Multiplex Interlayer Connectivity

• Multiplex interlayer connectivity is the study of patterned node alignments across network layers, shaping processes like percolation, diffusion, and synchronization.
• It employs analytical and computational models, such as supra-adjacency matrices and spectral methods, to quantify and optimize multilayer interactions.
• The concept drives advancements in applications including transportation, neuroscience, and social systems by predicting network robustness and critical transitions.

A multiplex network consists of a fixed set of entities (nodes), each of which can have multiple, distinct types of connectivity relations, realized as “layers.” Multiplex interlayer connectivity refers to the patterns, metrics, and mechanisms by which node replicas across layers are coupled, interact, or are aligned. In most formalizations, interlayer connectivity in a multiplex network is structurally defined by identity (“diagonal”) links, connecting each node in one layer to its replica in any other, but broader frameworks accommodate general, weighted, directed, adaptive, or dissimilar interlayer couplings. The precise form of interlayer connectivity critically shapes global processes such as percolation, diffusion, synchronization, link prediction, and collective behavior, distinguishing multiplex networks from single-layer or aggregated models.

1. Mathematical Formalism of Interlayer Connectivity

Multiplex networks are typically represented as a set of $M$ layers, each with $N$ aligned nodes. The canonical node-aligned multiplex is specified by

$$\mathcal{M} = \left\{\{A^{[\alpha]}\}_{\alpha=1}^M;\; C_1, \dots, C_N\right\}$$

where $A^{[\alpha]} \in \{0,1\}^{N\times N}$ are intra-layer adjacency matrices and $C_i = (c_i^{\alpha\beta})_{\alpha,\beta=1}^M$ are inter-layer coupling matrices for each node.

The most common multiplex, used in transportation, communication, and brain networks, places one-to-one, often uniform-weight, interlayer “identity edges” between each node $i$ in layer $\alpha$ and its replica in all other layers. The overall topology is encoded in the supra-adjacency matrix: $$A_{\text{multi}}(\gamma) = \operatorname{blkdiag}(A^{(1)},\ldots,A^{(L)}) + \gamma \big( 1_L 1_L^\top - I_L \big) \otimes I_N$$ where $\gamma$ is the interlayer coupling strength and $I_N$ the $N$0 identity.

In broader multilayer or interconnected networks, interlayer coupling can be non-diagonal (arbitrary off-diagonal blocks in $N$1), weighted, and/or directed. Adjacency tensors of order 4 or block matrix representations support such extensions, e.g., $N$2 with arbitrary nonzero entries (Lee et al., 2015).

A key structural variable is the interlayer degree vector of each node, capturing its activity profile across layers, and associated metrics such as overlap, participation coefficients and layer-layer Pearson degree correlations (Battiston et al., 5 May 2026).

2. Structural Metrics and Quantification

Quantitative characterization of interlayer connectivity employs both local and global measures:

• Edge overlap: $N$3; average edge overlap $N$4; pairwise layer edge overlap $N$5 (Battiston et al., 5 May 2026).
• Degree correlation: Layer-layer correlations $N$6 via Pearson coefficients, multiplexity $N$7 as the fraction of nodes active in both layers.
• Participation coefficient: $N$8 measuring the spread of node $N$9’s activity.
• Interlayer similarity: Metrics such as degree-degree correlation, betweenness-based similarity, clustering-coefficient similarity, average similarity of neighbors (ASN), and asymmetric ASN can serve as layer-layer similarity scores, which directly inform the effectiveness of information transfer, link prediction, or signaling across layers (Najari et al., 2019).

Many studies contextualize these metrics through parametrically variable dissimilarity—e.g., prime star multiplexes with relabeling parameter $$\mathcal{M} = \left\{\{A^{[\alpha]}\}_{\alpha=1}^M;\; C_1, \dots, C_N\right\}$$0 or spatially embedded multiplexes with characteristic link length $$\mathcal{M} = \left\{\{A^{[\alpha]}\}_{\alpha=1}^M;\; C_1, \dots, C_N\right\}$$1 (Allen-Perkins et al., 2019, Danziger et al., 2015).

3. Impact on Dynamical Processes and Critical Transitions

Multiplex interlayer connectivity fundamentally controls the behavior of network dynamical processes:

• Percolation and Robustness: With strict diagonal (identity) interlayer coupling, the mutual giant connected component (MGCC) is the set of nodes connected in all layers. The percolation threshold $$\mathcal{M} = \left\{\{A^{[\alpha]}\}_{\alpha=1}^M;\; C_1, \dots, C_N\right\}$$2 depends non-monotonically on spatial embedding and interlayer dependency. For spatially-embedded multiplexes, increasing the characteristic connectivity length $$\mathcal{M} = \left\{\{A^{[\alpha]}\}_{\alpha=1}^M;\; C_1, \dots, C_N\right\}$$3 initially increases $$\mathcal{M} = \left\{\{A^{[\alpha]}\}_{\alpha=1}^M;\; C_1, \dots, C_N\right\}$$4—due to the alignment and propagation of “holes” across layers—up to a critical $$\mathcal{M} = \left\{\{A^{[\alpha]}\}_{\alpha=1}^M;\; C_1, \dots, C_N\right\}$$5, beyond which $$\mathcal{M} = \left\{\{A^{[\alpha]}\}_{\alpha=1}^M;\; C_1, \dots, C_N\right\}$$6 decreases as the network approaches the random-graph limit (Danziger et al., 2015). Two percolation regimes exist: continuous (for $$\mathcal{M} = \left\{\{A^{[\alpha]}\}_{\alpha=1}^M;\; C_1, \dots, C_N\right\}$$7) and discontinuous (for $$\mathcal{M} = \left\{\{A^{[\alpha]}\}_{\alpha=1}^M;\; C_1, \dots, C_N\right\}$$8), corresponding respectively to second- and first-order transitions.
• Diffusion and Transport: The diffusion operator is given by the supra-Laplacian,

$$\mathcal{M} = \left\{\{A^{[\alpha]}\}_{\alpha=1}^M;\; C_1, \dots, C_N\right\}$$9

with $A^{[\alpha]} \in \{0,1\}^{N\times N}$0 the interlayer coupling. Eigenanalysis decouples the problem, allowing identification of interlayer and intralayer modes. Superdiffusion (relaxation faster than either layer in isolation) occurs above a critical interlayer strength, and nontrivial directionality in interlayer links can induce “jamming,” fragmenting the network into disconnected components (Bouchet et al., 26 Oct 2025, Sole-Ribalta et al., 2013, Allen-Perkins et al., 2019).

• Synchronization: Interlayer coupling, whether constant or adaptive (e.g., Hebbian plasticity), controls the onset, abruptness, and stability of synchronization. Adaptive, time-dependent interlayer links can induce first-order (explosive) transitions and hysteresis cycles not present in either single layers or static-coupling constructions (Kachhvah et al., 2020). Higher-order (hypergraph-based) intralayer structure further enhances the robustness and insensitivity of interlayer synchrony to link removals (Anwar et al., 2022).

4. Methods for Analysis, Design, and Optimization

Analytical and computational methods focus on understanding, forecasting, and optimizing the effects of multiplex interlayer connectivity.

• Spectral and Embedding Approaches: The spectrum of the supra-Laplacian governs timescales for diffusion and synchronizability. Optimization of interlayer weights to, e.g., maximize the algebraic connectivity ($A^{[\alpha]} \in \{0,1\}^{N\times N}$1), minimize spectral radius, or minimize spectral width, can be formulated as semidefinite programs (SDPs). For low budgets, uniform interlayer weights are optimal; for higher budgets, the optimal weights become non-uniform. The dual problem yields a geometric embedding whose dimension is governed by eigenvalue multiplicities, admitting a Euclidean realization of the multiplex (Shakeri et al., 2019).
• Global Efficiency and Communicability: Communication performance metrics, such as multiplex global efficiency (average inverse geodesic length) and total communicability (trace of the exponential of the supra-adjacency), are computed using supra-adjacency/path-length matrices and min-plus matrix products. Sensitivity analysis via Perron–Frobenius theory ranks intra-layer edges for reinforcement to optimize global efficiency, taking into account both intra- and interlayer pathways (Noschese et al., 2023, Noschese et al., 2024).
• Algorithmic Link Prediction: Tasks such as cross-layer link prediction exploit interlayer similarity measures, embedding-based frameworks (e.g., MulCEV), and iterative matrix-based matching (e.g., Iterative Degree Penalty, IDP). Methods fuse structural, embedding, and probabilistic evidence to recover or forecast interlayer alignments, with demonstrated superiority over monolayer or naïve approaches, especially in sparse, scale-free, weakly overlapping multiplexes (Najari et al., 2019, Tang et al., 2020, Tang et al., 2020, Wilson et al., 24 Apr 2025).

5. Functional Applications and Empirical Case Studies

Multiplex interlayer connectivity underpins model realism and performance in a spectrum of domains:

• Transport and Infrastructure: Vertex-aligned multiplexes capture multimodal transportation (e.g., airlines, urban transit), with interlayer coupling reflecting transfer ease. Empirical studies optimize route efficiency and resilience by ranking intralayer and interlayer modifications that maximize communication or minimize redundancy (Noschese et al., 2023, Noschese et al., 2024).
• Neuroscience: Analysis of the human connectome via core-periphery decomposition in multiplex (structural and functional layers) reveals integrative brain hubs otherwise missed by single-layer analyses. Aggregation of node-layer “richness” scores identifies composite cores sensitive to layer-weighting and interlayer coupling (Battiston et al., 2017).
• Social Systems: User identity mapping, interest propagation, cybercrime detection, and cross-network recommendation all leverage interlayer alignment, with method performance critically enhanced by structural interlayer similarity and embedding consistency (Tang et al., 2020, Tang et al., 2020).
• Synthetic and Model Networks: Parameterized construction schemes (e.g., prime-star relabeling, spatially embedded random graphs with variable link lengths) are used to probe the relation between interlayer dissimilarity, spectral gap, and resulting diffusive or critical phenomena (Allen-Perkins et al., 2019, Danziger et al., 2015).

6. Advanced Mechanisms: Directionality, Adaptivity, and Higher-Order Structure

Recent advances extend beyond uniform, undirected, static interlayer couplings:

• Directionality: Directed interlayer edges produce regimes such as directionality-induced jamming, where non-reciprocal coupling fragments the multiplex and suppresses diffusion. The orientation and asymmetry of interlayer links become control parameters for system-level behavior, both in toy models and real-world systems such as urban transportation (Bouchet et al., 26 Oct 2025).
• Adaptive Plasticity: Hebbian (activity-driven) adaptation of interlayer couplings induces abrupt or hysteretic transitions, which are analytically tractable under certain symmetry and mean-field assumptions (Kachhvah et al., 2020).
• Higher-Order (Hypergraph) Multiplexes: Integrating higher-order interactions in intralayer structure while retaining pairwise interlayer links enhances both intra- and interlayer synchronization, increases the spectral gap, and offers resilience to removal of interlayer coupling (Anwar et al., 2022).

7. Theoretical Insights and Open Problems

Multiplex interlayer connectivity governs the emergence of macroscopic collective behaviors not achievable in isolated or aggregated single-layer systems. Threshold phenomena (e.g., loss of monotonicity in the percolation threshold as a function of link length, first-order synchronization via adaptive interlayer coupling, and spectral “phase transitions” in optimization problems) are direct consequences of interlayer coupling design. Significant future directions include characterizing the universality classes of multiplex critical phenomena, extending core-periphery frameworks to capture higher-order interlayer link combinations, and developing scalable algorithms for real-world, large-scale, multilayer systems with heterogeneous interlayer architectures (Lee et al., 2015, Battiston et al., 5 May 2026, Battiston et al., 2017).