Adv. Geom.

6 (2006), 475–491 Advances in Geometry

DOI 10.1515/ADVGEOM.2006.029 6 de Gruyter 2006

Entropy of the geodesic flow for metric spaces and

Bruhat–Tits buildings
Enrico Leuzinger
(Communicated by T. Grundhöfer)

Abstract. Let ðX ; dX Þ be a geodesically complete Hadamard space endowed with a Borel-

measure m. Assume that there exists a group G of isometries of X which acts totally discontin-
uously and cocompactly on X and preserves m. We show that the topological entropy of the
geodesic flow on the space of (parametrized) geodesics of the compact quotient GnX is equal
to the volume entropy of m (if X satisfies a certain local uniformity condition). This extends a
result of Manning for riemannian manifolds of nonpositive curvature to the singular case. The
result in particular holds for Bruhat–Tits buildings, for which we also compute the entropy
explicitly.
Key words. topological entropy, geodesic flow, Hadamard spaces, Bruhat–Tits buildings, dis-
crete subgroups of Lie groups, p-adic groups.
2000 Mathematics Subject Classification. 11E95, 20E42, 22E40, 37B40, 51E24, 53D25

1 Introduction
The volume growth rate of a closed riemannian manifold ðM; gÞ with associated mea-
sure dvg is defined as

1
h vol ðdvg Þ :¼ lim sup log VolðBðx; rÞÞ;
r!y r

where Bðx; rÞ is the ball of radius r around a fixed point x in the universal covering
space of M. Remarkably, this rather coarse asymptotic invariant carries a lot of geo-
metric information. In fact, it is linked via inequalities to several important geometric
quantities like the simplicial volume, the bottom of the spectrum of the laplacian, the
Cheeger isoperimetric constant and the growth of the fundamental group (see [14], [7]
and [24], respectively). We also mention the entropy rigidity conjecture, which,
roughly, posits that if M carries a locally symmetric metric g0 , then h vol ðdvg0 Þ mini-
mizes h vol among all riemannian metrics on M with the same volume (see e.g. [2],
[11]). The quantity most closely related to the volume growth rate, however, is the
476 Enrico Leuzinger

topological entropy of the geodesic flow. Given a continuous flow j on a compact met-
ric space the topological entropy of j, h top ðjÞ, is an asymptotic invariant which mea-
sures the orbit complexity of the flow by a single number (see Section 2.3 for the def-
inition). Positivity of the topological entropy indicates ‘‘chaotic’’ behaviour.
Dinaburg and Manning proved in [12] and [22] that the topological entropy of
the geodesic flow j on (the unit tangent bundle of ) a compact riemannian manifold
ðM; gÞ is greater than or equal to the volume growth rate: h top ðjÞ d h vol ðdvg Þ. More-
over, equality holds if ðM; gÞ has nonpositive sectional curvature. The volume
growth rate h vol ðdvg Þ is therefore also called volume entropy of ðM; gÞ. The equality
result, h top ðjÞ ¼ h vol ðdvg Þ, was extended in [13] to manifolds without conjugate
points. Equality in particular holds for locally symmetric spaces of noncompact
type. For these manifolds it is possible to compute the volume growth rate (and hence
the topological entropy) explicitly: It equals 2krk, where r is the half sum of positive
roots counted with multiplicity (see [18]).
The geodesic flow (and its entropy) can also be defined in the more singular context
of metric spaces admitting compact quotients. In the present paper we are mainly in-
terested in Hadamard spaces, i.e., complete, simply connected geodesic metric spaces
of nonpositive curvature in the sense of Alexandrov (see Section 2 and also [1], [6] for
this nowadays well established notion). Even more generally we shall consider metric
measure spaces, ðX ; dX ; mÞ, where ðX ; dX Þ is a geodesic space and m a s-additive
Borel-measure on X .

Assumptions: Throughout this text we will assume that ðX ; dX Þ is geodesically com-

plete, i.e., every geodesic segment is the restriction of a geodesic line defined on all
of R, and that X is locally uniquely geodesic (see Section 2). In addition we will also
need the following local uniformity and global convexity conditions.

Property ( U): We say that a metric measure space ðX ; dX ; mÞ has property ðUÞ if
there is 0 < d0 c y such that for every 0 < d < d0 there are positive constants
Ci ðdÞ ði ¼ 1; 2Þ such that

0 < C1 ðdÞ ¼ inf mðBðx; dÞÞ c sup mðBðx; dÞÞ ¼ C2 ðdÞ < y;
xAX xAX

where Bðx; rÞ is the ball of center x and radius r in X .

Property (C): We say that a geodesically complete metric space ðX ; dX Þ has property
ðCÞ if for any two geodesics c1 : R ! X , c2 : R ! X (parametrized by arc-length)
the function t 7! dX ðc1 ðtÞ; c2 ðtÞÞ is convex.

Basic examples of spaces which satisfy property (C) are geodesically complete Ha-
damard spaces (see [1], I.5.4, [6], II.2.2). Note that (C) implies that X is geodesically
unique: between any two points of X there is a unique geodesic segment.
The geodesic flow on a compact metric space is defined by reparametrization of
arc-length parametrized geodesics (see Section 3.1). The main result of this paper is
 Entropy of the geodesic flow for metric spaces and Bruhat–Tits buildings 477

an extension of the Dinaburg–Manning result to metric measure spaces which have

properties (C) and (U) and which admit compact quotients. The proof is essentially
an adaption of Mannings’s arguments. However in the singular case various techni-
cal details are not obvious. The main di‰culty which occurs is the branching of geo-
desics: in contrast to the smooth case a geodesic is no longer determined by an initial
point and an initial direction.
An important class of examples of singular spaces with properties (C) and (U)
are the euclidean or a‰ne buildings of Bruhat and Tits associated to semisimple al-
gebraic groups defined over non-archimedean local fields. These metric spaces are lo-
cally finite poly-simplicial complexes which are often viewed as p-adic analogues of
globally symmetric spaces. In fact, our results below support that philosophy. The
existence of compact quotients is guaranteed by [4] at least in characteristic zero.
For such buildings we also compute the entropy explicitely in terms of root data
(for the definitions of the latter see Section 3; compare also [26]). The chief result of
the present paper is the following

Main Theorem. (a) Let ðX ; dX ; mÞ be a geodesic metric space which is geodesically

complete, locally uniquely geodesic and endowed with a Borel-measure m. Assume that
X has property ðUÞ and that there exists a group G of isometries of X that acts totally
discontinuously and cocompactly on X and preserves m. Then the topological entropy of
the geodesic flow j on the space of geodesics of the compact quotient GnX is bounded
from below by the volume entropy with respect to m:

h top ðjÞ d h vol ðmÞ:

(b) Let ðX ; dX ; mÞ, G and j be as in (a) and assume in addition that ðX ; dX Þ has prop-
erty ðCÞ, then equality holds:

h top ðjÞ ¼ h vol ðmÞ:

(c) Let ðB; dB Þ be the Bruhat–Tits building associated to a connected, simply con-
nected, semisimple linear algebraic group G defined over a non-archimedean local field
F of F-rank r. Let dv be the r-dimensional Hausdor¤–Lebesgue measure associated to
the metric dB and denote by 2r the sum of the positive roots of G with respect to some
Weyl chamber in a maximal F-split torus. Let G be a discrete cocompact subgroup of
the group G ¼ GðFÞ of F-rational points of G. Finally let j be the geodesic flow on the
space of geodesics of the finite polyhedron GnB. Then

h top ðjÞ ¼ h vol ðdvÞ ¼ 2krk:

We remark that in Part (c) of the above Main Theorem one has to normalize appro-
priately the logarithm used to define the entropy.
The plan of the paper is as follows. In Section 2 we prove Parts (a) and (b) of the
Main Theorem. We closely follows Manning’s arguments in [22] with appropriate
modifications taking into account the singular geometry. In Section 3 we discuss
478 Enrico Leuzinger

some relevant facts about Bruhat–Tits theory and estimate the volume of balls in
buildings (with respect to the Hausdor¤ measure induced by the building metric).
These estimates rely on the structure theory of linear algebraic groups defined over
local fields. As an application we then prove Part (c) of the Main Theorem. In Sec-
tion 4 we discuss some connections with the growth of fundamental groups.

2 Geodesic flow and entropy for metric measure spaces

General references for this section are [1], [6], [10] and [17]. Let ðX ; dX Þ be a com-
plete metric space. A geodesic segment in X is a locally distance minimizing curve
c : I ! X , of an intervall I J R into X , parametrized by arc-length. If I ¼ R, a geo-
desic segment is called a geodesic (line). The space X is geodesically complete if every
geodesic segment is the restriction of a (arc-length parametrized) geodesic defined on
R (see [10], 2.5). We say that X is geodesic if any two points x; y A X can be con-
nected by a geodesic segment of length dX ðx; yÞ.

2.1 Volume entropy of metric measure spaces. In the following we will use the loga-
rithm loga with (arbitrary) basis a > 1 instead of the usual natural logarithm. This
choice is justified by the formula in the Main Theorem (c) (and its proof ) where we
take a ¼ q, the order of the (finite) residue class field of the local field F. Classically,
for manifolds, one takes a ¼ e.

Proposition 1. Let ðX ; dX ; mÞ be a complete (noncompact) geodesic metric measure

space which is locally uniquely geodesic and satisfies property (U ) of Section 1 and
assume that there exists a group G of isometries that acts totally discontinuously and
cocompactly on X and preserves m. For any x A X let Bðx; rÞ be the ball of center x
and radius r in X . Then the limit

1
h vol ðmÞ :¼ lim loga mðBðx; rÞÞ
r!y r

exists and is independent of x.

Proof. Since X is locally uniquely geodesic, X nG has positive injectivity radius (see
[6], I.7.53). In view of property (U) and since G is cocompact, the proof is then
word-for-word the same as Manning’s proof in the case of manifolds (see [22]).

The limit h vol ðmÞ is called the volume entropy of the metric measure space ðX ; dX ; mÞ
(or of any of its compact quotients GnX ).

2.2 The geodesic flow for metric spaces and topological entropy. Let ðX ; dX Þ be a
geodesic metric space which is geodesically complete and locally uniquely geodesic.
Let G be a group which acts totally discontinuously and isometrically on X with
compact quotient GnX . Let d be the metric on GnX induced by dX . Since X is locally
isometric to GnX the latter is also geodesically complete. By GðX Þ and GðGnX Þ we
 Entropy of the geodesic flow for metric spaces and Bruhat–Tits buildings 479

denote the set of all geodesics (defined on R) of X and of the compact quotient GnX ,
respectively. We then define the geodesic flow j~ :¼ fj~s j s A Rg on GðX Þ by repara-
metrization, i.e. by,

j~s : GðX Þ ! GðX Þ; c 7! j~s ðcÞ ðs A RÞ

where the parametrized geodesic j~s ðcÞ : R ! X is given by j~s ðcÞðtÞ :¼ cðs þ tÞ. A
geodesic in GnX is the image of a geodesic in X under the canonical projection
p : X ! GnX and the geodesic flow on GnX is given by j ¼ p
j~. On the set
GðGnX Þ we define a metric d G by
ðy
ejtj
d G ðc1 ; c2 Þ :¼ dðc1 ðtÞ; c2 ðtÞÞ dt
y 2

(compare [15], 8.3 B).

The space of geodesics of a compact manifold is compact. This also holds in the
present situation.

Lemma 1. The metric space ðGðGnX Þ; d G Þ is compact.

Proof. Let c1 ; c2 A GðGnX Þ. A simple computation using twice the triangle inequality
and the definition of d G yields for any T d 0:

d G ðc1 ; c2 Þ c eT þ sup dðc1 ðtÞ; c2 ðtÞÞ:

t A ½T; T

The claim then follows from Arzelà–Ascoli (see [6]).

The topological entropy h top is defined for a continuous flow on a compact metric
space (see [17] and, for the geodesic flow on manifolds, also [25]). We are going to
define that notion for the geodesic flow j on the metric space ðGðGnX Þ; d G Þ (which
is compact by Lemma 1).
For any real number T > 0 define a new metric d G; T on GðGnX Þ by

d G; T ðc1 ; c2 Þ :¼ max d G ðjs ðc1 Þ; js ðc2 ÞÞ:

0cscT

A subset A H GðGnX Þ is called ðr; dÞ-separated if for any two di¤erent points c; c 0 A A
one has d G; r ðc; c 0 Þ d d. Let maxðr; dÞ be the maximal cardinality of an ðr; dÞ-separated
subset of GðGnX Þ. Then the toplogical entropy of the geodesic flow j is the following
limit:

1
h top ðjÞ :¼ lim lim sup loga maxðr; dÞ:
d!0 r!y r

Equivalently, h top ðjÞ can also be defined as follows (see [17], 3.1). A subset B H
GðGnX Þ is called ðr; dÞ-spanning if for all b A GðGnX Þ there is b 0 A B such that
480 Enrico Leuzinger

d G; r ðb; b 0 Þ c d. Let minðr; dÞ be the minimal cardinality of an ðr; dÞ-spanning subset of

GðGnX Þ. Then the toplogical entropy of the geodesic flow j can also be obtained as

1
h top ðjÞ ¼ lim lim sup loga minðr; dÞ:
d!0 r!y r

2.3 Proof of the Main Theorem, Part (a) and (b). Proof of Part (a): We wish to
estimate the cardinality of certain separated sets. To that end we pick e > 0 and a
basepoint x A X . By Proposition 1 there exists r0 ¼ r0 ðeÞ such that for (large) r > r0
we have mðBðx; rÞÞ d aðh vol eÞr . Since X is locally uniquely geodesic the injectivity
radius InjðGnX Þ of the compact space GnX is positive (see [6], I.7.53, I.7.55). Take
0 < d < InjðGnX Þ and assume also that 2d < d0 of property (U). The above measure
estimate then implies that there exists an increasing and divergent sequence of radii
ðrk Þk A N such that

mðBðx; rk þ d=2Þ  Bðx; rk ÞÞ d aðh vol eÞrk :

Let Yrk be a maximal subset of the shell Bðx; rk þ d=2Þ  Bðx; rk Þ H X whose points
are pairwise 2d apart. Next note that by property (U) we have

sup mðBðp; 2dÞÞ ¼ C2 ð2dÞ:

pAX

Hence the cardinality of Yrk satisfies

jYrk j d C21 aðh vol eÞrk : ð1Þ

We now set d  :¼ 14 dð1  ed=4 Þ (note that d  is a strictly monotone function of d ).

We fix some rk and wish to construct an ðrk ; d  Þ separated set in GðGnX Þ. For sim-
plicity we write r for rk . By assumption X is geodesic, i.e., any two points in X can
be joined by a minimal geodesic segment. For each point in Yrk we can therefore
choose a geodesic segment joining it to x. Since X is assumed to be geodesically com-
plete there is, for each such segment, a complete geodesic line, i.e., an element of
GðX Þ, extending it. Let A~ðrÞ H GðX Þ be the set of these geodesics and, for the projec-
tion p : GðX Þ ! GðGnX Þ, set AðrÞ :¼ pðA~ðrÞÞ. We want to show that AðrÞ is ðr; d  Þ-
separated. Pick c~xq ; c~xq 0 A A~ðrÞ. We then have

cxq ðrÞ; c~xq 0 ðrÞÞ d dX ððq; q 0 Þ  dX ð~

dX ð~ cxq 0 ðrÞ; q 0 Þ
cxq ðrÞ; qÞ  dX ð~
> 2d  d=2  d=2 ¼ d: ð2Þ

Next consider the two geodesics c :¼ p
c~xq and c 0 :¼ p
c~xq 0 in GðGnX Þ. As

d < InjðGnX Þ, the estimate (2) implies that there is s A ½0; r such that dðcðsÞ; c 0 ðsÞÞ >
d. But then
 Entropy of the geodesic flow for metric spaces and Bruhat–Tits buildings 481

ðy
1
d G ðjs ðcÞ; js ðc 0 ÞÞ ¼ dðcðs þ tÞ; c 0 ðs þ tÞÞ ejtj dt
y 2
ð d=4
1
d ½dðcðsÞ; c 0 ðsÞÞ  2t et dt
0 2
ð d=4
1 t 1
> de dt ¼ dð1  ed=4 Þ ¼ d  :
0 4 4

This shows that d G; r ðc; c 0 Þ :¼ max 0ctcr d G ðjt ðcÞ; jt ðc 0 ÞÞ > d  , i.e., that AðrÞ H
GðGnX Þ is an ðr; d  Þ-separated set for the geodesic flow on the compact metric space
ðGðGnX Þ; d G Þ.
By (2) the cardinality of AðrÞ is equal to the cardinality of A~ðrÞ. The estimate (1)
then yields that the maximal cardinality of an ðrk ; d  Þ-separated subset is bounded
from below by C2 ð2dÞ1 aðh vol eÞrk (recall that we set r ¼ rk ). Hence, the definition of
h top yields h top d h vol  e. Finally, as e was arbitrarily chosen, we obtain h top ðjÞ d
h vol ðmÞ, which proves (a).
Proof of Part (b): In view of (a) we need to prove the converse inequality
h top ðjÞ c h vol ðmÞ. We again fix a basepoint x A X and consider a compact fundamen-
tal domain F of G in X which contains x. Let D be the diameter of F and as before
let d < InjðGnX Þ and also d < d0 of property (U). We also fix some integer n d 3 and
set r1 :¼ n ln d. Proposition 1 asserts that for given e > 0 there is r2 ¼ r2 ðeÞ such that
for all r > r2 we have

mðBðx; rÞÞ c aðh vol þeÞr : ð3Þ

Let ZðrÞ be a maximal d-separated subset of SðrÞ :¼ fz A X j r  D c

dX ðz; FÞ c rg (i.e., for any u, v with u 0 v holds dX ðu; vÞ > d). By property (U) and
(3) there is a constant 0 < C1 ðd=2Þ such that for all (large) r d r0 :¼ maxfr1 ; r2 g the
following estimate for the cardinality of ZðrÞ holds

jZðrÞj c C11 mðBðx; r þ D þ d=2ÞÞ c C11 aðh vol þeÞðrþDþd=2Þ : ð4Þ

Let F be a maximal d-separated subset of the fundamental domain F. Since X

is geodesic, property (C) implies that between any y A F and z A ZðrÞ there exists
a unique geodesic segment (parametrized by arc-length) of length between r  D
and r þ D. Since X is assumed to be geodesically complete, there exists a (possibly
non-unique) extension of such a segment to a geodesic line. For each y
A F and
z A ZðrÞ we can thus choose a geodesic line c~yz : R ! X with y ¼ c~yz  nr . Set
cyz j y A F ; z A ZðrÞg and let BðrÞ :¼ pðB~ðrÞÞ where p : GðX Þ ! GðGnX Þ is
B~ðrÞ :¼ f~
the canonical projection


mapping
  geodesics in X to geodesics in GnX . We want to
show that BðrÞ is a r 1  2n ; 6d -spanning set for the geodesic flow on GðGnX Þ.
To that end consider a geodesic segment in GnX of length r and choose a segment
from u A F to v in X which covers it (compare [10], Lemma 3.4.17). Then there are
points y A F and z A ZðrÞ with dX ðu; yÞ c d and dX ðv; zÞ c d. We also have r  2d c
dX ðy; zÞ c r þ 2d. Let c~1 be a geodesic in X which extends the segment between u and
482 Enrico Leuzinger



v and set
c~2:¼ c~yz A B~ðrÞ as above, such that u ¼ c~1  nr and y ¼ c~2  nr . Let
z 0 :¼ c~2 r  nr be the point at distance r from y along the segment of c~2 from y to z.
By assumption (property (C)) the function t 7! dX ð~ c1 ðtÞ; c~2 ðtÞÞ is convex. Thus we
r r
have for all t A  n ; r  n ,
         
r r r r
dX ð~
c1 ðtÞ; c~2 ðtÞÞ c dX c~1  ; c~2  þ dX c~1 r  ; c~2 r 
n n n n
¼ dX ðu; yÞ þ dX ðv; z 0 Þ:

Since

dX ðv; z 0 Þ c dX ðv; zÞ þ dX ðz; z 0 Þ c d þ 2d ¼ 3d;

we find that the projections ci ¼ p
c~i ði ¼ 1; 2Þ in GnX satisfy

 
r r
c1 ðtÞ; c~2 ðtÞÞ c 4d for all t A  ; r  :
dðc1 ðtÞ; c2 ðtÞÞ c dX ð~ ð5Þ
n n

We next estimate d G ðjs ðc1 Þ; js ðc2 ÞÞ. Using (5) and—for arbitrary t1 , t2 —the triangle
inequality

dðc1 ðt1 Þ; c2 ðt1 ÞÞ c dðc1 ðt2 Þ; c2 ðt2 ÞÞ þ 2jt2  t1 j;

 
we compute for s A 0; r  2n r :

d G ðjs ðc1 Þ; js ðc2 ÞÞ

ðy
1
¼ dðc1 ðs þ tÞ; c2 ðs þ tÞÞ ejtj dt
y 2
ð r=ns       
r r r 1 jtj
c d c1  ; c2  þ2 tþsþ e dt
y n n n 2
ð0 ð rr=ns
1 jtj 1
þ 4d e dt þ 4d et dt
r=ns 2 0 2
ðy       
r r r 1 t
þ d c1 r  ; c2 r  þ2 sþtrþ e dt
rr=ns n n n 2
ðy ðy   ð r=nþs
t r t
c 2d e dt þ t  s  e dt þ 2d et dt
r=nþs r=nþs n 0
ð rr=ns ðy ðy  
r
þ 2d et dt þ 2d et dt þ t þ s  r þ et dt
0 rr=ns rr=ns n

¼ 4d þ er=ns þ erþr=nþs c 4d þ 2er=n c 6d;

 Entropy of the geodesic flow for metric spaces and Bruhat–Tits buildings 483

where, for the last two inequalities, we used that 0 c s c r  2n r and that
r d r0 ðd; nÞ ¼ n ln d. In particular we have

d G; rð12=nÞ ðc1 ; c2 Þ ¼ sup d G ðjs ðc1 Þ; js ðc2 ÞÞ c 6d:

s A ½0; rð2=nÞr

 
c2 Þ A BðrÞ, we have shown that BðrÞ is a r 1  2n ; 6d -spanning set in
Since c2 ¼ pð~
GðGnBÞ. The estimate (4) now yields
  
2
min r 1  ; 6d c jBðrÞj c jB~ðrÞj c jZðrÞj c C1 ðdÞ1 aðh vol þeÞðrþDþd=2Þ :
n

Hence we get from the (second) definition of h top that

 1
2
h top ðjÞ c 1  ðh vol þ eÞ:
n

Finally we use that e and the integer n d 3 are arbitrary and get h top ðjÞ c h vol ðmÞ.
Together with Part (a) this completes the proof of Part (b) of the Main Theorem.

3 Basic examples: Bruhat–Tits buildings

The most important examples of metric spaces for which the results of Section 2 hold
are Bruhat–Tits buildings. The purpose of the present section is to review the relevant
details in order to verify that these buildings in fact have all the required properties.
We also compute the volume entropy explicitly.

3.1 Bruhat–Tits theory. General references are the paper of Bruhat–Tits [9] and the
books of Borel [3], Brown [8], Macdonald [21], Margulis [23] and Ronan [27].
Let F be a complete, locally compact, non-archimedean local field, i.e., a finite ex-
tension either of the the p-adic numbers Qp or of a formal power series field over a
finite field ([30], Theorems I.5, I.8). Let v : F ! Z be the discrete valuation of F ,
where F is the multiplicative group of non-zero elements of F. Let O ¼ fx A F j
vðxÞ d 0g be the ring of integers. Then P ¼ fx A F j vðxÞ d 1g is a maximal ideal in
O, and since both O and P are open and compact, the quotient O=P is a finite field,
the residue class field of F. Let q denote its cardinality, q :¼ jO=Pj. The ultrametric
absolute value of x A F is jxj :¼ qvðxÞ .
Let G be a connected, simply connected, semisimple linear algebraic group defined
over F and let G ¼ GðFÞ be the group of F-rational points of G. Then G is a locally
compact group. Let S H G be a maximal F-split torus, i.e., an algebraic subgroup
which is F-isomorphic to ðF Þ r for some r A N. Any two such tori are conjugate
and r is called the F-rank of G (or G). We will always assume that F-rank d 1. Write
N and Z respectively for the normalizer and centralizer of S in G, and set S :¼ SðFÞ,
N :¼ NðFÞ, Z :¼ ZðFÞ for the respective groups of F-rational points. Finally define
H :¼ fz A Z j jwðzÞj ¼ 1 for all characters w of Zg.
484 Enrico Leuzinger

Let X ðSÞ ¼ HomF ðS; F Þ be the set of characters of S. To a (multiplicative) root

a A X ðSÞ of the pair ðG; SÞ (see [23], 0.27) is associated a unique (additive) root a de-
~
fined by aðsÞ :¼ logq j~aðsÞj A Z ðs A SÞ. Any such a is an element of the real vector
space a  :¼ X ðSÞ nZ R of dimension r :¼ rankF G, and thus defines a linear form
on the dual R-vector space a. There is a unique continuous homomorphism n : Z !
a which satisfies aðnðsÞÞ ¼ vð~ aðsÞÞ, for each s A S and a~ A X ðSÞ. Moreover, one has
Ker n ¼ H and hence nðZÞ G Z=H (see [29], 1.2).
The group W :¼ N=H is the a‰ne Weyl group associated to a reduced root system
S of rank r ¼ rankF G (see [5], Ch. 6.2 for the definitions). The corresponding Weyl
group W0 is the Weyl group of the pair ðS; GÞ, i.e., W0 G N=Z. However, in general,
S is di¤erent from the relative root system F ¼ FðS; GÞ. Nevertheless every root of S
is proportional to some root of F and vice versa (see [29], 1.7). We now endow a with
a W0 -invariant inner product h ; i. Then the a‰ne Weyl group W acts as an a‰ne
Coxeter group, i.e., a group of isometries of the euclidean space a generated by reflec-
tions in a‰ne hyperplanes belonging to a W -invariant, locally finite set H of a‰ne
hyperplanes. The complex obtained by the partition of the space by the elements of
H is an a‰ne Coxeter complex. Its maximal cells are called chambers. Every codi-
mension 1 cell is contained in exactly two chambers.
In a group G as above there further exists a so-called Iwahori subgroup B for which
B V N ¼ H and such that ðB; NÞ is a BN-pair (or Tits system) for G (see e.g. [8] or
[21]). Associated to such a pair is an a‰ne Bruhat–Tits building B (see [9]). This is a
locally finite polysimplicial complex given as the product of the Bruhat–Tits build-
ings associated to the almost F-simple factors of G. If one fixes a basic chamber
C H B then one can choose B as the stabilizer of C in G.
We recall some basic properties of such buildings (see e.g. [6], [8], [9] and [21] for
details).
(a) The building B has a collection of subcomplexes called apartments. Every apart-
ment is an a‰ne Coxeter complex of dimension r ¼ rankF G whose chambers are
also maximal cells of B. The building is thick, i.e., every codimension 1 face is
contained in at least three chambers.
(b) One can identify a basic apartment A H B with the euclidean space ða; h ; iÞ
([21], 2.4.1.) and thus obtains a well-defined structure of euclidean space on
each apartment. The euclidean distance dA ðx; yÞ defined for x; y A A can be ex-
tended to a complete metric dB on B ([21], 2.4.8). Moreover ðB; dB Þ is a geodesi-
cally complete Hadamard space (see [6], II.5.10, II.10.A.4 and [8], VI.3). In par-
ticular such a building B has the convexity property (C) defined in Section 1.
(c) The group G acts isometrically and strongly transitively on B (see [8], V.3.) The
stabilizers of chambers are conjugates of B and stabilizers of vertices are maximal
compact and open subgroups of G (see [9]).
Under the identification of the vector space a with the apartment A the origin
0 A a corresponds to a base point x0 A A (which is also a vertex of B ). Let K be the
maximal parahoric subgroup K :¼ stabðx0 Þ of G. Then K is open and maximal com-
pact and can be written as K ¼ BW0 B ([21], 2.6.8).
 Entropy of the geodesic flow for metric spaces and Bruhat–Tits buildings 485

For each root a A a  there is a unique ~a A a such that aðnðsÞÞ ¼ h~

a; nðsÞi ðs A SÞ.
We also set a4 :¼ 2~a 2 . The a‰ne Weyl group W is a semidirect product W ¼ W0 T
k~
ak
where T is the abelian group generated by fa4 j a A Sg (see [5], VI, 2.1). With our
identifications a G A and 0 1 x0 we have—by some abuse of notation—for all
t A T, t 1 0 þ t 1 t  x0 . Furthermore, the kernel of the map n : Z ! a is H and hence
nðZÞ G Z=H G T.
Fix a basic Weyl chamber C0 G aþ H a. Let T þ consist of those t A T for which
t  x0 is in the Weyl chamber C0 . Put Z þ :¼ n1 ðT þ Þ. Note that for z1 ; z2 A Z with
nðz1 Þ ¼ nðz2 Þ we have z1 z1
2 A H and therefore Kz1 K ¼ Kz2 K. Thus, for any z A Z
the double coset KzK depends only on t ¼ nðzÞ and we denote it by Kn1 ðtÞK. With
that notation the following proposition is the p-adic version of a Cartan decomposi-
tion of G (see [9] and also [21], 2.6.11).

Proposition 2. For subgroups Z þ and K of the semisimple group G defined as above,

there is a Cartan decomposition G ¼ KZ þ K. Moreover there is a one-to-one correspon-
dence from T þ onto the set of double cosets KnG=K given by t 7! Kn1 ðtÞK.

3.2 Measures on buildings. Let o be the left-invariant Haar measure on the locally
compact group G normalized such that oðKÞ ¼ 1. Let dv be the r-dimensional
Hausdor¤–Lebesgue measure, r ¼ rankF G, associated to the building metric dB .
The dv-measure (or volume) of a dv-measurable subset A H B will be denoted by
V ðAÞ. We also denote by Bðx; rÞ the ball of radius r > 0 and center x in ðB; dB Þ.
We have the following elementary estimate for the measures of balls:

Lemma 2. Let x0 be the base point of B (with stabðx0 Þ ¼ K) and C H A a chamber

with vertex x0 . Set D :¼ DiamðK  CÞ, the diameter of the compact set K  C H B of
the finitely many chambers in B with common vertex x0 . Then one has for all r > 0

V ðBðx0 ; rÞÞ c V ðK  CÞoðBðx0 ; rÞ V G  x0 Þ c V ðBðx0 ; r þ DÞÞ:

The following lemma asserts that buildings have property (U).

1
Lemma 3. Let C H B be a chamber of the building B. Assume that d < 100 DiamðCÞ.
Then there are positive constants Ci ¼ Ci ðdÞ ði ¼ 1; 2Þ such that

0 < C1 ¼ inf V ðBðx; dÞÞ c sup V ðBðx; dÞÞ ¼ C2 :

xAB xAB

Proof. Since G acts strongly transitively on B and preserves the measure dv we can
assume that x A C. The size of d guarantees that there is at least one wall, say E, of C
with dB ðx; EÞ > d. Let e be the vertex of the simplex C opposite to E. Then let C 0 be
the a‰ne translate of C (in the vector space a) from e to x. By construction the inter-
section of the simplex C 0 with the ball Bðx; dÞ is completely contained in C V Bðx; dÞ.
Since C has only finitely many vertices the volume of that intersection is uniformly
bounded from below. This argument also yields an upper bound since the building
486 Enrico Leuzinger

is locally finite (there is m A N such that any x A B is contained in at most m

chambers).

3.3 The volume entropy of buildings. The goal of this section is to compute the vol-
ume entropy for Bruhat–Tits buildings. It turns out that the value for the volume en-
tropy (see Lemma 4 below) is completely analogous to the case of symmetric spaces
mentioned in the introduction (compare also [26]).
Recall from Section 3.1 that the translational part T of the a‰ne Weyl group W is
isomorphic
P to the lattice in a spanned by fa4 j a A Sg and that T G T  x0 H A. Let
2r ¼ a A Fþ ðS; GÞ ma a be the sum of the positive roots (with multiplicities) of G with
respect to some Weyl chamber in a maximal F-split torus S. Then, again by 3.2, there
is t > 0 such that x1 :¼ tr4 A T. Note also that the inner product on a defines an in-
ner product on a  and we have ha; bi ¼ h~ a; ~
b i for all roots a, b.

Lemma 4. Let F be a non-archimedean local field whose residue class field has finite
order q. Let G be a connected, simply connected, semisimple linear algebraic group
defined over F with associated Bruhat–Tits building B. Let K be the stabilizer of the
base point x0 A B. The Haar measure on G (normalized such that K has measure one)
induces a left-invariant measure o on G=K. For k A N, set rk :¼ 2ktkrk1 . Then

1
lim logq oðBðx0 ; rk Þ V G  x0 Þ ¼ 2krk:
rk !y rk

Proof. For t ¼ nðzÞ A nðZþ Þ the point x :¼ t  x0 is in the apartment A. We want

to determine the measure of the K-orbit of x, i.e., oðK  xÞ ¼ jK  xj (recall that
oðKÞ ¼ 1). To that end we will use the Cartan decomposition of G (see Proposition
2). Set L :¼ t1 Kt V K; then the map K=L ! Kt1 K=K given by hL 7! ht1 K is a bi-
jection and hence we have

K  x G K=ðt1 Kt V KÞ G Kt1 K=K:

Next by [21], 2.3.5, we have

KtK ¼ BW0 tW0 B ¼ Kðwtw1 ÞK

for any w A W0 . Thus, for any t A T and w A W0 , we get for the index of K in KtK:

jKtK : Kj ¼ jKwtw1 K : Kj:

For t A T þ that index has been computed in [21], 3.2.15. There is a uniformly
bounded rational function Rt ðqÞ, i.e., there are constants 0 < C3 < C4 such that
C3 c Rt ðqÞ c C4 for all t A T, such that jKtK : Kj ¼ Rt ðqÞDðtÞ. Here DðtÞ denotes
the modular function associated to the minimal parabolic subgroup of G which cor-
responds to the positive roots Fþ ¼ Fþ ðS; GÞ. An explicit formula P for DðtÞ is also
determined in [28], Lemma [Link] as DðtÞ ¼ qh2~r; nðzÞi , with 2r :¼ a A Fþ ma a. Using
 Entropy of the geodesic flow for metric spaces and Bruhat–Tits buildings 487

these facts, we obtain, for t A nðZ þ Þ and w0 the element of maximal length in W0 ,
that

jKt1 K : Kj ¼ jKw0 t1 w0 K : Kj ¼ Rt ðqÞDðw0 t1 w1

0 ÞÞ
1
¼ Rt ðqÞDðw0 tw1
0 Þ ¼ Rt ðqÞDðtÞ ¼ Rt ðqÞqh2~r; nðzÞi :

Summarizing we get from the above that the cardinality of the K-orbit of x ¼ t  x0
for t ¼ nðzÞ A nðZþ Þ can be estimated by

C3 qh2~r; nðzÞi c jK  xj ¼ jK=ðt1 Kt V KÞj c C4 qh2~r; nðzÞi ; ð6Þ

Recall that the map T ¼ nðZÞ ! a, t 7! t  x0 is an isomorphism from T onto the

lattice spanned by a4 ¼ 2~a 2 , a A Sþ . For x A nðZ þ Þ  x0 , we have
kak

r; nðzÞi c k~
h~ rk knðzÞk ¼ krkdB ðx0 ; nðzÞ  x0 Þ: ð7Þ

For k A N we now set xk :¼ ktr4  x0 A T  x0 and set rk :¼ kxk k ¼ 2ktkrk1 . The in-
tersection of the balls Bðx0 ; rk Þ in B with the apartment A G a contains a finite num-
ber of vertices in G  x0 . Since G  x0 V A is quasi-isometric to A the cardinality of
that intersection is bounded by a polynomial P in rk :

jBðx0 ; rk Þ V G  x0 V Aj c Pðrk Þ:

By the Cartan decomposition, Proposition 2, Bðx0 ; rk Þ V G  x0 is the union of the K-

orbits of all points in the previous intersection. Combining this with (6), (7) and since
r; ktr4i ¼ 4kt, we eventually obtain the upper bound
2h~

jBðx0 ; rk Þ V G  x0 j c Pðrk ÞC4 q 4kt :

On the other hand, by definition of rk , Bðx0 ; rk Þ V G  x0 contains the K-orbit of

xk ¼ ktr4  x0 . Therefore using (6) we have also the lower bound

C3 q 4kt c jBðx0 ; rk Þ V G  x0 j:

Hence
   
1 1 1
krk logq C3 þ 2 c logq jBðx0 ; rk Þ V G  x0 j c krk logq ðPðrk ÞC4 Þ þ 2 :
2kt rk 2kt

Taking the limit, k ! y, yields the claim.

3.4 Proof of Part (c) of the Main Theorem. The building B is a geodesically
complete Hadamard space (see Section 3.1) and has property (U) by Lemma 3.
Hence the assumptions of the Main Theorem Part (a) and (b) are satisfied and we
488 Enrico Leuzinger

obtain h top ðjÞ ¼ h vol ðdvÞ. Next Lemma 2 implies that h vol ðdvÞ ¼ h vol ðoÞ. And finally
Lemma 4 yields h vol ðoÞ ¼ k2rk.

Remark. If the local field has F has characteristic zero, then a theorem of Borel and
Harder asserts that G contains cocompact lattices (see [4]). Moreover, in characteris-
tic zero, a lattice is necessarily compact (see [23], IX 3.7). If the characteristic of F
is positive, cocompact lattices do not always exist (see [23], IX 1.6, IX 4.5). If
rankF G ¼ 1, then the building B is a tree. In that case cocompact lattices always
exist, also in positive characteristic (see [20]).

4 An application: The growth of fundamental groups

4.1 Critical exponents of lattices. Consider a geodesic space ðX ; dX Þ and a group G

which acts isometrically and properly discontinuously on X . For x; y A X denote by
Nðx; y; RÞ the number of orbit points of y A X under G contained in the ball Bðx; RÞ.
The exponent of growth (or critical exponent) of G is defined as

1
dðGÞ :¼ lim sup loga Nðx; y; RÞ:
R!y R

This number is independent of the chosen points x; y A X .

Proposition 3. Let ðX ; dX ; mÞ be a complete, geodesic metric measure space. Assume

that X has property (U ) and that there exists a group G of isometries of X which acts
totally discontinuously and cocompactly on X and preserves the measure m. Then the
critical exponent of G satisfies

dðGÞ ¼ h vol ðmÞ:

Proof. Pick a point x A X and let F be a fundamental domain for G in X which con-
tains x. For R > 0 set GR :¼ fg A G j g  x A Bðx; RÞg. Then Nðx; x; RÞ ¼ jGR j, the car-
dinality of GR . If D is the diameter of F we have by the triangle inequality that
GR  F J Bðx; R þ DÞ. On the other hand, using that F is a fundamental domain,
we get again by the triangle inequality and for R d D, Bðx; R  DÞ J GR  F. In con-
clusion we have for large R

mðBðx; R  DÞÞ c jGR jmðFÞ c mðBðx; R þ DÞÞ;

and the claim follows from the definitions and Proposition 1.

Remark. Critical exponents of arbitrary discrete subgroups of real and p-adic Lie
groups have been investigated in [19] and [26].

The Main Theorem Part (c) and Proposition 3 immediately yield the
 Entropy of the geodesic flow for metric spaces and Bruhat–Tits buildings 489

Corollary 1. Let ðB; dB Þ be the euclidean Bruhat–Tits building associated to an alge-

braic group G as in the Main Theorem Part (c). Let G be a cocompact lattice in
G ¼ GðFÞ. Then the critical exponent of G satisfies dðGÞ ¼ h vol ðdvÞ ¼ 2krk.

4.2 An extension of a theorem of Dinaburg. Let ðX ; dX Þ be a geodesic metric

space and let ðY ; dY Þ be a compact metric space whose universal covering space
is X . Let G be the fundamental group of Y and let F be a fundamental domain
of G in X . The set S :¼ fg A G j gF V F 0 qg is a finite set of generators for G.
For each positive integer m let bðG; S; mÞ be the number of distinct group elements
which can be expressed as word of length c m in the elements of S. Then the limit
limm!y m1 log bðG; S; mÞ ¼: hðG; SÞ exists and the property hðG; SÞ > 0 is independent
of the specific set S (see [24] or [25], Lemma 5.16). The number hðG; SÞ is called ex-
ponential growth rate of G or the entropy of G with respect to S (see [16], 5.11). The
following proposition is proved in the same way as in the case of manifolds (compare
[22], 5.17).

Proposition 4. Let ðX ; dX ; mÞ and G be as in Proposition 1. Let D be the diameter of a

fundamental domain F H X of G. Then the exponential growth rate of G with respect
to a finite generating set S ¼ SðFÞ satisfies

1
h vol ðmÞ d hðG; SÞ:
2D

We have the following extension of a theorem of Dinaburg (see [12], [25], 5.18).

Corollary 2. Let ðX ; dX ; mÞ, G, S and D be as in Proposition 4. Then the topological

entropy of the geodesic flow j on the space of geodesics ( parametrized by arc-length)
of the compact quotient GnX satisfies

1
h top ðjÞ d hðG; SÞ:
2D

Proof. The claim directly follows from Part (a) of the Main Theorem and Proposition
4.

Corollary 3. Let ðB; dB Þ be the euclidean Bruhat–Tits building associated to an alge-

braic group G as in the Main Theorem (c). Let G be a cocompact lattice in G ¼ GðFÞ
and let S ¼ SðFÞ be a ( finite) generating set for G with respect to some fundamental
domain F of diameter D. Then the entropy of G satisfies

hðG; SÞ c 4Dkrk ¼ 2DdðGÞ:

Proof. This is a direct consequence of Part (c) of the Main Theorem, Corollary 1 and
Corollary 2.
490 Enrico Leuzinger

Acknowledgement. I am grateful to the referee, whose critical and constructive com-

ments very much helped me to improve this work.

References

[1] W. Ballmann, Lectures on spaces of nonpositive curvature. Birkhäuser 1995.

MR1377265 (97a:53053) Zbl 0834.53003
[2] G. Besson, G. Courtois, S. Gallot, Minimal entropy and Mostow’s rigidity theorems. Er-
godic Theory Dynam. Systems 16 (1996), 623–649. MR1406425 (97e:58177)
Zbl 0887.58030
[3] A. Borel, Linear algebraic groups. Springer 1991. MR1102012 (92d:20001)
Zbl 0726.20030
[4] A. Borel, G. Harder, Existence of discrete cocompact subgroups of reductive groups over
local fields. J. Reine Angew. Math. 298 (1978), 53–64. MR483367 (80b:22022)
Zbl 0385.14014
[5] N. Bourbaki, Groupes et algèbres de Lie. Chapitres IV–VI. Hermann, Paris 1968.
MR0240238 (39 #1590) Zbl 0483.22001
[6] M. R. Bridson, A. Haefliger, Metric spaces of non-positive curvature. Springer 1999.
MR1744486 (2000k:53038) Zbl 0988.53001
[7] R. Brooks, A relation between growth and the spectrum of the Laplacian. Math. Z. 178
(1981), 501–508. MR638814 (83a:58089) Zbl 0458.58024
[8] K. S. Brown, Buildings. Springer 1989. MR969123 (90e:20001) Zbl 0715.20017
[9] F. Bruhat, J. Tits, Groupes réductifs sur un corps local. Inst. Hautes Études Sci. Publ.
Math. no. 41 (1972), 5–251. MR0327923 (48 #6265) Zbl 0254.14017
[10] D. Burago, Y. Burago, S. Ivanov, A course in metric geometry. Amer. Math. Soc. 2001.
MR1835418 (2002e:53053) Zbl 0981.51016
[11] C. Connell, B. Farb, Some recent applications of the barycenter method in geometry. In:
Topology and geometry of manifolds (Athens, GA, 2001), volume 71 of Proc. Sympos. Pure
Math., 19–50, Amer. Math. Soc. 2003. MR2024628 (2005e:53058) Zbl 1049.53030
[12] E. Dinaburg, The relation between topological entropy and metric entropy. Sov. Math.
Dokl. 11 (1970), 13–16. Zbl 0196.26401
[13] A. Freire, R. Mañé, On the entropy of the geodesic flow in manifolds without conjugate
points. Invent. Math. 69 (1982), 375–392. MR679763 (84d:58063) Zbl 0476.58019
[14] M. Gromov, Volume and bounded cohomology. Inst. Hautes Études Sci. Publ. Math. no.
56 (1982), 5–99 (1983). MR686042 (84h:53053) Zbl 0516.53046
[15] M. Gromov, Hyperbolic groups. In: Essays in group theory, volume 8 of Math. Sci. Res.
Inst. Publ., 75–263, Springer 1987. MR919829 (89e:20070) Zbl 0634.20015
[16] M. Gromov, Metric structures for Riemannian and non-Riemannian spaces. Birkhäuser
1999. MR1699320 (2000d:53065) Zbl 0953.53002
[17] A. Katok, B. Hasselblatt, Introduction to the modern theory of dynamical systems. Cam-
bridge Univ. Press 1995. MR1326374 (96c:58055) Zbl 0878.58020
[18] G. Knieper, On the asymptotic geometry of nonpositively curved manifolds. Geom. Funct.
Anal. 7 (1997), 755–782. MR1465601 (98h:53055) Zbl 0896.53033
[19] E. Leuzinger, Kazhdan’s property (T), L 2 -spectrum and isoperimetric inequalities for lo-
cally symmetric spaces. Comment. Math. Helv. 78 (2003), 116–133.
MR1966754 (2004a:53047) Zbl 1027.22015
[20] A. Lubotzky, Lattices in rank one Lie groups over local fields. Geom. Funct. Anal. 1
(1991), 406–431. MR1132296 (92k:22019) Zbl 0786.22017
 Entropy of the geodesic flow for metric spaces and Bruhat–Tits buildings 491

[21] I. G. Macdonald, Spherical functions on a group of p-adic type. Ramanujan Institute,

Centre for Advanced Study in Mathematics, University of Madras 1971.
MR0435301 (55 #8261) Zbl 0302.43018
[22] A. Manning, Topological entropy for geodesic flows. Ann. of Math. (2) 110 (1979),
567–573. MR554385 (81e:58044) Zbl 0426.58016
[23] G. A. Margulis, Discrete subgroups of semisimple Lie groups. Springer 1991.
MR1090825 (92h:22021) Zbl 0732.22008
[24] J. Milnor, A note on curvature and fundamental group. J. Di¤erential Geometry 2 (1968),
1–7. MR0232311 (38 #636) Zbl 0162.25401
[25] G. P. Paternain, Geodesic flows. Birkhäuser 1999. MR1712465 (2000h:53108)
Zbl 0930.53001
[26] J.-F. Quint, Divergence exponentielle des sous-groupes discrets en rang supérieur. Com-
ment. Math. Helv. 77 (2002), 563–608. MR1933790 (2003j:22015) Zbl 1010.22018
[27] M. Ronan, Lectures on buildings. Academic Press 1989. MR1005533 (90j:20001)
Zbl 0694.51001
[28] A. J. Silberger, Introduction to harmonic analysis on reductive p-adic groups. Princeton
Univ. Press 1979. MR544991 (81m:22025) Zbl 0458.22006
[29] J. Tits, Reductive groups over local fields. In: Automorphic forms, representations and L-
functions (Proc. Sympos. Pure Math., Corvallis, Ore., 1977), Part 1, 29–69, Amer. Math.
Soc. 1979. MR546588 (80h:20064) Zbl 0415.20035
[30] A. Weil, Basic number theory. Springer 1995. MR1344916 (96c:11002) Zbl 0823.11001

Received 28 June, 2005; revised 18 October, 2005

E. Leuzinger, Math. Institut II, Universität Karlsruhe, 76128 Karlsruhe, Germany
Email: [Link]@[Link]