Positive quandle homology and its applications in knot theory (Q2346728)

A quandle is a set \(Q\) together with a binary operation \(*:Q\times Q\to Q\) satisfying \(x* x=x\), \((x* y)* z=(x* z)* (y* z)\) (for all \(x,y,z\in Q\)), and \(-* x:Q\to Q\) bijective for all \(x\in Q\). The canonical example is a group \(G\) with operation \(x*y=y^{-1}xy\). One of the most interesting points is that quandles can be used for knot coloring. A powerful knot invariant induced by a (finite) quandle is the so called state-sum invariant, that essentialy makes an average of a weight given by a ``2-cocycle'', over all crossings and all possible colorings, so it is very interesting to know non-trivial 2-cocycles. The way to produce cocycles is well known (see for instance [\textit{J. S. Carter} et al., Trans. Am. Math. Soc. 355, No. 10, 3947--3989 (2003; Zbl 1028.57003)]): given a quandle, a cohomology theory is defined, and its differential is given by a difference of two differentials. More precisely, for \(n\in\mathbb N\) consider the free abelian group with generators \(Q^n\), and \(d_1,d_2\) the maps given by \[ d_1(q_1,\dots,q_n)=\sum_{i=1}^n (-1)^i (q_1,\dots,q_{i-1},\widehat{q_i},\dots,,q_n), \] \[ d_2(q_1,\dots,q_n)=\sum_{i=1}^n(-1)^i (q_1*q_i,\dots,q_{i-1}*q_i,\widehat{q_i},\dots,,q_n) \] Then \(\partial:=d_1-d_2\) happens to be a differential, and the (rack) cohomology with coefficients in an abelian group \(A\) is defined as the cohomology of \(\partial^*:A^{Q^*}\to A^{Q^{*+1}}\) \[ \begin{multlined} \partial^*(f)(q_1,\dots,q_{n+1}) = \sum_{i=1}^{n+1} (-1)^i f(q_1,\dots,q_{i-1},\widehat{q_i},\dots,q_{n+1})\\ - \sum_{i=1}^{n+1} (-1)^i f(q_1*q_i,\dots,q_{i-1}*q_i,\widehat{q_i},\dots,q_{n+1}).\end{multlined} \] The quandle cohomology is defined in a similar way, considering the subcomplex generated by maps \(f:Q^n\to A\) vanishing on elements of type \((q_1,\dots,q_{i-1},q,q,q_{i+1},\dots,q_n)\), these cocycles are the ones that preserve Reidemeister moves, so they can be used to compute state-sum invariants. Also, given a quandle \(Q\) and an abelian group together with an automorphism \(T\), in [\textit{J. S. Carter} et al., Algebr. Geom. Topol. 2, 95--135 (2002; Zbl 0991.57005)] a twisted version is defined, considering the differential \[ \begin{multlined} \partial^*(f)(q_1,\dots,q_{n+1})= \sum_{i=1}^{n+1} (-1)^i T(f(q_1,\dots,q_{i-1},\widehat{q_i},\dots,q_{n+1})) \\ - \sum_{i=1}^{n+1} (-1)^i f(q_1*q_i,\dots,q_{i-1}*q_i,\widehat{q_i},\dots,q_{n+1}).\end{multlined} \] In the paper under review, the authors consider positive and negative cohomology as the cohomology corresponding to the differentials \(\partial_{\pm}:=d_1\pm d_2\), the negative cohomology being the usual cohomology, and the positive cohomology a new one. It is quite surprising that the authors do not notice that positive cohomology is the same as twisted cohomology, simply choosing \(T=-\mathrm{Id}\), the differential is the same up to a global sign, that of course does not change cohomology. The proofs of the standard facts such as \(\partial_\pm^2=0\), well-definedness of the subcomplex giving quandle cohomology, or the state-sum procedure in order to obtain knot invariants with \(H^2\) becomes unnecessary since they were already proven in [Zbl 0991.57005] for the general twisted version. We notice that also the twisted cohomology is a particular case of taking coefficients in \(G_Q\)-modules, where \(G_Q\) is the group freely generated by \(Q\) with the relations \(x*y=y^{-1}xy\), which is also a particular case of cohomology with coefficients in \(\mathbb Z(Q)\)-modules, where the algebra \(\mathbb Z(Q)\) (and the corresponding generalized (co)homology) was defined by \textit{N. Andruskiewitsch} and \textit{M. Graña} [Adv. Math. 178, No. 2, 177--243 (2003; Zbl 1032.16028)]; the algebra \(\mathbb Z(Q)\) admits a cannonical epimorphism \(\mathbb Z(Q)\to \mathbb Z[G_Q]\), so every \(\mathbb Z(Q)\)-module is a \(G_Q\)-module. The genuine new results of the paper under review are computations of twisted cohomology for two examples. First, the trivial quandle of two elements: \(T_2\). It is well-known that, for a non trivial 2-cocycle, \(T_2\) gives the linking number, so for example it does not ``see'' the Borromean rings. In this paper a particular twisted 2-cocycle is presented that detects the nontriviality of the Borromean rings. The second example is the dihedral quandle \(D_3\): the set of reflections of the dihedral group. It is well-known that the usual (that is with trivial coeficients) cohomology vanishes in degree 2, but it is shown that twisted 2-cohomology is not trivial, so one can use it to compute non trivial knot invariants, the authors also give an interpretation of this invariant in terms of trivially 3-colored crossing points.