Crossing number of an alternating knot and canonical genus of its Whitehead double (Q820659)

Let \(K\) be a non-trivial knot and \(W_K\) be some of its Whitehead doubles. We write \(c(K)\) for the crossing number of \(K\), and \(g_c\) for the canonical genus (the minimal genus of all surfaces obtained by Seifert's algorithm). The following conjecture has been around for some time, attributed here to [\textit{J. J. Tripp}, J. Knot Theory Ramifications 11, No. 8, 1233--1242 (2002; Zbl 1024.57013)]. Conjecture 1. \(c(K)=g_c(W_K)\). It is very easy to observe that \(c(K) \ge g_c(W_K)\) holds, so the question is how to estimate \(g_c(W_K)\) from below. The only really viable tool to estimate \(g_c(W_K)\) (in such case) is \textit{H. R. Morton}'s inequality [Math. Proc. Camb. Philos. Soc. 99, 107--109 (1986; Zbl 0588.57008)] \(2g_c(K')\ge \max\operatorname{deg}_mP(K')\) for the skein (HOMFLY) polynomial \(P(K')(l,m)\). Various partial results are known, but all seem kind of subsumed by the work in an unfortunately still unpublished paper of \textit{M. Brittenham} and \textit{J. Jensen} [``Canonical genus and the Whitehead doubles of families of alternating knots'', \url{https://www.math.unl.edu/~mbrittenham2/papers/}]. Although they state their results in a far more restrictive way, the core of their argument (Proposition 4) is that when Conjecture 1 holds for an alternating Conway polyhedron, then it holds for any alternating knot derived from that polyhedron. There is a caveat here: even if a knot is derived, the polyhedron may be a link. This is best fixed by removing Whitehead double clasps, and working with satellite links, the patterns for each component being a reverse 2-cable. In particular it is very easy to test Conjecture 1 then for every alternating arborescent knot (not only pretzel knots). In the paper under review, the authors establish two points. First, they prove that Conjecture 1 is false for composite knots, in that it fails for connected sums of \((2,l)\)-torus knots. They then propose a modified version of Conjecture 1, as follows. Conjecture 2. If \(K\) is a knot with \(k\) prime factors, \(c(K)-k+1=g_c(W_K)\). Second, they prove that the conjecture holds for alternating 3-braid knots. This consists in a complicated evaluation of the skein polynomial of the (reverse 2-cable of the closed) braids \((\sigma_1\sigma_2^{-1})^n\). With Brittenham-Jensen this then establishes Conjecture 1 more generally for all (prime) alternating knots derived from the above Conway 3-braid polyhedra. In should be pointed out that the counterexamples to Conjecture 1 and the formulation of Conjecture 2 appear natural in light of a very related conjecture made (as a question) in a paper by \textit{M. E. Kidwell} and the reviewer [Mich. Math. J. 51, No. 1, 3--12 (2003; Zbl 1030.57009)]. We write \(F(K)(a,z)\) for Kauffman's (2-variable) polynomial. Conjecture 3. If \(K\) is a non-trivial knot, then \(2\max\operatorname{deg}_z F(K)+2= \max\operatorname{deg}_m P(W_K)\). For alternating knots, Conjecture 3 is equivalent to Conjecture 2, so we see it confirmed for many alternating knots. However, the truth of Conjecture 3 for (all tested so far) non-alternating knots also makes clear why Morton's inequality is not strong enough to yield Conjecture 1 (or Conjecture 2) for non-alternating knots. The only modest insight in that case comes from a very recent paper of the reviewer [J. Symb. Comput. 101, 242--269 (2020; Zbl 1453.57008)] showing (with more diversified methods and in a somewhat broader setting) that \(g_c(W_K)\ge 3\).