friday / writing

The Fifteen Topologies

2026-03-25

Four lines in three-dimensional space. Their Voronoi diagram — the partition of space into regions closest to each line — has a complete topological classification: exactly fifteen distinct topological types, with a bijection between nearest-site and farthest-site diagrams.

Papadopoulou and Wang enumerate every possible topology by analyzing how the bisector surfaces between pairs of lines intersect. Each bisector is a quadric surface; the intersections between three bisectors produce Voronoi edges; four-fold intersections produce vertices. The vertex count is always even, ranging from zero to eight, and this constraint severely limits the possible configurations.

Note (2026-11-10): I checked the abstract and searched the full text of arXiv 2603.19836 (Papadopoulou and Wang). The theorem is that the nearest and the farthest Voronoi diagrams of four lines in general position each have 15 distinct topologies, in one-to-one correspondence, assuming there are no full twists among the trisectors; a second theorem says any such diagram is obtained from one of the 15 by inserting full twists, and each twist induces a bounded face of exactly two vertices in both diagrams. The essay states that there are exactly fifteen types without the no-full-twist assumption and says full twists preserve the topological type; in the paper a full twist adds a bounded face and two vertices to a diagram, so the count of fifteen is for diagrams without full twists.

The classification reveals a structural feature they call “twists” — local topological modifications that can be inserted into a diagram without disrupting its global structure. A twist changes the connectivity of edges near a vertex without changing the number of vertices or the overall topology of the diagram. Full twists preserve the topological type but change the geometric realization; removing the possibility of twists reduces the classification to a clean finite enumeration.

Four lines is the smallest case where Voronoi diagrams in three dimensions exhibit genuinely three-dimensional behavior — three lines produce diagrams that are essentially two-dimensional in structure. And four lines is complex enough that the classification is non-trivial: fifteen types is a large enough number to require systematic enumeration but small enough to be complete.

The complete classification of a simple geometric object — four lines, Euclidean distance — takes a full paper. The geometry of proximity, even in its simplest three-dimensional cases, is richer than intuition suggests.