Homotopy on spatial graphs and generalized Sato-Levine invariants (Q2269956)

Let \(G\) be a finite graph. An embedding \(f\) of \(G\) into the 3-sphere \(S^3\) is called a spatial embedding of \(G\), or simply a spatial graph. Two spatial embeddings of \(G\) are edge-homotopic if they can be transformed into each other by self crossing changes and ambient isotopies, where a self crossing change is a crossing change on the same spatial edge, and vertex homotopic if they can be transformed into each other by crossing changes on two adjacent spatial edges and ambient isotopies. These equivalence relations were introduced by \textit{K. Taniyama} [Topology 33, No.~3, 509--523 (1994; Zbl 0823.57006)], as generalizations of Milnor's link-homotopy on links. In the paper under review the author defines some new edge-homotopic (resp. vertex-homotopic) invariants of disconnected spatial graphs, i.e. of graphs \(G=G_1\cup G_2\), where \(G_1\), \(G_2\) are connected and disjoint. The invariants are certain weighted sums of generalized Sato-Levine invariants of constituent 2-component links of \(G\), i.e. of links \(L=\gamma_1\cup \gamma_2\), where \(\gamma_i\) is a cycle in \(G_i\), \(i=1,2\); these are defined if the graphs satisfy certain conditions. There is no restriction on the linking numbers of the constituent links, as there is in some invariants previously defined by the author and \textit{T. Fleming} [Trans. Am. Math. Soc. 361, No.~4, 1885--1902 (2009; Zbl 1163.57002)]. Those invariants are used to show that there are infinitely many embeddings of a certain graph which are not edge-homotopic (resp. vertex-homotopic).