Identifying non-invertible knots

Abstract

IN THIS PAPER, the thirty-six non-invertible knots with ten crossings or fewer are identified. These are the knots listed by Conway[6] as non-invertible on the basis of empirical observation. The method used involves the calculation of covering invariants of certain irregular metabelian covering spaces. For most of the knots, this is the first proof of non-invertibility to appear, however I understand that Bonahon and Siebenmann (in work discussed below) have an alternative proof for a large number of them. However, details of this work are apparently not yet available. Although most small knots are easily observed to be invertible (defined below) the existence of non-invertible knots has long been presumed [ I]. However the existence of even a single non-invertible knot was not proven until Trotter[24] gave examples, the smallest of which had 15 crossings. Further examples were given by Whitten[26]. Howeve; until recently, no method has been available for determining whether any “naturally occurring” knot is invertible. The knot 8,,, for instance, has long been thought to be non-invertible, but this was only recently proven by Kawauchi [ 161 though prior proofs are claimed by Riley (using methods similar to those of [23]) and Bonahon and Siebenmann [3]. Kawauchi[l6] gave a method which applied to amphicheiral hyperbolic knots, and he was able to identify the amphicheiral non-invertible knots with up to ten crossings (seven in all). The assumption that the knot be hyperbolic was removed by the present author[9]. This provided a useful criterion for an amphicheiral knot to be invertible. Bonahon and Siebenmann[3] and Boileau[4,5] have studied algebraic knots using their 2-fold branched covering space, and are able to classify them and determine their complete symmetry group (including invertibility). Algebraic knots are those which have a projection with a I* type diagramtheir Conway notation contains no dots. As Conway observes ([6], p. 335), those knots which he lists as 6** types-their notation begins with a dot-may be given a l* type projection at the cost of adding two extra crossings. Hence, they are also algebraic, and are amenable to the methods of BonahonlSiebenmann and Boileau. Of the 36 non-invertible knots, 27 are algebraic (see table). Of the remaining ones, three are covered by Kawauchi[l6]. For the invertibility of the knots 10loz, 1O1M, 10107, 10,,o, lo,,,, 10119, the methods of this paper give the only known proof. The method used involves a calculation of covering invariants of the knot, that is, the homology groups of the branched and unbranched covering spaces and linking numbers between components of the covering link in some suitable covering. This has been dubbed “the universal method” by Riley[22] and has had many applications over a long period. The 2and 3-fold cyclic coverings were used by Alexander and Briggs [ l] to distinguish most knots to nine crossings, but the importance of homology groups of cyclic covers diminished with the invention of the Alexander polynomial. Linking numbers in dihedral covering spaces were first used by Reidemeister[21] and Bankwitz and Schumann[2] to complete the classification of knots to nine crossings. The method has been applied to the classification of 10 and 11 crossing knots by Perko [191 (dihedral and Sq coverings) and Riley[22] (PSL coverings), and also as a powerful tool for showing non_amphicheirality[7, 191. The present article is the first