Graphical Small Cancellation Groups

• Graphical small cancellation groups are defined by using graph-encoded relators, allowing refined control over pieces, overlaps, and curvature in presentations.
• They employ two standard models—the immersion model and S-labelled graphs—to flexibly realize CAT(0), hyperbolic, and acylindrical actions.
• The framework underpins constructions of groups with prescribed subgraphs, expanders, unique algebraic traits, and diverse finiteness properties.

Graphical small cancellation groups are groups defined by relators encoded in graphs rather than only in cyclic words. In the formulations used across the literature, one either starts from an immersion $f:\Gamma\to\Theta$ and imposes relations coming from immersed cycles in $\Gamma$, or from an $S$-labelled graph $\Gamma$ and defines $$G(\Gamma)=\langle S \mid \text{labels of simple closed paths in }\Gamma\rangle$$. This framework extends classical small cancellation theory, was pioneered by Gromov and Rips, and is used to construct groups whose Cayley graphs contain prescribed subgraphs and whose large-scale geometry exhibits phenomena inaccessible in the strictly word-based setting (Gui, 12 Jan 2026, Gruber, 2012).

1. Foundational formulations and basic objects

Graphical small cancellation is built from combinatorial data richer than a set of relator words. In the immersion model, a graphical presentation is written $G=\langle f:\Gamma\to\Theta\rangle$, where $f$ is a graph immersion. The associated thickened graphical complex $X_t$ is obtained by attaching 2-cells to $\Theta$ along the images of all immersed cycles in $\Gamma$, while the non-thickened graphical complex $\Gamma$0 is formed by attaching, for each component $\Gamma$1 of $\Gamma$2, a cone over $\Gamma$3 to $\Gamma$4, triangulated so that all 2-cells are triangles (Gui, 12 Jan 2026).

A second standard formulation uses an $\Gamma$5-labelled graph $\Gamma$6. In that setting,

$\Gamma$7

where $\Gamma$8 is the word read along $\Gamma$9. This labelled-graph formalism is central in the development of graphical $S$0, $S$1, $S$2, and $S$3 conditions, and in the construction of infinitely presented groups and quotients over free products (Gruber et al., 2014, Gruber, 2012).

The key local object is the piece. In the immersion model, a piece in $S$4 is an immersed path in $S$5 admitting two distinct lifts to $S$6. In the labelled-graph model, a piece is a labelled path that embeds into the defining graph in two distinct ways; for $S$7-conditions one uses essential pieces, where the two maps are required to be essentially distinct, i.e. not related by a label-preserving automorphism of the graph. This shift from repeated subwords to repeated immersed paths is the decisive generalization from classical to graphical small cancellation (Gui, 12 Jan 2026, Gruber, 2012).

A common misconception is that graphical small cancellation is merely a notational reformulation of classical presentations. The cited work instead treats relators as immersed cycles, relator graphs, or cone-cells over finite graphs, and this increased flexibility is precisely what allows one to plant expanders, prescribe peripheral subgraphs, or organize infinitely many relations while retaining strong diagrammatic control (Esperet et al., 2023).

2. Small cancellation conditions and combinatorial control

The non-metric graphical conditions parallel the classical $S$8- and $S$9-conditions. In the immersion formalism, graphical $\Gamma$0 means that no immersed loop in $\Gamma$1 mapping to $\Gamma$2 is a concatenation of fewer than $\Gamma$3 pieces, while graphical $\Gamma$4 requires that every reduced disk diagram in $\Gamma$5 has all interior vertices either of degree $\Gamma$6 or at least $\Gamma$7. In labelled graphs, $\Gamma$8 and $\Gamma$9 assert that no nontrivial closed path is a concatenation of fewer than $$G(\Gamma)=\langle S \mid \text{labels of simple closed paths in }\Gamma\rangle$$0 pieces or essential pieces, respectively (Gui, 12 Jan 2026, Gruber, 2012).

Metric versions are expressed by length inequalities. Graphical $$G(\Gamma)=\langle S \mid \text{labels of simple closed paths in }\Gamma\rangle$$1 and $$G(\Gamma)=\langle S \mid \text{labels of simple closed paths in }\Gamma\rangle$$2 require that whenever a piece occurs in a simple closed path $$G(\Gamma)=\langle S \mid \text{labels of simple closed paths in }\Gamma\rangle$$3, its length is strictly less than $$G(\Gamma)=\langle S \mid \text{labels of simple closed paths in }\Gamma\rangle$$4. In the free-product setting, the relevant length is free-product syllable length rather than ordinary word length, and one obtains graphical $$G(\Gamma)=\langle S \mid \text{labels of simple closed paths in }\Gamma\rangle$$5 and $$G(\Gamma)=\langle S \mid \text{labels of simple closed paths in }\Gamma\rangle$$6 conditions adapted to quotients of $$G(\Gamma)=\langle S \mid \text{labels of simple closed paths in }\Gamma\rangle$$7 (Gruber et al., 2014, Steenbock, 2013).

Several structural consequences depend on how these conditions constrain overlaps. For example, if $$G(\Gamma)=\langle S \mid \text{labels of simple closed paths in }\Gamma\rangle$$8 holds with $$G(\Gamma)=\langle S \mid \text{labels of simple closed paths in }\Gamma\rangle$$9, then all pieces have length $G=\langle f:\Gamma\to\Theta\rangle$0, which sharply simplifies the local geometry of the non-thickened complex. In the $G=\langle f:\Gamma\to\Theta\rangle$1-$G=\langle f:\Gamma\to\Theta\rangle$2 regime, the links of vertices become the primary combinatorial carrier of curvature information: for vertices in $G=\langle f:\Gamma\to\Theta\rangle$3, the link is bipartite, and short cycles in links are ruled out by the small cancellation hypotheses (Gui, 12 Jan 2026).

The theory also extends beyond ordinary presentations to complexes of groups. A graphical complex of groups is built over a finite 1-dimensional poset $G=\langle f:\Gamma\to\Theta\rangle$4, and when $G=\langle f:\Gamma\to\Theta\rangle$5 is $G=\langle f:\Gamma\to\Theta\rangle$6-huge—equivalently $G=\langle f:\Gamma\to\Theta\rangle$7—the associated Basic Construction satisfies the graphical $G=\langle f:\Gamma\to\Theta\rangle$8-condition; for $G=\langle f:\Gamma\to\Theta\rangle$9, this yields a $f$0 graphical small cancellation structure with pieces of length at most $f$1 (Prytuła, 2020).

The scope of graphical small cancellation should not be overstated. A related metric condition, $f$2, was introduced as a framework that contains the classical metric conditions $f$3, $f$4-$f$5, and $f$6-$f$7, and is explicitly described as more general than standard graphical small cancellation because it controls total normalized overlap at vertices rather than only pairwise overlap data (Blufstein et al., 2020). This suggests that graphical small cancellation is best viewed as a powerful, but not exhaustive, corner of a broader diagrammatic-curvature theory.

3. Curvature, isoperimetry, and nonpositive geometry

One of the principal themes in the subject is the conversion of diagrammatic overlap bounds into metric curvature. For groups defined by graphical $f$8-presentations, the associated presentation satisfies the linear isoperimetric inequality

$f$9

while graphical $X_t$0-presentations satisfy the quadratic inequality

$X_t$1

For graphical $X_t$2-labelled graphs, the corresponding presentation complex is aspherical, and the resulting group is torsion-free with cohomological dimension at most $X_t$3 (Gruber, 2012).

A major recent advance identifies explicit nonpositively curved metrics in the graphical borderline cases. If $X_t$4 satisfies graphical $X_t$5 with $X_t$6, then the associated non-thickened graphical complex $X_t$7 admits a locally CAT metric in three cases: locally CAT(0) for $X_t$8, and locally CAT$X_t$9 for $\Theta$0 or $\Theta$1. In the CAT(0) case each triangle is given a Euclidean metric with angles $\Theta$2; in the negative-curvature cases one uses hyperbolic triangles with prescribed angles. A further corollary replaces the girth assumption by full graphical $\Theta$3-$\Theta$4, yielding a locally CAT(0) metric after subdivision (Gui, 12 Jan 2026).

The mechanism is local. The link condition requires every injective closed path in a vertex link to have total angle at least $\Theta$5. Graphical $\Theta$6-$\Theta$7 conditions rule out short immersed cycles in links, so the link condition can be checked combinatorially and then promoted to a piecewise Euclidean or piecewise hyperbolic metric statement (Gui, 12 Jan 2026).

Graphical complexes of groups provide a parallel source of nonpositive curvature. For $\Theta$8-huge $\Theta$9 with $\Gamma$0, the Basic Construction $\Gamma$1 admits a piecewise Euclidean metric making it locally CAT(0), hence globally CAT(0); the same complexes carry a $\Gamma$2 graphical small cancellation structure and an associated Wise complex that is systolic, implying biautomaticity. If $\Gamma$3, the fundamental group is automatically word-hyperbolic, while for $\Gamma$4 hyperbolicity depends on the absence of proper triples, and non-hyperbolic examples with flats also occur (Prytuła, 2020).

This corrects another frequent oversimplification: graphical small cancellation does not automatically imply hyperbolicity. Some families are CAT(0) but not hyperbolic, some are relatively hyperbolic rather than hyperbolic, and some are engineered specifically to contain large embedded subgraphs that obstruct coarse negative curvature in the ordinary sense (Prytuła, 2020, Han, 2020).

4. Hyperbolic actions, acylindricity, and boundary geometry

For infinitely presented groups, hyperbolicity is often recovered not from the ordinary Cayley graph but from coned-off or partially coned-off models. Infinitely presented graphical $\Gamma$5 small cancellation groups are either virtually cyclic or acylindrically hyperbolic. In particular, infinitely presented classical $\Gamma$6-groups, and hence classical $\Gamma$7-groups, are acylindrically hyperbolic. The proof constructs a hyperbolic coned-off Cayley graph and then a WPD element, placing these groups inside the general theory of acylindrical actions on hyperbolic spaces (Gruber et al., 2014).

The space of such actions can be highly non-rigid. For small cancellation groups satisfying $\Gamma$8-type hypotheses, Abbott and Hume define the poset $\Gamma$9 of thin cones, whose smallest element is the Gruber–Sisto coned-off graph. In almost all cases this poset is extremely rich: it has a largest element if and only if it has exactly one element, and between comparable distinct actions one can embed order-theoretic copies of $\Gamma$00 or $\Gamma$01. Thus even when a universal acylindrical action exists, there need not be a largest hyperbolic structure (Abbott et al., 2018).

Relative hyperbolicity in graphical $\Gamma$02 groups is governed by a precise overlap criterion. The Cayley graph is asymptotically tree-graded with respect to the collection of embedded components of the defining graph if and only if all pieces have uniformly bounded length; consequently the group is hyperbolic relative to the stabilizers associated to finitely many components up to label-preserving automorphism exactly in that uniformly bounded-piece regime (Han, 2020).

Boundary actions furnish a more recent dynamical refinement. For graphical $\Gamma$03 presentations with finite generating set, countable disjoint union of finite defining graphs, and the additional extreme fineness hypothesis that every edge lies in at most $\Gamma$04 simple closed paths, the action on the Gromov boundary of the coned-off Cayley graph has hyperfinite orbit equivalence relation. The class includes infinitely presented classical small cancellation groups, because there each edge appears in a unique relator cycle (Karpinski et al., 6 Sep 2025).

Growth theory also interacts with acylindricity in this setting. Uniform uniform exponential growth is preserved under suitable geometric small cancellation quotients of acylindrically hyperbolic groups, and the paper explicitly states that the conclusions apply directly to graphical small cancellation groups whenever the relevant hyperbolicity and acylindricity hypotheses are satisfied. This places graphical small cancellation inside a stability theory for quantitative growth under quotient operations (Legaspi et al., 2024).

5. Coarse geometry, analytic properties, and prescribed subgraphs

Graphical small cancellation became especially prominent through constructions of groups with prescribed coarse features. Using $\Gamma$05-small cancellation labellings of infinite sequences of finite bounded-degree graphs, one obtains finitely generated groups whose Cayley graphs contain each graph in the sequence isometrically. Applied to expander families, this yields finitely generated groups with expanders embedded isometrically into their Cayley graphs, strengthening earlier “monster” constructions in which expanders were not even coarsely embedded (Osajda, 2014).

The same labelling technology can be combined with compatible wall structures. For finitely generated groups defined by infinite graphical $\Gamma$06-presentations satisfying a lacunary walling condition, the associated $\Gamma$07-complexes enjoy the linear separation property, the group acts properly on a space with walls, and therefore has the Haagerup property. As recorded in the paper, this implies coarse embeddability into Hilbert space and, by Higson–Kasparov, the strong Baum–Connes conjecture and hence the Baum–Connes conjecture with arbitrary coefficients (Arzhantseva et al., 2014).

This produces a notable contrast within the subject. On one hand, graphical small cancellation can generate groups containing isometrically embedded expanders and therefore displaying strong coarse obstructions; on the other, with lacunary walling it yields groups with the Haagerup property and coarse embeddability into Hilbert space. The framework is therefore not tied to a single coarse-geometric behavior but functions as a construction technology whose outputs depend on the combinatorics of the defining graphs (Osajda, 2014, Arzhantseva et al., 2014).

The combinatorics of labels itself has become an optimization problem. A later paper replaces a Lovász Local Lemma approach by a counting argument and proves that for graph sequences of maximum degree $\Gamma$08, girth tending to infinity, and diameter bounded by $\Gamma$09, an even number

$\Gamma$10

is sufficient for a $\Gamma$11-labelling. For cubic Ramanujan graphs with refined counting, the number of required generators can be reduced to $\Gamma$12 (Esperet et al., 2023).

Graphical $\Gamma$13 groups also support a refined internal theory of coarse negative curvature. Contracting and Morse geodesics can be characterized through their intersections with embedded components of the defining graph. In the classical $\Gamma$14 case, bounded intersection lengths with relator cycles characterize strong contraction, while sublinear intersection growth characterizes Morse behavior. The same paper shows that every sublinear contraction rate occurs in a finitely generated group, that strong contraction may depend on the chosen finite generating set, and that when the defining graph has finite components, translation lengths are rational and bounded away from zero (Arzhantseva et al., 2016).

6. Algebraic consequences, subgroup phenomena, and frontier constructions

Graphical small cancellation has repeatedly been used to produce groups with striking algebraic properties. Infinitely presented classical $\Gamma$15-groups are SQ-universal, and this extends to graphical $\Gamma$16-groups over free products of infinite groups under stated finiteness hypotheses on the components. The same paper proves that for every $\Gamma$17 there are uncountably many pairwise non-quasi-isometric groups admitting classical $\Gamma$18-presentations but no graphical $\Gamma$19-presentation with finite components and finite label set, using distorted cyclic subgroups as the obstruction (Gruber, 2014).

Free-product graphical small cancellation was developed in part to reinterpret the Rips–Segev construction. In that setting one obtains torsion-free groups without the unique product property, including the first hyperbolic examples of torsion-free groups without unique product. The theory also yields uncountably many non-isomorphic torsion-free groups without unique product, and a graphical version of Comerford’s construction shows that for every subgroup $\Gamma$20 of a graphical small cancellation group there exists a free group $\Gamma$21 such that $\Gamma$22 again admits a graphical small cancellation presentation. This was used to build torsion-free hyperbolic groups all of whose subgroups up to any prescribed finite index still fail unique product (Steenbock, 2013, Gruber et al., 2014).

The theory extends into periodic group theory. For sufficiently large odd exponent $\Gamma$23, classical $\Gamma$24-small cancellation, its graphical generalization, and the free-product version all produce infinite $\Gamma$25-periodic groups. The graphical Burnside quotient $\Gamma$26 can be arranged so that components of $\Gamma$27 embed in its Cayley graph, and the framework yields periodic analogues of expander-containing groups, Rips constructions, and SQ-universality inside the Burnside variety (Coulon et al., 2017).

Graphical small cancellation has also entered finiteness-property theory. Brown and Leary construct an uncountable family of groups of type $\Gamma$28 using graphical $\Gamma$29-presentations on specially designed acyclic 2-complexes, explicitly avoiding the Morse-theoretic methods on cubical complexes that had dominated earlier constructions. In their framework, the groups $\Gamma$30 are of type $\Gamma$31, are finitely presented if and only if $\Gamma$32 is finite, and occur in continuously many isomorphism types (Brown et al., 2020).

These developments indicate both the reach and the limits of the subject. Graphical small cancellation can produce CAT(0) groups, acylindrically hyperbolic groups, relatively hyperbolic groups, Haagerup groups, periodic groups, SQ-universal groups, and groups of type $\Gamma$33; it can also encode failure of unique product, prescribed divergence, and isometric embeddings of expanders. A plausible implication is that graphical small cancellation should be understood less as a single geometric class than as a controlled presentation machine whose small-overlap axioms permit many distinct global behaviors, provided one keeps track of the auxiliary structures—walls, conings, bounded pieces, girth, and component geometry—that determine which large-scale properties survive in the quotient (Gruber, 2012, Arzhantseva et al., 2016).