A 2026 persistent-homology study of the Roman–Byzantine trade network reports a 150–300× jump in cross-network topological divergence after the 1082 Chrysobull

A July-2026 arXiv preprint (Bernal-Alvarado, Delepine & Pinedo Guadarrama, "Structural Divergence of the Roman–Byzantine Trade Network, 0–1453 CE") applies persistent homology — an algebraic-topology method — to a reconstructed 2,599-node trade network spanning 0–1453 CE, calibrated against Stanford's ORBIS model of the Roman world. Its headline quantitative finding is that "the cross-network Wasserstein ratio increases by 150–300×" after the Chrysobull of 1082 CE (Alexios I Komnenos's trade grant to Venice). The authors read this not as simple territorial loss but as a sudden economic and geographic decoupling of the network's cyclic structure — a topological signature of imperial fragmentation distinct from mere shrinkage.

The same paper also reports a candidate percolation threshold (H* ≈ 0.524) it argues is shared by both the Western collapse (476 CE) and the Byzantine endpoint (1453 CE), framing collapse as a criticality phenomenon rather than a sequence of contingent events.

The claim is Tier 1 in provenance (arXiv primary, abstract independently re-read) but rests on a single, unreplicated preprint whose central move is an interpretation — reading a Wasserstein-distance ratio as "decoupling." Whether the 150–300× jump reflects a real economic transition or an artifact of how a 1,400-year network is reconstructed and sampled is precisely the question the paper's own robustness section wrestles with (see claim-hub-selection-artifact-can-reverse-network-breakpoint-signal), which is why this note stays seedling.

This is the vault's first note on cliodynamics — the quantitative-history field this method belongs to; see claim-seshat-moralizing-gods-finding-disputed-as-data-artifact for the same field's public "pattern vs. artifact" fight.

The same persistent-homology / Wasserstein-distance machinery is a standard neural-network diagnostic (bounding generalization from a training trajectory); see observation-persistent-homology-gradient-free-bridge-widrow-byzantine for the bridge to Widrow-era backpropagation history.