A proof of the Kauffman-Harary conjecture (Q1035315)

Let \(p\) be an integer greater than 1. A \(p\)-coloring of a diagram \(\mathcal D\) of a knot \(k\) is an assignment of integers mod \(p\) to the arcs of \(\mathcal D\) so that the equation \(2x\equiv y+z\) mod \(p\) holds at each crossing, with \(x\) being the color of the over arc, and \(y\) and \(z\) being the colors of the under arcs. A coloring is trivial if only one color is used. We say that \(k\) is \(p\)-colorable if there is a diagram of \(k\) which admits a non-trivial \(p\)-coloring. It is an elementary fact to check that if \(k\) is \(p\)-colorable then every diagram of \(k\) has a non-trivial \(p\)-coloring. Let \(\det\, k=| \Delta_k(-1)|\) be the determinant of \(k\), where \(\Delta_k\) denotes the Alexander polynomial of \(k\). It is known that \(k\) is \(p\)-colorable if and only if \(p\) and \(\det\, k\) have a prime factor in common. \textit{F. Harary} and \textit{L. Kauffman} [Adv. Appl. Math. 22, No.~3, 312--337 (1999; Zbl 1128.57301)] conjectured that if \(k\) is an alternating knot having prime determinant \(p\), then every non-trivial \(p\)-coloring of a reduced alternating diagram of \(k\) assigns different colors to different arcs. This conjecture has been proved for some classes of knots by several groups of mathematicians. In the paper under review a proof of the conjecture is given. The proof is combinatorial, and roughly speaking is as follows. Let \(\mathcal D\) be a diagram with \(n\) crossings; to it we can associate an \(n\times n\) matrix \(C'\), called the crossing matrix. Each row of \(C'\) corresponds to a crossing, has one entry 2, two entries -1, and the other entries are 0. Columns of \(C'\) correspond to arcs of \(\mathcal D\), and if a \(p\)-coloring is represented by a column vector \(X\), then \(C'X\equiv 0\) mod \(p\). Let \(C\) be a matrix obtained by removing one row and one column of \(C'\). It is known that \(| \det\, C|=\det\, k\). Assume now that \(\mathcal D\) is a reduced alternating diagram, and that \(p=\det\, k\) is a prime number. Let \(L=pC^{-1}\) be the coloring matrix. The main argument in the proof is to show that each column of \(L\) has some entries that are not 0 mod \(p\). In turn this implies that \(\mathcal D\) must have a \(p\)-coloring where different arcs get different colors. Then it is shown that any \(p\)-coloring must have that property.