-connected node in the multiplex average is better connected than in any single layer (Torres-Hugas et al., 2024).
• Layer dissimilarity: Minimal overlap (edge-disjointness) between layers maximizes the potential superdiffusion index, while overlap does not affect the onset threshold (Cencetti et al., 2019).
• Interlayer assignment: Negatively correlated (disassortative) matching of node replicas boosts $\alpha$2 and superdiffusion probability compared to random or positively correlated assignment (Torres-Hugas et al., 2024).
• Interlayer coupling strength: Superdiffusion typically emerges above a critical interlayer coupling $\alpha$3, which for two identical layers satisfies $\alpha$4 (Cencetti et al., 2019).
• Layer asymmetry and resource allocation: Balanced allocation of intra-layer diffusion coefficients is optimal for superdiffusion, especially when layers are structurally similar (Cencetti et al., 2019).

For directed or asymmetric interlayer links, new regimes are accessible. For instance, directionality alone (even with symmetric individual layers) can reconstruct superdiffusion, a non-monotonic “prime regime,” or, under strong directionality, a jamming transition where the multiplex fragments dynamically (Bouchet et al., 26 Oct 2025, Wang et al., 2020, Tejedor et al., 2017).

3. Directionality, Optimal Coupling, and Jamming

When at least one layer or the interlayer links are directed, the diffusion time scale becomes non-monotonic with respect to interlayer coupling (Tejedor et al., 2017, Wang et al., 2020). There exists an optimal coupling value $\alpha$5 where the algebraic connectivity (real part of the second eigenvalue of the supra-Laplacian) is maximized, yielding maximal mixing speed; both lower and higher coupling decrease performance. Analytical results describe the behavior as:

• Linear regime $\alpha$6: $\alpha$7
• Sublinear/intermediate: Maximum at $\alpha$8
• Strong coupling $\alpha$9: Convergence to $L^{[\alpha]}$0 of the (symmetrized or averaged) network (Tejedor et al., 2017, Wang et al., 2020)

When the interlayer links are highly directed, directionality-induced jamming can occur: above a critical interlayer strength, the multiplex decomposes into disconnected components in the dynamical sense, the second supra-Laplacian eigenvalue vanishes, and the network fails to reach a global steady state. The jamming order parameter $L^{[\alpha]}$1 approaches unity as this fragmentation is realized (Bouchet et al., 26 Oct 2025).

Design implications include the ability to tune coupling directionality to accelerate, confine, or “jam” diffusion, with applications in transport, information control, and robustness (Bouchet et al., 26 Oct 2025, Wang et al., 2020).

4. Beyond Linear Diffusion: Replicator and Higher-Order Dynamics

Diffusion on multiplexes extends to nonlinear and higher-order processes:

• Replicator-diffusion equations incorporate local selection and mutation with linear or nonlinear diffusion dynamics. When state fractions, rather than counts, are the variables, the normalization constraint induces non-linearities in the effective diffusion terms, and improper implementation (fixing local population sizes) introduces spurious selective pressures in favor of faster-diffusing strategies (Requejo et al., 2016).
• Reaction-diffusion (pattern formation): Multiplexes admit coarse-grained, layer-homogeneous stationary patterns absent in single-layer systems. Instability analysis reveals distinct Turing-like regimes and tricritical bifurcation points, determined by both intra- and inter-layer diffusivities (Busiello et al., 2018).
• Hyper-diffusion and higher-order Laplacians: If layer overlap is allowed to mediate higher-order interactions, e.g., via four-body (multilink) “hyper-Laplacians,” the global spectrum and relaxation times are altered, producing coupled and, above threshold strength, synchronous dynamics for all layers—even when average mass transfer across layers is forbidden (Ghorbanchian et al., 2022). Multiplex Hodge Laplacians, coupling through simplex overlap, lead to topologically controlled diffusion with Fiedler gaps (mixing times) modulated by interlayer overlap and coupling parameters (Krishnagopal et al., 2023).

5. Diffusion of Contagion, Influence, and Innovations

Models of contagion and influence propagation on multiplexes—such as SIR and SIS epidemics, linear-threshold games, and Hawkes processes—require explicit treatment of edge types, clustering, and interdependence of spreading rules:

• Complex contagions: Content-dependent threshold dynamics on clustered multiplexes reveal that clustering can suppress or promote global cascades depending on degree, and that monoplex projections fail to capture multiple phase transitions when interlayer degree correlations are high (Zhuang et al., 2016).
• Epidemic-information co-dynamics: Asymmetrically coupled SIR/SIS models on communication-contact multiplexes display optimal information transmission rates for disease suppression, phase diagrams with bifurcations, and nontrivial cross-layer feedback; awareness diffusion effectively elevates epidemic thresholds only if it induces substantial behavioral change (Wang et al., 2016, Wu et al., 2022, Mei et al., 17 Apr 2025).
• Diffusion geometry: Random-walk-based metrics on the supra-Laplacian define the mesoscale diffusion geometry of multiplexes, including the effect of random-walk type, isolated nodes, and the transition between bottlenecked and integrated regimes (Bertagnolli et al., 2020).
• Innovation and strategy adoption: Multiplex coordination games and threshold diffusion models show that additional layers can either inhibit or promote adoption, depending on utility parameters, initial conditions, and decision rules; lower bounds and convergence rates for cascades are derived analytically and verified via simulation (Ramezanian et al., 2014, Kobayashi et al., 2021).
• Information flow and inequality: Empirical evidence from social experiments demonstrates that increased network multiplexing can suppress simple diffusion, but has a non-monotonic impact on complex contagions. Distinct layers, overlapping relationships, and demographic features interplay to shape observed diffusion and can impact access to information and innovation (Chandrasekhar et al., 2024).

6. Analytical and Experimental Results

Theoretical analysis is buttressed by extensive analytical frameworks for:

• Spectral solutions of the supra-Laplacian and related operators, separating weak- and strong-coupling limits (Gomez et al., 2012, Sole-Ribalta et al., 2013, Cencetti et al., 2019).
• Finite-size effects in synthetic and real-world datasets for superdiffusion predictions, with accuracy metrics exceeding 95% based on minimum-strength criteria (Torres-Hugas et al., 2024).
• Tricritical points and bifurcations in pattern-formation regimes (Busiello et al., 2018).
• Closed-form and message-passing solutions for threshold and coordination dynamics on sparse multiplexes, with cascade prediction via local Jacobians and fixed-point equations (Kobayashi et al., 2021, Zhuang et al., 2016).
• Empirical validation: Regression of diffusion centrality predicts observed adoption in field experiments, and varying the degree of multiplexing alters collective outcomes (Chandrasekhar et al., 2024).

Key findings include the invariance of the superdiffusion threshold with respect to edge overlap, the optimality of disassortative interlayer assignment, the direct impact of directionality on attainable regimes, and the necessity of explicit multiplex modeling in systems with heterogeneous interdependency structure (Cencetti et al., 2019, Torres-Hugas et al., 2024, Wang et al., 2020, Bouchet et al., 26 Oct 2025, Chandrasekhar et al., 2024).

7. Broader Implications and Applications

Multiplex diffusion dynamics underpin the design and control of engineered and natural systems involving multiple interaction modalities: transportation, neural signal integration, information spread in multilayer societies, epidemics mediated by behavioral feedback, and infrastructure coupling. The governing principles—rooted in spectral multiplex theory, directionality-induced phenomena, non-linear and higher-order couplings—inform optimal resource allocation, robustness enhancement, and targeted interventions.

Limitations remain with respect to dynamic/topological adaptation, computational scalability for high-order and large-scale multiplexes, and complete characterization of non-linear, non-normal, and strongly heterogeneous systems. Future developments are likely to address dynamic multiplexity, temporal coupling, and multilayer reconstruction from observed diffusion patterns (Suny et al., 2018).

In summary, diffusion on multiplex networks constitutes a mature but rapidly evolving field, integrating advanced spectral theory, nonlinear dynamics, network reconstruction, and empirical validation, with wide-ranging consequences for both theory and applications.