Johnson homomorphisms (Q2656614)

For a compact connected surface \(S\) of genus \(g\) with \(n\) marked points and \(r\) boundary components, let \(\Gamma_{g,n+\vec{r}}\) be the \textit{mapping class group} of isotopy classes of diffeomorphisms fixing the marked points and boundary components pointwise. The \textit{Torelli group} \(T_{g,n+\vec{r}} \subset \Gamma_{g,n+\vec{r}}\) is the subgroup consisting of those isotopy classes which act as the identity on \(H_1(\bar{S};\mathbb{Z})\), where \(\bar{S}\) is obtained from \(S\) by filling in the boundary components with discs. Torelli groups are intimately related to many topics in geometry. Let us give two examples. In algebraic geometry, the homotopy fibre of the period map \(\mathcal{M}_g \to \mathcal{A}_g\) from the moduli space of genus \(g\) curves to the moduli space of principally polarised varieties of dimension \(g\), is essentially the classifying space of \(T_g\). In 3-manifold topology, Torelli groups are related to the study of homology 3-spheres, since each of these arises as two handlebodies glued along an element of the Torelli group of their common boundary. References for these connections, as well as other ones, can be found in the survey under review. \textit{D. Johnson} [Math. Ann. 249, 225--242 (1980; Zbl 0409.57009)] constructed a homomorphism \(T_{g,\vec{1}} \to \Lambda^3 H_1(S;\mathbb{Z})\), which is surjective for \(g \geq 3\) and necessarily factors over the abelianisation to yield a surjective homomorphism \[ \overline{\tau}_1 \colon H_1(T_{g,\vec{1}}) \longrightarrow \Lambda^3 H_1(S;\mathbb{Z}), \] which has a finite 2-torsion kernel [\textit{D. Johnson}, Topology 24, 127--144 (1985; Zbl 0571.57010)]. This is the first of a collection of \textit{higher Johnson homomorphisms}. For the remainder of this review we take \(r \geq 1\), pick a basepoint \(x \in \partial S\), and let \(g\) be sufficiently large. We set \(\pi_1(S,\vec{v}) := \pi_1(\hat{S},x)\) with \(\hat{S}\) for the surface obtained from \(S\) by filling in all boundary components with discs except the one containing \(x\) (the notation reflects that this is equal to an instance of Deligne's fundamental group with tangential basepoint). The \textit{Johnson filtration} [\textit{D. Johnson}, Contemp. Math. 20, 165--179 (1983; Zbl 0553.57002)] \[ T_{g,n+\vec{r}} = J^1 T_{g,n+\vec{r}} \supset J^2 T_{g,n+\vec{r}} \supset \cdots \] is the descending filtration of the Torelli group defined in terms of the lower central series \(L^k \pi_1(S,\vec{v})\) as \[ J^k T_{g,n+\vec{r}} = \ker\left[T_{g,n+\vec{r}} \to \mathrm{Aut}\,\pi_1(S,\vec{v})/L^{k+1} \pi_1(S,\vec{v})\right]. \] Occasionally, it is better to think of this as a filtration of \(\Gamma_{g,n+\vec{r}}\) starting with \(J^0 \Gamma_{g,n+\vec{r}} = \Gamma_{g,n+\vec{r}}\). Passing to the associated graded, we obtain a graded Lie algebra \(\mathrm{Gr}^\bullet_J T_{g,n+\vec{r}}\) over \(\mathbb{Z}\) which comes with an injective homomorphism \[ \tau_\bullet \colon \mathrm{Gr}^\bullet_J T_{g,n+\vec{r}} \longrightarrow \mathrm{Der}\,\mathrm{Gr}^\bullet_L \pi_1(S,\vec{v}) \] to the derivation Lie algebra of the graded Lie algebra \(\mathrm{Gr}^\bullet_L \pi_1(S,\vec{v})\). Its restriction \(\tau_k\) to degree \(k\) is the \textit{\(k\)th higher Johnson homomorphism}; \(\tau_1\) for \((n,\vec{r}) = (0,1)\) recovers Johnson's original homomorphism. It is better to work with filtered objects than their associated gradeds, in particular after we base-change from \(\mathbb{Z}\) to \(\mathbb{Q}\). This approach originates in \textit{R. Hain}'s work [J. Am. Math. Soc. 10, No. 3, 597--651 (1997; Zbl 0915.57001)]. The relevant filtered objects are obtained by replacing \(\Gamma_{g,n+\vec{r}}\) with its \textit{relative unipotent completion} and \(\pi_1(S,\vec{v})\) by its \textit{unipotent completion} (also called \textit{Malcev completion}). These are proalgebraic groups over \(\mathbb{Q}\), and their Lie algebras are denoted \(\mathfrak{g}_{g,n+\vec{r}}\) and \(\mathfrak{p}(S,\vec{v})\) respectively. Then the filtered analogue of the above is a map of pronilpotent Lie algebras \[ \mathfrak{g}_{g,n+\vec{r}} \longrightarrow \mathrm{Der}\,\mathfrak{p}(S,\vec{v}), \] with filtration on the right induced by the lower central series of \(\mathfrak{p}(S,\vec{v})\) and the filtration on the left defined similarly to the Johnson filtration. Its image is contained in a certain sub-Lie algebra \(\mathrm{Der}^\theta\,\mathfrak{p}(S,\vec{v}) \subset \mathrm{Der}\,\mathfrak{p}(S,\vec{v})\). This survey is about the resulting map \(\mathfrak{g}_{g,n+\vec{r}} \to \mathrm{Der}^\theta\,\mathfrak{p}(S,\vec{v})\), called the \textit{geometric Johnson homomorphism}. It discusses three main themes: \textbf{1. Mixed Hodge structures and arithmetic Johnson homomorphisms}. The filtered objects in question come (after making some choices) with mixed Hodge structures. This has two consequences. On the one hand, the existence of mixed Hodge structures means that passing to associated gradeds does not nearly lose as much information as one might expect. On the other hand, it may be used to extend the geometric Johnson homomorphism to an \textit{arithmetic Johnson homomorphism}. Here, the domain \(\mathfrak{g}_{g,n+\vec{r}}\) is extended by a motivic Lie algebra, described explicitly using work of \textit{N. Takao} [J. Inst. Math. Jussieu 11, No. 1, 161--188 (2012; Zbl 1279.14037)] and \textit{F. Brown} [Ann. Math. (2) 175, No. 2, 949--976 (2012; Zbl 1278.19008)]. \textbf{2. Bounding the Johnson image using the Goldman-Turaev Lie bialgebra}. It is an open problem to determine the image of the higher Johnson homomorphism. To give an upper bound on its image, which is generated in degree 1 by work of \textit{R. Hain} [J. Am. Math. Soc. 10, No. 3, 597--651 (1997; Zbl 0915.57001)], one constructs homomorphisms out of its target which vanish on the image. The first such maps were given by \textit{S. Morita}'s trace map [Duke Math. J. 70, No. 3, 699--726 (1993; Zbl 0801.57011)]. Many refinements have been constructed since, and this survey focuses on a recent and powerful one due to \textit{N. Kawazumi} and \textit{Y. Kuno} [Ann. Inst. Fourier 65, No. 6, 2711--2762 (2015; Zbl 1370.57009)] and extended by Alekseev-Kawazumi-Kuno-Naef (cf. [\textit{A. Alekseev} et al., Adv. Math. 326, 1--53 (2018; Zbl 1422.57053) and ``The Goldman-Turaev Lie bialgebra and the Kashiwara-Vergne problem in higher genera'', Preprint, \url{arXiv:1804.09566}]). They proved there is an inclusion \(\mathrm{Der}^\theta\,\mathfrak{p}(S,\vec{v}) \hookrightarrow \mathbb{Q} \lambda(S)^\vee\) of the codomain of the geometric Johnson homomorphism into the \textit{completed Goldman-Turaev Lie bialgebra}, and that its image is almost annihilated by the cobracket. \textbf{3. A cohomological approach to Johnson homomorphisms.} There is also a conjectural cohomological approach to giving an upper bound of the Johnson image, which proceeds by understanding its domain. The Lie algebra \(\mathfrak{g}_{g,n+\vec{r}}\) fits in an extension \[ 0 \longrightarrow \mathfrak{u}_{g,n+\vec{r}} \longrightarrow \mathfrak{g}_{g,n+\vec{r}} \longrightarrow \mathfrak{sp}(H) \longrightarrow 0 \] with \(H = H_1(\overline{S};\mathbb{Q})\). Though \(\mathfrak{u}_{g,n+\vec{r}}\) admits an explicit quadratic presentation [Hain, Zbl 0915.57001], it is not easy to determine its decomposition into irreducible \(\mathrm{Sp}(H)\)-representations. For each \(\mathrm{Sp}(H)\)-representation \(V\) there is a map \[ F \colon \left(H^*(\mathfrak{u}_{g,n+\vec{r}}) \otimes V\right)^{\mathrm{Sp}(H)} \longrightarrow H^*(\Gamma_{g,n+\vec{r}};V) \] which is stably surjective by the Madsen-Weiss theorem [\textit{I. Madsen} and \textit{M. Weiss}, Ann. Math. (2) 165, No. 3, 843--941 (2007; Zbl 1156.14021)] and work of \textit{N. Kawazumi} and \textit{S. Morita} [Math. Res. Lett. 3, No. 5, 629--641 (1996; Zbl 0889.14009)]. The cohomology of \(\mathfrak{u}_{g,n+\vec{r}}\) inherits a mixed Hodge structure and if it were stably pure, \(F\) would be stably injective as well. This can then be used to determine the decomposition of \(\mathfrak{u}_{g,n+\vec{r}}\) in irreducible \(\mathrm{Sp}(H)\)-representations stably, as was done in unpublished work of Garoufalidis-Getzler. This survey is an invaluable resource for those who want to get an overview of the results in this field (up to the date of publication), which previously have been scattered through the literature. The problems, conjectures, and questions which appear throughout will be of particular interest to those working in this field.