Gordian Graph


Background

A knot is an embedding of \(S^1\) into \(S^3\). Knots are represented by knot diagrams, which are projections of the knot onto a plane with intersections labeled with “over” and “under” strands. Every knot has infinitely many diagrams, and determining how knots relate to their diagrams is a major guiding question in knot theory.

We say that two knots \(K_1\) and \(K_2\) are related by crossing change if there exists a diagram \(D_1\) of \(K_1\) and a crossing in that diagram such that if we change which strand is the “over” strand in that diagram, we get a diagram \(D_2\) for \(K_2\). This is related to the unknotting number of a knot, which is the minimal number of crossing changes necessary to unknot a knot, minimized over all possible diagrams. In general, this is hard to compute, and there exist \(10\)-crossing knots for which the unknotting number is still unknown.

However, we can also ask which knots are related to each other by crossing change. For instance, nontrivial knots \(K_1\) and \(K_2\) might both have diagrams that are only one crossing change away from a diagram for the unknot (and hence have unknotting number one), but that does not imply that \(K_1\) and \(K_2\) are related to each other by a single crossing change.

The data of how knots are related by crossing change is usually described as the Gordian graph. The Gordian graph is constructed as follows. The vertices are (isotopy classes of) knots. Two vertices \(K_1\) and \(K_2\) are joined by an edge if there is some diagram of \(K_1\) where a single crossing change turns it into a diagram for \(K_2\).

Our approach to drawing the Gordian graph involves brute force searching the space of all knot diagrams under, say, \(8\) crossings, systematically changing each crossing, and recording the resulting knot. Determining which knot we have as a result can be done by SnapPy, a program for low-dimensional topology capable of basic knot identification. Knot diagrams would likely be encoded by their planar diagram (PD) codes.

Having an interactive, searchable Gordian graph could be of enormous help to researchers looking for examples and counterexamples in low-dimensional topology to test their conjectures against. It may also help illuminate relationships between knots that are related by a single crossing change (for instance, every knot is at most one crossing change away from an Alexander polynomial \(1\) knot1). Also, many results about the Gordian graph give existence results that are not constructive, meaning that the actual diagram and crossing that is changed is not given by the theorem.

The Graph

🔗 Gordian Graph

The graph displays all knots corresponding to braid words with length 8 or less.

Clicking on a node displays a picture of the knot. Clicking on an edge provides an example crossing change connecting the respective nodes. Entering the names of two knots gives the shortest path between them.

References


  1. Yasutaka Nakanishi and Yuki Okada. Differences of alexander polynomials for knots caused by a single crossing change. Topology and its Applications, 159:1016–1025, 03 2012.↩︎