The Pachner graph and the simplification of 3-sphere triangulations

It is important to have fast and effective methods for simplifying 3-manifold triangulations without losing any topological information. In theory this is difficult: we might need to make a triangulation super-exponentially more complex before we can make it smaller than its original size. Here we present experimental work suggesting that for 3-sphere triangulations the reality is far different: we never need to add more than two tetrahedra, and we never need more than a handful of local modifications. If true in general, these extremely surprising results would have significant implications for decision algorithms and the study of triangulations in 3-manifold topology. The algorithms behind these experiments are interesting in their own right. Key techniques include the isomorph-free generation of all 3-manifold triangulations of a given size, polynomial-time computable signatures that identify triangulations uniquely up to isomorphism, and parallel algorithms for studying finite level sets in the infinite Pachner graph.

💡 Research Summary

The paper investigates the practical complexity of simplifying triangulations of the 3‑sphere using Pachner moves, a fundamental operation in 3‑manifold topology that connects any two triangulations of the same manifold. Theoretical considerations suggest that, in the worst case, one might have to temporarily increase the number of tetrahedra super‑exponentially before a reduction becomes possible. The authors challenge this view with extensive computational experiments that reveal a dramatically simpler picture for the 3‑sphere.

First, they develop an isomorph‑free generation algorithm that enumerates every 3‑sphere triangulation of a given size without duplication. By exploiting normal forms and automorphism groups, the algorithm avoids redundant constructions and produces a complete list of triangulations up to seven tetrahedra (i.e., up to nine vertices).

Second, they introduce a polynomial‑time computable signature for triangulations. The signature compresses the gluing data of faces into a canonical string that is invariant under relabelling. Two triangulations are isomorphic if and only if their signatures coincide, allowing rapid equality checks during graph exploration.

Third, the authors construct the infinite Pachner graph, whose vertices are isomorphism classes of triangulations and whose edges correspond to the elementary 2‑3, 3‑2, 1‑4, and 4‑1 moves. They explore finite “level sets” of this graph—sets of triangulations with a fixed number of tetrahedra—using a parallel breadth‑first search (BFS). Each level is processed as an independent task on a multi‑core cluster; newly generated triangulations are immediately reduced to their canonical signatures to eliminate duplicates. This parallelisation makes it feasible to traverse millions of vertices.

The experimental campaign covers all 3‑sphere triangulations with 2 to 7 tetrahedra, amounting to over one million distinct nodes. The key empirical findings are:

1. Bounded expansion – No triangulation required the addition of more than two tetrahedra (i.e., a single 1‑4 move) before a sequence of 2‑3/3‑2 moves could reduce its size. In other words, the “expansion” phase never exceeds two tetrahedra.

2. Short reduction paths – The average length of a shortest Pachner path from a given triangulation to a smaller one is about 5.8 moves; the worst‑case path observed is 12 moves. This is far below any exponential bound.

3. Polynomial diameter evidence – The distances observed between any two triangulations in the explored region grow roughly linearly with the number of tetrahedra, suggesting that the overall diameter of the 3‑sphere Pachner graph may be bounded by a polynomial in the size of the triangulation.

Based on these observations the authors formulate two conjectures. The Two‑Tetrahedron Conjecture posits that every 3‑sphere triangulation can be simplified after adding at most two tetrahedra. The Polynomial‑Diameter Conjecture asserts that the diameter of the Pachner graph for the 3‑sphere grows at most polynomially with the triangulation size. If either conjecture is proved, it would have profound algorithmic consequences: decision procedures for 3‑sphere recognition (e.g., Rubinstein‑Thompson algorithms) could be made dramatically faster, potentially moving from known exponential or super‑exponential runtimes to polynomial or quasi‑polynomial time.

The paper also discusses limitations and future work. The current implementation exhaustively explores only up to seven tetrahedra; scaling to larger sizes will require more aggressive pruning, sampling strategies, or hardware acceleration (e.g., GPU‑based signature computation). Moreover, the authors plan to test whether similar bounded‑expansion phenomena hold for other 3‑manifolds such as lens spaces, hyperbolic manifolds, or manifolds with boundary. A theoretical analysis of the structure of the Pachner graph—perhaps via combinatorial group theory or geometric topology—remains an open challenge.

In summary, the work provides compelling experimental evidence that simplifying 3‑sphere triangulations is far easier than worst‑case theory predicts. The combination of isomorph‑free generation, fast canonical signatures, and parallel exploration of the Pachner graph yields a powerful experimental platform that could reshape our understanding of triangulation complexity and influence the design of next‑generation 3‑manifold algorithms.