J. Math. Soc.

Japan
Vol. 56, No. 1, 2004

Entropy of subshifts and the Macaev norm

By Rui Okayasu

(Received May 15, 2002)

(Revised Aug. 14, 2002)

Abstract. We obtain the exact value of Voiculescu’s invariant ky ðtÞ, which is an

obstruction of the existence of quasicentral approximate units relative to the Macaev ideal
in perturbation theory, for a tuple t of operators in the following two classes: (1)
creation operators associated with a subshift, which are used to define Matsumoto al-
gebras, (2) unitaries in the left regular representation of a finitely generated group.

1. Introduction.
In the remarkable serial works [Voi1], [Voi2], [Voi3] and [DV] on perturbation of
Hilbert space operators, Voiculescu investigated a numerical invariant kF ðtÞ for a family
t of bounded linear operators on a separable Hilbert space, where kF ðtÞ is the obstruc-
ð0Þ
tion of the existence of quasicentral approximate units relative to the normed ideal SF
corresponding to a symmetric norming function F, (see definitions in Section 2). The
invariant kF ðtÞ is considered to be a kind of dimension of t with respect to the normed
ð0Þ
ideal SF (see [Voi1] and [DV]).
In the present paper, we study the invariant kF ðtÞ for the Macaev ideal, which is
denoted by ky ðtÞ. It is known that ky ðtÞ possesses several remarkable properties: for
ð0Þ
instance, ky ðtÞ is always finite and kF ðtÞ ¼ 0 if SF is strictly larger than the Macaev
ideal. In [Voi3], Voiculescu investigated the invariant ky ðtÞ for several examples. He
proved that ky ðtÞ ¼ log N for an N-tuple t of isometries in extensions of the Cuntz
algebra ON . Here, log N can be interpreted as the value of the topological entropy of
the N-full shift. Inspired by this result, we show that ky ðtÞ ¼ htop ðX Þ for a general
subshift X with a certain condition, where htop ðX Þ is the topological entropy of X and
t is the family of creation operators on the Fock space associated with the subshift
X , which is used to define the Matsumoto algebra associated with X (e.g. see [Mat]).
In particular, we show that ky ðtÞ ¼ htop ðX Þ holds for every almost sofic shift X (cf.
[Pet]).
Let G be a countable finitely generated group and S its generating set. We also
study ky ððla Þa A S Þ, where l is the left regular representation of G. For the related
topic, see [Voi5], in which a relation between ky ððla Þa A S Þ and the entropy of random
walks on groups is discussed. By using a method introduced in [Oka], we can com-
pute the exact value of ky ððla Þa A S Þ for certain amalgamated free product groups.
Voiculescu proved that log N a ky ððla Þa A S Þ a logð2N 1Þ holds for the free group FN
with the canonical generating set S ([Voi3, Proposition 3.7. (a)]). As a particular case
of our results, we show that ky ððla Þa A S Þ ¼ logð2N 1Þ actually holds.

2000 Mathematics Subject Classification. Primary 47B10, 47D50; Secondary 47B06, 37B10, 37B40.
Key Word and Phases. perturbation theory, Macaev ideal, symbolic dynamics.
178 R. Okayasu

Acknowledgment. The author wishes to express his gratitude to Masaki Izumi

for his constant encouragement and important suggestions. He is grateful to Kengo
Matsumoto for useful comments about symbolic dynamics. He also thanks the referee
for careful reading.

2. Preliminary.

Let H be a separable infinite dimensional Hilbert space. By BðHÞ; KðHÞ; FðHÞ

and FðHÞþ 1 , we denote the bounded linear operators, the compact operators, the finite
rank operators and the finite rank positive contractions on H, respectively.
We begin by recalling some facts concerning normed ideal in [GK]. Let c0 be the
set of real valued sequences x ¼ ðxj Þj A N with lim j!y xj ¼ 0, and c0; 0 the subspace of c0
consisting of the sequences with finite support. A function F on c0; 0 is said to be a
symmetric norming function if F satisfies:
(1) F is a norm on c0; 0 ;
(2) Fðð1; 0; 0; . . .ÞÞ ¼ 1;
(3) Fððxj Þj A N Þ ¼ Fððjxpð jÞ jÞj A N Þ for any bijection p : N ! N.
For x ¼ ðxj Þj A N A c0 , we define

FðxÞ ¼ lim Fðx  ðnÞÞ A ½0; yŠ;

n!y

where x  ðnÞ ¼ ðx1 ; . . . ; xn ; 0; 0; . . .Þ A c0; 0 and x1 b x2 b


is the decreasing rearrange-
ment of the absolute value ðjxj jÞj A N . If T A KðHÞ and F is a symmetric norming
function, then let us denote

kTkF ¼ Fððsj ðTÞÞj A N Þ;

where ðsj ðTÞÞj A N is the singular numbers of T. We define two symmetrically normed
ideals

SF ¼ fT A KðHÞ j kTkF < yg;

ð0Þ ð0Þ
and SF by the closure of FðHÞ with respect to the norm k
kF . Note that SF does
not coincide with SF in general. If S is a symmetrically normed ideal, i.e. S is a ideal
of BðHÞ and a Banach space with respect to the norm k
kS satisfying:
(1) kXTY kS a kX k
kTkS
kY k for T A S and X ; Y A BðHÞ,
(2) kTkS ¼ kTk if T is of rank one,
where k
k is the operator norm in BðHÞ, then there exists a unique symmetric norming
ð0Þ
function F such that kTkS ¼ kTkF for T A FðHÞ and SF J S J SF .
We introduce some symmetrically normed ideals. For 1 < p a y, the symmetri-
cally normed ideal Cp ðHÞ is given by the symmetric norming function

X
y xj
Fp ðxÞ ¼ :
j¼1
j1 1=p

ð0Þ
We define Cp ðHÞ ¼ SFp . We remark that it coincides with SFp . For 1 a p < y, the
symmetrically normed ideal Cpþ ðHÞ is given by the symmetric norming function
 Entropy of subshifts and the Macaev norm 179

Pn 
j¼1 xj
Fpþ ðxÞ ¼ sup P n 1=p :
nAN j¼1 j

ð0Þ
We define Cpþ ðHÞ ¼ SFpþ . However SFþ is strictly smaller than Cpþ ðHÞ. For 1 a p <
p
q < r a y, we have

Cp ðHÞ Y Cq ðHÞ Y Cq ðHÞ Y Cqþ ðHÞ Y Cr ðHÞ;

where Cp ðHÞ is the Schatten p class.
For a given symmetric norming function F, which is not equivalent to the l 1 -norm,
ð0Þ
there is a symmetric norming function F  such that SF  is the dual of SF , where the
dual pairing is given by the bilinear form ðT; SÞ 7! TrðTSÞ. If 1=p þ 1=q ¼ 1, then
Cp ðHÞ  F Cq ðHÞ and Cp ðHÞ  F Cqþ ðHÞ. In particular, Cy ðHÞ and C1þ ðHÞ are called
the Macaev ideal and the dual Macaev ideal, respectively.
ð0Þ
Let SF be a symmetrically normed ideal with a symmetric norming function F.
If t ¼ ðT1 ; . . . ; TN Þ is an N-tuple of bounded linear operators, then the number kF ðtÞ is
defined by
kF ðtÞ ¼ lim infþ max k½u; Ta ŠkF ;
u A FðHÞ1 1aaaN

where the inferior limit is taken with respect to the natural order on FðHÞþ 1 and
½A; BŠ ¼ AB BA. Throughout this paper, we denote k
kFp by k
kp and kFp by
kp . A relation between the invariant kF and the existence of quasicentral approximate
ð0Þ
units relative to the symmetrically normed ideal SF is discussed in [Voi1]. A quasi-
ð0Þ
central approximate unit for t ¼ ðT1 ; . . . ; TN Þ relative to SF is a sequence fun gy n¼1 J
þ
FðHÞ1 such that un % I and lim n!y k½un ; Ta ŠkF ¼ 0 for 1 a a a N. Note that for an
N-tuple t ¼ ðT1 ; . . . ; TN Þ, there exists a quasicentral approximate unit for t relative to
ð0Þ
SF if and only if kF ðtÞ ¼ 0 (e.g. see [Voi2, Lemma 1.1]).
We use the following propositions to prove our theorem.
ð0Þ
Proposition 2.1 ([Voi1, Proposition 1.1]). Let t ¼ ðT1 ; . . . ; TN Þ A BðHÞ N and SF
be a symmetrically normed ideal with a symmetric norming function F. If we take a
þ
sequence fun gy
n¼1 J FðHÞ1 with w-lim n!y un ¼ I , then

kF ðtÞ a lim inf max k½un ; Ta ŠkF :

n!y 1aaaN

Proposition 2.2 ([Voi3, Proposition 2.1]). Let t ¼ ðT1 ; . . . ; TN Þ A BðHÞ N and

Xa A C1þ ðHÞ for a ¼ 1; . . . ; N. If

X
N
½Xa ; Ta Š A C1 ðHÞ þ BðHÞþ ;
a¼1

then we have
 !
 XN  XN
  e
 Tr ½Xa ; Ta Š  a k ðtÞ kXa kþ
1;
 a¼1
 y
a¼1

e
where kXa kþ
1 ¼ inf Y A FðHÞ kXa Y kF þ .
1
180 R. Okayasu

The following proposition was shown in the proof of [GK, Theorem 14.1].
Proposition 2.3. For T A C1þ ðHÞ, we have
Pn
e j¼1 sj ðTÞ
kTkþ1 ¼ lim sup P n :
n!y j¼1 1=j

3. Subshifts and Macaev norm.

Let A be a finite set with the discrete topology, which we call the alphabet, and AZ
Q
the two-sided infinite product space y i¼ y A endowed with the product topology. The
shift map s on A is given by ðsðxÞÞi ¼ xiþ1 for i A Z. The pair ðAZ ; sÞ is called
Z

the full shift. In particular, if the cardinality of the alphabet A is N, then we call it the
N-full shift.
Let X be a shift invariant closed subset of AZ . The topological dynamical system
ðX ; sX Þ is called a subshift of AZ , where sX is the restriction of the shift map s. We
sometimes denote the subshift ðX ; sX Þ by X for short. A word over A is a finite
sequence w ¼ ða1 ; . . . ; an Þ with ai A A. For x A AZ and a word w ¼ ða1 ; . . . ; an Þ, we say
that w occurs in x if there is an index i such that xi ¼ a1 ; . . . ; xiþn 1 ¼ an . The empty
word occurs in every x A AZ by convention. Let F be a collection of words over
AZ . We define the subshift XF to be the subset of sequences in AZ in which no word
in F occurs. It is well-known that any subshift X of AZ is given by XF for some
collection F of forbidden words over AZ . Note that for F ¼ q, the subshift XF is
the full shift AZ .
Let X be a subshift of AZ . We denote by Wn ðX Þ the set of all words with length
n that occur in X and we set

y
WðX Þ ¼ 6 Wn ðX Þ:
n¼0

Let j : Wmþnþ1 ðX Þ ! A be a map, which we call a block map. The extension of j

from X to AZ is defined by ðxi Þi A Z 7! ð yi Þi A Z , where

yi ¼ jððxi m ; xi mþ1 ; . . . ; xiþn ÞÞ:

We also denote this extension by j and call it a sliding block code. Let X ; Y be two
subshifts and j : X ! Y a sliding block code. If j is one-to-one, then j is called an
embedding of X into Y and we denote X J Y . If j has an inverse, i.e. a sliding block
code c : Y ! X such that c  j ¼ idX and j  c ¼ idY , then two subshifts X and Y are
topologically conjugate.
The topological entropy of a subshift X is defined by

1
htop ðX Þ ¼ lim logjWn ðX Þj;
n!y n

where jWn ðX Þj is the cardinality of Wn ðX Þ. The reader is referred to [LM] for an

introduction to symbolic dynamics.
 Entropy of subshifts and the Macaev norm 181

For a given subshift X , we next construct the creation operators on the Fock space
associated with X (cf. [Mat]). Let fxa ga A A be an orthonormal basis of N-dimensional
Hilbert space C N , where N is the cardinality of A. For w ¼ ða1 ; . . . ; an Þ A Wn ðX Þ, we
denote xw ¼ xa1 n


n xan . We define the Fock space FX for a subshift X by

FX ¼ Cx0 l 0 spanfxw j w A Wn ðX Þg;

nAN

where x0 is the vacuum vector. The creation operator Ta on FX for a A A is given by

T a x0 ¼ xa ;

xa n xw if aw A WðX Þ;
Ta xw ¼
0 otherwise.

Note that Ta is a partial isometry such that

X
P0 þ Ta Ta ¼ 1;
aAA

where P0 is the rank one projection onto Cx0 . We denote by Pn the projection onto
the subspace spanned by xw for all w A Wn ðX Þ. For w ¼ ða1 ; . . . ; an Þ A Wn ðX Þ, we set
Tw ¼ Ta1


Tan . The following proposition is essentially proved in [Voi3].

Proposition 3.1. If t ¼ ðTa Þa A A , then we have

ky ðtÞ a htop ðX Þ:

Proof. We first assume that the topological entropy of X is non-zero. Let us

denote h ¼ htop ðX Þ. By definition, for a given e > 1, there exists K A N such that for
any n b K, we have

1
logjWn ðX Þj < eh:
n

Thus

jWn ðX Þj < e neh ;

for all n b K. We set

n 1 
X j
Xn ¼ 1 Pj :
j¼0
n

One can show that

1
k½Xn ; Ta Šk a :
n

Since
182 R. Okayasu

n
X K 1
X n
X
rn ¼ rankð½Xn ; Ta ŠÞ a jWj ðX Þj a jWj ðX Þj þ e jeh
j¼1 j¼1 j¼K

for n b K, we obtain
P rn
j¼1 1=j
ky ðtÞ a lim sup max k½Xn ; Ta Šky a lim sup a eh:
n!y aAA n!y n

In the case of h ¼ 0, for any e > 0, we have

jWn ðX Þj < e ne

for su‰ciently large n. By the same argument, we can get

ky ðtÞ a lim sup max k½Xn ; Ta Šky a e;
n!y aAA

for arbitrary e > 0. r

Next we obtain the lower bound of ky ðtÞ by using Proposition 2.2. Before it, we
prepare some notations. For any m A Z and w ¼ ða1 ; . . . ; an Þ A Wn ðX Þ, let us denote

m ½wŠ ¼ fðxi Þi A Z A X j xm ¼ a1 ; . . . ; xmþn 1 ¼ an g:

We sometimes denote the cylinder set 0 ½wŠ by ½wŠ for short. Let m be a shift invariant
probability measure on X . The following holds:
P
(1) a A A mð½aŠÞ ¼ 1;P
(2) mð½a1 ; . . . ; an ŠÞ ¼ a0 A A mð½a0 ; a1 ; . . . ; an ŠÞ;
P
(3) mð½a1 ; . . . ; an ŠÞ ¼ anþ1 A A mð½a1 ; . . . ; an ; anþ1 ŠÞ.
For any partition b ¼ ðB1 ; . . . ; Bn Þ of X , we define a function on X by
X
Im ðbÞ ¼ log mðBÞwB ;
BAb

where wB is the characteristic function of B. Let b1 ; . . . ; b k be partitions of X . The

k
partition 4i¼1 bi is defined by
(  )
k 
7 Bi  Bi A bi ; 1 a i a k :

i¼1 

The value
X
Hm ðbÞ ¼ mðBÞ log mðBÞ
BAb

is called the entropy of the partition b. We define

!
1 n 1
hm ðb; sX Þ ¼ lim Hm 4 sX i ðbÞ :
n!y n
i¼0

The entropy of ðX ; sX ; mÞ is defined by

 Entropy of subshifts and the Macaev norm 183

hm ðsX Þ ¼ supfhm ðb; sX Þ j Hm ðbÞ < yg:

Note that hm ðsX Þ a htop ðX Þ in general. A shift invariant probability measure m is said
to be a maximal measure if htop ðX Þ ¼ hm ðsX Þ. The reader is referred to [DGS] for
details.
Theorem 3.2. Let t ¼ ðTa Þa A A be the creation operators for a subshift X. If there
exists a shift invariant probability measure m on X such that for any e > 0 we have
(  !  )!
X y  1 n 
 
m xAX : Im 4 sX i b ðxÞ hm ðsX Þ > e < y;
n¼0
 n þ 1 i¼0 

where b is the generating partition f½aŠga A A of X, then

hm ðsX Þ a ky ðtÞ:

In particular, if we can take a maximal measure m with the above condition, then we have
ky ðtÞ ¼ htop ðX Þ:

Proof. Let m be a shift invariant probability measure on X . For a A A, we set

X X

Xa ¼ mð½awŠÞTw P0 Taw :
nb0 w A Wn ðX Þ

Then
X XX X

Ta Xa ¼ mð½awŠÞTaw P0 Taw
aAA nb0 a A A w A Wn ðX Þ
X X
¼ mð½wŠÞTw P0 Tw ;
nb1 w A Wn ðX Þ

and
!
X X X X
Xa Ta ¼ mð½awŠÞ Tw P0 Tw
aAA nb0 w A Wn ðX Þ aAA
X X
¼ mð½wŠÞTw P0 Tw :
nb0 w A Wn ðX Þ

Hence we have
X
½Xa ; Ta Š ¼ P0 :
aAA

We assume that hm ðsX Þ 0 0 and denote it by h for short. To apply Proposition 2.2, we
need an estimate of kXa kþe
1 . Fix e > 0 and a A A. We set
ðnþ1ÞðhþeÞ ðnþ1Þðh eÞ
Dn ¼ fw A Wn ðX Þ j e a mð½awŠÞ a e g;

and
184 R. Okayasu

X
en ¼ mð½awŠÞ:
w A Wn ðX ÞnDn

If m satisfies the assumption, then we have

X
en < y: ð?Þ
nb0

Note that sj ðXa Þ ¼ sj ðXa Ta Þ for all j A N. e

Thus we have kXa kþ e
þ
1 ¼ kXa Ta k1 . We put
XX
fa ¼
X mð½awŠÞTw P0 Tw :
nb0 w A Dn

We remark that for each j A N, there are n A N, w A Wn ðX Þ such that sj ðXa Ta Þ ¼

mð½awŠÞ. By ð?Þ, we obtain
Pn
e
þ e
þ j¼1 sj ðXa Ta Þ
kXa k1 ¼ kXa Ta k1 ¼ lim sup P n
n!y j¼1 1=j
Py
f e
þ
a kXa k1 þ lim sup P nk¼0 ¼ kX
ek fa kþe
1:
n!y j¼1 1=j
e Pn
Hence it su‰ces to give an estimate of kX fa kþ
1 . Let dn ¼ j¼0 jDj j, where jDj j is the
cardinality of Dj . One can easily check that
P dn
f
e
þ
fa k a lim sup P j¼1 sj ðXa Þ
kX 1 dn
:
n!y j¼1 1=j
fa Þ ¼ mð½awŠÞ for some w A Dn , then we have
Note that if sj ðX

e ðnþ1ÞðhþeÞ fa Þ ¼ mð½awŠÞ a e
a sj ð X ðnþ1Þðh eÞ
:
fa Þ ¼ mð½awŠÞ for some w A Dm and j a dn .
Assume that there are m > n such that sj ðX
Then it holds that
ðmþ1Þðh eÞ ðnþ1ÞðhþeÞ
e be : ð??Þ
ðmþ1Þðh eÞ ðnþ1ÞðhþeÞ
Indeed, if e <e , then
fa Þ ¼ mð½awŠÞ a e
sj ð X ðmþ1Þðh eÞ
<e ðnþ1ÞðhþeÞ
a mð½auŠÞ;
fa Þ ¼
for all u A Dk ð1 a k a nÞ. However, by our assumption, we have mð½avŠÞ a sj ðX
mð½awŠÞ for some v A Dl and 1 a l a n. This is a contradiction.
Hence, by ð??Þ, we have
hþe
m þ 1 a ðn þ 1Þ :
h e
Let k A N with

hþe hþe
ðn þ 1Þ 1 < k þ 1 a ðn þ 1Þ :
h e h e
 Entropy of subshifts and the Macaev norm 185

Since
P dn Pk P
f
j¼1 sj ðXa Þ i¼0 w A Di mð½awÞ
P dn a
j¼1 1=j
log dn
Pk
i¼0mð½aÞ
a
log dn
nþ1 hþe
a  mð½aÞ;
log dn h
e

we obtain

e
fa kþ nþ1 hþe
kX 1 a lim sup  mð½aÞ:
n!y log dn h
e

Moreover, because
X X
mð½aÞ ¼ mð½awÞ þ mð½awÞ a jDn je
ðnþ1Þðh
eÞ þ en ;
w A Dn w A Wn ðX ÞnDn

we have

ðmð½aÞ
en Þeðnþ1Þðh
eÞ a jDn j:

Note that en ! 0 ðn ! yÞ by ð?Þ. Therefore

e
fa kþ nþ1 hþe
kX 1 a lim sup  mð½aÞ
n!y logjDn j h
e

nþ1 hþe
a lim sup  mð½aÞ
n!y logðmð½aÞ
en Þ þ ðn þ 1Þðh
eÞ h
e
hþe
¼ mð½aÞ:
ðh
eÞ 2

Since e is arbitrary, we have

e 1
kXa kþ
1 a mð½aÞ:
h

By Proposition 2.2, the proof is complete. r

We now give some examples of subshifts with a maximal measure satisfying the
condition in Theorem 3.2.
Corollary 3.3. Let A be a 0-1 N  N matrix. We denote by SA the Markov
shift associated with A, i.e.

SA ¼ fðai Þi A Z A S Z j Aðai ; aiþ1 Þ ¼ 1g;

186 R. Okayasu

where S ¼ f1; . . . ; Ng is an alphabet. If t ¼ ðTa Þa A S is the creation operators for the

Markov shift SA , then we have

ky ðtÞ ¼ htop ðSA Þ:

Proof. It su‰ces to show that the unique maximal measure of SA satisfies the
condition in Theorem 3.2. For simplicity, we may assume that A is irreducible with the
Perron value a. Note that the topological entropy htop ðSA Þ is equal to log a. If l and r
PN
are the left and right Perron vectors with a¼1 la ra ¼ 1, then the unique maximal
measure m is given by

l a 0 ra n
mð½a0 ; a1 ; . . . ; an ŠÞ ¼ ;
an

where ða0 ; a1 ; . . . ; an Þ A Wnþ1 ðSA Þ (e.g. see [Kit]). For any e > 0, there exists K A N such
that for any n b K, we have
 
log la rb a
 n þ 1  < e;
 

for all 1 a a; b a N. Therefore for any w A Wnþ1 ðSA Þ, we have

 
 1 

 nþ1 log mð½wŠÞ log a  < e;


for all n b K, i.e. the maximal measure m satisfies the condition in Theorem 3.2. r

More generally, there is a class of subshifts, which is called almost sofic (see [Pet]).
A subshift X is said to be almost sofic if for any e > 0, there is an SFT S J X such
that htop ðX Þ e < htop ðSÞ, where a shift of finite type or SFT is a subshift that can be
described by a finite set of forbidden words, i.e. a subshift having the form XF for some
finite set F of words.
Corollary 3.4. If t ¼ ðTa Þa A A is the creation operators for an SFT S, then we
have
ky ðtÞ ¼ htop ðSÞ:

Proof. We recall that every SFT S is topologically conjugate to a Markov shift

SA associated with a 0-1 matrix A. Now we give a short proof of this result. Let S be
an SFT that can be described by a finite set F of forbidden words. We may assume
½N Š
that all words in F have length N þ 1. We set AS ¼ WN ðSÞ and the block map
½N Š ½N Š
j : WN ðSÞ ! AS , w 7! w. We define the N-th higher block code bN : S ! ðAS ÞZ by
½N Š
ðbN ðxÞÞi ¼ ðxi ; . . . ; xiþN 1Þ A AS ;

for x ¼ ðxi Þi A N A S. Note that bN is the sliding block code with respect to j. The
subshift bN ðSÞ is given by a Markov shift, i.e. there is a 0-1 matrix A with bN ðSÞ ¼ SA .
Let m be the maximal measure of SA . The maximal measure of S is given by
n ¼ m  b N . We recall that m is the Markov measure given by the left and right eigen-
vectors l; r and the eigenvalue a. For w A Wn ðSÞ with n b N, we have
 Entropy of subshifts and the Macaev norm 187

nð½wŠÞ ¼ mð½jðw½1; NŠ Þ; . . . ; jðw½n Nþ1; nŠ ފÞ

l a rb
¼ ;
an N
where a ¼ jðw½1; NŠ Þ, b ¼ jðw½n Nþ1; nŠ Þ and w½k; l Š ¼ ðwk ; . . . ; wl Þ for k a l. Hence one
can show that the maximal measure n of S satisfies the condition in Theorem 3.2 by the
same argument as in the proof of Corollary 3.3. r

Corollary 3.5. Let X be an almost sofic shift. If t ¼ ðTa Þa A A is the creation

operators for X, then we have
ky ðtÞ ¼ htop ðX Þ:

Proof. Let e > 0. Since X is almost sofic, there is an SFT S J X such that
htop ðX Þ e < htop ðSÞ. Let j : S ! X be an embedding. Note that the subshift jðSÞ
is also an SFT. Thus we may identify jðSÞ with S. Let m be the unique maximal
measure of S. For a A A, we set
XX

Xa ¼ mð½awŠÞTw P0 Taw ;
nb0 w

where w runs over all elements in Wn ðSÞ with aw A WðSÞ. We have shown that the
maximal measure m of S satisfies the condition of Theorem 3.2 in the proof of Corollary
3.4. Hence by the same argument as in the proof of Theorem 3.2, we have
htop ðSÞ a ky ðtÞ:
Thus for arbitrary e > 0, the following holds:

htop ðX Þ e < htop ðSÞ a ky ðtÞ:

It therefore follows from Proposition 3.1 that htop ðX Þ ¼ ky ðtÞ if X is an almost sofic
shift. r
For b > 1, the b-transformation Tb on the interval ½0; 1Š is defined by the mul-
tiplication with b ðmod 1Þ, i.e. Tb ðxÞ ¼ bx ½ bxŠ, where ½tŠ is the integer part of t. Let
N A N with N 1 < b a N and A ¼ f0; 1; . . . ; N 1g. The b-expansion of x A ½0; 1Š is
a sequence dðx; bÞ ¼ fdi ðx; bÞgi A N of A determined by
di ðx; bÞ ¼ ½ bTbi 1 ðxފ:

We set
zb ¼ sup ðdi ðx; bÞÞi A N ;
x A ½0; 1Þ

where the above supremum is taken in the lexicographical order, and we define the shift
invariant closed subset Sbþ of the full one-sided shift AN by
Sbþ ¼ fx A AN j s i ðxÞ a zb ; i ¼ 0; 1; . . .g;

where a is the lexicographical order on AN ¼ f0; 1; . . . ; N 1gN . The b-shift Sb is the

natural extension given by
188 R. Okayasu

Sb ¼ fðxi Þi A Z A AZ j ðxi Þibk A Sbþ ; k A Zg:

It is known that htop ðSb Þ ¼ log b, (see [Hof]).
The following result might be known among specialists. However, we give a proof
here as we can not find it in the literature.

Proposition 3.6. For b > 1, the b-shift Sb is an almost sofic shift.

Proof. In [Par], it is shown that Sb is an SFT if and only if dð1; bÞ is finite, i.e.
there is K A N such that dk ð1; bÞ ¼ 0 for all k b K. Thus we may assume that dð1; bÞ is
not finite. Let zb ¼ ðxi Þi A N . For n A N, there is bðnÞ < b such that

x1 x x
1¼ þ 2 2 þ


þ n n:
bðnÞ bðnÞ bðnÞ

In [Par, Theorem 5], it is proved that

lim bðnÞ ¼ b:
n!y

Hence we may assume that N 1 a bðnÞ < b for su‰ciently large n. Since the maxi-
mal element zbðnÞ has the form
ðx1 ; x2 ; . . . ; ðxn 1Þ; x1 ; x2 ; . . . ; ðxn 1Þ; x1 ; . . .Þ;

we have zbðnÞ < z, where < is the lexicographical order. Therefore we obtain
þ
SbðnÞ J Sbþ J f0; 1; . . . ; N 1gN :

It follows that SbðnÞ is the shift invariant closed subset of Sb with topological entropy
log bðnÞ. Since dð1; bðnÞÞ is finite, the subshift SbðnÞ is an SFT. It therefore follows
from [Par, Theorem 5] that Sb is an almost sofic. r

Hence it holds that ky ðtÞ ¼ htop ðSb Þ for every b-shift by Corollary 3.5.
Corollary 3.7. Let Sb be the b-shift for b > 1. If t ¼ ðTa Þa A A is the creation
operators for Sb , then we have
ky ðtÞ ¼ htop ðSb Þ ¼ log b:

4. Groups and Macaev norm.

We discuss a relation between groups and the Macaev norm. Let G be a countable
finitely generated group, S a symmetric set of generators of G. We denote by j
jS the
word length and by Wn ðG; SÞ the set of elements in G with length n, with respect to the
system of generators S. The logarithmic volume of a group G in a given system of gen-
erators S is the number

logjWn ðG; SÞj

vS ¼ lim ;
n!y n

(cf. [Ver]). The following proposition can be proved in the same way as in the free
group case [Voi3, Proposition 3.7. (a)].
 Entropy of subshifts and the Macaev norm 189

Proposition 4.1. Let G be a finitely generated group with a finite generating set
S and l the left regular representation of G. If we set lS ¼ ðla Þa A S , then


ky ðlS Þa vS :

Proof. Let us denote by Pn the projection onto the subspace spanfdg A l 2 ðGÞ j
jgjS ¼ ng. If we set

n
1  
X j
Xn ¼ 1
Pj ;
j¼0
n

then we have

1
kXn la
la Xn k ¼ kla Xn la
Xn k a :
n

for a A S. Hence
Pn
log j¼0 jWn ðG; SÞj


ky ðlS Þ a lim sup max k½Xn ; la k

y a lim ¼ vS : r
n!y aAS n!y n


Now we compute the exact value of ky ðlS Þ for certain amalgamated free product
groups.
Proposition 4.2. Let A be a finite group, G1 ; . . . ; GM nontrivial finite groups
containing A as a subgroup and H1 ; . . . ; HN the product group of the infinite cyclic group
Z and the finite group A, ðN þ M > 1Þ. Let G be the amalgamated free product group
of G1 ; . . . ; GM , H1 ; . . . ; HN with amalgamation over A. Set S ¼ G1 U    U GM U ðS1  AÞ
U    U ðSN  AÞnfeg, where Sj is the canonical generating set fxj ; x
1 j g of the infinite
cyclic group Z and e is the group unit. Let l be the left regular representation of G and
lS ¼ ðla Þa A S . Then we have


ky ðlS Þ ¼ vS :

In particular, for the free group FN ðN b 2Þ, we have

ky ðlS Þ ¼ logð2N
1Þ:


Proof. By Proposition 4.1, it su‰ces to show that vS a ky ðlS Þ. Let Wi be the
set of the representatives of Gi =A with e A Wi for i ¼ 1; . . . ; M. We identify xj with
ðxj ; eÞ A Hj for j ¼ 1; . . . ; N, and set WMþj ¼ fxj ; x
1
j ; eg. Let

MþN
S~ ¼ 6 Wi nfeg:
i¼1

We define the 0-1 matrix A with index S~ by


1 if jabjS ¼ 2;
Aða; bÞ ¼
0 otherwise.
190 R. Okayasu

One can easily check that the above matrix A is irreducible and the topological entropy
htop ðSA Þ of the Markov shift SA coincides with the logarithmic volume vS of G with
respect to the generating set S.
We denote by G0 the subset of G consisting of the group unit e and elements
a1


an A G; ðn A NÞ of the form
8
<ak A Wik nfeg for k ¼ 1; . . . ; n;
>
ik 0 ikþ1 if 1 a ik a M;
>
ak ¼ akþ1 if M þ 1 a ik a M þ N; ik ¼ ikþ1 :
:

Note that the subspace l 2 ðG0 Þ can be identified with the Fock space FA of the Markov
shift SA by the following correspondence:

d e $ x0 ;
da1


an $ xa1 n


n xan :

Let us denote by Pn the projection onto the subspace

spanfdg A l 2 ðGÞ j jgjS ¼ ng:

For a A S, we define the partial isometry Ta A Bðl 2 ðGÞÞ by

X
Ta ¼ Pnþ1 la Pn :
nb0

Under the identification with FA , the partial isometry Ta jl 2 ðG0 Þ for a A S~ is the creation
operator on FA , (cf. [Oka]). We also identify G0 and WðSA Þ. For w ¼ a1


an A G0 ,
we set Tw ¼ Ta1


Tan . Let m be the maximal measure of SA . For a A S~, we put
XX

Xa ¼ mð½awŠÞTw P0 Taw ;
nb0 w

where w runs over all w A G0 with jwjS ¼ n and jawjS ¼ jwjS þ 1. For a A SnS~, we set
Xa ¼ 0. It can be easily checked that ½la ; Xa Š ¼ ½Ta ; Xa Š for a A S~. Therefore by the
same proof as in the subshift case, we obtain

vS ¼ htop ðSA Þ ¼ ky ðlS Þ: r

Remark 4.3. Let G be a finitely generated group with a finite generating set S.
In [Voi5], Voiculescu proved that if the entropy hðG; mÞ of a random walk m on G with
support S is non-zero, then ky ððla Þa A S Þ is non-zero. However the above proposition
suggests that the volume vS of G is more related to the invariant ky ððla Þa A S Þ rather than
the entropy hðG; mÞ. It is an interesting problem to ask whether vS being non-zero
implies ky ððla Þa A S Þ being non-zero. We also remark here that there is a relation
between vS and hðG; mÞ: If hðG; mÞ 0 0, then vS 0 0, (see [Ver, Theorem 1]). If the
above mentioned problem was solved a‰rmatively, then it would follow from Prop-
osition 4.1 that ky ððla Þa A S Þ 0 0 if and only if vS 0 0, i.e. G has exponential growth.
 Entropy of subshifts and the Macaev norm 191

References
[DGS] M. Denker, C. Grillenberger and K. Sigmund, Ergodic theory on compact spaces, Lecture Notes in
Math., 527, Springer-Verlag, Berlin-New York, 1976.
[DV] G. David and D. Voiculescu, s-numbers of singular integrals for the invariance of absolutely
continuous spectra in fractional dimensions, J. Funct. Anal., 94 (1990), no. 1, 14–26.
[GK] I. C. Gohberg and M. G. Kreı̆n, Introduction to the theory of linear nonselfadjoint operators,
Translated from the Russian by A. Feinstein, Transl. Math. Monogr., 18, Amer. Math. Soc.,
Providence, R.I., 1969, xvþ378 pp.
[Hof ] F. Hofbauer, b-shifts have unique maximal measure, Monatsh. Math., 85 (1978), no. 3, 189–198.
[Kit] B. P. Kitchens, Symbolic dynamics. One-sided, two-sided and countable state Markov shifts,
Universitext, Springer-Verlag, Berlin, 1998.
[LM] D. Lind and B. Marcus, An introduction to symbolic dynamics and coding, Cambridge Univ.
Press, Cambridge, 1995, xviþ495 pp.
[Mat] K. Matsumoto, On C  -algebras associated with subshifts, Internat. J. Math., 8 (1997), no. 3,
357–374.
[Oka] R. Okayasu, Cuntz-Krieger-Pimsner algebras associated with amalgamated free product groups,
Publ. Res. Inst. Math. Sci., 38 (2002), no. 1, 147–190.
[Par] W. Parry, On the b-expansions of real numbers, Acta Math. Acad. Sci. Hungar., 11 (1960),
401–416.
[Pet] K. Petersen, Chains, entropy, coding, Ergodic Theory Dynam. Systems, 6 (1986), no. 3, 415–448.
[Ver] A. M. Vershik, Dynamic theory of growth in groups: entropy, boundaries, examples, Russian
Math. Surveys, 55 (2000), no. 4, 667–733.
[Voi1] D. Voiculescu, Some results on norm-ideal perturbations of Hilbert space operators, J. Operator
Theory, 2 (1979), no. 1, 3–37.
[Voi2] D. Voiculescu, Some results on norm-ideal perturbations of Hilbert space operators. II, J. Operator
Theory, 5 (1981), no. 1, 77–100.
[Voi3] D. Voiculescu, On the existence of quasicentral approximate units relative to normed ideals. Part I,
J. Funct. Anal., 91 (1990), no. 1, 1–36.
[Voi4] D. Voiculescu, Entropy of dynamical systems and perturbations of operators, Ergodic Theory
Dynam. Systems, 11 (1991), no. 4, 779–786.
[Voi5] D. Voiculescu, Entropy of random walks on groups and the Macaev norm, Proc. Amer. Math.
Soc., 119 (1993), no. 3, 971–977.

Rui Okayasu
Department of Mathematics
Osaka Kyoiku University
Kashiwara, Osaka 582-8582
Japan
E-mail: rui@cc.osaka-kyoiku.ac.jp