1 Background
Ideal boundaries of hyperbolic spaces Suppose that
is a hyperbolic metric space. Pick a base-point
. This defines the Gromov product
for points
. The ideal boundary
of
is the collection of equivalence classes
of sequences
in
where
if and only if

The topology on
is defined as follows. Let
. Define
-neighborhood of
to be
![$$U(\xi,r):= \{\eta\in \geo X: \exists (x_i), (y_i) \hbox{~~with~~} \xi=[x_i], \eta=[y_i], \liminf_{i, j\to \infty} (x_i, y_j)_o\ge r\}.$$ $$U(\xi,r):= \{\eta\in \geo X: \exists (x_i), (y_i) \hbox{~~with~~} \xi=[x_i], \eta=[y_i], \liminf_{i, j\to \infty} (x_i, y_j)_o\ge r\}.$$](/pggt/Boundaries/Section1?action=AttachFile&do=get&target=latex_63fb1885cab4dae5bb19ba423bae4f628724f2fd_p1.png)
Then the basis of topology at
consists of
. We will refer to the resulting ideal boundary
as the Gromov--boundary of
. One can check that the topology on
is independent of the choice of the base-point. Moreover, if
is a quasi-isometry then it induces a homeomorphism
. The Gromov product extends to a continuous function
![$$(\xi, \eta)_o: \geo X\times \geo X\to [0, \infty].$$ $$(\xi, \eta)_o: \geo X\times \geo X\to [0, \infty].$$](/pggt/Boundaries/Section1?action=AttachFile&do=get&target=latex_f794f8f618374a20dec0be0118defbf7d58b7a46_p1.png)
The geodesic boundary of
admits a family of visual metrics
defined as follows. Pick a positive parameter
. Given points
consider various chains
(where
varies) so that
. Given such a chain, define

where
. Finally,

where the infimum is taken over all chains connecting
and
. Taking different values of
results in H\"older--equivalent metrics. Each quasi-isometry
yields a quasi-symmetric homeomorphism (see section \ref{analytical} for the definition)

Conversely, each quasi-symmetric homeomorphism as above extends to a quasi-iso\-metry
, see MR1395067.
Then the ideal boundary of a Gromov--hyperbolic group
is defined as

where
is a Cayley graph of
. Hence
is well-defined up to a quasi-symmetric homeomorphism.
Ideal boundaries of
spaces Consider a
space
. Two geodesic rays
are said to be equivalent if there exists a constant
such that

The geodesic boundary
of
is defined to be the set of equivalence classes
of geodesic rays
in
. Fix a base-point
. If
is locally compact (which we will assume from now on), then there exists a unique a representative
in each equivalence class
so that
. With this convention the visual topology on
is defined as the compact-open topology on the space of maps
. One can check that this topology is independent of the choice of the base-point and that isometries
induce homeomorphisms
.
Example. If
then
is homeomorphic to
.
If
is a
space then it is also Gromov-hyperbolic. Then the two ideal boundaries of
(one defined via sequences and the other defined via geodesic rays) are canonically homeomorphic to each other. More specifically, each geodesic ray
defines sequences
, for
diverging to infinity. The equivalence class of such
is independent of
and one gets a homeomorphism from the
-boundary to the Gromov-boundary.
In general, quasi-isometries of
spaces to not extend to the ideal boundaries in any sense. Moreover, Bruce Kleiner and Chris Croke constructed examples MR1746908 of pairs of
spaces
which admit geometric (i.e. isometric, discrete, cocompact) actions by the same group
so that
are not homeomorphic.
Therefore, given a
--group
one can talk only of the collection of
boundaries of
, i.e. the set

where the actions
are geometric.
