Fractals in Probability and Analysis
CHRISTOPHER J. BISHOP
Stony Brook University, Stony Brook, NY
YUVAL PERES
Microsoft Research, Redmond, WA
University Printing House, Cambridge CB2 8BS, United Kingdom One Liberty Plaza, 20th Floor, New York, NY 10006, USA 477 Williamstown Road, Port Melbourne, VIC 3207, Australia 4843/24, 2nd Floor, Ansari Road, Daryaganj, Delhi  110002, India 79 Anson Road, #0604/06, Singapore 079906
Cambridge University Press is part of the University of Cambridge.
It furthers the University’s mission by disseminating knowledge in the pursuit of education, learning, and research at the highest international levels of excellence.
www.cambridge.org Information on this title: www.cambridge.org/9781107134119
10.1017/9781316460238
c Christopher J. Bishop and Yuval Peres 2017
This publication is in copyright. Subject to statutory exception and to the provisions of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press.
First published 2017
Printed in the United Kingdom by Clays, St Ives plc
A catalogue record for this publication is available from the British Library .
ISBN 9781107134119 Hardback
Cambridge University Press has no responsibility for the persistence or accuracy of URLs for external or thirdparty Internet Web sites referred to in this publication and does not guarantee that any content on such Web sites is, or will remain, accurate or appropriate.
Contents
Preface 
page 
1 Minkowski and Hausdorff dimensions 
1 
1.1 Minkowski dimension 
1 
1.2 Hausdorff dimension and the Mass Distribution Principle 
4 
1.3 Sets deﬁned by digit restrictions 
9 
1.4 Billingsley’s Lemma and the dimension of measures 
17 
1.5 Sets deﬁned by digit frequency 
21 
1.6 Slices 
26 
1.7 Intersecting translates of Cantor sets * 
29 
1.8 Notes 
34 
1.9 Exercises 
36 
2 Selfsimilarity and packing dimension 
45 
2.1 Selfsimilar sets 
45 
2.2 The open set condition is sufﬁcient 
51 
2.3 Homogeneous sets 
54 
2.4 Microsets 
57 
2.5 Poincar´e sets * 
61 
2.6 Alternative deﬁnitions of Minkowski dimension 
67 
2.7 Packing measures and dimension 
71 
2.8 When do packing and Minkowski dimension agree? 
74 
2.9 Notes 
76 
2.10 Exercises 
78 
3 Frostman’s theory and capacity 
83 
3.1 Frostman’s Lemma 
83 
3.2 The dimension of product sets 
88 
3.3 Generalized Marstrand Slicing Theorem 
90 
3.4
Capacity and dimension
93
viii 
Contents 

3.5 Marstrand’s Projection Theorem 
95 

3.6 Mapping a tree to Euclidean space preserves capacity 
100 

3.7 Dimension of random Cantor sets 
104 

3.8 Notes 
112 

3.9 Exercises 
115 

4 
Selfafﬁne sets 
119 
4.1 Construction and Minkowski dimension 
119 

4.2 The Hausdorff dimension of selfafﬁne sets 
121 

4.3 A dichotomy for Hausdorff measure 
125 

4.4 The Hausdorff measure is inﬁnite * 
127 

4.5 Notes 
131 

4.6 Exercises 
133 

5 
Graphs of continuous functions 
136 
5.1 Holder¨ continuous functions 
136 

5.2 The Weierstrass function is nowhere differentiable 
140 

5.3 estimates Lower Holder¨ 
145 

5.4 Notes 
149 

5.5 Exercises 
151 

6 
Brownian motion, Part I 
160 
6.1 Gaussian random variables 
160 

6.2 L´evy’s construction of Brownian motion 
163 

6.3 Basic properties of Brownian motion 
167 

6.4 Hausdorff dimension of the Brownian path and graph 
172 

6.5 Nowhere differentiability is prevalent 
176 

6.6 Strong Markov property and the reﬂection principle 
178 

6.7 Local extrema of Brownian motion 
180 

6.8 Area of planar Brownian motion 
181 

6.9 General Markov processes 
183 

6.10 Zeros of Brownian motion 
185 

6.11 Harris’ inequality and its consequences 
189 

6.12 Points of increase 
192 

6.13 Notes 
196 

6.14 Exercises 
199 

7 
Brownian motion, Part II 
201 
7.1 Dimension doubling 
201 

7.2 The Law of the Iterated Logarithm 
206 

7.3 Skorokhod’s Representation 
209 

7.4 Donsker’s Invariance Principle 
216 
7.5
Harmonic functions and Brownian motion in R ^{d}
221
Contents 
ix 

7.6 The maximum principle for harmonic functions 
226 

7.7 The Dirichlet problem 
227 

7.8 Polar points and recurrence 
228 

7.9 Conformal invariance * 
230 

7.10 Capacity and harmonic functions 
235 

7.11 Notes 
239 

7.12 Exercises 
241 

8 Random walks, Markov chains and capacity 
244 

8.1 Frostman’s theory for discrete sets 
244 

8.2 Markov chains and capacity 
250 

8.3 Intersection equivalence and return times 
254 

8.4 Lyons’ Theorem on percolation on trees 
258 

8.5 Dimension of random Cantor sets (again) 
260 

8.6 Brownian motion and Martin capacity 
264 

8.7 Notes 
266 

8.8 Exercises 
266 

9 Besicovitch–Kakeya sets 
270 

9.1 Existence and dimension 
270 

9.2 Splitting triangles 
276 

9.3 Fefferman’s Disk Multiplier Theorem * 
278 

9.4 Random Besicovitch sets 
286 

9.5 Projections of selfsimilar Cantor sets 
290 

9.6 The open set condition is necessary * 
297 

9.7 Notes 
302 

9.8 Exercises 
305 

10 The Traveling Salesman Theorem 
313 

10.1 Lines and length 
313 

10.2 The β numbers 
318 

10.3 Counting with dyadic squares 
322 

10.4 β and µ are equivalent 
325 

10.5 β sums estimate minimal paths 
329 

10.6 Notes 
334 

10.7 Exercises 
337 

Appendix A 
Banach’s FixedPoint Theorem 
343 
Appendix B 
Frostman’s Lemma for analytic sets 
353 
Appendix C 
Hints and solutions to selected exercises 
360 
References 
379 
Index
396
Preface
The aim of this book is to acquaint readers with some fractal sets that arise naturally in probability and analysis, and the methods used to study them. The book is based on courses taught by the authors at Yale, Stony Brook University, the Hebrew University and UC Berkeley. We owe a great debt to our advisors, Peter Jones and Hillel Furstenberg; thus the book conveys some of their per spectives on the subject, as well as our own. We have made an effort to keep the book selfcontained. The only prerequi site is familiarity with measure theory and probability at the level acquired in a ﬁrst graduate course. The book contains many exercises of varying difﬁculty. We have indicated with a “•” those for which a solution, or a hint, is given in Appendix C. A few sections are technically challenging and not needed for subsequent sections, so could be skipped in the presentation of a given chapter. We mark these with a “*” in the section title.
Acknowledgments: We are very grateful to Ton´ci Antunovi´c, Subhroshekhar Ghosh and Liat Kessler for helpful comments and crucial editorial work. We also thank Ilgar Eroglu, Hrant Hakobyan, Michael Hochman, Nina Holden, Pertti Mattila, Elchanan Mossel, Boris Solomyak, Perla Sousi, Ryokichi Tan aka, Tatiana Toro, B´alint Vir´ag, Samuel S. Watson, Yimin Xiao and Alex Zhai for useful comments. Rich´ard Balka carefully read the entire manuscript and provided hundreds of detailed corrections and suggestions. Many thanks to David Tranah and Sam Harrison at Cambridge University Press for numerous helpful suggestions. Finally, we dedicate this book to our families: Cheryl, David and Emily Bishop, and Deborah, Alon and Noam Peres; without their support and under standing, it would have taken even longer to write.
1
Minkowski and Hausdorff dimensions
In this chapter we will deﬁne the Minkowski and Hausdorff dimensions of a set
and will compute each in a few basic examples. We will then prove Billings ley’s Lemma and the Law of Large Numbers. These allow us to deal with more sophisticated examples: sets deﬁned in terms of digit frequencies, ran dom slices of the Sierpi´nski gasket, and intersections of random translates of the middle thirds Cantor set with itself. Both Minkowski and Hausdorff di mensions measure how efﬁciently a set K can be covered by balls. Minkowski dimension requires that the covering be by balls all of the same radius. This
makes it easy to compute, but it lacks certain desirable properties. In the def inition of Hausdorff dimension we will allow coverings by balls of different radii. This gives a better behaved notion of dimension, but (as we shall see) it
is
usually more difﬁcult to compute.
1.1 Minkowski dimension
A
be
is
denote the minimal number of sets of diameter at most ε needed to cover K . We deﬁne the upper Minkowski dimension as
the same as being a bounded set. For a totally bounded set K , let N (K , ε)
covered by a ﬁnite number of balls of diameter ε. For Euclidean space, this
subset K of a metric space is called totally bounded if for any ε > 0, it can
dim _{M} (K ) = limsup
ε→0
and the lower Minkowski dimension
dim _{M} (K ) = liminf
ε→0
1
log N (K , ε)
log1/ε
^{,}
log N (K , ε)
log1/ε
^{.}
05 Feb 2017 at 20:23:35,
02
2
Minkowski and Hausdorff dimensions
If the two values agree, the common value is simply called the Minkowski dimension of K and denoted by dim _{M} (K ). When the Minkowski dimension of a set K exists, the number of sets of diameter ε needed to cover K grows
_{l}_{i}_{k}_{e} _{ε} −dim _{M} ( K )+ o(1)
_{a}_{s} _{ε} _{→} _{0}_{.}
We get the same values of dim _{M} (K ) and dim _{M} (K ) if we replace N (K , ε)
by N _{B} (K , ε), which is the number of closed balls of radius ε needed to cover
K . This is because N _{B} (K , ε) ≤ N (K , ε) ≤ N (K , ε/2)
ball of at most twice the diameter and any ball of radius ε/2 has diameter at most ε; strict inequality could hold in a metric space). For subsets of Euclidean space we can also count the number of axisparallel squares of side length ε needed to cover K , or the number of such squares taken from a grid. Both pos sibilities give the same values for upper and lower Minkowski dimension, and for this reason Minkowski dimension is sometimes called the box counting dimension . It is also easy to see that a bounded set A and its closure A satisfy dim _{M} (A ) = dim _{M} (A ) and dim _{M} (A ) = dim _{M} (A). If X is a set and x , y ∈ X implies x − y  ≥ ε, we say X is ε separated. Let N _{s}_{e}_{p} (K , ε) be the number of elements in a maximal εseparated subset X of K . Clearly, any set of diameter ε/2 can contain at most one point of an ε separated set X , so N _{s}_{e}_{p} (K , ε) ≤ N (K , ε/2). On the other hand, every point of K is within ε of a maximal εseparated subset X (otherwise add that point to X ). Thus N (K , ε) ≤ N _{s}_{e}_{p} (K , ε). Therefore replacing N (K , ε) by N _{s}_{e}_{p} (K , ε) in the deﬁnition of upper and lower Minkowski dimension gives the same values (and it is often easier to give a lower bound in terms of separated sets).
(any set is contained in a
Example 1.1.1 Suppose that K is a ﬁnite set. Then N (K , ε) is bounded and dim _{M} (K ) exists and equals 0.
Suppose K = [0, 1]. Then at least 1/ε intervals of length ε are
needed to cover K and clearly ε ^{−}^{1} +1 sufﬁce. Thus dim _{M} (K ) exists and equals
Example 1.1.2
1. Similarly, any bounded set in R ^{d} with interior has Minkowski dimension d .
Example 1.1.3 Let C be the usual middle thirds Cantor set obtained as fol
lows. Let C ^{0} = [0, 1] and deﬁne C ^{1} = [0, _{3} ] ∪ [ ^{2} ,1] ⊂ C ^{0} by removing the
central interval of length
_{3} 1 . In general, C ^{n} is a union of 2 ^{n} intervals of length
3 ^{−} ^{n} and C ^{n} ^{+}^{1} is obtained by removing the central third of each. This gives a decreasing nested sequence of compact sets whose intersection is the desired set C. The construction gives a covering of C that uses 2 ^{n} intervals of length 3 ^{−} ^{n} . Thus for 3 ^{−} ^{n} ≤ ε < 3 ^{−} ^{n}^{+}^{1} we have
1
3
N (C ,ε) ≤ 2 ^{n} ,
05 Feb 2017 at 20:23:35,
02
1.1 Minkowski dimension
3
Figure 1.1.1 The Cantor middle thirds construction.
and hence
dim _{M} (C ) ≤ ^{l}^{o}^{g}^{2} log3 ^{.}
Conversely, the centers of the n th generation intervals form a 3 ^{−} ^{n} separated set
of size 2 ^{n} , so N _{s}_{e}_{p} (C ,3 ^{−} ^{n} ) ≥ 2 ^{n} . Thus
dim _{M} (C ) ≥ ^{l}^{o}^{g}^{2} log3 = log _{3} 2.
Therefore the Minkowski dimension exists and equals this common value. If
at each stage we remove the middle α (0 < α < 1) we get a Cantor set C _{α} with
Minkowski dimension log2/(log2 + log
1
1−α ^{)}^{.}
Example 1.1.4 Consider K = {0}∪{1,
1
2 ^{,}
1
1
3 ^{,} _{4} ^{,}
}. Observe that
1 
1 
1 
1 
n − 1 ^{−} n ^{=} n (n − 1) ^{>} 
n ^{2} ^{.} 
So, for ε > 0, if we choose n so that (n + 1) ^{−}^{2} < ε ≤ n ^{−}^{2} , then n ≤ ε ^{−}^{1}^{/}^{2} and
n distinct intervals of length ε are needed to cover the points 1,
_{n} . The
interval [0, _{n} _{+}_{1} ] can be covered by n + 1 additional intervals of length ε. Thus
1
2 ^{,}
1
,
1
ε ^{−}^{1}^{/}^{2} ≤ N (K ,ε) ≤ 2ε ^{−}^{1}^{/}^{2} + 1.
Hence dim _{M} (K ) = 1/2.
This example illustrates a drawback of Minkowski dimension: ﬁnite sets have dimension zero, but countable sets can have positive dimension. In par ticular, it is not true that dim _{M} ( ^{} _{n} E _{n} ) = sup _{n} dim _{M} (E _{n} ), a useful property for a dimension to have. In the next section, we will introduce Hausdorff di mension, which does have this property (Exercise 1.6). In the next chapter, we will introduce packing dimension, which is a version of upper Minkowski dimension forced to have this property.
05 Feb 2017 at 20:23:35,
02
4
Minkowski and Hausdorff dimensions
1.2 Hausdorff dimension and the Mass Distribution Principle
Given any set K in a metric space, we deﬁne the α dimensional Hausdorff content as
α (K ) = inf _{∑} U _{i}  ^{α} : K ⊂ ^{}
H
∞
i
i
U i ,
where {U _{i} } is a countable cover of K by any sets and E  denotes the diameter of a set E .
Deﬁnition 1.2.1
The Hausdorff dimension of K is deﬁned to be
dim(K ) = inf{α : H
∞
α (K ) = 0}.
More generally we deﬁne
α (K ) = inf _{∑} U _{i}  ^{α} : K ⊂ ^{} U _{i} ,U _{i}  < ε ,
H
ε
i
i
where each U _{i} is now required to have diameter less than ε. The α dimensional Hausdorff measure of K is deﬁned as
H ^{α} (K ) = lim
ε→0 ^{H}
ε
α
(K ).
This is an outer measure; an outer measure on a nonempty set X is a function μ ^{∗} from the family of subsets of X to [0, ∞] that satisﬁes
• μ ^{∗} (0/ ) = 0,
• μ
• μ ^{∗} ( ^{} ^{∞} _{=}_{1} A _{j} ) ≤ _{∑} ^{∞}
^{∗} (A ) ≤ μ ^{∗} (B ) if A ⊂ B ,
j
j _{=}_{1}
μ ^{∗} (A _{j} ).
For background on real analysis see Folland (1999). The αdimensional Haus dorff measure is even a Borel measure in R ^{d} ; see Theorem 1.2.4 below. When α = d ∈ N, then H ^{α} is a constant multiple of L _{d} , d dimensional Lebesgue measure.
α (K )
If we admit only open sets in the covers of K , then the value of H
ε
does not change. This is also true if we only use closed sets or only use convex
sets. Using only balls might increase H
α by at most a factor of 2 ^{α} , since any
ε
set K is contained in a ball of at most twice the diameter. Still, the values for
which H ^{α} (K ) = 0 are the same whether we allow covers by arbitrary sets or only covers by balls.
Deﬁnition 1.2.2 Let μ ^{∗} be an outer measure on X . A set K in X is μ ^{∗}  measurable , if for Every set A ⊂ X we have
H ^{α} (A ) = μ ^{∗} (A ∩ K )+ μ ^{∗} (A ∩ K ^{c} ).
05 Feb 2017 at 20:23:35,
02
1.2
Hausdorff dimension and the Mass Distribution Principle
5
Deﬁnition 1.2.3 Let (Ω, d ) be a metric space. An outer measure μ on Ω is called a metric outer measure if dist(A , B ) > 0 =⇒ μ(A ∪ B) = μ(A )+ μ(B ), where A and B are two subsets of Ω.
Since Hausdorff measure H ^{α} is clearly a metric outer measure, the follow ing theorem shows that all Borel sets are H ^{α} measurable. This implies that H ^{α} is a Borel measure (see Folland (1999)).
Theorem 1.2.4 Let μ be a metric outer measure. Then all Borel sets are μ measurable.
Proof It sufﬁces to show any closed set K is μmeasurable, since the measur able sets form a σ algebra. So, let K be a closed set. We must show for any set A ⊂ Ω with μ(A ) < ∞,
μ(A ) ≥ μ(A ∩ K )+ μ(A ∩ K ^{c} ). 
(1.2.1) 

Let B _{0} = 0/ and for n ≥ 1 deﬁne B _{n} = { x ∈ A : dist(x , K ) > 
1 _{n} }, so that 

∞ 


B _{n} = A ∩ K ^{c} 

n 
=1 

(since K is closed). Since μ is a metric outer measure and B _{n} ⊂ A \ K , 

μ(A ) ≥ μ[(A ∩ K ) ∪ B _{n} ] = μ(A ∩ K )+ μ(B _{n} ). 
(1.2.2) 
For all m ∈ N, the sets D _{n} = B _{n} \ B _{n}_{−}_{1} satisfy
m
∑ =1 μ(D _{2} _{j} ) = μ
j
=1 D _{2} _{j} ≤ μ(A ),∀m,
m
j
since if x ∈ B _{n} , and y ∈ D _{n} _{+}_{2} , then
dist(x , y ) ≥ dist(x , K ) − dist(y, K ) ≥ ^{1} _{n} −
^{1}
n + 1 ^{.}
Similarly _{∑} ^{m} _{=}_{1} μ(D _{2} _{j} _{−}_{1} ) ≤ μ(A ). So _{∑} ^{∞} _{=}_{1} μ(D _{j} ) < ∞. The inequality
j
j
μ(B _{n} ) ≤ μ(A ∩ K ^{c} ) ≤ μ(B _{n} )+
∞
∑
j = n +1
μ(D _{j} )
implies that μ(B _{n} ) → μ(A ∩ K ^{c} ) as n → ∞. Thus letting n → ∞ in (1.2.2) gives
(1.2.1).
05 Feb 2017 at 20:23:35,
02
6
Minkowski and Hausdorff dimensions
The construction of Hausdorff measure can be made a little more general by considering a positive, increasing function ϕ on [0, ∞) with ϕ(0) = 0. This is called a gauge function and we may associate to it the Hausdorff content
ϕ
H
∞
(K ) = inf _{∑}
i
ϕ(U _{i} ) : K ⊂ ^{}
i
i ;
U
then H
ϕ (K ) are deﬁned as before. The case
ϕ(t ) = t ^{α} is just the case considered above. We will not use other gauge func tions in the ﬁrst few chapters, but they are important in many applications, e.g., see Exercise 1.59 and the Notes for Chapter 6.
ϕ (K ), and H ^{ϕ} (K ) = lim _{ε}_{→}_{0} H
ε
ε
Lemma 1.2.5
If H ^{α} (K ) < ∞ then H ^{β} (K ) = 0 for any β > α .
Proof
α
It follows from the deﬁnition of H
ε
that
β
H
ε
(K ) ≤ ε ^{β}^{−}^{α} H
α
ε
(K
),
which gives the desired result as ε → 0.
Thus if we think of H ^{α} (K ) as a function of α, the graph of H ^{α} (K ) versus α shows that there is a critical value of α where H ^{α} (K ) jumps from ∞ to 0. This critical value is equal to the Hausdorff dimension of the set. More generally we have:
Proposition 1.2.6 For every metric space E we have
and therefore
H ^{α} (E ) = 0
⇔
α
H
∞
(E ) = 0
dim E = inf{α : H ^{α} (E ) = 0} = inf{α : H ^{α} (E ) < ∞}
= sup{α : H ^{α} (E ) > 0} = sup{α : H ^{α} (E ) = ∞}.
(E ) = 0,
Proof Since H ^{α} (E ) ≥
then for every δ > 0 there is a covering of E by sets { E _{k} } with _{∑} _{=}_{1} E _{k}  ^{α} < δ. These sets have diameter less than δ ^{1}^{/}^{α} , hence H _{1}_{/}_{α} (E ) < δ . Letting δ ↓ 0
α
H
∞
(E ), it sufﬁces to prove “⇐”. If H
α
∞
α
δ
∞
k
yields H ^{α} (E ) = 0, proving the claimed equivalence. The equivalence readily implies that dim E = inf{α : H ^{α} (E ) = 0} = sup{α : H ^{α} (E ) > 0}. The other conclusions follow from Lemma 1.2.5.
The following relationship to Minkowski dimension is clear
dim(K ) ≤ dim _{M} (K ) ≤ dim _{M} (K ).
05 Feb 2017 at 20:23:35,
02
(1.2.3)
1.2
Hausdorff dimension and the Mass Distribution Principle
7
Indeed, if B _{i} = B (x _{i} , ε/2) are N (K , ε) balls of radius ε/2 and centers x _{i} in K that cover K , then consider the sum
S ε =
N ( K ,ε)
∑  B _{i}  ^{α} = N (K ,ε)ε ^{α} = ε ^{α}^{−} ^{R} ^{ε} ,
i=1
_{w}_{h}_{e}_{r}_{e} _{R} ε _{=} log N ( K ,ε)
log(1/ε)
. If α > liminf _{ε}_{→}_{0} R _{ε} = dim _{M} (K ) then inf _{ε}_{>}_{0} S _{ε} = 0.
Strict inequalities in (1.2.3) are possible.
Example 1.2.7 Example 1.1.4 showed that K = {0} ^{} _{n} { ^{1} _{n} } has Minkowski
dimension _{2} . However, any countable set has Hausdorff dimension 0, for if we
} and cover the n th point by a ball of diameter
as small as we wish for any α > 0. Thus K is
a compact set for which the Minkowski dimension exists, but is different from
the Hausdorff dimension.
δ _{n} = ε2 ^{−} ^{n} we can make _{∑} _{n} δ ^{α}
enumerate the points { x _{1} , x _{2} ,
1
n
Lemma 1.2.8 (Mass Distribution Principle) If E supports a strictly positive Borel measure μ that satisﬁes
μ(B (x , r )) ≤ Cr ^{α} ,
for some constant 0 < C < ∞ and for every ball B (x , r ), then H ^{α} (E ) ≥ H
∞
α (E )
≥ μ(E )/C. In particular, dim(E ) ≥
α .
Let {U _{i} } be a cover of E . For { r _{i} }, where r _{i} > U _{i} , we look at the
Proof
following cover: choose x _{i} in each U _{i} , and take open balls B(x _{i} , r _{i} ). By as
sumption,
μ(U _{i} ) ≤ μ(B (x _{i} , r _{i} )) ≤ Cr _{i} ^{α} .
We deduce that μ (U _{i} ) ≤ C U _{i}  ^{α} , whence
∑
i
U _{i}  ^{α} ≥ _{∑}
i
Thus H ^{α} (E ) ≥ H
∞
α (E ) ≥ μ(E )/C .
μ(U _{i} )
C
_{≥}
μ(E )
C
^{.}
We note that upper bounds for Hausdorff dimension usually come from ﬁnd ing explicit coverings of the set, but lower bounds are proven by constructing an appropriate measure supported on the set. Later in this chapter we will gen eralize the Mass Distribution Principle by proving Billingsley’s Lemma (Theo rem 1.4.1) and will generalize it even further in later chapters. As a special case of the Mass Distribution Principle, note that if A ⊆ R ^{d} has positive Lebesgue d measure then dim(A ) = d .
05 Feb 2017 at 20:23:35,
02
8
Minkowski and Hausdorff dimensions
Example 1.2.9 Consider the Cantor set E obtained by replacing the unit square in the plane by four congruent subsquares of side length α < 1/2 and continuing similarly. See Figure 1.2.1. We can cover the set by 4 ^{n} squares of diameter ^{√} 2·α ^{n} . Thus
dim _{M} (E ) ≤ lim
n
→∞
log4 ^{n}
log4
log( ^{√} 2α ^{n} ) ^{=}
−
− logα ^{.}
On the other hand, it is also easy to check that at least 4 ^{n} sets of diameter α ^{n} are needed, so
dim _{M} (E ) ≥
log4
− logα ^{.}
Thus the Minkowski dimension of this set equals β = − log4/ logα.
Figure 1.2.1 Four generations of a Cantor set.
We automatically get dim(E ) ≤ β and we will prove the equality using Lemma 1.2.8. Let μ be the probability measure deﬁned on E that gives each nth generation square the same mass (namely 4 ^{−} ^{n} ). We claim that
μ(B (x , r )) ≤ Cr ^{β} ,
for all disks and some 0 < C < ∞. To prove this, suppose B = B (x , r ) is some disk hitting E and choose n so that α ^{n} ^{+}^{1} ≤ r < α ^{n} . Then B can hit at most 4 of the n th generation squares and so, since α ^{β} = 1/4,
μ(B ∩ E ) ≤ 4·4 ^{−} ^{n} = 4α ^{n}^{β} ≤ 16 r ^{β} .
Example 1.2.10 Another simple set for which the two dimensions agree and are easy to compute is the von Koch snowﬂake. To construct this we start with an equilateral triangle. At each stage we add to each edge an equilateral triangle pointing outward of side length 1/3 the size of the current edges and centered on the edge. See Figure 1.2.2 for the ﬁrst four iterations of this process. The boundary of this region is a curve with dimension log4/ log3 (see Theorem 2.2.2). We can also think of this as a replacement construction, in which at
05 Feb 2017 at 20:23:35,
02
1.3 Sets deﬁned by digit restrictions
9
each stage, a line segment is replaced by an appropriately scaled copy of a polygonal curve.
Figure 1.2.2 Four generations of the von Koch snowﬂake.
Even for some relatively simple sets the Hausdorff dimension is still un known. Consider the Weierstrass function (Figure 1.2.3)
f _{α}_{,}_{b} (x ) =
∞
∑
n=1
b ^{−} ^{n} ^{α} cos(b ^{n} x),
where b > 1 is real and 0 < α < 1. It is conjectured that the Hausdorff dimen sion of its graph is 2 − α, and this has been proven when b is an integer; see the discussion in Example 5.1.7. On the other hand, some sets that are more difﬁcult to deﬁne, such as the graph of Brownian motion (Figure 1.2.4), will turn out to have easier dimensions to compute (3/2 by Theorem 6.4.3).
Figure 1.2.3 The Weierstrass function with b = 2, α = 1/2. This graph has Minkowski dimension 3/2 and is conjectured to have the same Hausdorff dimen sion.
1.3 Sets deﬁned by digit restrictions
In this section we will consider some more complicated sets for which the Minkowski dimension is easy to compute, but the Hausdorff dimension is not
05 Feb 2017 at 20:23:35,
02
10
Minkowski and Hausdorff dimensions
Figure 1.2.4 1dimensional Brownian motion. This graph has dimension 3/2 al most surely.
so obvious, and will be left to later sections. These subsets of [0, 1] will be deﬁned by restricting which digits can occur at a certain position of a num ber’s b ary expansion. In a later section we will consider sets deﬁned by the asymptotic distribution of the digits. We start by adapting Hausdorff measures to b adic grids. Let b ≥ 2 be an integer and consider b adic expansions of real numbers, i.e.,
to each sequence { x _{n} }∈{0,1,
b − 1} ^{N} we associate the real number
,
∞
∑
x _{n} b ^{−} ^{n} ∈ [0,1].
x =
n =1
badic expansions give rise to Cantor sets by restricting the digits we are al lowed to use. For example, if we set b = 3 and require x _{n} ∈ {0, 2} for all n we get the middle thirds Cantor set C .
For each integer n let I _{n} (x ) denote the unique halfopen interval of the form
_{n} ) containing x . Such intervals are called b adic intervals of generation
b n ^{,} n ( dyadic if b = 2). It has been observed (by Frostman (1935) and Besicovitch (1952)) that we can restrict the inﬁmum in the deﬁnition of Hausdorff measure to coverings of the set that involve only b adic intervals and only change the value by a bounded factor. The advantage of dealing with these intervals is that they are nested, i.e., two such intervals either are disjoint or one is contained in the other. In particular, any covering by badic intervals always contains a subcover by disjoint intervals (just take the maximal intervals). Furthermore, the b adic intervals can be given the structure of a tree, an observation that we will use extensively in later chapters.
_{[} k −1
k
b
05 Feb 2017 at 20:23:35,
02
1.3 Sets deﬁned by digit restrictions
We deﬁne the grid Hausdorff content by
˜
H
∞
α
(A ) = inf{ _{∑} J _{i}  ^{α} , A ⊂ ^{} J _{i} },
i i
and the grid Hausdorff measures by
˜
H
α
ε
(A ) = inf{ _{∑}  J _{i}  ^{α} , A ⊂ ^{} J _{i} ,J _{i}  < ε},
i
H ^{α} (A ) = lim H
ε→0
α
ε
˜
˜
i
(A
),
11
(1.3.1)
(1.3.2)
(1.3.3)
where the inﬁmums are over all coverings of A ⊂ R by collections { J _{i} } of b  adic intervals. The grid measure depends on b , but we omit it from the notation; usually the value of b is clear from context. Clearly
˜ 

H 
^{α} (A ) ≥ H ^{α} (A ), 
since we are taking the inﬁmum over a smaller set of coverings. However, the two sides are almost of the same size, i.e.,
˜ 

H 
^{α} (A ) ≤ (b + 1)H ^{α} (A ), 
since any interval I can be covered by at most (b + 1) shorter badic intervals;
just take the smallest n so that b ^{−} ^{n} ≤  I  and take all b adic intervals of length
˜
b ^{−} ^{n} that hit I . Thus H ^{α} (A ) = 0 if and only if H ^{α} (A ) = 0. Similarly, we can
deﬁne N (K , ε) to be the minimal number of closed badic intervals of length ε
needed to cover K . We get
N (K ,ε) ≤ N (K ,ε) ≤ (b + 1)N (K ,ε).
Hence, the Hausdorff and the Minkowski dimensions are not changed.
(A ) and grid Hausdorff measures
H
ϕ (A ) and H ^{ϕ} (A ) with respect to any gauge function ϕ by replacing  J _{i}  ^{α}
with ϕ( J _{i} ) in (1.3.1) and (1.3.2). Furthermore, the deﬁnitions can be extended to subsets of R ^{n} by replacing the b adic intervals J _{i} in (1.3.1) and (1.3.2) with
b adic cubes (products of b adic intervals of equal length).
Deﬁnition 1.3.1
˜
˜
One can deﬁne grid Hausdorff content H
∞
˜
˜
ε
˜
ϕ
For S ⊂ N the upper density of S is
N })
#(S ∩{1,
d(S ) = limsup
N →∞
,
N
^{.}
Here and later #(E ) denotes the number of elements in E , i.e., # is counting measure. The lower density is
d(S ) = liminf
N
→∞
#(S ∩{1,
,
N })
N
^{.}
05 Feb 2017 at 20:23:35,
02
12
Minkowski and Hausdorff dimensions
If d(S ) = d(S ), then the limit exists and is called d (S ), the density of S .
Example 1.3.2
Suppose S ⊂ N, and deﬁne
A _{S} = x = _{∑} S x _{k} 2 ^{−} ^{k} : x _{k} ∈ {0,1} .
k
∈
The set A _{S} is covered by exactly
2
^{∑}
n
k
_{=}_{1} 1 _{S} ( k ) _{=} _{2} #( S ∩{1,
,n
})
closed dyadic intervals of generation n , where 1 _{S} is the characteristic function
of S , i.e., 1 _{S} (n ) = 1 for n ∈ S , and 1 _{S} (n ) = 0 for n
∈ S . So
Thus
which implies
˜
log _{2} N (A _{S} ,2 ^{−} ^{n} ) =
n
∑
1 _{S} (k ).
˜ log N (A _{S} ,2 ^{−} ^{n} ) = ^{1} 
k 

log2 
^{n} 
n 
k 

dim 
_{M} 
(A _{S} ) = 

dim 
_{M} 
=1
n
∑ 1 _{S} (k ),
=1
d(S ),
(A _{S} ) = d(S ).
It is easy to construct sets S where the liminf and limsup differ and hence we get compact sets where the Minkowski dimension does not exist (see Exercise 1.17). The Hausdorff dimension of A _{S} is equal to the lower Minkowski dimen sion. We shall prove this in the next section as an application of Billingsley’s Lemma.
We can also construct A _{S} as follows. Start with the interval [0, 1] and sub divide it into two equal length subintervals [0, 1/2] and [1/2, 1]. If 1 ∈ S then
∈ S then keep only the leftmost, [0, 1/2]. Cut each
keep both intervals and if 1
of the remaining intervals in half, keeping both subintervals if 2 ∈ S and only keeping the left interval otherwise. In general, at the n th step we have a set A ^{S} ⊂ A ^{S} _{−}_{1} that is a ﬁnite union of intervals of length 2 ^{−} ^{n} (some may be adjoin ing). We cut each of the intervals in half, keeping both subintervals if n ∈ S and
n
n
throwing away the righthand one if n
n ^{.} In Figure 1.3.1 we have drawn the ﬁrst generations of the construction corre
sponding to S = {1, 3, 4, 6, 8, 10,
∈ S . The limiting set is A _{S} = ^{} ^{∞}
n
=1 ^{A} ^{S}
}.
Example 1.3.3
The shifts of ﬁnite type are deﬁned by restricting which
05 Feb 2017 at 20:23:35,
02
1.3 Sets deﬁned by digit restrictions
13
Figure 1.3.1 First 8 generations of A _{S} for S = {1, 3, 4, 6, 8, 10,
}.
digits can follow other digits. Let A = (A _{i} _{j} ), 0 ≤ i, j < b , be a b × b matrix of 0s and 1s, and deﬁne
X A =
∞ n=1 ∑ x _{n} b ^{−} ^{n} : A _{x} _{n} _{x} _{n}_{+}_{1} = 1 for all n ≥ 1 .
We will also assume that if the j th column of A is all zeros, so is the j th row; this implies every ﬁnite sequence { x _{n} } ^{n} satisfying the condition above can be extended to at least one inﬁnite sequence satisfying the condition. Thus such ﬁnite strings correspond to b adic closed dyadic intervals that intersect X _{A} .
Notice that a b ary rational number r will belong to X _{A} if either of its two possible b ary expansions does (one is eventually all 0s, the other eventually all (b − 1)s). The condition on { x _{n} } described in the example is clearly shift invariant and is the intersection of countably many “closed” conditions, so the set X _{A} is compact and invariant under the map T _{b} , where T _{b} (x )=(bx ) mod 1. For example, if
1
A = ^{1}
1
0 ^{,}
^{1}
then x ∈ X _{A} if and only if the binary representation of x has no two consecutive 1s. The ﬁrst ten generations in the construction of the Cantor set X _{A} are shown in Figure 1.3.3. Any such matrix can be represented by a directed graph where
the vertices represent the numbers {0,
, i to j if A _{i} _{j} = 1. For example, the 2 × 2 matrix A above is represented by
Figure 1.3.2. An element of x = _{∑} _{n} x _{n} b ^{−} ^{n} ∈ X _{A} corresponds in a natural way to an inﬁnite
path { x _{1} , x _{2} ,
} on this graph and a b adic interval hitting X _{A} corresponds to
a ﬁnite path. Conversely, any ﬁnite path of length n corresponds to a b adic
b − 1} and a directed edge connects
˜
interval of length b ^{−} ^{n} hitting X _{A} . Thus N _{n} (X _{A} ) ≡ N (X _{A} , b ^{−} ^{n} ) is the number of
05 Feb 2017 at 20:23:35,
02
14
Minkowski and Hausdorff dimensions
Figure 1.3.2 The directed graph representing A .
Figure 1.3.3 Approximations of X _{A} .
distinct paths of length n in the graph (this uses our assumption about the rows and columns of A ). By the deﬁnition of matrix multiplication, the number of paths of length n from i to j is (A ^{n} ) _{i} _{j} (i.e., the (ith, j th) entry in the matrix A ^{n} ). (See Exercise 1.23.) Thus the total number of length n paths is N _{n} (X _{A} ) = A ^{n} , where the norm of a matrix is deﬁned as B = _{∑} _{i}_{,} _{j} B _{i} _{j} . (However, since any two norms on a ﬁnitedimensional space are comparable, the precise norm won’t matter.) Thus
where
dim _{M} (X _{A} ) = lim
n→∞
log N _{n} (X _{A} ) n log b
= lim
n
→∞
log(N _{n} (X _{A} )) ^{1}^{/} ^{n} log b
_{=} logρ(A)
log b
ρ(A ) = lim
n
_{→}_{∞} A ^{n} ^{1}^{/} ^{n}
^{,}
(1.3.4)
is, by deﬁnition, the spectral radius of the matrix A , and is equal to the ab solute value of the largest eigenvalue. That this limit exists is a standard fact about matrices and is left to the reader (Exercise 1.24). For the matrix
A = ^{1}
1
0 ^{,}
^{1}
the spectral radius is (1 + ^{√} 5)/2 (Exercise 1.25), so the Minkowski dimension
05 Feb 2017 at 20:23:35,
02
1.3 Sets deﬁned by digit restrictions
of X _{A} is
logρ(A )
log(1 + ^{√} 5) − log2
_{=}
log2
log2
^{.}
15
We shall see in Example 2.3.3, that for shifts of ﬁnite type the Hausdorff and Minkowski dimensions agree.
Example 1.3.4 Now we consider some sets in the plane. Suppose A is a b × b
Viel mehr als nur Dokumente.
Entdecken, was Scribd alles zu bieten hat, inklusive Bücher und Hörbücher von großen Verlagen.
Jederzeit kündbar.