Stable Commutator Length Overview

• Stable commutator length (scl) is a measure of the asymptotic efficiency of expressing elements as commutators, integrating algebraic, topological, and analytic perspectives.
• It is defined via limits on commutator length and interpreted topologically through negative Euler characteristics of bounding surfaces, with duality provided by homogeneous quasimorphisms.
• Current research explores scl's rationality, polyhedral structure, computational complexity, and applications in mapping class groups, braid groups, and random matrix theory.

Stable commutator length, usually denoted $\mathrm{scl}$, is the asymptotic commutator complexity of an element or chain in a group. It refines ordinary commutator length by stabilizing over powers, and it is simultaneously an algebraic, topological, and functional-analytic invariant: algebraically it measures efficient expression by commutators, topologically it is realized by normalized negative Euler characteristic of bounding surfaces, and analytically it is dual to homogeneous quasimorphisms via Bavard duality (Fournier-Facio, 18 Jul 2025). Across current research, stable commutator length appears in free groups, graphs of groups, mapping class groups, braid groups, right-angled Artin groups, one-relator groups, Baumslag–Solitar groups, big mapping class groups, and free $\mathbb{Q}$-groups, where rationality, positivity, spectral gaps, computability, and asymptotic distribution are studied in markedly different forms (Heuer et al., 2019).

1. Definitions and geometric formulations

For a group $G$, the commutator subgroup $[G,G]$ is generated by commutators $[x,y]=xyx^{-1}y^{-1}$. For $g\in [G,G]$, the commutator length is

$$\mathrm{cl}_G(g)=\min\Big\{n\;\Big|\; g=\prod_{i=1}^n [x_i,y_i],\ x_i,y_i\in G\Big\},$$

and the stable commutator length is

$$\mathrm{scl}_G(g)=\lim_{n\to\infty}\frac{\mathrm{cl}_G(g^n)}{n}.$$

Subadditivity implies that the limit exists and equals the infimum (Fournier-Facio, 18 Jul 2025). A standard extension used in several contexts assigns

$\mathrm{scl}_G(g)=\frac{\mathrm{scl}_G(g^k)}{k}$

when some power $g^k$ lies in $\mathbb{Q}$0, and $\mathbb{Q}$1 if no positive power does (Fournier-Facio, 18 Jul 2025).

The theory naturally extends from single elements to chains. If $\mathbb{Q}$2 denotes the real vector space with basis $\mathbb{Q}$3, then one may define commutator length and stable commutator length on integral chains $\mathbb{Q}$4, and then extend by homogeneity and continuity to rational and real chains. Quotienting $\mathbb{Q}$5 by the subspace generated by $\mathbb{Q}$6 and $\mathbb{Q}$7 yields $\mathbb{Q}$8; the image of boundaries gives $\mathbb{Q}$9, and scl induces a seminorm on $G$0 and a norm on $G$1, where $G$2 is the zero-scl subspace (Fournier-Facio, 18 Jul 2025).

The topological model is equally fundamental. For $G$3, $G$4 is the least genus of a once-punctured orientable surface mapping to $G$5 with boundary representing $G$6, and

$G$7

where $G$8 ranges over admissible maps whose boundary wraps with total degree $G$9 around a loop representing $[G,G]$0 (Brantner, 2011). In free groups this description is especially effective, and for one-relator groups it underlies the comparison between $[G,G]$1 and simplicial volume of the canonical $[G,G]$2-class (Heuer et al., 2019).

Basic formal properties recur throughout the literature: homogeneity $[G,G]$3, conjugacy invariance, and monotonicity under homomorphisms (Fournier-Facio, 18 Jul 2025). These properties are the common substrate for the very different geometric realizations seen in free products, mapping class groups, and rational completions.

2. Duality, quasimorphisms, and bounded cohomology

A quasimorphism on $[G,G]$4 is a function $[G,G]$5 with finite defect

$[G,G]$6

It is homogeneous if $[G,G]$7 for all $[G,G]$8. Homogeneous quasimorphisms are conjugacy invariant and additive on commuting elements (Fournier-Facio, 18 Jul 2025).

The classical dual formula is Bavard duality: $[G,G]$9 for $[x,y]=xyx^{-1}y^{-1}$0, with the supremum over non-homomorphisms since homomorphisms vanish on $[x,y]=xyx^{-1}y^{-1}$1 (Fournier-Facio, 18 Jul 2025). In chain form, the duality becomes a Banach-space identification: the dual Banach space of $[x,y]=xyx^{-1}y^{-1}$2, equipped with the scl norm, is naturally isomorphic to $[x,y]=xyx^{-1}y^{-1}$3, equipped with norm $[x,y]=xyx^{-1}y^{-1}$4 (Fournier-Facio, 18 Jul 2025). This identification makes homogeneous quasimorphisms modulo homomorphisms the continuous linear functionals on the scl-normed space of $[x,y]=xyx^{-1}y^{-1}$5-boundaries modulo zero-scl chains.

Several consequences are structurally decisive. If $[x,y]=xyx^{-1}y^{-1}$6, then scl vanishes on $[x,y]=xyx^{-1}y^{-1}$7; conversely, the existence of a nontrivial homogeneous quasimorphism detects positive scl on some chain (Fournier-Facio, 18 Jul 2025). If $[x,y]=xyx^{-1}y^{-1}$8 is isometric for scl on chains, then every homogeneous quasimorphism on $[x,y]=xyx^{-1}y^{-1}$9 extends to one on $g\in [G,G]$0 with the same defect, so scl-isometric embeddings induce surjections

$g\in [G,G]$1

(Fournier-Facio, 18 Jul 2025).

This dual perspective links scl to bounded cohomology. Bavard duality identifies scl as a norm dual to the defect-seminorm on bounded $g\in [G,G]$2-cocycles (Heuer et al., 2019), and for big mapping class groups the continuity of scl is proved by first showing that homogeneous quasimorphisms vanish on an open neighborhood and hence are continuous (Field et al., 2021). In braid groups, quasihomomorphisms into the concordance group give Lipschitz control on commutator length and thus on scl (Brandenbursky et al., 2014). The same general mechanism underlies explicit lower bounds in mapping class groups, Baumslag–Solitar groups, and hyperbolic actions on trees.

3. Rationality, polyhedrality, and computation

A large part of scl theory is organized around rationality and polyhedrality. For free groups, Calegari’s rationality theorem implies that scl of rational chains is always rational, and multiple later developments take this as the benchmark case (Fournier-Facio, 18 Jul 2025). In $g\in [G,G]$3, scl admits a polyhedral formulation in terms of flows on complete digraphs, Klein functions, and “sails,” leading to explicit asymptotic upper and lower bounds and to a sharp generic law: words of reduced length $g\in [G,G]$4 generically have scl near $g\in [G,G]$5 (Brantner, 2011). The same work shows that although extremal rays of the relevant polyhedra are completely classifiable, extremal points are universal enough to realize any connected directed multigraph, and computing scl on exponent-encoded words in $g\in [G,G]$6 is NP-hard unless $g\in [G,G]$7 (Brantner, 2011).

For graphs of groups, Chen proved that if a group acts on a tree with cyclic edge and vertex stabilizers, then scl is rational, and more generally that for graphs of groups whose vertex groups have trivial scl and whose incident edge images are central and mutually commensurable in each vertex group, scl is piecewise rational linear on $g\in [G,G]$8 and computable by linear programming (Chen, 2019). This covers generalized Baumslag–Solitar groups and subsumes several earlier rationality results for free products and amalgams. It also yields a Dehn-surgery-type phenomenon: in surgery families, scl varies predictably and converges to rational limits (Chen, 2019).

One-relator groups furnish a different but closely related polyhedral and surface-theoretic picture. For $g\in [G,G]$9 with $$\mathrm{cl}_G(g)=\min\Big\{n\;\Big|\; g=\prod_{i=1}^n [x_i,y_i],\ x_i,y_i\in G\Big\},$$0, the simplicial volume $$\mathrm{cl}_G(g)=\min\Big\{n\;\Big|\; g=\prod_{i=1}^n [x_i,y_i],\ x_i,y_i\in G\Big\},$$1 of the canonical $$\mathrm{cl}_G(g)=\min\Big\{n\;\Big|\; g=\prod_{i=1}^n [x_i,y_i],\ x_i,y_i\in G\Big\},$$2-class is often related linearly to $$\mathrm{cl}_G(g)=\min\Big\{n\;\Big|\; g=\prod_{i=1}^n [x_i,y_i],\ x_i,y_i\in G\Big\},$$3 by

$$\mathrm{cl}_G(g)=\min\Big\{n\;\Big|\; g=\prod_{i=1}^n [x_i,y_i],\ x_i,y_i\in G\Big\},$$4

and this relation holds exactly for free-product and HNN-type decomposable relators, approximately for proper powers and $$\mathrm{cl}_G(g)=\min\Big\{n\;\Big|\; g=\prod_{i=1}^n [x_i,y_i],\ x_i,y_i\in G\Big\},$$5 relators, and asymptotically for random relators (Heuer et al., 2019). This suggests that the rational polyhedral behavior of scl can control other $$\mathrm{cl}_G(g)=\min\Big\{n\;\Big|\; g=\prod_{i=1}^n [x_i,y_i],\ x_i,y_i\in G\Big\},$$6-type invariants of groups.

Baumslag–Solitar groups occupy an intermediate position. Clay–Forester–Louwsma established lower bounds for general elements, showed that scl is computable and rational for elements of alternating $$\mathrm{cl}_G(g)=\min\Big\{n\;\Big|\; g=\prod_{i=1}^n [x_i,y_i],\ x_i,y_i\in G\Big\},$$7-shape, and characterized exactly which such elements admit extremal surfaces (Clay et al., 2013). Chen later reinterpreted parts of this through rationality in graphs of groups and proved that for $$\mathrm{cl}_G(g)=\min\Big\{n\;\Big|\; g=\prod_{i=1}^n [x_i,y_i],\ x_i,y_i\in G\Big\},$$8-alternating words, scl in $$\mathrm{cl}_G(g)=\min\Big\{n\;\Big|\; g=\prod_{i=1}^n [x_i,y_i],\ x_i,y_i\in G\Big\},$$9 agrees with scl in $$\mathrm{scl}_G(g)=\lim_{n\to\infty}\frac{\mathrm{cl}_G(g^n)}{n}.$$0 (Chen, 2019).

4. Positivity, gaps, and zero loci

A central qualitative dichotomy in the subject is between positivity and vanishing. In some groups, every nontrivial element has positive scl; in others, large families have scl zero.

A recent example of strong positivity is the free $$\mathrm{scl}_G(g)=\lim_{n\to\infty}\frac{\mathrm{cl}_G(g^n)}{n}.$$1-group $$\mathrm{scl}_G(g)=\lim_{n\to\infty}\frac{\mathrm{cl}_G(g^n)}{n}.$$2, the uniquely divisible completion of a free group. Every non-identity element has positive scl, and the inclusion $$\mathrm{scl}_G(g)=\lim_{n\to\infty}\frac{\mathrm{cl}_G(g^n)}{n}.$$3 is isometric for scl on chains (Fournier-Facio, 18 Jul 2025). At the same time, there is no spectral gap: if $$\mathrm{scl}_G(g)=\lim_{n\to\infty}\frac{\mathrm{cl}_G(g^n)}{n}.$$4 has positive scl, then

$$\mathrm{scl}_G(g)=\lim_{n\to\infty}\frac{\mathrm{cl}_G(g^n)}{n}.$$5

so the positive spectrum accumulates at zero (Fournier-Facio, 18 Jul 2025). This shows that positivity and gap phenomena are logically distinct.

In mapping class groups of finite type, positivity is governed by Nielsen–Thurston decomposition. For a finite-index subgroup $$\mathrm{scl}_G(g)=\lim_{n\to\infty}\frac{\mathrm{cl}_G(g^n)}{n}.$$6, an element $$\mathrm{scl}_G(g)=\lim_{n\to\infty}\frac{\mathrm{cl}_G(g^n)}{n}.$$7 has positive scl exactly when some chiral equivalence class of pure components is essential; if $$\mathrm{scl}_G(g)=\lim_{n\to\infty}\frac{\mathrm{cl}_G(g^n)}{n}.$$8, then it is uniformly bounded away from zero by a constant depending only on $$\mathrm{scl}_G(g)=\lim_{n\to\infty}\frac{\mathrm{cl}_G(g^n)}{n}.$$9 and $\mathrm{scl}_G(g)=\frac{\mathrm{scl}_G(g^k)}{k}$0 (Bestvina et al., 2013). In particular, there exists a finite-index subgroup on which every nontrivial element has positive scl, and every nontrivial element of the Torelli group has positive scl (Bestvina et al., 2013). Hyperelliptic mapping class groups refine the quantitative side: the scl of a nonseparating Dehn twist satisfies

$\mathrm{scl}_G(g)=\frac{\mathrm{scl}_G(g^k)}{k}$1

while certain explicit elements $\mathrm{scl}_G(g)=\frac{\mathrm{scl}_G(g^k)}{k}$2 have exactly

$\mathrm{scl}_G(g)=\frac{\mathrm{scl}_G(g^k)}{k}$3

(Calegari et al., 2012).

Right-angled Artin groups also exhibit uniform positivity. If the defining graph has chromatic number $\mathrm{scl}_G(g)=\frac{\mathrm{scl}_G(g^k)}{k}$4, then every nontrivial element has

$\mathrm{scl}_G(g)=\frac{\mathrm{scl}_G(g^k)}{k}$5

and if the defining graph is triangle-free, then every nontrivial element has

$\mathrm{scl}_G(g)=\frac{\mathrm{scl}_G(g^k)}{k}$6

(Forester et al., 2017). The proofs are geometric, using surfaces in Salvetti complexes and a non-overlapping theorem for subwords. The same paper notes that stronger $\mathrm{scl}_G(g)=\frac{\mathrm{scl}_G(g^k)}{k}$7 bounds were obtained by Heuer via quasimorphisms (Forester et al., 2017).

For groups acting on trees, the universal lower bound is $\mathrm{scl}_G(g)=\frac{\mathrm{scl}_G(g^k)}{k}$8. Clay–Forester–Louwsma constructed homogeneous counting quasimorphisms for tree actions with defect at most $\mathrm{scl}_G(g)=\frac{\mathrm{scl}_G(g^k)}{k}$9, and proved that any well-aligned hyperbolic element satisfies

$g^k$0

(Clay et al., 2013). In $g^k$1, the element $g^k$2 has scl exactly $g^k$3, showing the bound is sharp (Clay et al., 2013). For Baumslag–Solitar groups $g^k$4 with $g^k$5, they furthermore proved a spectral gap: $g^k$6 for every element $g^k$7 (Clay et al., 2013).

5. Surface groups, braids, mapping classes, and large-scale behavior

Mapping class and braid groups provide some of the richest interactions between scl and topology. In hyperelliptic mapping class groups, scl bounds for Dehn twists are obtained from explicit $g^k$8-signature quasimorphisms, including exact defect computations for the Meyer function and exact values for certain high-complexity elements (Calegari et al., 2012). In finite-index subgroups of $g^k$9, the essential-chiral criterion and the uniform gap theorem show that scl is determined by the decomposition of a mapping class into pseudo-Anosov and multitwist components (Bestvina et al., 2013).

For infinite-type surfaces, the behavior is topological rather than purely algebraic. If $\mathbb{Q}$00 is an infinite-type surface whose mapping class group is locally coarsely bounded and whose maximal end classes satisfy the hypotheses of Mann–Rafi theory, then the commutator subgroup of $\mathbb{Q}$01 is open and closed, and

$\mathbb{Q}$02

is continuous (Field et al., 2021). The same work shows that in many such cases the abelianization is finitely generated and discrete (Field et al., 2021). This suggests that, in big mapping class groups, scl is compatible with the ambient Polish group topology in a way absent from discrete finite-type settings.

Braid groups connect scl to $\mathbb{Q}$03-dimensional topology. A quasihomomorphism

$\mathbb{Q}$04

to the concordance group satisfies defect $\mathbb{Q}$05 and is Lipschitz with respect to the conjugation-invariant word norm and commutator length (Brandenbursky et al., 2014). For $\mathbb{Q}$06, the stable four-ball genus obeys

$\mathbb{Q}$07

so vanishing scl forces uniform bounds on stable four-ball genus in the concordance group (Brandenbursky et al., 2014). In the infinite braid group $\mathbb{Q}$08, commutator length is bounded and hence scl vanishes identically, yet the commutator subgroup admits a stably unbounded conjugation-invariant norm, answering an open problem of Burago, Ivanov, and Polterovich (Brandenbursky et al., 2014).

Large-scale probabilistic behavior is now also well developed. For random geodesic words or random walks of length $\mathbb{Q}$09 in hyperbolic groups, and more generally in groups acting nondegenerately on hyperbolic spaces, scl is of order $\mathbb{Q}$10 with high probability (Calegari et al., 2010). In free groups, the normalized quantity $\mathbb{Q}$11 concentrates near $\mathbb{Q}$12 (Calegari et al., 2010). This situates scl among asymptotic random invariants and gives quantitative control on random finite-dimensional sections of the scl norm ball.

6. Current directions: rational completions, irrational values, and random matrices

Recent work has considerably expanded the frontier of the subject. Free $\mathbb{Q}$13-groups illustrate a new completion phenomenon: the free group embeds isometrically for scl into its $\mathbb{Q}$14-completion, every nontrivial element in the completion has positive scl, and the space $\mathbb{Q}$15 is infinite-dimensional (Fournier-Facio, 18 Jul 2025). The same paper conjectures that scl is rational on free $\mathbb{Q}$16-groups, and shows that this conjecture would imply rationality for many root extensions, including non-orientable surface groups isometrically embedded in free $\mathbb{Q}$17-groups (Fournier-Facio, 18 Jul 2025). Since rationality for surface groups is a longstanding open problem, this places rationality on free $\mathbb{Q}$18-groups at the center of a broader unresolved program.

The irrational side of the theory changed decisively with the construction of algebraic irrational scl in finitely presented groups. In the total lift of the golden ratio Thompson group $\mathbb{Q}$19, where $\mathbb{Q}$20, the scl spectrum contains

$\mathbb{Q}$21

providing the first finitely presented, indeed type $\mathbb{Q}$22, example with algebraic irrational scl values (Fournier-Facio et al., 2021). The computation uses the rotation quasimorphism, which spans the space of homogeneous quasimorphisms on the lifted group, so scl is exactly half the absolute rotation number there (Fournier-Facio et al., 2021). This resolves a question of Calegari and shows that rationality phenomena are not universal even in highly structured finitely presented groups.

A further conceptual development is the random-matrix reformulation of scl. Magee and Puder showed that for a word $\mathbb{Q}$23, stable commutator length can be expressed through stable Fourier coefficients of $\mathbb{Q}$24-random unitary matrices; subsequent work extended this viewpoint to other stable invariants of words and other families of groups (Puder et al., 21 Sep 2025). In particular, the infimum of decay exponents of suitable stable Fourier coefficients recovers $\mathbb{Q}$25 in the unitary case, while analogues involving symmetric groups and wreath products capture stable primitivity rank and related invariants (Puder et al., 21 Sep 2025). This suggests that scl belongs to a wider class of “stable invariants of words” detectable by asymptotic representation theory.

The present landscape is therefore sharply stratified. Rationality and piecewise rational linearity hold in free groups, many free products, graphs of groups with cyclic data, and large families of alternating words in Baumslag–Solitar groups [(Brantner, 2011); (Chen, 2019); (Clay et al., 2013)]. Exact positivity and spectral gaps are known in right-angled Artin groups, mapping class groups, Torelli groups, and Baumslag–Solitar groups [(Forester et al., 2017); (Bestvina et al., 2013); (Clay et al., 2013)]. At the same time, algebraic irrational values occur in finitely presented groups (Fournier-Facio et al., 2021), and rationality for surface groups remains open, with free $\mathbb{Q}$26-groups now offering a new, closely related test case (Fournier-Facio, 18 Jul 2025).