## Quasi-isometric rigidity of non-cocompact $S$-arithmetic lattices

### April 10, 2005

Throughout we let $K$  be an algebraic number field, ${V}_{K}$  the set of all inequivalent valuations on $K$  , and ${V}_{K}^{\infty }\subseteq {V}_{K}$  the subset of archimedean valuations. We will use $S$  to denote a finite subset of ${V}_{K}$  that contains ${V}_{K}^{\infty }$  , and we write the corresponding ring of $S$  -integers in $K$  as ${\mathcal{O}}_{S}$  .
In this paper, $\mathbf{G}$  will always be a connected non-commutative absolutely simple algebraic $K$  -group. Any group of the form $\mathbf{G}\left({\mathcal{O}}_{S}\right)$  is called an $S$  -arithmetic group. For example, if $m\in \mathbb{N}$  , then $\mathbf{P}\mathbf{G}{\mathbf{L}}_{\mathbf{n}}\left(\mathbb{Z}\left[1/m\right]\right)$  is an $S$  -arithmetic group.
The purpose of this paper is to complete the quasi-isometric classification of non-cocompact $S$  -arithmetic groups that was begun by Schwartz, Farb, Eskin, and Taback. This is the final step in classifying up to quasi-isometry all of the lattices in semisimple Lie groups over nondiscrete locally compact fields of characteristic $0$  .
Specifically, we show:
Main Result. If $\mathbf{G}\left({\mathcal{O}}_{S}\right)$  is a non-cocompact $S$  -arithmetic group that is not abstractly commensurable with $\mathbf{S}{\mathbf{L}}_{2}\left(\mathbb{Z}\right)$  , then any quasi-isometry of $\mathbf{G}\left({\mathcal{O}}_{S}\right)$  is a finite distance in the sup-norm from a commensurator of $\mathbf{G}\left({\mathcal{O}}_{S}\right)$  .
For a more precise statement of the Main Result, see Theorem A below.
Examples of the Main Result that had been unknown include that the quasi-isometry group of $\mathbf{P}\mathbf{G}{\mathbf{L}}_{2}\left(\mathbb{Z}\left[1/m\right]\right)$  is isomorphic to $\mathbf{P}\mathbf{G}{\mathbf{L}}_{2}\left(\mathbb{Q}\right)$  when $m$  is composite.
$\mathbf{S}$  -arithmetic groups are distinguished among all finitely-generated groups. Corollary B below shows that any finitely-generated group that is quasi-isometric to a non-cocompact $S$  -arithmetic group is isomorphic to that $S$  -arithmetic group “up to finite groups”.
Our restrictive definition of $\mathbf{S}$  -arithmetic groups. The definition of an $S$  -arithmetic group that is being used in this paper is a bit more restrictive than the standard definition which allows $\mathbf{G}$  to have a nontrivial center.
However, our result still determines the quasi-isometry group of any group that fits the standard definition of a non-cocompact $S$  -arithmetic group since each $S$  -arithmetic group in the standard and more inclusive definition is quasi-isometric to an $S$  -arithmetic group that fits our definition.
For example, $\mathbf{S}{\mathbf{L}}_{\mathbf{n}}$  has nontrivial center, but $\mathbf{S}{\mathbf{L}}_{\mathbf{n}}\left(\mathbb{Z}\left[1/m\right]\right)$  and $\mathbf{P}\mathbf{G}{\mathbf{L}}_{\mathbf{n}}\left(\mathbb{Z}\left[1/m\right]\right)$  differ only by finite groups. Therefore, $\mathcal{Q}\mathcal{ℐ}\left(\mathbf{S}{\mathbf{L}}_{\mathbf{n}}\left(\mathbb{Z}\left[1/m\right]\right)\right)\sim =\mathbf{P}\mathbf{G}{\mathbf{L}}_{\mathbf{n}}\left(\mathbb{Q}\right)\mathbb{\setminus }C$  where $C=1$  if $n=2$  , and $C=\mathbb{Z}/2\mathbb{Z}$  otherwise.
Past contributions to the Main Result. The Main Result was proved in several steps beginning with arithmetic groups, that is, with $S$  -arithmetic groups $\mathbf{G}\left({\mathcal{O}}_{S}\right)$  where ${\mathcal{O}}_{S}=\mathcal{O}$  is a ring of integers in a number field.
Schwartz proved the first case of the Main Result, the case when $\mathbf{G}\left(\mathcal{O}\right)$  is a lattice in a real rank one simple Lie group [Sch1. The quasi-isometry types of lattices in higher-rank real Lie groups was first explored by Farb-Schwartz who proved the Main Result for the case when $\mathbf{G}\left(\mathcal{O}\right)$  is quasi-isometric to a Hilbert modular group such as $\mathbf{S}{\mathbf{L}}_{2}\left(\mathbb{Z}\left[\sqrt{2}\right]\right)$  [Fa-Sch. After Schwartz determined the quasi-isometry groups of arithmetic groups of the form $\mathbf{P}\mathbf{G}{\mathbf{L}}_{2}\left(\mathcal{O}\right)$  [Sch2, Eskin—using results of Eskin-Farb [Es-Faand Lubotzky-Mozes-Raghunathan [L-M-R—proved the Main Result for any arithmetic group that is a lattice in a real semisimple Lie group with no rank one factors, including all arithmetic groups of higher $\mathbb{Q}$  -rank [Es. The final pieces of the full quasi-isometric classification of non-cocompact arithmetic groups were put into place by Farb in [Fa.
Taback extended the arithmetic version of the Main Result to the $S$  -arithmetic groups $\mathbf{P}\mathbf{G}{\mathbf{L}}_{2}\left(\mathbb{Z}\left[1/p\right]\right)$  , showing that their quasi-isometry groups are isomorphic to $\mathbf{P}\mathbf{G}{\mathbf{L}}_{2}\left(\mathbb{Q}\right)$  [Ta.
In [W1and [W2, the Main Result was proved in the case when $S$  contains a nonarchimedean place, and the relative rank of $\mathbf{G}$  is greater than $1$  for each completion of $K$  that corresponds to a valuation in $S$  .
The general case. This paper is based on a technique of combining a large-scale viewpoint of reduction theory with results from [W2and Kleiner-Leeb's [K-Lto allow each of the previously known cases of the Main Result to be commissioned in our proof for the remaining open cases. Thus, our proof avoids the analysis of some of the details that might have been needed in a proof that started from scratch.
Quasi-isometries. For constants $L\ge 1$  and $C\ge 0$  , an $\left(L,C\right)$  quasi-isometric embedding of a metric space $X$  into a metric space $Y$  is a function $\phi :X\to Y$  such that for any ${x}_{1},{x}_{2}\in X$  :
$\frac{1}{L}d\left({x}_{1},{x}_{2}\right)-C\le d\left(\phi \left({x}_{1}\right),\phi \left({x}_{2}\right)\right)\le Ld\left({x}_{1},{x}_{2}\right)+C$  We call $\phi$  an $\left(L,C\right)$  quasi-isometry if $\phi$  is an $\left(L,C\right)$  quasi-isometric embedding and there is a number $D\ge 0$  such that every point in $Y$  is within distance $D$  of some point in the image of $X$  .
Quasi-isometry groups. For a metric space $X$  , we define the relation $\sim$  on the set of functions $X\to X$  by $\phi \sim \psi$  if ${sup}_{x\in X}d\left(\phi \left(x\right),\psi \left(x\right)\right)<\infty$  In this paper we will call two functions equivalent if they are related by $\sim$  .
For a finitely generated group with a word metric $\Gamma$  , we form the set of all quasi-isometries of $\Gamma$  , and denote the quotient space modulo $\sim$  by $\mathcal{Q}\mathcal{ℐ}\left(\Gamma \right)$  . We call $\mathcal{Q}\mathcal{ℐ}\left(\Gamma \right)$  the quasi-isometry group of $\Gamma$  as it has a natural group structure arising from function composition.
In addition to a group structure, we also endow $\mathcal{Q}\mathcal{ℐ}\left(\Gamma \right)$  with the quotient of the compact-open topology.
Commensurators. For any valuation $v\in {V}_{K}$  , we let ${K}_{v}$  be the completion of $K$  with respect to $v$  . For any set of valuations ${S}^{\prime }\subseteq {V}_{K}$  , we define ${G}_{{S}^{\prime }}={\prod }_{v\in {S}^{\prime }}\mathbf{G}\left({K}_{v}\right)$  We identify $\mathbf{G}\left({\mathcal{O}}_{S}\right)$  as a discrete subgroup of ${G}_{S}$  using the diagonal embedding. The quotient space $\mathbf{G}\left({\mathcal{O}}_{S}\right)\{G}_{S}$  is not compact if and only if $\mathbf{G}$  is $K$  -isotropic.
We let $\text{Aut}\left({G}_{S}\right)$  be the group of topological group automorphisms of ${G}_{S}$  .
An automorphism $\psi \in \text{Aut}\left({G}_{S}\right)$  commensurates $\mathbf{G}\left({\mathcal{O}}_{S}\right)$  if $\psi \left(\mathbf{G}\left({\mathcal{O}}_{S}\right)\right)\cap \mathbf{G}\left({\mathcal{O}}_{S}\right)$  is a finite index subgroup of both $\psi \left(\mathbf{G}\left({\mathcal{O}}_{S}\right)\right)$  and $\mathbf{G}\left({\mathcal{O}}_{S}\right)$  . Define the commensurator group of $\mathbf{G}\left({\mathcal{O}}_{S}\right)$  to be the subgroup of $\text{Aut}\left({G}_{S}\right)$  consisting of automorphisms that commensurate $\mathbf{G}\left({\mathcal{O}}_{S}\right)$  . Denote this group as $\text{Comm}\left(\mathbf{G}\left({\mathcal{O}}_{S}\right)\right)$  and notice that it is different from the standard definition of the commensurator group of $\mathbf{G}\left({\mathcal{O}}_{S}\right)$  in that we have not restricted ourselves to inner automorphisms.
Precise statement of the Main Result. We will show that for all but essentially one of the non-cocompact $S$  -arithmetic lattices, the quasi-isometry group of the lattice is isomorphic to the commensurator group of the lattice.
Theorem A. Suppose $\mathbf{G}$  is $K$  -isotropic. If either $K\mathbb{\ncong }\mathbb{Q}$  , $S\ne {V}_{K}^{\infty }$  , or $\mathbf{G}$  is not $\mathbb{Q}$  -isomorphic to $\mathbf{P}\mathbf{G}{\mathbf{L}}_{2}$  , then there is an isomorphism of topological groups $\mathcal{Q}\mathcal{ℐ}\left(\mathbf{G}\left({\mathcal{O}}_{S}\right)\right)\sim =\text{Comm}\left(\mathbf{G}\left({\mathcal{O}}_{S}\right)\right)$  As a routine exercise, we can deduce from Theorem A that any quasi-isometry from a finitely-generated group to a lattice as in Theorem A is close to being a group isomorphism.
Corollary B. Suppose $K$  , $S$  , and $\mathbf{G}$  are as in Theorem A. Assume that $\Gamma$  is a finitely-generated group and that there is a quasi-isometry $\phi :\Gamma \to \mathbf{G}\left({\mathcal{O}}_{S}\right)$  Then there exists a finite-index subgroup $\Lambda$  of $\Gamma$  and a homomorphism $\phi :\Lambda \to \mathbf{G}\left({\mathcal{O}}_{S}\right)$  with a finite kernel and finite co-image such that $\phi$  is equivalent to $\phi$  restricted to $\Lambda$  . Function fields. For all of our definitions above, we could replace $K$  with a global function field $L$  (we would need to add the assumption that $S\ne \varnothing$  since ${V}_{L}^{\infty }=\varnothing$  ). A proof similar to our proof of Theorem A shows that Theorem A holds if $K$  is replaced by $L$  , and the assumption that either $K\mathbb{\ncong }\mathbb{Q}$  , $S\ne {V}_{K}^{\infty }$  , or $\mathbf{G}$  is not $\mathbb{Q}$  -isomorphic to $\mathbf{P}\mathbf{G}{\mathbf{L}}_{2}$  is replaced by the two assumptions that ${\text{rank}}_{L}\left(\mathbf{G}\right)=1$  and ${\text{rank}}_{{L}_{v}}\left(\mathbf{G}\right)>1$  for at least one valuation $v\in S$  .
The two assumptions above made on the relative ranks of $\mathbf{G}$  can probably be replaced with the single assumption that ${\sum }_{v\in S}{\text{rank}}_{{L}_{v}}\left(\mathbf{G}\right)>1$  . Note that this single condition on the relative ranks of $\mathbf{G}$  is equivalent to $\mathbf{G}\left({\mathcal{O}}_{S}\right)$  being finitely generated.
Organization of the paper. In Section 2 we will give a description of horoballs in a product of a symmetric space and a Bruhat-Tits tree. This description will be used in Section 1 to prove Theorem A. Acknowledgements. I am grateful to the following mathematicians who contributed to this paper: Tara Brendle, Kariane Calta, Indira Chatterji, Alex Eskin, Benson Farb, Dan Margalit, Dave Witte Morris, Jennifer Taback, and Kevin Whyte. 1. Proof of Theorem A Let $\mathbf{G}$  , $K$  , and $S$  be as in Theorem A, and let $\phi :\mathbf{G}\left({\mathcal{O}}_{S}\right)\to \mathbf{G}\left({\mathcal{O}}_{S}\right)$  be a quasi-isometry. If $S={V}_{K}^{\infty }$  , then Theorem A reduces to the already existing quasi-isometric rigidity theorem for arithmetic groups (see [Fa), so we shall always assume that ${V}_{K}^{\infty }$  is a proper subset of $S$  .
Geometric models. For each valuation $v$  of $K$  , we let ${X}_{v}$  be the symmetric space or Euclidean building corresponding to $\mathbf{G}\left({K}_{v}\right)$  . If ${S}^{\prime }$  is a finite set of valuations of $K$  , we let ${X}_{{S}^{\prime }}={\prod }_{v\in {S}^{\prime }}{X}_{v}$  Recall that there is a natural inclusion of topological groups $\text{Aut}\left({G}_{{S}^{\prime }}\right)↪\text{Isom}\left({X}_{{S}^{\prime }}\right)$  .
Let $\mathcal{O}$  be the ring of integers in $K$  , and fix a connected subspace ${\Omega }_{{V}_{K}^{\infty }}\subseteq {X}_{{V}_{K}^{\infty }}$  that $\mathbf{G}\left(\mathcal{O}\right)$  acts cocompactly on. Let ${D}_{\infty }\subseteq {X}_{{V}_{K}^{\infty }}$  be a fundamental domain for this action.
For each nonarchimedean valuation $w\in S-{V}_{K}^{\infty }$  , we denote the ring of integers in ${K}_{w}$  by ${\mathcal{O}}_{w}$  . The group $\mathbf{G}\left({\mathcal{O}}_{w}\right)$  is bounded in $\mathbf{G}\left({K}_{w}\right)$  , so $\mathbf{G}\left({\mathcal{O}}_{w}\right)$  fixes a point ${x}_{w}\in {X}_{w}$  . We choose a bounded set ${D}_{w}\subseteq {X}_{w}$  containing ${x}_{w}$  with $\mathbf{G}\left({\mathcal{O}}_{S}\right){D}_{w}={X}_{w}$  and such that $g{x}_{w}\in {D}_{w}$  for $g\in \mathbf{G}\left({\mathcal{O}}_{S}\right)$  implies that $g{x}_{w}={x}_{w}$  .
For any set of valuations ${S}^{\prime }$  satisfying ${V}_{K}^{\infty }\subseteq {S}^{\prime }\subseteq S$  , we define the space ${\Omega }_{{S}^{\prime }}=\mathbf{G}\left({\mathcal{O}}_{{S}^{\prime }}\right)\left({D}_{\infty }×{\prod }_{w\in {S}^{\prime }-{V}_{K}^{\infty }}{D}_{w}\right)$  Note that ${\Omega }_{{S}^{\prime }}$  is a subspace of ${X}_{{S}^{\prime }}$  .
We endow ${\Omega }_{{S}^{\prime }}$  with the path metric. Since $\mathbf{G}\left({\mathcal{O}}_{{S}^{\prime }}\right)$  acts cocompactly on ${\Omega }_{{S}^{\prime }}$  , we have the following observation:
Lemma 1.1. For ${V}_{K}^{\infty }\subseteq {S}^{\prime }\subseteq S$  , the space ${\Omega }_{{S}^{\prime }}$  is quasi-isometric to the group $\mathbf{G}\left({\mathcal{O}}_{{S}^{\prime }}\right)$  .
Fibers of projections to buildings are $S$  -arithmetic. In the large-scale, the fibers of the projection of ${\Omega }_{S}$  onto building factors of ${X}_{S}$  are also $S$  -arithmetic groups (or more precisely, ${S}^{\prime }$  -arithmetic groups). This is the statement of Lemma 1.3 below, but we will start with a proof of a special case.
Lemma 1.2. The Hausdorff distance between ${\Omega }_{S}\cap \left({X}_{{S}^{\prime }}×{\prod }_{w\in S-{S}^{\prime }}\left\{{x}_{w}\right\}\right)$  and ${\Omega }_{{S}^{\prime }}×{\prod }_{w\in S-{S}^{\prime }}\left\{{x}_{w}\right\}$  is finite.
Proof. There are three main steps in this proof. First, if $y\in {\Omega }_{{S}^{\prime }}$  , then $y=gd$  for some $g\in \mathbf{G}\left({\mathcal{O}}_{{S}^{\prime }}\right)$  and some $d\in {D}_{\infty }×{\prod }_{w\in {S}^{\prime }-{V}_{K}^{\infty }}{D}_{w}$  Since $\mathbf{G}\left({\mathcal{O}}_{{S}^{\prime }}\right)\le \mathbf{G}\left({\mathcal{O}}_{w}\right)$  for all $w\in S-{S}^{\prime }$  , it follows from our choice of the points ${x}_{w}$  that
 $\begin{array}{cc}\left\{y\right\}×{\prod }_{w\in S-{S}^{\prime }}\left\{{x}_{w}\right\}& =g\left(\left\{d\right\}×{\prod }_{w\in S-{S}^{\prime }}\left\{{x}_{w}\right\}\right)\subseteq {\Omega }_{S}\end{array}$
 $\begin{array}{}\end{array}$
Therefore,
 $\begin{array}{cc}{\Omega }_{{S}^{\prime }}×{\prod }_{w\in S-{S}^{\prime }}\left\{{x}_{w}\right\}\subseteq {\Omega }_{S}\cap \left({X}_{{S}^{\prime }}×{\prod }_{w\in S-{S}^{\prime }}\left\{{x}_{w}\right\}\right)& \end{array}$ (1)
 $\begin{array}{}\end{array}$
Second, we suppose $\left\{y\right\}×{\prod }_{w\in S-{S}^{\prime }}\left\{{x}_{w}\right\}\subseteq {\Omega }_{S}$  for some $y\in {X}_{{S}^{\prime }}$  . Then there exists a $g\in \mathbf{G}\left({\mathcal{O}}_{S}\right)$  such that $gy\in {D}_{\infty }×{\prod }_{w\in {S}^{\prime }-{V}_{K}^{\infty }}{D}_{w}$  and $g{x}_{w}\in {D}_{w}$  for all $w\in S-{S}^{\prime }$  . Notice that our choice of ${D}_{w}$  implies $g{x}_{w}={x}_{w}$  for all $w\in S-{S}^{\prime }$  . Thus, $g$  is contained in the compact group ${H}_{w}=\left\{h\in \mathbf{G}\left({K}_{w}\right)|h{x}_{w}={x}_{w}\right\}$  for all $w\in S-{S}^{\prime }$  . Consequently, g is contained in the discrete group $\mathbf{G}\left({\mathcal{O}}_{S}\right)\cap \left({G}_{{S}^{\prime }}×{\prod }_{w\in S-{S}^{\prime }}{H}_{w}\right)$  We name this discrete group ${\Gamma }_{{S}^{\prime }}$  .
Note that we have shown $\left\{y\right\}×{\prod }_{w\in S-{S}^{\prime }}\left\{{x}_{w}\right\}\subseteq {\Gamma }_{{S}^{\prime }}\left({D}_{\infty }×{\prod }_{w\in S-{V}_{K}^{\infty }}{D}_{w}\right)$  Therefore,
 $\begin{array}{cc}{\Omega }_{S}\cap \left({X}_{{S}^{\prime }}×{\prod }_{w\in S-{S}^{\prime }}\left\{{x}_{w}\right\}\right)\subseteq {\Gamma }_{{S}^{\prime }}\left({D}_{\infty }×{\prod }_{w\in S-{V}_{K}^{\infty }}{D}_{w}\right)& \end{array}$ (2)
 $\begin{array}{}\end{array}$
Third, we recall that $\mathbf{G}\left({\mathcal{O}}_{{S}^{\prime }}\right)=\mathbf{G}\left({\mathcal{O}}_{S}\right)\cap \left({G}_{{S}^{\prime }}×{\prod }_{w\in S-{S}^{\prime }}\mathbf{G}\left({\mathcal{O}}_{w}\right)\right)$  and use the definition of ${\Gamma }_{{S}^{\prime }}$  coupled with the fact that $\mathbf{G}\left({\mathcal{O}}_{w}\right)\le {H}_{w}$  to see that ${\Gamma }_{{S}^{\prime }}$  contains $\mathbf{G}\left({\mathcal{O}}_{{S}^{\prime }}\right)$  . Since, ${\Gamma }_{{S}^{\prime }}$  and $\mathbf{G}\left({\mathcal{O}}_{{S}^{\prime }}\right)$  are lattices in ${G}_{{S}^{\prime }}×{\prod }_{w\in S-{S}^{\prime }}{H}_{w}$  , the containment $\mathbf{G}\left({\mathcal{O}}_{{S}^{\prime }}\right)\le {\Gamma }_{{S}^{\prime }}$  is of finite index. Therefore, the Hausdorff distance between ${\Gamma }_{{S}^{\prime }}\left({D}_{\infty }×{\prod }_{w\in {S}^{\prime }-{V}_{K}^{\infty }}{D}_{w}×{\prod }_{w\in S-{S}^{\prime }}\left\{{x}_{w}\right\}\right)$  and ${\Omega }_{{S}^{\prime }}×{\prod }_{w\in S-{S}^{\prime }}\left\{{x}_{w}\right\}=\mathbf{G}\left({\mathcal{O}}_{{S}^{\prime }}\right)\left({D}_{\infty }×{\prod }_{w\in {S}^{\prime }-{V}_{K}^{\infty }}{D}_{w}×{\prod }_{w\in S-{S}^{\prime }}\left\{{x}_{w}\right\}\right)$  is finite. Combined with (1) and (2) above, the lemma follows.
$■$  The more general form of Lemma 1.2 that we will use in our proof of Theorem A is the following lemma. We will use the notation of ${x}_{S-{S}^{\prime }}$  for the point $\left({x}_{w}{\right)}_{w\in S-{S}^{\prime }}\in {X}_{S-{S}^{\prime }}$  .
Lemma 1.3. Suppose ${V}_{K}^{\infty }\subseteq {S}^{\prime }\subseteq S$  . If $y\in {X}_{S-{S}^{\prime }}$  and $y\in \mathbf{G}\left({\mathcal{O}}_{S}\right){x}_{S-{S}^{\prime }}$  , then the Hausdorff distance between ${\Omega }_{S}\cap \left({X}_{{S}^{\prime }}×\left\{y\right\}\right)$  and ${\Omega }_{{S}^{\prime }}×\left\{y\right\}$  is finite.
Remark. Our assumption in Lemma 1.3 that $y\in \mathbf{G}\left({\mathcal{O}}_{S}\right){x}_{S-{S}^{\prime }}$  is not a serious restriction over the assumption that $y\in {X}_{S-{S}^{\prime }}$  . Indeed, $\mathbf{G}\left({\mathcal{O}}_{S}\right)$  is dense in ${G}_{S-{S}^{\prime }}$  , so the orbit $\mathbf{G}\left({\mathcal{O}}_{S}\right){x}_{S-{S}^{\prime }}$  is a finite Hausdorff distance from the space ${X}_{S-{S}^{\prime }}$  .
Proof. Let $g\in \mathbf{G}\left({\mathcal{O}}_{S}\right)$  be such that $y=g{x}_{S-{S}^{\prime }}$  . Then
 $\begin{array}{cc}\left\{h\in \mathbf{G}\left({\mathcal{O}}_{S}\right)|h{x}_{S-{S}^{\prime }}=y\right\}& =g\left\{h\in \mathbf{G}\left({\mathcal{O}}_{S}\right)|h{x}_{S-{S}^{\prime }}={x}_{S-{S}^{\prime }}\right\}\end{array}$
 $\begin{array}{cc}& =g\left(\mathbf{G}\left({\mathcal{O}}_{S}\right)\cap \left({G}_{{S}^{\prime }}×{\prod }_{w\in S-{S}^{\prime }}{H}_{w}\right)\right)\end{array}$
 $\begin{array}{cc}& =g{\Gamma }_{{S}^{\prime }}\end{array}$
 $\begin{array}{}\end{array}$
where ${H}_{w}$  and ${\Gamma }_{{S}^{\prime }}$  are as in the proof of the previous lemma.
Now by our choice of the points ${x}_{w}\in {X}_{w}$  for $w\in S-{V}_{K}^{\infty }$  at the beginning of this section, we have
 $\begin{array}{cc}{\Omega }_{S}\cap \left({X}_{{S}^{\prime }}×\left\{y\right\}\right)& =\mathbf{G}\left({\mathcal{O}}_{S}\right)\left({D}_{\infty }×{\prod }_{w\in S-{V}_{K}^{\infty }}{D}_{w}\right)\cap \left({X}_{{S}^{\prime }}×\left\{y\right\}\right)\end{array}$
 $\begin{array}{cc}& =g{\Gamma }_{{S}^{\prime }}\left({D}_{\infty }×{\prod }_{w\in {S}^{\prime }-{V}_{K}^{\infty }}{D}_{w}\right)×\left\{y\right\}\end{array}$
 $\begin{array}{}\end{array}$
Notice that the final space from the above chain of equalities is a finite Hausdorff distance from $g\mathbf{G}\left({\mathcal{O}}_{{S}^{\prime }}\right)\left({D}_{\infty }×{\prod }_{w\in {S}^{\prime }-{V}_{K}^{\infty }}{D}_{w}\right)×\left\{y\right\}$  since ${\Gamma }_{{S}^{\prime }}$  is commensurable with $\mathbf{G}\left({\mathcal{O}}_{{S}^{\prime }}\right)$  .
Because $g$  commensurates $\mathbf{G}\left({\mathcal{O}}_{{S}^{\prime }}\right)$  , the above space is also a finite Hausdorff distance from ${\Omega }_{{S}^{\prime }}×\left\{y\right\}$  .
$■$  Extending quasi-isometries of ${\Omega }_{S}$  to ${X}_{S}$  . Applying Lemma 1.1, we can regard our quasi-isometry $\phi :\mathbf{G}\left({\mathcal{O}}_{S}\right)\to \mathbf{G}\left({\mathcal{O}}_{S}\right)$  as a quasi-isometry of ${\Omega }_{S}$  .
Our goal is to show that $\phi$  is equivalent to an element of $\text{Comm}\left(\mathbf{G}\left({\mathcal{O}}_{S}\right)\right)$  , and we begin by extending $\phi$  to a quasi-isometry of ${X}_{S}$  .
Lemma 1.4. There is a permutation of $S$  , which we name $\tau$  , and there are quasi-isometries ${\phi }_{v}:{X}_{v}\to {X}_{\tau \left(v\right)}$  such that the restriction of the quasi-isometry ${\phi }_{S}={\prod }_{v\in S}{\phi }_{v}:{X}_{S}\to {X}_{S}$  to ${\Omega }_{S}$  is equivalent to $\phi$  .
Proof. In Proposition 10.1 of [Es, Eskin proved this lemma for the case $S={V}_{K}^{\infty }$  . In [W2, much of Eskin's work on arithmetic lattices from [Eswas generalized to $S$  -arithmetic lattices in [W2, including all of the tools needed to apply the proof of Proposition 10.1 of [Esto prove Lemma 1.4.
Note that the statement of Proposition 10.1 from [Esclaims that ${X}_{v}$  and ${X}_{\tau \left(v\right)}$  are isometric for $v\in {V}_{K}^{\infty }$  . This is because quasi-isometric symmetric spaces are isometric up to scale.
$■$  Our goal now is to show that ${\phi }_{S}$  is equivalent to an element of $\text{Comm}\left(\mathbf{G}\left({\mathcal{O}}_{S}\right)\right)$  .
At this point, the proof breaks into two cases.
Case 1: ${G}_{{V}_{K}^{\infty }}$  is not locally isomorphic to $\mathbf{P}\mathbf{G}{\mathbf{L}}_{2}\left(\mathbb{R}\right)$  .
Notice that $\text{Comm}\left(\mathbf{G}\left({\mathcal{O}}_{S}\right)\right)$  acts by isometries on ${X}_{S}$  . So a good first step toward our goal is to show that ${\phi }_{S}$  is equivalent to an isometry. First, we will show that ${\phi }_{{V}_{K}^{\infty }}$  is equivalent to an isometry.
Lemma 1.5. The quasi-isometry ${\phi }_{{V}_{K}^{\infty }}:{X}_{{V}_{K}^{\infty }}\to {X}_{{V}_{K}^{\infty }}$  is equivalent to an isometry of the symmetric space ${X}_{{V}_{K}^{\infty }}$  . Indeed, it is equivalent to an element of $\text{Comm}\left(\mathbf{G}\left(\mathcal{O}\right)\right)$  .
Proof. Notice that ${\phi }_{{V}_{K}^{\infty }}$  is simply the restriction of ${\phi }_{S}$  to ${X}_{{V}_{K}^{\infty }}×\left\{{x}_{S-{V}_{K}^{\infty }}\right\}$  .
Since $\mathbf{G}\left({\mathcal{O}}_{S}\right)$  is dense in ${G}_{S-{V}_{K}^{\infty }}$  , the Hausdorff distance between $\mathbf{G}\left({\mathcal{O}}_{\mathcal{S}}\right){x}_{S-{V}_{K}^{\infty }}$  and ${X}_{S-{V}_{K}^{\infty }}$  is finite. Thus, by replacing ${\phi }_{S}$  with an equivalent quasi-isometry, we may assume that ${X}_{{V}_{K}^{\infty }}×\left\{{x}_{S-{V}_{K}^{\infty }}\right\}$  is mapped by ${\phi }_{S}$  into a space ${X}_{{V}_{K}^{\infty }}×\left\{y\right\}$  for some $y\in {X}_{S-{V}_{K}^{\infty }}$  with $y\in \mathbf{G}\left({\mathcal{O}}_{S}\right){x}_{S-{V}_{K}^{\infty }}$  .
Since the Hausdorff distance between ${\phi }_{S}\left({\Omega }_{S}\right)$  and ${\Omega }_{S}$  is finite, we have by Lemmas 1.3 and 1.1 that ${\phi }_{{V}_{K}^{\infty }}$  induces a quasi-isometry of $\mathbf{G}\left(\mathcal{O}\right)$  . The lemma follows from the existing quasi-isometric classification of arithmetic lattices using our assumption in this Case that ${G}_{{V}_{K}^{\infty }}$  is not locally isomorphic to $\mathbf{P}\mathbf{G}{\mathbf{L}}_{2}\left(\mathbb{R}\right)$  ; see [Fa.
$■$  At this point, it is not difficult to see that Theorem A holds in the case when every nonarchimedean factor of ${G}_{S}$  is higher rank:
Lemma 1.6. If ${\text{rank}}_{{K}_{v}}\left(\mathbf{G}\right)>1$  for all $v\in S-{V}_{K}^{\infty }$  then ${\phi }_{S}$  is equivalent to an element of $\text{Comm}\left(\mathbf{G}\left({\mathcal{O}}_{S}\right)\right)$  .
Proof. By Kleiner-Leeb's Theorem 1.1.3 in [K-L, ${\phi }_{v}$  is equivalent to an isometry for all $v\in S-{V}_{K}^{\infty }$  . Combined with Lemma 1.5, we know that ${\phi }_{S}$  is equivalent to an isometry.
That ${\phi }_{S}$  is equivalent to an element of $\text{Comm}\left(\mathbf{G}\left({\mathcal{O}}_{S}\right)\right)$  follows from Proposition 7.2 of [W2. Indeed, any isometry of ${X}_{S}$  that preserves ${\Omega }_{S}$  up to finite Hausdorff distance corresponds in a natural way to an automorphism of ${G}_{S}$  that preserves $\mathbf{G}\left({\mathcal{O}}_{S}\right)$  up to finite Hausdorff distance, and any such automorphism of ${G}_{S}$  is shown in Proposition 7.2 of [W2to be a commensurator.
$■$  For the remainder of Case 1, we are left to assume that there is at least one $w\in S-{V}_{K}^{\infty }$  such that ${\text{rank}}_{{K}_{w}}\left(\mathbf{G}\right)=1$  .
Before beginning the proof of the next and final lemma for Case 1, it will be best to recall some standard facts about boundaries.
Tree boundaries. If $w$  is a nonarchimedean valuation of $K$  , and ${\text{rank}}_{{K}_{w}}\left(\mathbf{G}\right)=1$  , then ${X}_{w}$  is a tree.
For any minimal ${K}_{w}$  -parabolic subgroup of $\mathbf{G}$  , say $\mathbf{M}$  , we let ${\varepsilon }_{\mathbf{M}}$  be the end of ${X}_{w}$  such that $\mathbf{M}\left({K}_{w}\right){\varepsilon }_{\mathbf{M}}={\varepsilon }_{\mathbf{M}}$  . Notice that the space of all ends of the form ${\varepsilon }_{\mathbf{P}}$  where $\mathbf{P}$  is a minimal $K$  -parabolic subgroup of $\mathbf{G}$  forms a dense subset of the space of ends of ${X}_{w}$  .
Tits boundaries. For any minimal $K$  -parabolic subgroup of $\mathbf{G}$  , say $\mathbf{P}$  , we let ${\delta }_{\mathbf{P}}$  be the simplex in the Tits boundary of ${X}_{{V}_{K}^{\infty }}$  corresponding to the group ${\prod }_{v\in {V}_{K}^{\infty }}\mathbf{P}\left({K}_{v}\right)$  .
If $\delta$  is a simplex in the Tits boundary of ${X}_{{V}_{K}^{\infty }}$  , and $\varepsilon$  is an end of the tree ${X}_{w}$  , then we denote the join of $\delta$  and $\varepsilon$  by $\delta *\varepsilon$  . It is a simplex in the Tits boundary of ${X}_{T}$  where $T={V}_{K}^{\infty }\cup \left\{w\right\}$  .
Lemma 1.7. Let $w\in S-{V}_{K}^{\infty }$  be such that ${\text{rank}}_{{K}_{w}}\left(\mathbf{G}\right)=1$  . Then ${\phi }_{w}:{X}_{w}\to {X}_{\tau \left(w\right)}$  is equivalent to an isometry that is induced by an isomorphism of topological groups $\mathbf{G}\left({K}_{w}\right)\to \mathbf{G}\left({K}_{\tau \left(w\right)}\right)$  .
Proof. Below, we will denote the set of valuations ${V}_{K}^{\infty }\cup \left\{\tau \left(w\right)\right\}$  by ${T}^{\tau }$  . We begin by choosing a minimal $K$  -parabolic subgroup of $\mathbf{G}$  , say $\mathbf{P}$  , and a geodesic ray $\rho :{\mathbb{R}}_{\ge 0}\to {X}_{T}$  that limits to the interior of the simplex ${\delta }_{\mathbf{P}}*{\varepsilon }_{\mathbf{P}}$  .
By Lemma 1.5, the image of ${\phi }_{T}\circ \rho$  under the projection ${X}_{T}\to {X}_{{V}_{K}^{\infty }}$  limits to a point in the interior of ${\delta }_{\mathbf{Q}}$  for some minimal $K$  -parabolic subgroup $\mathbf{Q}$  .
Similarly, ${\phi }_{w}$  is a quasi-isometry of a tree, so it maps each geodesic ray into a bounded neighborhood of a geodesic ray that is unique up to finite Hausdorff distance. Thus, the image of ${\phi }_{T}\circ \rho$  under the projection ${X}_{T}\to {X}_{\tau \left(w\right)}$  limits to ${\varepsilon }_{{\mathbf{Q}}^{\prime }}$  for some minimal ${K}_{\tau \left(w\right)}$  -parabolic subgroup ${\mathbf{Q}}^{\prime }$  . Together, these results imply that ${\phi }_{T}\circ \rho$  is a finite Hausdorff distance from a geodesic ray that limits to a point in the interior of ${\delta }_{\mathbf{Q}}*{\varepsilon }_{{\mathbf{Q}}^{\prime }}$  .
By Lemma 2.4, there is a subspace ${\mathcal{ℋ}}_{\mathbf{P}}$  of ${X}_{T}$  corresponding to $\mathbf{P}$  (called a “ $T$  -horoball”) such that $t↦d\left(\rho \left(t\right),{X}_{T}-{\mathcal{ℋ}}_{\mathbf{P}}\right)$  is an unbounded function. Thus, $t↦d\left({\phi }_{T}\circ \rho \left(t\right),{X}_{{T}^{\tau }}-{\phi }_{T}\left({\mathcal{ℋ}}_{\mathbf{P}}\right)\right)$  is also unbounded.
Using Lemma 2.1 and the fact that ${\phi }_{T}\left({\Omega }_{T}\right)$  is a finite Hausdorff distance from ${\Omega }_{{T}^{\tau }}$  , we may replace ${\phi }_{T}$  with an equivalent quasi-isometry to deduce that ${\phi }_{T}\left({\mathcal{ℋ}}_{\mathbf{P}}\right)$  is contained in the union of all ${T}^{\tau }$  -horoballs in ${X}_{{T}^{\tau }}$  .
It will be clear from the definition given in Section 2 that each $T$  -horoball is connected. Thus, it follows from Lemma 2.2 that ${\phi }_{T}\left({\mathcal{ℋ}}_{\mathbf{P}}\right)$  is a finite distance from a single ${T}^{\tau }$  -horoball ${\mathcal{ℋ}}_{\mathbf{M}}\subseteq {X}_{{T}^{\tau }}$  where $\mathbf{M}$  is a minimal $K$  -parabolic subgroup of $\mathbf{G}$  . Therefore, $t↦d\left({\phi }_{T}\circ \rho \left(t\right),{X}_{{T}^{\tau }}-{\mathcal{ℋ}}_{\mathbf{M}}\right)$  is unbounded.
Because the above holds for all $\rho$  limiting to ${\delta }_{\mathbf{P}}*{\varepsilon }_{\mathbf{P}}$  , and because ${\phi }_{T}\circ \rho$  limits to ${\delta }_{\mathbf{Q}}*{\varepsilon }_{{\mathbf{Q}}^{\prime }}$  , we have by Lemma 2.5, that $\mathbf{Q}=\mathbf{M}={\mathbf{Q}}^{\prime }$  . That is, ${\phi }_{{V}_{K}^{\infty }}$  completely determines the map that ${\phi }_{w}$  induces between the ends of the trees ${X}_{w}$  and ${X}_{\tau \left(w\right)}$  that correspond to $K$  -parabolic subgroups of $\mathbf{G}$  .
Recall that by Lemma 1.5, ${\phi }_{{V}_{K}^{\infty }}$  is equivalent to a commensurator of $\mathbf{G}\left(\mathcal{O}\right)\le {G}_{{V}_{K}^{\infty }}$  . Using Lemma 7.3 of [W2, ${\phi }_{{V}_{K}^{\infty }}$  (regarded as an automorphism of ${G}_{{V}_{K}^{\infty }}$  ) restricts to $\mathbf{G}\left(K\right)$  as a composition $\delta \circ {\sigma }^{\circ }:\mathbf{G}\left(K\right)\to \mathbf{G}\left(K\right)$  where $\sigma$  is an automorphism of $K$  , ${\sigma }^{\circ }:\mathbf{G}\left(K\right){\to }^{\sigma }\mathbf{G}\left(K\right)$  is the map obtained by applying $\sigma$  to the entries of elements in $\mathbf{G}\left(K\right)$  , and $\delta {:}^{\sigma }\mathbf{G}\to \mathbf{G}$  is a $K$  -isomorphism of $K$  -groups.
Thus, if $\partial {X}_{w}$  and $\partial {X}_{\tau \left(w\right)}$  are the space of ends of the trees ${X}_{w}$  and ${X}_{\tau \left(w\right)}$  respectively, and if $\partial {\phi }_{w}:\partial {X}_{w}\to \partial {X}_{\tau \left(w\right)}$  is the boundary map induced by ${\phi }_{w}$  , then we have shown that $\partial {\phi }_{w}\left({\varepsilon }_{\mathbf{P}}\right)={\varepsilon }_{\delta {\left(}^{\sigma }\mathbf{P}\right)}$  for any $\mathbf{P}\le \mathbf{G}$  that is a minimal $K$  -parabolic subgroup of $\mathbf{G}$  .
Our next goal is to show that the valuation $w\circ {\sigma }^{-1}$  is equivalent to $\tau \left(w\right)$  .
If this is the case, then $\delta \circ {\sigma }^{\circ }$  extends from a group automorphism of $\mathbf{G}\left(K\right)$  to a topological group isomorphism ${\phi }_{w}:\mathbf{G}\left({K}_{w}\right)\to \mathbf{G}\left({K}_{\tau \left(w\right)}\right)$  If $\partial {\phi }_{w}:\partial {X}_{w}\to \partial {X}_{\tau \left(w\right)}$  is the map induced by $\phi$  , then $\partial {\phi }_{w}$  equals $\partial {\phi }_{w}$  on the subset of ends in $\partial {X}_{w}$  corresponding to $K$  -parabolic subgroups of $\mathbf{G}$  since ${\phi }_{w}$  extends $\delta \circ {\sigma }^{\circ }$  . Therefore, $\partial {\phi }_{w}=\partial {\phi }_{w}$  on all of $\partial {X}_{w}$  by the density of the “ $K$  -rational ends” in $\partial {X}_{w}$  . Thus, ${\phi }_{w}$  determines ${\phi }_{w}$  up to equivalence.
This would prove our lemma.
So to finish the proof of this lemma, we will show that $w\circ {\sigma }^{-1}$  is equivalent to $\tau \left(w\right)$  .
For any maximal $K$  -split torus $\mathbf{S}\le \mathbf{G}$  , we let ${\gamma }_{\mathbf{S}}^{w}\subseteq {X}_{w}$  (resp. ${\gamma }_{\mathbf{S}}^{\tau \left(w\right)}\subseteq {X}_{\tau \left(w\right)}$  ) be the geodesic that $\mathbf{S}\left({K}_{w}\right)$  (resp. $\mathbf{S}\left({K}_{\tau \left(w\right)}\right)$  ) acts on by translations.
Fix $\mathbf{S}$  and $\mathbf{T}$  , two maximal $K$  -split tori in $\mathbf{G}$  such that ${\gamma }_{\mathbf{S}}^{w}\cap {\gamma }_{\mathbf{T}}^{w}$  is nonempty and bounded. We choose a point $a\in {\gamma }_{\mathbf{S}}^{w}\cap {\gamma }_{\mathbf{T}}^{w}$  .
Since $\mathbf{S}\left({\mathcal{O}}_{T}\right)$  is dense in $\mathbf{S}\left({K}_{w}\right)$  , there exists a group element ${g}_{n}\in \mathbf{S}\left({\mathcal{O}}_{T}\right)$  for each $n\in \mathbb{N}$  such that $d\left({g}_{n}\left({\gamma }_{\mathbf{T}}^{w}\right),a\right)>n$  Note that ${g}_{n}\left({\gamma }_{\mathbf{T}}^{w}\right)={\gamma }_{{g}_{n}\mathbf{T}{g}_{n}^{-1}}^{w}$  . Thus $d\left({\phi }_{w}\left({\gamma }_{{g}_{n}\mathbf{T}{g}_{n}^{-1}}^{w}\right),{\phi }_{w}\left(a\right)\right)$  is an unbounded sequence.
As ${g}_{n}\mathbf{T}{g}_{n}^{-1}$  is $K$  -split, ${\phi }_{w}\left({\gamma }_{{g}_{n}\mathbf{T}{g}_{n}^{-1}}^{w}\right)$  is a uniformly bounded Hausdorff distance from ${\gamma }_{\delta \circ {\sigma }^{\circ }\left({g}_{n}\mathbf{T}{g}_{n}^{-1}\right)}^{\tau \left(w\right)}=\delta \circ {\sigma }^{\circ }\left({g}_{n}\right){\gamma }_{\delta \circ {\sigma }^{\circ }\left(\mathbf{T}\right)}^{\tau \left(w\right)}$  because a geodesic in ${X}_{\tau \left(w\right)}$  is determined by its two ends.
We finally have that $d\left(\delta \circ {\sigma }^{\circ }\left({g}_{n}\right){\gamma }_{\delta \circ {\sigma }^{\circ }\left(\mathbf{T}\right)}^{\tau \left(w\right)},{\phi }_{w}\left(a\right)\right)$  is an unbounded sequence. It is this statement that we shall contradict by assuming that $w\circ {\sigma }^{-1}$  is inequivalent to $\tau \left(w\right)$  .
Note that ${g}_{n}\in \mathbf{G}\left({\mathcal{O}}_{v}\right)$  for all $v\in {V}_{K}-T$  since ${g}_{n}\in \mathbf{S}\left({\mathcal{O}}_{T}\right)$  . Thus, ${\sigma }^{0}\left({g}_{n}\right){\in }^{\sigma }\mathbf{G}\left({\mathcal{O}}_{v\circ {\sigma }^{-1}}\right)$  for all $v\in {V}_{K}-T$  . If it were the case that $w\circ {\sigma }^{-1}$  is inequivalent to $\tau \left(w\right)$  , then it follows that ${\sigma }^{0}\left({g}_{n}\right){\in }^{\sigma }\mathbf{G}\left({\mathcal{O}}_{\tau \left(w\right)}\right)$  . Hence, $\delta \circ {\sigma }^{0}\left({g}_{n}\right)$  defines a bounded sequence in $\mathbf{G}\left({K}_{\tau \left(w\right)}\right)$  . Therefore, $d\left(\delta \circ {\sigma }^{\circ }\left({g}_{n}\right){\gamma }_{\delta \circ {\sigma }^{\circ }\left(\mathbf{T}\right)}^{\tau \left(w\right)},{\phi }_{w}\left(a\right)\right)$  is a bounded sequence, our contradiction.
$■$  The proof of Theorem A in Case 1 is complete with the observation that applications of Lemma 1.7 to tree factors, allows us to apply the Proposition 7.2 of [W2as we did in Lemma 1.6.
Case 2: ${G}_{{V}_{K}^{\infty }}$  is locally isomorphic to $\mathbf{P}\mathbf{G}{\mathbf{L}}_{2}\left(\mathbb{R}\right)$  .
It follows that ${V}_{K}^{\infty }$  contains a single valuation $v$  , and that ${K}_{v}\sim =\mathbb{R}$  . Thus $K=\mathbb{Q}$  , and ${V}_{K}^{\infty }$  is the set containing only the standard real metric on $\mathbb{Q}$  .
Our assumption that $\mathbf{G}$  is absolutely simple implies that ${G}_{{V}_{K}^{\infty }}$  is actually isomorphic to $\mathbf{P}\mathbf{G}{\mathbf{L}}_{2}\left(\mathbb{R}\right)$  . Thus, $\mathbf{G}$  is a $\mathbb{Q}$  -form of $\mathbf{P}\mathbf{G}{\mathbf{L}}_{2}$  . As we are assuming that $\mathbf{G}$  is $\mathbb{Q}$  -isotropic, it follows from the classification of $\mathbb{Q}$  -forms of $\mathbf{P}\mathbf{G}{\mathbf{L}}_{2}$  that $\mathbf{G}$  and $\mathbf{P}\mathbf{G}{\mathbf{L}}_{2}$  are $\mathbb{Q}$  -isomorphic (see e.g. page 55 of [Ti).
From our assumptions in the statement of Theorem A, $S\ne {V}_{K}^{\infty }$  . As the only valuations, up to scale, on $\mathbb{Q}$  are the real valuation and the $p$  -adic valuations, $\mathbf{G}\left({\mathcal{O}}_{S}\right)$  is commensurable with $\mathbf{P}\mathbf{G}{\mathbf{L}}_{2}\left(\mathbb{Z}\left[1/m\right]\right)$  for some $m\in \mathbb{N}$  with $m\ne 1$  .
If $p$  is a prime dividing $m$  , and ${v}_{p}$  is the $p$  -adic valuation on $\mathbb{Q}$  , then we can substitute ${\phi }_{{V}_{K}^{\infty }\cup \left\{{v}_{p}\right\}}$  for ${\phi }_{{V}_{K}^{\infty }}$  in the proof of Lemma 1.5 by replacing our use of the quasi-isometric classification of arithmetic groups with Taback's theorem that $\mathcal{Q}\mathcal{ℐ}\left(\mathbf{P}\mathbf{G}{\mathbf{L}}_{2}\left(\mathbb{Z}\left[1/p\right]\right)\right)\sim =\mathbf{P}\mathbf{G}{\mathbf{L}}_{2}\left(\mathbb{Q}\right)$  [Ta. Then we can carry out the remainder of the proof of Theorem A as in the proof of Lemma 1.7 by again substituting ${\phi }_{{V}_{K}^{\infty }\cup \left\{{v}_{p}\right\}}$  for ${\phi }_{{V}_{K}^{\infty }}$  as long as we choose $w$  from the proof of Lemma 1.7 to be an $\ell$  -adic valuation with $p\ne \ell$  . Finally, we again apply Proposition 7.2 of [W2.
Our proof of Case 2 and of Theorem A is complete.
2. Horoball patterns in a product of a tree and a symmetric space In this section we will study the components of ${X}_{S}-{\Omega }_{S}$  when ${X}_{S}$  is a product of a symmetric space and a tree.
Setting notation. We let $w$  be a nonarchimedean valuation on $K$  such that ${\text{rank}}_{{K}_{w}}\left(\mathbf{G}\right)=1$  . Then we set $T$  equal to ${V}_{K}^{\infty }\cup \left\{w\right\}$  .
Horoballs in rank one symmetric spaces. Let $\mathbf{P}$  be a minimal $K$  -parabolic subgroup of $\mathbf{G}$  . As in the previous section, we let ${\delta }_{\mathbf{P}}$  be the simplex in the Tits boundary of ${X}_{{V}_{K}^{\infty }}$  corresponding to the group ${\prod }_{v\in {V}_{K}^{\infty }}\mathbf{P}\left({K}_{v}\right)$  .
Note that $\mathbf{G}$  being $K$  -isotropic and ${\text{rank}}_{{K}_{w}}\left(\mathbf{G}\right)=1$  together implies that ${\text{rank}}_{K}\left(\mathbf{G}\right)=1$  . From this latter equality, it is a well-known consequence of reduction theory that ${X}_{{V}_{K}^{\infty }}-{\Omega }_{{V}_{K}^{\infty }}$  is a disjoint collection of horoballs.
To any horoball of ${X}_{{V}_{K}^{\infty }}-{\Omega }_{{V}_{K}^{\infty }}$  , say $H$  , there corresponds a unique ${\delta }_{\mathbf{P}}$  as above such that any geodesic ray $\rho :{\mathbb{R}}_{\ge 0}\to {X}_{{V}_{K}^{\infty }}$  that limits to ${\delta }_{\mathbf{P}}$  defines an unbounded function $t↦d\left(\rho \left(t\right),{X}_{{V}_{K}^{\infty }}-H\right)$  $T$  -horoballs in ${X}_{T}$  . Let $y\in {X}_{w}$  and suppose $y\in \mathbf{G}\left({\mathcal{O}}_{T}\right){x}_{w}$  . Recall that by Lemma 1.3, the space ${\Omega }_{T}\cap \left({X}_{{V}_{K}^{\infty }}×\left\{y\right\}\right)$  is a finite Hausdorff distance from ${\Omega }_{{V}_{K}^{\infty }}×\left\{y\right\}$  .
For any minimal $K$  -parabolic subgroup of $\mathbf{G}$  , say $\mathbf{P}$  , we let ${\mathcal{ℋ}}_{\mathbf{P},\infty }^{y}\subseteq {X}_{{V}_{K}^{\infty }}×\left\{y\right\}$  be the horoball of ${\Omega }_{T}\cap \left({X}_{{V}_{K}^{\infty }}×\left\{y\right\}\right)$  that corresponds to ${\delta }_{\mathbf{P}}$  .
For arbitrary $x\in {X}_{w}$  , we define ${\mathcal{ℋ}}_{\mathbf{P},\infty }^{x}={\mathcal{ℋ}}_{\mathbf{P},\infty }^{y}$  where $y\in \mathbf{G}\left({\mathcal{O}}_{T}\right){x}_{w}$  minimizes the distance between $x$  and $\mathbf{G}\left({\mathcal{O}}_{T}\right){x}_{w}$  . We let ${\mathcal{ℋ}}_{\mathbf{P}}={\bigcup }_{x\in {X}_{w}}\left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{x}×\left\{x\right\}\right)$  Each of the spaces ${\mathcal{ℋ}}_{\mathbf{P}}$  is called a $T$  -horoball.
Let $\mathcal{P}$  be the set of all minimal $K$  -parabolic subgroups of $\mathbf{G}$  . The following lemma follows directly from our definitions. It will be used in the proof of Lemma 1.7.
Lemma 2.1 The Hausdorff distance between ${X}_{T}-{\Omega }_{T}$  and ${\bigcup }_{\mathbf{P}\in \mathcal{P}}{\mathcal{ℋ}}_{\mathbf{P}}$  is finite.
We record another observation to be used in the proof of Lemma 1.7.
Lemma 2.2 If $\mathbf{P}$  and $\mathbf{Q}$  are minimal $K$  -parabolic subgroups of $\mathbf{G}$  , then ${\mathcal{ℋ}}_{\mathbf{P}}\cap {\mathcal{ℋ}}_{\mathbf{Q}}\ne \varnothing$  .
Proof. The horoballs comprising $\left({X}_{{V}_{K}^{\infty }}-{\Omega }_{{V}_{K}^{\infty }}\right)×\left\{{x}_{w}\right\}$  are pairwise disjoint, and are a finite Hausdorff distance from the horoballs of ${\Omega }_{T}\cap \left({X}_{{V}_{K}^{\infty }}×\left\{{x}_{w}\right\}\right)$  by Lemma 1.2. Hence, if $y=g{x}_{w}$  for some $g\in \mathbf{G}\left({\mathcal{O}}_{T}\right)$  , then the horoballs determined by ${\Omega }_{T}\cap \left({X}_{{V}_{K}^{\infty }}×\left\{y\right\}\right)=g\left[{\Omega }_{T}\cap \left({X}_{{V}_{K}^{\infty }}×\left\{{x}_{w}\right\}\right)\right]$  are disjoint.
$■$  Deformations of horoballs over geodesics in ${X}_{w}$  . We let $\pi :{X}_{T}\to {X}_{{V}_{K}^{\infty }}$  be the projection map. Note that if $x\in {X}_{w}$  and $\mathbf{P}$  is a minimal $K$  -parabolic subgroup of $\mathbf{G}$  , then $\pi \left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{x}\right)$  is a horoball in ${X}_{{V}_{K}^{\infty }}$  that is based at ${\delta }_{\mathbf{P}}$  . Recall that for any minimal ${K}_{w}$  -parabolic subgroup of $\mathbf{G}$  , say $\mathbf{Q}$  , we denote the point in the boundary of the tree ${X}_{w}$  that corresponds to $\mathbf{Q}\left({K}_{w}\right)$  by ${\varepsilon }_{\mathbf{Q}}$  .
Lemma 2.3 Suppose $\mathbf{P}$  is a minimal $K$  -parabolic subgroup of $\mathbf{G}$  and that $\mathbf{Q}$  is a minimal ${K}_{w}$  -parabolic subgroup of $\mathbf{G}$  . If $\gamma :\mathbb{R}\to {X}_{w}$  is a geodesic with $\gamma \left(\infty \right)={\varepsilon }_{\mathbf{P}}$  and $\gamma \left(-\infty \right)={\varepsilon }_{\mathbf{Q}}$  then
• (i) $s\le t$  implies $\pi \left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{\gamma \left(s\right)}\right)\subseteq \pi \left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{\gamma \left(t\right)}\right)$  (ii) ${\cup }_{t\in \mathbb{R}}\pi \left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{\gamma \left(t\right)}\right)={X}_{{V}_{K}^{\infty }}$  (iii) ${\cap }_{t\in \mathbb{R}}\left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{\gamma \left(t\right)}\right)=\varnothing$  (iv) There exists constants ${L}_{\mathbf{P}},C>0$  such that if $h\left(s,t\right)=d\left(\pi \left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{\gamma \left(s\right)}\right),\pi \left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{\gamma \left(t\right)}\right)\right)$  then $|h\left(s,t\right)-{L}_{\mathbf{P}}|s-t||\le C$
Proof. As the ends of ${X}_{w}$  corresponding to $K$  -parabolic subgroups are a dense subset of the full space of ends, it suffices to prove this lemma when $\mathbf{Q}$  is defined over $K$  . In this case, the image of $\gamma$  corresponds to a $K$  -split torus $\mathbf{S}\le \mathbf{G}$  that is contained in $\mathbf{P}$  .
Let $\alpha$  be a root of $\mathbf{G}$  with respect to $\mathbf{S}$  such that $\alpha$  is positive in $\mathbf{P}$  . Since the diagonal embedding of $\mathbf{S}\left({\mathcal{O}}_{T}\right)$  in the group ${\prod }_{v\in {V}_{K}^{\infty }}\mathbf{S}\left({K}_{v}\right)$  has a dense image, there is some $b\in \mathbf{S}\left({\mathcal{O}}_{T}\right)$  such that $|\alpha \left(\mathbf{S}\left(b\right)\right){|}_{v}<1$  for all $v\in {V}_{K}^{\infty }$  .
Thus, $\mathbf{S}\left(b\right)\pi \left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{\gamma \left(0\right)}\right)$  is a horoball in ${X}_{{V}_{K}^{\infty }}$  that strictly contains $\pi \left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{\gamma \left(0\right)}\right)$  .
Generally, we have $\mathbf{S}\left(b{\right)}^{m}\pi \left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{\gamma \left(0\right)}\right)\mathbb{⊊}\mathbf{S}\left(b{\right)}^{n}\pi \left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{\gamma \left(0\right)}\right)$  for all $m,n\in \mathbb{Z}$  with $m  .
By the product formula we have $|\alpha \left(\mathbf{S}\left(b\right)\right){|}_{w}>1$  . Thus, there is a positive number $\lambda >0$  such that $\gamma \left(n\lambda \right)=\mathbf{S}\left(b{\right)}^{n}\gamma \left(0\right)$  for any $n\in \mathbb{Z}$  . It follows for $m  that $\pi \left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{\gamma \left(m\lambda \right)}\right)=\mathbf{S}\left(b{\right)}^{m}\pi \left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{\gamma \left(0\right)}\right)\mathbb{⊊}\mathbf{S}\left(b{\right)}^{n}\pi \left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{\gamma \left(0\right)}\right)=\pi \left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{\gamma \left(n\lambda \right)}\right)$  We let ${L}_{\mathbf{P}}=\frac{1}{\lambda }d\left(\pi \left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{\gamma \left(0\right)}\right),\pi \left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{\gamma \left(\lambda \right)}\right)\right)$  so that $h\left(m\lambda ,n\lambda \right)={L}_{\mathbf{P}}\lambda |m-n|={L}_{\mathbf{P}}|m\lambda -n\lambda |$  Then we take ${C}^{\prime }={max}_{0\le s\le t\le \lambda }\left\{d\left(\pi \left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{\gamma \left(s\right)}\right),\pi \left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{\gamma \left(t\right)}\right)\right)\right\}$  and, say, $C=2{C}^{\prime }+{L}_{\mathbf{P}}d\left(\gamma \left(0\right),\gamma \left(\lambda \right)\right)$  $■$  Basepoints in the Tits boundary for $T$  -horoballs. The Tits boundary for $X$  is the spherical join of the Tits boundary for the symmetric space ${X}_{{V}_{K}^{\infty }}$  and the Tits boundary for the tree ${X}_{w}$  .
The purpose of the following two lemmas—and of this entire section—is to show that each $T$  -horoball ${\mathcal{ℋ}}_{\mathbf{P}}$  is geometrically associated with the join of ${\delta }_{\mathbf{P}}$  and ${\varepsilon }_{\mathbf{P}}$  , denoted ${\delta }_{\mathbf{P}}*{\varepsilon }_{\mathbf{P}}$  . Lemma 2.4 Let $\mathbf{P}$  be a minimal $K$  -parabolic subgroup of $\mathbf{G}$  . Any geodesic ray $\rho :{\mathbb{R}}_{\ge 0}\to {X}_{T}$  that limits to the simplex ${\delta }_{\mathbf{P}}*{\varepsilon }_{\mathbf{P}}$  in the Tits boundary of ${X}_{T}$  defines an unbounded function when composed with the distance from the complement of ${\mathcal{ℋ}}_{\mathbf{P}}$  in ${X}_{T}$  :
$t↦d\left(\rho \left(t\right),{X}_{T}-{\mathcal{ℋ}}_{\mathbf{P}}\right)$  Proof. Any such geodesic ray $\rho$  is a product of a geodesic ray $b:{\mathbb{R}}_{\ge 0}\to {X}_{{V}_{K}^{\infty }}$  that limits to ${\delta }_{\mathbf{P}}$  and a geodesic ray $c:{\mathbb{R}}_{\ge 0}\to {X}_{w}$  that limits to ${\varepsilon }_{\mathbf{P}}$  .
Let $Y=\pi \left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{c\left(0\right)}\right)×c\left({\mathbb{R}}_{\ge 0}\right)$  . Since $t↦d\left(b\left(t\right),{X}_{{V}_{K}^{\infty }}-\pi \left({\mathcal{ℋ}}_{\mathbf{P},\infty }^{c\left(0\right)}\right)\right)$  is unbounded, $t↦d\left(\rho \left(t\right),{X}_{T}-Y\right)$  is unbounded. The lemma follows from Lemma 2.3 which guarantees that $Y\subseteq {\mathcal{ℋ}}_{\mathbf{P}}$  .
$■$  Lemma 2.5 Suppose $\mathbf{Q}$  and $\mathbf{M}$  are minimal $K$  -parabolic subgroups of $\mathbf{G}$  , and that ${\mathbf{Q}}^{\prime }$  is a minimal ${K}_{w}$  -parabolic subgroup of $\mathbf{G}$  with $\mathbf{M}\ne \mathbf{Q}$  or $\mathbf{M}\ne {\mathbf{Q}}^{\prime }$  .
Then there is a geodesic ray $\rho :{\mathbb{R}}_{\ge 0}\to X$  with $\rho \left(\infty \right)\in {\delta }_{\mathbf{Q}}*{\varepsilon }_{{\mathbf{Q}}^{\prime }}$  such that the function $t↦d\left(\rho \left(t\right),{X}_{T}-{\mathcal{ℋ}}_{\mathbf{M}}\right)$  is bounded. Proof. Choose a geodesic ray $b:{\mathbb{R}}_{\ge 0}\to {X}_{{V}_{K}^{\infty }}$  that limits to ${\delta }_{\mathbf{Q}}$  and a geodesic ray $c:{\mathbb{R}}_{\ge 0}\to {X}_{w}$  that limits to ${\varepsilon }_{{\mathbf{Q}}^{\prime }}$  . Let $r$  be the ratio of the speed of $b$  to the speed of $c$  .
If $\mathbf{M}\ne {\mathbf{Q}}^{\prime }$  , then after ignoring at most a bounded interval of $c$  , we can extend $c$  to a bi-infinite geodesic with $c\left(-\infty \right)={\varepsilon }_{\mathbf{M}}$  . With ${L}_{\mathbf{M}}$  as in Lemma 2.3, $\rho \left(t\right)=\left(b\left({L}_{\mathbf{M}}t\right),c\left(rt\right)\right)$  defines a geodesic ray satisfying the lemma.
In the remaining case, $\mathbf{M}\ne \mathbf{Q}$  and $\mathbf{M}={\mathbf{Q}}^{\prime }$  .
The distance from $b\left(t\right)$  to $\pi \left({\mathcal{ℋ}}_{\mathbf{M},\infty }^{b\left(0\right)}\right)$  is a convex function in $t$  . Since $\mathbf{M}\ne \mathbf{Q}$  , this function has a positive derivative, $u>0$  , for some large value of $t$  . Then $\rho \left(t\right)=\left(b\left({L}_{\mathbf{M}}t\right),c\left(urt\right)\right)$  defines a geodesic ray satisfying the lemma.
$■$  References

1. Eskin, A., Quasi-isometric rigidity of nonuniform lattices in higher rank symmetric spaces. J. Amer. Math. Soc., 11 (1998), 321-361.
2. Eskin, A., and Farb, B., Quasi-flats and rigidity in higher rank symmetric spaces. J. Amer. Math. Soc., 10 (1997), 653-692.
3. Farb, B., The quasi-isometry classification of lattices in semisimple Lie groups. Math. Res. Letta., 4 (1997), 705-717.
4. Farb, B., and Schwartz, R., The large-scale geometry of Hilbert modular groups. J. Diff. Geom., 44 (1996), 435-478.
5. Kleiner, B., Leeb, B., Rigidity of quasi-isometries for symmetric spaces and Euclidean buildings. Inst. Hautes Études Sci. Publ. Math., 86 (1997), 115-197.
6. Lubotzky, A., Mozes, S., and Raghunathan, M. S., The word and Riemannian metrics on lattices of semisimple groups. Inst. Hautes Études Sci. Publ. Math., 91 (2000), 5-53.
7. Schwartz, R., The quasi-isometry classification of rank one lattices. Inst. Hautes Études Sci. Publ. Math., 82 (1995), 133-168.
8. Schwartz, R., Quasi-isometric rigidity and Diophantine approximation. Acta Math., 177 (1996), 75-112.
9. Taback, J., Quasi-isometric rigidity for $PS{L}_{2}\left(\mathbb{Z}\left[1/p\right]\right)$  . Duke Math. J., 101 (2000), 335-357.
10. Tits, J., Classification of algebraic semisimple groups. 1966 Algebraic Groups and Discontinuous Subgroups (Proc. Sympos. Pure Math., Boulder, Colo., 1965) p. 33-62 Amer. Math. Soc., Providence, R.I., 1966.
11. Wortman, K., Quasiflats with holes in reductive groups. Preprint.
12. Wortman, K., Quasi-isometric rigidity of higher rank $S$  -arithmetic lattices. Preprint.

Kevin Wortman Department of Mathematics Cornell University Malott Hall Ithaca, NY 14853 Email: wortman@math.cornell.edu