Ginestra Bianconi and Christoph Rahmede (2017), Scientific Reports 7, 41974 — ARXIV-1607.05710, DOI-10.1038/srep41974
Key takeaways
- The model. Start from one d-simplex. At each step pick a (d − 1)-face α with probability ∝ 1 + s n_α, where n_α is the number of d-simplices on it minus one, and glue a new d-simplex to it through one new node. Flavor s = −1 allows only faces with n_α = 0, so the complex is a discrete manifold; s = 0 is uniform; s = 1 is preferential attachment on faces. The rule refers to no coordinate or distance.
- Degrees, derived. A master equation gives the degree distribution in closed form: geometric, P(k) = (d/(d+1))^{k−d}/(d+1), when d + s = 1, and a ratio of Gamma functions with a power-law tail k^{−γ}, γ = 2 + 1/(d + s − 1) ≤ 3, when d + s > 1 (d = 1, s = 1 is the Barabási–Albert γ = 3). Simulations match (Fig. 2). This is the part of the paper that is derived and could have failed.
- Geometry, by construction. Each new d-simplex is glued to one face, so the dual (simplices as nodes, shared faces as links) is a tree. With every link taken to have the same length, the paper draws each simplex as an ideal simplex of the Poincaré ball, with the first d + 1 nodes on the sphere summing to zero and each new node at the normalised sum of its parent face’s node positions. This is offered as the unique embedding that fills the space as t → ∞, which is asserted, not proved. The case is also argued from the small-world property (N ≃ e^D cannot fit a finite-dimensional Euclidean space with equal links), which the paper itself says is not sufficient.
- What the hyperbolicity rests on. The authors say it is a consequence of the equal-length “proto-geometry”: with unequal lengths the same complex embeds in a sphere. In the ideal embedding all node distances are infinite, and for s = 0, 1 sibling nodes from one face coincide, so the embedding is clean only for the manifold flavor s = −1. For s = −1, d = 3 the complexes are stacked polytopes tied to Apollonian packings, whose symmetry group is a discrete subgroup of SO(3,1); for d = 2 the paper links them to Farey sequences without a derivation.
- Network signatures together. For d ≥ 2, modularity 0.80–0.97 and average clustering 0.65–0.84 (N = 10⁴, 20 realisations, Table I), with the small-world property for all flavors except the chain (d = 1, s = −1).
- A transition in the geometry, numerically. Giving faces a fitness e^{−βε_α} from random node energies and attaching ∝ η_α(1 + s n_α), at large β the centroid of node positions R approaches 1 and their spread σ vanishes with size: the complex grows in one direction of the ball. Shown by simulation (Fig. 4, N up to 10⁴, d = 2, 3, s = −1), with the maximum degree peaking at the transition for d = 3 only.
Standing in the record
Filed on 2026-10-09 at the owner’s request, in the second part of the batch on hierarchy and hyperbolic geometry asked for after the record opened QUESTION-025. The owner gave only the link.
Read the same day (NOTE-678). It is the companion to Krioukov et al. (LIT-865, THEORY-188), whose model it cites as its foil: there nodes are sprinkled in a hyperbolic space and the network follows from distance; here the network grows by a combinatorial rule and the hyperbolic picture follows from its tree-shaped dual. The reading is the source of THEORY-193. It is also a worked instance of the higher-order network geometry the Battiston et al. review (LIT-091) maps only in passing, and of the “emergent space” theme that LIT-102 gestures at without a construction. It does not answer QUESTION-025; the NOTE says what it offers it.