Building Three-Dimensional Differentiable Manifolds Numerically II: Limitations
Lee Lindblom, Oliver Rinne
TL;DR
The paper extends the method of constructing reference metrics on 3-manifolds by relaxing the dihedral-angle uniformity constraint and expanding the class of triangulations considered. It formulates a set of edge-sum and vertex- continuity constraints on dihedral angles to obtain a globally $C^0$ metric that can be smoothed to $C^1$, and uses a BFGS-based minimization of the constraint norm $||C||$ with a practical tolerance. Results show only a small fraction of the Regina catalog multicubes admit constraint-satisfying dihedral angles, yielding 23 uniformly constrained and 80 non-uniformly constrained cases, of which 17 boldface entries admit non-singular $C^1$ metrics. The findings reveal substantial limitations of the existing approach and suggest that achieving broader applicability will require fundamental changes to how triangulations are transformed into multicube coverings. The work thus clarifies the current boundary of numerically constructing differentiable structures on 3-manifolds and points to directions for overcoming these barriers.
Abstract
Methods were developed in Ref. [1] for constructing reference metrics (and from them differentiable structures) on three-dimensional manifolds with topologies specified by suitable triangulations. This note generalizes those methods by expanding the class of suitable triangulations, significantly increasing the number of manifolds to which these methods apply. These new results show that this expanded class of triangulations is still a small subset of all possible triangulations. This demonstrates that fundamental changes to these methods are needed to further expand the collection of manifolds on which differentiable structures can be constructed numerically.
