Torelli Groups in Surface Topology

• The Torelli group is defined as the subgroup of the mapping class group that acts trivially on the first homology of a surface, serving as a fundamental invariant in low-dimensional topology.
• It is generated by separating twists and bounding pair maps, with explicit finite generating sets established for orientable surfaces of genus 3 and higher.
• Research on Torelli groups bridges geometric group theory, quantum topology, and representation theory, while addressing open problems in finite presentation and representation stability.

The Torelli group of a surface is the subgroup of the mapping class group consisting of those isotopy classes of diffeomorphisms that act trivially on the integral first homology. Torelli groups are fundamental invariants in low-dimensional topology, connecting mapping class group theory, quantum topology, and representation theory. Their structure, generation, finite presentability, and representation stability are central research themes in modern geometric group theory and algebraic geometry.

1. Definition and Core Properties

Let $\Sigma_{g,n}$ denote a compact, connected, oriented surface of genus $g$ with $n$ boundary components. The mapping class group $\mathrm{Mod}(\Sigma_{g,n})$ consists of isotopy classes of orientation-preserving homeomorphisms fixing the boundary pointwise. Torelli groups are defined as the kernel of the induced action on the first homology: $$I_{g,n} = \ker[\mathrm{Mod}(\Sigma_{g,n}) \to \mathrm{Sp}_{2g}(\mathbb{Z})],$$ where $\mathrm{Sp}_{2g}(\mathbb{Z})$ denotes the integral symplectic group acting on $H_1(\Sigma_{g,n};\mathbb{Z})$ via the algebraic intersection form. For surfaces with two boundary components, Putman's extension uses the relative homology $H_1(\Sigma_{g,2}, Q; \mathbb{Z})$ and defines $I_{g,2}$ as the kernel of the natural action preserving the distinguished isotropic line spanned by $\partial$-classes (Stylianakis, 9 Jan 2026).

For non-orientable surfaces $N_g$, the Torelli group $\mathcal{I}(N_g)$ is the subgroup of the mapping class group acting trivially on $H_1(N_g;\mathbb{Z})$ (Hirose et al., 2014). Torelli groups also generalize naturally to infinite-type surfaces, where they are closed normal subgroups of the mapping class group (Aramayona et al., 2018).

Torelli groups are deeply connected with the structure of the mapping class group: $$1 \to I_{g,n} \to \mathrm{Mod}(\Sigma_{g,n}) \to \mathrm{Sp}_{2g}(\mathbb{Z}) \to 1.$$

2. Generating Sets and Structural Results

For orientable closed (and once-bordered) surfaces of genus $g \geq 3$, the Torelli group is finitely generated (Putman, 2011, Stylianakis, 9 Jan 2026, Boldsen et al., 2011). Classical results, due to Birman–Powell and Putman, establish that it is generated by separating twists (Dehn twists about separating curves) and bounding-pair (BP) maps, i.e., products $T_cT_d^{-1}$ where $c,d$ are disjoint, homologous, nonseparating curves whose union separates the surface (Hatcher et al., 2011, Aramayona et al., 2018). In genus $g \geq 3$, every separating twist can be expressed as a product of BP maps, so BP maps suffice to generate the Torelli group for $g\geq3$ (Hatcher et al., 2011).

Putman constructed explicit generating sets of size $O(g^3)$ ("57 · $\binom{g}{3}$" elements) via the action on the handle graph, settling Johnson's conjecture and providing cubic growth in genus (Putman, 2011). For surfaces with two boundary components ($g\geq 3$), the Torelli group $I_{g,2}$ is also finitely generated, with a generating set of cubic size in $g$ (Stylianakis, 9 Jan 2026).

For non-orientable closed surfaces of genus $g\geq 4$, Hirose–Kobayashi gave infinite normal generating sets for $\mathcal{I}(N_g)$ consisting of bounding simple closed curve maps and suitable bounding-pair maps, but finite generation remains open (Hirose et al., 2014).

On infinite-type surfaces, the Torelli group is topologically generated by the subgroup of compactly supported separating twists and BP maps, and is the closure of this subgroup in the mapping class group topology (Aramayona et al., 2018).

3. Homological and Representation-Theoretic Properties

The cohomology and representation theory of Torelli groups form a rich subject. Johnson constructed, for $g\geq3$, the first Johnson homomorphism: $$\tau_1: I_{g,1} \rightarrow \bigwedge^3 H_1(\Sigma_g;\mathbb{Z}),$$ which is $Sp$–equivariant and completely determined by its values on BP maps. Higher Johnson homomorphisms map to successive quotients of the lower central series of $\pi_1(\Sigma_g)$. Tsuji constructed an embedding of $I_{g,1}$ into the completed Kauffman bracket skein algebra, which recovers the first Johnson homomorphism skein-theoretically and provides a topological quantum field theoretic framework for Johnson theory (Tsuji, 2016).

For the second rational homology, it is known that $H_2(I_{g,1};\mathbb{Q})$ is, for $g>7$, generated as an $Sp(2g,\mathbb{Z})$-module by the image from genus 6 via stabilization. The quotient of the arc complex modulo Torelli is $(g-2)$-connected, enabling a spectral sequence calculation showing representation stability for $H_2$ (Boldsen et al., 2011).

Church, Ellenberg, Farb, and Stylianakis have established degree bounds on polynomial growth of the (stable) rational cohomology of Torelli groups and their images under the Johnson homomorphism. For example, the dimension of the image of the induced map $$\psi_n : H_n(I_{g,1};\mathbb{Q}) \to \wedge^n(\wedge^3 H)$$ grows as a degree $3n$ polynomial in $g$ for large $g$ (Lindell, 2020).

Representation stability for the sequence of $Sp$-modules $\{H_k(I_{g,1};\mathbb{Q})\}_{g}$ remains partially open: surjectivity (generation from small genus) is established for $k=2$ (Boldsen et al., 2011), while injectivity and eventual stabilization are conjectural.

4. Filtrations, Further Algebraic Structures, and Graph Complexes

There are deep links between the Torelli group and filtered/algebraic structures:

• Johnson filtration: A central descending filtration of the mapping class group by the kernel of its action on nilpotent quotients of $\pi_1(\Sigma)$, with $I_g$ as the first term.
• Magnus kernel and filtration: The kernel of the Magnus representation (action on $F/F''$ for $F=\pi_1(\Sigma_{g,1})$) denoted $\mathrm{Mag}(\Sigma)$, admits a descending infinite filtration by subgroups $M_k(\Sigma)$ (higher Magnus kernels) whose quotients are shown to be large: each $M_k/M_{k+1}$ contains a copy of a lower central series quotient of a free group and admits an infinite rank abelian quotient (McNeill, 2013).
• Graph complex descriptions and symplectic character: The graded Lie algebra associated to the lower central series of $I_g$ (via $$t_g^k = \Gamma^k I_g / \Gamma^{k+1} I_g \otimes \mathbb{Q}$$) can be modeled through graph complexes, with representations governed by the quadratic dual of its presentation. For $g\gg k$,

$$t_g = \mathbb{L}(\langle 1^3 \rangle) / (R), \qquad R + \langle 2^2 \rangle + \langle 0 \rangle = \wedge^2(\langle 1^3 \rangle),$$

and explicit Sp-character tables are computed via plethystic formulas and graph enumeration (Garoufalidis et al., 2017). The conjecture that the associated algebra $A$ is Koszul would yield exact duality formulas for stable representation decompositions.

5. Topological and Quantum Aspects

The embedding of Torelli groups into completed skein algebras delivers a diagrammatic, topological quantum field theory perspective, which packages Dehn twists and BP maps as formal skein elements and connects classical Johnson theory to quantum invariants. This embedding organizes all generators of the Torelli group as skein elements and directly realizes the Johnson homomorphism as a graded piece (Tsuji, 2016).

Skein algebra embeddings and the connection to quantum representations (e.g., SO(3)-representations, Casson-type invariants) provide tools for constructing quantum invariants of 3-manifolds and studying the topological origin of Johnson-type homomorphisms.

6. Torelli Groups for Non-Orientable and Infinite-Type Surfaces

For non-orientable closed surfaces $N_g$, the Torelli group $\mathcal{I}(N_g)$ is normally generated by specified bounding simple closed curve (BSCC) maps and bounding pair maps with separating supports—explicitly, for $g\geq4$, by the minimal normal subgroup containing all BSCC maps cutting off $N_2^1$ and BP maps cutting off $\Sigma_1^2$ in their complements (Hirose et al., 2014). A finite generating set is not known.

For infinite-type surfaces, the Torelli group is topologically generated (with closure taken in the compact-open topology) by separating twists and BP maps supported on compact subsurfaces. The commensuration group of the Torelli group is isomorphic to the full mapping class group (Aramayona et al., 2018), mirroring known rigidity results for finite-type surfaces.

7. Open Problems and Recent Progress

Recent breakthroughs include:

• Finite generation of $I_{g,2}$ for $g\geq3$, resolving a longstanding question, and finite generation of the stabilizer of a nonseparating curve for $g\geq4$ (Stylianakis, 9 Jan 2026).
• Cubic-size explicit generating sets for $I_{g,n}$ in various settings (Putman, 2011, Stylianakis, 9 Jan 2026).
• Stable polynomial growth predictions for the homology and cohomology of Torelli groups, with explicit bounds on the range of stability and degree growth (Lindell, 2020).
• Modelings of the lower central series and stable representation via graph complexes and symmetric functions (Garoufalidis et al., 2017).

Central open problems include determining finite presentability of $I_g$ for $g\geq3$, finite generation of the Torelli group for non-orientable surfaces, explicit full description of higher homology, and full representation stability.

---

Summary Table: Torelli Group Generation

Surface	Generators	Finite Generation?	Reference
Orientable $g\geq3$	BP maps (and sep. twists)	Yes	(Putman, 2011, Hatcher et al., 2011)
$\Sigma_{g,1}$, $g\geq3$	BP maps, sep. twists	Yes	(Putman, 2011, Boldsen et al., 2011)
$\Sigma_{g,2}$, $g\geq3$	Chain maps, BP maps	Yes	(Stylianakis, 9 Jan 2026)
Non-orientable $g\geq4$	BSCC and BP maps	Unknown	(Hirose et al., 2014)
Infinite-type	Sep. twists, BP maps (closure)	Topological gen.	(Aramayona et al., 2018)

Key: BP = bounding pair, BSCC = bounding simple closed curve, sep. = separating.

---

• "Small generating sets for the Torelli group" (Putman, 2011)
• "The Torelli group and the Kauffman bracket skein module" (Tsuji, 2016)
• "Abelian Cycles in the Homology of the Torelli group" (Lindell, 2020)
• "Towards representation stability for the second homology of the Torelli group" (Boldsen et al., 2011)
• "Generating the Torelli group" (Hatcher et al., 2011)
• "Big Torelli groups: generation and commensuration" (Aramayona et al., 2018)
• "A normal generating set for the Torelli group of a non-orientable closed surface" (Hirose et al., 2014)
• "A new filtration of the Magnus kernel of the Torelli group" (McNeill, 2013)
• "Finiteness properties of the Torelli group of surfaces with 2 boundary components" (Stylianakis, 9 Jan 2026)
• "Graph complexes and the symplectic character of the Torelli group" (Garoufalidis et al., 2017)