The Ihara zeta function as a partition function for network structure characterisation

Jianjia Wang, Edwin R. HancockScientific 2024 · DOI

Abstract

Statistical characterizations of complex network structures can be obtained from both the Ihara Zeta function (in terms of prime cycle frequencies) and the partition function from statistical mechanics. However, these two representations are usually regarded as separate tools for network analysis, without exploiting the potential synergies between them. In this paper, we establish a link between the Ihara Zeta function from algebraic graph theory and the partition function from statistical mechanics, and exploit this relationship to obtain a deeper structural characterisation of network structure. Specifically, the relationship allows us to explore the connection between the microscopic structure and the macroscopic characterisation of a network. We derive thermodynamic quantities describing the network, such as entropy, and show how these are related to the frequencies of prime cycles of various lengths. In particular, the n-th order partial derivative of the Ihara Zeta function can be used to compute the number of prime cycles in a network, which in turn is related to the partition function of Bose-Einstein statistics. The corresponding derived entropy allows us to explore a phase transition in the network structure with critical points at high and low-temperature limits. Numerical experiments and empirical data are presented to evaluate the qualitative and quantitative performance of the resulting structural network characterisations.

Introduction: Bridging Two Worlds

In network science, a persistent challenge lies in connecting a network's "microscopic" connectivity—the specific arrangement of nodes and edges—with its "macroscopic" properties, such as robustness, complexity, or functional classification. While statistical mechanics provides a robust framework for describing physical systems through macroscopic variables like entropy, applying these concepts to graphs has often relied on loose analogies. The primary hurdle has been defining what constitutes the physical "microstates" of a graph.

This paper provides a formal mathematical bridge between algebraic graph theory and statistical mechanics. By establishing a link between the Ihara Zeta function and the partition function, the authors demonstrate that a network's structure can be rigorously analyzed as a thermodynamic system. In this framework, the network's "prime cycles"—loops that do not backtrack or repeat—act as the microscopic states, allowing for the derivation of thermodynamic quantities that serve as a unique fingerprint for network topology.

Methodology: Cycles as Microstates

The core innovation of this research is the mapping of graph cycles to energy states in a Bose–Einstein system. The authors utilize the Hashimoto matrix (a non-backtracking matrix) to represent the network as an oriented line graph. This transformation is essential for calculating the Ihara Zeta function, which counts the number of prime cycles of various lengths.

Figure 1: The transformation process from a simple undirected graph (a) to a symmetric directed graph (b) and finally to an oriented line graph (c). This sequence allows the network to be represented in a way that facilitates the calculation of the Ihara Zeta function via the Hashimoto matrix.

By treating these prime cycles as "particles" (specifically bosons) and their lengths as "energy levels," the authors derive a Network Partition Function. From this, they calculate fundamental thermodynamic quantities:

  1. Network Entropy ($S$): A measure of the structural complexity and diversity of cycles within the graph.
  2. Internal Energy ($E$): Related to the distribution of cycle lengths.

This mapping allows researchers to identify "phase transitions"—sudden, dramatic shifts in network behavior—by locating critical points where the partition function reaches specific thresholds.

Key Results and Topology Fingerprinting

The framework was validated using both synthetic models and real-world datasets. One of the most significant findings is that different network topologies occupy distinct, non-overlapping regions in the entropy space.

Figure 2: Histograms demonstrating the clear separation of Random Graphs, Scale-free Networks, and Small-world Networks based on the derived thermodynamic characterizations. This separation proves that the Ihara-based entropy acts as a reliable "topology fingerprint."

As shown in the experimental results, scale-free networks (which typically possess many star-like hubs but fewer cycles) exhibit lower entropy compared to small-world networks, which are rich in local clusters and cycles. This sensitivity allows the method to detect structural anomalies that traditional metrics might miss.

Real-World Applications

  • Financial Markets: Analyzing the S&P 500 from 2010–2020, the network entropy successfully identified sharp peaks corresponding to global crises, including the 2010 Flash Crash and the onset of COVID-19. These peaks represent sudden reorganizations of market correlations.
  • Neuroscience: The entropy components differentiated between Alzheimer’s Disease (AD) patients and healthy controls. Healthy brains exhibited higher entropy, suggesting a more complex and robust cycle structure compared to the degraded networks of AD patients.
  • Biological Classification: Using entropy as a feature in Principal Component Analysis (PCA), the authors clustered different tumor mutation networks (Ovarian, Uterine, and Lung cancers), demonstrating the utility of network thermodynamics in bioinformatics.

Conclusion and Future Directions

By unifying the Ihara Zeta function with the partition function, this work provides a rigorous physical interpretation for network thermodynamics. The resulting entropy and energy metrics offer a powerful tool for classifying networks and detecting phase transitions in complex systems.

While highly effective at distinguishing broad classes of networks, future work will focus on increasing sensitivity to subtle structural variations within specific datasets, such as early-stage disease progression. Additionally, the authors plan to investigate whether other statistical models—such as Fermi–Dirac or Maxwell–Boltzmann statistics—might reveal different hidden layers of network architecture.

Summary updated

Abstract

Statistical characterizations of complex network structures can be obtained from both the Ihara Zeta function (in terms of prime cycle frequencies) and the partition function from statistical mechanics. However, these two representations are usually regarded as separate tools for network analysis, without exploiting the potential synergies between them. In this paper, we establish a link between the Ihara Zeta function from algebraic graph theory and the partition function from statistical mechanics, and exploit this relationship to obtain a deeper structural characterisation of network structure. Specifically, the relationship allows us to explore the connection between the microscopic structure and the macroscopic characterisation of a network. We derive thermodynamic quantities describing the network, such as entropy, and show how these are related to the frequencies of prime cycles of various lengths. In particular, the n-th order partial derivative of the Ihara Zeta function can be used to compute the number of prime cycles in a network, which in turn is related to the partition function of Bose-Einstein statistics. The corresponding derived entropy allows us to explore a phase transition in the network structure with critical points at high and low-temperature limits. Numerical experiments and empirical data are presented to evaluate the qualitative and quantitative performance of the resulting structural network characterisations.
BibTeX
@inproceedings{wang2024ihara,
  author = {Jianjia Wang and Edwin R. Hancock},
  title = {{The Ihara zeta function as a partition function for network structure characterisation}},
  booktitle = {Scientific},
  year = {2024},
  doi = {10.1038/s41598-024-68882-x},
}
Figure