Algebraic characterization of binary graphs

One of the fundamental concepts in the statistical mechanics field is that of ensemble. Ensembles of graphs are collections of graphs, defined according to certain rules. The two most used ensembles in network theory are the microcanonical and the grandcanonical (whose definitions mimick the classical ones, originally proposed by Boltzmann and Gibbs), even if the latter is far more used than the former to carry on the analytical calculations. For binary (undirected or directed) networks, the grandcanonical ensemble is defined by considering all the graphs with the same number of vertices and a variable number of links, ranging from 0 to the maximum: N(N-1)/2 for binary, undirected graphs and N(N-1) for binary, directed graphs. Even if it is commonly used almost exclusively as a tool to calculate the average of some topological quantity of interest, its structure is so rich to deserve an analysis on its own. In this paper a logic-algebraic characterization of the grandcanonical ensemble of binary graphs is provided.

💡 Research Summary

The paper provides a rigorous logical‑algebraic description of the grand‑canonical ensemble of binary graphs, both undirected and directed. Starting from the representation of a graph as a 0‑1 adjacency matrix, the authors define the set Ω_N of all such matrices for a fixed number of vertices N. This set constitutes the grand‑canonical ensemble 𝔾_N, containing every possible edge configuration: 2^{N(N‑1)/2} graphs for the undirected case and 2^{N(N‑1)} for the directed case.

Three algebraic structures are introduced on 𝔾_N. First, Boolean operations are defined element‑wise: the join (⊔) corresponds to the logical OR of adjacency entries, the meet (⊓) to the logical AND, and complementation (¬) to the logical NOT, which yields the complement graph. These operations satisfy associativity, commutativity, distributivity, and possess identity elements (the empty graph and the complete graph), thereby turning (𝔾_N, ⊔, ⊓) into a finite Boolean lattice.

Second, the partial order induced by entry‑wise inequality (A ⊆ B ⇔ A_{ij} ≤ B_{ij} for all i, j) coincides with the subgraph relation, making the lattice a graded poset where each level contains all graphs with the same number of edges. The level sets L_M = {G ∈ 𝔾_N | |E(G)| = M} are precisely the micro‑canonical ensembles.

Third, because the Boolean lattice is closed under countable unions, intersections, and complements, 𝔾_N forms a σ‑algebra. This permits a probability measure p(G) to be assigned to each graph, enabling the definition of events, conditional probabilities, and the Shannon entropy S = −∑_{G} p(G) log p(G). In the grand‑canonical ensemble the probability of a graph is taken as a Bernoulli product p(G) = θ^{|E(G)|}(1−θ)^{E_max−|E(G)|}, where θ is the link‑presence parameter. The lattice structure ensures that differentiation with respect to θ and summation over graphs are compatible operations, simplifying the calculation of ensemble averages, variances, and higher moments.

The authors discuss how this formalism clarifies the relationship between micro‑canonical and grand‑canonical ensembles: the former are simply the rank‑M slices of the lattice, while the latter are weighted mixtures of these slices. By expressing constraints (e.g., the presence of a particular motif) as Boolean expressions, one can define constrained sub‑ensembles directly within the lattice. Moreover, the Boolean‑algebraic viewpoint provides a natural framework for entropy maximization under such constraints, leading to analytically tractable maximum‑entropy graph models.

Finally, the paper argues that this logical‑algebraic perspective not only deepens the theoretical understanding of binary graph ensembles but also offers practical tools for network modeling, inference, and the extension to more complex settings such as multilayer, weighted, or dynamic graphs. The authors suggest that future work could explore these extensions, leveraging the same lattice‑σ‑algebra foundation.