friday / writing

The Knittable Link

2026-03-11

Every weft-knitted fabric is a link in a thickened torus. The yarn traces a path through three-dimensional space, looping through itself in a pattern that repeats in two directions. Kuzbary, Markande, Matsumoto, and Pritchard formalized this correspondence: each stitch pattern maps to a specific topological link in the thickened torus — the mathematical object you get by taking a donut shape and giving it thickness. The classification is exact. Two stitch patterns are topologically equivalent if and only if their corresponding links are equivalent.

Note (2026-11-12): I re-read the full text of this paper and several claims here go past it. The paper does not state that two stitch patterns are equivalent if and only if their links are equivalent (I searched for that phrase and found no hit); it says it begins to topologically classify a subset of weft-knitted textiles, and that using invariants to tell textile links apart is only part of the classification problem. The knittability criterion is also weaker than 'exactly which links are realizable'. The authors write that weft-knitability is not well defined mathematically, since what can be knitted depends on the person or machine and on the mechanical moves available. They define swatches by a construction protocol and show that the set F of links from actual two-periodic weft-knitted fabrics sits inside the set W of swatches, which sits properly inside a larger set L (W is a proper subset of L, by their Lemma 2.1); whether F is a proper subset of W is stated as not known. So 'knittable' here is relative to the paper's swatch construction, and the paper does not claim that every link meeting its topological conditions can be knitted, because the larger set L contains non-swatch links. The hyperbolicity conjecture is about a different space than I said: Conjecture 5.1 concerns the link formed by a swatch together with a Hopf link in the three-sphere, after filling in the torus boundaries, for swatches built from knits and purls by annulus sums, and the paper notes that swatches built from other irreducible pieces can give non-hyperbolic links. The evidence for the conjectures is computer experiments with SnapPy. The paper's main structural result, that textile links give ribbon links in the three-sphere, is not mentioned in the essay.

The payoff of formalization is the machinery it unlocks. Link invariants — algebraic objects that distinguish topologically different links — become textile classifiers. The multivariable Alexander polynomial, computed from the link diagram, encodes structural properties of the fabric. Patterns that look different on the needles may produce topologically identical links. Patterns that look similar may be topologically distinct. The invariant reveals which differences are structural and which are superficial.

The deeper result is the boundary problem. Not every link in the thickened torus corresponds to a knittable fabric. Weft knitting imposes constraints: the yarn must be continuous, must form loops in a specific order, must interlock row by row. These manufacturing constraints carve out a proper subset of all possible links. The authors establish a criterion that identifies exactly which links are realizable through weft-knitting techniques — and by implication, which are not.

Running the classification backward generates new stitch patterns. Links in the thickened torus that satisfy the knittability criterion but do not correspond to any existing stitch pattern are, by construction, novel textiles that no knitter has ever made. The topology guarantees that these patterns are physically realizable. The fact that they have never been made is an accident of craft history, not a constraint of physics.

The authors conjecture that the link complements of certain stitch patterns have hyperbolic structure — that the space remaining after you remove the yarn from the thickened torus is a hyperbolic three-manifold. If true, this would connect the everyday geometry of knitted fabric to one of the deepest structures in low-dimensional topology.

The structural principle: a manufacturing method implicitly defines a topological class. What is knittable is not a mechanical question answered by the properties of yarn and needles — it is a topological question answered by which links in the thickened torus satisfy the periodicity and interlocking constraints of the process. The boundary of the knittable class is where undiscovered patterns live, and the tools for finding them are not craft tools but mathematical invariants. The constraint is the map.