I have read on wikipedia that a Gromov hyperbolic group which is solvable is elementary (i.e. virtually cyclic). Where can I find a proof of this fact?

There is a proof of a similar fact in Bridson-Haefliger that if a solvable group $\Gamma$ acts properly and cocompactly on a CAT(0) space, then it is virtually abelian. Is the proof similar to this one?

  • 2
    $\begingroup$ Take a look at Bridson-Haefliger, III.$\Gamma$.3.20. $\endgroup$
    – Steve D
    Commented Oct 7, 2019 at 21:33

1 Answer 1


I do not know an exact reference. I think it is a folklore. Here is a proof using basic properties of hyperbolic groups which can be found in any book on hyperbolic groups. Let $G$ be solvable and hyperbolic. Then it has an Abelian normal subgroup $H$, the last nontrivial member of the derived series. If $H$ is finite, then $G/H$ is quasi-isometric to $G$, so it is hyperbolic and we can proceed by induction on the solvability class. So $H$ is infinite. It cannot contain an infinite locally finite subgroup because all finite subgroups of a hyperbolic group have uniformly bounded orders. $H$ cannot contain a subgroup isomorphic to $\mathbb Z^2$ because $G$ is hyperbolic. Hence $H$ is virtually cyclic and contains a characteristic infinite cyclic subgroup $H_0$. Then $H_0$ is normal in $G$. So $G$ has a homomorphism into $Out(\mathbb Z)$ which is a group of order 2. Hence $G$ has a subgroup $N$ of index at most 2 which centralizes $H_0$. A centralizer of an infinite cyclic subgroup in a hyperbolic group is virtually cyclic. Hence $G$ is virtually cyclic.

PS. I was talking about Gromov hyperbolic, hence finitely generated discrete groups. The (solvable) group of $2 \times 2 $ upper triangular matrices with determinant 1, with Riemannian metric is quasi-isometric to the hyperbolic plane, hence is hyperbolic itself.

PPS. Another proof is based on Druţu, Cornelia; Sapir, Mark Tree-graded spaces and asymptotic cones of groups. With an appendix by Denis Osin and Mark Sapir. Topology 44 (2005), no. 5. If $G$ is solvable, it satisfies a non-trivial law. Hence if $G$ is finitely generated but not virtually cyclic, its asymptotic cones do not contain cut points. But if $G$ is hyperbolic then all asymptotic cones are trees where every point is a cut-point. Hence $G$ is virtually cyclic.

This proof is of course an overkill, but it works for every group satisfying a non-trivial law, not just solvable groups. Also "hyperbolic " can be replaced by "relatively hyperbolic", "acylindrically hyperbolic" and even by "lacunary hyperbolic".

PPPS Of course the easiest proof is using the fact that every nonelementary hyperbolic group contains a noncyclic free subgroup (a solvable group cannot contain such a subgroup). I do not remember who proved it first. Gromov, probably. Maybe Olshanskiy or Delzant.

  • 3
    $\begingroup$ Gromov proved everything in the 1987 book. He proved results for an arbitrary action by isometries on a geodesic hyperbolic space, one has the 5-chotomy (a) bd orbits (b) horocyclic: no loxodromic, unbd orbits, (c) axial: there's a loxo and all have the same endpoints (d) focal: not axial, exists loxo and they have common endpoint (e) general type: there are 2 loxo with 4 distinct endpoints. He proved that (e) implies a discrete nonabelian free subgroup. Also (b) excludes cobounded action. Eventually he proved that (d) is impossible for a discrete hyperbolic group acting on itself. $\endgroup$
    – YCor
    Commented Oct 8, 2019 at 10:04
  • $\begingroup$ The Gromov reference is Remark 3.1.A(b) of Gromov, M. "Hyperbolic groups", Essays in group theory, pp. 75–263, Math. Sci. Res. Inst. Publ., 8, Springer, New York, 1987 $\endgroup$
    – HJRW
    Commented Oct 16, 2019 at 11:08
  • $\begingroup$ Gromov did not prove PPS. $\endgroup$
    – user6976
    Commented Oct 16, 2019 at 11:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.