Bratteli diagrams encode the combinatorial structure of dynamical systems on Cantor sets — each level of the diagram corresponds to a partition of the space, and the edges between levels describe how partitions refine. The diagram captures the system's large-scale dynamics while abstracting away metric details.
Separated graphs extend this framework by disentangling two things that Bratteli diagrams conflate: the topological structure of the space and the dynamics of the homeomorphism acting on it. In a separated graph, the space's partition structure is encoded independently from the map's combinatorial action. The separation allows you to read off whether a dynamical system is minimal (every orbit is dense) or essentially minimal (minimal on a large subset) directly from the graph structure, without computing orbits.
Note (2026-11-08): the abstract (Claramunt) matches the essay on the construction, the four examples and the reading of minimality and essential minimality from the separated graph. The descriptions of the shift's graph as branching and reflecting exponential orbit growth, and of the odometer's graph as nearly linear, are not in the abstract.
The paper demonstrates the construction on four canonical examples: the two-sided shift, the bitwise NOT map, the classical odometer, and the shift on the one-point compactification of the integers. Each produces a visually distinct separated graph that makes the system's dynamical properties immediately readable. The shift's graph has a characteristic branching structure reflecting its exponential orbit growth; the odometer's graph is nearly linear, reflecting its rigid rotation.
The structural claim: dynamics and topology are separable aspects of a dynamical system, and encoding them separately makes both more transparent. The Bratteli diagram's power is that it encodes everything; the separated graph's power is that it encodes things independently. Separating what's combined clarifies what each part contributes.