Upload
teturo-kamae
View
212
Download
0
Embed Size (px)
Citation preview
Numeration Systems as Dynamical Systems: IntroductionAuthor(s): Teturo KamaeSource: Lecture Notes-Monograph Series, Vol. 48, Dynamics & Stochastics (2006), pp. 198-211Published by: Institute of Mathematical StatisticsStable URL: http://www.jstor.org/stable/4356373 .
Accessed: 18/06/2014 00:23
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp
.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].
.
Institute of Mathematical Statistics is collaborating with JSTOR to digitize, preserve and extend access toLecture Notes-Monograph Series.
http://www.jstor.org
This content downloaded from 62.122.72.154 on Wed, 18 Jun 2014 00:23:08 AMAll use subject to JSTOR Terms and Conditions
IMS Lecture Notes?Monograph Series Dynamics & Stochastics Vol. 48 (2006) 198-211 ? Institute of Mathematical Statistics, 2006 DOI: 10.1214/074921706000000220
Numeration systems as dynamical
systems ? introduction
Teturo Kamae1
Matsuyama University
Abstract: A numeration system originally implies a digitization of real num- bers, but in this paper it rather implies a compactification of real numbers as a result of the digitization.
By definition, a numeration system with G, where G is a nontrivial closed multiplicative subgroup of R+, is a nontrivial compact metrizable space O admitting a continuous (?a; + ?)-action of (?,?) G G ? ? to ? G O, such that the (? + ?)-action is strictly ergodic with the unique invariant probability measure ?^, which is the unique G-invariant probability measure attaining the topological entropy | logA| of the transformation ? ??? ?a; for any ? f 1.
We construct a class of numeration systems coming from weighted substi- tutions, which contains those coming from substitutions or ?-expansions with algebraic ?. It also contains those with G = R+.
We obtained an exact formula for the ?-function of the numeration systems coming from weighted substitutions and studied the properties. We found a lot of applications of the numeration systems to the /^-expansions, Fractal geometry or the deterministic self-similar processes which are seen in [10].
This paper is based on [9] changing the way of presentation. The complete version of this paper is in [10].
1. Numeration systems
By a numeration system, we mean a compact metrizable space O with at least 2
elements as follows:
(til) There exists a nontrivial closed multiplicative subgroup G of R+ and a
continuous action ?a; +1 of (?, t) G G x R to u; G O such that ?'(?a; + t) +1' =
?'?a; + X't + f.
(?)2) The (? + ?)-action of t G R to ? G O is strictly ergodic with the unique in-
variant probability measure ??? called the equilibrium measure on O. Consequently, it is invariant under the (?a; + ?)-action of (?, t) G G x R to ? G O as well.
(tt3) For any fixed ?? G G, the transformation ? ???- ??a> on O has the | log ??|-
topological entropy. For any probability measure ? on O other than pq which is
invariant under the Attraction of ? G G to a;, and 1 f ?? G G, it holds that
???) < ???(??) = I log ?01 -
The (? + ?)-action of t G R to ? G O is called the additive action or R-action, while the ?a^^??? of ? G G to a; G O is called the multiplicative action or G-action.
Note that if O is a numeration system, then O is a connected space with the
continuum cardinality. Also, note that the multiplicative group G as above is either
R_i_ or {??; n G ?} for some ? > 1. Moreover, the additive action is faithful, that
is, ? +1 = ? implies t = 0 for any o; G O and t G R.
1Matsuyama University, 790-8578 Japan, e-mail: kamaeGapost.plala.or.jp AMS 2000 subject classifications: primary 37B10. Keywords and phrases: numeration system, weighted substitution, fractal function and set,
self-similar process, ?-function.
198
This content downloaded from 62.122.72.154 on Wed, 18 Jun 2014 00:23:08 AMAll use subject to JSTOR Terms and Conditions
Numeration systems as dynamical systems 199
This is because if there exist ?? G O and ti f ? such that ???-ti = ??, then
take a sequence ?? in G such that ?? ?? O and ??a>? converges as ? ?? co. Let
a;?? := linin?oo ??a;?. For any t G R, let an be a sequence of integers such that
a????? ?? t as ? ?? co. Then we have
^oo + t ? lim (???? + ??a?*?) ??>?? = lim ?? (?? + Oriti) = lim ??a;? = ??,.
??>?? ??>cx)
Thus, a;^ becomes a fixed point of the (? + t)-action of t G R to o; G O. Since
this action is minimal, we have O = {a;^}, contradicting with that O has at least
2 elements.
An example of a numeration system is the set {0,1}Z with the product topology divided by the closed equivalence relation ~ such that
(...,a_2,a_?;a?,a?,a2,...) ~ (... ,/3-2,0-?? A)?AiA2, ? ? ?)
if and only if there exists N G Z U {?00} satisfying that an = ?n (Vn > ?), aN = ?? + 1 and an = 0, ?? = 1 (Vn < N) or the same statement with a
and ? exchanged. Let O(2) := {0,1}Z/ ~ and the equivalence class containing
(...,a_2,a_i;ao,ai,a2,-..) ? {0,1}Z is denoted by S?=-??a?2? G O(2). Then,
O(2) is an additive topological group with the addition as follows:
00 00 00
n= ? oo n=?00 n= ? oo
if and only if there exists (..., n_2,n_i; 770, r/i,7/2,...) G {0,1}Z satisfying that
2r/n+1 -f 7n = an ?+? /3n -f r/n (Vn G ?).
This is isomorphic to the 2-adic solenoidal group which is by definition the pro-
jective limit of the projective system ? : R/Z -+ R/Z with ?(a) = 2a (a G R/Z). Moreover, R is imbedded in O(2) continuously as a dense additive subgroup in
the way that a nonnegative real number a is identified with ?^??=_00 cxn2n such
that a = X)n=_00 an2n and an = 0 (Vn > N) for some N G ?, while a negative real number ?a with a as above is identified with X^^L.^l
? an)2n. Then, R
acts additively to O(2) by this addition. Furthermore, G := {2fc; k G ?} acts
multiplicatively to O(2) by
CX) 00
2fc ? a?2"= ? an_fc2". n= ? cx) n= ? oo
Thus, we have a group of actions on O(2) satisfying ((tl), (tf2)and (fl3) with G :=
{2k; fc G ?} and the equilibrium measure (1/2, l/2)z.
Theorem 1.1. O(2) is a numeration system with G = {2n; n G ?}.
We can express O(2) in the following different way. By a partition of the upper half plane EI := {z = ? + iy\ y > 0}, we mean a disjoint family of open sets
such that the union of their closures coincides with H. Let us consider the space
O(2)' of partitions a; of EI by open squares of the form (xi,X2) x (^1,2/2) with
X2 - x\ = ?/2 - y\ = y\ and yi G G such that (xx,x2) ? (2/1,2/2) ? ? implies
(xi,(xi +X2)/2) x (2/1/2,7/1) G a; (type 0) and
(0*1 +*2)/2,x2) x (yi/2,2/i) Go; (type 1).
This content downloaded from 62.122.72.154 on Wed, 18 Jun 2014 00:23:08 AMAll use subject to JSTOR Terms and Conditions
200 T. Kamae
y
I I I I 1 I 1 I I I I I M I 1 ? ? ? ? I I I I I I I I I I I I I I I I I I I 13
Fig 1. The tiling corresponding to ? ? 01.101 ?
An example of o; G O(2)' is shown in Figure 1. For ? G O(2)', let (a?,a?,...) be
the sequence of the types defined in (1) of the squares in ? intersecting with the
half vertical line from +0 + i to +0 + ?co and let (a_?, ct-2,...) be the sequence of the types of the squares in ? intersecting with the line segment from +0 + i to
+0. Then, ? is identified with S??:=_00 an2n. Note that replacing +0 by -0, we get
E!?L-oo?*2n suchthat (... ,a_2,a_?;a0,a?,...) ~ (... ,?-2,?-?\?^,??,...). The topology on O(2)' is defined so that ?? G O(2)' converges to a; G O(2)' as
? ?? co if for every R G ?, there exist Rn G ?? such that limn_>oo ?(?, Rn) = 0, where ? is the Hausdorff metric between sets R, R' C EI
p(R, R') := max{sup inf \z-z'\, sup inf \z - z'\).
z?RZfeR' z'^R'Z^R (2)
For ? G O(2)', t G R and ? G {2n; n G R}, ? + ? G O(2); and ?a; G O(2)' are
defined as the partitions
and
? + t := {(xi -t,x2 -t) ? (2/1,2/2); (^1^2) x (yi,y2) ? a;}
?a; := {(???,??2) ? (Xyi,Xy2); (^1,^2) x (2/1,2/2) ? a;}.
Let ? : O(2)' ?? O(2) be the identification mapping defined above. Then, ? is
a homeomorphism between O(2)' and O(2) such that ?(? + t) = ?(?) + t and
?(??) = ??(?) for any ? G O(2)', t G R and ? G {2n; n G Z}. Thus, O(2)' is
isomorphic to O(2) as a numeration system and will be identified with O(2).
This content downloaded from 62.122.72.154 on Wed, 18 Jun 2014 00:23:08 AMAll use subject to JSTOR Terms and Conditions
Numeration systems as dynamical systems 201
We generalize this construction. Let A be a nonempty finite set. An element in
A is called a color. An open rectangle (xi,x2) x (2/1*2/2) in EI is called an admissible
tile if
X2 - xi = 2/1 (3)
is satisfied (see Figure 2). In another word, an admissible tile is a rectangle (xi, ?2) x
(2/1,2/2) in EI such that the lower side has the hyperbolic length 1. Let Tl be the set
of admissible tiles in H.
A colored tiling a; is a subset of TZ ? A such that
(1) R ? Rf = 0 for_any (R, a) and (R', a') in ? with (R, a) f (Ri\ a'), and
(2) Ua?A U(?,e)?w R = EL
An element in Tl ? A is called a colored tile. We denote
dom(u>) := {R; (R, a) ? ? for some a G A}.
For R G ??p?(?), there exists a unique a G A such that (R, a) G a;, which is denoted
by u)(R) and is called the color of the tile R (in a;). Let R = (??, x2) ? (2/1,2/2)? We
call 2/2/2/1 the vertical size of the tile R which is denoted by S(R). Let O(?) be the set of colored tilings with colors in A. A topology is introduced
on O(?) so that a net {??}?e? C O(?) converges to o; G O(?) if for every (R, a) G ?, there exists (Rn,an) G ?? such that
an = a for any sufficiently large ? ? I and lim p(R, Rn) = 0, ?????
where ? is the Hausdorff metric defined in (2). For an admissible tile R := (xi,x2) x (2/1,2/2), t ? R and A6i+,we denote
R + t := (xi+t,x2-\-t) ? (2/1,2/2)
XR :- (Xxi,Xx2) x (?2/?,?2/2).
Note that they are also admissible tiles.
FIG 2. Admissible tiles.
This content downloaded from 62.122.72.154 on Wed, 18 Jun 2014 00:23:08 AMAll use subject to JSTOR Terms and Conditions
202 T. Kamae
For ? G O(?), t G R and ? G R+, we define ? + t G O(?) and ?a; G O(?) as follows:
? + t = {(R-t,a)', (R,a)eu>}
?? = {(??,a); (?, a) G a;}.
Thus, we define a continuous group action ?? + t of (?, t) G R+ ? R to ? G O(?). We construct compact metrizable subspaces of O (A) corresponding to weighted substitutions which are numeration systems. Though JJA > 2 is assumed in [7], we consider the case JjA = 1 as well.
2. Remarks on the notations
In this paper, the notations are changed in a large scale from the previous papers
[7], [8] and [9] of the author. The main changes are as follows:
(1) Here, the colored tilings are defined on the upper half plane EI, not on R2 as in
the previous papers. The multiplicative action here agree with the multiplication on
EI, while it agree with the logarithmic version of the multiplication at one coodinate
in the previous papers. Here, the tiles are open rectangles, not half open rectangles as in the previous papers.
(2) Here, we simplified the proof in [9] for the space of colored tilings coming from
weighted substitutions to be numeration systems by omitting the arguments on the
topological entropy.
(3) The roles of x-axis and y-axis for colored tilings are exchanged here and in [9] from those in [7] and [8].
(4) Here and in [9], the set of colors is denoted by A instead of S. Colors are denoted
by a, ol, a? (etc.) instead of s, s', G? (etc.).
(5) Here and in [9], the weighted substitution is denoted by (s, r) instead of (f, ?).
(6) Here and in [9], admissible tiles are denoted by R, R', Ri, R1 (etc.) instead of
S, S', Si, S* (etc.).
(7) Here and in [9], the terminology "primitive" for substitutions is used instead of
"mixing" in [7] and [8].
3. Weighted substitutions
A substitution s on a set A is a mapping A ?? A"1", where A+ = U?li &*. For
? G A+, we denote |?| := ? if ? G Ae, and ? with \?\ = ? is usually denoted by
?o?i ? * '?i-i with & G A. We can extend s to be a homomorphism A+ ?? A+ as
follows:
s(?):=s(??)s(??)'??s(??-?),
where ? G A^ and the right-hand side is the concatenations of tf(??)'s. We can define
s2, s3,... as the compositions of s : A+ ?? A+.
A weighted substitution (s,t) on A is a mapping A ?? A+ ? (0,1)+ such that
\s(a)\ = \t(a)\ and S*<|t(a)| T(a)i = * f?r anv ? ? A. Note that s is a substitution
on A. We define rn : A ?? (0,1)+ (n = 2,3,...) (depending on s) inductively by
Tn(a)k = r(a)irn-1(a(a)i)j
for any a G A and i, j, k with
0 < i < |s(a)|, 0 < j < \cn-l(*(a)i)\, k = ? \an-\a(a)h)\ + j. h<i
This content downloaded from 62.122.72.154 on Wed, 18 Jun 2014 00:23:08 AMAll use subject to JSTOR Terms and Conditions
Numeration systems as dynamical systems 203
Then, (s?,t?) is also a weighted substitution for ? = 2,3, ?
A substitution s on A is called primitive if there exists a positive integer ? such
that for any a, o! G A, s?(a)? = o! holds for some i with 0 <i < \s?(a)\. For a weighted substitution (s, t) on A, we always assume that
the substitution s is primitive. (4)
We define the base set ?(s, t) as the closed, multiplicative subgroup of R+ generated
by the set
G rn(a)i ; a G ?, ? = 0,1,... and 0 < i < \s?(a)\ ?
]^ such that &n(a)i = a J '
Example 3.1. Let A = {+, -} and (s,t) be a weighted substitution such that
+ -> (+,4/9)(- l/9)(+,4/9) - - (_,4/9)(+,l/9)(-4/9),
where we express a weighted substitution (s, t) by
a -> (s(a)?, t(a)0)(s(a)?,t(a)?) ? ? ? (a G A).
Then, 4/9 G ?(s,t) since s(+)0 = + and t(+)0 = 4/9. Moreover, 1/81 G ?(s,t) since s2(+)4 = 4- and t2(+)4 = 1/81. Since 4/9 and 1/81 do not have a common
multiplicative base, we have ?(s,t) = R+. This weighted substitution is discussed
in the following sections. The repetition of this weighted substitution starting at + is shown in Figure 3 by colored tiles. The substituted word of a color is represented as the sequence of colors of the connected tiles in below in order from left. The horizontal (additive) sizes of tiles are proportional to the weights and the vertical
(multiplicative) sizes are the inverse of the weights.
Let G := ?(s, t). Then, there exists a function g : A ?? R+ such that
g(a(a)i)G = g(a)r(a)iG (5)
for any a G A and 0 < i < \s(a)\. Note that if G = R+, then we can take g = 1. In the other case, we can define g by g(ao) = 1 and g(a) := rn(ao)? for some ? and i such that s?(a?)? = a, where ?q is any fixed element in A.
Let (s,t) be a weighted substitution satisfying (4). Let G = ?(s,t). Let g satisfy (5). Let O(s, t, g)f be the set of all elements ? in O (A) such that for any
((xi,x2) x (2/1,2/2),?) ?o;, we have
(I) 2/1 ? g(a)G, and
(II) (R\ s(a)?) G ? holds for i = 0,1,..., \s(a)\ - 1, where
i?? i
R1 := (?? -l? (x2 -
x\)^2r(a)j, ??-l? (x2 - X\)^2r(a)j)
j=0 j=0
? (T(a)iyi, 2/1).
A vertical line 7 := {?} ? (?co, co) is called a separating line of a; G O(s, r, g)' if for any (R, a) G ?, R ? ? = 0. Let O(s, t, g)" be the set of all ? G O(s, r, g)' which do not have a separating line and O(s, t, g) be the closure of O(s, T,g)n. Then, the action of G x R on O(s, t,g) satisfies (jjl). We usually denote O(s, r, 1) simply by
O(s,t).
This content downloaded from 62.122.72.154 on Wed, 18 Jun 2014 00:23:08 AMAll use subject to JSTOR Terms and Conditions
204 T. Kamae
4/9
+
1/9 4/9
+ +
Fig 3. The weighted substitution in Example 1.
Remark 3.2 ([7]). A nontrivial primitive substitution s : A ?? A+, where "non- trivial" means Sa?? ?s(a)? ^ 2, is considered as a weighted substitution in a canon- ical way. Let
M := (Jt{0 < i < \s(a)\; s(a)? = a'})a,a'eA
be the associate matrix. Let ? be the maximum eigen-value of M and ? := (?a)aeA be a positive column vector such that ?? = ??. Define weight t by
r(a)i = ?s(a)?
??a '
which is called the natural weight of s. Thus, we get a weighted substitution (s, r) which admits weight 1. We modify (s, t) if necessary in the following way. If there exists a G A with \s(a)\ = 1, so that a ?? (a', 1) is a part of (cr, r), then we
replace all the occurrences of a in the right hand side of "??" by o! and remove a from A together with the rule a ?? (a', 1) from (s,t). We continue this process until no a G A satisfies \s(a)\ = 1. After that if there exist a, a' G A such that
(s(a),t(a)) ? (s(a'),t(a')), then we identify them.
For example, the 2-adic expansion substitution 1 ?? 12, 2 ?? 12 corresponds to the weighted substitution 1 ?? (1,1/2)(1,1/2). The Thue-Morse substitution
This content downloaded from 62.122.72.154 on Wed, 18 Jun 2014 00:23:08 AMAll use subject to JSTOR Terms and Conditions
Numeration systems as dynamical systems 205
1 -? 12, 2 -> 21 corresponds to the weighted substitution 1 -? (1, l/2)(2,1/2), 2 -?
(2,1/2)(1,1/2). The Fibonacci substitution 1 ?? 12, 2 ?> 1 corresponds to the
weighted substitution 1 -? (1,?_1)(1,?~2), where ? = (1 + Vb)/2. The weighted substitution (s,t) obtained in this way satisfies that ?(s,t) =
{??; n G ?} and that g in (5) can be defined by g(a) = ?a (a G A). Dynami- cal systems coming from substitutions are discussed by many authors (see [2], for
example). Our weighted substitutions are a generalization of them.
Let (s, t) be a weighted substitution on A satisfying (4). Let g satisfy (5). Con-
sider O(s, t, g). We call the tile Rl in (II) the ?-th child of the tile (xi,x2) x (2/1,2/2), and (xi,x2) x (2/1,2/2) the mother of Rl. Note that the vertical size S(Rl) of
Rl coincides with the inverse of the weight t(a)?. If Rj is a child of Rj+i for
j = 0,1,..., fc ? 1. Then, the tile Rq is called a fc-th descendant of the tile Rk. If ?o is the z-th tile among the set of the fc-th descendants of Rk counting as 0,1,2, ? ? ?
from left, we call ?o the (k,i)-descendant of the tile i?*. In this case, we also say that Rk is the fc-th ancestor of ?o ?
Theorem 3.3. The space O(s, t, g) is a numeration system with G = ?(s,t).
Proof. We have already proved (#1) and (fl2) in Theorem 3 of [7]. Here we prove
(#3). Let O := O(s,t, g) and ??? be the equilibrium measure. Since ??. is the unique invariant probability measure under the additive action, it is also invariant under
the multiplicative action.
By Goodman [4], it is sufficient to prove that for any X E G with ? f 1, the
transformation ? ??? ?a; on O has the metrical entropy |^?| under ???, while it
has the metrical entropy less than | log ?| under any other G-invariant probability measure. D
Lemma 3.4. Let
S := {a; G O; ? has a separating line}
S? := {a; G O; y-axis is the separating line of ?}.
Then, we have
(i) S \ S? is dissipative with respect to the G-action, so that ?/(S \ S?) = 0 for any G-invariant probability measure ? on Vi.
(ii) For any o; G S?, a; restricted to the right quarter plane (?,??) ? (?,??) and to the left quarter plane (?00,0) ? (?,??) are cyclic individually with respect to the G-action. Hence, Ga; with respect to the G-action is either cyclic or conjugate to a 2-dimensional irrational rotation with a multiplicative time parameter.
(iii) S? is a finite union of minimal and equicontinuous sets with respect to the G-action. In fact, there is a mapping from the set of pairs a G A and i with
0<z<z+l< \s(a)\ onto the set of minimal sets in S??
Proof, (i) If the line ? = u is the separating line of ? G O, then ? = Xu is the
separating line of ?a;. Hence, S \ S? is dissipative.
(ii) Let o; G S?? Denote by a;"1" the restriction of ? to the right quarter plane
(0,00) ? (0,00), while by ?~ the restriction of a; to the left quarter plane (?00,0) ?
(?,??). Let (Rf)i^z be the sequence of tiles in ?sp?(?) such that Rf intersects with the upper half lines of ? = ?0, and Rf is a child of Rf+i for any z G ?
(db respectively). Let af :? u(Rf) be the colors of Rf (? respectively). Define
mappings s? from A to A by s+(a) = s(a)? and s_(a) = s(a)|<t(a)|_?. Since
0?(af) = af-i (i ? Z) (? respectively), the sequence (af)iez is periodic, which
This content downloaded from 62.122.72.154 on Wed, 18 Jun 2014 00:23:08 AMAll use subject to JSTOR Terms and Conditions
206 T. Kamae
also implies that the vertical sizes S(Rf) of Rf, which coincide with the inverses
of the weights r(a?+i)?, are also periodic in z G ? with the period, say r* which
is the minimum period of (af)iez (? respectively). Then, ?+ := rr (aj)^* is
the minimum (multiplicative) cycle of a;"1", while ?~ := tG (a$ )7\-( _.. is the
minimum (multiplicative) cycle of ?~, that is, ??? = a;* holds for ? = ?* and ?*
is the minimum among ? > 1 with this property (? respectively).
Therefore, ? is cyclic with respect to the G-action if ?+ and ?~ have a common
multiplicative base. In this case, the minimum cycle of ? is the minimum positive number ? such that ? = (?+)? = (X~)m holds for some positive integers ?,p?.
Otherwise, the G-action to Go; is conjugate to an 2-dimensional irrational rotation
with a multiplicative time parameter.
(iii) We use the notations in the proof of (zz). Take any pair (a, i) with a G A and
0<z<z + l< |s(a)|. Take any a/ G O having a tile R G ??t?(?') with o/(?) = a
such that the z/-axis passes in between the z-th child of R and the ? + 1-th child of
?. Let ?(a, i) be the set of limit points of ?a/ as ? G G tends to co. Note that this
does not depend on the choice of a/. Then, ?(a,?) is a closed G-invariant subset
of S0. Moreover, since the sequence (s"(s(a)?),s+(s(a)?+?))?=0,?,2,... enter into a
cycle after some time, ip(a,i) is minimal and equicontinuous with respect to the
G-action.
To prove that the mapping ip is onto, take any ? G S?? There exists ?? G
O(s, t, g)" which converges to ? as ? ?? co. We may assume that there exists a
pair (a, i) such that for any ? = 1,2,???, there exists R G ??p?(??) with a = o/n(?) such that the y-axis separetes the z-th child of R and the z + 1-th child of R. Then, ? G f(a, i), which proves that f is a mapping from the set of pairs (a, i) with a G A
and 0<z<z + l< \s(a)\ onto the set of minimal sets in S0 with respect to the
G-action. D
Example 3.5. Let ? with 0 < ? < 1 satisfy that logp/log(l - p) is irrational.
Let (s,t) be a weighted substitution on A = {1} such that 1 ?? (l,p)(l,l - p).
Then, ?(s,t) = R+ holds. Let O = O(s, t). In this case, elements in S0 are not
periodic, but almost periodic as shown in Figure 4. Then, the dynamical system
(S0, ? (? G R+)) is isomorphic to ((R/Z)2, ?? (X G R+)) with
Tx(x,y) = (x + logA/log(l/p), y -r logA/log(l/(l - p))).
Lemma 3.6. It holds that h?n(X) = \ ̂ ?| for any X G G. Let ?f 1 and ? be any other X-invariant probability measure on O, then h?(X) < \ logA|.
Proof. To prove the lemma, it is sufficient to prove the statements for ? > 1. Take
any G-invariant probability measure ? on O which attains the topological entropy of the multiplication by ?? G G with ?? > 1, that is, hv(Xi) = log ??. We assume
also that the G-action to O is ergodic with respect to v. Then by Lemma 3.4, either
?(S?) = 1 or ?(O \ S) = 1. In the former case, h?(X) = 0 holds for any Xe G since
the G-action on S? is equicontinuous by Lemma 3.4, which contradicts with the
assumption. Thus, we have ?/(O \ S) = 1.
For ? G O, let ?o(o;) G ??p?(?) be such that ?o (a;) = (^1,^2) x (2/1,2/2) with
Xi < 0 < x2 and 2/1 < 1 < 2/2? Take a0 G A such that
?/({? G O; ?(??(?)) = a0}) > 0.
This content downloaded from 62.122.72.154 on Wed, 18 Jun 2014 00:23:08 AMAll use subject to JSTOR Terms and Conditions
Numeration systems as dynamical systems 207
vertical sizes
1/(1-p)
vertical sizes
i/p
IF
Fig 4. An element in S? in Example 2.
Take oo '?= max{6 < 1; b G g(ao)G} (see (5)). Let
O? := { ? G O; the set {? G G; ?a;(??(?a;)) = a0}
is unbounded at 0 and co simultaneously }
O? := { ? G O?; ?o (?) = (xi,x2) x (2/1,2/2)
with yi = 60 and o>(?o(o;)) = a? }.
For a; G O?, let ??(a;) be the smallest X ? G with ? > 1 such that ?a; G O?? Define
a mapping ? : O? ?? O? by ?(?) := ??(a;)a;. For fc = 0,1,2, ? ? ? and i1 = 0,1, ? ? ? , |s*(a0)| - 1, let
P(k,i) := { ? G O0; ?0(?)~1 ??(?0(?)?)
is the (fc, z)-descendant of ?o(o;)}
(see Figure 5) and let
V:={P(k,i); fc = l,2,??? , 0<z<|afe(ao)| }
be a measurable partition of O0. Note that ??(a;) = Tk(a0)~l if o; G P(fc, z). Since ?^(O?) = 1 by the ergodicity and
O?= jj (J AP(fc,z),
P(k,i)eV 1<?<t*=(a0)-1 xeG
there exists a unique ?-invariant probability measure v$ on O? such that for any Borei set ? C O, we have
/boTfc(ao)t_1 iy0(X'1BnP(k,i))dX/X
This content downloaded from 62.122.72.154 on Wed, 18 Jun 2014 00:23:08 AMAll use subject to JSTOR Terms and Conditions
208 T. Kamae ?
ao
6?
a?
Fig d. ? e P(3,4) with ?0(?) = 13/3
with
if G = R+ and
C(v) := S -?oETkMi vo(P(k,i)) < co
P(kyi)?P
v(B) = C(v)-1 S S ??-^????))
with
P(k,i)ev ^G 60<?<60t*(a0). X
<?("):= S (-logrfc(ao)?/log/3)i/o(P(M))<oo
?(*:,?)???
if G = {/?"; ? ? ?} with /? > 1.
Since
S rfc(ao)? = land ? m(P(fc,t)) = l,
p(fc,t)ep p(fc,?)?P
we have
i/JP) := - JZ logn,(P(M))-i*(P(M))
p(k,i)ep
< - S logrfc(ao)?-^o(P(fc,?))
P(M)?T>
by the convexity of - log ?. The equality in (8) holds if and only if
H>(P(fc, i)) = Tk(a0)i (VP(fc, t) ? ?).
(6)
(7)
(8)
(9)
This content downloaded from 62.122.72.154 on Wed, 18 Jun 2014 00:23:08 AMAll use subject to JSTOR Terms and Conditions
Numeration systems as dynamical systems 209
By (6), (7), and (8), we have
H,0(V) = - S log^(P(M))-"o(P(M))<co.
p(k,i)er
For any o;, a/ G O? such that A/c(o;) and ?* (a/) belong to the same element in
V for fc = 0,1,2, ? ? ?, the horizontal position of ?o(o;), say (xi,x2), coincides with
that of ?o(o/). Therefore, ? and a/ restricted to (ari,x2) x (0, &o) coincide. In the
same way, if Afc(o;) and Afc(o/) belong to the same element in V for any fc G ?, then ?o := ?o (a;) = ?o (a/) holds and all the ancestors of ?o in ? and a/ coincide
as well as their colors. Therefore, ? and a/ restricted to the region covered by the
ancestors of ?o coincide. Hence, if ? or a/ does not have the separating lines, then
? = ?' holds.
Since ?/(S) = 0, we have ??(S ? O?) = 0. Hence, the above argument implies that V is a generator of the system (O?, ?, ?). Thus, /??0(?) = hUQ(A,V). It follows
from (8) that
K0(A) = K0(A,V)
<??0(?)
<- S logrfc(ao)i^o(P(fc,z)). (10)
p(k,i)ev
The equality in the above that
?^(?) = - S logTk(ao)i'Vo(P(k,i))
P(k,i)ev
holds if and only if (AnV)n?.z is an independent sequence with respect to i/o satis-
fying (9). Since
MAi)/logAi = ??)
?Oo ???<??)
?,?(?)
-Ep(fc,i)eplog^fc(?o)i ? vo(P(k,?))
'
hv(Xi) < ^?? follows from (10), while the equality holds if and only if (AnV)n?:Z is an independent sequence with respect to i/o satisfying (9). Let this probability measure be ?. Then, it is not difficult to prove that ? is invariant under the additive
action. Hence, the uniqueness of such measure ([7]) proves ? = ?^, which completes the proof of Lemma 3.6 and Theorem 3.3. D
The following Theorem 3.7 follows from a known result about the spectrum of unitary operators corresponding to the affine action (Lemma 11.6 of [13], for
example).
Theorem 3.7 ([10]). Let O be a numeration system with G = R+, that is, with
the multiplicative R+-action. Then, the additive action on the probability space O
with respect to ??\ has a pure Lebesgue spectrum.
This content downloaded from 62.122.72.154 on Wed, 18 Jun 2014 00:23:08 AMAll use subject to JSTOR Terms and Conditions
210 T. Kamae
4. ?-function
Here, we listed only the results on the ?-functions. For the proof, refer [10]. Let O := O(s,t,a) satisfying (4) and (5). For a G C, we define the associated
matrices on the suffix set A ? A as follows:
Ma := ? ? T^A (?)
?a,+ := (?s(?)?=a< r(a)J)a|Q/?A
ia,- := (la(a)k(a)|_1=a' T(a)k(?)|-l)a^,? ?'?A
Let ? be the set of closed orbits of O with respect to the action of G. That is, ? is the family of subsets ? of O such that ? = Gu for some o; G O with ?? = ? for
some X ? G with ? > 1. We call ? as above a multiplicative cycle ?? ?. The minimum
multiplicative cycle of ? is denoted by c(?). Note that c(?) exists since ?? f ? for
any a; G O and XeG with 1 < ? < min{r(a)~1; a G A, 0 < z < |r(a)|}. We say that ? G ? has a separating line if o; G ? has a separating line. Note that
in this case, the separating line is necessarily the y-axis and is in common among a; G ?. Denote by ?? the set of ? G ? with the separating line.
Define the ?-function of G-action to O by
&(*):= H(l-c(t)-*r\ (12)
where the infinite product converges for any a G C with ?1(a) > 1. It is extended
to the whole complex plane by the analytic extension.
Theorem 4.1. We have
det(/-Ma,+)det(/-MQ)_) ??
=-det(/-MQ)- ?S?(a)'
where
CE0(a):= ??1"^)"0)"1 ????
is a finite product with respect to ? e ??.
Theorem 4.2. (i) ??(a) f 0 i}Tl(a) f 0.
(ii) In the region Tl(a) f 0, a is a pole of ??(a) with multiplicity k if and only
if it is a zero of det(I ? Ma) with multiplicity fc for any fc = 1,2, ?
(iii) 1 is a simple pole of ??(a).
Theorem 4.3. For O = O(s,t/,#), if ? (s,t) = {??; ? G ?} with ? > 1, then
there exist polynomials p,q G Z[z] such that ??(a) = p(X?t)/q(Xa)- Conversely,
if ??(a) = p(XQ)/q(Xa) holds for some polynomials p,q G Z[z] and X > 1, then
?(s,t) = {Xkn; n G ?} for some positive integer fc.
Theorem 4.4. If ?(s,t) = {Xn; n G ?}, then X is an algebraic number.
Acknowledgment
The author thanks his old friend, Prof. Mike Keane for his useful discussions and
encouragements to develop this research for more than 10 years.
This content downloaded from 62.122.72.154 on Wed, 18 Jun 2014 00:23:08 AMAll use subject to JSTOR Terms and Conditions
Numeration systems as dynamical systems 211
References
[1] Arnoux, P. and Ito, S. (2001). Pisot substitutions and Rauzy fractals. Bull.
Belg. Math. Soc. Simon Stevin 8, 2, 181-207. MR1838930
[2] Fogg, N. P. (2002). Substitutions in Dynamics, Arithmetics and Combi-
natorics. Lecture Notes in Mathematics, Vol. 1794. Springer-Verlag, Berlin.
MR1970385
[3] Dunford, N. AND Schwartz, J.T. (1963). Linear Operators II, Interscience
Publishers John Wiley & Sons, New York-London. MR0188745
[4] Goodman, T. N. T. (1971). Relating topological entropy and measure en-
tropy. Bull. London Math. Soc. 3, 176-180. MR289746
[5] Gjini, N. and Kamae, T. (1999). Coboundary on colored tiling space as
Rauzy fractal. Indag. Math. (N.S.) 10, 3, 407-421. MR1819898
[6] Ito, S. and Takahashi, Y. (1974). Markov subshifts and realization of /?-
expansions. J. Math. Soc. Japan 26, 33-55. MR346134
[7] Kamae, ?. (1998). Linear expansions, strictly ergodic homogeneous cocycles and fractals. Israel J. Math. 106, 313-337. MR1656897
[8] Kamae, T. (2001). Stochastic analysis based on deterministic Brownian mo-
tion. Israel J. Math. 125, 317-346. MR1853816
[9] Kamae, T. (2006). Numeration systems, fractals and stochastic processes, Israel J. Math, (to appear).
[10] Kamae, T. (2005) Numeration systems as dynamical systems. Preprint, avail-
able at http://wwwl4.plala.or.jp/kamae.
[11] Petersen, K. (1983). Ergodic Theory. Cambridge Studies in Advanced Math-
ematics, Vol. 2. Cambridge University Press, Cambridge. MR833286
[12] Plessner, A. (1941). Spectral theory of linear operators I, Uspekhi Matem.
Nauk 9, 3-125. MR0005798
[13] Starkov, A. N. (2000). Dynamical Systems on Homogeneous Spaces, Trans-
lations of Mathematical Monographs 190, Amer. Math. Soc. MR1746847
[14] Walters, P. (1975). Ergodic Theory ? Introductory Lectures, Lecture Notes
in Mathematics 458. Springer-Verlag, Berlin-New York. MR0480949
This content downloaded from 62.122.72.154 on Wed, 18 Jun 2014 00:23:08 AMAll use subject to JSTOR Terms and Conditions