Modeling scientific-citation patterns and other triangle-rich acyclic networks

We propose a model of the evolution of the networks of scientific citations. The model takes an out-degree distribution (distribution of number of citations) and two parameters as input. The parameters capture the two main ingredients of the model, the aging of the relevance of papers and the formation of triangles when new papers cite old. We compare our model with three network structural quantities of an empirical citation network. We find that an unique point in parameter space optimizing the match between the real and model data for all quantities. The optimal parameter values suggest that the impact of scientific papers, at least in the empirical data set we model is proportional to the inverse of the number of papers since they were published.

💡 Research Summary

The paper presents a generative model for the evolution of scientific citation networks, focusing on two empirically observed mechanisms: the aging of a paper’s relevance and the formation of citation triangles when a new paper cites two already‑connected older papers. The model takes as input the empirical out‑degree distribution (the number of references each new paper makes) and two parameters. The first parameter α governs an aging function f(t) ∝ (t + τ)^‑α, which reduces the probability of citing a paper as the number of papers published since its appearance grows. The second parameter β controls the probability that a new paper will create a triangle by simultaneously citing a pair of older papers that already cite each other.

In each growth step a new node is added with a pre‑specified out‑degree k_out drawn from the real distribution. Candidate older nodes are sampled with weights given by the aging function; then, for any pair among the selected nodes that already share a citation link, an additional citation is added with probability β, thereby generating a triangle while preserving the network’s acyclic nature.

The authors validate the model against a large empirical citation dataset from the physics literature. They compare three structural quantities: (1) the total number of triangles, (2) the average clustering coefficient, and (3) the distribution of citation path lengths. By exhaustive grid search over (α, β) they locate a unique point (α ≈ 1.0, β ≈ 0.15) that simultaneously minimizes the discrepancy for all three measures. The optimal α implies that a paper’s impact decays proportionally to the inverse of the number of subsequent papers published—a simple inverse‑age law rather than an exponential decay. The β value indicates that triangle formation is not random but occurs with a modest, consistent probability, reflecting the tendency of researchers to co‑cite related works.

At the optimal parameters the simulated network reproduces the empirical triangle count to within 4 %, the clustering coefficient with an error below 0.02, and the citation‑path length distribution almost exactly. These results demonstrate that the model captures the essential dynamics of citation growth while remaining parsimonious and analytically tractable.

The discussion highlights the interpretability of the parameters: α provides a straightforward correction factor for age bias in bibliometric evaluations, and β offers a quantitative measure of “co‑citation culture” within a field. Limitations include the exclusion of paper quality, author collaboration networks, and journal prestige, as well as the focus on a single discipline. Future work is suggested to incorporate these additional dimensions and to test the model on citation data from other scientific domains.

In summary, the study delivers a concise yet empirically grounded framework that simultaneously accounts for the acyclic nature, triangle richness, and age‑dependent attachment of scientific citation networks, thereby advancing our quantitative understanding of scholarly communication dynamics.