The tangle-valued 1-cocycle for knots

Abstract

This paper contains the strongest and at the same time most calculable knot invariant ever. Let $\Theta$ be the topological moduli space of all ordered oriented tangles in 3-space. We construct a non-trivial combinatorial 1-cocycle $\mathbb{L}$ for $\Theta$ that takes its values in $H_0(\Theta;\mathbb{Z})$. The 1-cocycle $\mathbb{L}$ has a very nice property, called the {\em scan-property}: if we slide a tangle $T$ over or under a given crossing $c$ of a fixed tangle $T'$, then the value of $\mathbb{L}$ on this arc $scan(T)$ in $\Theta$ is already an isotopy invariant of $T$. In particular, let $D$ be a framed long knot diagram. We take the product with a fixed long knot diagram $K$ and we consider the 2-cable, with a fixed crossing $c$ in $2K$. $\mathbb{L}(scan(2D))$ gives an element in $H_0(\Theta)$. To this element we associate the {\em set of Alexander vectors}, consisting of the corresponding integer multiples of the one-variable Alexander polynomials of (the standard closures) of all sub-tangles of each of the tangles. We can vary the knots $(K,c)$ and moreover we can iterate our construction by starting now again the $scan$ with the tangles in $\mathbb{L}(scan(2D))$ and so on. The result is the infinite {\em Alexander tree}, which is an isotopy invariant of the knot represented by $D$. {\em As examples we show with just one edge of the Alexander tree that the knot $8_{17}$ and the Conway knot are not invertible!} This makes the Alexander tree a very promising candidate for a complete and "locally" calculable knot invariant, because the tangles in $\mathbb{L}(scan(2D))$ can be drawn with linear complexity and their Alexander polynomials can be calculated with quartic complexity with respect to the number of crossings of $D$.