Finite type invariants of knots in homology 3-spheres with respect to null LP-surgeries (Q2318711)

The focus of this paper lies in the theory of finite type invariants. For 3-manifolds, this theory traces back to \textit{T. Ohtsuki} [J. Knot Theory Ramifications 5, No. 1, 101--115 (1996; Zbl 0942.57009)] (for integral homology spheres) before being generalised independently by \textit{S. Garoufalidis} et al. [Geom. Topol. 5, 75--108 (2001; Zbl 1066.57015)] and \textit{K. Habiro} [ibid. 4, 1--83 (2000; Zbl 0941.57015)]. In this theory, in order to study the free abelian group \(\mathcal{F}\) generated by all 3-manifolds, one introduces a filtration \[ \mathcal{F}=\mathcal{F}_0 \supset \mathcal{F}_1 \supset \cdots \supset \mathcal{F}_n \supset \mathcal{F}_{n+1} \supset \cdots \] whose \(n\)-th level is defined using surgery along (links associated to) a disjoint union of \(m\) \(Y\)-graphs with \(m \geq n\). A central aim of this theory is to describe the associated graded space \( \mathcal{G}_n=\mathcal{F}_n/\mathcal{F}_{n+1}\); this usually involves describing \( \mathcal{G}_n\) in terms of a certain algebra of trivalent graphs. Motivated by the Kontsevich integral, \textit{S. Garoufalidis} and \textit{L. Rozansky} later considered the free abelian group \(\mathcal{F}^{\text{null}}\) generated by all pairs \((M,K)\), where \(M\) is an integral homology sphere and \(K \subset M\) is a knot [Topology 43, No. 5, 1183--1210 (2004; Zbl 1052.57011)]. Understanding \(\mathcal{F}^{\text{null}}\) again centers around a filtration \(\lbrace \mathcal{F}_n^{\text{null}} \rbrace\) (defined using surgeries on certain ``null'' \(Y\)-graphs) and on the associated graded \(\mathcal{G}_n^{\text{null}}\). These sets further split along \(S\)-equivalence classes of pairs \((M,K)\): \[ \mathcal{G}_n^{\text{null}} =\bigoplus_{[M,K]} \mathcal{G}_n^{\text{null}}(M,K). \] Garoufalidis and Rozansky focus on knots in integral homology spheres with Alexander polynomial \(1\) (i.e. the case \([M,K]=[S^3,\operatorname{Unknot}]\)). They define an algebra \(\mathcal{A}(\mathbb{Z}[t^{\pm 1}])\) of trivalent graphs (with edges decorated by elements of \(\mathbb{Z}[t^{\pm 1}]\)) and establish the existence of a surjection \[ \mathcal{A}(\mathbb{Z}[t^{\pm 1}])\otimes \mathbb{Z}[1/2] \twoheadrightarrow \mathcal{G}^{\text{null}}[S^3,\operatorname{Unknot}] \otimes \mathbb{Z}[1/2]. \] The paper under review consists of a vast generalisation of this result: from knots in integral homology spheres to nullhomologous knots in rational homology spheres and from \([S^3,\operatorname{Unknot}]\) to arbitrary \(S\)-equivalence classes. Instead of using \(S\)-equivalence classes however, the author works with the isometry type of the Blanchfield pairing. Here, roughly speaking, a rational Blanchfield module \((H,b)\) consists of a torsion \(\mathbb{Q}[t^{\pm 1}]\)-module \(H\) and a \(\mathbb{Q}(t)/\mathbb{Q}[t^{\pm 1}]\)-valued Hermitian \(b\) on~\(H\). Previous work of the author [Bull. Soc. Math. Fr. 143, No. 2, 403-431 (2015; Zbl 1325.57006)] implies that each filtration level \(\mathcal{F}^{\text{null}}_n\) splits along isomorphism classes of Blanchfield modules: \[ \mathcal{F}_n^{\text{null}} =\bigoplus_{[H,b]} \mathcal{F}_n^{\text{null}}[H,b]. \] In this language, the work of Garoufalidis-Rozansky studies the trivial Blanchfield module, for knots in integral homology spheres. For nullhomologous knots in arbitrary rational homology spheres, and an arbitrary Blanchfield module \((H,b)\), Theorem 2.7 of the paper under review establishes a surjection from an algebra \(\mathcal{A}^{\text{aug}}(H,b)\) of decorated trivalent trees to the associated graded \(\mathcal{G}_n^{\text{null}}(H,b)\): \[ \mathcal{A}^{\text{aug}}(H,b) \twoheadrightarrow \mathcal{G}_n^{\text{null}}(H,b). \] For knots in integral homology spheres, this result improves upon the aforementioned theorem of Garoufalidis-Rozansky (Theorem 2.10). Explicit descriptions of \(\mathcal{G}_n^{\text{null}}(H,b)\) in degrees \(n=0,1\) can be found in Theorem 2.3. In Theorem 2.11, the author applies these results to relate invariants of Kricker and Lescop (both of which arise in the study of the Kontsevich integral). For further results of this nature, we direct the reader to \textit{B. Audoux} and \textit{D. Moussard} [Int. J. Math. 30, No. 5, Article ID 1950021, 37 p. (2019; Zbl 1457.57015)].