Artin groups: complete topological models, controlled subgroups, and a curvature obstruction

← All digests

Family 254 · Group theory
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-22

1. The question

Ordinary braids give a vivid example of a group. Place strands between two rows of points, multiply braids by stacking them, and identify motions that can be continuously deformed into one another. Artin groups generalize the familiar relations between adjacent crossings without requiring an underlying picture of strands.

Their definitions are short, but three basic questions have resisted uniform answers. Does the standard space built from an Artin presentation have a contractible universal cover? Do the natural subgroups remain natural after taking intersections? Can every Artin group act geometrically on a space of nonpositive curvature?

This family claims affirmative answers to the first two questions for every Artin group with finitely many standard generators, and a negative answer to the third. Together they draw a useful boundary: having a complete topological model and well-behaved subgroups does not force the strongest expected metric geometry.

2. Background and essential definitions

The defining relations

Choose a finite set $S$ and symmetric labels $m_{st}$, with $m_{ss}=1$ and, for distinct indices, $m_{st}\in\{2,3,4,\ldots,\infty\}$. The Artin group has a generator $\sigma_s$ for each $s\in S$. When $m_{st}$ is finite, the two alternating words of that length are equal. For example,

$$\begin{aligned} m_{st}=2&:\quad \sigma_s\sigma_t=\sigma_t\sigma_s,\\ m_{st}=3&:\quad \sigma_s\sigma_t\sigma_s =\sigma_t\sigma_s\sigma_t. \end{aligned}$$

An infinite label imposes no relation on that pair. Finite rank means that $S$ is finite; it does not mean the group itself is finite.

Adding the relations $\sigma_s^2=1$ gives the associated Coxeter group $W$. A subset $T\subseteq S$ is spherical when its Coxeter subgroup $W_T$ is finite. The Artin subgroup $A_T$ generated by those same indices is generally infinite even when $T$ is spherical.

A parabolic subgroup is a conjugate $gA_Tg^{-1}$. Thinking of $A_T$ as using selected generators is helpful, but conjugation makes intersections far less transparent than intersections of subsets of $S$.

Topology versus curvature

The Salvetti complex is a space made by attaching cells. It has one cell of dimension $|T|$ for every spherical subset $T$, including a vertex for the empty subset. Its loops represent generators and its two-dimensional cells record braid relations. Higher-dimensional cells are essential parts of the construction.

A connected space is a $K(G,1)$ if its fundamental group is $G$ and its universal cover is contractible. Contractibility means that the identity map can be continuously deformed to a constant map. Equivalently, for this cell complex there are no higher homotopy groups: every sphere of dimension at least two maps into the space in a null-homotopic way.

A CAT(0) space instead satisfies a metric condition: distances inside each geodesic triangle are no greater than the corresponding distances in a Euclidean comparison triangle. Such spaces are contractible, but contractibility alone does not provide a CAT(0) metric.

Here a geometric action is a proper, cocompact isometric action on a nonempty proper CAT(0) space. Properness of the space means that closed bounded subsets are compact. Properness of the action limits how many group elements can move a compact set back into itself; cocompactness means that translates of one compact set cover the space.

3. Development history and previous partial results

The classifying-space question grew from configuration spaces and hyperplane complements. Classical braid groups supplied early positive examples. Deligne proved the spherical Artin case in 1972, using the structure of finite reflection arrangements. Salvetti’s constructions turned the relevant spaces into explicit cellular models.

Charney and Davis developed a complex built from spherical parabolic cosets and connected its contractibility to the general conjecture. They proved important classes, including FC type and the case where the largest spherical subset has size at most two. This dimension need not equal the number of generators.

Paolini and Salvetti resolved all affine types. More recently, Huang and Przytycki treated the dimension-at-most-three case and further classes, while Hoda and Huang treated groups whose irreducible spherical parabolics have specified types. These are broad positive results, but the first manuscript claims to remove the restrictions on the finite-rank Coxeter matrix altogether.

The intersection problem developed alongside this topology. Van der Lek’s embedding theorem ensures that standard parabolic subgroups have the expected Artin presentation. That does not say that intersections of arbitrary conjugates are parabolic. Earlier intersection theorems covered right-angled groups, spherical type, large type, selected affine types, and other substantial classes. In FC type, some theorems required one or both intersected parabolics to be spherical; those hypotheses should not be confused with a theorem about arbitrary parabolics.

The CAT(0) question has its own positive history. Right-angled Artin groups act on natural CAT(0) cube complexes. Results also cover several large-label classes, including Haettel’s XXL-type groups, and classical braid groups with up to seven strands through work by several authors. Brady and Crisp found examples whose smallest CAT(0) dimension exceeds the dimension of a classifying space. Their obstruction was dimensional; the new counterexample claims to exclude geometric CAT(0) actions in every dimension.

4. What the new manuscripts claim

All three manuscripts are dated September 23, 2026.

M1 claims the Artin $K(\pi,1)$ conjecture in finite rank. For every finite Coxeter matrix, the full standard Salvetti complex has contractible universal cover. Finite and infinite labels, disconnected diagrams, and nonspherical Coxeter groups are all allowed.

The manuscript also records consequences using established theorems: torsion-freeness; a description of the center as $\mathbb Z^k$, where $k$ counts the spherical irreducible diagram components; and the classifying-space comparison with the positive Artin monoid. These are consequences of the asphericity input combined with earlier work, rather than independent new arguments for every conclusion.

M3 claims the parabolic intersection conjecture in the same finite-rank scope. In particular,

$$gA_Xg^{-1}\cap hA_Yh^{-1}=kA_Zk^{-1}$$

for suitable $Z\subseteq S$ and $k\in A$. Neither intersected subgroup needs to be spherical. The paper further claims that every intersection of parabolics is already the intersection of at most $|S|$ members of the family. The empty intersection is understood to be the whole group.

Consequently, every subset of $A$ has a unique smallest containing parabolic subgroup. M3 also claims a uniform algorithm deciding membership in a supplied conjugate of a standard parabolic, with the Coxeter matrix and words as input. Taking the trivial parabolic gives a solution of the word problem. No running-time bound is asserted.

M2 claims an explicit counterexample to the CAT(0) conjecture. Its Artin group has 116 generators and only the off-diagonal labels $2,3,\infty$. It admits no geometric action on any nonempty proper CAT(0) space, without a dimension bound. The other manuscripts’ asphericity statements are not inputs to this obstruction.

5. Why the advance matters

Importance: 5/5, provisional. The family addresses several central structural questions about the entire finite-rank class, rather than extending a theorem to one additional diagram type.

If M1 is correct, a concrete finite cell complex becomes a classifying space for every such Artin group. That turns the defining diagram into a uniform topological tool for computing and studying group invariants.

If M3 is correct, parabolic subgroups acquire a robust closure operation. Intersections and smallest containing parabolics become available without repeatedly checking special-type hypotheses. The claimed membership algorithm adds an effective counterpart to the structural theorem.

M2 would establish that this strong topology cannot always be upgraded to geometric CAT(0) structure. Its force comes from excluding all dimensions and metrics allowed by the definition. It therefore rules out repairing the example simply by choosing a larger complex or a more ingenious nonpositively curved metric.

6. The main idea

Measuring spherical cosets by harmonic heights

M1 replaces the universal-cover problem by a classical equivalent problem: contract the space assembled from spherical cosets $A_Tx$, ordered by inclusion. The challenge is that the whole group need not be spherical even though the pieces indexing this space are.

The paper builds an algebraic action of the Artin group on complexes of modules. A complex is a sequence of vector-space-like objects with differentials whose successive compositions vanish. Counting generators in its graded layers produces nonnegative numerical data, one coordinate for each original generator and an extra framing coordinate.

Inside a spherical subset, the exterior coordinates are extended harmonically. The extension solves equations of the form

$$2p_i=\sum_{j\ne i} b_{ij}p_j,$$

where the coefficients encode Coxeter labels, with $b_{ij}=2\cos(\pi/m_{ij})$ for finite labels and $2$ for infinite labels. Sphericity gives the positive-definiteness needed for a unique extension in the relevant subsystem.

These vectors define a height on cosets. The paper orders cosets by height and studies how each new vertex attaches to the earlier part. Two algebraic estimates control the process: a comparison preventing competing minimal candidates, and an obstruction to an isolated positive harmonic layer. Special equality cases require a modified tie rule, not just strict descent everywhere. The resulting earlier links are empty or contractible, and connectedness then yields the claimed contraction.

Turning intersections into stabilizers

M3 uses a related categorical framework with a relative frame adapted to a chosen standard parabolic $A_X$. The framing object is arranged so that its stabilizer is exactly $A_X$. Transporting that object by a group element converts an intersection of conjugate parabolics into a stabilizer problem inside another parabolic.

Harmonic heights now organize the set of minimal configurations. The paper claims that these minima are connected, including paths with tied heights. This forces a stabilizer into a smaller parabolic unless a whole factor acts trivially. A separate positive-height argument handles spherical cases where the first reduction would not lower rank. Induction then identifies the stabilizer as parabolic.

For the algorithm, the finite algebra has computable algebraic coefficients. Tensoring the complexes corresponding to an input word and cancelling invertible differential entries reduces membership to a finite exact calculation. The claim is termination and correctness as a decision procedure, not practical efficiency. M3 develops its needed stabilizer argument without assuming M1’s asphericity theorem.

A short angular circuit that cannot exist

M2 begins with the three-generator group

$$T=\langle a,b,c\mid aba=bab,\ bcb=cbc,\ cac=aca\rangle.$$

It attaches a 40-strand braid group along each of the three braid pairs. Each attachment contributes 37 new generators. Two further generators centralize $a,b,c$, giving the total $3+3\cdot37+2=116$.

Suppose a geometric CAT(0) action existed. The stable translation length

$$\ell(g)=\lim_{n\to\infty}\frac{d(x,g^nx)}{n}$$

measures the distance moved per iterate at large scale. For commuting elements, these lengths behave like lengths of Euclidean translation vectors. The long braid blocks impose quantitative constraints on the lengths associated with each of the three braid pairs.

Those constraints yield six asymptotically Euclidean sectors, each with angle less than $\pi/3$, forming a circuit with total angle less than $2\pi$. Converting that circuit into a contradiction in arbitrary dimension is the key step. The paper compares a sector disk with a CAT(0) geodesic disk, transfers the comparison to finite triangular diagrams, and uses a winding count: one part forces nonzero signed multiplicity, while a folding argument forces it to vanish.

One further issue matters. A subgroup of a group acting geometrically need not itself act cocompactly. The two extra generators make $T$ their simultaneous centralizer. The paper uses this to obtain a suitable closed convex region on which $T$ does act geometrically, so the length constraints and the angular obstruction apply to the same action.

7. Scope, limitations, and open questions

The universal statements in M1 and M3 concern finite rank. Their wording does not by itself supply a theorem for arbitrary infinite generating sets. M1 concerns the full Salvetti complex, not merely its two-dimensional presentation complex.

The counterexample excludes geometric CAT(0) actions. It does not exclude every action on every CAT(0) space, such as actions lacking properness or cocompactness. Its 116-generator construction is explicit, but the headline does not establish that this size is minimal.

The membership result supplies no complexity bound. Likewise, the intersection theorem does not mean that simply intersecting the visible generator sets finds the answer for conjugated parabolics.

Finally, M3 explicitly credits earlier work for acylindrical hyperbolicity in its stated non-clique scope. Its deduction through the new intersection theorem is another route to that conclusion, not a new class of groups with that property.

8. Sources and further reading

M1 and M3 give detailed historical references for spherical, affine, FC, and large-type results. For the categorical background, see Heng–Licata’s Artin–Tits actions and stability conditions. The earlier three-dimensional asphericity result and ABI-type theorem show how broad partial results preceded the general claim.

For the metric side, compare Haettel’s XXL-type theorem, Jeong’s seven-strand braid theorem, and Brady–Crisp’s dimension obstruction. Bridson–Haefliger’s book on nonpositive curvature develops the CAT(0) background.

Source manuscripts

• M1. OpenAI, Harmonic heights and the Artin K(pi,1) conjecture (September 23, 2026).
• M2. OpenAI, An Artin group with no geometric CAT(0) action (September 23, 2026).
• M3. OpenAI, Parabolic intersections in Artin groups (September 23, 2026).

Sources consulted on 22 October 2026 at repository revision 1581d4c. Download the manuscript citations.