Full Geometry and Analysis of Metric Spaces Via Weighted Partitions Jun Kigami Ebook All Chapters
Full Geometry and Analysis of Metric Spaces Via Weighted Partitions Jun Kigami Ebook All Chapters
Full Geometry and Analysis of Metric Spaces Via Weighted Partitions Jun Kigami Ebook All Chapters
com
https://textbookfull.com/product/geometry-
and-analysis-of-metric-spaces-via-weighted-
partitions-jun-kigami/
textbookfull
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
https://textbookfull.com/product/functional-analysis-an-
introduction-to-metric-spaces-hilbert-spaces-and-banach-algebras-
second-edition-joseph-muscat/
https://textbookfull.com/product/a-primer-on-hilbert-space-
theory-linear-spaces-topological-spaces-metric-spaces-normed-
spaces-and-topological-groups-2nd-edition-carlo-alabiso/
https://textbookfull.com/product/geometry-and-martingales-in-
banach-spaces-1st-edition-wojbor-a-woyczynski/
Fundamentals of Signal Processing in Metric Spaces with
Lattice Properties: Algebraic Approach 1st Edition
Andrey Popoff
https://textbookfull.com/product/fundamentals-of-signal-
processing-in-metric-spaces-with-lattice-properties-algebraic-
approach-1st-edition-andrey-popoff/
https://textbookfull.com/product/an-invitation-to-alexandrov-
geometry-cat-0-spaces-stephanie-alexander/
https://textbookfull.com/product/spaces-an-introduction-to-real-
analysis-tom-lindstrom/
https://textbookfull.com/product/riemannian-geometry-and-
geometric-analysis-7th-edition-jurgen-jost-auth/
https://textbookfull.com/product/analysis-on-fock-spaces-and-
mathematical-theory-of-quantum-fields-1st-edition-asao-arai/
Lecture Notes in Mathematics 2265
Jun Kigami
Geometry
and Analysis
of Metric Spaces
via Weighted
Partitions
Lecture Notes in Mathematics
Volume 2265
Editors-in-Chief
Jean-Michel Morel, CMLA, ENS, Cachan, France
Bernard Teissier, IMJ-PRG, Paris, France
Series Editors
Karin Baur, University of Leeds, Leeds, UK
Michel Brion, UGA, Grenoble, France
Camillo De Lellis, IAS, Princeton, NJ, USA
Alessio Figalli, ETH Zurich, Zurich, Switzerland
Annette Huber, Albert Ludwig University, Freiburg, Germany
Davar Khoshnevisan, The University of Utah, Salt Lake City, UT, USA
Ioannis Kontoyiannis, University of Cambridge, Cambridge, UK
Angela Kunoth, University of Cologne, Cologne, Germany
Ariane Mézard, IMJ-PRG, Paris, France
Mark Podolskij, University of Luxembourg, Esch-sur-Alzette, Luxembourg
Sylvia Serfaty, NYU Courant, New York, NY, USA
Gabriele Vezzosi, UniFI, Florence, Italy
Anna Wienhard, Ruprecht Karl University, Heidelberg, Germany
This series reports on new developments in all areas of mathematics and their
applications - quickly, informally and at a high level. Mathematical texts analysing
new developments in modelling and numerical simulation are welcome. The type of
material considered for publication includes:
1. Research monographs
2. Lectures on a new field or presentations of a new angle in a classical field
3. Summer schools and intensive courses on topics of current research.
Texts which are out of print but still in demand may also be considered if they fall
within these categories. The timeliness of a manuscript is sometimes more important
than its form, which may be preliminary or tentative.
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland
AG 2020
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG.
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
This monograph reflects what I have been pursuing for the last 10 years or more.
The main questions are what is the natural way to comprehend the “structure” of
a space? and how can one develop “analysis” on a space from its “structure”?
These are, of course, vague questions since the meanings of the words “structure”
and “analysis” are not quite clear. The former can be replaced by “geometry” but
this does not make things better. One of the more precise formulations is, for a given
space, what is a natural counterpart of the Euclidean metric on Rn ? Needless to say,
rich analysis on Rn has been developed under the Euclidean metric. A prototype is
a class of self-similar sets, including the Sierpinski gasket and carpet, where there is
a natural notion of partitions associated with trees and on which Brownian motion
has been constructed and studied. This monograph is just a tiny step toward this vast
frontier of geometry and analysis of metric spaces but, hopefully, it is a right one at
least.
The origin of the story is the notion of successive divisions of compact metric
spaces, which appear in many different areas of mathematics such as the construc-
tion of self-similar sets, Markov partitions associated with hyperbolic dynamical
systems, and dyadic cubes associated with a doubling metric space. The common
feature in these is to divide a space into a finite number of subsets, then divide
each subset into finitely many pieces, repeating this process again and again. In
this monograph, we generalize such successive divisions and call them partitions.
Given a partition, we consider the notion of a weight function assigning a “size”
to each piece of the partition. Intuitively, we believe that a partition and a weight
function should provide a “geometry” and an “analysis” on the space of interest. We
pursue this idea in Chaps. 2–4. In Chap. 2, the metrizability of a weight function,
i.e. the existence of a metric “adapted to” a given weight function, is shown to be
equivalent to the Gromov hyperbolicity of the graph associated with the weight
function. In Chap. 3, notions such as bi-Lipschitz equivalence, Ahlfors regularity,
the volume doubling property, and quasisymmetry are shown to be equivalent to
certain properties of weight functions. In particular, we find that quasisymmetry
and the volume doubling property are the same notion in the world of weight
functions. In Chap. 4, a characterization of the Ahlfors regular conformal dimension
v
vi Preface
vii
viii Contents
1.1 Introduction
for any m ≥ 0 and i1 . . . im ∈ {0, 1}m . In this example, there are two notable
properties.
The first one is the role of the (infinite) binary tree
Tb = {φ, 0, 1, 00, 01, 10, 11, 000, 001, 010, 011, . . .} = {0, 1}m ,
m≥0
where {0, 1}0 = {φ}. The vertex φ is called the root or the reference point and Tb is
called the tree with the root (or the reference point) φ. Note that the correspondence
i1 . . . im → Ki1 ...im determines a map from the binary tree to the collection of
compact subsets of [0, 1] with the property (1.1.1).
Secondly, note that Ki1 ⊇ Ki1 i2 ⊇ Ki1 i2 i3 ⊇ . . . and
Ki1 ...im is a single point (1.1.2)
m≥1
© The Editor(s) (if applicable) and The Author(s), under exclusive license 1
to Springer Nature Switzerland AG 2020
J. Kigami, Geometry and Analysis of Metric Spaces via Weighted Partitions,
Lecture Notes in Mathematics 2265, https://doi.org/10.1007/978-3-030-54154-5_1
2 1 Introduction and a Showcase
0 Kφ 1 φ
1
K0 2 K1
0 1
1 3
K00 4 K01 K10 4 K11
00 01 10 11
1 3 5 7
8 8 8 8
Fig. 1.1 A partition of the unit interval [0, 1] and the associated tree
See Fig. 1.2 for an illustration of the idea. Note that the properties (1.1.3)
and (1.1.4) correspond to (1.1.1) and (1.1.2) respectively. In this monograph such
{Xp }p∈T is called a partition of X parametrized by the tree T . (We will give the
precise definition in Sect. 2.2.) In addition to the “vertical” edges, which are the
edges of the tree, we provide “horizontal” edges to T to describe the combinatorial
structure reflecting the topology of X as is seen in Fig. 1.2. More precisely, a pair
(p, q) ∈ T × T is a horizontal edge if p and q have the same distance from the root
φ and Xp ∩ Xq = ∅. We call T with horizontal and vertical edges the resolution of
X associated with the partition.
Another key notion is a weight function on the tree T . Note that a metric and
a measure give weights of the subsets of X. More precisely, let d be a metric on
X inducing the original topology of X and let μ be a Radon measure on X where
μ(Xp ) > 0 for any p ∈ T . Define ρd : T → (0, 1] and ρμ : T → (0, 1] by
diam(Xp , d) μ(Xp )
ρd (p) = and ρμ (p) = ,
diam(X, d) μ(X)
.................
...........
..
...............................................................................................................................
.... ............
.........
φ
......... ... .......
.....
. ................................... ...
... .......
.....
. ........ ... .. ...........
.
.. ... ..
. ..
..... ......
..
...
.. ..
. .....
....
.. ... ....
.... ....
.....
....
....
X11 .. ......
.... ......
.....
X21 X22 ..
.
.
.
..
..
....
.....
....
....
....
....
...
... . .. ... .
. ... ...
.... ...... . .
.
.
. .
. ..
..... ... . ... ...
...
...
.... ... .
....... ...
. ...
....
.
.
.
.
.
. ....... ... . ..
.. .
.. ... ...
..
2
........................................... ..
X32
... .. . ...
.. . ........
. .
. . . . ..
..
..
...
...
X12 ....
..
.
..
.........................
...
......
......... ...
...
...
. .
..
.
..
..
. 1 3
... ........................................... .... ...
...
... ........ ......... ...
.... ....... .. ... ...
.... ....
X31
.. ...
....
.... . .....
. ..
. .. ......
X13
. ... ... ...
..... .... ... ... ..... 21 22
......
...... .... ... ..........
....... ..... ...
.. ... ......... 11
......... .... .....
......... . ........
........... .. ...........
.................
.................................. .. .................
......................................................................
12 32
X = Xφ
13 31
A partition of X The associated graphs T
Solid lines are vertical edges.
Dotted lines are horizontal edges.
Fig. 1.2 A partition and the associated graphs (up to the 2nd stage)
4 1 Introduction and a Showcase
where diam(A, d) is the diameter of A with respect to the metric d. Then ρd (resp.
ρμ ) is though of as a natural weight of Xp associated with d (resp. μ). In both cases
where = d or = μ, the function ρ : T → (0, ∞) satisfies
if π(pi+1 ) = pi for any i ≥ 1. (To have the second property (1.1.6) in case of
= μ, we must assume that the measure μ is non-atomic, i.e. μ({x}) = 0 for any
x ∈ X.)
As we have seen above, given a metric or a measure, we have obtained a weight
function ρ satisfying (1.1.5) and (1.1.6). In this monograph, we are interested in the
opposite direction. Namely, given a partition of a compact metrizable topological
space parametrized by a tree T , we define the notion of weight functions as the
collection of functions from T to (0, 1] satisfying the properties (1.1.5) and (1.1.6).
Then our main object of interest is the space of weight functions including those
derived from metrics and measures. Naively we believe that a partition and a weight
function essentially determine a “geometry” and/or an“analysis” of the original set
no matter where the weight function comes. It may come from a metric, a measure
or else. Keeping this intuition in mind, we are going to develop a basic theory of
weight functions in three closely related directions in this monograph.
The first direction is to study when a weight function is naturally associated
with a metric. In brief, our conclusion will be that a power of a weight function is
naturally associated with a metric if and only if the rearrangement of the resolution
T associated the weight function is Gromov hyperbolic. To be more precise, let
{Xw }w∈T be a partition of a compact metrizable topological space X with no
isolated points and let ρ : T → (0, 1] be a weight function. In Sect. 2.3, we
ρ
will define δM (·, ·), which is called the visual pre-metric associated with ρ, in the
ρ
following way: let s be the collection of w’s in T where the size ρ(w) is almost s.
ρ ρ
Define a horizontal edge of s as (w, v) ∈ s with Xw ∩ Xv = ∅. For r ∈ (0, 1),
the rearranged resolution Tρ,r associated with the weight function ρ is defined as
ρ
the vertices ∪m≥0 r m with the vertical edges from the tree structure of T and the
ρ ρ
horizontal edges of r m . Then the visual pre-metric δM (x, y), where M ≥ 1 is a
natural number, for x, y ∈ X is given by the infimum of s where x and y can be
ρ
connected by an M-consecutive horizontal edges in s . We think that a metric d
ρ
is naturally associated with the weight function ρ if and only if d and δM are bi-
Lipschitz equivalent on X × X. More precisely, we are going to use a phrase“d is
adapted to ρ” instead of “d is naturally associated with ρ”. The notion of visual
pre-metric in this monograph is a counterpart of that of visual pre-metric on the
boundary of a Gromov hyperbolic metric space, whose detailed account can be
seen in [12, 27] and [18] for example. Now the main conclusion of the first part
1.1 Introduction 5
is Theorem 2.5.12 saying that the hyperbolicity of the rearranged resolution Tρ,r is
equivalent to the existence of a metric adapted to some power of the weight function.
Moreover, if this is the case, the metric adapted to some power of the weight function
is shown to be a visual metric in Gromov’s sense.
The second direction is to establish relationships of various relations between
weight functions, metrics and measures. For example, Ahlfors regularity and the
volume property are relations between measures and metrics. For α > 0, a measure
μ is α-Ahlfors regular with respect to a metric d if and only if there exist c1 , c2 > 0
such that
where Bd (x, r) = {y|y ∈ X, d(x, y) < r}, for any r ∈ (0, diam(X, d)] and x ∈ X.
See Definition 3.3.3 for the precise definition of the volume doubling property. On
the other hand, bi-Lipschitz and quasisymmetry are equivalence relations between
two metrics. (The precise definitions of bi-Lipschitz equivalence and quasisymmetry
are given in Definitions 3.1.9 and 3.6.1 respectively.) Regarding those relations, we
are going to claim the following relationships
and
in the framework of weight functions. To illustrate the first claim more explicitly, let
us introduce the notion of bi-Lipschitz equivalence of weight functions. Two weight
functions ρ1 and ρ2 are said to be bi-Lipschitz equivalent if and only if there exist
c1 , c2 > 0 such that
for any p ∈ T . Now the first statement (1.1.7) can be resolved into three parts as
follows: let ρ1 and ρ2 be two weight functions.
Claim 1: Suppose that ρ1 = ρd1 and ρ2 = ρd2 for metrics d1 and d2 on X. Then
ρ1 and ρ2 are bi-Lipschitz equivalent if and only if d1 and d2 are bi-
Lipschitz equivalent as metrics.
Claim 2: Suppose that ρ1 = ρd and ρ2 = ρμ for a metric d on X and a Radon
measure μ on X. Then (ρ1 )α and ρ2 are bi-Lipschitz equivalent if and
only if μ is α-Ahlfors regular with respect to d.
Claim 3: Suppose that ρ1 = ρd for a metric d on X, then ρ1 and ρ2 are bi-
Lipschitz equivalent if and only if the metric d is adapted to the weight
function ρ2 .
6 1 Introduction and a Showcase
One can find the precise statement in Theorem 1.2.11 in the case of partitions
of S 2 . The second statement (1.1.8) is rationalized in the same manner. See
Theorem 1.2.12 for the exact statement in the case of S 2 for example.
The third direction is a characterization of Ahlfors regular conformal dimension.
The Ahlfors regular conformal dimension, AR conformal dimension for short, of a
metric space (X, d) is defined as
where “ρ ∼ d” means that the two metrics ρ and d are quasisymmetric to each
QS
other. Note that α is the Hausdorff dimension of (X, d) if there exists a measure
μ which is α-Ahlfors regular with respect to d. In [13], Carrasco Piaggio has
given a characterization of Ahlfors regular conformal dimension in terms of the
critical exponent of p-combinatorial modulus of discrete path families. Given the
results from the previous part, we have obtained the ways to express the notions
of quasisymmetry and Ahlfors regularity in terms of weight functions. So we are
going to translate Carrasco Piaggio’s work into our framework. However, we are
going to use the critical exponent of p-energy instead of p-combinatorial modulus
in our work.1 Despite the difference between p-energy and p-modulus, we will
make essential use of Carrasco Piaggio’s ideas, which have been quietly embedded
in his paper [13]. Furthermore, we are going to define the notion of p-spectral
dimension and present a relation between Ahlfors regular conformal dimension
and p-spectral dimension. In particular, for p = 2, the 2-spectral dimension has
been know to appear in the asymptotic behavior of the Brownian motion and the
eigenvalue counting function of the Laplacian on certain fractals like the Sierpinski
gasket and the Sierpinski carpet. See [4, 5] and [23] for example. For the Sierpinski
carpet, we will show that the 2-spectral dimension gives an upper bound of Ahlfors
regular conformal dimension.
One of the main reasons for recent interest in Ahlfors regular conformal
dimension is its close connection with the Cannon’s conjecture, although we are
not going to pursue such a direction in this monograph.
Conjecture (Cannon) Let G be a Gromov hyperbolic group whose boundary ∂∞ G
is homeomorphic to the two dimensional sphere S 2 . Let d be a visual metric on
∂∞ G. Then (∂∞ G, d) is quasisymmetric to S 2 with the Euclidean metric.
For this conjecture, Bonk and Kleiner have given the following partial answer
in [7].
Such situations may happen if our target space (X, d) is not connected.
In this section, we summarize our main results in this monograph in the case of a
2-dimensional sphere S 2 (or in other words, the Riemann sphere), which is denoted
by X in what follows. We use ds to denote the standard spherical geodesic metric
on X. Set
N0
X= Xi .
i=1
1.2 Summary of the Main Results; the Case of 2-dim. Sphere 9
for any k = 0, 1, . . ., where T0 is a one point set {φ}. Let T = ∪k≥0 Tk . Then T is
naturally though of as a (non-directed) tree whose edges are given by the totality of
(i1 . . .ik , i1 . . .ik ik+1 ). We regard the correspondence w ∈ T to Xw ∈ U as a map
from T to U, which is denoted by X . Namely, X (w) = Xw for any w ∈ T . Note
that X (φ) = X. Define
is a single point, which is denoted by σ (i1 i2 . . .). Note that σ is a map from to
X. This assumption corresponds to (1.1.4) and hence the map X is a partition of X
parametrized by the tree T , i.e. it satisfies the conditions (i), (ii), (iii), (iv) and (v) in
the introduction. Since X = ∪w∈Tk Xw for any k ≥ 0, this map σ is surjective.
In [8, Chapter 5], the authors have constructed “cell decomposition” associated
with an expanding Thurston map. This “cell decomposition” is, in fact, an example
of a partition formulated above.
Throughout this section, for simplicity, we assume the following conditions (SF)
and (TH), where (SF) is called strong finiteness in Definition 2.2.4 and (TH) ensures
the thickness of every exponential weight function. See Definition 3.1.19 for the
“thickness” of a weight function.
(SF)
(TH) There exists m ≥ 1 such that for any w = i1 . . .in ∈ T , there exists v =
i1 . . . in in+1 . . .in+m ∈ T such that Xv ⊆ int(Xw ).
The main purpose of this monograph is to describe metrics and measures of X
from a given weight assigned to each piece Xw of the partition X .
Definition 1.2.1 A map g : T → (0, 1] is called a weight function if and only if it
satisfies the following conditions (G1), (G2) and (G3).
(G1) g(φ) = 1
(G2) g(i1 . . .ik ) ≥ g(i1 . . .ik ik+1 ) for any i1 . . .ik ∈ T and i1 . . .ik ik+1 ∈ T .
(G3)
Define
Moreover, we define following conditions (SpE) and (SbE), which represent “super-
exponential” and “sub-exponential” respectively:
(SpE) There exists λ ∈ (0, 1) such that
and
For any d ∈ D(X), define gd : T → (0, 1] by gd (w) = diam(Xw , d) and for any
μ ∈ M(X), define gμ : T → (0, 1] by gμ (w) = μ(Xw ) for any w ∈ T .
From Proposition 2.3.5, we have the following fact.
Proposition 1.2.3 If d ∈ D(X) and μ ∈ M(X), then gd and gμ are weight
functions.
So a metric d ∈ D(X) has associated weight function gd . How about the converse
direction, i.e. for a given weight function g, is there a metric d such that g = gd ?
To make this question more rigorous and flexible, we define the notion of “visual
g
pre-metric” δM (·, ·) associated with a weight function g.
Definition 1.2.4 Let g ∈ G(T ). Define
g
s = {i1 . . .ik |i1 . . .ik ∈ T , g(i1 . . .ik−1 ) > s ≥ g(i1 . . .ik )}
g g g
Although δM (x, y) ≥ 0, δM (x, y) = 0 if and only if x = y and δM (x, y) =
g g
δM (y, x), the quantity δM may not satisfy the triangle inequality in general. The
g
visual pre-metric δM (x, y) is a counterpart of the visual metric defined in [8]. See
Sect. 2.3 for details.
g
If the pre-metric δM (·, ·) is bi-Lipschitz equivalent to a metric d, we consider d
as a metric which is naturally associated with the weight function g.
12 1 Introduction and a Showcase
for any x, y ∈ X.
(2) A metric d is said to be M-adapted if and only if it is M-adapted to gd and it is
said to be adapted if it is M-adapted for some M ≥ 1.
(3) Define
g
h
Eg,r = {(w, v)|w, v ∈ r m for some m ≥ 0, w = v, Xw ∩ Xv = ∅}
and
g g
v
Eg,r = {(w, v)|w ∈ r m , v ∈ r m+1 for some m ≥ 0, Xw ⊇ Xv }
respectively.
The following theorem is a special case of Theorem 2.5.12.
Theorem 1.2.7 Let g ∈ G(X). There exist M ≥ 1, α > 0 and a metric d ∈ D(X)
such that d is M-adapted to g α if and only if the graph (Tg,r , Eg,r
h ∪E v ) is Gromov
g,r
hyperbolic for some r ∈ (0, 1). Moreover, if this is the case, then X is homeomorphic
to the hyperbolic boundary of (Tg,r , Eg,r
h ∪ E v ) and the adapted metric d is a
g,r
visual metric in the Gromov sense.
Next, we define two equivalent relations ∼ and ∼ on the collection of
BL GE
exponential weight functions. Later, we are going to identify these with known
relations according to the types of weight functions.
1.2 Summary of the Main Results; the Case of 2-dim. Sphere 13
(4) For d ∈ DA,e (X), μ ∈ M(X) and α > 0, (gd )α ∼ gμ and gd is uniformly finite
BL
if and only if μ is α-Ahlfors regular with respect to d, i.e. there exist c1 , c2 > 0
such that
and
g g
M (w) = {v|v ∈ r m , there exists (v(0), v(1), . . . , v(M)) such that
v(0) = w and (v(i), v(i + 1)) ∈ Eg,r
h
for any i = 0, . . . , M − 1}.
g g
The set S n (A) corresponds to the refinement of A in r m+n and the set M (w) is
g
the M-neighborhood of w in r m .
g,r , M ≥ 1
Definition 1.2.14 Let g ∈ Ge (T ) and let r ∈ (0, 1). For p > 0, w ∈ T
and n ≥ 0, define
g
EM,p,w,n = inf |f (u) − f (v)|p
h ,u,v∈ g
(u,v)∈Eg,r
r m+n
g
f : r m+n → R, f |S n (w) = 1, f |g g
\S n (M (w)) =0 ,
r m+n
g
where the integer m is chosen so that w ∈ r m , and
g g
EM,p = lim inf sup EM,p,w,m .
m→∞
w∈Tg,r
In this section, we review basic notions and notations on a tree with a reference
point.
Definition 2.1.1 Let T be a countably infinite set and let A : T × T → {0, 1}
which satisfies A(w, v) = A(v, w) and A(w, w) = 0 for any w, v ∈ T . We call the
pair (T , A) a (non-directed) graph with the vertices T and the adjacent matrix A.
An element (u, v) ∈ T × T is called an edge of (T , A) if and only if A(u, v) = 1.
We will identify the adjacent matrix A with the collection of edges {(u, v)|u, v ∈
T , A(u, v) = 1}.
(1) The set {v|A(w, v) = 1} is called the neighborhood of w in (T , A). (T , A) is
said to be locally finite if the neighborhood of w is a finite set for any w ∈ T .
(2) For w0 , . . . , wn ∈ T , (w0 , w1 , . . . , wn ) is called a path between w0 and wn if
A(wi , wi+1 ) = 1 for any i = 0, 1, . . . n − 1. A path (w0 , w1 , . . . , wn ) is called
simple if and only if wi = wj for any i, j with 0 ≤ i < j ≤ n and |i − j | < n.
(3) (T , A) is called a (non-directed) tree if and only if there exists a unique simple
path between w and v for any w, v ∈ T with w = v. For a tree (T , A), the
unique simple path between two vertices w and v is called the geodesic between
w and v and denoted by wv. We write u ∈ wv if wv = (w0 , w1 , . . . , wn ) and
u = wi for some i.
In this monograph, we always fix a point in a tree as the root of the tree and call
the point the reference point.
© The Editor(s) (if applicable) and The Author(s), under exclusive license 17
to Springer Nature Switzerland AG 2020
J. Kigami, Geometry and Analysis of Metric Spaces via Weighted Partitions,
Lecture Notes in Mathematics 2265, https://doi.org/10.1007/978-3-030-54154-5_2
18 2 Partitions, Weight Functions and Their Hyperbolicity
Remark Strictly, the notations like π and | · | should be written as π (T ,A,φ) and
| · |(T ,A,φ) respectively. In fact, if we will need to specify the tree in question, we are
going to use such explicit notations.
One of the typical examples of a tree is the infinite binary tree. In the next
example, we present a class of trees where #(S(w)) is independent of w ∈ T .
(N)
Example 2.1.3 Let N ≥ 2 be an integer. Let Tm = {1, . . . , N}m for m ≥ 0.
(N)
(We let T0 = {φ}, where φ represents an empty sequence.) We customarily write
(N) (N)
(i1 , . . . , im ) ∈ Tm as i1 . . . im . Set T (N) = ∪m≥0 Tm . Define π : T (N) → T (N)
by π(i1 . . . im im+1 ) = i1 . . . im for m ≥ 0 and π(φ) = φ. Furthermore, define
Then (T (N) , A(N) , φ) is a locally finite tree with a reference point φ. In particular,
(T (2) , A(2) , φ) is called the infinite binary tree.
It is easy to see that for any infinite geodesic ray (w0 , w1 , . . .), there exists a
geodesic ray originated from φ that is equivalent to (w0 , w1 , . . .). In fact, adding
the geodesic φw0 to (w0 , w1 , . . .) and removing a loop, one can obtain the infinite
geodesic ray having required property. This fact shows the following proposition.
2.1 Tree with a Reference Point 19
Through this map, we always identify the collection of ends and the collection
of infinite geodesic rays originated from φ, φ .
Hereafter in this monograph, (T , A, φ) is assumed to be a locally finite tree with
a fixed reference point φ ∈ T . If no confusion can occur, we omit φ in the notations.
φ
For example, we use , and v in place of φ and v respectively.
Example 2.1.5 Let N ≥ 2 be an integer. In the case of (T (N) , A(N) , φ) defined
in Example 2.1.3, the collection of the ends is (N) = {1, . . . , N}N =
{i1 i2 i3 . . . , |ij ∈ {1, . . . , N} for any m ∈ N}. With the natural product topology,
(N) is a Cantor set, i.e. perfect and totally disconnected.
Tw = {v|v ∈ T , w ∈ φv}.
w ∧ v = wmax{i|i=0,...,|w|,[w]i =[v]i }
ω ∧ τ = [ω]max{m|[ω]m =[τ ]m } .
2−|ω∧τ | if ω = τ,
ρ∗ (ω, τ ) =
0 if ω = τ.
w if m ≤ n,
w∧v =
v if m ≤ n.
2.2 Partition
In this section, we formulate exactly the notion of a partition introduced in Sect. 1.1.
A partition is a map from a tree to the collection of nonempty compact subsets of
a compact metrizable topological space with no isolated point and it is required
to preserve the natural hierarchical structure of the tree. Consequently, a partition
induces a surjective map from the Cantor set, i.e. the collection of ends of the tree,
to the compact metrizable space.
Throughout this section, T = (T , A, φ) is a locally finite tree with a reference
point φ.
Definition 2.2.1 (Partition) Let (X, O) be a compact metrizable topological space
having no isolated point, where O is the collection of open sets, and let C(X, O)
be the collection of nonempty compact subsets of X. If no confusion can occur, we
write C(X) in place of C(X, O).
(1) A map K : T → C(X, O), where we customarily denote K(w) by Kw for
simplicity, is called a partition of X parametrized by (T , A, φ) if and only if
it satisfies the following conditions (P1) and (P2), which correspond to (1.1.3)
and (1.1.4) respectively.
(P1) Kφ = X and for any w ∈ T , #(Kw ) ≥ 2, Kw has no isolated point and
Kw = Kv .
v∈S(w)
Ow = Kw \ Kv ,
v∈(T )|w| \{w}
Bw = Kw ∩ Kv .
v∈(T )|w| \{w}
So Ow is open. The rest of the statement is immediate from the property (P2).
(2) By Lemma 2.1.7, if u = w ∧ v, then |u| < |w| and |u| < |v|. Let w =
[w]|u|+1 and let v = [v]|u|+1 . Then w , v ∈ S(u) and w = v . Since Ow ⊆
Kw \Kv , it follows that Ow ∩ Kv = ∅. Using (1), we see Ow ∩ Kv = ∅.
(3) This follows immediately by the definition of Bw .
The condition (P2) provides a natural map from the ends of the tree to the
space X.
Another random document with
no related content on Scribd:
— Je ne peux m’empêcher, en effet, reprit Gaspard, de trouver
bien sots les princes de la terre, les puissants du monde, qui
s’imaginent que le peuple ne pense pas, parce qu’il est muet. Le
peuple a l’air de ne point penser, parce qu’il ne sait pas s’exprimer
avec des mots savants, des mots de livres, mais sa pensée s’agite
dans son cœur ; — et quand il la reconnaît dans la parole des
grands hommes, il comprend qu’elle est juste et il sent se délier sa
langue… Or çà, je vois que l’ami Pablo donne des signes
d’impatience. N’oublions pas qu’il a un miracle à nous conter et qu’il
brûle de faire briller son éloquence à nos yeux.
CHAPITRE X