One relation, negative curvature, and finite special covers

← All digests

Family 258 · Group theory
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-22

1. The question

How complicated can a group become when just one equation is imposed on freely generated elements? One-relator groups include free groups, surface groups, and groups with severe geometric distortion. A short presentation is no guarantee of simple geometry.

Gersten’s conjecture proposes a precise dividing line. Among finitely generated one-relator groups, are the familiar Baumslag–Solitar groups the only subgroup obstructions to hyperbolicity? A second question asks whether every hyperbolic one-relator group has, after passing to finite index, the particularly useful geometry of a compact special cube complex.

The two manuscripts claim yes to both. They also combine the second theorem with earlier work of Kielak and Linton to obtain virtual free-by-cyclic structure. The conclusion is unusually broad, but its wording requires care: the free kernel for a general one-relator group may have infinite rank.

2. Background and essential definitions

One relation and the obstruction groups

A one-relator group has a presentation

$$G=\langle X\mid r\rangle,$$

where $X$ is a finite set of generators and $r$ is one word in those generators and their inverses. The relation means that $r$, every conjugate of $r$, and all their consequences become trivial.

For nonzero integers $m,n$, the Baumslag–Solitar group is

$$\operatorname{BS}(m,n) =\langle a,t\mid ta^mt^{-1}=a^n\rangle.$$

The relation identifies two possibly different powers of $a$ by conjugation. For example, $\operatorname{BS}(1,2)$ contains the identity $t^ka t^{-k}=a^{2^k}$, exhibiting exponential distortion of powers. The case $\operatorname{BS}(1,1)$ is $\mathbb Z^2$, whose flat geometry is another obstruction.

A word-hyperbolic group has uniformly thin geodesic triangles in a Cayley graph. Such groups contain no Baumslag–Solitar subgroup with both parameters nonzero. The first manuscript claims the converse within the one-relator class.

Finite special cubical geometry

A cube complex is built by gluing Euclidean cubes along faces. Nonpositive curvature is a local condition ensuring that the universal cover is CAT(0), with geodesic triangles no thicker than Euclidean comparison triangles.

A special cube complex has additional control over its hyperplanes, the surfaces formed by joining cube midsections. An equivalent description useful here is that it admits a cubical local isometry to the Salvetti complex of a right-angled Artin group. A right-angled Artin group has one generator per vertex of a graph, with commuting relations for its edges. The local isometry preserves the relevant local cubical geometry and induces an embedding of fundamental groups.

A group is virtually compact special if a finite-index subgroup is the fundamental group of a compact special cube complex. Both qualifiers matter: the special model may require a finite cover, and compactness is stronger than merely embedding in a right-angled Artin group.

A subgroup is quasiconvex if geodesics between its elements stay within a fixed distance of it. This controls how subgroup geometry sits in the ambient hyperbolic group and is the condition needed in many cubical combination theorems.

Free-by-cyclic, with a rank convention

The manuscripts use a subgroup-closed meaning of free-by-cyclic:

$$1\longrightarrow K\longrightarrow H\longrightarrow Q\longrightarrow1, \qquad K\text{ free},\quad Q\leq\mathbb Z.$$

The free group $K$ may have infinitely many generators, and $Q$ may be trivial. “Virtually” means that such an $H$ has finite index in the original group. This is weaker than saying that a finite-index subgroup is a mapping torus of an automorphism of a finite-rank free group.

3. Development history and previous partial results

Magnus’s Freiheitssatz established that omitting an appropriate generator appearing in a one-relator presentation leaves a free subgroup. His subsequent solution of the word problem made induction through simpler presentations a central method in the subject.

Newman’s spelling theorem handles proper-power relators, the source of torsion in the classical one-relator setting, and yields hyperbolicity there. Thus the main new difficulty in the first manuscript lies in the non-power case, even though its final theorem includes all relators.

Gersten formulated the subgroup-obstruction conjecture in the early 1990s. Recent progress made the remaining gap more precise. Louder and Wilton developed negative-immersion methods, and Linton established hyperbolicity and virtual specialness in broad classes using one-relator hierarchies. Further criteria involved additional stability or primitivity-rank conditions. These did not automatically settle hyperbolicity for every Baumslag–Solitar-free one-relator group.

Combination theorems of Bestvina and Feighn explain when negatively curved pieces remain negatively curved after gluing. Brinkmann’s expansion theorem treats mapping tori of free-group automorphisms, and Mutanguha extended the Baumslag–Solitar criterion to ascending HNN extensions of finite-rank free groups under injective endomorphisms. The first manuscript extends the needed combination analysis to the more general splittings occurring in its one-relator induction.

The special-cube-complex story developed through Haglund–Wise, Wise’s quasiconvex hierarchy theorem, and Agol’s virtual special theorem. Wise had already proved virtual compact specialness for one-relator groups with torsion, resolving and strengthening an earlier residual-finiteness conjecture of Baumslag in that class.

For torsion-free groups, Linton’s work reduced the remaining problem to particular primitive extension groups and specified how a terminal result would propagate up a hierarchy. Kielak and Linton showed that virtual compact specialness would also give Wise’s proposed virtual free-by-cyclic conclusion for hyperbolic one-relator groups. The new second manuscript claims to provide that missing geometric hypothesis.

4. What the new manuscripts claim

Both manuscripts are dated September 25, 2026.

M1 claims Gersten’s conjecture: every finitely generated one-relator group containing no subgroup isomorphic to $\operatorname{BS}(m,n)$, for any nonzero integers $m,n$, is word-hyperbolic. Negative parameters are included. In particular, excluding only visibly distorted examples while allowing $\mathbb Z^2$ would not satisfy the hypothesis.

The theorem includes trivial relators and proper-power relators and has no torsion-free assumption. Under its non-power graph-presentation hypotheses, M1 also records injectivity and quasiconvexity of Magnus subgraphs, crediting the established subgroup theorems used for that conclusion.

M2 claims that every word-hyperbolic one-relator group is virtually compact special. Precisely, some finite-index subgroup is the fundamental group of a finite nonpositively curved cube complex admitting a cubical local isometry into a finite right-angled Artin Salvetti complex.

M2 also claims the same conclusion for every hyperbolic ascending mapping torus

$$F*_{\phi}=\langle F,t\mid t^{-1}ft=\phi(f),\ f\in F\rangle,$$

where $F$ is free of finite rank and $\phi:F\to F$ is injective, possibly nonsurjective.

Combining the one-relator theorem with Kielak–Linton gives a finite-index embedding

$$H\hookrightarrow F_d\rtimes_{\alpha}\mathbb Z,$$

where $d$ is finite and $\alpha$ is an automorphism. Restricting projection to $\mathbb Z$ gives the free-by-cyclic sequence above. Its kernel $H\cap F_d$ may have infinite rank: subgroups of finitely generated free groups need not be finitely generated.

For the ascending mapping tori, M2 gives a stronger separate conclusion: a finite-index subgroup maps onto $\mathbb Z$ with a finite-rank free kernel. The new homomorphism is not identified with the original ascending height.

5. Why the advance matters

Importance: 5/5, provisional. If correct, the family supplies both a central hyperbolicity criterion and a strong geometric structure theorem throughout the hyperbolic one-relator class.

The first result identifies exactly what prevents hyperbolicity in this class through a family of embedded subgroup obstructions. The second connects hyperbolicity to finite covers, cube complexes and right-angled Artin groups.

Using established implications, M2 obtains integral linearity and residual finiteness for the whole group, not merely the chosen finite-index subgroup. Residual finiteness means that every nonidentity element survives in some finite quotient. It also records conjugacy-separation and quasiconvex-subgroup consequences, with their appropriate hypotheses.

This restricted positive theorem contrasts with the claimed general hyperbolic counterexamples in families 252 and 257. The one-relator assumption is doing substantial work; those broader phenomena cannot simply be imported into this narrower class.

6. The main idea

An induction that needs genuine expansion

The Magnus method rewrites a one-relator group as simpler pieces joined through free subgroups. An HNN extension, for example, adds a stable letter that conjugates one specified subgroup to another. Such a decomposition acts on a tree whose vertices represent the pieces and whose edges represent the gluing subgroups.

M1 uses a decreasing complexity measure to reach simpler one-relator vertex groups. Once a vertex group is hyperbolic, existing results provide quasiconvexity of the relevant free edge groups. The remaining task is to prove annular flaring: long corridors passing through successive pieces must expand rather than maintaining uniformly narrow sides indefinitely.

Failure of flaring is encoded by a free-group current, an invariant measure on pairs of distinct points at infinity, or equivalently on unoriented geodesic lines. Long nonexpanding corridors produce a stationary current whose typical lines can be transferred through the tree.

Intersection and rank bounds organize persistent path stabilizers into finitely many subgroup carriers. Returns to a carrier induce automorphisms with no periodic nontrivial conjugacy classes; Brinkmann’s theorem then forces expansion. This makes the carrier-line sets have zero current mass.

Zero mass is not the same as absence from the support of a measure. M1 addresses this distinction separately. It transports positive-mass neighborhoods along carrier returns. A failed transfer produces a long periodic overlap; a successful one would concentrate mass on a set already shown to have zero mass. Further control of whole periods excludes the periodic alternative.

The remaining support has only finitely many relevant branching-ray types. A return then creates infinitely many disjoint positive-mass translates inside a compact region, contradicting local finiteness. This supplies flaring and closes the hyperbolicity induction.

Building special quotients and recovering the original group

M2 first reduces the problem, using Linton’s hierarchy, to primitive extension groups. It uses M1 to establish hyperbolicity of certain embedded terminal groups and the geometry of their Magnus subgroups. This is a specific dependency: it does not assume that every finitely generated subgroup of a hyperbolic group is hyperbolic, which would be false.

The construction proceeds through suitable fillings, meaning quotients obtained by imposing extra relations in designated subgroups. The aim is to make these quotients virtually compact special while retaining enough information to recover specialness of the unfilled group.

Persistent intersections along the hierarchy are controlled by joint injections into finite-dimensional homology. Keeping the joint map, rather than only separate rank inequalities, distinguishes contributions from different paths. Periodic end groups are organized into peripheral subgroups, and exact intersection control makes the filled hierarchy sufficiently acylindrical: long tree segments have tightly constrained stabilizers.

Within the filled pieces, binary labels and attached panels build separating walls. These walls give enough boundary separation for a known cubulation criterion. Agol’s theorem then supplies virtual compact specialness of the filled groups.

Good quotients alone do not automatically imply that the original group is special. A separate recovery criterion uses selective fillings that retain prescribed finite coefficient data. That information is transferred into two opposite Novikov completions, formal series rings organized by increasing or decreasing values of a character to $\mathbb Z$. The two-sided homological control produces a finite-index subgroup with a finite-rank free kernel in the cases needed by the construction. Established cubulation and hierarchy results then recover specialness and propagate it upward.

The later general one-relator free-by-cyclic conclusion uses Kielak–Linton. It retains the possibly infinite-rank kernel, rather than extending the stronger intermediate fibering conclusion to every one-relator group.

7. Scope, limitations, and open questions

The subgroup hypothesis excludes actual embedded Baumslag–Solitar groups, not just presentations that visibly contain their defining relation. The theorem is a structural criterion; it does not by itself give a stated efficient algorithm for detecting every such subgroup.

M1’s technical combination theorem has hypotheses on hyperbolic vertex groups, free edge groups, their intersections and rank behavior. It is not a theorem that every Baumslag–Solitar-free HNN extension is hyperbolic.

M2’s compactness and finite-index qualifications should both be retained. An abstract right-angled Artin embedding alone does not provide the claimed finite special model. Quasiconvex-subgroup consequences also cannot be silently upgraded to all finitely generated subgroups.

Finally, the finite rank of the ambient free group $F_d$ does not make the general subgroup kernel finite rank. The stronger virtual finite-rank fibering statement is separately established for the hyperbolic ascending tori, and its character may differ from their canonical height. The related quadratic-filling problem for nonhyperbolic ascending tori lies outside these claims.

8. Sources and further reading

For the classical starting point, see Magnus’s Freiheitssatz and Newman’s one-relator spelling theorem. Linton’s hierarchy paper and primitive-extension reduction explain the recent reductions used here.

Haglund–Wise’s special cube complexes, Wise’s quasiconvex hierarchy monograph, and Kielak–Linton’s virtual free-by-cyclic theorem supply the geometric and fibering background. Mutanguha’s endomorphism theorem gives the earlier ascending-torus criterion.

Source manuscripts

• M1. OpenAI, Baumslag-Solitar-free one-relator groups are hyperbolic (September 25, 2026).
• M2. OpenAI, Virtual compact specialness of hyperbolic one-relator groups (September 25, 2026).

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