Knot theory yields a knit stitch that shrinks a dropped loop's damage until it stops

A torus is the right shape for this problem. Knitting leaves both ends of the yarn outside the fabric, and the loops stay in place by passing through one another without ever being tied ^[2]. A flat swatch also has edges. Wrapping the stitch model around a doughnut keeps the repeating pattern and deletes the edges, so the fabric becomes one closed knot that knot theory can analyse ^[3]. With no border in the model, edge effects drop out of the comparison ^[3].

Stacking several rows on the torus let the team follow a defect from one row into the next until the whole fabric came apart ^[4]. The researchers call this defect propagation and regard it as a defining trait of knittable fabric ^[4]. "It's a very original approach," said Omelchenko, who was not involved in the study ^[6].

Plain jersey fails without limit. "If one unit cell is disentangled, the cell just below that cell is disentangled, and it is repeated, and it propagates endlessly," Shimamoto said ^[8].

He then used moves from basic knot theory to rework the stitch so that propagation can still begin but cannot continue forever ^[9]. In the resulting robust knit, described in Physical Review X, a defect that propagates gets smaller with each row until it stops ^{[1] [11]}.

Both stitches went through the same torus analysis, and they differ in how the loops cross, so in this model the stitch's structure decides whether one mistake ends or runs the length of the fabric ^[15]. The thing this doesn't tell you is the effect size in real cloth. The account does not say how many rows a defect survives in the new stitch, what the more complex pattern costs in yarn or knitting time, whether knitted samples were tested, or what role yarn fibre plays.

Shimamoto learned to knit from his mother, and he gave her an author credit on the paper as thanks ^[14].