Unveiling the Dimensionality of Networks of Networks

Abstract: "Every object that biology studies is a system of systems." (Fran\c{c}ois Jacob, 1974). Most networks feature intricate architectures originating from tinkering, a repetitive use of existing components where structures are not invented but reshaped. Still, linking the properties of primitive components to the emergent behavior of composite networks remains a key open challenge. Here, by composing scale-invariant networks, we show how tinkering decouples Fiedler and spectral dimensions, hitherto considered identical, providing valuable insights into mesoscopic and macroscopic collective regimes.

• The paper demonstrates that modular networks exhibit decoupled spectral and Fiedler dimensions, challenging traditional homogeneous assumptions.
• It applies the Laplacian Renormalization Group framework to analyze diffusion, synchronization, and transport across various network scales with extensive numerical validation.
• The study shows that tuning base and fiber scaling laws induces anomalous Fiedler dimensions, offering new design principles for complex, modular systems.

Decoupling of Spectral and Fiedler Dimensions in Networks of Networks

Introduction

The paper "Unveiling the Dimensionality of Networks of Networks" (2510.20520) addresses the fundamental problem of how the composition of modular, scale-invariant networks leads to emergent collective behaviors, with a particular focus on the concept of network dimensionality. The authors rigorously demonstrate that, contrary to the prevailing assumption in homogeneous systems, the spectral dimension ($d_s$) and the Fiedler dimension ($d_g$) can decouple in composite, inhomogeneous, or modular networks. This decoupling has significant implications for the understanding of mesoscopic and macroscopic dynamical phenomena, such as diffusion, synchronization, and transport, in both natural and engineered systems.

Theoretical Framework: Laplacian Renormalization Group and Network Dimensionality

The analysis is grounded in the Laplacian Renormalization Group (LRG) formalism, which extends statistical mechanics concepts to complex networks. The LRG leverages the Laplacian operator $\hat{L} = \hat{D} - \hat{A}$, where $\hat{A}$ is the adjacency matrix and $\hat{D}$ is the degree matrix, to define a diffusion process on the network. The time-evolution operator $e^{-\tau \hat{L}}$ acts as a propagator, and the associated Laplacian density matrix $\hat{\rho}(\tau)$ enables the definition of entropy $S(\tau)$ and entropic susceptibility (heat capacity) $C(\tau)$ as a function of the diffusion time $\tau$.

The spectral dimension $d_s$ is extracted from the scaling of the Laplacian eigenvalue density at small eigenvalues, governing asymptotic and thermodynamic behaviors. In contrast, the Fiedler dimension $d_g$ is defined via the scaling of the smallest nonzero Laplacian eigenvalue (the Fiedler eigenvalue, $\lambda_1$) with system size, and is crucial for characterizing mesoscopic relaxation and synchronization timescales.

Scale-Invariance in Regular Fractal Networks

The authors first validate their approach on regular fractal networks, such as the Sierpinski gasket and carpet. The heat capacity $C(\tau)$ exhibits plateaus corresponding to scale-invariant regimes, with the plateau value directly related to half the spectral dimension. The scaling of the Fiedler eigenvalue with system size yields a Fiedler dimension consistent with the spectral dimension in these homogeneous cases.

Figure 1: (a) Third generation of the Sierpinski gasket. (b) Heat capacity, $C$, versus diffusion time, $\tau$, for the Sierpinski gasket of $n^\text{th}$ generation.

Emergent Dimensions in Bundled and Composite Networks

The central contribution of the paper is the analysis of bundled (network-of-networks) structures, where a "fiber" network is attached to each node of a "base" network. In the thermodynamic limit, the spectral dimension of the composite is determined by the fiber, as the base becomes negligible in the average spectral density. However, the heat capacity $C(\tau)$ for finite systems reveals multiple plateaus, each associated with the spectral dimension of a different structural component (fiber or base), and a final decay governed by the Fiedler eigenvalue.

Figure 2: (a) Dirac comb network. (b) Heat capacity, $C$, versus diffusion time, $\tau$, for the periodic Dirac comb, showing distinct plateaus for fiber and base dimensions.

The authors derive, via perturbative analysis of the Laplacian spectrum, a general expression for the Fiedler dimension of bundled networks:

$$d_g = 2 \frac{d_{f,f} + d_{f,b}}{d_{f,f} + 2 \frac{d_{f,b}}{d_{g,b}}}$$

where $d_{f,f}$ and $d_{f,b}$ are the fractal dimensions of the fiber and base, and $d_{g,b}$ is the Fiedler dimension of the base. This formula predicts that $d_g$ can be anomalous and does not, in general, coincide with either the fiber or base dimension, nor with the spectral dimension of the composite.

Figure 3: (a) Local slope of the Fiedler scaling minus the expected value given by the analytical formula, confirming the theoretical prediction across network families.

Numerical Validation and Anomalous Fiedler Dimension

Extensive numerical simulations on various composite architectures—including Dirac combs, Dirac brushes, and networks with Barabási-Albert or random tree bases—confirm the theoretical predictions. The heat capacity profiles and Fiedler eigenvalue scaling consistently reveal the presence of multiple structural regimes and the emergence of an anomalous Fiedler dimension, especially in cases where the base and fiber have different scaling laws.

Figure 4: (a) Heat capacity, $C$, versus diffusion time, $\tau$, for a periodic Dirac comb with sublinear base growth, illustrating the disappearance of the base-induced plateau.

A key result is that the Fiedler dimension can be tuned by varying the relative growth rates of the base and fiber, leading to regimes where the Fiedler dimension is dominated by the base, the fiber, or an intermediate anomalous value. The analysis also clarifies the conditions under which the perturbative approach is valid and when the low-energy spectrum is determined by the fiber rather than the base.

Implications and Future Directions

The decoupling of spectral and Fiedler dimensions in modular and composite networks has significant implications for the design and analysis of complex systems. In biological, social, and technological networks, where modularity and hierarchical composition are ubiquitous, the mesoscopic relaxation and synchronization properties cannot be inferred from the spectral dimension alone. The Fiedler dimension provides essential complementary information about large-scale connectivity and compartmentalization.

From a theoretical perspective, the results motivate a refined classification of network geometries based on their compositional rules and scaling exponents. Practically, the framework suggests design principles for synthetic systems—such as neuromorphic circuits, adaptive materials, or modular software—where function emerges from the interplay of structurally diverse units. The recursive application of the analytical results enables the characterization of hierarchical architectures at multiple scales.

The techniques developed also offer a powerful lens for analyzing large-scale empirical networks, where the interplay between function and geometry is central to understanding emergent behaviors.

Conclusion

This work rigorously establishes that the spectral and Fiedler dimensions, while coinciding in homogeneous networks, can decouple in composite, modular, or inhomogeneous structures. The emergent Fiedler dimension in such systems is governed by a nontrivial interplay between the dimensions of the constituent components, leading to anomalous scaling regimes. These findings have broad implications for the analysis, design, and understanding of complex networks across disciplines, and open new avenues for both theoretical investigation and practical application in the study of multi-scale, modular systems.