---
title: 'Small Artin Groups: Definitions and Frameworks'
url: https://www.emergentmind.com/topics/small-artin-groups
type: topic
---

# Small Artin Groups: Definitions and Frameworks

“Small Artin groups” is not a uniform term. In the usual Artin-group terminology, it refers to **small-type** Artin groups, meaning that all defining labels are \(2\) or \(3\). In other parts of the literature, however, “small” is used for low-rank Artin groups, for complete parabolic subgroups that are small in graph diameter, or for a representation-theoretic class defined by integrality of a specialization of Squier’s generalized Burau representation. The subject therefore consists less of a single canonical class than of a family of closely related low-complexity regimes in the general theory of Artin groups [2507.16770][1105.1048][2201.13044][2510.23002].

## 1. Terminology and competing meanings

The principal source of ambiguity is that different papers isolate different notions of “smallness.”

| Usage | Defining condition | Context |
|---|---|---|
| Standard small type | all finite labels are \(2\) or \(3\) | usual Artin-group terminology |
| Low-rank small groups | rank at most \(2\) | subgroup separability and structural decompositions |
| Small parabolic pieces | complete induced subgraphs | parabolic intersection and Bass–Serre reductions |
| Squier-integral small groups | exponents in \(\{1,2,3,\infty\}\) | congruence subgroup theory |

In the standard usage, a small Artin group has only labels \(2\) and \(3\). This is the sense explicitly contrasted with Juh’s 2025 class of Artin groups with **no** \(3\)-labels: those groups have labels in \(\{0,2,4,5,6,\dots\}\), so they are not the same class, and their intersection with standard small Artin groups is essentially the right-angled case [2507.16770].

A second usage treats “small Artin groups” as the decisive **rank-\(\le 2\)** pieces from which certain higher-rank groups are assembled. This is particularly visible in subgroup separability, where rank \(1\) and rank \(2\) Artin groups are the atomic LERF building blocks [2010.13222].

A third usage concerns **complete standard parabolic subgroups**. These are “small” not in rank, but in graph diameter: their defining induced subgraphs are complete, hence of diameter at most \(1\), and they function as the “puzzle pieces” in reduction arguments for intersections, fixed point properties, and automatic continuity [2201.13044].

A fourth, explicitly terminological, definition appears in the congruence-subgroup literature: a group is called a **small Artin group** if it admits a Coxeter system with all exponents in \(\{1,2,3,\infty\}\), equivalently if Squier’s generalized Burau representation becomes integral after specializing \(s=1\) and \(t=-1\) [2510.23002].

A probabilistic perspective sharpens how restrictive these families are. In the random Artin-group model with rank \(N\) and permitted labels controlled by a function \(f(N)\), \(FC\)-type groups, RAAGs, triangle-free Artin groups, and spherical Artin groups are **uniformly small**, whereas \(2\)-dimensional Artin groups have threshold scale \(N^{3/2}\), and large-type, extra-large-type, and free-of-infinity groups have threshold scale \(N^2\) [2301.04211].

## 2. Presentations and canonical subclasses

An Artin group is defined from a finite simplicial graph \(\Gamma\) with vertex set \(V(\Gamma)\) and edge labels \(m_{uv}\ge 2\). If \(u,v\in V(\Gamma)\) are adjacent, the defining relation is the usual alternating braid relation
\[
\underbrace{uvuv\cdots}_{m_{uv}\text{ letters}}
=
\underbrace{vuvu\cdots}_{m_{uv}\text{ letters}},
\]
and if there is no edge, there is no Artin relation between those generators. Equivalently, the label data may be encoded by a Coxeter matrix \((m_{uv})\), with standard parabolic subgroups obtained by passing to full induced subgraphs [2010.13222].

In this framework, the main standard subclasses are:

- **Small type**: all finite labels are \(2\) or \(3\).
- **Right-angled Artin groups**: all labels are \(2\) or \(\infty\).
- **Even Artin groups**: every finite label is even.
- **Large type**: all finite labels are at least \(3\).
- **Extra-large type**: all finite labels are at least \(4\).
- **Spherical type**: the associated Coxeter group is finite.

Among irreducible spherical Artin groups, the small-type cases are precisely
\[
A_n,\quad D_n,\quad E_6,\quad E_7,\quad E_8.
\]
This list is important because it isolates the simply laced finite-type families inside the broader spherical classification [2010.06046].

Rank is the number of vertices of the defining graph. In low rank, the dihedral groups
\[
A(m)=\langle a,b\mid \underbrace{abab\cdots}_{m}=\underbrace{baba\cdots}_{m}\rangle
\]
form the complete rank-\(2\) family, with special cases \(m=2\) giving \(\mathbb Z^2\), \(m=3\) giving type \(A_2\), \(m=4\) giving type \(B_2\), and \(m\ge 5\) giving type \(I_2(m)\) [2010.13222].

## 3. Low-rank small Artin groups as building blocks

One of the clearest senses in which small Artin groups are fundamental is subgroup separability. The graph-theoretic class \(\mathcal S\) introduced for this purpose starts from graphs with at most two vertices and is closed under disjoint union and **2-cone extension**. The associated Artin groups are exactly those obtained from rank-\(\le 2\) Artin groups by repeated free products and direct products with \(\mathbb Z\), and these are precisely the subgroup separable Artin groups [2010.13222].

The structural statement is:
\[
\boxed{\text{LERF Artin groups}=\text{closure of rank-}\le 2\text{ Artin groups under }*\text{ and }\times\mathbb Z.}
\]
Thus every subgroup separable Artin group is assembled from the smallest Artin groups by the two operations
\[
G,H\mapsto G*H,\qquad G\mapsto G\times\mathbb Z.
\]

The low-rank pieces are completely controlled. Rank \(1\) gives \(\mathbb Z\), hence LERF. Every rank-\(2\) Artin group \(A(m)\) is also LERF: the case \(m=2\) is abelian, and for \(m>2\) the group is an irreducible spherical rank-\(2\) Artin group with infinite cyclic center and a one-relator presentation, so Moldavanskii–Timofeeva’s theorem applies [2010.13222].

At rank \(3\), the classification already becomes rigid. A connected rank-\(3\) Artin group is LERF only when its graph is a **2-cone** over a rank-\(2\) graph, equivalently when
\[
A_\Gamma\cong A(m)\times \mathbb Z.
\]
Graph-theoretically, these are exactly the connected three-vertex cases in which one vertex is \(2\)-adjacent to the other two [2010.13222].

This low-rank picture is not merely illustrative. It says that, for subgroup separability, small Artin groups are not peripheral examples but the complete set of irreducible building blocks.

## 4. Small parabolic pieces and reduction principles

A different but equally important notion of smallness is encoded by **parabolic subgroups**. For a subset \(X\subseteq V(\Gamma)\), the subgroup \(A_X\) generated by \(X\) is a standard parabolic subgroup. If every two distinct vertices of \(X\) are joined by an edge, then \(X\) is **free of infinity**, equivalently \(\Gamma_X\) is complete, and \(A_X\) is a **complete standard parabolic subgroup** [2201.13044].

These complete parabolics are small in graph diameter rather than in rank. They are singled out by the reduction principle:
> if all complete standard parabolic subgroups of \(A_\Gamma\) have a property \(P\), then \(A_\Gamma\) should have property \(P\).

The paper on parabolic intersections makes this precise for several intersection conditions. Its main theorem shows that if intersections of parabolic subgroups are parabolic inside every complete Artin subgroup \(A_Y\), then in the ambient group the intersection of a complete parabolic subgroup with an arbitrary parabolic subgroup is again parabolic:
\[
(\mathrm{Int})\Rightarrow (\mathrm{Int+-}).
\]
This result is proved by repeated Bass–Serre decompositions
\[
A_\Gamma = A_I *_{A_K} A_J
\]
along missing edges, so that complete parabolic subgroups appear as the terminal pieces of the decomposition [2201.13044].

A related reduction philosophy appears in the survey of basic questions on Artin–Tits groups. There, torsion-freeness, triviality of center in the connected non-spherical case, and solvability of the word problem all reduce from arbitrary Artin groups to the **free-of-infinity** case. In particular:
- if all free-of-infinity Artin groups are torsion-free, then all Artin groups are torsion-free;
- if all non-spherical connected free-of-infinity Artin groups have trivial center, then all non-spherical connected Artin groups do;
- if all free-of-infinity Artin groups have solvable word problem, then all Artin groups do [1105.1048].

This is directly relevant to small type, because small-type Artin groups are free of infinity. In that sense, standard small Artin groups lie in the reduced core case for several foundational conjectures.

## 5. Small type versus the no-\(3\)-label phenomenon

A persistent misconception in recent literature is to identify standard small Artin groups with the class treated by Juh’s polynomial Dehn-function theorem. That identification is false. The theorem concerns Artin groups whose defining graph has **no edge labeled \(3\)**, not Artin groups whose labels are only \(2\) or \(3\) [2507.16770].

More precisely, if \(A(\Gamma)\) has labels \(n_{ij}\neq 3\) for every edge, then
\[
f(n)=n^6
\]
is an isoperimetric function for the standard Artin presentation. Equivalently, every null-homotopic word of length \(n\) bounds a van Kampen diagram of area at most \(n^6\), and therefore the word problem is solvable. The paper records the corollary:
\[
\textit{If }n_{ij}\neq 3\textit{ for every }i,j,\textit{ then }A(\Gamma)\textit{ has solvable word problem. In particular, even Artin groups have solvable word problem.}
\]

The class is
\[
\{0,2,4,5,6,\dots\},
\]
where \(0\) means “no edge.” This includes all even Artin groups and all right-angled Artin groups, but it excludes the genuinely small-type braid relations with label \(3\). In standard terminology, small Artin groups have labels only \(2\) or \(3\), so Juh’s class intersects them essentially only in the right-angled case [2507.16770].

The reason the label \(3\) case is excluded is structural rather than cosmetic. Juh’s proof works by splitting regions of a van Kampen diagram into commuting \(2\)-regions and noncommuting \(4^+\)-regions,
\[
Reg(M)=Reg_2(M)\sqcup Reg_{4^+}(M),
\]
and then applying relative \(C(4)\)-\(T(4)\) methods after passing to Howie diagrams and collapsed diagrams. The label \(3\) case produces length-\(6\) braid relators, which do not fit this dichotomy or the associated band arguments. The theorem therefore clarifies a genuine boundary in current diagrammatic methods: standard small-type groups are not covered precisely because the short \(3\)-braid relations remain the essential obstruction [2507.16770].

## 6. Congruence and representation-theoretic smallness

A substantially different notion of small Artin group appears in the study of congruence subgroups. Starting from Squier’s representation
\[
\sigma:A[\Gamma]\to GL(V)
\]
over \(\mathbb Z[s^{\pm},t^{\pm}]\), one specializes to
\[
s=1,\qquad t=-1.
\]
At this specialization, the representation is integral exactly when every exponent satisfies
\[
m_{ij}\in\{1,2,3,\infty\}.
\]
The paper therefore defines a **small Artin group** to be an Artin group admitting such a Coxeter system [2510.23002].

This definition includes:
- **braid groups**;
- **right-angled Artin groups**.

It is the Artin analogue of the corresponding notion for Coxeter groups, where the Tits representation is integral precisely for the same exponent set. Under this definition, standard small type and RAAGs both count as small, because labels \(2\), \(3\), and \(\infty\) are all permitted [2510.23002].

The associated principal congruence subgroups are defined by reducing the specialized representation modulo \(m\):
\[
\tilde{\sigma}_m:A[\Gamma]\to GL(n,\mathbb Z_m),\qquad
A[\Gamma][m]:=\ker(\tilde{\sigma}_m).
\]

The main global result is negative. If a small Coxeter graph has a connected component that is **not** affine, then the corresponding Artin group does **not** have the congruence subgroup property. Thus the only cases not ruled out are those whose connected components are all affine [2510.23002].

At low levels, the paper identifies several principal congruence subgroups explicitly. For spherical type,
\[
A[\Gamma][2]=\ker(A[\Gamma]\to W[\Gamma]/Z(W[\Gamma])),
\]
so
\[
A[\Gamma]/A[\Gamma][2]\cong W[\Gamma]/Z(W[\Gamma]).
\]
For right-angled Artin groups, level \(2\) is trivial in the sense that
\[
A[\Gamma][2]=A[\Gamma],
\]
while level \(4\) satisfies
\[
A[\Gamma][4]=\ker(A[\Gamma]\to W[\Gamma]/W[\Gamma]').
\]
This places braid groups, RAAGs, and other Squier-integral small Artin groups inside a unified congruence framework [2510.23002].

## 7. Model theory and geometric realizations of small examples

Small and low-rank Artin groups also play a prominent role in model theory. A recent paper proves that **two spherical Artin groups are elementarily equivalent if and only if they are isomorphic**, so within spherical type the first-order theory determines the isomorphism class exactly [2507.21575].

The same paper reduces the superstability problem for a non-abelian Artin group to a rank-\(2\) question: one needs a dihedral parabolic subgroup \(\langle a,b\rangle\cong A(I_2(m))\) to be \(n\)-pure in the ambient group for some sufficiently large prime \(n\). This reduction makes dihedral Artin groups the decisive small test cases for a global model-theoretic property [2507.21575].

The dihedral groups are analyzed explicitly. For \(A(I_2(m))\), the quotient by the center is
\[
A(I_2(m))/Z \cong
\begin{cases}
C_{m/2}*\mathbb Z, & m \text{ even},\\[2mm]
C_2*C_m, & m \text{ odd}.
\end{cases}
\]
These computations are used both in the superstability reduction and in distinguishing low-rank spherical types by first-order means [2507.21575].

On the affine side, the paper proves that the groups of type \(\tilde A_n\), \(n\ge 4\), can be separated from the other simply laced affine Artin groups by existential sentences. The argument uses strong co-Hopfianity and second homology computations such as
\[
H_2(A(\tilde A_n);\mathbb Z)\cong \mathbb Z\oplus \mathbb Z/2\mathbb Z \qquad (n\ge 4),
\]
contrasting with the finite \(2\)-group second homology of \(\tilde D_n\) and \(\tilde E_n\) [2507.21575].

Small Artin groups also admit concrete geometric realizations as fundamental groups of plane-curve complements. The deltoid gives the Artin group of the triangle,
\[
\langle x_1,x_2,x_3\mid x_1x_2x_1=x_2x_1x_2,\ x_2x_3x_2=x_3x_2x_3,\ x_3x_1x_3=x_1x_3x_1\rangle,
\]
and a parabola together with two parallel lines yields the Euclidean Artin group \((4,4,2)\),
\[
\langle x,y,z\mid xy=yx,\ (yz)^2=(zy)^2,\ (xz)^2=(zx)^2\rangle.
\]
More generally, for the hypocycloids \(\mathcal C_{k,k-1}\), the fundamental group of the complement is the Artin group of the polygon on
\[
N=2k-1
\]
vertices [1707.08353].

These realizations are significant because they turn small and low-rank Artin groups into explicit topological objects: cusp and node combinatorics become braid and commutation relations, and polygonal real diagrams become affine cycle-type Artin groups. In that sense, the study of small Artin groups sits at the intersection of Coxeter combinatorics, subgroup structure, geometric group theory, arithmetic representations, model theory, and the topology of plane curves.

Source: https://www.emergentmind.com/topics/small-artin-groups