Johnson Kernel in Mapping Class Groups

• The Johnson kernel is the second stage of the Johnson filtration, defined as the kernel of the first Johnson homomorphism and characterized by its action on nilpotent quotients of the fundamental group.
• It is generated by Dehn twists about separating curves, reflecting its geometric and algebraic significance in controlling mapping class group structures.
• Studies reveal that the Johnson kernel exhibits finite generation in high genus with intricate homological properties, playing a central role in Torelli theory and quantum representation stability.

The Johnson kernel is the second stage of the Johnson filtration of a mapping class group, lying between the Torelli group and the deeper Johnson subgroups. For a genus $g$ surface with one boundary component, it is $\mathcal I_g^1(2)$, the kernel of the action of $\mathrm{Mod}_g^1$ on $\pi_1(\Sigma_g^1)/\gamma_3(\pi_1(\Sigma_g^1))$; for a closed genus $g$ surface, it is the subgroup $\mathcal K_g\subset \mathcal I_g$ generated by Dehn twists about separating curves and, equivalently, the kernel of the first Johnson homomorphism (Church et al., 2013, Spiridonov, 2021). It is a central object in Torelli theory because it is the first term of the Johnson filtration not detected by the first Johnson homomorphism, yet it remains large enough to control substantial geometry, topology, and representation-theoretic structure.

1. Definitions, conventions, and place in the Johnson filtration

For a compact oriented surface $\Sigma_g^1$ with one boundary component and basepoint $\ast\in \partial \Sigma_g^1$, the lower central series of $\pi=\pi_1(\Sigma_g^1,\ast)$ is

$$\gamma_1(\pi)=\pi,\qquad \gamma_{k+1}(\pi)=[\gamma_k(\pi),\pi].$$

The Johnson filtration of $\mathcal I_g^1(2)$0 is

$\mathcal I_g^1(2)$1

with $\mathcal I_g^1(2)$2 the Torelli group and $\mathcal I_g^1(2)$3 the Johnson kernel (Church et al., 2013). In the closed case, the corresponding subgroup is denoted $\mathcal I_g^1(2)$4, and in the notation $\mathcal I_g^1(2)$5 it is the second term of the Johnson filtration for $\mathcal I_g^1(2)$6 (Ershov et al., 2017, Detcherry, 31 Mar 2026).

Several standard notational conventions coexist in the literature.

Setting	Notation	Description
One boundary component	$\mathcal I_g^1(2)$7, $\mathcal I_g^1(2)$8	Second Johnson term; kernel of the first Johnson homomorphism
Closed surface	$\mathcal I_g^1(2)$9, $\mathrm{Mod}_g^1$0 with $\mathrm{Mod}_g^1$1	Subgroup generated by separating twists
General $\mathrm{Mod}_g^1$2 notation	$\mathrm{Mod}_g^1$3	Johnson kernel as second filtration term
Mod-$\mathrm{Mod}_g^1$4 analogues	$\mathrm{Mod}_g^1$5, $\mathrm{Mod}_g^1$6	Kernels of first mod-$\mathrm{Mod}_g^1$7 Johnson homomorphisms

For the bordered surface $\mathrm{Mod}_g^1$8, the first Johnson homomorphism is

$\mathrm{Mod}_g^1$9

with image $\pi_1(\Sigma_g^1)/\gamma_3(\pi_1(\Sigma_g^1))$0, and its kernel is

$\pi_1(\Sigma_g^1)/\gamma_3(\pi_1(\Sigma_g^1))$1

(Cooper, 2014). For closed surfaces, Johnson constructed a surjective homomorphism

$\pi_1(\Sigma_g^1)/\gamma_3(\pi_1(\Sigma_g^1))$2

whose kernel is precisely $\pi_1(\Sigma_g^1)/\gamma_3(\pi_1(\Sigma_g^1))$3 (Spiridonov, 2021). This dual description—filtration-theoretic and homomorphism-theoretic—is one reason the Johnson kernel is simultaneously a geometric and algebraic invariant.

2. Classical geometric structure and generators

The Johnson kernel is classically characterized by separating twists. In the one-boundary convention used in the Johnson-filtration literature, a genus $\pi_1(\Sigma_g^1)/\gamma_3(\pi_1(\Sigma_g^1))$4 separating twist is a Dehn twist $\pi_1(\Sigma_g^1)/\gamma_3(\pi_1(\Sigma_g^1))$5 about a separating curve $\pi_1(\Sigma_g^1)/\gamma_3(\pi_1(\Sigma_g^1))$6 cutting $\pi_1(\Sigma_g^1)/\gamma_3(\pi_1(\Sigma_g^1))$7 into subsurfaces homeomorphic to $\pi_1(\Sigma_g^1)/\gamma_3(\pi_1(\Sigma_g^1))$8 and $\pi_1(\Sigma_g^1)/\gamma_3(\pi_1(\Sigma_g^1))$9. Johnson proved that $g$0 is generated by genus $g$1 and $g$2 separating twists (Church et al., 2013). In the closed-surface case, $g$3 is generated by all Dehn twists about separating simple closed curves (Gaifullin, 2019).

The kernel characterization is equally fundamental. In the closed case,

$g$4

so $g$5 is exactly the part of the Torelli group invisible to the first Johnson homomorphism (Spiridonov, 2021). In the bordered case,

$g$6

and Johnson’s computation of $g$7 on genus-$g$8 bounding pair maps identifies the first nontrivial quotient of the Torelli group, with the Johnson kernel as its kernel (Cooper, 2014).

The action of the Johnson kernel on curves is much more rigid than the action of the full Torelli group. For $g$9, two homologous oriented nonseparating curves $\mathcal K_g\subset \mathcal I_g$0 are equivalent under $\mathcal K_g\subset \mathcal I_g$1 if and only if, for representatives $\mathcal K_g\subset \mathcal I_g$2, the class

$\mathcal K_g\subset \mathcal I_g$3

lies in $\mathcal K_g\subset \mathcal I_g$4, where $\mathcal K_g\subset \mathcal I_g$5. For separating curves cutting off the same symplectic subspace $\mathcal K_g\subset \mathcal I_g$6, equivalence under $\mathcal K_g\subset \mathcal I_g$7 is controlled by the condition

$\mathcal K_g\subset \mathcal I_g$8

and this is also the criterion for the separating twists $\mathcal K_g\subset \mathcal I_g$9 and $\Sigma_g^1$0 to be conjugate inside $\Sigma_g^1$1 (Church, 2011). These criteria show that Johnson-kernel orbits are governed by nilpotent invariants of $\Sigma_g^1$2, not merely by homology.

3. Uniform generation and support complexity

A central development in the modern theory is that the Johnson kernel participates in a uniform bounded-support generation phenomenon. For each $\Sigma_g^1$3, there exists $\Sigma_g^1$4 such that for all $\Sigma_g^1$5, the subgroup $\Sigma_g^1$6 is generated by elements supported on homologically standard subsurfaces of genus at most $\Sigma_g^1$7, with $\Sigma_g^1$8 depending only on $\Sigma_g^1$9, not on $\ast\in \partial \Sigma_g^1$0 (Church et al., 2013). Applied to $\ast\in \partial \Sigma_g^1$1, this gives a uniform support bound for the Johnson kernel: $\ast\in \partial \Sigma_g^1$2 is generated by mapping classes supported on homologically standard subsurfaces of bounded genus independent of $\ast\in \partial \Sigma_g^1$3 (Church et al., 2013).

For the Johnson kernel itself, classical work is sharper: Johnson’s generation theorem by genus $\ast\in \partial \Sigma_g^1$4 and $\ast\in \partial \Sigma_g^1$5 separating twists shows that one may take $\ast\in \partial \Sigma_g^1$6 in the bordered setting, although the FI-module proof does not recover this explicit value (Church et al., 2013). At the same time, the lower-bound theorem in the same paper shows that for all $\ast\in \partial \Sigma_g^1$7, $\ast\in \partial \Sigma_g^1$8 is not generated by elements supported on subsurfaces of genus $\ast\in \partial \Sigma_g^1$9. For $\pi=\pi_1(\Sigma_g^1,\ast)$0, this implies that the Johnson kernel is not generated by genus $\pi=\pi_1(\Sigma_g^1,\ast)$1 supports, so any uniform generation must involve positive-genus pieces (Church et al., 2013).

The significance is conceptual. The proof uses central filtrations of weak FI-groups, FI-modules, and central stability over $\pi=\pi_1(\Sigma_g^1,\ast)$2, and is non-constructive because it relies on the Noetherian property for FI-modules (Church et al., 2013). This suggests that the Johnson kernel is part of a broader representation-stability regime: its generators may be chosen with uniformly bounded topological complexity across genus, even though explicit optimal bounds remain a separate problem.

4. Finiteness properties and abelianization

The finiteness theory of the Johnson kernel has two distinct aspects: finite generation of the group itself, and finite generation of its abelianization. For $\pi=\pi_1(\Sigma_g^1,\ast)$3 with $\pi=\pi_1(\Sigma_g^1,\ast)$4, any subgroup containing $\pi=\pi_1(\Sigma_g^1,\ast)$5 is finitely generated. Since

$\pi=\pi_1(\Sigma_g^1,\ast)$6

this implies that the Johnson kernel is finitely generated in genus $\pi=\pi_1(\Sigma_g^1,\ast)$7 (Ershov et al., 2017). In the same range, its abelianization is finitely generated, and more precisely

$\pi=\pi_1(\Sigma_g^1,\ast)$8

is nilpotent (Ershov et al., 2017).

The abelianization is also accessible from a different direction. For all $\pi=\pi_1(\Sigma_g^1,\ast)$9, $$\gamma_1(\pi)=\pi,\qquad \gamma_{k+1}(\pi)=[\gamma_k(\pi),\pi].$$0 is a non-trivial unipotent $$\gamma_1(\pi)=\pi,\qquad \gamma_{k+1}(\pi)=[\gamma_k(\pi),\pi].$$1-module, and for $$\gamma_1(\pi)=\pi,\qquad \gamma_{k+1}(\pi)=[\gamma_k(\pi),\pi].$$2 it admits an explicit presentation as a $$\gamma_1(\pi)=\pi,\qquad \gamma_{k+1}(\pi)=[\gamma_k(\pi),\pi].$$3-module (Dimca et al., 2011). A complementary calculation determines the rational abelianization explicitly: for $$\gamma_1(\pi)=\pi,\qquad \gamma_{k+1}(\pi)=[\gamma_k(\pi),\pi].$$4,

$$\gamma_1(\pi)=\pi,\qquad \gamma_{k+1}(\pi)=[\gamma_k(\pi),\pi].$$5

where the $$\gamma_1(\pi)=\pi,\qquad \gamma_{k+1}(\pi)=[\gamma_k(\pi),\pi].$$6-summand is generated by the secondary class $$\gamma_1(\pi)=\pi,\qquad \gamma_{k+1}(\pi)=[\gamma_k(\pi),\pi].$$7, and the other summands arise from a refinement of the second Johnson homomorphism (Morita et al., 2017).

These results separate the Johnson kernel from naive expectations of homological smallness. Its first homology is finite-dimensional over $$\gamma_1(\pi)=\pi,\qquad \gamma_{k+1}(\pi)=[\gamma_k(\pi),\pi].$$8 in sufficiently large genus, but it is not trivial; it carries nontrivial $$\gamma_1(\pi)=\pi,\qquad \gamma_{k+1}(\pi)=[\gamma_k(\pi),\pi].$$9-module structure and is closely related to infinitesimal Alexander invariants and the discrepancy between the lower central series of the Torelli group and the Johnson filtration (Dimca et al., 2011, Morita et al., 2017).

5. Homology, cohomology, and geometric applications

The Johnson kernel has striking higher-homological behavior. Bestvina–Bux–Margalit showed that its cohomological dimension is

$\mathcal I_g^1(2)$00

and Gaifullin proved that the top homology $\mathcal I_g^1(2)$01 is not finitely generated: it contains a free abelian subgroup of infinite rank, so $\mathcal I_g^1(2)$02 is infinite-dimensional (Gaifullin, 2019). He further showed that this top homology is not finitely generated as a $\mathcal I_g^1(2)$03-module (Gaifullin, 2019).

A more refined description of one geometric piece of top homology is obtained from maximal collections of disjoint separating curves. For any $\mathcal I_g^1(2)$04 disjoint separating curves on $\mathcal I_g^1(2)$05, the associated commuting separating twists determine an abelian cycle in $\mathcal I_g^1(2)$06. The subgroup $\mathcal I_g^1(2)$07 generated by these simplest abelian cycles satisfies

$\mathcal I_g^1(2)$08

as a $\mathcal I_g^1(2)$09-module, where $\mathcal I_g^1(2)$10 is free abelian of rank $\mathcal I_g^1(2)$11. It has a presentation with generators indexed by trivalent trees and relations

$\mathcal I_g^1(2)$12

for cyclic triples of trees, and balanced trees furnish a basis (Spiridonov, 2021). This gives a concrete combinatorial model for a large, natural part of the top homology.

The Johnson kernel also has strong consequences for surface bundles. If a surface bundle has monodromy contained in $\mathcal I_g^1(2)$13, then its cohomology ring is isomorphic to that of a trivial bundle,

$\mathcal I_g^1(2)$14

and all higher Johnson invariants vanish upon restriction to $\mathcal I_g^1(2)$15 (Salter, 2016). In a related direction, any nontrivial surface bundle over a surface with monodromy in the Johnson kernel fibers in a unique way (Salter, 2014). This suggests that the Johnson kernel, despite being large as a group, is cohomologically rigid in the context of surface-bundle topology.

6. Variants, low-genus developments, and current directions

The Johnson-kernel paradigm has natural mod-$\mathcal I_g^1(2)$16 analogues. Using the Stallings and Zassenhaus mod-$\mathcal I_g^1(2)$17 central series of $\mathcal I_g^1(2)$18, one defines mod-$\mathcal I_g^1(2)$19 Johnson filtrations $\mathcal I_g^1(2)$20 and $\mathcal I_g^1(2)$21, with first kernels

$\mathcal I_g^1(2)$22

For $\mathcal I_g^1(2)$23, the Zassenhaus mod-$\mathcal I_g^1(2)$24 Johnson kernel is generated by separating twists and $\mathcal I_g^1(2)$25-th powers of Dehn twists, while the Stallings mod-$\mathcal I_g^1(2)$26 Johnson kernel is generated by separating twists, $\mathcal I_g^1(2)$27-th powers of bounding pair maps, and $\mathcal I_g^1(2)$28-powers of Dehn twists (Cooper, 2014). These groups are natural mod-$\mathcal I_g^1(2)$29 thickenings of the classical Johnson kernel.

Recent work also connects the Johnson kernel to quantum representations. For any nontrivial element

$\mathcal I_g^1(2)$30

the $\mathcal I_g^1(2)$31-WRT quantum representation $\mathcal I_g^1(2)$32 has infinite order for all sufficiently large prime $\mathcal I_g^1(2)$33. More generally, for any nontrivial $\mathcal I_g^1(2)$34, $\mathcal I_g^1(2)$35 has either order $\mathcal I_g^1(2)$36 or infinite order for large enough prime $\mathcal I_g^1(2)$37 (Detcherry, 31 Mar 2026). In the same paper, this implies that every nontrivial element of the derived subgroup $\mathcal I_g^1(2)$38 has a pseudo-Anosov part (Detcherry, 31 Mar 2026).

Low-genus behavior remains delicate. The torsion subgroup of the abelianized Johnson kernel is non-trivial for genus $\mathcal I_g^1(2)$39, and explicit lower bounds for its cardinality are obtained by infinitesimal methods based on the action on the Malcev Lie algebra (Faes et al., 2022). More recently, the abelianizations $\mathcal I_g^1(2)$40 for $\mathcal I_g^1(2)$41 were proved finitely generated, settling the genus $\mathcal I_g^1(2)$42 abelianization problem while leaving finite generation of $\mathcal I_g^1(2)$43 itself open (Gaifullin, 28 Jul 2025).

Several open directions recur across the literature. One asks for effective support-genus bounds in uniform generation theorems (Church et al., 2013). Another asks whether the subgroup $\mathcal I_g^1(2)$44 of top homology generated by simplest abelian cycles exhausts $\mathcal I_g^1(2)$45 (Spiridonov, 2021). Low-genus finiteness questions remain central, especially the finite generation of $\mathcal I_g^1(2)$46 itself (Gaifullin, 2019, Gaifullin, 28 Jul 2025). Together these problems indicate that the Johnson kernel is now understood simultaneously as a filtration term, a geometrically generated subgroup, a homologically large object, and a testing ground for stability, quantum, and arithmetic methods.