Atlantis Studies in Mathematics

Series Editor: Jan van Mill

Shou Lin
Ziqiu Yun

Generalized
Metric Spaces and
Mappings
Atlantis Studies in Mathematics

Volume 6

Series editor
Jan van Mill, VU University, Amsterdam, The Netherlands
Aims and Scope

With this book series, we aim to publish monographs of high quality in all areas of
mathematics. Both research monographs and books of an expository nature are
welcome. This series is the continuation of the “Mathematics Studies”, previously
published by Elsevier. All books published after November 2010 are promoted,
distributed and sold by Springer, both as e-books and in print. The books are also
part of SpringerLink and included in the relevant Springer subject collections. All
book proposals submitted to this series will be reviewed by the Series Editor. After
the manuscript has been completed, it will be entirely reviewed by one of our
editors or reviewers. Only after this review will the book be published.

More information about this series at http://www.springer.com/series/10070

Shou Lin Ziqiu Yun
•

Generalized Metric Spaces

and Mappings
Shou Lin Ziqiu Yun
School of Mathematics and Statistics Department of Mathematics
Minnan Normal University Soochow University
Zhangzhou Suzhou
China China

ISSN 1875-7634 ISSN 2215-1885 (electronic)

Atlantis Studies in Mathematics
ISBN 978-94-6239-215-1 ISBN 978-94-6239-216-8 (eBook)
DOI 10.2991/978-94-6239-216-8
Library of Congress Control Number: 2016951688

Translation and revision from the Chinese language edition: 广义度量空间与映射 by Shou Lin
© Science Press, Beijing 2007. All Rights Reserved.
© Atlantis Press and the author(s) 2016
This book, or any parts thereof, may not be reproduced for commercial purposes in any form or by any
means, electronic or mechanical, including photocopying, recording or any information storage and
retrieval system known or to be invented, without prior permission from the Publisher.

Printed on acid-free paper

Foreword

What are generalized metric spaces? Most often, this expression denotes a system
of classes of topological spaces, each of which is defined with the help of some
typical topological property of metrizable spaces. For example, the classes of
paracompact spaces, Moore spaces, spaces with a point-countable base, sub-
metrizable spaces, perfectly normal spaces, first-countable spaces belong to this
system. We also include in it the class of symmetrizable spaces and the class of D-
metrizable spaces. The topologies of these spaces are generated by generalized
metrics. In this way, we obtain a rich system of classes of topological spaces which
are all emerging, growing and spreading in many directions, from the same pow-
erful germ—the concept of the topology generated by a metric.
The system of generalized metrizable spaces (below, we often use an abbrevi-
ation “GMS-system” to denote it) also includes various relations between its
members. The simplest among them is, of course, the inclusion relation. But much
deeper, often unpredictable links between classes of topological spaces are estab-
lished by means of mappings. Mappings of various kinds serve as instruments by
means of which the abstract geometric objects, such as topological spaces, topo-
logical groups, metric spaces, function spaces and so on, can be compared between
one another and further be classified. This can be done at the level of individual
spaces, but similarly, natural classes of mappings can be used to establish funda-
mental connections between classes of topological spaces.
The ideas and problems of the theory of generalized metrizable spaces, in par-
ticular, the general metrization problem, greatly influenced all domains of
set-theoretic topology. The goal of this book is to describe the present structure
of the theory of generalized metric spaces and the modern tendencies in this theory.
An important fact: this is the first book on generalized metrizable spaces. Some
brilliant survey papers on this topic were written by prominent topologists R.
Hodel, G. Gruenhage, D. Burke, D. Lutzer, H.H. Wicke and J. Worrell, W.

v
vi Foreword

Fleissner, G.M. Reed who themselves contributed greatly to the field. But a book on
this rapidly developing subject was lacking. It should have appeared at least 20
years earlier. This has happened in China, where a version of this book has
appeared in Chinese in 1995. So, it is not astonishing that a strong input in the
set-theoretic topology was made by in the last 20 or 30 years by Guoshi Gao,
Zhimin Gao, Shou Lin, Chuan Liu, Fucai Lin, Ziqiu Yun, Liang-Xue Peng, Lei
Mou, Wei-Xue Shi, Li-Hong Xie, Jing Zhang and many others.
It is an excellent news that, finally, an expanded and polished international
edition of this book will appear in English. The contents are well selected. The book
has three chapters and two appendices. The main part, consisting of the three
chapters, is written in a very systematic way. Theorems and examples form a part
of the big picture; they are supplied with detailed proofs. Stratifiable spaces, spaces
with a r-discrete network, bases of countable order and their theory developed by
H.H. Wicke and J. Worrell, uniform bases introduced by P.S. Alexandroff,
monotonically normal spaces of R.W. Heath, p-spaces and their theory, quotients of
separable metrizable spaces, A.H. Stone’s theorem on paracompactness of metric
spaces, are presented with a great force. All major theorems obtained in the past 50
years in GMS-theory, in particular, V.I. Ponomarev’s theorem characterizing
first-countable spaces as open continuous images of metric spaces, V.V. Filippov’s
theorem on metrizability of paracompact p-spaces with a point-countable base,
generalizing A.S. Miščenko’s theorem on compacta with such a base, Michael’s
theorem on preservation of paracompactness by closed continuous mappings, a
characterization of perfect preimages of metrizable spaces as paracompact p-spaces
(by A.V. Arhangel’skiǐ), another, independent, characterization of the same spaces
by K. Morita (as paracompact M-spaces), all these basic results of the GMS-theory
the reader will find in the book. Appendices A and B are written in the form of
survey, they provide the reader with many further advanced and more special facts
and open questions, with metamathematical discussions and attractive philosophical
insights, omitting detailed proofs. Much of additional information on the history
of the subject is also provided here. An attractive feature of this book is its inter-
national character. The list of references is very rich, it includes about 500 entries,
and will be of great value for readers. Besides the papers of mathematicians we
have already mentioned above, the list includes the classical ground-breaking
papers on GMS-theory, written by R. Bing, C. Dowker, J. Nagata, E. Michael, M.E.
Rudin, A.H. Stone, Yu.M. Smirnov, K. Nagami, Z. Frolík, R. Engelking, Z.
Balogh, J. van Mill, M. Hušek, H. Tamano, P. Nyikos, J. Chaber, S. Nedev, C.
Borges, T. Hoshina, R. Stephenson, Y. Tanaka, H. Junnila, N.V. Veličko, M. Itō,
D. Shakhmatov, M. Sakai, Y. Yajima and others.
The book, written by Professors Shou Lin and Ziqiu Yun who are active experts
in the GMS-theory with a long list of original contributions, will be a most valuable
source of information for graduate and undergraduate topology students wishing to
Foreword vii

start their own research in this field or to apply it to other mathematical disciplines.
It will be also very helpful to experts in topology already working in GMS-theory,
as well as to those who want to enrich their research in other topics by linking it to
GMS-theory. The book will also help to develop new original special courses in
general topology.

April 2016 Alexander V. Arhangel’skiǐ

Distinguished Professor Emeritus
Ohio University, and Professor of Moscow University
Preface

The uniqueness of this book lies in describing generalized metric spaces by means
of mappings. Back in 1961, Alexandroff [3] proposed the idea of using the mapping
method to study spaces in the first Prague Topological Symposium. The survey
paper “Mappings and spaces” written by Arhangel’skiǐ [31] in 1966 inherited and
developed the idea. We were greatly interested in this paper. Professors L. Wu and
B. Chen then translated it into Chinese (originally in Russian) and published it in
“Mathematics” (1981–1982), and wanted to arouse the interest of Chinese scholars.
For a comprehensive introduction to the theory of generalized metric spaces, we
recommend the books written by Burke, Lutzer [88] and Gruenhage [162], and two
chapters written by Nagata [378] and Tamano [449] in the book “Topics in General
Topology” [364].
There are roughly three perspectives of investigating spaces by using mappings:
(1) Which classes of generalized metric spaces can be represented as images or
preimages of metric spaces under certain mappings? For example, the
M-spaces introduced by Morita [360] for investigating the normality of pro-
duct spaces can be expressed as preimages of metric spaces under
quasi-perfect mappings. This has opened up a new way of investigating
M-spaces and established connection between this class of spaces and metric
spaces.
(2) What are the intrinsic characterizations of images of metric spaces under
certain mappings? For example, the closed images of metric spaces (usually
called Lašnev spaces) are characterized, by Foged [130], as regular
Fréchet-Urysohn spaces with a r-hereditarily closure-preserving k-network.
Thus, it can be compared with the Burke-Engelking-Lutzer metrization
theorem [87] and one can connected these spaces with some generalized
metric spaces defined by k-networks, for example @-spaces, etc.
(3) Certain generalized metric spaces are preserved under what kinds of map-
pings? Take the class of metrizable spaces as an example, by the
Hanai-Morita-Stone theorem [363, 441], we know that metrizability is
invariant under perfect mappings. Michael [336] further proved that

ix
x Preface

metrizability is invariant under countably bi-quotient closed mappings, which

shows the charm of bi-quotient mappings.
These simple examples in the above three aspects reveal great colorfulness and
attractiveness of infiltrations of mapping method for investigating generalized
metric spaces.
Following the ideas and methods of Alexandroff-Arhangel’skiǐ, this book pre-
sents the relevant results in the theory of generalized metric spaces in the past three
decades from the perspective of mappings particularly the achievements of Chinese
scholars in recent years. As an interesting scientific research reading material, it
further elaborates these ideas and methods.
This book is also suitable as an elective course material or a teaching reference
book for the graduates and senior undergraduates of mathematics major, or as a
reading material for graduate students of general topology. It also can be used as
a reference book for mathematicians and scientific researchers in other fields.

Suzhou, China Guoshi Gao

September 1992
Preface to the English Edition

The concept of paracompactness introduced by Dieudonné [112] in 1944 is a

significant sign of general topology to enter the peak period. The extraordinary
Bing-Nagata-Smirnov metrization theorem [62, 370, 428], established in the years
1950 and 1951, created a fundamental change for the full exploration of the nature
of metrizability. Meanwhile, the theorem disclosed a bright future of the investi-
gation on properties of generalized metric spaces. The deep study on paracom-
pactness and metrizability performs the prelude to the research on the theory of
generalized metric spaces.
In 1961, at the international topological symposium named “General Topology
and its Relationship with Modern Analysis and Algebra” in Prague, Alexandroff [3]
put forward the idea of investigating spaces by mappings, namely, to connect
various classes of spaces by using mappings as a linkage. In this way, the funda-
mental research framework and whole structure of the theory of topological spaces
can be reflected by the relationships between mappings and spaces. Mappings thus
became a powerful tool for revealing the internal properties of various classes of
spaces. In 1966, Arhangel’skiǐ [31] published the historical literature titled
“Mappings and Spaces” which presented a series of constructive concrete steps of
how to operate the Alexandroff idea. As a milepost in the road of the vigorous
development of general topology, it ushered in an innovation era for investigating
spaces by mappings. Since then, the Alexandroff idea became an indispensable
method for the research. It promoted a rapid development of general topology,
particularly the theory of generalized metric spaces. In a word, the Alexandroff idea
of mutual classifications of spaces and mappings has constituted an important part
of contemporary general topology [4].
According to Alexandroff and Arhangel’skiǐ, the core of investigating spaces by
means of mappings is to establish the extensive connections, with the aid of
mappings, between the class of metric spaces and classes of spaces having specific
topological properties, to study the intrinsic characterizations of images of metric
spaces under various mappings, and to discuss which kinds of mappings can pre-
serve certain classes of spaces. This framework determines the goals of this book: to

xi
xii Preface to the English Edition

comprehensively describe characterizations of images of metric spaces under var-

ious mappings and to establish mapping theorems for several important generalized
metric spaces.
The most prominent feature of this book is to discuss the theory of generalized
metric spaces in a systematic way from the perspective of mappings. Through the
mapping theorems of generalized metric spaces, this book tries to point out that the
principle of classifying spaces by mappings is undeniably a powerful research tool
that has decisive significance in general topology, by which one can peep the full
linkage between spaces and mappings. Most of the contents of this book are from
hundreds of research papers on spaces and mappings published in the recent
50 years. While paying attention to the innovation and unique features of the
contents, this book highlights the outstanding achievements of Chinese scholars in
recent years. The authors hope this book can guide readers to grasp the principle
of classifying spaces by mappings and catch up the new developments of the theory
of generalized metric spaces. We also hope this book can provides solid support for
readers in their further research work and can magnify the impacts from China’s
general topological circles to the world.
This book is composed of three chapters, two appendices and a list of more than
400 references. To provide necessary preparation for the descriptions of relation-
ships between spaces and mappings in the following two chapters, Chap. 1 briefly
introduces some basic concepts of generalized metric spaces, the properties of
fundamental operations on these spaces and their simple characterizations. By
means of several families of sets with specific properties and the concepts of bases
or their generalizations, Chap. 2 presents the intrinsic characterizations of images or
preimages of metric spaces under quotient mappings, pseudo-open mappings,
countably bi-quotient mappings, open mappings, closed mappings and
compact-covering mappings or under these mappings with some additional con-
ditions, for example the fibers are separable or compact. Chapter 3 introduces
several classes of generalized metric spaces defined by specific bases, weak bases,
k-networks, networks and (mod k)-networks etc., such as M-spaces, p-spaces,
g-metrizable spaces, @-spaces, k-semi-stratifiable spaces, r-spaces, R-spaces and so
on. Characterizations and mapping theorems of these spaces are also obtained in
this chapter. There are two appendices at the end of this book. The first one aims to
help readers to better understand some results on the covering properties used in the
text. The second one aims to help readers to obtain a more comprehensive vision
of the theory of generalized metric spaces, particularly to help them clearly rec-
ognize the place of Alexandroff’s idea in contemporary general topology.
The first two editions of this book were written in Chinese and published in 1995
and 2007, respectively [271, 282]. They have played an active role on the devel-
opment of the theory of spaces and mappings. The primary goal of this new edition
is to update recent developments in the area and to continue to revise the text for the
sake of clarity and accessibility for readers. In order to arouse interest about the
theory of generalized metric spaces and promote contributions of Chinese mathe-
maticians in a broader international society, this edition is published in English.
Preface to the English Edition xiii

We benefited from the input of many talented people in writing this book.
The most important person in our academic career is Professor G. Gao.1 He is our
mentor. His outstanding work in spaces and mappings and his actively promoting
Alexandroff-Arhangel’skiǐ ideas have created opportunities for us to deeply engage
in the study of this subject [140, 142]. We are very grateful to Prof. Gao’s thorough
coaching, invaluable advice, altruistic helps and constant encouragements since
1979. He has laid down the learning and research foundations for us, pointed out
the direction of further exploration to us and given us courage and confidence in our
research. This book reflects Professor Gao’s research style and academic thinking to
a certain extent. We also thank Professor Alexander V. Arhangel’skiǐ for his
valuable comments on improving the writing of this book.
This edition is supported by the fund projects “The three spaces problems in
paratopological groups” and “The coverage problems in sensor networks based on
local information and rough set theory” of the National Natural Science Foundation
of China (Project Numbers 11471153 and 61472469).

Zhangzhou, China Shou Lin

Suzhou, China Ziqiu Yun
May 2016

1
G. Gao, see “Biographies of Modern Chinese Mathematicians in the 20th Century, V. 3”, edited
by M. Chen, Jiangsu Education Press, Nanjing, 1998, 287–297.
Contents

1 The Origin of Generalized Metric Spaces . . . . . . . . . . . . . . . . . . . . . . 1

1.1 Notations and Terminologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Distance Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Bases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.4 Stratifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.5 Networks and (mod k)-Networks . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.6 k-Networks and Weak Bases . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.7 Generalized Countably Compact Spaces . . . . . . . . . . . . . . . . . . . . 34
1.8 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2 Mappings on Metric Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.1 Classes of Mappings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.2 Perfect Mappings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.3 Quotient Mappings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.4 Open Mappings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
2.5 Closed Mappings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
2.6 Compact-Covering Mappings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
2.7 s-Mappings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
2.8 ss-Mappings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
2.9 p-Mappings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
2.10 Compact Mappings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
2.11 r-Locally Finite Mappings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
3 Generalized Metric Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
3.1 Spaces with Point-Countable Covers . . . . . . . . . . . . . . . . . . . . . . 148
3.2 R-Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
3.3 r-Spaces and Semi-Stratifiable Spaces . . . . . . . . . . . . . . . . . . . . . 180
3.4 k-Semi-Stratifiable Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
3.5 Mi -Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
3.6 Developable Spaces and p-Spaces . . . . . . . . . . . . . . . . . . . . . . . . 212

xv
xvi Contents

3.7 M-Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225

3.8 @-Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
3.9 g-Metrizable Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
3.10 Several Open Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
Appendix A: Characterizations of Several Covering Properties . . . . . . . 259
Appendix B: The Formation of the Theory of Generalized
Metric Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
Chapter 1
The Origin of Generalized Metric Spaces

Metrization theory is the core in the study of general topology and the theory of
generalized metric spaces is a generalization of this theory. By the opinion of Burke
and Lutzer [88], the theory of generalized metric spaces has its roots mainly in
three problems: the metrization problem, the problem of paracompactness in product
spaces and the mutual classification problem of spaces and mappings suggested by
Alexandroff in 1961 [3].
What is a “generalized metric space”? Perhaps, all the topological properties
which are weaker than those for metrization can be called generalized metric prop-
erties. However, this interpretation is too broad. Roughly speaking, the class of gen-
eralized metric spaces are such classes of spaces that they posses some properties
which are useful in characterizing metrizability, and inherit many delicate natures of
metric spaces, and some of the theory or techniques of metric spaces can be carried
over to these classes [162, 193].
Hodel [193] pointed out: “There are a number of reasons why generalized metriz-
able spaces are worthy of study. Perhaps the most important reason is that such
classes increase our understanding of the metrizable spaces. But in addition, topolo-
gists are continually seeking broader classes of spaces for which especially important
results hold.” For example, the Katětov–Morita dimension theorem, the Dugundji
extension theorem and the Borsuk homotopy extension theorem etc. can be estab-
lished in these spaces. Because of these reasons, the theory of generalized metric
spaces has been one active research direction in general topology since the 1960s. A
large number of problems of this theory which are formed by the mutual blending
of other branches, such as axiomatic set theory, mathematical logic, combinatorics
mathematics, functional analysis, topological algebra, dynamic system and computer
science etc., have been included in the books “Open Problems in Topology” [342],
“Open Problems in Topology II” [392] and “Problems from Topology Proceedings”
[391]. The achievements of the theory of generalized metric spaces between 1960
and 2015 have been summarized in some important works, such as Arhangel’skiı̌
[31, 34], Burke and Lutzer [88], G. Gao [140, 141], Gruenhage [162, 163, 164, 166],
© Atlantis Press and the author(s) 2016 1
S. Lin and Z. Yun, Generalized Metric Spaces and Mappings,
Atlantis Studies in Mathematics 6, DOI 10.2991/978-94-6239-216-8_1
2 1 The Origin of Generalized Metric Spaces

Hodel [193], Kodama and Nagami [230], S. Lin [282, 285], Morita and Nagata [364],
Nagata [373, 376]. Numbers of challenging problems proposed by many scholars,
together with some of long-term unresolved problems, have become the birthplaces
of further development of the theory of generalized metric spaces. Just as Hodel
[193] pointed out after summing up the staggering achievement of the theory of
generalized metric spaces: “But more important perhaps is the fact that the study of
generalized metrizable spaces is by no means complete; rather, it continues to grow
with many new and important results appearing every year.”
In this chapter, we derive most of the generalized metric spaces that will be
discussed in this book from three perspectives: distance functions, bases or their
extensions and generalized countable compactness. At the same time, we give some
simple characterizations of these spaces, describe their basic operation properties,
such as hereditary properties, productive properties and properties preserved under
topological sums, and discuss their preliminary relationships with some other classes
of spaces.

1.1 Notations and Terminologies

Through the Book, Spaces are All Hausdorff Topological Spaces

In this section, we first give definitions of the notations and terminologies used
frequently in this book, and then list several classic results.

1.1.1 Sets of Real Numbers and Cardinal Numbers

We denote the real line by R. The sets of natural numbers, positive integers, ratio-
nal numbers, irrational numbers and nonnegative real numbers are denoted as ω,
N, Q, P and R+ respectively. The letter ω also denotes the smallest infinite ordinal.
The unit closed interval is denoted by I. Denote S1 = {0} ∪ {1/n : n ∈ N}. The car-
dinalities of N and R are denoted by ℵ0 and c respectively. The smallest uncountable
cardinality (resp. ordinal) is denoted as ℵ1 (resp. ω1 ).

1.1.2 Operations on Subsets of Topological Spaces

The topology and the set of closed subsets of a space X are denoted by τ (X ) and
τ c (X ) respectively, and when there is no ambiguity, τ (X ) and τ c (X ) are denoted as
τ and τ c respectively. Let A and (Y, τ  ) be a subset and a subspace of X respectively
and let Z ⊂ Y . Then
1.1 Notations and Terminologies 3

the closure of A in X is denoted as A or cl(A);

the interior of A in X is denoted as A◦ or int(A);
the boundary of A in X is denoted as ∂ A;
the set of accumulation points of A in X is denoted as Ad ;
the closure of Z in Y is denoted as clY (Z ) or clτ  (Z );
the interior of Z in Y is denoted as intY (Z ) or intτ  (Z ).

1.1.3 Families of Sets in Topological Spaces

In the sequel, a family always means a family of sets. For any space X , denote

K (X ) = {K ⊂ X : K is a compact set in X }, and

S (X ) = {S ⊂ X : S is a convergent sequence in X with its limit point}.

Each finite subset of X is regarded as a trivial convergent sequence. A sequence {xn }

in X is called a nontrivial sequence, if xn = xm whenever n = m.
Let P be a family in X . Then

P <ω = {F ⊂ P : F is finite};
P F = {∪F : F ∈ P <ω };
∪ P = ∪{P : P ∈ P} is called the union of P;
∩ P = ∩{P : P ∈ P} is called the intersection of P;
P − = P = {P : P ∈ P} is called the closure of P;
P ◦ = {P ◦ : P ∈ P} is called the interior of P;
 
P= {P : P ∈ P} is called the topological sum of P.

For every A ⊂ X and x ∈ X , denote

(P) A = {P ∈ P : P ∩ A = ∅}, (P)x = (P){x} ;

st(A, P) = ∪(P) A , st(x, P) = ∪(P)x ;
stn+1 (A, P) = st(stn (A, P), P), n ∈ N;
P|A = {P ∩ A : P ∈ P}.

If F is also
 a family in X , denote P ∧ F = {P ∩ F : P ∈ P, F ∈ F }. The
definition of α∈Γ Pα is similar.
4 1 The Origin of Generalized Metric Spaces

1.1.4 Mappings on Spaces

Assume that X and Y are spaces and f : X → Y is a mapping.

Suppose A ⊂ X , the restriction of f to A to be the function f |A : A → f (A)
whose rule is f |A (x) = f (x), for each x ∈ A.
For each B ⊂ Y , f | f −1 (B) is denoted by f B and called the restriction of f to B.
If P is a family in X , then f (P) = { f (P) : P ∈ P} is called the image of P
under f .
If F is a family in Y , then f −1 (F ) = { f −1 (F) : F ∈ F } is called the inverse
image of F under f or the preimage of F under f .
Suppose X, Y, Z and W are spaces. If f : X → Y, g : X → Z , and h : W → Z ,
then the diagonal mapping f Δg : X → Y × Z and the product mapping

f ×h : X ×W →Y × Z

are defined as ( f Δg)(x) = ( f (x), g(x)) and

 ( f × h)(x, w)
 = ( f (x), 
h(w)) respec-
tively.
 The mappings Δ α∈Γ f α : X → α∈Γ Yα and α∈Γ f α : α∈Γ X α →
α∈Γ Yα can be defined similarly.
 X onto X is denoted by id X .
The identity mapping from 
For the product space α∈Γ X α and any β ∈Γ , the projection of α∈Γ X α to
the β-th coordinate space X β is denoted as πβ : α∈Γ X α → X β .

1.1.5 Operations of Spaces

Suppose Φ is a topological property.


(1) The property Φ is said to be additive if the topological sum α∈Λ X α has Φ
whenever {X α }α∈Λ is a family of spaces having Φ.
(2) The property Φ is said to be hereditary (resp. open hereditary, closed hereditary)
if any subspace (resp. open subspace, closed subspace) of X has Φ whenever
X has Φ.
(3) The property Φ is said to be productive
 (resp. finitely productive, countably
productive) if the product space α∈Λ X α has Φ whenever {X α }α∈Λ is a family
(resp. finite family, countable family) of spaces with the property Φ.
(4) The property Φ is said to be preserved or invariant (resp. inversely preserved or
inversely invariant) under a class of mappings L , provided that if f : X → Y is
an onto mapping with f ∈ L and X (resp. Y ) has the property Φ, then Y (resp.
X ) also has Φ.
For the sake of convenience, we will use terms “Φ space”, “Φ property” and “Φ
property of space” without distinction in the sequel.
1.1 Notations and Terminologies 5

1.1.6 Several Classic Results

For future reference, we list in this section several classic lemmas and theorems in
general topology [119] which will be used in the subsequent sections.
(1) The Zermelo theorem on well-ordering. On every set X there exists a relation
which well-orders X .
(2) The continuum hypothesis (abbreviated as CH). c = ℵ1 .
(3) The Urysohn-Tychonoff metrization theorem. Every regular space with a count-
able base can be embedded in the Hilbert cube Iℵ0 , and hence is a metrizable
space.
(4) The Urysohn lemma. A space X is a normal space if and only if for every
pair A, B of disjoint closed subsets of X , there exists a continuous function
f : X → I such that f (A) ⊂ {0} and f (B) ⊂ {1}.
(5) The Urysohn extension theorem. A space X is a normal space if and only if every
real valued continuous function from a closed subspace of X is continuously
expandable over X .
(6) The Tychonoff theorem. Every Cartesian product of compact spaces is compact.
(7) The Tychonoff compact extension theorem. A space is a completely regular
space if and only if there exists a compact extension of it.
(8) The Baire category theorem. The space R is second category, i.e. the intersection
of countably many dense open subsets of R is dense in R, and hence R is not
the union of countably many nowhere dense closed subspaces of R.
(9) The diagonal theorem. If every f α : X → Yα is continuous mapping and
the family { f α }α∈Λ
 separates points from closed sets in X , then the diagonal
Δα∈Λ f α : X → α∈Λ Yα is an embedding, i.e. Δα∈Λ f α : X → Δα∈Λ f α (X ) is
a homeomorphic mapping.

1.2 Distance Functions

One of the reasons for metric spaces being accepted easily by mathematicians is the
existence of a metric. By distance functions, we introduce in this section the concepts
of metric spaces, symmetric spaces, semi-metric spaces and relevant developable
spaces, quasi-developable spaces. The Stone theorem is proved in this section.
To prove proposition “(A) ⇒ (B) for a space X ”, sometimes we assume that
the condition of (A) is true in X . In the sequel, we shall do in that way without any
explanation.

Definition 1.2.1 A function d : X × X → R+ is called a distance (resp. pseudo-

distance) on X if d satisfies the following conditions for all x, y, z ∈ X :
(1) d(x, y) = 0 if and only if x = y (resp. d(x, x) = 0),
(2) d(x, y) = d(y, x),
(3) d(x, y)  d(x, z) + d(z, y).
6 1 The Origin of Generalized Metric Spaces

For every A, B ⊂ X and x ∈ X , define

d(A, B) = inf{d(x, y) : x ∈ A, y ∈ B};

d(x, B) = d(B, x) = d({x}, B).

For each ε > 0, let

B(x, ε) = {y ∈ X : d(x, y) < ε}.

Then X is called a metric space (resp. pseudo-metric space) if X is a topological

space generated by the base {B(x, ε) : x ∈ X, ε > 0}. This topology is also called
the metric topology generated by d. If the topology for X is generated by a distance
d, then X is called a metrizable space and d is called a metric on X . 
Suppose {(X α , dα )}α∈Λ is a family of metric spaces. Let X = α∈Λ X α and
define d : X × X → R+ as follows:

min{dα (x, y), 1}, there exists α ∈ Λ such that x, y ∈ X α ,
d(x, y) =
1, otherwise.

Then (X, d) is a metric space and d is called the standard topological sum metric.
Metric spaces are general mathematics research objects which have many good
properties. For example, metric spaces are additive, hereditary and countably pro-
ductive. In the theory of metric spaces, the A.H. Stone theorem is one of the most
important results.

Definition 1.2.2 ([62]) A family P in a space X is said to be discrete at a point

x ∈ X if there is a neighborhood V of x such that V meets at most one element of
P. If P is discrete at every point x ∈ X , then P is said to be discrete in X . The
union of countably many discrete families is called a σ -discrete family. Generally,
if Φ is a property of families, then the union of countably many families having Φ
is called a σ -Φ family.

Theorem 1.2.3 [440] (The Stone theorem) Every metric space is a paracompact
space.

Proof Suppose {Uα }α∈Λ is an open cover of a metric space (X, d). For every α ∈ Λ
and n ∈ N, define
(3.1) Uα,n = {x ∈ X : B(x, 1/2n ) ⊂ Uα }.
Then Uα = n∈N Uα,n , and x ∈ Uα,n if and only if d(x, X − Uα )  1/2n .
Hence,
(3.2) if x ∈ Uα,n and y ∈
/ Uα,n+1 , then d(x, y) > 1/2n+1 .
By the Zermelo theorem on well-ordering, we may assume < is a well-ordering
on Λ. Let 
∗
(3.3) Uα,n = Uα,n − γ <α Uγ ,n+1 , α ∈ Λ.
Then for every α, β ∈ Λ with α = β, according to α < β or β < α, by (3.3)
we obtain
1.2 Distance Functions 7

∗ ∗
(3.4) Uβ,n ⊂ X − Uα,n+1 or Uα,n ⊂ X − Uβ,n+1 .
∗ ∗
If x ∈ Uα,n and y ∈ Uβ,n , then by (3.3) and (3.4), when α < β, x ∈ Uα,n and
y ∈ / Uα,n+1 ; when β < α, y ∈ Uβ,n and x ∈ / Uβ,n+1 . Thus, we always have
d(x, y) > 1/2n+1 by (3.2), i.e.
∗ ∗
(3.5) d(Uα,n , Uβ,n )  1/2n+1 .
For each x ∈ X , let α be the least element in Λ such that x ∈ Uα . Take n ∈ N
∗
such that x ∈ Uα,n . By (3.3), x ∈ Uα,n
 , and it means that
∗
(3.6) α∈Λ,n∈N Uα,n = X .
For every α ∈ Λ and n ∈ N, define
+ ∗
(3.7) Uα,n = {x ∈ X : d(x, Uα,n ) < 1/2n+3 }.
Then
∗ +
(3.8) Uα,n ⊂ Uα,n ⊂ Uα .
+ +
By (3.5), (3.7) and the triangle inequality, it is easy to prove d(Uα,n , Uβ,n ) 
1/2 n+2
for every α, β ∈ Λ with α = β. So for each x ∈ X , B(x, 1/2 ) meets at
n+3
+ +
most one element of {Uα,n }α∈Λ , and it follows that {Uα,n }α∈Λ is a discrete family in
+
X consisting of open sets. Hence, {Uα,n }α∈Λ,n∈N is a σ -discrete open refinement of
{Uα }α∈Λ , and that proves X is a paracompact space. 

Definition 1.2.4 A function d : X × X → R+ is called a symmetric [382] on a set

X if for every x, y ∈ X ,
(1) d(x, y) = 0 if and only if x = y;
(2) d(x, y) = d(y, x).
A space (X, τ ) is said to be a symmetrizable space [7] if there is a symmetric d on X
such that the topology τ given on X is generated by the symmetric d, that is, a subset
U ∈ τ if and only if for each x ∈ U , there exists ε > 0 such that B(x, ε) ⊂ U .
Suppose d is a symmetric on X . Then a necessary and sufficient condition for
(X, d) being a symmetrizable space is that for any A ⊂ X , A is closed in X if and
only if for each x ∈ X − A, d(x, A) > 0. It is easy to verify that symmetrizability
is an additive, open hereditary and closed hereditary property.
A space X is called an ℵ1 -compact space [214] if every discrete closed subspace of
X is countable. Obviously, Lindelöf spaces and closed hereditarily separable spaces
are ℵ1 -compact spaces, and every locally finite family in an ℵ1 -compact space is
countable.

Theorem 1.2.5 ([380, 381]) Every symmetrizable ℵ1 -compact space is a hereditar-

ily Lindelöf space.

Proof Suppose (X, d) is a symmetrizable ℵ1 -compact space. To prove X is a hered-

itarily Lindelöf space, we only need to prove every open subspace of X is a Lindelöf
space. Suppose Y is an open subset of X an U = {Uα : α ∈ Λ} is an open cover of
Y , where Λ is a well-ordering set. For every n ∈ N and α ∈ Λ, let

Mαn = {x ∈ Y : x ∈ Uα − Uβ , B(x, 1/n) ⊂ Uα }.
β<α
8 1 The Origin of Generalized Metric Spaces

If Mαn = ∅, then pick yα,n ∈ Mαn . Let Dn = {yα,n : α ∈ Λ}.

Claim. Dn is a closed discrete set in the subspace Y . 
Let E ⊂ Dn and x ∈ Y − E. Take λ ∈ Λ and m  n such that x ∈ Uλ − β<λ Uβ
and B(x, 1/m) ⊂ Uλ . We prove B(x, 1/m)  ∩ E ⊂ {yλ,n }. In fact, suppose yα,n ∈
B(x, 1/m). If λ < α, then yα,n ∈ Uα − β<α Uβ ⊂ Uα − Uλ , so yα,n ∈ / B(x, 1/m).
It follows that α  λ. If α < λ, then x ∈ B(yα,n , 1/m) ⊂ B(yα,n , 1/n) ⊂ Uα , a
contradiction. Thus B(x, 1/m) ∩ E ⊂ {yλ,n }, which means E is a closed subset of
Y . Therefore, Dn is 
a closed discrete set in the subspace Y .
Obviously, Y = n∈N,α∈Λ Mαn .
Let F = X − Y . Then F is closed in X , and hence

F= {x ∈ X : d(x, F) < 1/m}.
m∈N

If there is k ∈ N such that {Mαk : α ∈ Λ} is uncountable, then there exist an

uncountable subset Λ of Λ, m ∈ N and a subset {xα : α ∈ Λ } of X , such that,
xα ∈ Mαk and d(xα , F)  1/m. For every α, β ∈ Λ , if β < α, then xβ ∈ Uβ
and xα ∈ / Uβ , so xα = xβ . Let M = {xα : α ∈ Λ }. Since X is ℵ1 -compact, the
uncountable set M is not a closed discrete subset of X . We may assume there is
y ∈ X − M such that d(y, M) = 0. Then there is α ∈ Λ such that d(y, xα ) < 1/m.
Since d(xα , F)  1/m, y ∈ X − F = Y . By the above claim, M is a closed set in Y ,
and it follows from d(y, M) = 0 that y ∈ M, a contradiction. That means for each
n ∈ N, the family {Mαn : α ∈ Λ} is countable. So, there is a countable subfamily of
U covering Y . Therefore, Y is a Lindelöf space. 

Definition 1.2.6 ([477]) A symmetrizable space (X, d) is called a semi-metrizable

space and d is called a semi-metric on X if, for each x ∈ X , {B(x, ε) : ε > 0} forms
a (not necessarily open) neighborhood base at x.
If (X, d) is a symmetrizable space, then a necessary and sufficient condition for
(X, d) being a semi-metrizable space is that for each A ⊂ X , x ∈ A if and only if
d(x, A) = 0. It is easy to verify that semi-metrizability is an additive and hereditary
property.

Definition 1.2.7 Suppose X is a space.

(1) X is said to be a Fréchet–Urysohn space [133] if, for every A ⊂ X and x ∈ A,
there is a sequence in A converging to x.
(2) A function g : N × X → τ is called a g-function [176] (or decreasing g-function
[491]) on X if, for every x ∈ X and n ∈ N, x ∈ g(n + 1, x) ⊂ g(n, x). If
 g-functions are always denoted by g. For each A ⊂ X , define
not specified,
g(n, A) = x∈A g(n, x).
(3) If there exists a sequence {Fn } of covers of a space X , such that, for each
x ∈ X , {st(x, Fn )}n∈N is a neighborhood base of x in X , then X is called a
semi-developable space [1] and {Fn } is called a semi-development for X .
1.2 Distance Functions 9

Theorem 1.2.8 The following are equivalent:

(1) X is a semi-metrizable space.
(2) There is a semi-metrizable function on X , i.e. there is a g-funct-
ion on X satisfying the following condition: for any x ∈ X and any sequence
{xn } in X with xn ∈ g(n, x) or x ∈ g(n, xn ), xn → x [176].
(3) There is a semi-development for X [1].
(4) X is a first countable (or Fréchet–Urysohn) symmetrizable space [31].

Proof (1) ⇒ (3). Suppose X is a semi-metrizable space. For each n ∈ N, let

Fn = {A ⊂ X : diam A < 1/n},

where diam A = sup{d(x, y) : x, y ∈ A}. Then for each x ∈ X , st(x, Fn ) =

B(x, 1/n). So {st(x, Fn )}n∈N is a neighborhood base of x.
(3) ⇒ (2). Suppose {Fn } is a semi-development for X . We may assume that
Fn+1 refines Fn . Define g : N × X → τ by g(n, x) = st(x, Fn )◦ . Then g is a
semi-metrizable function on X .
(2) ⇒ (4). Suppose g is a semi-metrizable function on X . Obviously, X is a first
countable space. For every x, y ∈ X with x = y, let

m(x, y) = min{n ∈ N : y ∈
/ g(n, x), x ∈
/ g(n, y)}.

Define d : X × X → R+ as follows:

0, x = y,
d(x, y) =
1/m(x, y), x = y.

Then d is a semi-metric on X .
(4) ⇒ (1). Suppose X is a Fréchet–Urysohn space and d is a symmetric on X . If
A ⊂ X and x ∈ A − A, then there is a sequence {xn } in A converging to x. Assume
d(xn , x)  0. Take a subsequence Z of {xn } such that d(x, Z ) > 0. Let T = Z ∪{x}.
Then T is a closed subspace of X . It follows that (T, d) is a symmetrizable space,
and hence x is an isolated point in T , a contradiction. Thus, d(xn , x) → 0, so
d(x, A) = 0. 

Corollary 1.2.9 Semi-metrizability is countably productive.

Proof Suppose gn is a semi-metrizable function on X n for each n ∈ N. Define



 
g :N× Xn →τ Xn
n∈N n∈N

  
by g(m, x) = ( nm gn (m, xn ))× n>m X n , where x = (xn ) ∈ n∈N X n . Then g
is a semi-metrizable function on n∈N X n . 
10 1 The Origin of Generalized Metric Spaces

Definition 1.2.10 ([113]) A neighborhood assignment for a space (X, τ ) is a func-

tion φ : X → τ satisfying x ∈ φ(x) for each x ∈ X . X is called a D-space if, for
each neighborhood assignment φ for X , there is a discrete closed set D in X such
that {φ(d) : d ∈ D} covers X .
van Douwen and Pfeffer [114] raised several interesting questions on D-spaces.
Among them, the question that whether every regular Lindelöf space is a D-space is
still open. However, numbers of generalized metric spaces were found to be D-spaces
[36, 165].

Theorem 1.2.11 ([84]) Symmetrizable spaces are hereditarily D-spaces.

Proof Suppose (Z , d) is a symmetrizable space and X ⊂ Z with the subspace

topology τ . To show X is a D-space, let φ : X → τ be a neighborhood assignment
for X and let U = {φ(x) : x ∈ X }. Since open subspaces of symmetrizable
spaces are symmetrizable,
 we may assume each φ(x), x ∈ X , is actually open in
Z andZ = x∈X φ(x). We need to find a closed discrete subset D of X such that
X ⊂ x∈D φ(x).
For each x ∈ X let k x ∈ N such that B(x, 1/k x ) ⊂ φ(x). For each n ∈ N, let
In = {x ∈ X : k x = n}. Well-order X so that when y ∈ Ii , z ∈ I j for every i < j
then y < z.
Now, recursively find Jn ⊂ In as follows: for each x ∈ In ,

x ∈ Jn ⇔ x = min((In − ∪{φ(y) : i < n, y ∈ Ji }) − ∪{φ(y) : y ∈ Jn , y < x}).


Let D = n∈N Jn .

Claim 1. W = {φ(x) : x ∈ D} covers X .

For any y ∈ X find m ∈ N such that y ∈ Im . If y ∈ Jm , then certainly y ∈ φ(y) ⊂
∪W . If y ∈
/ Jm , then
 
y∈( {φ(x) : i < m, x ∈ Ji }) ∪ ( {φ(x) : x ∈ Jm }) ⊂ ∪W .

Claim 2. D is a closed discrete set in ∪W (and hence in X ).

This follows if we show that for any t ∈ D, D − {t} is closed. To this end
we may assume Z = ∪W and let x ∈ Z − (D − {t}). It suffices to prove that
B(x, 1/n) ∩ (D − {t}) = ∅ for some n ∈ N. Let m be the first element of N such that
there is a first element y of Jm with x ∈ φ(y). Now, for all z ∈ D with y < z we have
z∈/ φ(y). Also, for all z ∈ D with z < y we have B(z, 1/m) ⊂ B(z, 1/k z ) ⊂ φ(z)
and x ∈/ φ(z), so z ∈ / B(x, 1/m). That is, φ(y) ∩ B(x, 1/m) ∩ (D − {y}) = ∅, then
B(x, 1/n)∩(D −{y}) = ∅ for some n ∈ N. If t = y, then we are done. If t = y, then
since y ∈ D we know x = y and there is j ∈ N such that y ∈ / B(x, 1/j). This gives
that B(x, 1/j)∩φ(y)∩ B(x, 1/m)∩(D −{t}) = ∅, then B(x, 1/n)∩(D −{t}) = ∅
for some n ∈ N. 

Definition 1.2.12 A sequence {Un } of open covers on X is called a development

for X if it is a semi-development for X . A space with a development is called a
1.2 Distance Functions 11

developable space [8] and a regular developable space is called a Moore space [357].
A sequence {Un } of families consisting open sets in X is called a quasi-development
for X if for every x ∈ U ∈ τ , there is n ∈ N such that x ∈ st(x, Un ) ⊂ U . A space
with a quasi-development is called a quasi-developable space [55].
A space is called a perfect space if each closed set in this space is a G δ -set.
Theorem 1.2.13 The following are equivalent:
(1) X is a developable space.
(2) There is a developable function on X , i.e. there is a g-function on X satisfying
the following condition: for each x ∈ X and any sequences {xn }, {yn } in X with
{x, xn } ⊂ g(n, yn ), xn → x [176].
(3) X is a quasi-developable perfect space [55].
Proof (1) ⇒ (2). Suppose {Un } is a development for X . For every x ∈ X and n ∈ N,
take Un ∈ Un such that x ∈ Un . Let g(n, x) = in Ui . Then g is a developable
function on X .
(2) ⇒ (3). Suppose g is a developable function on X . Obviously, for any closed
subset A of X , A = n∈N g(n, A), and hence X is perfect. For every n ∈ N, let
Un = {g(n, x) : x ∈ X }. Then {Un } is a development, hence a quasi-development
for X .
(3) ⇒ (1). Suppose {Un } is a quasi-development for X and X is a perfect space.
 each n ∈ N, there is a sequence {Fn, j } j∈N of closed sets in X such that ∪Un =
For
j∈N Fn, j . Let Hn, j = Un ∪ {X − Fn, j }. Then {Hn, j }n, j∈N is a development for
X. 
Developable spaces and quasi-developable spaces are additive, hereditary and
countably productive. Every developable space is a semi-metric subparacompact
space (see Theorem A.3.3 in Appendix A). The following theorem gives a simple
sufficient condition for a semi-metric space to be a developable space. A base B for
a space X is said to be point-countable if each point of X only belongs to at most
countably many elements of B.
Theorem 1.2.14 ([178]) Every semi-metric space with a point-countable base is a
developable space.
Proof Suppose U is a point-countable base for a semi-metric space X and  is a
well-ordering on X . For each x ∈ X , denote (U )x = {Un (x)}n∈N . For each n ∈ N,
let

Vn (x) = B(x, 1/n)◦ ,

h(n, x) = Un (x) ∩ Vn (x),
p(n, x) = min{y ∈ X : x ∈ h(n, y)},
g(n, x) = Vn (x) ∩ (∩{h(i, p(i, x)) : i  n})
∩ (∩{U j ( p(i, x)) : j, i  n, x ∈ U j ( p(i, x))}),
Gn = {g(n, x) : x ∈ X }.
12 1 The Origin of Generalized Metric Spaces

Then {Gn } is a development for X . Because otherwise, there exist x ∈ X and an

open neighborhood W of x satisfying the following condition: for each i ∈ N, there
is xi ∈ X such that x ∈ g(i, xi ) ⊂ W . Since x ∈ Vi (xi ), xi → x. Pick l, m ∈ N
such that B(x, 1/l) ⊂ Um (x) ⊂ W . If x ∈ h(l, y) ⊂ Vl (y) for some y ∈ X , then
y ∈ B(x, 1/l) ⊂ Um (x), and hence p(l, x) ∈ Um (x). Thus, there is k ∈ N such that
Um (x) = Uk ( p(l, x)). Since Uk ( p(l, x)) ∩ h(l, p(l, x)) is an open neighborhood of
x, there is i 0 ∈ N such that xi ∈ Uk ( p(l, x))∩h(l, p(l, x)) when i  i 0 , so p(l, xi ) 
p(l, x) when i  i 0 . On the other hand, if i  l, then x ∈ g(i, xi ) ⊂ h(l, p(l, xi )),
and hence p(l, x)  p(l, xi ). So p(l, xi ) = p(l, x) when i  max{i 0 , l}, it follows
that xi ∈ Uk ( p(l, xi )). Hence g(i, xi ) ⊂ Uk ( p(l, xi )) = Uk ( p(l, x)) = Um (x) ⊂
W when i  max{i 0 , l, k}, a contradiction. This contradiction shows that X is a
developable space. 

1.3 Bases

The problem of seeking metrization theorems for spaces, i.e. looking for character-
izations of metrizable spaces, is a general problem produced after proposing of the
concepts of metric spaces and topological spaces. This problem was solved perfectly
in the 1950s.
Definition 1.3.1 A family P in a space X is said to be locally finite at a point
x ∈ X [5], if there is a neighborhood V of x such that V meets at most finitely many
elements of P. If P is locally finite at every point x ∈ X , then P is said to be
locally finite in X .
Bing, Nagata and Smirnov obtained their classic metrization theorem after the
Stone theorem.
Theorem 1.3.2 (The Bing–Nagata–Smirnov metrization theorem) The following
are equivalent for every regular space X :
(1) X is a metrizable space.
(2) X has a σ -discrete base [62].
(3) X has a σ -locally finite base [370, 428].
Proof (1) ⇒ (2). Suppose (X, d) is a metrizable space. For each n ∈ N, let Bn =
{B(x, 1/2n )}x∈X . Then st(x, Bn ) ⊂ B(x, 1/n). By the Stone theorem (see Theorem
1.2.3), Bn has a σ -discrete open refinement Vn . It is easy to verify that n∈N Vn is
a σ -discrete base of X .
(2) ⇒ (3) is obvious. We prove (3) ⇒ (1). Suppose X has a σ -locally finite base
B. For any open cover U of X , let V = {B ∈ B : there is U ∈ U such that B ⊂
U }. Then V is a σ -locally finite refinement
 of U . So X is a paracompact space,
hence a normal space. Denote B = n∈N Bn , where each Bn = {Bα : α ∈ Λn } is
locally finite. For every n, m ∈ N and α ∈ Λn , let

Aα = ∪{A ∈ Bm : A ⊂ Bα }.
1.3 Bases 13

Then Aα ⊂ Bα . By the Urysohn lemma, there is a continuous function f α : X → I

such that f α (Aα ) ⊂ {1} and f α (X − Bα ) ⊂ {0}. For every x, y ∈ X , define
  
dn,m (x, y) = min 1, | f α (x) − f α (y)| .
α∈Λn

Then dn,m is a pseudo-distance on X . Rearrange {dn,m }n,m∈N as {dk }k∈N . Define

ρ : X × X → R+ by
∞
dk (x, y)
ρ(x, y) = .
k=1
2k

Then ρ is a distance on X and the metric topology generated by ρ is just the original
topology for X . Thus X is a metrizable space. 

(1) ⇔ (2) and (1) ⇔ (3) of Theorem 1.3.2 are called the Bing metrization
theorem and the Nagata-Smirnov metrization theorem respectively. By them we can
obtain more metrization theorems.

Definition 1.3.3 ([63]) A family P of subsets of X is said to be point-finite if (P)x

is finite for every x ∈ X ; P is said to be compact-finite if (P) K is finite for every
K ∈ K (X ).
Obviously, locally finite ⇒ compact-finite ⇒ point-finite.

Corollary 1.3.4 ([63]) A space X is a metrizable space if and only if X is a regular

space with a σ -compact-finite base.

Proof We only need  to prove the sufficiency. Suppose B is a σ -compact-finite base

for X . Let B = n∈N Bn , where each Bn is compact-finite. We first prove each Bn
is locally finite. Otherwise, there is n ∈ N such
that Bn is not locally finite at some
point x ∈ X . Let (B)x = {Bi }i∈N and Pk = ik Bi for each k ∈ N. Then (Bn ) Pk
is infinite for each k ∈ N, and hence there exist an infinite subset {Q k }k∈N of Bn and
a sequence {xk } in X such that xk ∈ Pk ∩ Q k . Let F = {x} ∪ {xk : k ∈ N}. Then F
is a compact set in X , which contradicts the assumption that Bn is compact-finite.
So B is σ -locally finite base of X , and hence X is a metrizable space. 

Let X be a space and A ⊂ X . A family P of X is call a neighborhood base of A

in X if, A ⊂ P ◦ for each P ∈ P, and for each open set U in X containing A, there
exists P ∈ P such that P ⊂ U .

Theorem 1.3.5 The following are equivalent for any space X :

(1) X is a metrizable space.
(2) There is a sequence {Un } of open covers of X such that for each x ∈
X, {st2 (x, Un )}n∈N is a neighborhood base of x in X [358].
(3) There is a sequence {Un } of open covers of X such that for each K ∈
K (X ), {st(K , Un )}n∈N is a neighborhood base of K in X [215].
14 1 The Origin of Generalized Metric Spaces

(4) There is a development {Un } for X such that Un+1 is a star-refinement of Un

[469].
(5) X is a collectionwise normal developable space [62] (the Bing metrization cri-
terion).

Proof (1) ⇒ (2), (3) and (4). Suppose X is a metrizable space. For each n ∈ N, let

Un = {U ∈ τ (X ) : diamU < 1/2n }.

Then {Un } satisfies conditions (2) and (3). Since X is a paracompact  space, we can
take H1 = U1 and let Hn+1 be an open star-refinement of Un+1 ∧ ( in Hi ). Then
{Hn } satisfies condition (4).
(4) ⇒ (2). It follows from the fact that if Un+1 is a star-refinement of Un , then
st2 (x, Un+1 ) ⊂ st(x, Un ).
(3) ⇒ (2). We may assume that Un+1 refines Un and there exist x ∈ X and U ∈ τ
with x ∈ U ∈ τ , such that, st2 (x, Un ) ⊂ U for each n ∈ N. Pick xn ∈ st2 (x, Un )−U .
Then there is yn ∈ st(x, Un ) such that xn ∈ st(yn , Un ), and hence yn → x. We may
assume yn ∈ U for each n ∈ N. Let K = {x} ∪ {yn : n ∈ N}. Then K ∈ K (X ), so
there is m ∈ N such that st(K , Um ) ⊂ U . It follows that xm ∈ U , a contradiction.
(2) ⇒ (5). We may assume that Un+1 refines Un . Obviously, X is a developable
space. Let H = {Hα }α∈Λ be a discrete family of closed sets in X . For every α ∈ Λ
and x ∈ H α , there is n(x) ∈ N such that st (x, Un(x) ) ⊂ X − ∪{Hβ : α  = β ∈ Λ}.
2

Let Uα = x∈Hα st(x, Un(x) ). Then Hα ⊂ Uα ∈ τ and {Uα }α∈Λ is a disjoint family.
So X is a collectionwise normal space.
(5) ⇒ (1). Let {Un } be a development for X and X be a collectionwise  normal
space. Since X is a subparacompact space, Un has a closed refinement m∈N Fn,m
such that Fn,m is a discrete family of closed sets in X . Since X is a collectionwise
normal space, we can take a discrete family Bn,m = {B F : F ∈ Fn,m } of open sets in
 condition: for each F ∈ Fn,m , there is U ∈ Un such that
X satisfying the following
F ⊂ B F ⊂ U . Then n,m∈N Bn,m is a σ -discrete base of X . So X is a metrizable
space.

Definition 1.3.6 (1) A family P in X is said to be closure-preserving at x ∈ X if,

for any P  ⊂ P, x ∈ ∪P  whenever x ∈ ∪P  . For a subset A of X , if P is
closure-preserving at every x ∈ A, then P is said to be closure-preserving in A
[326]. A regular space with a σ -closure-preserving base is called an M1 -space
[92].
(2) A family B in X is called a quasi-base for X if, for every x ∈ X and U ∈ τ
with x ∈ U ∈ τ , there is B ∈ B such that x ∈ B ◦ ⊂ B ⊂ U . A regular space
with a σ -closure-preserving quasi-base is called an M2 -space [92].
(3) A family P of ordered pairs of subsets of X is called a pair-base for X if P
satisfies the following conditions:
(i) (P1 , P2 ) ∈ P ⇒ P1 ⊂ P2 and P1 ∈ τ ;
(ii) for each x ∈ U ∈ τ , there is (P1 , P2 ) ∈ P such that x ∈ P1 ⊂ P2 ⊂ U .
1.3 Bases 15

A family P of ordered pairs of sets in X is said to be cushioned in X [327] if,

for every P  ⊂ P, ∪{P1 : (P1 , P2 ) ∈ P  } ⊂ ∪{P2 : (P1 , P2 ) ∈ P  }. A space with
a σ -cushioned base is called an M3 -space [92].
A σ -cushioned base is usually called a σ -cushioned pair-base. Since a cushioned
family is always a family of ordered pairs, so we call it a σ -cushioned base. In this
book, we give the same treatment in all other places when we mention a cushioned
family.
M1 -spaces, M2 -spaces and M3 -spaces are unified called Mi -spaces [92]. It is
obvious that metrizable ⇒ M1 ⇒ M2 ⇒ M3 ⇒ paracompact. Below we investigate
the properties of Mi -spaces. Obviously, Mi -spaces are additive.

Proposition 1.3.7 (1) M1 -spaces are open hereditary.

(2) M2 -spaces and M3 -spaces are hereditary.

Proof (1) Assume that B is a σ -closure-preserving base for an M1 -space X . If Z is

an open subspace of X , then {B ∈ B : B ⊂ Z } is a σ -closure-preserving base for
Z , so Z is also an M1 -space.
(2) We prove it only for the M2 -case. Suppose B is a σ -closure-preserving quasi-
base for an M2 -space X . If Z is a subspace of X , then B |Z a σ -closure-preserving
quasi-base for Z , and hence Z is an M2 -space. 
For families P1 and P2 of sets in spaces X and Y respectively, define

P1 × P2 = {P1 × P2 : P1 ∈ P1 , P2 ∈ P2 }.

For families P1 and P2 of ordered pairs of sets in spaces X and Y respectively,

define

P1 × P2 = {(P1 × P2 , Q 1 × Q 2 ) : (Pi , Q i ) ∈ Pi , i = 1, 2}.


We define α∈Λ Pα in a similar way.

Lemma 1.3.8 If Pi (i = 1, 2) is a locally finite (resp. closure-preserving, cush-

ioned) family for X i , then P1 × P2 is a locally finite (resp. closure-preserving,
cushioned) family for space X 1 × X 2 .

Proof We prove it only for the closure-preserving case.

Let P1 = {Fα }α∈A and P2 = {Hβ }β∈B . We prove that for any C ⊂ A × B

∪{Fα × Hβ : (α, β) ∈ C} ⊂ ∪{F α × H β : (α, β) ∈ C}.

Assume (x, y) ∈ X 1 × X 2 − ∪{F α × H β : (α, β) ∈ C}. Let

U = X 1 − ∪{F α : x ∈ X 1 − F α , (α, β) ∈ C},

V = X 2 − ∪{H β : y ∈ X 2 − H β , (α, β) ∈ C}.
16 1 The Origin of Generalized Metric Spaces

Then (x, y) ∈ U × V ∈ τ (X 1 × X 2 ) and (U × V )∩(∪{Fα × Hβ : (α, β) ∈ C}) = ∅,

/ ∪{Fα × Hβ : (α, β) ∈ C}. Therefore, P1 × P2 is a closure-preserving
so (x, y) ∈
family in X 1 × X 2 . 

Proposition 1.3.9 Mi -spaces are countably productive for i = 1, 2, 3.


Proof We prove it only for the M1 -case. For each m ∈ N, let n∈N Bn,m be a base
for an M1 -space X m , where each Bn,m is closure-preserving. For each n ∈ N, let
⎛ ⎛ ⎞⎞  
  
Bn = ⎝ ⎝ B j,m ⎠⎠ × Xm .
mn jn m>n

 
Then n∈N Bn is a σ -closure-preserving base for m∈N Xm. 

In the last part of this section, we give some basic characterizations of Mi -spaces.

Definition 1.3.10 A subset D of a space X is called a regular closed set in X if

there is an open subset W of X such that D = W . The complement of a regular
closed set is called a regular open set.
By the definition, D is a regular closed set in X if and only if D = D ◦− , and D is
a regular open set in X if and only if D = D −◦ . If P is a closure-preserving family
of regular closed sets in X , then P ◦ is a closure-preserving family of open sets in
X.

Proposition 1.3.11 A space X is an M1 -space if and only if X is a regular space

with a σ -closure-preserving quasi-base consisting of regular closed sets.

Proof If X is an M1 -space and B is a σ -closure-preserving quasi-base for X , then

B − is a σ -closure-preserving quasi-base for X consisting of regular closed sets. On
the contrary, if B is a σ -closure-preserving quasi-base for X consisting of regular
closed sets, then B −◦ is a σ -closure-preserving base for X , and hence X is an M1 -
space. 

Proposition 1.3.12 ([373]) A space X is an M2 -space if and only if there is a

g-function on X satisfying the following conditions:
(1) If y ∈ g(n, x), then g(n, y) ⊂ g(n, x).
/ g(m, H ).
(2) If y ∈ X − H ∈ τ , then there is m ∈ N such that y ∈

Proof Suppose X is an M2 -space and n∈N Bn is a quasi-base for X , where each
Bn is closure-preserving. Define g : N × X → τ by

g(n, x) = X − ∪{P ∈ B i : x ∈
/ P, i  n}.

Then g is a g-function on X satisfying conditions (1) and (2).

On the other hand, if there is a g-function on X satisfying (1) and (2), then X is
a regular space. For each n ∈ N, let
1.3 Bases 17

Bn = {X − g(n, A) : A ⊂ X }.

Since
 g satisfies (1), Bn is closure-preserving. By the assumption that g satisfies (2),
n∈N Bn is a quasi-base for X . So X is an M2 -space. 
Proposition 1.3.13 ([182]) A space X is an M3 -space if and only if there is a
g-function on X satisfying the condition: if y ∈ X − H ∈ τ , then there is m ∈ N
/ g(m, H ).
such that y ∈

Proof Suppose X is an M3 -space and n∈N Pn is a quasi-base for X such that each
Pn is a cushioned family. Define g : N × X → τ as follows:

g(n, x) = X − ∪{P1 : (P1 , P2 ) ∈ Pi , x ∈

/ P2 , i  n}.

Then g is a g-function on X we need.

On the other hand, suppose there is a g-function on X satisfying the condition
/ g(m, H ). Then X is a
that for each y ∈ X − H ∈ τ , there is m ∈ N such that y ∈
regular space. For every n ∈ N, let

Pn = {(X − g(n, X − U ), U ) : U ∈ τ }.

If τ  ⊂ τ , then

∪{X − g(n, X − U ) : U ∈ τ  } ⊂ X − g(n, X − ∪τ  ) ⊂ ∪τ  ,

and hence Pn is a cushioned family in X . For every x ∈ X and U ∈ τ with x ∈ U ∈

τ , there / g(m, X − U ), and hence x ∈ X −g(m, X − U ) ⊂ U .
 is m ∈ N such that x ∈
So n∈N Pn is a σ -cushioned quasi-base for X . 
Remark 1.3.14 For a g-function on X , consider the following conditions:
(1) If y ∈ X − H ∈ τ , then there is m ∈ N such that y ∈
/ g(m, H ).
(2) If x ∈ X and {xn } is a sequence in X such that x ∈ g(n, xn ) for each n ∈ N,
then xn → x.
(3) For each x ∈ X , {x} = n∈N g(n, x).
Then (1) ⇔ (2) ⇒ (3).
The family of spherical open sets of a metric space is the most natural base for the
metric topology. Due to the metrization theorems in this section, it is necessary for
us to investigate the local finiteness or compact-finiteness for the family of spherical
open sets. A family U is said to be star-finite if for every U ∈ U , (U )U is finite.
Theorem 1.3.15 ([53]) Suppose (X, d) is a metric space. For every ε > 0, denote
Bd (ε) = {Bd (x, ε) : x ∈ X }. Then the following are equivalent:
(1) Bd (ε) is locally finite.
(2) Bd (ε) is point-finite.
(3) Bd (ε) is star-finite.
18 1 The Origin of Generalized Metric Spaces

Proof We only need to prove (2) ⇒ (3). Suppose Bd (ε) is point-finite. For each
B ∈ Bd (ε), let B ∗ = {x ∈ B : B = Bd (x, ε)}, i.e. B ∗ is the set of centers of B.
Then B ∗ meets at most finitely many elements of Bd (ε). Because otherwise, there is
a sequence {Bn } consisting of different elements of Bd (ε) such that Bn ∩ B ∗ = ∅.
Fix xn ∈ Bn ∩ B ∗ , y ∈ B ∗ and z n ∈ Bn∗ . Then d(xn , z n ) < ε, and hence z n ∈
Bd (xn , ε) = B, so d(y, z n ) < ε, it follows that y ∈ Bn , a contradiction. If Bd (ε) is
not star-finite, then there is an element B of Bd (ε) meeting infinitely many Bn∗ s for
Bn ∈ Bd (ε). Fix xn ∈ B ∩ Bn∗ and x ∈ B ∗ . Then d(x, xn ) < ε, and hence x ∈ Bn , a
contraction. 

For each metric space X , there is a compatible metric d such that for every
ε > 0, Bd (ε) is closure-preserving [375]. However, on any hedgehog space J with
uncountably many spininess, there is no compatible metric d for J , such that for
every ε > 0, Bd (ε) is locally finite [53, 495].

1.4 Stratifications

In 1966, Borges [65] gave an important characterization of M3 -spaces in terms of

“stratification”. Borges’s method is an important way for generating generalized
metric spaces.

Definition 1.4.1 ([65]) A space X is called a stratifiable space if, there is a function
G : N × τ → τ satisfying the following conditions:

(1) n ∈ N, U ∈ τ ⇒ G(n, U ) ⊂ U = m∈N G(m, U );
(2) V ⊂ U ⇒ G(n, V ) ⊂ G(n, U ).
G is called a stratification on X and we may assume that G is increasing on n ∈ N.

Proposition 1.4.2 ([65]) For every space X , the following are equivalent:
(1) X is an M3 -space.
(2) X is a stratifiable space.
(3) There is a function H : N × τ c → τ satisfying the following conditions:

(i) n ∈ N, F ∈ τ c ⇒ H (n, F) ⊃ F = m∈N H (m, F);
(ii) L ⊂ F ⇒ H (n, L) ⊂ H (n, F).

Proof (2) ⇔ (3) is obvious.

(1) ⇔ (3). If there is a g-function satisfying the condition of Proposition 1.3.13,
then the function H : N × τ c → τ defined by H (n, F) = g(n, F) satisfies (3). On
the other hand, if H : N × τ c → τ satisfies (3), then the function g : N × X → τ
defined by g(n, x) = H (n, {x}) satisfies the condition of Proposition 1.3.13. 

In order to narrative brevity, a g-function satisfying the condition in Proposition

1.3.13 is called a stratifiable function.
1.4 Stratifications 19

Definition 1.4.3 A space X is called a semi-stratifiable space [109] if, there is a

function F : N × τ → τ c satisfying the following conditions:

(1) U ∈ τ ⇒ U = n∈N F(n, U );
(2) V ⊂ U ⇒ F(n, V ) ⊂ F(n, U ).
Further assume X is a regular space and satisfies the following condition, then X
is called a k-semi-stratifiable space [316].

(3) For every compact subset K of X with K ⊂ U ∈ τ , there is m ∈ N such that

K ⊂ F(m, U ).
We may assume F is increasing on n ∈ N. The function F defined above is called
a semi-stratification and a k-semi-stratification on X respectively,

Proposition 1.4.4 ([110]) For every space X , the following are equivalent:
(1) X is a semi-stratifiable space.
(2) There is a semi-stratifiable function on X , i.e. there is a g-function on X , such
that, for any point x and any sequence {xn } in X with x ∈ g(n, xn ), xn → x.
(3) There is a function G : N × τ c → τ satisfying the following conditions:

(i) F ∈ τ c ⇒ F = n∈N G(n, F);
(ii) L ⊂ F ⇒ G(n, L) ⊂ G(n, F).

Proof (1) ⇔ (3) is obvious. By Remark 1.3.14, (2) ⇔ (3). 

By Theorem 1.2.8 and Proposition 1.4.4, we have the following corollaries.

Corollary 1.4.5 ([110]) A space X is a semi-metrizable space if and only if X is a

first countable semi-stratifiable space.

Corollary 1.4.6 ([110]) Semi-stratifiable spaces are additive, hereditary and count-
ably productive.

Definition 1.4.7 A space X is said to have a G δ -diagonal [92] (resp. G ∗δ -diagonal

[189]) if, there is a sequence {Un } of
open covers of X , such that, for every x ∈ X ,
{x} = n∈N st(x, Un ) (resp. {x} = n∈N st(x, Un )). The sequence {Un } is called a
G δ -diagonal sequence (resp. G ∗δ -diagonal sequence) for X .
A space X is said to have a quasi-G δ -diagonal [192] if, there is a sequence {Un }
of open subsets of X , such that, for every x ∈ X , {x} = n∈N {st(x, Un ) : x ∈ ∪Un }.
The sequence {Un } is called a quasi-G δ -diagonal sequence for X .
Obviously, every G ∗δ -diagonal is a G δ -diagonal, and every G δ -diagonal is a quasi-
G δ -diagonal.

Lemma 1.4.8 ([235]) Assume that Y is a submetacompact subspace of a regular

space X . If {Un } is a sequence of families of open sets in X covering Y , then there is
a sequence {Vn } of
of open sets in X such that for each y ∈ Y ,
covers of Y consisting
n∈N st(y, V n ) = n∈N st(y, V n ) ⊂ n∈N st(y, Un ).
20 1 The Origin of Generalized Metric Spaces

Proof Since X is regular and Y is submetacompact, for every m ∈ N, there is a

sequence {Vm,n } of families of open sets of X covering Y satisfying the following
conditions:
 
(1) {Vm,n|Y }n is a θ -refinable sequence of ( i, j<m Vi, j|Y ) ∧ ( km Uk|Y );
(2) for every V ∈ Vm,n and i, j < m, there is W ∈ Vi, j such that V ⊂ W .

Now suppose y ∈ Y and x ∈ i, j∈N st(y, Vi, j ). Fix i, j ∈ N and pick m > max{i, j}.
Then there is n ∈ N such that y is only in finitely many elements of Vm,n , and hence

st(y, Vm,n ) = ∪{V : y ∈ V ∈ Vm,n } ⊂ st(y, Vi, j ).

So i, j∈N st(y, Vi, j ) = i, j∈N st(y, Vi, j ) ⊂ n∈N st(y, Un ). 
Proposition 1.4.9 (1) A space X has a G δ -diagonal if and only if the set Δ =
{(x, x) : x ∈ X } is a G δ -set in X 2 [92].
(2) Every regular submetacompact space with a G δ -diagonal has a G ∗δ -diagonal
[189].
Proof (2) is obtained by Lemma 1.4.8. We prove (1). Suppose {Un } is a G δ -diagonal
sequence for X . For every n ∈ N, let

G n = ∪{U × U : U ∈ Un }.

Then G n ∈ τ (X 2 ) and Δ = n∈N G n .
On the contrary, assume {G n } is a sequence of open sets in X such that Δ =
2

n∈N G n . For every n ∈ N, let

Un = {U ∈ τ (X ) : U × U ⊂ G n }.

Then {Un } is a G δ -diagonal sequence for X . 

A subparacompact space with a G δ -diagonal may not have a G ∗δ -diagonal [281].
Theorem 1.4.10 [93] (The Chaber theorem)
(1) Every countably compact space with a quasi-G δ -diagonal is compact.
(2) Every countably compact space with a G δ -diagonal is compact and metrizable.
Proof (1) Let X be a countably compact space and {Un } be a quasi-G δ -diagonal
sequence for X .
We prove X is compact first. Let U be an open cover of X . Assume that no
finite subfamily of U covers X . One can define, by induction, a decreasing sequence
{Fn }n∈ω of closed subsets of X such that F0 = X and, for each n ∈ N, Fn = Fn−1 −
∪Un , if this set cannot be covered by any finite subfamily of U ; Fn = Fn−1 − ∪Vn ,
if Vn is a finite subfamily of U which covers Fn−1 − ∪Un .
Since the set F0 cannot be covered by any finite subfamily of U , it is easy to see
that none of the sets Fn can be covered by any finite subfamily of U . Using that fact
1.4 Stratifications 21

that X is countably compact, we infer the closed set F = n∈ω Fn cannot be covered
by any finite subfamily of U , i.e., F cannot be covered by any finite subfamily of
U|F .
From the assumption that, for each x ∈ X , n∈N {st(x, Un ) : x ∈ ∪Un } = {x}, it
follows that
(1.1) If a set H ⊂ F cannot be covered by any countable subfamily of U|F and
x ∈ F, then there exists n ∈ N such that x ∈ ∪Un , and H − st(x, Un ) cannot be
covered by any countable subfamily of U|F .
Because otherwise, for each n ∈ N with x ∈ ∪Un , there is a countable subfamily of
U|F covering H − st(x, Un ), it follows that there is a countable subfamily of U|F
covering H − {x}, and hence H can be covered by a countable subfamily of U|F , a
contradiction.
We shall construct, by induction, a transfinite sequence {xα }α<ω1 of elements of F
and a sequence {n α }α<ω1 of natural
 numbers such that for each α < ω1
(1.2) xα ∈ (∪U
 α n ) ∩ F − β<α st(x β , Un β ),
(1.3) F − βα st(xβ , Un β ) cannot be covered by any countable subfamily of
U|F .
Let x0 ∈ F be arbitrary. Using (1.1), we can choose n 0 ∈ N such that (1.2) and (1.3)
are satisfied for α0 = 0.
Assume that α0 < ω1 and {x α }α<α0 , {n α }α<α0 are defined and satisfy (1.2) and (1.3)
for each α < α0 . Then F − α<α0 st(xα , Un α ) cannot be covered by any countable
subfamily of U|F . Because otherwise, {st(xα , Un α ) : α < α0 } ∪ U would contain a
finite subcover of F. Let γ be the largest ordinal α such that st(xα , Un α ) appears in this
finite subcover. Then U would contain a finite subcover of F − αγ st(xα , Un α ),
a contradiction with (1.3). Hence we can use (1.1) in order to obtain xα0 and n α0
satisfying (1.2) and (1.3) for α = α0 . This completes the construction.
Let m ∈ N such that the set A = {α < ω1 : n α = m} is uncountable. Take an α ∈ A.
Then xα ∈ ∪Un α = ∪Um , and Fm = Fm−1 − ∪Vm , where Vm is a finite subfamily
of U which covers Fm−1 − ∪Um , thus F ⊂ Fm ⊂ ∪Um . Since each element of Um
contains at most one point of {xα : α ∈ A}, the set {xα : α ∈ A} is an uncountable
discrete closed subset of F, a contradiction. Therefore, X is compact.
(2) Let X be a countably compact space with a G δ -diagonal. By (1) and Proposition
1.4.9, X has a G ∗δ -diagonal. Assume that {Gn } is a G ∗δ -diagonal sequence for X such
that Gn+1 refines Gn for each n ∈ N. For every x ∈ O ∈ τ ,{O} ∪ {X − st(x, Gn ) :
n ∈ N} is an open cover of X which contains a finite subcover. So there exists m ∈ N
such that st(x, Gm ) ⊂ O. It follows that {Gn } is a development for X . By the Bing
metrization criterion, X is metrizable. 

The condition “quasi-G δ -diagonal” in Theorem 1.4.10(1) can be weaken. For

example, J. Orihuela, R.J. Smith and S. Troyanski [389] claimed that a space X has
(∗) if and only if there is a sequence {Un }n∈N of families of open subsets of X ,
such that, given x, y ∈ X with x = y, there is n ∈ N such that (1) U ∩ {x, y} is a
singleton for some U ∈ Un ; (2) no U ∈ Un contains both x and y. They proved that
a countably compact space with (∗) is compact. A further result of above Theorem
1.4.10(2) see Theorem 2.4.6.
22 1 The Origin of Generalized Metric Spaces

Proposition 1.4.11 Every stratifiable space is a k-semi-stratifiable space, every

k-semi-stratifiable space is a semi-stratifiable space and every semi-stratifiable
space is a submetacompact perfect space with a G δ -diagonal. Therefore, every semi-
stratifiable regular space has a G ∗δ -diagonal.

Proof Obviously, stratifiable ⇒ k-semi-stratifiable ⇒ semi-stratifiable. Suppose X

is a semi-stratifiable space. By Definition 1.4.3, X is a perfect space. By Corollary
1.4.6 and Proposition 1.4.9, X has a G δ -diagonal. For any open cover U of X and
each n ∈ N, let
Un = {F(n, U ) : U ∈ U },

where F : N × τ → τ c is a semi-stratification on X . Then n∈N Un is a σ -cushioned
refinement of U , and hence X is a submetacompact space (see Theorem A.3.3 in
Appendix A). 

Corollary 1.4.12 Every countably compact semi-stratifiable space is a compact

metrizable space.

Definition 1.4.13 ([186]) A space X is called a monotonically normal space if, for
each pair F, K of disjoint closed subsets of X , one can assign an open set D(F, K )
such that
(1) F ⊂ D(F, K ) ⊂ D(F, K ) ⊂ X − K ;
(2) if F ⊂ F  , K  ⊂ K , then D(F, K ) ⊂ D(F  , K  ).
D is called a monotone normality operator for X .

Theorem 1.4.14 ([186]) A space X is a stratifiable space if and only if X is a

monotonically normal and semi-stratifiable space.

Proof Suppose X is a stratifiable space. Then there is a function H : N × τ c → τ

satisfying the conditions of Proposition 1.4.2(3), and we may assume that H (n +
1, F) ⊂ H (n, F). For disjoint closed subsets F, K ⊂ X , define

D(F, K ) = ∪{H (n, F) − H (n, K ) : n ∈ N}.

Then F ⊂ D(F, K ) ∈ τ . If y ∈ K , then there is m ∈ N such that y ∈ / H (m, F),

and hence (X − H (m, F)) ∩ H (m, K ) is an open neighborhood of y which does
not meet D(F, K ). So D(F, K ) ⊂ X − K . It is easy to verify that D also satisfies
Definition 1.4.13(2) by the definition of D.
On the other hand, suppose X is a monotonically normal and semi-stratifiable
space. Assume that D is a monotone normality operator for X and G is a function
satisfying Proposition 1.4.4(3). Define H : N × τ c → τ by H (n, F) = D(F, X −
G(n, F)). Then H satisfies Proposition 1.4.2(3), and hence X is a stratifiable
space. 

In the following, we turn to investigate properties of k-semi-stratifiable spaces.

1.4 Stratifications 23

Proposition 1.4.15 For every regular space X , the following are equivalent:
(1) X is a k-semi-stratifiable space.
(2) There exists a k-semi-stratifiable function on X , i.e. there is a g-function on X ,
such that, for every x ∈ X and sequences {xn }, {yn } in X with xn ∈ g(n, yn ) and
xn → x, yn → x [145, 254].
(3) There is a function G : N × τ c → τ such that

(i) F ∈ τ c ⇒ F = n∈N G(n, F);
(ii) L ⊂ F ⇒ G(n, L) ⊂ G(n, F);
(iii) if K ∩ F = ∅ with K compact and F closed in X , then there is m ∈ N such
that K ∩ G(m, F) = ∅ [316].
Proof (1) ⇒ (2). Let F : N × τ → τ c be a k-semi-stratification on X and we may
assume F(n, U ) ⊂ F(n + 1, U ). Define g : N × X → τ by

g(n, x) = X − F(n, X − {x}).

Then g is a k-semi-stratifiable function on X . Because in fact, for every x ∈ X and

sequences {xn }, {yn }, if xn ∈ g(n, yn ) and xn → x ∈ U ∈ τ , then we may assume
xn ∈ U for each n ∈ N, and hence there is m ∈ N such that {x} ∪ {xn : n ∈ N} ⊂
F(m, U ). Thus xn ∈ F(n, U ) − F(n, X − {yn }) when n  m. So yn ∈ U , and hence
yn → x.
(2) ⇒ (3). Suppose g is a k-semi-stratifiable function on X . Define G : N×τ c → τ
by G(n, F) = g(n, F). Obviously, G satisfies the conditions (i) and (ii) of (3). If
K is compact in X and K ∩ F = ∅, but for some n ∈ N, K ∩ g(n, yn ) = ∅,
then yn ∈ F and xn ∈ K ∩ g(n, yn ). By Corollary 1.4.12, {xn } has a convergent
subsequence {xn k }. Suppose xn k → x ∈ K . Then yn k → x ∈ F, a contradiction. So
there is m ∈ N such that K ∩ G(m, F) = ∅.
(3) ⇒ (1) is obvious. 
Corollary 1.4.16 ([316]) k-semi-stratifiable spaces are additive, hereditary and
countably productive.
Definition 1.4.17 ([319]) A space X is said to be a c-semi-stratifiable space if, for
every compact subset C of X , there is a sequence {G(n, C)} of open sets of X such
that

(1) C = n∈N G(n, C);
(2) C ⊂ F ⇒ G(n, C) ⊂ G(n, F).
Obviously, every semi-stratifiable space is a c-semi-stratifiable space. It is easy to
verify that a space X
is a c-semi-stratifiable space if and only if there is a g-function
on X such that C = n∈N g(n, C) for every compact subset C of X [319].
Proposition 1.4.18 ([58]) A space is c-semi-stratifiable if and only if there is a
g-function on X , such that

(1) {x} = n∈N g(n, x), x ∈ X ;
24 1 The Origin of Generalized Metric Spaces

a sequence {xn } of distinct points of X converges to some x ∈ X , then

(2) if
n∈N g(n, x n ) ⊂ {x}.

Proof Suppose X is c-semi-stratifiable. For any x ∈ X , let g(n, x) = G(n, {x}),

where the function G is defined in Definition 1.4.17. Obviously, (1) of this theorem
is satisfied. We verify assertion (2). Suppose {x n } is a sequence of distinct points of
x ∈ X . Let q ∈ n∈N g(n, xn ). With C = {xn : n ∈
X that converges to a point
N} ∪ {x} we have q ∈ n∈N g(n, xn ) ⊂ n∈N G(n, C) = C, so either q = x as
required, or else q = xm for some m ∈ N. If q = xm define yi = xm+i for each
Then {yi } converges
i ∈ N. L = {yi : i ∈ N} ∪ {x}, we know that
to x. Writing
q ∈ n∈N g(n, xn ) ⊂ n>m g(n, xn ) = i∈N g(i, yi ) ⊂ i∈N G(i, L) = L so that
q = xm , a contradiction. Therefore q = x, as required.
Conversely, suppose X has a g-function with properties (1) and (2). For any
compact subset C of X , let G(n, C) = g(n, C). Clearly, (2) of Definition 1.4.17 is
satisfied. We verify (1). Suppose q ∈ n∈N G(n, C). If q ∈ / C, then choose xn ∈ C
with q ∈ g(n, xn ) for each n ∈ N. Because xn = q for each n, by passing to a
subsequence if necessary, we may assume that the points xn ’s are pairwise distinct.
Note that because C is compact and each point of C is a G δ -set, the subspace C is first
is a subsequence {xn k } that converges to some point z ∈ C.
countable. Hence there
Consequently, q ∈ k∈N g(n k , xn k ) ⊂ {z} ⊂ C. That contradiction establishes part
(1) of Definition 1.4.17 and completes the proof. 

Corollary 1.4.19 ([400]) c-semi-stratifiable spaces are additive, hereditary and

countably productive.

Proposition 1.4.20 ([319]) Every space with a G ∗δ -diagonal is c-semi-stratifiable.

Proof Let {Un } be a G ∗δ -diagonal sequence for X such that Un+1 refines Un for each
n ∈ N. Define G(n, C) = st(C, Un ) for each n ∈ N and each compact subset C
of X . If x ∈ X − C, then there is k ∈ N such that C ⊂ X − st(x, Uk ), because
− st(x, Un )}n∈N covers the compact set C. So x ∈
{X / st(C, Uk ), and hence C =
n∈N st(C, U n ) = n∈N G(n, C). 

Question 1.4.21 ([58]) Is it true that each quasi-developable regular space is c-

semi-stratifiable?

1.5 Networks and (mod k)-Networks

In this section, we mainly discuss the classes of generalized metric spaces defined
by networks.

Definition 1.5.1 A family P in a space X is called a network for X [21, 22] if, for
each x ∈ U ∈ τ , there is P ∈ P such that x ∈ P ⊂ U . A regular space with a
σ -locally finite network is called a σ -space [385].
1.5 Networks and (mod k)-Networks 25

By the regularity, every σ -space has a σ -locally finite closed network.

Proposition 1.5.2 σ -spaces are additive, hereditary and countably productive.

Definition 1.5.3 ([231]) A family P of ordered pairs of subsets of X is called a

pair-network for X if P satisfies the following conditions:
(1) (P1 , P2 ) ∈ P ⇒ P1 ⊂ P2 ;
(2) for each x ∈ U ∈ τ , there is (P1 , P2 ) ∈ P such that x ∈ P1 ⊂ P2 ⊂ U .

Proposition 1.5.4 ([231]) A space X is a semi-stratifiable space if and only if X has

a σ -cushioned network.

Proof Suppose F : N × τ → τ c is a semi-stratification for X . For each n ∈ N, let

Pn = {(F(n, U ), U ) : U ∈ τ }.

Then n∈N P n is a σ -cushioned network for X .
Suppose n∈N Pn is a pair-network for X such that each Pn is a cushioned
family in X . Define g : N × X → τ by

g(n, x) = X − ∪{P1 : (P1 , P2 ) ∈ Pi , x ∈

/ P2 , i  n}.

Then g is a g-function on X . If there exist x ∈ X and a sequence {xn } in X such that

x ∈ g(n, xn ) and xn  x, then we may assume x ∈ / Z , where Z = {xn : n ∈ N}.
Hence there exist m ∈ N and (P1 , P2 ) ∈ Pm such that x ∈ P1 ⊂ P2 ⊂ X − Z . For
every n  m, g(m, xn )∩ P1 = ∅, which contradicts to the fact that x ∈ g(m, xn )∩ P1 .
So g is a semi-stratification for X , and hence X is a semi-stratifiable space. 

Proposition 1.5.5 Every Moore space is a σ -space and every σ -space is a semi-
stratifiable spaces.

Proof Suppose {Un } is a development for a Moore space X . Since X is a submeta-

compact space, Un has a σ -locally finite refinement Fn . Then n∈N Fn is a σ -locally
finite network for X , and hence X is a σ -space.
By Proposition 1.5.4, every σ -space is a semi-stratifiable spaces. 

Definition 1.5.6 ([331]) A regular space with a countable network is called a cosmic
space.

Proposition 1.5.7 ([331]) Cosmic spaces are hereditary and countably productive.

Proposition 1.5.8 A space X is a cosmic space if and only if X is an ℵ1 -compact

σ -space.

Below we introduce several generalizations of networks.

Definition 1.5.9 Suppose P is a family in a space X .

26 1 The Origin of Generalized Metric Spaces

(1) For any A ⊂ X, P is call a network of A in X if, A ⊂ P and for each open
set U in X containing A, there exists P ∈ P such that P ⊂ U . When A is a
singleton {x}, P is called a net at x in X rather than a network of {x} in X .
(2) P is called a (mod k)-network for X [333] if, there is a cover K of X by
compact sets such that for each K ∈ K , {P ∈ P : K ⊂ P} is a network
of K in X . Then P is also called a (mod k)-network w.r.t. K . A space with a
σ -locally finite closed (mod k)-network is called a strong Σ-space [368].
Obviously, every σ -space is a strong Σ-space.
Proposition 1.5.10 ([368]) Strong Σ-spaces are additive, hereditary and countably
productive.
Proof
 We only prove the countably productive case. For each i ∈ N, let Pi =
j∈N P i, j be a closed (mod k)-network w.r.t. Ki for a strong Σ-space X i , where
each Pi, j is locally finite and Pi, j ⊂ Pi, j+1 . Let X = i∈N X i . Define
⎛ ⎞
   
K = Ki , Fn = ⎝ Pi,n ⎠ × X i , n ∈ N.
i∈N in i>n


Then n∈N Fn is a σ -locally finite closed (mod k)-network w.r.t K in X . 
A space X is said to be expandable [255] if, for every locally finite family {Fα }α∈Λ
of closed sets in X , there is a locally finite family {Uα }α∈Λ of open sets in X such
that Fα ⊂ Uα for each α ∈ Λ.
Theorem 1.5.11 ([368]) Paracompact strong Σ-spaces are countably productive.
Proof We still use the notations in the proof of Proposition 1.5.10. Suppose each
X i is also a paracompact space. We prove X is a paracompact space. Since every
paracompact space is an expandable space (see Theorem A.1.11 in Appendix A), for
every i, j ∈ N and P ∈ Pi, j , there is an open set V (P) in X i containing P such
that {V (P) : P ∈ Pi, j } is locally finite in X i . For any open cover U of X , define

F = {F ∈ Fn : n ∈ N and there is U  ∈ U <ω such that F ⊂ ∪U  }.

For each F ∈ F , pick U F ∈ U <ω such that F ⊂ ∪U F . Let

U F = {U j (F) : j  k F }, where k F = |U F |.

Then the family

W (n, j) = {U j (F) ∩ V (F) : F ∈ Fn ∩ F }


is locally finite, and hence n, j∈N W (n, j) is a σ -locally finite open refinement of
U . So X is a paracompact space. 
1.5 Networks and (mod k)-Networks 27

Definition 1.5.12 ([338]) A family P in a space X is called an almost (mod k)-

network for X if, for each x ∈ X , there is K x ∈ K (X ), such that, when K x ⊂ U ∈ τ ,
there is P ∈ P with x ∈ P ⊂ U . If “K x ∈ K (X )” in the definition is replaced with
“a countably compact closed set K x ”, then P is called an almost quasi-(mod k)-
network for X .
Obviously, every (mod k)-network is an almost (mod k)-network for a space X .

Lemma 1.5.13 ([338]) Suppose P is a point-countable family of closed sets in a

space X which is closed under finite intersections. If P is an almost (mod k)-network
for X , then there is a cover K of X with K ⊂ K (X ) such that
(1) P is a (mod k)-network w.r.t. K ;
(2) if x ∈ P ∈ P, then there is K ∈ K such that x ∈ K ⊂ P.

Proof For each x ∈ X , define

C x = (∩(P)x ) ∩ K x ,

where K x satisfies the conditions of Definition 1.5.12. Let

K = {C x : x ∈ X }.

Then the cover K of X satisfies (2). Let C x ⊂ U ∈ τ and (P)x = {Pn }n∈N . We
may assume Pn+1 ⊂ Pn . Let L = K x − U . Then {X − Pn }n∈N is an increasing open
cover of the compact set L in X , so there is m ∈ N such that L ⊂ X − Pm , and hence
K x ⊂ U ∪ (X − Pm ), thus there is n ∈ N such that x ∈ Pn ⊂ U ∪ (X − Pm ). Take
j  max{n, m}. Then C x ⊂ P j ⊂ U . Consequently, P is a (mod k)-network w.r.t.
K. 

Proposition 1.5.14 ([338]) A space X is a strong Σ-space if and only if X has a

σ -locally finite closed almost (mod k)-network.

Definition 1.5.15 A family P in a space X is called a pseudo-network for X if, for

every x, y ∈ X with x = y, there is P ∈ P such that x ∈ P ⊂ X − {y}. A space
with a σ -closure-preserving closed pseudo-network is called a σ  -space [426].
A pseudo-network for X is also called a T1 -separating cover of X [333] or a
ct-network for X [426].

Proposition 1.5.16 ([189]) A space X is a σ  -space if and only if there is a σ  -

function on X , i.e. there is a g-function on X such that

(1) {x} = n∈N g(n, x) for every x ∈ X ;
(2) if y ∈ g(n, x), then g(n, y) ⊂ g(n, x).

Proof Suppose X is a σ  -space and n∈N Pn is a closed pseudo-network for X such
that each Fn is closure-preserving. Define g : N × X → τ by

g(n, x) = X − ∪{F ∈ Fi : x ∈
/ F, i  n}.
28 1 The Origin of Generalized Metric Spaces

Then g is a σ  -function on X .
Conversely, suppose g is a σ  -function on X . Then {X −g(n, A) : n ∈ N, A ⊂ X }
is a σ -closure-preserving closed pseudo-network for X . 
Corollary 1.5.17 σ  -spaces are additive, hereditary and countably productive.
Corollary 1.5.18 ([319]) Every σ  -space is c-semi-stratifiable.
Proof Suppose X is a σ  -space. Let g be a σ  -function on X . Take the sequence
{g(n, C)} for each compact
subset C of X . We prove that C = n∈N g(n, C).
Otherwise, there is z ∈ n∈N g(n, C) − C, and hence there is a sequence {xn } in C
such that z ∈ g(n, xn ), n ∈ N. Let x be an accumulation point of {xn } in C because
of the compactness of C. Then x = z. Since g satisfies Proposition 1.5.16(1), there
/ g(m, x), so there is n > m such that xn ∈ g(m, x). Further by
is m ∈ N such that z ∈
the assumption that g satisfies Proposition 1.5.16(2), g(m, xn ) ⊂ g(m, x), and hence
z ∈ g(n, xn ) ⊂ g(m, xn ) ⊂ g(m, x), a contradiction. So X is c-semi-stratifiable. 
Question 1.5.19 ([58, 319]) Is there a regular c-semi-stratifiable space that is not a
σ  -space?
Proposition 1.5.20 ([43]) Every submetacompact space with a G δ -diagonal is a
σ  -space.
Proof Suppose {Un } is a G δ -diagonal sequence in a submetacompact space X . By
the submetacompactness of X , for every n ∈ N, there is a σ -closure-preserving
closed cover Pn of X satisfying that for every x ∈ X , there exist P ∈ Pn and
<ω
 ∈ (Un )x such that x ∈ P ⊂ ∪U (see Theorem A.4.8 in Appendix A). So
U
n∈N Pn is a σ -closure-preserving closed pseudo-network for X , and hence X is a
σ  -space. 
Corollary 1.5.21 ([422]) Every semi-stratifiable space is a σ  -space.

1.6 k-Networks and Weak Bases

In this section, we investigate some properties of families between bases and net-
works.
Definition 1.6.1 ([331]) A family P in a space X is called a strict k-network for X if,
for every K ⊂ U ∈ τ with K compact in X , there is P ∈ P such that K ⊂ P ⊂ U .
A regular space with a countable strict k-network is called an ℵ0 -space.
Strict k-networks were called pseudo-bases in [331]. We use the term “strict
k-network” rather than “pseudo-base” in Definition 1.6.1, because the following
reasons: (1) the term “pseudo-base” already has been used with a different meaning by
Arhangel’skiı̌ and Projzvolov [41] (see Definition 3.1.1); (2) the concept of pseudo-
bases in the sense of Michael [331] is stronger than that of k-networks (see Definition
1.6.5); and (3) the term “strong k-network” already also has been used with a different
meaning by H. Chen [102] and C. Liu, M. Dai [308].
1.6 k-Networks and Weak Bases 29

Proposition 1.6.2 ([331]) Every separable metric space is an ℵ0 -space and every
ℵ0 -space is a cosmic spaces.

Proposition 1.6.3 ([331]) ℵ0 -spaces are hereditary and countably productive.

A space X is of the point G δ -property if every singleton of X is a G δ -set in X .

Theorem 1.6.4 ([256, 473]) A regular space with a point-countable strict k-network
is an ℵ0 -space.

Proof Suppose P is a point-countable strict k-network for a regular space X .

(4.1) X is of the point G δ -property.
For each x ∈ X , pick y ∈ X − {x}. Let

F = {X − P : y ∈ P ⊂ P ⊂ X − {x}, P ∈ P}.

Then F is a countable family of open sets in X and x ∈ ∩F . If z ∈ X − {x}, take

V ∈ τ such that {y, z} ⊂ V ⊂ V ⊂ X − {x}. Then there exists P ∈ P such that
{y, z} ⊂ P ⊂ V . So {y, z} ⊂ P ⊂ P ⊂ X − {x}, and hence z ∈ / ∩F . Therefore,
{x} = ∩F .
(4.2) X is a separable space.

Fix x ∈ X and choose {Vn : n ∈ N} ⊂ τ such that {x} = n∈N Vn . Denote
(P)x = {Pm }m∈N . For every n, m ∈ N, if Pm − Vn = ∅, take an,m ∈ Pm − Vn ; if
Pm − Vn = ∅, let an,m = x. Let D = {an,m : n, m ∈ N}. We prove X = D. Suppose
G ∈ τ − {∅}. If x ∈ G, then G ∩ D = ∅; if x ∈ / G, take y ∈ G, then there is
n ∈ N such that y ∈ X − Vn . Since {x, y} ⊂ G ∪ Vn , there exists m ∈ N such that
{x, y} ⊂ Pm ⊂ G ∪ Vn , and hence y ∈ Pm − Vn ⊂ G. So an,m ∈ G, and it follows
that G ∩ D = ∅.
(4.3) X has a countable strict k-network.
Suppose D is a countable dense subset of X . Let

H = {P ∈ P : P ∩ D = ∅}.

Then H is countable. If K ⊂ G ∈ τ − {∅} with K compact in X , take x ∈ D ∩ G,

then K ∪ {x} ⊂ G, and hence there is P ∈ H such that K ⊂ P ⊂ G. So H is a
countable strict k-network for X . Thus X is an ℵ0 -space. 

In general, a base may not be a strict k-network.

Definition 1.6.5 Suppose P is a family of subset sets in a space X .

(1) P is called a k-network for X [387] if, for every K ⊂ U ∈ τ with K compact
in X , there exists F ∈ P <ω such that K ⊂ ∪F ⊂ U . A regular space with a
σ -locally finite k-network is called an ℵ-space [387].
30 1 The Origin of Generalized Metric Spaces

(2) P is called a cs-network for X [169] if, for each x ∈ X , any sequence {xn } in
X converging to x and x ∈ U ∈ τ , there exist m ∈ N and P ∈ P such that
{x}∪{xn : n  m} ⊂ P ⊂ U . A regular space with a σ -locally finite cs-network
is called a cs-σ -space [170].
(3) P is called a cs ∗ -network for X [146] if, for each x ∈ X , any sequence {xn }
in X converging to x and x ∈ U ∈ τ , there are subsequence {xni } of {xn } and
P ∈ P such that {x} ∪ {xni : i ∈ N} ⊂ P ⊂ U .
Obviously, both closed k-networks and cs-networks are cs ∗ -networks.
Proposition 1.6.6 ([170, 388]) ℵ-spaces and cs-σ -spaces are additive, hereditary
and countably productive.
Proposition 1.6.7 ([455]) Suppose every compact subspace of a space X is sequen-
tially compact. If P is a point-countable cs ∗ -network for X , then P is also a
k-network for X .
Proof For every K ⊂ U ∈ τ with K compact in X , denote

H = {P ∈ P : P ⊂ U };
(H )x = {Pn (x) : n ∈ N}, x ∈ K .

If there is no F ∈ H <ω such that K ⊂ ∪F , then we can choose a sequence

{xk } ⊂ K such that xk ∈ / Pn (x j ) whenever n, j < k. Suppose {xki } is a subsequence
of {xk } converging to a point x ∈ K . Then there is P ∈ H such that P contains
infinitely many xki , a contradiction. So there exists F ∈ H <ω such that K ⊂ ∪F ,
and hence P is a k-network for X . 
In the proof of Proposition 1.6.7, we only use the following condition which is
weaker than being a cs ∗ -network: if a sequence {xn } in X converges x and x ∈ U ∈ τ ,
then there is P ∈ P such that P contains infinitely many terms of {xn } and P ⊂ U .
Corollary 1.6.8 ([170]) Every cs-σ -space is an ℵ-space.
Proof Let P be a σ -locally finite cs-network for X . For each K ∈ K (X ), P|K is a
countable network for K , so K is metrizable. By Proposition 1.6.7, P is a k-network
for X , and hence X is an ℵ-space. 
Corollary 1.6.9 ([169, 170]) For every regular space X , the following are equiva-
lent:
(1) X is an ℵ0 -space.
(2) X has a countable cs ∗ -network.
(3) X is an ℵ1 -compact ℵ-space.
(4) X is an ℵ1 -compact cs-σ -space.
Proof Obviously, (1) ⇒ (4) ⇒ (3) ⇒ (2). We only need to prove (2) ⇒ (1). Let
P be a countable cs ∗ -network for X . By the proof of Proposition 1.6.7, P F is a
countable strict k-network for X , so X is an ℵ0 -space. 
1.6 k-Networks and Weak Bases 31

Similar to Definition 1.5.3, we can define pair-strict k-networks, pair-k-networks,

pair-cs-networks and pair-cs ∗ -networks.

Proposition 1.6.10 ([129]) For every regular space X , the following are equivalent:
(1) X is a k-semi-stratifiable space.
(2) X has a σ -cushioned strict k-network.
(3) X has a σ -cushioned cs ∗ -network.

Proof By the proof of Proposition 1.5.4 we obtain (1) ⇒ (2). (2) ⇒ (3) is obvious.
Below we prove (3) ⇒ (1).
Let n∈N Pn be a pair-cs ∗ -network for X , where each Pn is cushioned. Define
g : N × X → τ by

g(n, x) = X − ∪{P1 : (P1 , P2 ) ∈ Pi , x ∈

/ P2 , i  n}.

Then g is a g-function on X . For every x ∈ X and sequences {xn }, {yn } in X ,

if xn ∈ g(n, yn ) and xn → x, but yn → x does not hold, then without loss of
generality, we may assume that there is an open neighborhood U of x such that
yn ∈/ U for every n ∈ N. Since xn → x ∈ U ∈ τ , there exist a subsequence {xni }
of {xn }, k ∈ N and (P1 , P2 ) ∈ Pk such that {x} ∪ {xni : i ∈ N} ⊂ P1 ⊂ P2 ⊂ U .
By xni ∈ g(n i , yni ), yni ∈ P2 ⊂ U , a contradiction. So yn → x, and hence g is a
k-semi-stratifiable function on X . Therefore, X is a k-semi-stratifiable space. 

Corollary 1.6.11 ([316]) Every ℵ-space is a k-semi-stratifiable space.

In the rest part of this section, we investigate some properties of families between
bases and cs-network.

 1.6.12 A family P in a space X is called a weak base for X [31], if

Definition
P = x∈X Px satisfies the following conditions:
(1) x ∈ ∩Px ;
(2) if U, V ∈ Px , then there exists W ∈ Px such that W ⊂ U ∩ V ;
(3) for each G ⊂ X , G ∈ τ if and only if for each x ∈ G, there is P ∈ Px such
that P ⊂ G.
Px is called a weak base of x (in X ). A subset of X is called a weak neighborhood of
x if it contains an element of Px . If Px is countable for each x ∈ X , then X is called
a g-first countable space (or g f -countable space, weakly first countable space) [31].
If P is countable, then X is called a g-second countable space. A regular space with
a σ -locally finite weak base is called a g-metrizable space [425].
If P is a weak base for X and A is an open or closed subspace of X , then P|A is a
weak base for A. So, both g-first countability and g-metrizability are open hereditary
and closed hereditary.
Weak neighborhoods are generalizations of spherical neighborhoods in sym-
metrizable spaces. Obviously, every symmetrizable space is g-first countable. It is
not strange to characterize symmetrizable spaces in terms of weak bases.
32 1 The Origin of Generalized Metric Spaces

Definition 1.6.13 ([472]) A sequence {Un } of covers of a space X is called a weak

development for X , if {st(x, Un )}n∈N is a weak base of x in X for each x ∈ X .

Proposition 1.6.14 ([472]) A space X is a symmetrizable space if and only if X has

a weak development.

Proof Suppose (X, d) is a symmetrizable space. For every n ∈ N, let

Un = {A ⊂ X : diamA < 1/n}.

Then for each x ∈ X , st(x, Un ) = B(x, 1/n). So {Un } is a weak development for
X.
Conversely, let {Un } be a weak development for X . For every x, y ∈ X with
/ st(y, Un ). Define
x = y, let n(x, y) be the least positive integer n such that x ∈
d : X × X → R+ as follows:

0, x = y,
d(x, y) =
2−n(x,y) , x = y.

Then d is a symmetric on X and for every x ∈ X and n ∈ N, st(x, Un ) = B(x, 1/2n ).

It follows that X is a symmetrizable space. 

Definition 1.6.15 ([133]) A space X is called a sequential space if, a subset F of

X is closed in X if and only if for every Z ∈ S (X ), Z ∩ F is closed in X .
Obviously, sequential spaces are additive, open hereditary and closed hereditary.

Proposition 1.6.16 ([425]) Every g-first countable space is a sequential space.

Proof Suppose X is a g-first countable space. Let F ⊂ X and suppose the intersec-
tion of F and any sequence with its limit point is closed in X . For each x ∈
/ F, take
a decreasing weak base {Pn }n∈N of x. Then there is m ∈ N such that Pm ∩ F = ∅.
Because otherwise, there exists a sequence {xn } in F such that xn ∈ Pn , it follows
that xn → x. As a consequence, F ∩ ({x} ∪ {xn : n ∈ N}) is not closed in X , a
contradiction. Thus F is closed in X , and hence X is a sequential space. 

Proposition 1.6.17 Let X be a Fréchet–Urysohn space. If W is a weak neighborhood

of x in X , then x ∈ W ◦ .

Proof Suppose P = x∈X Px is a weak base for X as defined in Definition 1.6.12
and W ⊃ P ∈ Px . If x ∈ X − W ◦ , then x ∈ X − P ◦ = X − P, thus there is
a sequence {xn } in X − P such that xn → x. Denote Z = {xn : n ∈ N}. Then
Z is not closed in X . On the other hand, for each y ∈ X − Z , if y = x, then
P ⊂ X − Z ; if y = x, then y ∈ X − Z . As a consequence, there is G ∈ P y such
that G ⊂ X − Z ⊂ X − Z . So Z is closed in X , a contradiction. Hence, x ∈ W ◦ . 

Corollary 1.6.18 ([425]) A space X is a first countable space if and only if X is a

g-first countable Fréchet–Urysohn space.
1.6 k-Networks and Weak Bases 33

Let X be a space. Suppose F ⊂ W ⊂ X . Then W is called a sequential neighbor-

hood of F in the space X if, for each sequence {xn } in X converging to a point in F,
{xn } is eventually in W , i.e. there is m ∈ N such that {xn : n  m} ⊂ W . When F is
a singleton {z}, instead of saying that W is a sequential neighborhood of {z}, we say
that W is a sequential neighborhood of z. A subset Z of X is called a sequentially
open set in X if for each z ∈ Z , Z is a sequential neighborhood of z; Z is called a
sequentially closed set in X if X − Z is a sequentially open set. Obviously, X is a
sequential space if and only if every sequentially open set in X is open, if and only
if every sequentially closed set in X is closed [133].
Every weak base Px of a point x in the Definition 1.6.12 can be regarded as
a net at x consisting of weak neighborhoods of x. Similarly, we can define a net
of x consisting of sequential neighborhoods of x by the definition of a sequential
neighborhood [273].

Corollary 1.6.19 If Bx is a weak neighborhood of x in a space X , then Bx is a

sequential neighborhood of x.

Proof Let {xn } be a sequence in X such that xn → x. Denote Z = {x}∪{xn : n ∈ N}.

Then Z is a Fréchet–Urysohn closed subspace of X , and hence x ∈ int Z (Bx ∩ Z ).
So {xn } is eventually in Bx . 

Corollary 1.6.20 ([425]) If P is a weak base for a space X , then P is a cs-network

for X .

Proposition 1.6.21 ([291]) Let X be a g-first countable space. If P is a point-

countable cs-network for X which is closed under finite intersections, then there
exists a subfamily of P which is a weak base for X .

Proof Let {B(n, x)}n∈N be a decreasing weak base of x for each x ∈ X . Define

Px = {P ∈ P : there is n ∈ N such that B(n, x) ⊂ P}.

We prove Px is a weak base of x. Obviously, x ∈ ∩Px and Px is closed under finite

intersections. Let G be a subset of X . If for each x ∈ G, there is P ∈ Px such that
P ⊂ G, then there is n ∈ N such that B(n, x) ⊂ G, and hence G ∈ τ . On the other
hand, if x ∈ G ∈ τ , then there exists P ∈ Px such that P ⊂ G. Because otherwise,
denote {P ∈ P : x ∈ P ⊂ G} = {P(m, x)}m∈N . Then B(n, x) ⊂ P(m, x) for any
n, m ∈ N. Fix xn,m ∈ B(n, x) − P(m, x). For every n  m, let xn,m = yk , where
k = m + n(n−1)
2
. Then yk → x. It follows that there exist m, i ∈ N such that

{x} ∪ {yk : k  i} ⊂ P(m, x) ⊂ G.

Take j  i such that y j =

 xn,m for some n  m. Then xn,m ∈ P(m, x), a contradic-
tion. As a consequence, x∈X Px is a weak base for X . 

By Proposition 1.6.21 and Corollary 1.6.20, we have the following theorem.

34 1 The Origin of Generalized Metric Spaces

Theorem 1.6.22 ([425]) A space X is a g-metrizable space if and only if X is a

g-first countable cs-σ -space.
Therefore, a space X is a g-second countable regular space if and only if X is a
g-first countable ℵ0 -space [425].

Example 1.6.23 ([305]) A weak base for a compact space which is not a k-network.
Let X = [0, ω1 ] and give X the ordered topology. Then X is a compact space.
Let L be the set of all accumulation points in X . For each α < ω 1 , let Pα be a local
base of α. Denote Pω1 = {(β, ω1 ] ∩ L : β < ω1 }. Let P = αω1 Pα . Since X
is compact and X is not covered by any finitely many elements of P, P is not a
k-network for X . We prove that P is a weak base for X . Let U be a subset of X
satisfying that for each α ∈ U , there is P ∈ Pα such that P ⊂ U . If α ∈ U − {ω1 },
then U is a neighborhood of α; if ω1 ∈ U , there is β < ω1 such that (β, ω1 ]∩ L ⊂ U .
It follows that there is α ∈ (β, ω1 ) such that (α, ω1 ] ⊂ U . Because otherwise, there
is a strictly increasing sequence {αn } in (β, ω1 ) − U . Let γ = sup{αn : n ∈ N}. Then
γ ∈ U , and hence αn ∈ U for some αn , a contradiction. As a consequence, U is an
open set in X . So P is a weak base for X .
Sakai [419] proved that for a sequential space X , every weak base for X is a
k-network for X if and only if every compact subset of X is Fréchet–Urysohn.

1.7 Generalized Countably Compact Spaces

Many classes of generalized metric spaces defined or characterized in the previous

sections has the following characteristics: they can described by some convergent
sequences satisfying certain conditions either in terms of g-functions, such as semi-
metrizable spaces, developable spaces and k-semi-stratifiable spaces etc.; or in terms
of sequences of open covers of spaces, such as metrizable spaces and developable
spaces etc. If we replace the convergence of sequences in their descriptions with the
existence of accumulation points of the sequences, then we can define a group of
new generalized metric spaces and they can be collectively referred as generalized
countably compact spaces [195] or spaces defined by sequences of open covers which
guarantee that certain sequences has cluster points [190].

Definition 1.7.1 For every space X and any g-function on X , consider the following
conditions:
(1) Suppose x ∈ X and {xn } is a sequence in X . If xn ∈ g(n, x) for each n ∈ N,
then {xn } has an accumulation point in X .
(2) Suppose x ∈ X and {xn } is a sequence in X . If x ∈ g(n, xn ) for each n ∈ N,
then {xn } has an accumulation point in X .
(3) Suppose x ∈ X and {xn }, {yn } are sequences in X . If {x, xn } ⊂ g(n, yn ) for each
n ∈ N, then {xn } has an accumulation point in X .
1.7 Generalized Countably Compact Spaces 35

If a space X has a g-function satisfying condition (1) (resp. (2), (3)), then X is
called a q-space [330] (resp. β-space [189], wΔ-space [66]) and the g-function is
called a q-function (resp. β-function, wΔ-function).
By the above definitions, every first countable space is a q-space, every semi-
stratifiable space is a β-space and every developable space is a wΔ-space.

Proposition 1.7.2 ([464]) For every space X , the following are equivalent:
(1) X is a β-space.
(2) There is a function F : N × τ → τ c satisfying the following conditions:
(i) U ∈ τ ⇒ F(n, U ) ⊂ U ;
(ii) V ⊂ U ⇒ F(n, V ) ⊂ F(n, U ); 
(iii) if {Un }n∈N is an increasing cover of open subsets of X , then n∈N F(n, Un ) =
X.
(3) There is a function G : N × τ c → τ satisfying the following conditions:
(i) F ∈ τ c ⇒ F ⊂ G(n, F);
(ii) L ⊂ F ⇒ G(n, L) ⊂ G(n, F);
(iii) if {Fn }n∈N is a decreasing
sequence of closed subsets of X with the empty
intersection, then n∈N G(n, Fn ) = ∅.

Proof We only need to prove (1) ⇔ (3).

(1) ⇒ (3). Let g be a β-function on X . Define G : N × τ c → τ by G(n, F) =
g(n, F). Then G satisfies (3). Because in fact, if {Fn }n∈N is a decreasing
sequence of
closed subsets of X with the empty intersection and there is x ∈ n∈N G(n, Fn ), then
there is xn ∈ F n such that x ∈ g(n, x n ), and it follows that {x n } has an accumulation
point in X , so n∈N Fn = ∅.
(3) ⇒ (1). We may assume that G is decreasing on n ∈ N. Define g : N × X → τ
by g(n, x) = G(n, {x}). Then g is a β-function on X . Because in fact, suppose x ∈ X
and {xn } is a sequence in X such that x ∈ g(n, xn ) and {xn } has no accumulation
point in X . Let
Fn = {xi : i  n}, n ∈ N.

Then {Fn } is a decreasing

sequence of closed subsets
of X with the empty
intersection.
It follows that n∈N G(n, Fn ) = ∅. But x ∈ n∈N g(n, Fn ) ⊂ n∈N G(n, Fn ), a
contradiction. 

As a consequence, every β-space is a countably metacompact space (see Propo-

sition A.2.8 in Appendix A).

Proposition 1.7.3 ([189]) A space X is a wΔ-space if and only if X there is a

sequence {Un } of open covers of X such that if x ∈ X and {xn } is a sequence in X
with xn ∈ st(x, Un ), then {xn } has an accumulation point in X .

The proof of Proposition 1.7.3 is similar to that of Theorem 1.2.13. The sequence
of open covers of X in the above proposition is called a wΔ-sequence. Generally, if a
36 1 The Origin of Generalized Metric Spaces

sequence {Fn } of open covers of X satisfying that for each x ∈ X and any sequence
{xn } in X , {xn } has an accumulation point in X whenever xn ∈ st(x, Fn ), then we
say that {Fn } satisfies wΔ-condition.

Definition 1.7.4 Let {Un } be a sequence of open covers of a space X . Consider the
following conditions:
(1) For every n ∈ N, Un+1 is a star-refinement of Un and {Un } is a wΔ-sequence in
X.
(2) For each x ∈ X and any sequence {xn } in X with xn ∈ st2 (x, Un ), {xn } has an
accumulation point in X .
A space satisfying condition (1) (resp. (2)) is called an M-space [360] (resp. a
wM-space [202]), and the sequence {Un } is called an M-sequence (resp. a wM-
sequence).
Every countably compact space or metrizable space is an M-space, every M-
space is a wM-space, every wM-space is a wΔ-space and every wΔ-space is both a
β-space and a q-space.

Lemma 1.7.5 ([422]) Every wM-space is an expandable space.

Proof Let {Un } be a wM-sequence in X . We may assume that Un+1 refines Un .

(5.1) If x ∈ X and {xn } is a sequence in X with xn ∈ st3 (x, Un ), then {xn } has an
accumulation point in X .
If the sequence {xn } has no accumulation point in X , then {{xn } : n ∈ N}
is a locally finite family in X , and hence {st(xn , Un ) : n ∈ N} is locally finite.
Because otherwise, there is z ∈ X such that for each n ∈ N, there is i(n)  n with
st(z, Un ) ∩ st(xi(n) , Ui(n) ) = ∅, and hence xi(n) ∈ st(st(z, Un ), Ui(n) ) ⊂ st2 (z, Un ).
As a consequence, {xi(n) } has an accumulation point in X , a contradiction. Since
xn ∈ st3 (x, Un ), st2 (x, Un ) ∩ st(xn , Un ) = ∅. Take yn ∈ st2 (x, Un ) ∩ st(xn , Un ).
Then the sequence {yn } has no accumulation point in X , again a contradiction.

(5.2) X is a countably paracompact space.

Let {G n }n∈N be an increasing cover of open subsets of X . Define

Fn = X − st2 (X − G n , Un ), n ∈ N.
 
Then X = n∈N Fn . Because in fact, if z ∈ X − n∈N Fn = n∈N st2 (X − G n , Un ),
then st2 (z, Un ) ∩ (X − G n ) = ∅, and hence n∈N (X − G n ) = ∅, a contradiction.
Now let
Hn = X − st(X − G n , Un ), n ∈ N.

Then Fn ⊂ Hn ⊂ H n ⊂ G n , so X = n∈N Hn , and it follows that X is a countably
paracompact space (see Proposition A.1.13 in Appendix A).
1.7 Generalized Countably Compact Spaces 37

(5.3) X is an expandable space.

Let {Fλ }λ∈Λ be a locally finite family of closed sets in X . Then for every x ∈ X ,
there is n ∈ N such that {λ ∈ Λ : st2 (x, Un ) ∩ Fλ = ∅} is finite. For each n ∈ N, let

An = {x ∈ X : {λ ∈ Λ : st2 (x, Un ) ∩ Fλ = ∅} is finite},

Bn = A◦n .

Then Bn ⊂ Bn+1 . We prove X = n∈N Bn . In fact, for any z ∈ X , by (5.1), there is
n ∈ N such that {λ ∈ Λ : st3 (z, Un ) ∩ Fλ = ∅} is finite, and hence st(z, Un ) ⊂ An ,
so z ∈ Bn . By (5.2), there is a locally finite open cover {G n }n∈N of X such that
G n ⊂ Bn . Let

G λ,n = st(Fλ , Un ) ∩ G n , Hλ = ∪{G λ,n : n ∈ N}.

Then Fλ ⊂ Hλ ∈ τ . We prove that {Hλ }λ∈Λ is locally finite. For every x ∈ X , denote

U = X − ∪{G n : x ∈
/ G n , n ∈ N},
{G n : n ∈ N, x ∈ G n } = {G n(i) : i  k},
m = max{n(i) : i  k},
V = st(x, Um ) ∩ U.

Then x ∈ V ∈ τ . For each i  k, since x ∈ Bn(i) ,

{λ ∈ Λ : st(x, Um ) ∩ G λ,n(i) = ∅} ⊂ {λ ∈ Λ : st2 (x, Un(i) ) ∩ Fλ = ∅}

is a finite set, and hence V only meets finitely many elements of {Hλ }λ∈Λ . So X is
an expandable space. 

Lemma 1.7.6 (The convergence

Let {G n } be a decreasing sequence of sub-
lemma)
sets of a space X such that n∈N G n = n∈N G n . Then the following are equivalent:
(1) If {xn } is a sequence in X with xn ∈ G n , then {xn } has an accumulation point in
X.

(2) n∈N G n is a countably compact set and {G n }n∈N is a network of n∈N G n in
X.

Proof (1) ⇒ (2). Obviously, n∈N G n is a countably compact closed set in X . Let

n∈N G n ⊂ V ∈ τ . If there is a sequence
n
{x } in X such that xn ∈ G n − V , let x be an
accumulation point of {xn }, then x ∈ n∈N G n − V = ∅, a contradiction. It follows
that there is m ∈ N such that G m ⊂ V , i.e. {G n }n∈N is a network for n∈N G n in X .
(2) ⇒ (1). Suppose xn ∈ G n for each n ∈ N. If {x n } has no accumulation point,
then there is m ∈ N such that n∈N G n ⊂ X − {xn : n  m}, and hence there exists
k  m such that xk ∈ G k ⊂ X − {xn : n  m}, a contradiction. 
38 1 The Origin of Generalized Metric Spaces

Theorem 1.7.7 (1) If a regular space X is a q-space of the point G δ -property, then
X is a first countable space [336].
(2) Suppose X has a G ∗δ -diagonal or X is a σ  -space. If X is a β-space, then X is a
semi-stratifiable space [189]; if X is a wΔ-space, then X is a developable space
[189]; if X is a wM-space, then X is a metrizable space [422].
Proof (1) Suppose g : N × X → τ is a q-function on X such that g(n + 1, x) ⊂
g(n, x) and {x} = n∈N g(n, x). By Lemma 1.7.6, {g(n, x)}n∈N is a neighborhood
base of x, and hence X is a first countable space.
(2) If X has a G ∗δ -diagonal, then take a G ∗δ -diagonal sequence {Un } of X such that
Un+1 refines Un .
When X is a β-space, let g be a β-function on X . Define h : N × X → τ by
h(n, x) = g(n, x)∩st(x, Un ). Then h is a semi-stratifiable function on X . Because in
fact, if x ∈ X and {xn } is a sequence in X with x ∈ h(n, xn ), then each subsequence
of {xn } has an accumulation point. Let y be an accumulation point of {xn } and
y = x. Then there is n ∈ N such that y ∈ / st(x, Un ), and hence there exists m  n
such that xm ∈ / st(x, Um ). It follows that x ∈ / st(xm , Um ), a contradiction. So x
is an accumulation point of any subsequence of {xn }, and hence xn → x. As a
consequence, X is a semi-stratifiable space. When X is a wΔ-space, let {Vn } be a
wΔ-sequence in X . Then {Un ∧ Vn } is a development of X . If X is a wM-space, then
X is a developable expandable space, and hence X is a developable paracompact
space (see Theorem A.4.10 in Appendix A), so X is a metrizable space.
Now assume that X is a σ  -space. When X is a β-space, denote a σ  -function
and a β-function on X by g and h respectively. Define l : N × X → τ by l(n, x) =
g(n, x) ∩ h(n, x). We prove l is a semi-stratifiable function on X . If x ∈ X and {xn }
is a sequence in X with x ∈ l(n, xn ), then {xn } has an accumulation point y, because
h is a β-function. We may assume that xn ∈ g(n, y), so x ∈ g(n, xn ) ⊂ g(n, y),
and hence x = y, which proves that any subsequence of {xn } has an accumulation
point and x must be the only one accumulation point, and hence xn → x. So X is a
semi-stratifiable space. Similarly, we can prove that when X is a wΔ-space, X is a
developable space and when X is a wM-space, X is a developable wM-space, and
hence X is a metrizable space. 

Question 1.7.8 ([77]) Is every wΔ-space with a point-countable base developable?

Question 1.7.9 ([125]) Is every quasi-developable β-space developable?

Question 1.7.10 ([202]) Is every wM-space with a G δ -diagonal metrizable?

It is proved that every wM-space with a G δ -diagonal is a first countable and

regular space [246].

Definition 1.7.11 ([124]) A family P in a space X is said to be interior-preserving

if, for any F ⊂ P, (∩F )◦ = ∩F ◦ .
Obviously, a family P in X is interior-preserving if and only if {X − P} P∈P
is closure-preserving. It should be noted that for any g-function on X , if g(n, x) ⊂
g(n, y) whenever x ∈ g(n, y), then {g(n, A) : A ⊂ X } is interior-preserving. On
1.7 Generalized Countably Compact Spaces 39

the other hand, if an open family Bn in X is interior-preserving, define g : N × X →

τ by g(n, x) = ∩(Bn )x , then g satisfies the condition that if x ∈ g(n, y), then
g(n, x) ⊂ g(n, y).

Proposition 1.7.12 Every space with a σ -interior-preserving base is a σ  -space.


Proof Let n∈N Bn be a base for a space X such that each Bn is interior-preserving
and X ∈ Bn ⊂ Bn+1 . Define g : N × X → τ by g(n, x) = ∩(Bn )x . Then g is a
σ  -function on X , and hence X is a σ  -space. 

Corollary 1.7.13 ([60]) Every β-space with a σ -point-finite base is developable.

Proof Let X be a β-space with a σ -point-finite base. By Proposition 1.7.12 and

Theorem 1.7.7, X is semi-stratifiable. By Corollary 1.4.5 and Theorem 1.2.14, X is
developable. 

1.8 Examples

In the first part of this section, we list several spaces which were frequently used
in the literatures of general topology as counterexamples. We not only use these
spaces to illustrate that there are no implication relationships among some classes of
spaces discussed in the previous sections, but also use them as counterexamples in
the following two chapters. The citation of each example only refers to the literature
constructing the example originally.

Example 1.8.1 ([177]) The V -space.

Let X = {(x, y) ∈ R2 : y  0} and define a V -topology for X as follows: for
each y > 0, (x, y) is an isolated point in X ; for y = 0, a basic neighborhood of
(x, y) has the form

V (x, n) = {(t, s) ∈ X : t = x ± s, 0  s  1/n}, n ∈ N.

X is called the V -space or the Heath V-space.

It is easy to verify that X is a completely regular metacompact space. Since R×{0}
is second category for the Euclidean topology, we can prove that X is not a normal
space by using the standard category method. Let

Un = {V (x, n) : x ∈ R} ∪ {(x, y) ∈ X : y > 1/n}, n ∈ N.

Then {Un } is a development for X , so X is a Moore space.

Example 1.8.2 ([181]) The bowtie space.

Let X = R2 . For every x = (t, s) ∈ X and n ∈ N, denote a bowtie set Bt (x, 1/n)
of the form
40 1 The Origin of Generalized Metric Spaces

{x} ∪ {(t  , s  ) ∈ X : 0 < |t − t  | < 1/n, |s − s  |/|t − t  | < 1/n}.

The topology generated on X by taking {Bt (x, 1/n)}n∈N as a neighborhood base of

x is called the bowtie topology on X and X is called the bowtie space. Obviously, X
is a regular space.
(2.1) X is a semi-metrizable space.
Define g : N × X → τ by g(n, x) = Bt (x, 1/n). Then g is a semi-metrizable
function on X , and hence X is a semi-metrizable space. We can define a concrete
semi-metric d on X as follows:
⎧
⎨ 0, t = t  and s = s  ,
d((t, s), (t  , s  )) = 1, t = t  and s = s  ,
⎩
|t − t  | + |(s − s  )/(t − t  )|, otherwise.

(2.2) X is not a σ -space [162].


Otherwise, there is a closed network n∈N Fn for X such that every Fn is discrete
(see Theorem 3.3.1). For every x ∈ X , there exist n(x) ∈ N and Fx ∈ Fn(x) such that
x ∈ Fx ⊂ Bt (x, 1). Since R2 is second category for the Euclidean topology, there
exist m ∈ N and Y ⊂ X , such that, n(x) = m whenever x ∈ Y and Y is dense in a
Euclidean open subset of R2 . Pick p ∈ U and let W = X − ∪{F ∈ Fm : p ∈ / F}.
Then p ∈ W ∈ τ (X ). Take y, z ∈ Y ∩W ∩U such that y ∈ / Bt (z, 1) and z ∈
/ Bt (y, 1).
Then Fy = Fz ∈ Fm , so Fy ∩ Fz = ∅, and hence there is F ∈ Fm such that p ∈ /F
and F ∩ W = ∅, which contradicts the definition of W . As a consequence, X is not
a σ -space.
By Theorem 1.7.7, X is not a wΔ-space.

Example 1.8.3 ([323]) The butterfly space.

Let X = R2 . Define the butterfly topology on X as follows: let x = (t, s) ∈ X ,
if s = 0, then the neighborhoods of x are the same as in the Euclidean topology; if
s = 0, then the basic neighborhoods of x have the bowtie form Bt (x, 1/n), n ∈ N.
A space with the butterfly topology is called the butterfly space. Obviously, X is a
regular space.
(3.1) X is a cosmic space.
Since the subspaces R × {0} and X − (R × {0}) of X have the Euclidean topology,
we may assume P and F are their countable bases respectively. Then P ∪ F is a
countable network for X , and hence X is a cosmic space.
(3.2) X is a semi-metrizable space.
Obviously, X is a first countable space. By Corollary 1.4.5, X is a semi-metrizable
space.
(3.3) The compact subset I × {0} of X has no countable neighborhood base in X .
1.8 Examples 41

Otherwise, let {Un }n∈N be a countable neighborhood base of I × {0} in X . For

each x ∈ I × {0}, there is n ∈ N such that Un ⊂ Bt (x, 2), and hence there exist
m ∈ N and an infinite subset Z of I × {0} such that Um ⊂ x∈Z Bt (x, 2). Let a be an
accumulation point of Z in I × {0}. Then there is k ∈ N such that Bt (a, 1/k) ⊂ Um .
Take b ∈ Z − {a} such that the distance between the first coordinates of b and a
smaller than 1/k, it follows that Bt (a, 1/k) ⊂ Bt (b, 2), a contradiction.
(3.4) X is not a wΔ-space.
Otherwise, X is a paracompact developable space, and hence X is a metrizable
space, which contradicts (3.3).
(3.5) Ceder [92] proved that X is an M1 -space. Amono and Mizokami [13] gave a
simple proof for X being an M1 -space.

Example 1.8.4 ([154, 366]) The Mrówka space ψ(D).

Let D be an infinite set. A family A of countable infinite subsets of D is said to be
almost disjoint if the intersection of any two different elements of A is a finite set.
If A is a maximal almost disjoint family of D,then A is uncountable. Otherwise,
denote A = {Ai }i∈N . For each i ∈ N,  A i − n<i A n = A i − n<i (A n ∩ Ai ) is
infinite, and hence there is xi ∈ Ai − n<i An . Let E = {xi : i ∈ N}. Then for each
i ∈ N, E ∩ Ai is finite, it follows that E ∈
/ A and A ∪ {E} is also an almost disjoint
family, a contradiction.
Let ψ(D) = A ∪ D. The set ψ(D) with a topology described in the following
way is called a Mrówka space: the points of D are isolated points in ψ(D); basic
neighborhoods in ψ(D) of a point A ∈ A are sets of the form {A} ∪ (A − F), F ∈
A<ω . Then ψ(D) is a locally compact space.
(4.1) ψ(D) has a σ -interior-preserving base.
Denote A = {Aα }α∈Λ . For each α ∈ Λ, let Aα = {x(α, n) : n ∈ N}. Define

B1 = {{x} : x ∈ D};
Bn+1 = {{Aα } ∪ {x(α, m) : m  n} : α ∈ Λ}, n ∈ N.

Then n∈N Bn is a σ -interior-preserving base for ψ(D).
(4.2) ψ(D) is a quasi-developable space.
For each x ∈ ψ(D), {st(x, Bn ) : n ∈ N, st(x, Bn ) = ∅} is a neighborhood base
of x in ψ(D), and hence ψ(D) is a quasi-developable space.
(4.3) ψ(D) is a pseudo-compact space.
Let f : ψ(D) → R be a continuous function. Since any infinite subset of D
has an accumulation point in A , f |D is a bounded function. So f is also a bounded
function, because D is a dense subset of ψ(D). Thus ψ(D) is a pseudo-compact
space.
(4.4) ψ(D) is not a meta-Lindelöf space.
42 1 The Origin of Generalized Metric Spaces

Since the open cover {{Aα } ∪ D}α∈Λ of ψ(D) has no point-countable open refine-
ment, ψ(D) is not a meta-Lindelöf, and hence ψ(D) has no point-countable base. It
follows that ψ(D) is neither a cosmic space nor a wM-space (see Theorem 1.7.7).
When D has certain cardinalities, the Mrówka space ψ(D) has some special prop-
erties.
(4.5) ψ(N) is a developable space.
We may assume that ∪A = N. For each n ∈ N, define

Fn = {1, 2, . . . , n},
Un = {{Aα } ∪ (Aα − Fn ) : α ∈ Λ} ∪ {{x} : x ∈ Fn }.

If x ∈ ψ(N), then

{Aα } ∪ (Aα − Fn ), x = Aα ,
st(x, Un ) =
{x}, x ∈ Fn .

Thus ψ(N) is a developable space.

(4.6) If |D| > c, then ψ(D) has no σ -locally finite closed pseudo-network [94].

Otherwise, let n∈N Pn be a closed pseudo-network for ψ(D) such that each Pn
is locally finite. By the maximality of A , Pn|D is a finite set, and hence n∈N Pn|D is
a countable pseudo-network for D. So |D|  ℵℵ0 0 = c, a contradiction. We can prove
that ψ(D) is not a β-space. Otherwise, by Theorem 1.7.7, ψ(D) is a semi-stratifiable
space, it follows that ψ(D) is a subparacompact space with a G δ -diagonal, and hence
ψ(D) has a σ -locally finite closed pseudo-network, also a contradiction.

Example 1.8.5 ([329]) The Michael line.

Suppose X = I and B is a Bernstein set in X [119, 230], i.e. B has the following
properties:
(1) Both B and I − B have the cardinality c and are dense in I with the Euclidian
topology.
(2) Any subset of B or I− B is countable whenever it is closed in I with the Euclidian
topology.
The set X with the following topology is called the Michael line (or the Michael
line about B): a set G is open in X if and only if there is a Euclidian open subset U
of I and a subset Z of B such that G = U ∪ Z . It is obvious that X is a regular space.
(5.1) X is a Lindelöf space.
Let {Vα }α∈Λ be an open cover of X . For every α ∈ Λ, there exist a Euclidian open
subset Uα of I and a subset Z α of B such that Vα = Uα ∪ Z α . Let

U = ∪{Uα : α ∈ Λ}.
1.8 Examples 43


Then there is a countable subset Λ1 of Λ such that U = α∈Λ1 Uα . By the property
 X − U is countable. Thusthere is a countable subset Λ2 of Λ such
of Bernstein sets,
that X − U ⊂ α∈Λ2 Vα , and hence X = α∈Λ1 ∪Λ2 Vα . So X is a Lindelöf space.
(5.2) X has a σ -disjoint base.
Let B be a countable base for I with the Euclidian topology. Then

B ∪ {{x} : x ∈ B}

is a σ -disjoint base for X . By Proposition 1.7.12, X is a σ  -space.

(5.3) X is not a β-space.
Otherwise, by Theorem 1.7.7, X is a semi-stratifiable space. By Corollary 1.4.5
and Theorem 1.2.14, X is a developable space, and it follows that X is a metrizable
space. However, X is a non-separable Lindelöf space, a contradiction. So X is not a
β-space.
Example 1.8.6 ([19]) The Arens space S2 .
Let X = {0} ∪ N ∪ (N × N) and F = N<ω . Denote the family of all functions
from N to N by NN . For every n, m ∈ N, F ∈ F and f ∈ NN , let

V (n, m) = {n} ∪ {(n, k) ∈ N × N : k  m},

H (F, f ) = ∪{V (n, f (n)) : n ∈ N − F}.

The set X with the following topology is called the Arens space and denoted briefly
as S2 : for each x ∈ X , take
⎧
⎨ {{x}}, x ∈ N × N,
Nx = {V (x, m) : m ∈ N}, x ∈ N,
⎩
{{x} ∪ H (F, f ) : F ∈ F , f ∈ NN }, x = 0

as a neighborhood base of x. Then X is a regular space.

(6.1) X is a g-second countable space.
For every n ∈ N, let Fn = {1, 2, . . . , n}. For every x ∈ X , define

Nx , x ∈ N ∪ (N × N),
Px =
{{x} ∪ (N − Fn ) : n ∈ N}, x = 0.

Then Px is a weak base of x, so X is g-first countable, and hence X is a g-second

countable space.
(6.2) X is a symmetrizable space.
For each n ∈ N, let

Un = {{0} ∪ (N − Fn )} ∪ {V (x, n) : x ∈ Fn } ∪ {{x} : x ∈ N × N}.

44 1 The Origin of Generalized Metric Spaces

Then Un is a cover of X . If x ∈ X , then

⎧
⎨ {0} ∪ (N − Fn ), x = 0,
st(x, Un ) = V (x, n), x = m < n,
⎩
{x}, x = (m, k), n > max{m, k}.

Thus {Un } is a weak development for X . By Proposition 1.6.14, X is a symmetriz-

able space.
It should be noted that the subspace X −N of X neither is discrete nor contains any
nontrivial convergent sequence, so X − N is not a sequential space. This shows that
none of g-first countability, g-second countability, g-metrizability, symmetrizability
and being a sequential space is hereditary. By Corollary 1.6.18, X is not a Fréchet–
Urysohn space.
(6.3) X × (P ∪ {0}) is not a sequential space.
Let
Y = X × (P ∪ {0}), F = {((i, j), 1/i + π/j) : i, j ∈ N}.

Then F ⊂ Y and (0, 0) ∈ F − F, so F is not a closed set in Y . On the other hand, for
any S ∈ S (Y ), denote F ∩ S = {yn }n∈N . Since the sequence {π1 (yn )} is convergent
in X , the first coordinates of {π1 (yn )} only can take finitely many values. Moreover,
the second coordinates of {π1 (yn )} also only can take finitely many values, because
the sequence {π2 (yn )} is convergent in P∪{0}. So, F ∩ S is actually a finite set, hence
a closed set in Y . It follows that Y is not a sequential space. Thus symmetrizability
is not finitely productive. Since Y is not g-first countable at (0, 0), the product of a
countable weak base of 0 in X and a countable weak base of 0 in P ∪ {0} is not a
countable weak base of (0, 0) in Y .

Example 1.8.7 ([37]) The sequential fan Sω .

Let X = {0} ∪ (N × N). For every n, m ∈ N and f ∈ NN , let

W (n, m) = {(n, k) ∈ N × N : k  m},

L( f ) = ∪{W (n, f (n)) : n ∈ N}.

The set X with the following topology is called a sequential fan (or sequential fan
space) and denoted briefly as Sω : for each x ∈ X , take

{{x}}, x ∈ N × N,
Nx =
{{x} ∪ L( f ) : f ∈ NN }, x = 0

as a neighborhood base of x. Then X is a regular space.

(7.1) X is an ℵ0 -space.
Let P = {{x} : x ∈ X } ∪ {W (n, m) : n, m ∈ N}. Then P is a countable
k-network for X , and hence X is an ℵ0 -space.
1.8 Examples 45

(7.2) X is a Fréchet–Urysohn space.

If A ⊂ X and x ∈ A, we may assume that x ∈ A − A, then x = 0. Since there is
n ∈ N such that W (n, 1) ∩ A is infinite, we can take a sequence {xm } ⊂ W (n, 1) ∩ A
consisting of infinitely many points. Then xm → x. So X is a Fréchet–Urysohn
space.
(7.3) X is an M1 -space.
Since N × N consists of isolated points of X , N0 is a closure-preserving family of
open sets of X , so X has a σ -closure-preserving base, and hence X is an M1 -space.
(7.4) X is not a first countable space, because there is no countable neighborhood
base of 0 in X .
X is homeomorphic to a quotient space of the topological sum of countably many
convergent sequences. From this point of view, we can construct the fan space in a
similar way. To each α < ω1 , assume that a space X α is homeomorphic to S1 . Denote
the
 limit point of X α as x α . For the
 subspace A = α<ω1 {x α } of the metric space
α<ω1 X α , the quotient space ( α<ω1 X α )/A is called the fan space, and denoted
by Sω1 .
(7.5) Sω1 has no point-countable cs ∗ -network [453].
Let s be the accumulation point of Sω1 . For each α < ω1 , let Yα = X α − {s}. If
Sω1 has a point-countable cs ∗ -network P, denote {P ∈ P : s ∈ P and there are
infinitely many α < ω1 such that Yα ∩ P = ∅} as {Pn }n∈N , then we can inductively
choose a subset {yn : n ∈ N} of Sω1 , such that, yn ∈ Pn − {s} and different yn belongs
to different Yα . Then {yn : n ∈ N} is a closed set in Sω1 . Let

V = Sω1 − {yn : n ∈ N},

F = {P ∈ P : P ⊂ V },
H = ∪{F ∈ F : s ∈ F}.

If s ∈ P ∈ F , then P ∈/ {Pn : n ∈ N}, so P only meets finitely many Yα , and hence

H also only meets countably many Yα . As a consequence, there is β < ω1 such that
Yβ ∩ H = ∅. Let V ∩ Yβ = {xn : n ∈ N}. Then xn → s ∈ V , so there exist a
subsequence {xni } of {xn } and P ∈ F such that {s} ∪ {xni : i ∈ N} ⊂ P, and hence
Yβ ∩ H = ∅, a contradiction. Therefore, Sω1 has no point-countable cs ∗ -network.

Example 1.8.8 ([326]) The Michael space.

Let βN be the Čech-Stone compactification of N. Take p ∈ βN − N, and let
X = N ∪ { p}. The set X with the subspace topology of βN is called the Michael
space. Obviously, X is a regular space. Since each compact set in βN either is finite
or has the cardinality 2c [119], every compact set in X is a finite set, so X is not a
sequential space.
(8.1) X is an ℵ0 -space.
46 1 The Origin of Generalized Metric Spaces

Since each compact set in X is finite, {{x} : x ∈ X } is a countable k-network of

X , and hence X is an ℵ0 -space.
(8.2) X is an M1 -space.
Since N is a set of isolated points in X , any open neighborhood base of p is a
closure-preserving family in X , so X has a σ -closure-preserving base, and hence X
is an M1 -space.

Example 1.8.9 ([9]) The Alexandroff double-arrow space.

Let X = I × {0, 1} − {(0, 0), (1, 1)}. Define the lexicographic order < on X as
follows: for every (a, b), (s, t) ∈ X, (a, b) < (s, t) if and only if a < s, or a = s
and b < t. The totally ordered set (X, <) with the ordered topology is called the
Alexandroff double-arrow space or the Alexandroff-Urysohn double-arrow space.
For each x = (a, b) ∈ X , a basic neighborhood of x has the following form: if
b = 0, then ((a − 1/n, 1), x] = {x} ∪ {(s, t) : a − 1/n < s < a, t = 0, 1}; if b = 1,
then [x, (a + 1/n, 0)) = {x} ∪ {(s, t) : a < s < a + 1/n, t = 0, 1}.
We recall the right half-open interval topology before analyzing the properties
of X . Let S be the real line. Then the topology generated by taking B = {[a, b) :
a, b ∈ S} as a base is called the right half-open interval topology or the Sorgenfrey
line topology for S [433,437], and S is called the Sorgenfrey line. Obviously, S is a
separable regular space.
(9.1) S is a hereditarily Lindelöf space.
∗
topology for R by τ . Let A ⊂ S. If a subfamily {Uα }α∈Λ of
Denote the Euclidian
B covers A, let U = α∈Λ intτ ∗ (Uα ), then some countable family {intτ ∗ (Uαn )}n∈N
covers U . Let B = A − U . If b ∈ B, then there is sb > b such that (b, sb ) ∩ B = ∅.
So {(b, sb ) : b ∈ B} is a family of disjoint open intervals in R, and hence B is
countable. As a consequence, there is a countable subfamily of {Uα }α∈Λ covering A.
Therefore, A is a Lindelöf space.
(9.2) S is a monotonically normal space [186].
Let F, K be a pair of disjoint closed subsets of S. For each x ∈ F, take

+∞, (x, +∞) ∩ K = ∅,
sx =
inf((x, +∞) ∩ K ), otherwise.

Then x < sx and [x, sx ) ∩ K = ∅. Let D(F, K ) = ∪{[x, sx ) : x ∈ F}. Then

F ⊂ D(F, K ) ∈ τ (S). To prove that D is a monotone normality operator on S,
we only need to verify that D(F, K ) ⊂ S − K . If y ∈ K , then there is z > y
such that [y, z) ∩ F = ∅. It follows for each x ∈ F, [x, sx ) ∩ [y, z) = ∅, so
/ D(F, K ). Thus S is a monotonically normal
[y, z) ∩ D(F, K ) = ∅, and hence y ∈
space.
(9.3) S is not a β-space.
1.8 Examples 47

We first prove that S is not a metrizable space. Otherwise, S is a second countable

space, and hence some countable subfamily {[an , bn )}n∈N of B is a base for S. Take
c ∈ S − {an : n ∈ N}. Then there is m ∈ N such that c ∈ [am , bm ) ⊂ [c, c + 1), and
hence c = am , a contradiction. So S is not a metrizable space.
If S is a β-space, then there exists a β-function g on S such that g(n, x) ⊂ [x, x +
1/n), thus g is a developable function on S. Because in fact, if a point x and sequences
{xn }, {yn } in S satisfy that {x, xn } ⊂ g(n, yn ), then any subsequence of {yn } has an
accumulation point and yn  x, xn < yn +1/n, so yn → x, and hence xn → x. Thus,
S is a developable space, and it follows that S is a metrizable space, a contradiction.
The topology for the set S generated by taking {(a, b] : a, b ∈ S} as a base is called
the left half-open interval topology. Below we turn to investigate the properties of
the Alexandroff double-arrow space X .
(9.4) X is a compact non-metrizable space.
Since the subspace (0, 1] × {1} of X has the right half-open interval topology,
X is not metrizable. Because X is a totally ordered set with the least upper bound
property (An ordered set A is said to have the least upper bound property if every
nonempty bounded above subset A0 of A has a least upper bound) and X itself is a
closed interval [(0, 1), (1, 0)], X is a compact space. Of course, we can directly prove
that X is compact as follows. Since a space with either the right half-open interval
topology or the left half-open interval topology is a hereditarily Lindelöf space, X
is a hereditarily Lindelöf space. Thus, to prove that X is compact, we only need to
prove that any infinite subset of X has an accumulation point. For any infinite subset
Y of X , the set π1 (Y ), as an infinite subset of I, has an accumulation point a for the
Euclidian topology, and hence (a, 0) or (a, 1) is an accumulation point of Y in X .
(9.5) Each closed subset of X has a countable neighborhood base in X .
By the convergence lemma, we only need to verify that X is a perfect space. For
each nonempty open subset G of X , there is a cover G of G consisting of open
subsets of X such that the closure of each element of G is contained in G. Since G
is a Lindelöf subspace of X , some countable subfamily of G covers G. So, G is an
Fσ -set in X .
Example 1.8.10 ([83]) There is a locally compact space X satisfying the following
conditions:
(1) X has a locally finite closed cover consisting of finite sets.
(2) X has a point-countable cover consisting of compact clopen sets.
(3) X is not a subparacompact space.
Let ω2 be the smallest ordinal number with the cardinality ℵ2 and let X = ω2 × ω2 −
{(0, 0)}. For each α ∈ ω2 − {0}, put

Hα = ω2 × {α}, Vα = {α} × ω2 .

Define a topology for X as follows: for each α ∈ ω2 − {0}, a basic neighborhood

of (0, α) has the form {(0, α)} ∪ (Hα − F), F ∈ Hα<ω , and a basic neighborhood of
48 1 The Origin of Generalized Metric Spaces

(α, 0) has the form {(α, 0)} ∪ (Vα − F), F ∈ Vα<ω ; the other points of X are isolated.
Obviously, X is a locally compact space.
For each (α, β) ∈ X , define F(α, β) = {(α, β), (α, 0), (0, β)}.
Let F = {F(α, β) : 0 < α, β < ω2 }. Then F is a locally finite closed cover
of X .
(10.1) X is a metacompact space.
Let
P = {Hα : α ∈ ω2 − {0}} ∪ {Vα : α ∈ ω2 − {0}}.

Then P is a point-finite cover of X consisting of compact clopen sets. Thus X is

metacompact space.
(10.2) If A, B ⊂ X and every A ∩ Vα and B ∩ Hα are countable, then X = A ∪ B.
Otherwise, let
β0 = sup{β : (α, β) ∈ A ∩ Vα , α ∈ ω1 }.

Then β0 < ω2 and the nonempty subset ω1 × (ω2 − (β0 + 1)) of X does not
meet A, and hence ω1 × (ω2 − (β0 + 1)) ⊂ B. Take δ ∈ ω2 − (β0 + 1). Then
ω1 × {δ} ⊂ B ∩ Hδ , a contradiction. 
Assume that the open cover P of X has a refinement n∈N P n such that each
Pn is discrete. Denote the set of all the elements of Pn contained in Vα and in Hα
by Vn and Hn respectively. Then Pn = Vn ∪ Hn . For each n ∈ N, let An = ∪Hn
and Bn = ∪Vn . For each α ∈ ω2 , there is a neighborhood of (α, 0) only meeting at
most one element of Hn , so Vα only meets finitely many elements ofHn , and hence
An ∩Vα is finite. Similarly, Bn ∩ Hα is finite. Thus, by (10.2), X = n∈N (An ∪ Bn ),
i.e. n∈N Pn is not a cover of X , a contradiction. So X is not a subparacompact
space.
In the next part of this section, we give three examples to show that some classes
of spaces are not preserved under certain topological operations. First, we discuss
the hereditary problem of topological property.

Example 1.8.11 Any topological property between compactness and being a β-

space is not open hereditary.
Take a set D such that |D| > c. By Example 1.8.4, ψ(D) is a locally compact
space but is not a β-space. Let Z be the Alexandroff one point compactification of
ψ(D). Then Z is a compact space. Since the open subspace ψ(D) of Z is not a
β-space, any topological property between compactness and being a β-space is not
open hereditary.
Next, the following example shows why we only consider the countable case
when discussing the products of generalized metric spaces.

Example 1.8.12 The space Nω1 has none of the following properties: the point G δ -
property, being a q-space or being a β-space.
Denote X = Nω1 . Obviously, X does not have the point G δ -property.
1.8 Examples 49

(12.1) X is not a q-space.

Because otherwise, let g be a q-function on X . We may assume that each g(n, x)
is a basic open set in X . Take z ∈ X such that for every α < ω1 , πα (z) = 1. There
exists β < ω1 such that πβ (g(n, z)) = N for each n ∈ N. For each n ∈ N, pick
xn ∈ X such that 
1, α ∈ ω1 − {β},
πα (xn ) =
n, α = β.

Then xn ∈ g(n, z), but the sequence {xn } has no accumulation point, a contradiction.
So X is not a q-space.
(12.2) X is not a β-space.
Because otherwise, let g be a β-function on X . For each basic open set V in X ,
define
R(V ) = {α < ω1 : πα (V ) = N}.

Take x1 ∈ X such that for each α < ω1 , πα (x1 ) = 1. Let V1 be a basic neighborhood
(i.e. a neighborhood with the form of basic open sets) of x1 and V1 ⊂ g(1, x1 ). Take
x2 ∈ X such that 
πα (x1 ), α ∈ R(V1 ),
πα (x2 ) =
2, α∈/ R(V1 ).

Let V2 be a basic neighborhood of x2 such that V2 ⊂ g(2, x2 ) and R(V1 ) ⊂ R(V2 ).

We can choose inductively a sequence {xn } and a sequence {Vn } of basic open sets
such that
(i) xn ∈ Vn ⊂ g(n, xn ),
(ii) R(Vn ) ⊂ R(Vn+1 ),
(iii) 
πα (xn ), α ∈ R(Vn ),
πα (xn+1 ) =
n + 1, α ∈ / R(Vn ).

Now, pick x ∈ X such that


πα (xn ), there is n ∈
N such that α ∈ R(Vn ),
πα (x) =
1, α ∈ ω1 − n∈N R(Vn ).

Then for every n ∈ N, x ∈ Vn ⊂ g(n, xn ). If we pick β ∈ ω1 − n∈N R(Vn ), then
πβ (xn ) = n, and hence {xn } has no accumulation point, a contradiction. Thus X is
not a β-space.
The last example in this section shows that in Theorem 1.5.11, the assumption of
strong Σ-spaces can not be replaced with that of semi-metrizable spaces. We first
prove the following lemma [334].
Lemma 1.8.13 (CH) Let {τλ }λ∈Λ be a collection of topologies with weight not larger
than c for a set X , where |Λ|  c. If a countable set A in X is not the intersection
50 1 The Origin of Generalized Metric Spaces


of countably many elements of λ∈Λ τλ , then there is an uncountable subset Y of X
containing A such that for every τλ , Y is a Lindelöf subspace.

Proof For each λ ∈ Λ, let Uλ be a base of (X, τλ ) with cardinality not greater than
c, and we may assume Uλ is closed under countable unions. Let

U = {U ∈ Uλ : A ⊂ U, λ ∈ Λ}.

By CH, |U |  ℵ1 . Denote U = {Uα }α<ω1 . Since for each α < ω1 , β<α Uβ − A
is uncountable, we can choose a subset {xα : α < ω1 } of X such that

xα ∈ ( Uβ − {xβ : β < α}) − A.
β<α

Let Y = A ∪ {xα : α < ω1 }. We prove that for each λ ∈ Λ, Y is a Lindelöf subspace

of (X, τλ ). Assume that a subset V of τλ covers Y , and we may assume V ⊂ Uλ .
Pick a countable subset W of V covering A and let W = ∪W . Then W ∈ U , and
hence there is α < ω1 such that W = Uα . Since Y − W is countable, there is a
countable subset F of V covering Y − W . Then the countable subset W ∪ F of V
covers Y , and hence Y is a Lindelöf subspace for τλ . 

Example 1.8.14 There exist topologies τ1 and τ2 for R2 satisfying the following
conditions:
(1) (R2 , τ1 ) and (R2 , τ2 ) are homeomorphic semi-metrizable regular spaces.
(2) If Q2 ⊂ Z ⊂ R2 and |Z | = c, then (Z , τ1 ) × (Z , τ2 ) is not a normal space.
(3) Q2 is not the intersection of countably many elements of τ1 ∪ τ2 .
(4) (CH) There exists an uncountable subset Z of R2 containing Q2 such that (Z , τ1 )
and (Z , τ2 ) are homeomorphic hereditarily Lindelöf spaces.
For convenience, we call Bt (x, 1/n) defined in Example 1.8.2 a horizontal bowtie
neighborhood of x, and a vertical bowtie neighborhood of x is defined in a similar
way. Denote the topologies for R2 generated by the horizontal bowtie neighborhoods
and the vertical bowtie neighborhoods by τ1 and τ2 respectively.
(1) By Example 1.8.2, (R2 , τ1 ) is a semi-metrizable regular space. The corre-
sponding (x, y) → (y, x) is a homeomorphism from (R2 , τ1 ) onto (R2 , τ2 ), so
(R2 , τ2 ) is also a semi-metrizable regular space.
(2) Let E = (Z , τ1 ) × (Z , τ2 ). Then Q2 × Q2 is a countable dense subset of E
and {(x, x) : x ∈ Z } is a closed discrete subspace of E with the cardinality c. By the
Urysohn extension theorem, E is not a normal space.
(3) Denote the Euclidean topology on R2 by τ . If Q2 ⊂ U ∈ τ1 ∪ τ2 , then intτ (U )
is dense in (R2 , τ ). By the Baire category theorem, Q2 is not the intersection of
countably many dense open subsets of (R2 , τ ), and Q2 is not the intersection of
countably many subsets of τ1 ∪ τ2 .
(4) (CH) By (3) and Lemma 1.8.13, there is an uncountable subset Y of R2
containing Q2 such that Y is a Lindelöf subspace for both τ1 and τ2 . Let Z = Y ∪ Y ∗ ,
1.8 Examples 51

where Y ∗ = {(y, x) : (x, y) ∈ Y }. Since both Y and Y ∗ are Lindelöf subspaces for
τ1 , it follows from Theorem 1.2.5 that Z is a hereditarily Lindelöf space for τ1 . It
is obvious that (Z , τ1 ) is homeomorphic to (Z , τ2 ), so (Z , τ2 ) is also a hereditarily
Lindelöf space.
By (2) and (4) of Example 1.8.14, we have the following corollary.

Corollary 1.8.15 [61, 334] (CH) There is a semi-metrizable hereditarily Lindelöf

regular space Z such that Z 2 is not a normal space.

Question 1.8.16 ([334]) Is there any semi-metrizable hereditarily Lindelöf regular

space X such that X 2 is not a normal space?

Question 1.8.17 ([162]) Is there any semi-metrizable hereditarily Lindelöf regular

space Z which is not a σ -space?

Moore [356] solved the following famous L-space problem positively: Is there
any hereditarily Lindelöf regular space which is not separable?
Chapter 2
Mappings on Metric Spaces

In process of the development of general topology, because of the participation of

various branches of mathematics and the driving of internal problems, a vast diver-
sity of classes of topological spaces were evolved. Detailed analysis and extensive
classification of these objects appear in different poses and with different expressions
are the necessary conditions for the existence of general topology and form its main
internal tasks [31]. At the 1961 Prague Topological Symposium, Alexandroff [3]
raised two kinds of questions:

Question 2.0.1 ([3]) Which spaces can be represented as images of “nice” (e.g.
metric or zero-dimensional, etc.) spaces under “nice” continuous mappings?

Question 2.0.2 ([3]) Which spaces can be mapped onto “nice” spaces by “nice”
mappings?

The idea of classifying spaces by mappings caused by Alexandroff’s questions

leads to a method of mutual classifications of spaces and mappings. Alexandroff [4]
and Arhangel’skiı̌ [31] thought that the essence of this method is the following three
fundamental problems which are closely tied up with one another:

Question 2.0.3 ([31]) Under what circumstances can each space of a given class
A be mapped onto a space of a class B by means of a mapping belonging to a class
F?

Question 2.0.4 ([31]) If the class F (A ) of spaces are images of spaces of type A
by mappings of type F , then what internal properties can the spaces belonging to
F (A ) have?

Question 2.0.5 ([31]) Let F (A , B) denote the class of mappings whose domain
is a member of the class A and whose range is a member of the class B. Let H
be some other class of mappings. What are the properties of mappings of the class
F (A , B) ∩ H ?

These general formulations involve, in particular, the following question:

© Atlantis Press and the author(s) 2016 53
S. Lin and Z. Yun, Generalized Metric Spaces and Mappings,
Atlantis Studies in Mathematics 6, DOI 10.2991/978-94-6239-216-8_2
54 2 Mappings on Metric Spaces

Question 2.0.6 ([31]) What topological properties are preserved by these and other
mappings?
The significance of the idea of mutual classifications of spaces and mappings,
i.e. Alexandroff-Arhangel’skiı̌’s questions, is that to reveal the internal properties of
various classes of topological spaces by mappings, and to use mappings as a link
connecting the multifarious topological spaces in one. Practice shows this principle
not only instilled fresh blood to general topology in many classic topics, but also
produced many new research directions, and brought the prosperity of general topol-
ogy from the late 60s to the whole 80s in the 20th century [4, 238]. As a supplement,
readers can read the survey papers [268, 298].
Which classes of spaces can be regarded as the “nice” classes of spaces in Alexan-
droff’s questions?
Which classes of mappings can be regarded as the “nice” classes of mappings in
Alexandroff’s questions?
We start from the class of metrizable spaces, and to show that quotient map-
pings, pseudo-open mappings, open mappings and closed mappings really satisfy
the requirements of being “nice” mappings. In this chapter, we follow the idea of
Alexandroff and look for some important intrinsic properties of images or preimages
of metric spaces.
All the mappings in Sect. 2.1 are assumed to be continuous mappings, and all
the mappings in Sects. 2.2–3.9 are continuous onto mappings.

2.1 Classes of Mappings

In this section, we give definitions for certain classes of mappings and investigate
their basic properties. The relationships between different mappings are very exten-
sive. Here we only describe several properties needed in the subsequent sections.
The mapping lemma (see Proposition 2.1.12) reflects the relationships between the
transformations of mappings.
Definition 2.1.1 Let f : X → Y be a mapping.
(1) f is called a quotient mapping [50] if, for each U ⊂ Y with f −1 (U) ∈ τ (X),
U ∈ τ (Y ).
(2) f is called a pseudo-open mapping [26] if, for each y ∈ Y with f −1 (y) ⊂ V ∈
τ (X), y ∈ f (V )◦ .
(3) f is called a countably bi-quotient mapping [424] if, for each y ∈ Y and each
countable open family U in X covering f −1 (y), there is P ∈ U F such that
y ∈ f (P)◦ .
(4) f is called an open mapping [42] if f (V ) ∈ τ (Y ) whenever V ∈ τ (X).
(5) f is called a closed mapping [196] if f (F) ∈ τ c (Y ) whenever F ∈ τ c (X).
We first point out two basic relationships between mappings. Let f : X → Y be
a mapping. If A ⊂ X, B ⊂ Y , then
2.1 Classes of Mappings 55

(1) f −1 (B) ⊂ A ⇔ B ⊂ Y − f (X − A);

(2) f (A ∩ f −1 (B)) = f (A) ∩ B.

Proposition 2.1.2 (1) Every open mapping is a countably bi-quotient mapping,

every countably bi-quotient mapping is a pseudo-open mapping and every
pseudo-open mapping is a quotient mapping.
(2) Every closed mapping is a pseudo-open mapping.

Proof We only need to prove that every closed mapping is a pseudo-open mapping
and every pseudo-open mapping is a quotient mapping.
Let f : X → Y be a mapping.
Assume that f is a closed mapping. If y ∈ Y and f −1 (y) ⊂ V ∈ τ (X), then
y ∈ Y − f (X − V ) ⊂ f (V ), so y ∈ f (V )◦ . Thus f is a pseudo-open mapping.
Assume that f is a pseudo-open mapping. If U ⊂ Y and f −1 (U) ∈ τ (X), then for
any y ∈ U, f −1 (y) ⊂ f −1 (U), so y ∈ U ◦ , and hence U ∈ τ (Y ). Thus f is a quotient
mapping. 

There exist examples to show that not every pseudo-open mapping is a composi-
tion of an open mapping and a closed mapping [38].
Below we consider mappings with some additional conditions on fibers.

Definition 2.1.3 Let f : X → Y be a mapping.

(1) f is called a compact mapping [471] (resp. a countably compact mapping, an L-
mapping, an s-mapping) if, for every y ∈ Y , f −1 (y) is a compact (resp. countably
compact, Lindelöf, separable) set in X.
(2) f is called a peripherally compact mapping [471] (resp. peripherally countably
compact mapping, peripheral L-mapping) if, for each y ∈ Y , ∂f −1 (y) is a com-
pact (resp. countably compact, Lindelöf) set in X.
(3) f is called a perfect mapping [471] if f is a closed compact mapping.
(4) f is called a quasi-perfect mapping if f is a closed countably compact mapping.

Obviously, every quasi-perfect mapping is a countably bi-quotient mapping.

Definition 2.1.4 Let f : X → Y be a mapping.

(1) f is called a compact-covering mapping [331] if, for every K ∈ K (Y ), there is
L ∈ K (X) such that f (L) = K.
(2) f is called a sequence-covering mapping [424] if, for every S ∈ S (Y ), there is
L ∈ K (X) such that f (L) = S.
(3) f is called a sequentially quotient mapping [64] if, for every S ∈ S (Y ), there is
L ∈ S (X) such that f (L) is a subsequence of S.

We should note that in the different literatures, the definitions of mappings crown
to “sequence-covering mappings” in the name may be different [167].
56 2 Mappings on Metric Spaces

Remark Suppose φ is a mapping property. An image (resp. a preimage) of a space

under a mapping having the property φ is briefly called a φ image (resp. preimage)
of a space. For example, an image of a space under a compact mapping is called a
compact image of a space. A φ image of a space and the range space is regular (resp.
completely regular, normal) is called a regular (resp. completely regular, normal) φ
image of a space. The definition of a regular (resp. completely regular, normal) φ
preimage of a space is similar.

In the following, we give characterizations of some classes of mappings.

Proposition 2.1.5 ([26]) Let f : X → Y be a mapping. Then f is a pseudo-open

mapping if and only if for every B ⊂ Y , fB is a quotient mapping.

Proof Suppose f is a pseudo-open mapping. Then for every B ⊂ Y , fB : f −1 (B) → B

is a pseudo-open mapping, and hence fB is a quotient mapping.
Conversely, for each y ∈ Y with f −1 (y) ⊂ V ∈ τ (X), if y ∈ / f (V )◦ , let Z =
Y − f (V ); if y ∈ Z − Z, let F = Z ∪ {y}, then Z is not closed in F. Since fF is a
quotient mapping, f −1 (Z) is not a closed set in f −1 (F), so there is x ∈ f −1 (Z)∩f −1 (y),
and hence y = f (x) ∈ f (f −1 (Z)), as a consequence, Z ⊂ f (f −1 (Z)). Thus,

y ∈ Y − f (V ) ⊂ f (f −1 (Y − f (V ))) ⊂ f (X − V ),

it follows that f −1 (y) ∩ (X − V ) = ∅, a contradiction. Therefore f is a pseudo-open

mapping. 
A mapping satisfying the equivalent condition of being a pseudo-open mapping
in the above proposition is called a hereditarily quotient mapping [26].
Proposition 2.1.6 ([64]) Let f : X → Y be a mapping. Then f is a sequentially
quotient mapping if and only if for each F ⊂ Y , F is a sequentially closed (resp.
sequentially open) set in Y whenever f −1 (F) is a sequentially closed (resp. sequen-
tially open) set in X.

Proof Suppose f is a sequentially quotient mapping. Suppose F ⊂ Y and f −1 (F) is a

sequentially closed set in X. If F is not a sequentially closed set in Y , then F contains
a sequence {yn } such that yn → y ∈ Y − F, and hence there exist a sequence {xi } in
X and a subsequence {yni } of {yn } such that xi ∈ f −1 (yni ) and xi → x ∈ f −1 (y). Then
x ∈ f −1 (F), and hence y ∈ F, a contradiction.
Conversely, let {yn } be a sequence in Y such that yn → y. We may assume yn = y
for every n ∈ N. Since {yn : n ∈ N} is not a sequentially closed set in Y , ∪{f −1 (yn ) :
n ∈ N} is not a sequentially closed set in X. However, f −1 (y) ∪ (∪{f −1 (yn ) : n ∈ N})
is a sequentially closed set in X, so there is {pk } ⊂ ∪{f −1 (yn ) : n ∈ N} such that
pk → x ∈ f −1 (y). Since each f −1 (yn ) is a sequentially closed set, there are at most
finitely many pk belong to any fixed f −1 (yn ), so we can pick a subsequence {yni } of
{yn } and pki ∈ f −1 (yni ). Let L = {x} ∪ {pki : i ∈ N}. Then L ∈ S (X) and f (L)
contains a subsequence of {y} ∪ {yn : n ∈ N}. Thus f is a sequentially quotient
mapping. 
2.1 Classes of Mappings 57

Proposition 2.1.7 ([119]) Let f : X → Y be a mapping. Then f is a closed mapping

if and only if for each y ∈ Y with f −1 (y) ⊂ U ∈ τ (X), there is V ∈ τ (Y ) such that
y ∈ V and f −1 (V ) ⊂ U.

Proof Suppose f is a closed mapping. For each y ∈ Y with f −1 (y) ⊂ U ∈ τ (X),

take V = Y − f (X − U). Then y ∈ V ∈ τ (Y ) and f −1 (V ) ⊂ U.
Conversely, suppose F ∈ τ c (X). Then for any y ∈ Y − f (F), f −1 (y) ⊂ X − F ∈
τ (X) holds, and hence there is V ∈ τ (Y ) such that y ∈ V and f −1 (V ) ⊂ X − F, so
V ∩ f (F) = ∅, and it follows that f (F) ∈ τ c (Y ). Thus f is a closed mapping. 

Proposition 2.1.8 ([32]) Let f : X → Y and g : X → Z be mappings. If f is a

perfect mapping, then f
g is a perfect mapping.

Proof Let h = f
g. Then h can be expressed as the composition of the following
two mappings:

idX
g : X → X × Z, f × idZ : X × Z → Y × Z.

Since idX separates points from closed sets in X, by the diagonal theorem idX
g is
a closed embedding, hence a perfect mapping. Because both f and idZ are perfect
mappings, f × idZ is a perfect mapping. So h is a perfect mapping. 

Corollary 2.1.9 Assume that Φ is a closed hereditary and finitely productive topo-
logical property. If there exist a one-to-one mapping f : X → Y and a perfect
mapping g : X → Z, where Y and Z have Φ, then X also has Φ.

Proof Put h = f
g : X → Y × Z. By Proposition 2.1.8, h is a perfect mapping.
Since h is a one-to-one mapping, h is a closed embedding, so X also has Φ. 

Proposition 2.1.10 ([70]) If X is a compact space, then for any space Y , π1 : Y ×

X → Y is a perfect mapping.

Proof Obviously, π1 is a compact mapping. Let F ∈ τ c (Y × X). If y ∈ / π1 (F),

then ({y} × X) ∩ F = ∅, so there is an open subset V of Y containing y such that
(V × X) ∩ F = ∅, and hence V ∩ π1 (F) = ∅. As a consequence, π1 (F) ∈ τ c (Y ).
Thus π1 is a closed mapping. 

In the second part of this section, we discuss more precise relationships between
different classes of mappings under certain conditions. We first introduce two con-
cepts.

Definition 2.1.11 (1) A space X is called a strongly Fréchet–Urysohn  space [424]

if, for every decreasing sequence {An } of sets in X with x ∈ n∈N An , there is
xn ∈ An for each n ∈ N such that xn → x.
(2) X is called a k-space [136] if, for each A ⊂ X, A ∈ τ c (X) provide K ∩ A ∈ τ c (X)
for any K ∈ K (X).
58 2 Mappings on Metric Spaces

Obviously, first countable space ⇒ strongly Fréchet–Urysohn space ⇒ Fréchet–

Urysohn space ⇒ sequential space ⇒ k-space. By the definitions, strongly Fréchet–
Urysohn spaces are additive and hereditary; k-spaces are additive, open hereditary
and closed hereditary.

Proposition 2.1.12 (The mapping lemma) Let f : X → Y be a mapping.

(1) If Y is a k-space and f is a compact-covering mapping, then f is a quotient
mapping [331].
(2) If Y is a sequential space and f is a sequentially quotient mapping or a sequence-
covering mapping, then f is a quotient mapping [64, 424].
(3) If Y is a Fréchet–Urysohn space and f is a quotient mapping, then f is a pseudo-
open mapping [26, 133].
(4) If Y is a strongly Fréchet–Urysohn space and f is a quotient mapping, then f is
a countably bi-quotient mapping [336].
(5) If X is a sequential space and f is a quotient mapping, then f is a sequentially
quotient mapping [64].

Proof (1) Suppose F ⊂ Y and f −1 (F) ∈ τ c (X). For each K ∈ K (Y ), there is L ∈

K (X) such that f (L) = K. Since f −1 (F)∩L ∈ K (X), F ∩K = f (f −1 (F)∩L) ∈
K (Y ) ⊂ τ c (Y ), so F ∈ τ c (Y ). Thus f is a quotient mapping.
(2) If f is a sequence-covering mapping, then by replacing K (Y ) with S (Y ), the
proof for f being a quotient mapping is similar to that in (1).
Let f be a sequentially quotient mapping. If F ⊂ Y and f −1 (F) ∈ τ c (X), then
f −1 (F) is a sequentially closed set in X. By Proposition 2.1.6, F is a sequentially
closed set in Y , so F ∈ τ c (Y ), thus f is quotient mapping.
(3) Suppose y ∈ Y and f −1 (y) ⊂ V ∈ τ (X). If y ∈ Y − f (V )◦ , then y ∈ Y − f (V ),
thus there is a sequence {yn } ⊂ Y − f (V ) such that yn → y. Let

Z = {yn : n ∈ N}, F = f −1 (Z).

Then F ⊂ f −1 (Z) = F ∪f −1 (y). Since f −1 (y) ⊂ V and V ∩F = ∅, f −1 (y)∩F =

∅, so F ∈ τ c (X), and hence f −1 (Y − Z) = X − F ∈ τ (X), thus Y − Z ∈ τ (Y ),
a contradiction. As a consequence, y ∈ f (V )◦ , so f is a pseudo-open mapping.
(4) If f is not a countably bi-quotient mapping, then there exist y ∈ Y and a countable
family {Ui }i∈N of open subsets in X  covering f −1 (y) such that for every n ∈
◦
N, y ∈ Y − f ( in Ui ) = Y − f  ( in Ui ). Since Y is a strongly Fréchet–
Urysohn space, there is yn ∈ Y − f ( in Ui ) for each n ∈ N such that yn → y.
Let E = {yn : n ∈ N} and F = {y} ∪ E. By (3) and Proposition 2.1.5, fF is a
quotient mapping. Since E is not a closed subset of F, f −1 (E) is not a closed
subset of f −1 (F), so there is x ∈ f −1 (E) ∩ f −1 (y), and hence there is m ∈ N such
that x ∈ Um . It follows that there is k  m such that f −1 (yk ) ∩ Um = ∅, thus
yk ∈ f (Um ), a contradiction. So f is a countably bi-quotient mapping.
2.1 Classes of Mappings 59

(5) If F ⊂ Y and f −1 (F) is a sequentially closed set in X, then f −1 (F) is a closed

set in X, so F is a closed set in Y , and hence F is a sequentially closed set in Y .
By Proposition 2.1.6, f is a sequentially quotient mapping. 

Proposition 2.1.13 Let f : X → Y , where X is of the point G δ -property. If one of

the following is true, then f is a sequentially quotient mapping.
(1) f is a sequence-covering mapping.
(2) f is a closed mapping and X is a regular space [291].

Proof (1) If S ∈ S (Y ), then there is K ∈ K (X) such that f (K) = S. Denote

S = {y} ∪ {yn : n ∈ N}, where yn → y. Pick xn ∈ f −1 (yn ) ∩ K for each
n ∈ N. Since X is a space of the point G δ -property, by Theorem 1.7.7, there is
a subsequence {xni } of {xn } such that xni → x ∈ f −1 (y). Thus f is a sequentially
quotient mapping.
(2) Let {yn } be any nontrivial sequence converging to y in Y . Pick xn ∈ f −1 (yn )
for each n ∈ N. Since f is a closed mapping, each subsequence of {xn } has
an accumulation point. Let x be an accumulation point of {xn } and let  {Vn } be a
sequence of open neighborhoods of x, such that, V n+1 ⊂ Vn and {x} = n∈N Vn .
Take a subsequence {tk } of {xn } such that
 tk ∈ Vk . Let p be an accumulation point
of any subsequence of {tk }. Then p ∈ k∈N V k , so p = x, which means that x is
the only accumulation point of {tk }, and hence {tk } is a convergent subsequence
of {xn }. Thus f is a sequentially quotient mapping. 

Lemma 2.1.14 ([330, 336]) Let f : X → Y be a closed mapping, where X is a

normal space or a countably paracompact space. If Y is a q-space or a strongly
Fréchet–Urysohn space, then f is a peripherally countably compact mapping.

Proof If there is y ∈ Y such that ∂f −1 (y) is not a countably compact set in X, then
∂f −1 (y) contains a discrete countable closed subspace {xi : i ∈ N}. Since X is a
normal space or a countably paracompact space, there is a locally finite family {Vi :
i ∈ N} of open subsets in X such that xi ∈ Vi (see Theorem A.1.11 in Appendix A).
(1) When Y is a q-space, let g be a q-function on Y . Since xi ∈ ∂f −1 (y), we can
select by the inductive method

z1 ∈ V1 ∩ f −1 (g(1, y)) − f −1 (y), and

zi+1 ∈ Vi+1 ∩ f −1 (g(i + 1, y)) − f −1 ({y, f (z1 ), . . . , f (zi )}).

Since f (zi ) ∈ g(i, y), {f (zi )} has an accumulation point. But zi ∈ Vi and f is a
closed mapping, {f (zi )} has no accumulation point, a contradiction.
(2) Suppose Y is a strongly Fréchet–Urysohn space. Since xi ∈ Vi ∩ ∂f −1 (y),

y ∈ f (Vi ) − {y} ⊂ f ( Vj ) − {y}.
ji
60 2 Mappings on Metric Spaces


So there is yi ∈ f ( ji Vj ) − {y} for each i ∈ N such that yi → y. Since
{Vi : i ∈ N} is a locally finite family and f is a closed mapping, there is a
subsequence {yik } of {yi } such that {yik : k ∈ N} is a discrete closed set in Y , a
contradiction. 

Lemma 2.1.15 ([330]) If f : X → Y is an onto mapping, then there is a closed

subspace Z of X, such that, for each y ∈ Y , (f|Z )−1 (y) is either a singleton or the
nonempty set ∂f −1 (y).

Proof For each y ∈ Y , pick py ∈ f −1 (y). Let

Z = ∪{∂f −1 (y) : ∂f −1 (y) = ∅} ∪ {py : ∂f −1 (y) = ∅}.

Since

X − Z = ∪ {f −1 (y)◦ : ∂f −1 (y) = ∅}
∪ {f −1 (y)◦ − {py } : ∂f −1 (y) = ∅},

Z ∈ τ c (X). Obviously, for each y ∈ Y , (f|Z )−1 (y) is either a singleton or the nonempty
set ∂f −1 (y). 

Proposition 2.1.16 ([330]) Let X be a paracompact space. If f : X → Y is a closed

onto mapping, then f is a compact-covering mapping.

Proof For each nonempty compact set K in Y , by Lemma 2.1.14, fK is a peripherally

compact mapping. By Lemma 2.1.15, there is a closed subspace L of f −1 (K) such
that fK|L is a perfect onto mapping, so L ∈ K (X) and f (L) = K. 

Since every Lindelöf regular space is a paracompact space, by Proposition 2.1.16,

every closed onto L-mapping of a regular space is a compact-covering mapping.

Question 2.1.17 ([341]) Characterize the class of spaces Y such that every closed
onto mapping f : X → Y is a countably bi-quotient mapping.

Lemma 2.1.14 derives a general question:

Question 2.1.18 ([460]) Let f : X → Y be a closed mapping. Find additional

conditions for X or Y , such that, with these conditions, ∂f −1 (y) has nice properties
for each y ∈ Y .

Theorem 2.2.2, Lemmas 2.7.20 and 2.7.21, Theorems 3.4.16, 3.8.16 and Corollary
3.8.17 etc. are all associated with this problem.
2.2 Perfect Mappings 61

2.2 Perfect Mappings

The importance of perfect mappings in the mapping theory, same as the role of
compact spaces in general topology, is without rebuke. The main purpose of this
section is to give characterizations of images and preimages of metric spaces under
perfect mappings. To this end, we introduce a series of generalized metric spaces,
such as p-spaces and so on. We also introduce a metrization theorem of adjunction
spaces obtained by Y. Liu and L. Liu [315] and other relevant metrization theorems.

Theorem 2.2.1 ([363, 441]) Perfect mappings preserve metrizability.

Proof Let f : X → Y be a perfect mapping, where X is a metrizable space. In virtue

of the Stone theorem, X is a paracompact space, so Y is a paracompact space (see
Corollary A.1.3 in Appendix A). To prove that Y is a metrizable space, we only need
to prove that Y is a developable space. By Theorem 1.3.5, there is a development
{Un } for X such that Un+1 is a star-refinement of Un . Then,
(1.1) for each nonempty compact set K in X, {st(K, Un )}n∈N is a neighborhood base
of K in X.
For every y ∈ Y and n ∈ N, put

Uy,n = st(f −1 (y), Un ),

Wy,n = Y − f (X − Uy,n ),
Vy,n = f −1 (Wy,n ).

Then f −1 (y) ⊂ Uy,n . It follows that

(1.2) y ∈ Wy,n , and f −1 (y) ⊂ Vy,n ⊂ Uy,n .
(1.3) {Wy,n }n∈N is a local base of y.
Since f is a closed mapping, Wy,n is an open set in Y . If W is an open neigh-
borhood of y, then f −1 (y) ⊂ f −1 (W ). By (1.1), there is n ∈ N such that
st(f −1 (y), Un ) ⊂ f −1 (W ), i.e. Uy,n ⊂ f −1 (W ), so Wy,n ⊂ W . Thus (1.3) is
proved.
By (1.2), f −1 (y) ⊂ Vy,n+1 , and by (1.1), there is m  n + 1 such that
Uy,m ⊂ Vy,n+1 .
(1.4) If y ∈ Wz,m , then Wz,m ⊂ Wy,n .
Fix x ∈ f −1 (y). Since f −1 (y) ⊂ Vz,m ⊂ Uz,m , x ∈ st(f −1 (z), Um ), so there is
Ux ∈ Um such that x ∈ Ux and

∅ = f −1 (z) ∩ Ux ⊂ f −1 (z) ∩ Uy,m ⊂ f −1 (z) ∩ Vy,n+1 ,

and hence z ∈ f (Vy,n+1 ) = Wy,n+1 . As a consequence, we obtain (∗): f −1 (z) ⊂

Vy,n+1 . Below we prove Wz,m ⊂ Wy,n .
Suppose t ∈ Wz,m . Since f −1 (t) ⊂ Uz,m , s ∈ st(f −1 (z), Um ) whenever s ∈
f −1 (t), and hence there is Us ∈ Um such that s ∈ Us and f −1 (z) ∩ Us = ∅.
Take s ∈ f −1 (z) ∩ Us . By (∗), f −1 (z) ⊂ Uy,n+1 , so s ∈ st(f −1 (y), Un+1 ),
62 2 Mappings on Metric Spaces

and hence there is Us ∈ Un+1 such that s ∈ Us and f −1 (y) ∩ Us = ∅. It

follows that s ∈ Us ∩ Us . Take s ∈ f −1 (y) ∩ Us . Since Um refines Un+1 ,
s, s ∈ Us ∪ Us ⊂ st(Us , Un+1 ). Because Un+1 is a star-refinement of Un ,
there is Vs ∈ Un such that

s ∈ st(Us , Un+1 ) ⊂ Vs ⊂ st(f −1 (y), Un ) = Uy,n .

So f −1 (t) ⊂ Uy,n , and hence t ∈ Wy,n . It follows that Wz,m ⊂ Wy,n .

For each n ∈ N, let Wn = {Wy,n }y∈Y . If y ∈ Y and W is a neighborhood of y,
then by (1.3), there is n ∈ N such that Wy,n ⊂ W . By (1.4), there is m ∈ N
such that st(y, Wm ) ⊂ Wy,n . Thus {st(y, Wn )}n∈N is a local base of y, so {Wn }
is a development for Y , and hence Y is a developable space. 

It should be noticed that when f : X → Y is a perfect mapping, for a cover U of

X, {Y −f (X −U) : U ∈ U } may not be a cover of Y . For example, let X = I×{0, 1}.
Take X as a subspace of R2 with the Euclidean topology. Let A = {(0, 0), (0, 1)},
Y = X/A and let f : X → Y be the quotient mapping. Then f is a perfect mapping.
Take U = {I × {0}, I × {1}}. Then U is an open cover of X, and for any U ∈ U ,
f (0, 0) ∈
/ Y − f (X − U).
The closed image S1 × N/({0} × N) of the metric space S1 × N is a copy of the
sequential fan Sω . To make a closed image of a metric spaces being metrizable, some
additional conditions are necessary.

Theorem 2.2.2 (The Hanai–Morita–Stone theorem) If f : X → Y is a closed

mapping and X is a metric space, then the following are equivalent:
(1) Y is a metrizable space.
(2) Y is a strongly Fréchet–Urysohn space [336].
(3) f is a peripherally compact mapping [363, 441].
(4) f is a countably bi-quotient mapping [336].

Proof By Theorem 2.2.1, Lemmas 2.1.14, 2.1.15 and Proposition 2.1.12 (the map-
ping lemma), we only need to prove that countably bi-quotient mappings preserve the
 Fréchet–Urysohn property. Let {An } be a decreasing
strongly  sequence of sets in Y .
If y ∈ n∈N An , then there is x ∈ f −1 (y) such that x ∈ n∈N f −1 (An ). Because other-

wise, f −1 (y) ⊂ n∈N (X −f −1 (An )), so there is m ∈ N such that y ∈ f (X −f −1 (Am ))◦ ,
and hence Am ∩ f (X − f −1 (Am )) = ∅, a contradiction. Thus there is xn ∈ f −1 (An )
such that xn → x, so f (xn ) ∈ An and f (xn ) → y. 

The above result is a deformation of the Hanai–Morita–Stone theorem. In fact, the

Hanai–Morita–Stone theorem can be described as follows: if f : X → Y is a closed
mapping and X is a metric space, then Y is a metric space if and only if Y is a first
countable space, if and only if f is a peripherally compact mapping. Its prototype is
a result of Vaı̌nšteı̌n [471]: every closed mapping from a metric space onto a metric
space is a peripherally compact mapping. The Hanai–Morita–Stone theorem is also
called the Morita–Hanai–Stone theorem.
2.2 Perfect Mappings 63

Assume that X and Y are two disjoint spaces. Let A be a closed subset of X and
let f : A → Y be a continuous mapping. The quotient space Z of X ⊕ Y obtained
by mapping x and f (x) (∀x ∈ A) to one point is called the adjunction space and
is often denoted as X ∪f Y . The quotient mapping p : X ⊕ Y → Z is called the
adjunction mapping [69]. The following metrization theorem of adjunction spaces
can be regarded as an application of Theorem 2.2.2.
a natural mapping (or an obvious mapping). Let {Xα }α∈Λ be
Recall the concept of
a cover of X. Put Z = α∈Λ Xα . For each α ∈ Λ, denote the homeomorphism from
the subspace Xα of Z onto the subspace Xα of X as hα . Then h : Z → X is called a
natural mapping (or an obvious mapping) if h|Xα = hα for each α ∈ Λ.

Theorem 2.2.3 ([67, 315]) If X and Y are metric spaces, then X ∪f Y is a metrizable
space if and only if X ∪f Y is a first countable space.

Proof Let Z = X ∪f Y . Assume that Z is a first countable space. By the mapping

lemma (see Proposition 2.1.12), p is a pseudo-open mapping. Let X1 = clX (X − A),
Z1 = clZ p(X − A).
(3.1) If B is a closed set of X1 , then f (B) is a closed set of Y .
Suppose y ∈ Y − f (B). Then there are disjoint open sets U and V in X
containing B and f −1 (y) respectively. Let W = V ∪ Y , G = p(W )◦ . Then W
is a neighborhood of p−1 (p(y)) = {y} ∪ f −1 (y) in X ⊕ Y , so p(y) ∈ G, and
hence p−1 (G) ∩ (U − A) = ∅. Since B ⊂ clX (U − A), p−1 (G) ∩ f (B) = ∅.
Thus f (B) is a closed set in Y .
(3.2) p1 = p|X1 : X1 → Z1 is a closed mapping.
It is easy to verify that Z1 = p(X − A) ∪ {y ∈ Y : f −1 (y) ∩ ∂A = ∅}, and
hence p(X1 ) = Z1 . By (3.1) and p1|X−A = idX−A , p1 is a closed mapping.
(3.3) X ∪f Y is a metrizable space.
By (3.2) and Theorem 2.2.2, Z1 is a metrizable closed subspace of X ∪f Y .
Since p(Y ) is homeomorphic to Y , p(Y ) is a metrizable closed subspace of
X ∪f Y . Let q : p(Y ) ⊕ Z1 → X ∪f Y be the natural mapping. Then q is a
perfect mapping. By Theorem 2.2.1, X ∪f Y is a metrizable space. 

Example 2.2.4 ([315]) A non-metrizable adjunction space.

(1) Let X = I2 − {(0, 0), (1, 0)} and Y = I. Give X and Y the Euclidean subspace
topology. Let A = {(x, 0) : 0 < x < 1}, and define f : A → Y by f (x, 0) = x.
Then f −1 (y) is a compact set for every y ∈ Y . Since f (A) = (0, 1) is not a closed
set in Y , by (3.1) in the proof of Theorem 2.2.3, X ∪f Y is not a first countable
space.
(2) The definitions of X and A are same as that in (1). Take Y = {0}. Define f : A → Y
by f (A) = {0}. By using the notions in the proof of Theorem 2.2.3, we have
X1 = X, Z1 = Z and p1 : X1 → Z1 is a closed mapping. Since ∂p−1 1 (0) = A is
not compact in X1 , Z1 is not a metrizable space.

Below we introduce a characterization of perfect preimages of metric spaces.

64 2 Mappings on Metric Spaces

Lemma 2.2.5 ([162, 469]) If {Un } be a sequence of open covers of a space X such
that Un+1 star-refines Un , then there is a pseudo-distance d on X such that

(1) for any y ∈ X, y ∈ n∈N st(x, Un ) if and only if d(x, y) = 0;
(2) U is an open subset of (X, d) if and only if for each x ∈ U, there is m ∈ N such
that st(x, Um ) ⊂ U.

Proof Denote U0 = {X}. For every x, y ∈ X, define

D(x, y) = inf{2−n : y ∈ st(x, Un )},

 n 

d(x, y) = inf D(xi , xi+1 ) : n ∈ N, xi ∈ X, x0 = x and xn+1 = y .
i=0

Then d is a pseudo-distance on X, and by using the inductive method, we know that,

whenever n  2,


n−1
D(x, y)  2D(x, x1 ) + 4 D(xi , xi+1 ) + 2D(xn , y).
i=1

So D(x, y)/4  d(x, y)  D(x, y), and hence (1) holds. Since for every x ∈ X and
n ∈ N,
st(x, Un+2 ) = B(x, 1/2n+2 ) ⊂ B(x, 1/2n ) ⊂ st(x, Un ),

(2) holds. 

Theorem 2.2.6 [360] (The Morita theorem) A space X is a quasi-perfect preimage

of a metric space if and only if X is an M-space.

Proof The necessity. Let f : X → Y be a quasi-perfect mapping, where Y is a metric

space. Then there is a development {Un } for Y such that Un+1 star-refines Un . For
each n ∈ N, define Fn = f −1 (Un ). Then Fn+1 star-refines Fn . If x ∈ X and {xn }
is any sequence in X with xn ∈ st(x, Fn ), then f (xn ) ∈ st(f (x), Un ), and hence
f (xn ) → f (x). Since f is a quasi-perfect mapping, {xn } has an accumulation point.
Thus {Fn } is an M-sequence in X, and hence X is an M-space.
The sufficiency. Let {Un } be an M-sequence in X. By Lemma 2.2.5, there is a
pseudo-distance d on X such that

(6.1) for each y ∈ X, y ∈ n∈N st(x, Un ) if and only if d(x, y) = 0;
(6.2) U is an open set in (X, d) if and only if for each x ∈ U, there is m ∈ N such
that st(x, Um ) ⊂ U.
Define an equivalence relation “∼” on X as follows: for every x, y ∈ X, x ∼ y
if and only if d(x, y) = 0. Let Y = X/∼ be the quotient set, and define
ρ : Y × Y → R+ by ρ([x], [y]) = d(x, y), then (Y , ρ) is a metric space.
We prove that f : X → Y is a quasi-perfect mapping. Since f −1 (Bρ ([x], ε)) =
Bd (x, ε) ∈ τ (X) for every x ∈ X and ε > 0, f is a continuous function.
2.2 Perfect Mappings 65

 we prove that {st(x, Un )}n∈N is a neighborhood base of 2the set {x} =

First,
n∈N st(x, U n ) in X. In fact, since Un+1 star-refines Un , st (x, Un+1 ) ⊂
st(x, Un ), so st(x, Un+1 ) ⊂ st(x, Un ). By the convergence lemma, {st(x,
Un )}n∈N is a neighborhood base of the countably compact closed set [x] in
X. Now, we can prove f is a quasi-perfect mapping. In fact, on the one
hand, f −1 ([x]) = [x] is a countably compact set in X. On the other hand,
if x ∈ / f −1 (f (H)) for any H ∈ τ c (X), then [x] ∩ H = ∅, and hence
there is m ∈ N such that st(x, Um ) ∩ H = ∅. If y ∈ st(x, Um+1 ), then
st(y, Um+1 ) ⊂ st(x, Um ), thus [y] ∩ H = ∅, so st(x, U m+1 ) ∩ f −1 (f (H)) = ∅,
and hence f −1 (f (H)) ∈ τ c (X). It follows that f (H) ∈ τ c (Y ). Therefore f is a
quasi-perfect mapping. 

By the properties of paracompact spaces, we obtain the following corollaries.

Corollary 2.2.7 ([360]) A space X is a perfect preimage of a metric space if and

only if X is a paracompact M-space.

Corollary 2.2.8 ([368]) Every paracompact M-space is a strong Σ-space.

Proof If X is a paracompact M-space, then there is a metric space Y and a perfect

mapping f : X → Y . Let B be a σ -locally finite base for Y and let K = {f −1 (y) :
y ∈ Y }. Then f −1 (B)− is a σ -locally finite closed (mod k)-network w.r.t. K in X.
So X is a strong Σ-space. 

M-spaces and paracompact M-spaces characterize the quasi-perfect preimages

and the perfect preimages of metric spaces respectively. These spaces are just the
spaces sought in Alexandroff’s questions (see Question 2.0.2). In the second part of
this section, we give characterizations of these spaces and discuss the metrization
problems of them.

Definition 2.2.9 ([17]) A space X is called a submetrizable space if X is the preimage

of a metric space under a one-to-one mapping.

Obviously, X is a submetrizable space if and only if there exists a metrizable

topology on X which is coarser than the topology on X. Every submetrizable space is
a σ  -space with a G ∗δ -diagonal. Submetrizability is additive, hereditary and countably
productive.

Proposition 2.2.10 ([320]) A space X is a submetrizable space if and only if X has

a G δ -diagonal sequence {Un } such that Un+1 star-refines Un .

Proof If X is a submetrizable space, then there exist a metric space Y and a one-to-
one mapping f : X → Y . Let {Fn } be a development for Y such that Fn+1 star-refines
Fn . For each n ∈ N, define Un = f −1 (Fn ). Then {Un } is a G δ -diagonal sequence in
X and Un+1 star-refines Un .
66 2 Mappings on Metric Spaces

Conversely, there is a pseudo-distance d on X satisfying the conditions (1) and

(2) of Lemma 2.2.5. Since {Un } is a G δ -diagonal sequence in X, for every x, y ∈ X,
d(x, y) = 0 if and only if x = y. So d is a distance on X. Because each Bd (x, ε) is an
open set in X, the metrizable topology derived by d is coarser than the topology for
X. 

Corollary 2.2.11 ([65]) Every paracompact space with a G δ -diagonal is a sub-

metrizable space.

Theorem 2.2.12 ([93]) A space X is a metrizable space if and only if X is an M-space

with a G δ -diagonal.

Proof We only need to prove the sufficiency. By the Morita theorem and Chaber
theorem, there exist a metric space Z and a perfect mapping g : X → Z, and hence
X is a paracompact space. By Corollary 2.2.11, there exist a metric space Y and a
one-to-one mapping f : X → Y . By Corollary 2.1.9, X is a metrizable space. 

The above theorem can be restated as follows: a perfect (or quasi-perfect) preimage
of a metric space is metrizable if and only if it has a G δ -diagonal. For the sake of
convenience, we say that a topological property Φ satisfies the perfect preimage
G δ -diagonal theorem provided that if f : X → Y is a perfect mapping, X is a regular
space with a G δ -diagonal and Y has the property Φ, then X also has Φ.

Example 2.2.13 A preimage of I under a finite-to-one closed mapping may have no

G δ -diagonal, the Alexandroff double-arrow space is such an example.
Suppose X is the Alexandroff double-arrow space (see Example 1.8.9 and we still
use the notations in Example 1.8.9). Let f : X → I be the projection mapping, where
I has the Euclidean topology. Then f is an at most two-to-one closed mapping. It is
easy to verify that X has no G δ -diagonal.
Because of the above example, when discussing the problem whether a perfect
preimage of a space in the class C belongs to C , where C is a class of gener-
alized metric spaces, generally the condition that “having a G δ -diagonal” should
be attached. This is the origin of “the perfect preimage G δ -diagonal theorem”. Of
course, we also can discuss so called “the perfect preimage σ  -theorem” and so on.
Since the situation is similar, we shall not make the repeat.
In the last part of this section, we give some characterizations of paracompact
M-spaces.

Definition 2.2.14 A regular space X is called a p-space [25] if there is a sequence

{Un } of families of open sets in βX such that
(1) each Un covers X;
(2) for each x ∈ X, n∈N st(x, Un ) ⊂ X.
If we also have  
(3) for each x ∈ X, n∈N st(x, Un ) = n∈N st(x, Un ),
then X is called a strict p-space [31]. The sequence {Un } above is said to be a
pluming or a strict pluming in X.
2.2 Perfect Mappings 67

Since every locally compact space is an open set in its Čech–Stone compactifi-
cation, every locally compact space is a p-space. Some internal characterizations of
strict p-spaces and p-spaces make them more convenient to be used.

Proposition 2.2.15 ([91]) A completely regular space X is a strict p-space if and

only if there is a strict p-sequence in X, i.e. there is a sequence {Un } of open covers
of X such that

(1) for each x ∈ X, Cx = n∈N st(x, Un ) is a compact subset of X;
(2) for each x ∈ X, {st(x, Un )}n∈N is a neighborhood base of Cx in X.

Proof The necessity. Let {Fn } be a strict pluming on X. We may assume that Fn+1
partially refines Fn . For each n ∈ N, put Un = Fn|X . We prove the sequence {Un }
of open covers of X is a strict p-sequence in X.
 
(1) For each x ∈ X, Cx = n∈N st(x, Fn ) = n∈N st(x, Fn ) is a closed subset of
βX contained in X, so Cx is a compact set in X.
(2) For each x ∈ X and Cx ⊂ U ∈ τ (X), there is G ∈ τ (βX) such that G ∩ X = U.
Then {G} ∪ {βX − st(x, Fn ) : n ∈ N} covers the compact space βX, so there is
m ∈ N such that st(x, Fm ) ⊂ G, and hence st(x, Um ) ⊂ G, thus {st(x, Un )}n∈N
is a neighborhood base of Cx in X.
The sufficiency. Let {Un } be a strict p-sequence in X. For each n ∈ N, take
Fn ⊂ τ (βX) such that Fn|X = Un . Then Fn is a cover of X and for each x ∈ X,
   
n∈N st(x, Un ) ⊂ n∈N st(x, Fn ) ⊂ n∈N st(x, Fn ). If y ∈ n∈N st(x, Fn ) − Cx ,
pick W ∈ τ (βX) such that y ∈ W and W ∩ Cx = ∅, then there is m ∈ N such
that st(x, Um ) ∩ W = ∅,so st(x, Fm ) ∩ W = ∅, and hence  y ∈/ st(x, Fm ), a
contradiction. Therefore, n∈N st(x, Fn ) ⊂ Cx . Consequently, n∈N st(x, Fn ) =

n∈N st(x, Fn ) ⊂ X. 

Thus, every space with a strict p-sequence is a wΔ-space. Obviously, the property
of having a strict p-sequence is countably productive [158].

Theorem 2.2.16 ([77]) A completely regular space X is a p-space if and only if there
is a p-sequence in X, i.e. there is a sequence {Un } of open covers of X such that for
every x ∈ X and n ∈ N, if x ∈ Un ∈ Un , then

(1) D x = n∈N U n is a compact set in X;
(2) { nk U n }k∈N is a network of Dx in X.

Proof The necessity. Let {Pn } be a pluming in X. Choose a sequence {Un } of open
covers of X satisfying that for every n ∈ N and U ∈ Un , there is P ∈ Pn such
that clβX U ⊂ P. We prove that {Un } is a p-sequence in X. Foreach x ∈ X and any
sequence {Un } of open sets in X with x ∈ Un ∈ Un , put Dx = n∈N U n .
68 2 Mappings on Metric Spaces

 
(1) Since n∈N clβX Un ⊂ n∈N st(x, Pn ) ⊂ X,

Dx = (X ∩ clβX Un ) = clβX Un
n∈N n∈N

is a compact set in X.
(2) Let Dx ⊂ U ∈ τ (X). Take G ∈ τ (βX) such that U = G  ∩ X. Then {G} ∪ {βX −
clβX Un : n ∈ N} covers βX, so there is k ∈ N such that nk clβX Un ⊂ G, and
 
hence nk U n ⊂ U. Thus, { nk U n }k∈N is a network of Dx in X.
The sufficiency. Let {Un } be a p-sequence in X. For each n ∈ 
N, take Pn ⊂ τ (βX)
such that Pn|X = Un . Then Pn covers X. Let x ∈ X. If y ∈ n∈N st(x, Pn ) − X,
 there is a sequence {Pn } of subsets of βX such that {x, y} ⊂ Pn ∈ Pn , and hence
then
n∈N Pn ∩ X is a compact set in X, so there exist G ∈ τ (βX) and k ∈ N such that
nk Pn ∩ X ⊂ G ⊂ clβX G ⊂ βX − {y}. Let

⎛ ⎞

W =⎝ Pn ⎠ ∩ (βX − clβX G).
nk

Then W ∩ X = ∅. However, y ∈ W ∈ τ (βX), so W ∩ X = ∅, a contradiction. Thus


n∈N st(x, Pn ) ⊂ X. Therefore, {Pn } is a pluming in X. 

Obviously, the property of having a p-sequence is countably productive [158].

A space X is called an isocompact space [49] if every countably compact closed
set in X is a compact set. Every submetacompact space is an isocompact space (see
Theorem A.4.2 in Appendix A).

Corollary 2.2.17 Let X be an isocompact regular space. If X is a wΔ-space, then

X has a p-sequence.

Proof For the wΔ-sequence {Un } in X, take a sequence {Vn } of open covers of X
such that Vn refines Un . By the convergence lemma, {Vn } is a p-sequence in X. 

Theorem 2.2.18 ([77, 156]) For every submetacompact regular space X, the fol-
lowing are equivalent:
(1) X has a strict p-sequence.
(2) X has a p-sequence.
(3) X is a wΔ-space.

Proof We only need to prove (2) ⇒ (3) ⇒ (1).

(2) ⇒ (3). Let {Un } be a p-sequence in X, where Un+1 refines
 Un . For each n ∈ N,
take a θ -sequence {Vn,j }j∈N in X such that Vn,j refines Un ∧ ( m,i<n Vm,i ). For each
k ∈ N, let Vk = Vk,1 . We prove {Vk } is a wΔ-sequence in X. Suppose x ∈ X and
{xk } is a sequence in X satisfying xk ∈ st(x, Vk ). Then there are strictly increasing
2.2 Perfect Mappings 69

sequences {ki }, {ji } in N such that ki+1 > ji and 1  |(Vki ,ji )x | < ℵ0 . For each n ∈ N,
since {xki : i > n} ⊂ st(x, Vkn+1 ) ⊂ st(x, Vkn ,jn ) and Vkn ,jn refines Un , we may assume
that there is Un ∈ Un such that {xki : i > n} ∪ {x} ⊂ Un , so {xki } has an accumulation
point, and hence {xk } has an accumulation point.
(3) ⇒ (1). Let {Un } be a wΔ-sequence in X. By Lemma 1.4.8, there is a sequence
 n } of open covers
{V  of X such that Vn+1 refines Vn ∧ Un , and for each x ∈ X,
n∈N st(x, V n ) = n∈N st(x, Vn ) ⊂ n∈N st(x, Un ). By the convergence lemma,
{Vn } is a strict p-sequence in X. 

Corollary 2.2.19 For every paracompact space X, the following are equivalent:
(1) X is an M-space.
(2) X is a wΔ-space.
(3) X is a p-space.

By Corollaries 2.2.19, 2.2.8 and Theorem 1.5.11, paracompact p-spaces and para-
compact M-spaces are countably productive [25, 360].

Corollary 2.2.20 ([25]) Every cosmic p-space is metrizable.

Proof Let X be a cosmic p-space. Then X is a paracompact p-space with a G δ -

diagonal. It follows from Corollary 2.2.19 and Theorem 2.2.12 that X is
metrizable. 

An earlier version of the above corollary is a result proved by Arhangel’skiı̌ [21]

that every cosmic compact space is metrizable.
A Tychonoff space X is said to be Čech-complete if it is a G δ -set in some Hausdorff
compactification of it. It is easy to see that a Tychonoff space X is Čech-complete if
and only if it is a G δ -set in every Hausdorff compactification of it. A space is a Čech-
complete paracompact space if and only if it can be mapping by a perfect map onto
a complete metric space [135]. The problem of characterizing the perfect preimages
of metric spaces was posed by Alexandroff [3]. To solve this problem, Arhangel’skiı̌
[25] introduced the concept of p-spaces and proved the following theorem.

Theorem 2.2.21 [25] (The Arhangel’skiı̌ theorem) A space is a paracompact p-

space if and only if it can be mapped onto a metrizable space by a perfect mapping.

Proof Firstly, if X is a Tychonoff space, Y is a p-space and f : X → Y is a perfect

mapping, then X is also a p-space. Indeed, let {Un } be a pluming of the space Y
in the Čech–Stone compactification βY . Since f : X → Y is perfect, the extension
F : βX → βY of the mapping f satisfies the condition F(βX−X) ⊂ βY −Y . For each
n ∈ N, let Vn =F −1 (Un ). Then {Vn } is a pluming of the space 
X in βX. In fact, for
each x ∈ X, F( n∈N st(x, Vn )) ⊂ n∈N st(f (x), Un ) ⊂ Y , thus n∈N st(x, Vn ) ⊂ X.
The scheme of the proof of the main part of the theorem is as follows. Let {Un }
be a pluming of a paracompact space X in the Čech–Stone compactification βX. We
construct by induction a sequence {Vn } of open covers of X satisfying the condi-
tions: (1) for each x ∈ X and each n ∈ N the closure clX (st(x, Vn )) is contained in
70 2 Mappings on Metric Spaces

some element of Un ; (2) Vn+1 is a refinement of Vn ; (3) 

Vn+1 is a star refinement
of Vn . Points x, y ∈ X are said to be equivalent if y ∈ n∈N st(x, Un ). Then the
canonical mapping from X to the space Y of equivalent classes corresponding to
the above equivalence is the required perfect mapping from X onto the metrizable
space Y . 
This theorem generalizes the characterization of Čech-complete paracompactness
given by Frolík [135]. It shows that the p-space property is a “genuine” generalization
of Čech-completeness.
Example 2.2.22 (1) There is a locally compact space (hence a p-space) which is
not isocompact, for example ω1 .
(2) There is a p-space which is not locally compact, for example the space P of
irrationals.
(3) These is a non-regular developable space without any p-sequence [156].
Let S = {(x, y) : x, y ∈ R, y > 0}, L = {(x, 0) : x ∈ R} and X = S ∪ L. Denote
the Euclidean subspace topology on X by τ ∗ . Define the half-disc topology on
X as follows [437]:

τ = {τ ∗ } ∪ {{x} ∪ (S ∩ U) : x ∈ L, x ∈ U ∈ τ ∗ }.

Then (X, τ ) is called a half-disc topological space. It is easy to verify that X is

not a regular space, and by using spherical neighborhoods of points in R2 , it is
also easy to show X is a developable space.
If X has a p-sequence {Un }, then for x = (0, 0) ∈ X and each n ∈ N, fix
Un ∈ (Un )x and take a neighborhood {x} ∪ (S ∩ B(x, 2rn )) of x contained in
U, where B(x, 2rn ) is a spherical
 neighborhood of x in R and rn+1 < rn .
2

Let xn = (rn , 0). Then xn ∈ kn U k , and hence the sequence {xn } has an
accumulation point, which contradicts the fact that {xn : n ∈ N} is a discrete
closed subset of X. Consequently, X has no p-sequence.
(4) There is a locally compact, submetrizable space which is not a β-space [162].
Let B be the Bernstein set in the real line R (see Example 1.8.5). Then every
uncountable closed set in R meets both B and R − B. Let {Bα : α < 2ω } be the
family of all the countable sets in B the closures of which in R are uncountable.
For each α < 2ω , by the transfinite induction, pick xα ∈ Bα − (B ∪ {xβ : β < α})
and xα,n ∈ Bα such that xα,n → xα . Let X = B∪{xα : α < 2ω }. Define a topology
for X as follows: each point in B is isolated; the elements of a neighborhood base
of xα have the form {xα } ∪ {xα,n : n  m}, m ∈ N.
Obviously, X is a locally compact submetrizable space. Let H = {xα : α < 2ω }.
Then H is a closed set in X. If X is a β-space, then by Theorem 1.7.7, X is a
semi-stratifiable
 space, and hence X has a sequence {Un } of open sets such that
H = n∈N Un . If some B − Un is uncountable, then there is α < 2ω such that
Bα ⊂ B − Un , and which contradicts the fact that  xα,n → xα ∈ Un , thus every
B−Un is countable. Since B is uncountable, B∩( n∈N Un ) = ∅, a contradiction.
(5) p-spaces and M-spaces are independent.
The space X in the above (4) is a p-space which is not an M-space. On the other
2.2 Perfect Mappings 71

hand, let X be the Frolík space [134], i.e. N ⊂ X ⊂ βN and X is a countably

compact subspace of βN with cardinality not greater than c. Then X is an M-
space. Since every compact set in X is a finite set, X is not a k-space, hence X is
not a p-space (see Propositions 2.4.10 and 2.4.11).

Question 2.2.23 ([162]) Is every normal Moore space submetrizable?

Question 2.2.24 ([158]) Is the product of a strict p-space and a wΔ-space a wΔ-
space?

2.3 Quotient Mappings

By taking quotient mappings as its core, in this section, we give the characterizations
of quotient images, pseudo-open images and countably bi-quotient images of metric
spaces, and show that they are just sequential spaces, Fréchet–Urysohn spaces and
strongly Fréchet–Urysohn space respectively. As applications, we give the character-
izations of quotient images of normally metric spaces and connected metric spaces.
This topic relies on the mapping properties of generalized sequentiality properties.

Proposition 2.3.1 (1) k-spaces and sequential spaces are preserved by quotient
mappings [133, 228].
(2) Fréchet–Urysohn spaces are preserved by pseudo-open mappings [26, 133].
(3) Strongly Fréchet–Urysohn spaces are preserved by countably bi-quotient map-
pings [424].

Proof Let f : X → Y be a mapping.

(1) Suppose f is a quotient mapping and X is a k-space. If a subset A of Y satisfies that
for each K ∈ K (Y ), A∩K ∈ τ c (Y ), then for each L ∈ K (X), A∩f (L) ∈ τ c (Y ),
so f −1 (A) ∩ L = f −1 (A ∩ f (L)) ∩ L ∈ τ c (X), and hence f −1 (A) ∈ τ c (X), thus
A ∈ τ c (Y ). Consequently, Y is a k-space.
In the above argument, actually we proved the following stronger result: Let A
be any subset of Y . If A ∩ f (L) ∈ τ c (Y ) for each L ∈ K (X), then A ∈ τ c (Y ).
Replace K (Y ) and K (X) in the above proof with S (Y ) and S (X) respectively,
we obtain the result that sequential spaces are preserved by quotient mappings.
(2) Suppose f is a pseudo-open mapping and X is a Fréchet–Urysohn space. If
y ∈ A ⊂ Y and f −1 (y) ∩ f −1 (A) = ∅, then y ∈ f (X − f −1 (A))◦ ⊂ Y − A, a
contradiction. So there is x ∈ f −1 (y) ∩ f −1 (A), it follows that there is a sequence
{xn } ⊂ f −1 (A) such that xn → x, thus {f (xn )} ⊂ A and f (xn ) → y. Consequently,
Y is a Fréchet–Urysohn space.
(3) was proved in Theorem 2.2.2. 

Definition 2.3.2 Let F be a cover of a space X. We say that X has the weak topology
with respect to F [117], or say that X is determined by F [167] provided that for
any A ⊂ X, A ∈ τ c (X) if and only if for every F ∈ F , A ∩ F ∈ τ c (F).
72 2 Mappings on Metric Spaces

Obviously, X is a k-space (resp. sequential space) if and only if X has the weak
topology with respect to K (X) (resp. S (X)).

Proposition 2.3.3 ([167]) Suppose F is a cover of X and Z = F . Let f : Z → X
be the natural mapping. Then f is a quotient mapping if and only if X has the weak
topology with respect to F .

Proof Suppose f is a quotient mapping. For each A ⊂ X, if A ∩ F ∈ τ c (F) whenever

F ∈ F , then f −1 (A) ∈ τ c (Z), and hence A ∈ τ c (X), thus X has the weak topology
with respect to F . If X has the weak topology with respect to F and A ⊂ X satisfies
f −1 (A) ∈ τ c (Z), then for each F ∈ F , A ∩ F ∈ τ c (F), and hence A ∈ τ c (X), so f is
a quotient mapping. 

Theorem 2.3.4 ([136]) For every space X, the following are equivalent:
(1) X is a k-space.
(2) X is a quotient image of a paracompact locally compact space.
(3) X is a quotient image of a locally compact space.

Proof (1) ⇒ (2). If X is a k-space, then X has the weak topology with respect
to K (X). By Proposition
 2.3.3, X is a quotient image of the paracompact locally
compact space K (X).
(2) ⇒ (3) is obvious.
(3) ⇒ (1). We only need to prove that every locally compact space is a k-space.
Suppose Y is a locally compact space and A ⊂ Y . If A ∈/ τ c (Y ), then there is y ∈ A−A.
Let V be an open neighborhood of y such that V ∈ K (Y ). Then y ∈ V ∩ A − V ∩ A,
so V ∩ A ∈/ τ c (Y ), and hence Y is a k-space. 

Corollary 2.3.5 Let X be a k-space. If every compact subset of X is of the point

G δ -property, then X is a sequential space.

Proof By Theorem 1.7.7, K (X) is a first countable space. Since X is a quotient
image of K (X), by Proposition 2.3.1, X is a sequential space. 

Replace K (X) in the proof of Theorem 2.3.4 with S (X), we get some charac-
terizations of quotient images of metric spaces.

Theorem 2.3.6 ([133]) For every space X, the following are equivalent:
(1) X is a sequential space.
(2) X is a quotient image of a locally compact metric space.
(3) X is a quotient image of a metric space.

By using the mapping lemma (see Proposition 2.1.12), we obtain the following
characterizations of pseudo-open images and countably bi-quotient images of metric
spaces.

Theorem 2.3.7 ([26, 133]) For every space X, the following are equivalent:
2.3 Quotient Mappings 73

(1) X is a Fréchet–Urysohn space.

(2) X is a pseudo-open image of a locally compact metric space.
(3) X is a pseudo-open image of a metric space.

Theorem 2.3.8 ([424]) For every space X, the following are equivalent:
(1) X is a strongly Fréchet–Urysohn space.
(2) X is a countably bi-quotient image of a locally compact metric space.
(3) X is a countably bi-quotient image of a metric space.

In the second part of this section, we investigate quotient images of two special
classes of spaces. We first introduce one special class of metric spaces, the quotient
images of spaces in this class are metrizable spaces.

Definition 2.3.9 ([367]) A metric space (X, d) is call a normally metric space, if
X d is compact in X.
The normally metric space X defined by Mrówka [367] is that there is a metric d
on X such that for each pair A, B of disjoint closed sets in X, d(A, B) > 0. Mrówka
proved that this definition is equivalent to Definition 2.3.9. We do not use this fact
in this book.

Theorem 2.3.10 ([11, 210]) For every metric space X, the following are equivalent:
(1) X is a normally metric space.
(2) Every closed image of X is a metric space.
(3) Every quotient image of X is a metric space.

Proof (1) ⇒ (3). Let f : X → Y be a quotient mapping. Denote a countable dense

subset of the compact metric space X d by D = {ai : i ∈ N}. Define

B = {B(ai , 1/j) : i, j ∈ N}, U = B F .

Then U is a countable family in X. For each U ∈ τ (X), let

H(U) = (Y − f (X d − U)) ∩ f (U),

G 1 (U) = (X − f −1 (f (X d − U))) ∩ (f −1 (f (U)) − U),
G 2 (U) = (X − f −1 (f (X d − U))) ∩ U.

Then f −1 (H(U)) = G 1 (U) ∪ G 2 (U). Since G 1 (U) ⊂ X − X d , G 2 (U) ∈ τ (X) and

H(U) ∈ τ (Y ). Let

W = {H(U) : U ∈ U } ∪ {{y} : y ∈ Y − f (X d )}.

To prove Y is a metric space, we show that Y is a paracompact space and W is a

σ -locally finite base for Y .
74 2 Mappings on Metric Spaces

(10.1) Y is a paracompact space.

For any open cover V of Y , there is V ∈ V <ω such that f (X d ) ⊂ ∪V . Then
V ∪ {{y} : y ∈ Y − ∪V } is a locally finite open refinement of V , and hence
Y is a paracompact space.
(10.2) W is a base for Y .
For every y ∈ Y and y ∈ V ∈ τ (Y ), we may assume that there is x ∈ X d such
that y = f (x). Then there is U ∈ U such that X d ∩f −1 (f (x)) ⊂ U ⊂ f −1 (V ),
so y ∈ H(U) ⊂ V .
(10.3) W is a σ -locally finite base.
Denote the set of all the finite subfamilies of {H(U) : U ∈ U } covering
f (X d ) by {Pi }i∈N , because U is countable. For each i ∈ N, let

Wi = Pi ∪ {{y} : y ∈ Y − ∪Pi }.

Then W = i∈N Wi and Wi is locally finite.
(3) ⇒ (2) is obvious. Below we prove (2) ⇒ (1). If X d ∈/ K (X), then X d contains
a countable discrete closed subspace Z. Let f : X → X/Z be the quotient mapping.
Then f is a closed mapping and ∂f −1 ([Z]) = Z is not a compact set in X. By Theorem
2.2.2, X/Z is not a metric space, a contradiction. Thus X d ∈ K (X). 

Corollary 2.3.11 ([476]) A space X is a compact metric space if and only if each
image of X is a metric space.

Proof We only need to prove the sufficiency. By Theorem 2.3.10, X is a normally

metric space. If X is not a compact space, then X contains a countable discrete
closed subspace Z ⊂ X − X d . Let Y = X − Z. Take τ1 = τ (X)|Y and take a non-
metrizable Hausdorff topology on Z as τ2 . Since (Z, τ1|Z ) is a clopen subspace of X,
idX : X → (Y , τ1 ) ⊕ (Z, τ2 ) is a continuous mapping. However, (Y , τ1 ) ⊕ (Z, τ2 ) is
not a metric space, a contradiction. Consequently, X is a compact metric space. 

Corollary 2.3.12 ([210]) For every metric space X, the following are equivalent:
(1) Every quotient image of X is a metric space.
(2) Every quotient image of X is a first countable space.
(3) Every quotient image of X is a Fréchet–Urysohn space.
(4) Every quotient mapping on X is a countably bi-quotient mapping.
(5) Every quotient mapping on X is a pseudo-open mapping.

Proof (1) ⇒ (2) ⇒ (4) ⇒ (5) ⇒ (3) is obvious. We only need to prove (3) ⇒ (1).
If X d ∈
/ K (X), then X d contains a countable discrete closed subspace {xn : n ∈ ω}.
Take a discrete family {Un : n ∈ ω} of open sets in X such that xn ∈ Un . To each
n ∈ ω, take a sequence {xn,m }m ⊂ Un such that xn,m → xn . Let Y = X − {xn : n ∈ N}.
Define a mapping f : X → Y by

x0,n , x = xn , n ∈ N,
f (x) =
x, x ∈ Y .
2.3 Quotient Mappings 75

Give Y the quotient topology induced by f . Then the Hausdorff space Y contains a
copy of Arens space S2 , and hence Y is not a Fréchet–Urysohn space, a contradiction.
Consequently, X d ∈ K (X). By Theorem 2.3.10, every quotient image of X is a metric
space. 
Example 2.3.13 ([272]) There exist a metric space X and a quotient mapping f :
X → Y such that f is neither a closed mapping nor an open mapping.
Suppose X = I × ω and B is a countable base of the Euclidean topology for I.
Let

V (B, m) = B × ({0} ∪ {n ∈ N : n  m}), B ∈ B, m ∈ N;

P = {{x} : x ∈ I × N} ∪ {V (B, m) : B ∈ B, m ∈ N}.

Give X the topology generated by the base P. Then X is a regular space and P is a
σ -discrete base for X, and hence X is a metrizable space. Since the topology for X d
is the Euclidean subspace topology for I, X d is a compact set in X, and hence X is a
normally metric space. Let f : X → I be the projection mapping. Then the quotient
mapping f is neither a closed mapping nor an open mapping.
Example 2.3.13 shows that “pseudo-open mapping” and “countably bi-quotient
mapping” in Corollary 2.3.12 cannot be strengthened to “closed mapping” and “open
mapping” respectively. Example 2.3.14 below shows that “closed mapping” or “quo-
tient mapping” in Theorem 2.3.10 cannot be replaced with “countably bi-quotient
mapping” either.
Example 2.3.14 ([272]) There is a metric space which is not a normally metric space
such that every countably bi-quotient image
 of this space is a metric space.
For each n ∈ N, let In = I. Put X = n∈N In . Then the metric space X is not a
normally metric space. Let f : X → Y be a countably bi-quotient mapping. Then Y
is a Lindelöf, locally compact and locally metrizable space, hence a metric space.
Question 2.3.15 Characterize the class of spaces such that every countably bi-
quotient image of these spaces is metrizable.
Another special class of spaces is the class of connected sequential spaces. Con-
nectedness is invariant under mappings. Is every connected space an image of a
connected metric space [465]?
Definition 2.3.16 ([120]) A space X is called an s-connected space if X cannot
be represented as the union of two disjoint nonempty sequentially open sets. An
s-connected space is also called a sequentially connected space.
Obviously, every connected sequential space is an s-connected space and every
s-connected space is a connected space.
Theorem 2.3.17 ([278]) For every space X, the following are equivalent:
(1) X is a sequence-covering image of a connected metric space.
(2) X is an image of a connected metric space.
(3) X is an s-connected space.
76 2 Mappings on Metric Spaces

Proof (2) ⇒ (3). We only need to prove that if f : M → X is a mapping and M is

s-connected, then X is also s-connected. Otherwise, X is the union of two disjoint
nonempty sequentially open sets A and B. If {an } is a sequence in X converging to a
point a ∈ f −1 (A), then f (an ) → f (a) ∈ A, so {f (an )} is eventually in A, and hence
{an } is eventually in f −1 (A). It follows that f −1 (A) is a sequentially open set in M.
For the same reason, f −1 (B) is also a sequentially open set in M. Thus, M is the
union of two disjoint nonempty sequentially open sets, a contradiction.

(3) ⇒ (1). Suppose X is an s-connected space. Let M = S (X) and denote
the metric space induced by the topological sum as (M, d). Let q : (M, d) → X be
the natural mapping. Then q is continuous.
Define ρ : (M × I)2 → R+ as follows:

d(y1 , y2 ) + t1 + t2 , y1 = y2 ,
ρ((y1 , t1 ), (y2 , t2 )) =
|t2 − t1 |, y1 = y2 .

It is easy to verify that ρ is a distance on M × I and M is homeomorphic to the

subspace M × {0} of (M × I, ρ).
Define a binary relation R on M × I as follows: (y1 , t1 )R(y2 , t2 ) if and only if
q(y1 ) = q(y2 ) and t1 = t2 = 1, or y1 = y2 and t1 = t2 . Then R is an equivalent
relation. Let Z = (M × I)/R, and let p : M × I → Z be the quotient mapping. For
every y ∈ M, n ∈ N, put

By,n = p({(y, 1)} ∪ {(y , t) : q(y ) = q(y), 1 − 1/n < t < 1}).

Define a topology τ on Z as follows: when (y, t) ∈ M × I, if t = 1, then the

neighborhoods of p(y, t) in Z have the same form as the neighborhoods of (y, t) in
the product space M × I; if t = 1, each element of a neighborhood base of p(y, t)
in Z has the form By,n , n ∈ N. Then (Z, τ ) is a regular space and for each U ∈ τ ,
p−1 (U) is an open set in (M × I, ρ). Thus, p is continuous.
(17.1) (Z, τ ) is a metrizable space.
Assume that B is a σ -locally finite base for the metric space (M × I, ρ). Let
Q be the set of all rational numbers in the interval (1/3, 1) of I. Put

P1 = {p(B) : B ∈ B, B ⊂ M × [0, 1/2)},

P2 = {p({y} × (r1 , r2 )) : y ∈ M, r1 , r2 ∈ Q },
P3 = {By,n : y ∈ M, n ∈ N},
P = P1 ∪ P2 ∪ P3 .

Then P is a σ -locally finite base for (Z, τ ), and hence (Z, τ ) is a metrizable
space.
Define h : (M × I, ρ) → X and f : (Z, τ ) → X such that h(y, t) =
q(y) and f ◦ p = h.
(17.2) f is a sequence-covering mapping.
Notice that p(E ×[0, 1)) ∈ τ for each open subset E of (M, d) and p(q−1 (x)×
2.3 Quotient Mappings 77

−1
(0, 1]) ∈ τ for each x ∈ X. If A is an open set in X, then
 q (A) is an open set in
(M, d), and hence f (A) = p(q (A)×[0, 1))∪( x∈A p(q−1 (x)×(0, 1])) ∈
−1 −1

τ . Thus f is continuous. Assume that {xn } is a nontrivial sequence in X

converging to x. By the structure of q, there is a sequence {yn } in (M, d)
converging to y such that q(y) = x and q(yn ) = xn . Since ρ((yn , 0), (y, 0)) =
d(yn , y), (yn , 0) → (y, 0). Because p is continuous, p(yn , 0) → p(y, 0).
Since f (p(yn , 0)) = xn , f is a sequence-covering mapping.
(17.3) (Z, τ ) is an s-connected space.
Otherwise, Z is the union of two disjoint nonempty sequentially open sets C
and D. Obviously, f −1 (x) is an s-connected set in Z, so either f −1 (x) ⊂ C, or
f −1 (x) ⊂ D, and hencethere exist sets A, B in X such that X = A ∪ B, C =
 −1
x∈A f (x) and D = x∈B f −1 (x), i.e. C = f −1 (A), D = f −1 (B). Suppose
{xn } is a sequence in X converging to x ∈ A. By (17.2), there is a sequence {zn }
in Z converging to some point z ∈ f −1 (x) such that f (zn ) = xn . Then z ∈ C,
so {zn } is eventually in C, and hence {xn } is eventually in A. Consequently, A
is a sequentially open set in X. For the same reason, B is also a sequentially
open set in X, and which contradicts the s-connectivity of X. Thus, Z is an
s-connected space. 

By the mapping lemma (see Proposition 2.1.12), we have the following corollary.

Corollary 2.3.18 ([120, 278]) A space X is a quotient image (resp. pseudo-open

image) of a metric space if and only if X is a connected sequential space (resp.
Fréchet–Urysohn space).

Example 2.3.19 ([278]) There is a connected space which is not an s-connected

space.
Let βR be the Čech–Stone compactification of R. We first prove that for any
p ∈ βR − R, there does not exist nontrivial sequence in R converging to p. In fact,
if there is a nontrivial sequence {xn } in R converging to p, take A = {x2n : n ∈
N}, and B = {x2n−1 : n ∈ N}, then A, B are disjoint closed sets in R. Since R
is normal, clβR (A) ∩ clβR (B) = ∅ (see Corollary 3.6.4 of Engelking [119]), and
p ∈ clβR (A) ∩ clβR (B), a contradiction.
Since R is connected, βR is also connected. On the other hand, R is both a
sequentially open set and a sequentially closed set in βR, so βR is not an s-connected
space.
By Theorem 2.3.17, the connected space βR is not an image of any connected
metric space.
In the last proposition of this section, we give the relationship between Fréchet–
Urysohn spaces and strongly Fréchet–Urysohn spaces by means of products.

Proposition 2.3.20 ([336]) A space X is a strongly Fréchet–Urysohn space if and

only if X × S1 is a Fréchet–Urysohn space.

Proof Suppose X is a strongly Fréchet–Urysohn space. For any A ⊂ X ×S1 , if p ∈ A,

we may assume p = (x, 0). For each n ∈ N, define
78 2 Mappings on Metric Spaces

An = A ∩ (X × (S1 − {1/m : m < n})),

Bn = π1 (An ).

Then {Bn } is a decreasing sequence of sets in X and x ∈ n∈N Bn , so there is xn ∈ Bn
such that xn → x, and hence there is {zn } ⊂ A such that zn → p. Thus X × S1 is a
Fréchet–Urysohn space.
Conversely, suppose X × S1 is aFréchet–Urysohn space.  For any decreasing
sequence {An } of sets in X, if x ∈ n∈N An , take A = n∈N (An × {1/n}), then
(x, 0) ∈ A ⊂ X × S1 , so there is a sequence {zi } ⊂ A such that zi → (x, 0), and
hence there exist a subset {ni : i ∈ N} of N and xi ∈ Ani such that zi = (xi , 1/ni ) and
xi → x when ni → ∞. Consequently, there is yn ∈ An such that yn → x. Thus X is
a strongly Fréchet–Urysohn space. 

2.4 Open Mappings

In this section, we give characterizations of open images of metric spaces and para-
compact M-spaces. In 1960, Ponomarev [401] proved that a space X is a first count-
able space if and only if it is an open image of a metrizable space. The Ponomarev
theorem is one of the original motivations for the Alexandroff idea [35], and the par-
ticular method of representing a non-metrizable space as an image of a subspace of
a Baire’s zero-dimensional space created by Ponomarev, referred to as Ponomarev’s
method, is a remarkable contribution to the mapping theory of metric spaces. In the
following sections of this chapter, we introduce this method systematically.

Definition 2.4.1 ([340]) A family B of open sets in a space X is called an outer

base of a set A in X if for each x ∈ A and x ∈ U ∈ τ (X), there is B ∈ B such that
x ∈ B ⊂ U.

Lemma 2.4.2 (The König lemma) Let {Xi } be a sequence of nonempty finite sets.
If for each n < m, there is a correspondence  πn : Xm → Xnm such that πn =
m m

πn ◦ πk and πm = idXm , then there is (xi ) ∈ i∈N Xi such that πn (xm ) = xn .

k m m

Proof For each i ∈ N, give Xi the discrete topology. Then Xi is a compact space. Let

X= Xi ,
i∈N
Y = {(xi ) ∈ X : πnm (xm ) = xn , n < m}.

Then Y is a closed set in the compact space X. Because in fact, if y = (yi ) ∈ X − Y ,

then there is n < m such that πnm (ym ) = yn . Let V = {(xi ) ∈ X : xm = ym , xn = yn }.
Then y ∈ V ∈ τ (X) and V ∩ Y = ∅, so Y is a closed set in X. To complete our
proof, we only need to show Y = ∅. For each m ∈ N, take
2.4 Open Mappings 79

Ym = {(xi ) ∈ X : if n < m, then πnm (xm ) = xn }.

Then Ym is a nonempty closed set in X, and hence{Ym } is a sequence of closed sets

in X with the finite intersection property, so Y = m∈N Ym = ∅. 

Lemma 2.4.3 ([340]) Suppose X is a space and K ∈ K (X). If K has a countable

outer base U in X, then there is a sequence {Ui } of finite subsets of U such that
(1) for each i ∈ N, K ⊂ ∪Ui ;
(2) if x ∈ K and x ∈ Ui ∈ Ui for every i ∈ N, then {Ui }i∈N is a neighborhood base
of x;
(3) for every x ∈ K and i ∈ N, there is Ui ∈ Ui such that x ∈ Ui+1 ∩ K ⊂ Ui .

Proof Let {Vi } be the families of all finite subsets of U covering K. By the inductive
method, we can take a subsequence {Ui } of {Vi } such that for every i ∈ N, Ui partially
refines Vi and {U ∩ K : U ∈ Ui+1 } partially refines Ui . Below we verify that {Ui }
satisfies conditions (1)–(3).
(1) holds obviously. For each x ∈ K and x ∈ Ui ∈ Ui , let x ∈ W ∈ τ . Take
V ∈ U and V ∈ U <ω such that x ∈ V ⊂ W and K − V ⊂ ∪V ⊂ X − {x}.
Then there is m ∈ N such that V ∪ {V } = Vm , so Um ⊂ V , and hence {Ui }i∈N is a
neighborhood base of x. To verify (3), define Wi = (Ui )x for every x ∈ K and i ∈ N.
Then Wi is finite and satisfies that if Ui+1 ∈ Wi+1 , then there is Ui ∈ Wi such that
Ui+1 ∩ K ⊂ Ui . By the König lemma, there is Ui ∈ Ui such that x ∈ Ui+1 ∩ K ⊂ Ui
for each i ∈ N. 

Proposition 2.4.4 ([340]) Suppose X is a first countable space. If U is a base for

X, then there exist a metrizable space M and an open mapping f : M → X such that
(1) if K ∈ K (X) and K has a countable outer base UK ⊂ U , then there is
L ∈ K (M) such that f (L) = K;
(2) if (U )E is countable for some E ⊂ X, then f −1 (E) has a countable base.

Proof Denote U = {Uα }α∈Λ . For each i ∈ N, let Λi be the set Λ with the discrete
topology. Define
 
M = β = (αi ) ∈ Λi :
i∈N

{Uαi } is a neighborhood base of some point x(β) in X .

Then M is a metrizable space. For each β ∈ M, x(β) is the only point being
determined, so we can define a function f : M → X by f (β) = x(β).
(4.1) f is continuous.
By the first countability of X, f is an onto mapping. Let β = (αi ) ∈ M and
f (β) = x ∈ U ∈ τ (X). Then there is m ∈ N such that x ∈ Uαm ⊂ U. Put
80 2 Mappings on Metric Spaces

V = {γ ∈ M : πm (γ ) = αm }.

Then β ∈ V ∈ τ (M) and f (V ) ⊂ Uαm ⊂ U. Thus f is continuous.

(4.2) f is an open mapping.
For every n ∈ N and αi ∈ Λi (∀i  n), put

B(α1 , . . . , αn ) = {β ∈ M : πi (β) = αi , i  n}.

If β = (βi ) ∈ B(α1 , . . . , αn ), then

f (β) ∈ Uβi ⊂ Uαi .
i∈N in

 
So f (B(α1 , . . . , αn )) ⊂ in Uαi . If x ∈ in Uαi , for each i > n, take αi ∈ Λi
such that {Uαi }i>n is a neighborhood base  of x. Let β = (αi ) ∈ i∈N Λi . Then
β ∈ B(α1 , . . . , αn ) and f (β) = x, so in Uαi ⊂ f (B(α1 , . . . , αn )). Thus,

f (B(α1 , . . . , αn )) = Uαi .
in

Since {B(α1 , . . . , αn ) : αi ∈ Λi , i  n} is a base for M, f is an open mapping.

By the definition of f , (2) holds. Below we prove (1).
(4.3) If K ∈ K (X) and K has a countable outer base UK ⊂ U , then there is
a sequence {Ui } of finite subsets of UK satisfying all conditions of Lemma
2.4.3.
For each i ∈ N, there is Γi ∈ Λ<ω i such that Ui = {Uαi : αi ∈ Γi }. We may
assume that Uαi ∩ K = ∅. Let
 

L = β = (αi ) ∈ Γi : Uαi+1 ∩ K ⊂ Uαi for each i ∈ N .
i∈N

 
Then L is a closed set in i∈N Γi . Because in fact, if (αi ) ∈ i∈NΓi − L,
then there is m ∈ N such that Uαm+1 ∩ K ⊂ Uαm . Let W = {(βi ) ∈ i∈N Γi :
βm = αm }. Then (α
 i ) ∈ W ∈ τ ( i∈N Γi ) and W ∩ L = ∅. Thus L is a

compact set in i∈N Γ i . Since K ∩ ( i∈N Uαi ) = ∅ for each β = (αi ) ∈ L,
we can take x ∈ K ∩ ( i∈N Uαi ). Then β ∈ M and f (β) = x, so L ⊂ M and
f (L) ⊂ K. By Lemma 2.4.3(3), K ⊂ f (L). Consequently, L ∈ K (M) and
f (L) = K. 

Theorem 2.4.5 ([401]) (The Ponomarev theorem) A space X is a first countable

space if and only if X is an open image of a metrizable space.

Proof The necessity comes from Proposition 2.4.4 and the sufficiency is obtained
by the fact that first countability is invariant under open mappings. 
2.4 Open Mappings 81

Theorem 2.4.6 ([58]) Any countably compact subset of a space X having a quasi-
G δ -diagonal is a compact metrizable G δ -subset of X.

Proof Let {Un } be a quasi-G δ -diagonal sequence for X, and C be a countably compact
subset of X. It is easy to see that {Un|C } is a quasi-G δ -diagonal sequence for C. By
Theorem 1.4.10, C is compact.
(1) Every closed set of C is a G δ -set in C.
Let M be a closed subset of C, and assume that M has no isolated points. Let
I be the set of isolated points of M. When M − I is a G δ -set, it easily follows
that M itself is a G δ -set. For each x ∈ M, there is a strictly increasing sequence
{m(i, x)}i∈N of positive integers such that for all n ∈ N, x ∈ ∪Un if and only
if n = m(i, x) for some i  n. For each i ∈ N, pick U(i, x) ∈ Um(i,x) such that
x ∈ U(i, x).
Let H(1, x) = U(1, x) ∩ C for each x ∈ M. Then {H(1, x) : x ∈ M}, together
with C − M, covers C. Let W1 be a finite subcover of this cover. Note the
fact that if W ∈ W1 and W ∩ M = ∅, then there is some x ∈ M such that
W ⊂ U(1, x).
Next, for each x ∈ M, by the regularity of C, let V (2, x) be an open neighbor-
hood of x in C such that clC [V (2, x)] = V (2, x) is contained in some element
of W1 . Let H(2, x) = V (2, x) ∩ U(2, x). Then {H(2, x) : x ∈ M}, together
with C − M, covers C. Let W2 be a finite subcover of this cover.
Continue this process. Then for each n ∈ N we obtain a finite cover Wn of
open subsets of C and a family {H(n, x) : x ∈ M} of open sets, such that, the
following hold for all n ∈ N:
(1.1) if W ∈ Wn and W ∩ M = ∅,  then W ⊂ H(n, x) for some x ∈ M;
(1.2) for all x ∈ M, x ∈ H(n, x) ⊂ in U(i, x);
(1.3) for all x ∈ M and all j  n, H(n + 1, x) is contained in some element  of Wj .
Now let Vn = st(M, Wn ). Clearly each Vn is open in C and M ⊂ n∈N Vn . Let
p ∈ n∈N Vn − M. Let {Wi }it1 be all elements of W1 which contain p and
intersect M. Such elements of W1 exist since p ∈ V1 . By (1.1), for each r  t1 ,
there is some xr ∈ M such that Wr ⊂ H(1, xr ). Let T1 = {xr : r  t1 } and
j1 = 1.
Now let j2 = j1 + 1. Let {Wi }it2 be all elements of Wj2 which contain p and
intersect M. Such elements exist since p ∈ Vj2 . By (1.2), for r  t2 , there is
some xr ∈ M such that Wr ⊂ H(j2 , xr ). By (1.3), for each r  t2 , H(j2 , xr ) is
contained in some element of Wj1 . It easily follows that H(j2 , xr ) ⊂ H(j1 , xs )
for some xs ∈ T1 . Let T2 = {xr : r  t2 }.
Continuing this process, we obtain a strictly increasing sequence {jn }n∈N of
positive integers and a nonempty finite subset Tn of M for each n ∈ N, such
that, the following hold for all n ∈ N:
(1.4) if x ∈ Tn , then p ∈ H(jn , x);
(1.5) if x ∈ Tn+1 , then H(jn+1 , x) ⊂ H(jn , y) for some y ∈ Tn .
For each n ∈ N, let Hn = {H(jn , x) : x ∈ Tn }. Then by Lemma 2.4.2, there
is a sequence {xn } in M such that p ∈ H(jn , xn ) and H(jn+1 , xn+1 ) ⊂ H(jn , xn )
82 2 Mappings on Metric Spaces

for all n ∈ N. Since C is compact, {xn } has an accumulation point, say q.

Clearly q ∈ M, and it is easy to check that q ∈ H(jn , xn ), n ∈ N. (Recall that
xn ∈ H(jn , xn ) by (1.2), and that H(jn+1 , xn+1 ) ⊂ H(jn , xn ).) Now p ∈ / M, so
p = q. Hence there is some n0 ∈ N such that p ∈ / st(q, Un0 ) = ∅. Choose
n  n0 such that xn ∈ st(q, Un0 ). Since xn ∈ ∪Un0 , there is some i0  n0 such
that n0 = m(i0 , xn ). Note that
 U(i0 , xn ) ∈ Um(i0 ,xn ) = Un0 and i0  n  jn .
Now {p, q} ⊂ H(jn , xn ) ⊂ ijn U(i, xn ), so both p and q belong to U(i0 , xn ).
Thus p ∈ st(q, Un0 ), a contradiction. Hence p ∈ M, and so the proof that M is
a G δ -set in C is completed.
(2) C is metrizable.
Suppose {Vn } is a quasi-G δ -diagonal sequence for C. For every n∈ N, by
(1), there is a sequence {Cn,j } of closed sets in C such that ∪Vn = j∈N Cn,j .
Let Vn,j = Vn ∪ {C − Cn,j } for every n, j ∈ N. Then {Vn,j } is a G δ -diagonal
sequence for C. By Proposition 1.4.9, C has a G ∗δ -diagonal. Assume that {Gn }
is a G ∗δ -diagonal sequence for C such that Gn+1 refines Gn for each n ∈ N. For
every x ∈ O ∈ τ (C), {O}∪{C −st(x, Gn ) : n ∈ N} is an open cover of C which
contains a finite subcover. So there exists m ∈ N such that st(x, Gm ) ⊂ O. It
follows that {Gn } is a development of C. By the Bing metrization criterion, C
is metrizable.
(3) C is a G δ -subset in X.
We may assume that U1 = {X}. Being a compact and metrizable subset of X,
C is hereditarily
 Lindelöf, sothat there is a countable subfamily Un ⊂ Un that

covers C ∩ Un . Let U = n∈N Un . Because U is countable, we may  index

it as U = {Ui : i ∈  N}. For each x ∈ C and n ∈  N, let U (x, n) = in {Ui :
x ∈ Ui }. Let G n = x∈M U (x, n). Then  C ⊂ n∈N n G . For a contradiction,
suppose that there is some point z ∈ n∈N G n − C. Choose a point xi ∈ C
with z ∈ U (xi , i) for each i ∈ N. Because xi ∈ C, there is an accumulation
point p of the sequence {xi } in C. Because p = z,we may find  m ∈ N such
that p ∈ st(p, Um ) ⊂ X − {z}. Because p ∈ C ∩ Um ⊂ C ∩ Um , some
U ∈ Um has p ∈ U. Then U appears somewhere in the listing of U given
above, say U = Uk for some k ∈ N. Because p is an accumulation point of the
sequence {xi } and p ∈ U = Uk , there is some j > k with xj ∈ Uk . But then we
have z ∈ U (xj , j) ⊂ Uk ⊂ st(p, Um ) ⊂ X − {z} and that is impossible. Hence
C is a G δ -set in X. 

Question 2.4.7 ([465]) Is every first countable connected space an open image of
a connected metric space?
In the second part of this section, we introduce open images of paracompact
M-spaces.

Proposition 2.4.8 ([335, 475]) Let K and P be covers of a space Y , where the
elements of K are countably compact closed subsets of Y and P is closed under
finite intersections. If for every y ∈ P ∈ P, there is K ∈ K such that
2.4 Open Mappings 83

(i) y ∈ K ⊂ P,
(ii) some countable subfamily of P is a network for K in Y ,
then there exist a metrizable space M, a σ -discrete base B of M and a subspace X
of Y × M satisfying the following conditions: let f = π1|X and g = π2|X , then
(1) P = f (g −1 (B));
(2) if β ∈ M, then f (g −1 (β)) ∈ K ;
(3) g is a closed mapping;
(4) for each E ⊂ Y , |{B ∈ B : B ∩ g(f −1 (E)) = ∅}|  ℵ0 · |(P)E |.

Proof Denote P = {Pα }α∈Λ . For each i ∈ N, let Λi be the set Λ with the discrete
topology. Define
 
M = β = (αi ) ∈ Λi :
i∈N

{Pαi } is a decreasing network of some element Kβ of K in Y .

Then M is a metrizable space and for each β ∈ M, Kβ is the only point being
determined. Let
X = {(y, β) ∈ Y × M : y ∈ Kβ }.

For every n ∈ N and αi ∈ Λi (∀i  n), define

B(α1 , . . . , αn ) = {β ∈ M : πi (β) = αi , i  n},

B = {B(α1 , . . . , αn ) : αi ∈ Λi , i  n ∈ N}.

Then B is a σ -discrete base of M.

(1) f (g −1 (B(α1 , . . . , αn ))) = Pαn .
Obviously, f (g −1 (B(α1 , . . . , αn ))) ⊂ Pαn . If y ∈ Pαn , then there is K ∈ K such
that y ∈ K ⊂ Pαn and K has a countable network F ⊂ P in Y . We may denote
F = {Pαi }i>n , where αi ∈ Λi and Pαn ⊃ Pαi ⊃ Pαi+1 . Let β = (αi ). Then
β ∈ B(α1 , . . . , αn ) and y ∈ K = f (g −1 (β)), so Pαn ⊂ f (g −1 (B(α1 , . . . , αn ))).
Thus f (g −1 (B(α1 , . . . , αn ))) = Pαn .
(2) For each β ∈ M, f (g −1 (β)) = Kβ ∈ K by the definition.
(3) For each closed set C in X, let β = (αi ) ∈ g(C). Then for every n ∈ N,
there is (yn , β (n) ) ∈ (Y × B(α1 , . . . , αn )) ∩ C, so yn ∈ Pαn , and hence {yn } has
an accumulation point, we denote it by y0 . Since {β (n) } converges to β in M,
(y0 , β) ∈ C, thus β ∈ g(C). Consequently, g is a closed mapping.
(4) By (1), for each E ⊂ Y , |{B ∈ B : B ∩ g(f −1 (E)) = ∅}|  ℵ0 · |(PE )|. 

In the above proof of Proposition 2.4.8, by (1), one can see that f is an onto
mapping. Moreover, if P is an open cover of Y , then f is an open mapping. By (2),
g is an onto mapping. By (3), g is a quasi-perfect mapping, thus X is an M-space.
84 2 Mappings on Metric Spaces

Definition 2.4.9 ([28]) A space X is called a pointwise countable type space (or
space of pointwise countable type) if for each x ∈ X, there is a compact set K
containing x such that K has a countable neighborhood base in X.
Obviously, every first countable space is of pointwise countable type, and every
space of pointwise countable type is a q-space.

Proposition 2.4.10 ([28]) Every regular space with a p-sequence is of pointwise

countable type.

Proof Let {Un } be a p-sequence in a regular space X. For every x ∈ X and n ∈ N,

take Un ∈ (Un )x . By the regularity of X, there is a sequence {Vn } of
open subsets of
X such that x ∈ Vn+1 ⊂ V n+1 ⊂ Vn ⊂ Un for each n ∈ N. Let K = n∈N Vn . By the
convergence lemma, x ∈ K ∈ K (X) and {Vn }n∈N is a countable neighborhood base
of K in X. Thus X is of pointwise countable type. 

Proposition 2.4.11 ([28]) Every space of pointwise countable type is a k-space.

Proof Let X be a space of pointwise countable type. Suppose there is a set A which is
not closed in X such that K ∩A ∈ τ c for each K ∈ K (X). Take x ∈ A−A. Then there
is a compact set C in X containing x such that C in X has a decreasing neighborhood
base {Ui }i∈N . Choose an open neighborhood V of x such that V ∩ C ∩ A = ∅.
Let B = {xi : i ∈ N}, where xi ∈ A ∩ V ∩ Ui . Then B ∪ C ∈ K (X), and hence
B = A ∩ (V ∩ (B ∪ C)) ∈ K (X). Since C ∩ B = ∅, there is i ∈ N such that
Ui ∩ B = ∅, a contradiction. Consequently, X is a k-space. 

Lemma 2.4.12 ([475]) Let X be of pointwise countable type. If x ∈ U ∈ τ , then

there is K ∈ K (X) such that x ∈ K ⊂ U and K has a countable neighborhood base
in X.

Proof Choose L ∈ K (X) such that x ∈ L and L has a decreasing neighborhood

base {Ui }i∈N in X. By the inductive method, there is a decreasing sequence {Vi } of
open subsets in X such that x ∈ Vi ⊂ Ui ∩ U and V i+1 ∩ L ⊂ Vi ∩ L. We notice
that if Vn has been selected, then by the compactness of L, there is Vn+1 ∈τ such
that x ∈ Vn+1 ⊂ Vn ∩ Un+1 and V n+1 ∩ L − Vn = ∅. Define K = L ∩ i∈N Vi .

Then K ∈ K (X) and x ∈ K ⊂ U. If K ⊂ W ∈ τ , then L ⊂ W ∪ i∈N (X − V i ),
so there is m ∈ N such that L ⊂ W ∪ (X − V m ), and hence there is k  m such that
Uk ⊂ W ∪ (X − V m ), thus Vk ⊂ Uk ∩ V m ⊂ W . Therefore, {Vi }i∈N is a neighborhood
base of K in X. 

Theorem 2.4.13 ([475]) A space X is of pointwise countable type if and only if X

is an open image of a paracompact M-space.

Proof The necessity can be obtained by Lemma 2.4.12, Proposition 2.4.8 and the
remark after Proposition 2.4.8. To prove the sufficiency, we only need to show that
every paracompact M-space is of pointwise countable type and spaces of pointwise
countable type are preserved by open mappings.
2.4 Open Mappings 85

Suppose X is a paracompact space and {Un } is an M-sequence in X. Then for

every x ∈ X and n ∈ N, st(x, Un+1 ) ⊂ st(x, Un ). Bythe convergence lemma,
{st(x, Un )}n∈N is a neighborhood base of the compact set n∈N st(x, Un ) containing
x in X. So X is of pointwise countable type.
Let f : X → Y be an open mapping, where X is of pointwise countable type.
For each y ∈ Y , pick x ∈ f −1 (y). Then there is a compact set K containing x such
that K has a decreasing neighborhood base {Ui }i∈N in X, and hence {f (Ui )}i∈N is a
neighborhood base of the compact set f (K) containing y in Y . Thus Y is of pointwise
countable type. 

Arhangel’skiı̌ [28] also introduced the concept of countable type spaces: a space X
is called a countable type space or of countable type if, for each compact subset F of X,
there is a compact set K in X containing F such that K has a countable neighborhood
base in X. Obviously, every countable type space is of pointwise countable type, and
countable type spaces are preserved by compact-covering open mappings. Choban
[103] proved that every regular space with a p-sequence is of countable type. Wicke
[475] proved that a space X is of countable type if and only if X is a compact-covering
open image of a paracompact p-space. The Michael line (see Example 1.8.5) is of
countable type which is not a p-space.
The property of pointwise countable type, q-space property, sequential space
property, k-space property, strongly Fréchet–Urysohn property, Fréchet–Urysohn
property, g-first countability and so on are all topological properties weaker than the
first countability, and hence these properties are collectively called the generalized
sequentiality properties.

Example 2.4.14 None of the following spaces is preserved by perfect mappings: first
countable spaces, spaces of pointwise countable type, q-spaces or g-first countable
spaces. Hence, perfect mappings do not preserve open images of metric spaces or of
paracompact M-spaces.
If X is the butterfly space (see Example 1.8.3), then X is a first countable space.
Let K = I × {0}. Then K is a compact set in X. Let f : X → X/K be the quotient
mapping. Then f is a perfect mapping and X/K is not a first countable space. Since
X/K is a regular Fréchet–Urysohn space of the point G δ -property, X/K is neither a
q-space (see Theorem 1.7.7) nor a g-first countable space (see Corollary 1.6.18).

Example 2.4.15 Some implication relationships do not exist among the generalized
sequentiality properties.
(1) There is a g-first countable space which is neither a Fréchet–Urysohn space nor
a q-space. The Arens space S2 (see Example 1.8.6) is such a space.
(2) There is a compact space (hence a space of pointwise countable type) which is
not a sequential space. The compactification βN is such a space.
(3) There is a Fréchet–Urysohn space which is not a strongly Fréchet–Urysohn
space. The sequential fan Sω (see Example 1.8.7) is such a space.
(4) There is a q-space which is not a k-space. Let X be the Frolík space [134], i.e.
N ⊂ X ⊂ βN and X is a countably compact subspace of βN with cardinality
86 2 Mappings on Metric Spaces

not greater than c. Then every compact set in X is a finite set, and hence X is not
a k-space.
(5) There is a paracompact and first countable space which is not a p-space. The
butterfly space is such a space.
Example 2.4.16 A finite-to-one open image of a metric space may not be metrizable.
Denote the V -space (see Example 1.8.1) by X. For each r ∈ R, let Xr = {(x, y) ∈
R2 : y = |x − r|}. Then Xr is a metrizable clopen subspace of X and {Xr : r ∈ R}
is a point-finite open cover of X. Let M = r∈R Xr and let f : M → X be the
natural mapping. Then M is a metric space and f is a finite-to-one open mapping.
We prove that f is a compact-covering mapping. For each compact set K in X, let
K0 = K ∩ (R × {0}) and K1 = K − ∪{Xr : r ∈ π1 (K0 )}. Since R × {0} is a discrete
closed subspace of X, K0 is a finite set. Since K1 is a compact set of the discrete space
X − (R × {0}), K1 is finite. Hence there is a compact set L in M such that f (L) = K.
Question 2.4.17 [411] (R.C. Olson) Is every quotient L-mapping from a space with
a point-countable base onto a space of pointwise countable type a countably bi-
quotient mapping?

2.5 Closed Mappings

Seeking characterizations of the closed images of metric spaces is a problem raised by

Arhangel’skiı̌ [31]. Lašnev [240] studied this problem and gave the first solution, and
Foged [130] gave another answer to Arhangel’skiı̌’s problem. This section consists
of three parts. In the first part, we introduce the characterizations of closed images
of metric spaces, which contains the famous Foged theorem. In the second part, we
investigate properties of HCP families and several metrization theorems are obtained.
In the third part, we discuss the countable productivity of closed images of metric
spaces.
Definition 2.5.1 ([427]) A space X is called a Lašnev space if X is a closed image
of a metric space.
Obviously, Lašnev spaces are additive and hereditary. By Proposition 2.3.20,
Sω × S1 is not a Lašnev space, and hence Lašnev spaces are not finitely productive.
Definition 2.5.2 Suppose P is a family of sets in a space X.
(1) P is said to be hereditarily closure-preserving in X if, for every H(P) ⊂ P ∈ P,
{H(P) : P ∈ P} is closure-preserving [240].
(2) P is said to be weakly hereditarily closure-preserving if, for every x(P) ∈ P ∈
P, the family {{x(P)} : P ∈ P} is closure-preserving, i.e. {x(P) : P ∈ P} is a
discrete closed subspace in X [87].
(3) P is said to be countably hereditarily closure-preserving (resp. countably weakly
hereditarily closure-preserving) if every countable subfamily of P is hereditar-
ily closure-preserving (resp. weakly hereditarily closure-preserving) [138].
2.5 Closed Mappings 87

(4) P is said to be CF if the family {K ∩ P : P ∈ P} is finite for each compact set

K ⊂ X [347].
Hereditarily closure-preserving can be simply expressed as HCP [87]. Obviously,
locally finite ⇒ HCP ⇒ weakly HCP and closure-preserving; compact-finite ⇒
CF. It is easy to verify that HCP families and weakly HCP families are preserved
by closed mappings.

Proposition 2.5.3 ([257]) If P is an HCP family in a regular space X, then P is

also an HCP family in X.

Proof Let P = {Pα }α∈Λ . If P is not an HCP family in  X, then foreach α ∈ Λ,
there is Hα ⊂ Pα such that α∈Λ H α ∈ / τ c . Take x ∈ α∈Λ Hα − α∈Λ H α . For
each α ∈ Λ, there exist Vα , Uα ∈ τ such that x ∈ Vα , H α ⊂ Uα and Vα ∩ Uα = ∅,
and hence Hα ⊂ Uα ∩ Pα ⊂ Uα ∩ Pα . Thus

x ∈ ∪{Uα ∩ Pα : α ∈ Λ} = ∪{Uα ∩ Pα : α ∈ Λ}.

Hence there is β ∈ Λ such that x ∈ Uβ ∩ Pβ , so Uβ ∩ Pβ ∩ Vβ = ∅, a contradiction.

Consequently, P is an HCP family in X. 

Lemma 2.5.4 Suppose P is a countably weakly HCP family in a space X. Let

D = {x ∈ X : P is not point-finite at x}.

(1) If K is a countably compact subset of X, then there is a finite subset F of K such

that (P)K−F is finite [386].
(2) {P − D : P ∈ P} ∪ {{x} : x ∈ D} is compact-finite [405].
(3) P is a CF family [347].

Proof Let K be a nonempty countably compact subset of X. First, K ∩ D is a finite

set. Because otherwise, there exist an infinite subset {xi : i ∈ N} of K and an
infinite subfamily {Pi : i ∈ N} of P such that xi ∈ Pi for each i ∈ N. It follows
that {xi : i ∈ N} is a closed discrete subset of K, which contradicts the countable
compactness of K. Let F = K ∩ D. Then F is a finite subset of K. If there is an
infinite subfamily {Qi : i ∈ N} of P such that Qi ∩ (K − F) = ∅ for each i ∈ N,
then take a sequence {yi } in K such that yi ∈ Qi − D for each i ∈ N. Since P is
point-finite at each point yi and P is a countably weakly HCP family, {yi : i ∈ N}
is a closed discrete subset of K, a contradiction. So (P)K−F is finite, and hence (1)
is proved.
Since K ∩ (P − D) = (K − F) ∩ P, {P − D : P ∈ P} ∪ {{x} : x ∈ D} is
compact-finite, so (2) is proved, and hence (3) is also proved. 

Lemma 2.5.5 ([130, 222]) If P a countably weakly HCP family in a Fréchet–

Urysohn space X, then {∩P ∗ : P ∗ ∈ P <ω } is an HCP family.
88 2 Mappings on Metric Spaces

ProofOtherwise, there exist an index set Λ and Pα ∈ P <ω , G α ⊂ ∩P  α , such

that, α∈Λ G α is not a closed subset of X. So there is a sequence {xn } ⊂ α∈Λ G α

such that xn → x ∈ X − α∈Λ G α . For each n ∈ N, there is α(n) ∈ Λ such that
xn ∈ G α(n) . We may assume that all α(n) are different. Since Pα is finite, there is a

subsequence {α(nk )} of {α(n)} such that Pα(nk ) − i<k Pα(ni ) = ∅. For each k ∈ N,
pick Pk ∈ Pα(nk ) . Then xnk ∈ Pk . Since X is a Fréchet–Urysohn space, there is a
sequence {yk,m }m∈N in Pk converging to xnk . Then, x ∈ {yk,m : k, m ∈ N}, and hence
there is a sequence {yki ,mi }i∈N in {yk,m : k, m ∈ N} converging to x. Since P is a
countably weakly HCP family, we  may assume xnk ∈ / Pk , so {ki : i ∈ N} is infinite,
and hence for each m ∈ N, {P ∈ k∈N Pα(nk ) : {yki ,mi : i  m} ∩ P = ∅} is infinite,
a contradiction. 

Lemma 2.5.6 ([130]) Suppose there is a k-network n∈N Pn for a Fréchet–Urysohn
space X such that Pn ⊂ Pn+1 . If a sequence Z in X converges to x ∈ U − Z, where
U ∈ τ , then there is m ∈ N such that Z is eventually in int(∪{P ∈ Pm : P ⊂ U}).

Proof Define Pn∗ = {P ∈ Pn : P ⊂ U} for each n ∈ N. If the lemma is not true, then
we can choose a subsequence {zn } of Z such that zn ∈ U −int(∪Pn∗ ) ⊂ U − ∪Pn∗ . So
there is a sequence {zn,k }k ⊂ U−∪Pn∗ such that zn,k → zn . Thus x ∈ {zn,k : n, k ∈ N},
and hence there is a subsequence {znj ,kj }j such that znj ,kj → x, where nj < nj+1 . Since
 ∗
n∈N Pn is a k-network, there is m ∈ N such that {znj ,kj } is eventually in ∪Pm ,
which contradicts the fact that znj ,kj ∈ U − ∪Pm∗ whenever nj  m. 

Proposition 2.5.7 Let f : X → Y be a compact-covering mapping. If P is a k-

network for X, then f (P) is a k-network for Y .

Theorem 2.5.8 For every regular space X, the following are equivalent:
(1) X is a Lašnev space.
(2) X is a k-space with a σ -HCP k-network and contains no closed copy of S2 .
(3) X is a Fréchet–Urysohn space with a σ -HCP k-network [130].
(4) X is a Fréchet–Urysohn space with a σ -compact-finite k-network [299].
(5) X is a Fréchet–Urysohn space with a σ -CF k-network [347].

Proof (1) ⇒ (2). Let f : M → X be a closed mapping, where M is a metric space. By

Proposition 2.3.1, X is a Fréchet–Urysohn space. Since S2 is not a Fréchet–Urysohn
space (see Example 1.8.6), X contains no closed copy of S2 . Let P be a σ -locally
finite base for M. Since f is a closed mapping and HCP families are preserved by
closed mappings, by Propositions 2.1.16 and 2.5.7, f (P) is a σ -HCP k-network for
X.
(2) ⇒ (3) [454]. Obviously, X is of the point G δ -property. By Corollary 2.3.5, X
is a sequential space. If X is not a Fréchet–Urysohn space, then there is a subset A
of X such that A = 
A, where


A = {z ∈ X : there is a sequence {zn } in A such that zn → z}.
2.5 Closed Mappings 89

So 
A is not a closed set in X, and hence there is a nontrivial sequence {xn } in  A
convergingto x ∈ A−A. Take a sequence {Un } of open sets in X such that U n+1 ⊂ Un
and {x} = n∈N Un . We may assume xn ∈ Un . Choose a sequence {Vn } of disjoint
open sets in X such that xn ∈ Vn . For each n ∈ N, there is a sequence {xn,m }m in
A ∩ Un ∩ Vn converging to xn . Let

M = {x} ∪ {xn : n ∈ N} ∪ {xn,m : n, m ∈ N}.

Then M is a closed copy of S2 in X, a contradiction. Thus X is a Fréchet–Urysohn

space.
(3) ⇒ (1). Let P be a σ -HCP k-network
 in the Fréchet–Urysohn space X. By
Lemma 2.5.5, we may assume P = n∈N Pn , where each Pn is an HCP family
which is closed under finite intersections and Pn ⊂ Pn+1 . For each P ∈ Pn , define

Rn (P) = P − int(∪{Q ∈ Pn : P ⊂ Q}),

Rn = {Rn (P) : P ∈ Pn }.

(8.1) For every x ∈ U ∈ τ , let {xn } be a sequence in X converging to x ∈ X −{xn : n ∈

N}. If {xn } is eventually in int(∪{P ∈ Pm : P ⊂ U}), then {xn } is eventually in
int(∪Rm ) and ∪Rm ⊂ U, where Rm = {R ∈ Rm : R∩{xn : n ∈ N} is infinite}.
In fact, denote

L = {xn : n ∈ N},
V = int(∪Pm ) − ∪{Q ∈ Pm ∪ Rm : Q ∩ L is a finite set}.

Then V ∈ τ . By Lemma 2.5.4, {xn } is eventually in V . If y ∈ Q ∈ Pm for some

y ∈ V , then Q∩L is infinite, and hence Pm is point-finite at y (by Lemma 2.5.4),
so ∩(Pm )y ∈ Pm . Let P(y) = ∩(Pm )y . Then y ∈ / ∪{Q ∈ Pm : P(y) ⊂ Q}, so
y ∈ Rm (P(y)), thus Rm (P(y))∩L is infinite. As a consequence, Rm (P(y)) ∈ Rm ,
therefore y ∈ Rm (P(y)) ⊂ ∪Rm , which shows V ⊂ ∪Rm , and hence {xn } is
eventually in int(∪Rm ). For each Rm (P) ∈ Rm , there is Q ∈ Pm such that
Rm (P) ⊂ P ⊂ Q ⊂ U. Because otherwise,

int(∪{Q ∈ Pm : Q ⊂ U}) ⊂ int(∪{Q ∈ Pm : P ⊂ Q}) ⊂ X − Rm (P),

then {xn } is eventually in X −Rm (P), so Rm (P)∩L is a finite set, a contradiction.

Consequently, ∪Rm ⊂ U.
Now for each n ∈ N define Rn∗ = Rn ∪ {X − int(∪Rn )} and denote Rn∗ =
{Rα : α ∈ Λn }. Give Λn the discrete topology. Let
  
M = β = (αn ) ∈ Λn : {Rαn } is a net at some point x(β) in X .
n∈N

Then M is a metrizable space and x(β) is the only point determined by β for
each β ∈ M. Define f : M → X by f (β) = x(β).
90 2 Mappings on Metric Spaces

(8.2) f is an onto mapping.

For each x ∈ X, if x is isolated in X, then there is m ∈ N, such that {x} ∈ Pm ,
so Rm ({x}) = {x}, and hence there is β ∈ M such that f (β) = x. If x is an
accumulation point of X, then there is a sequence {xk } in X − {x} converging
to x. For each n ∈ N, pick αn ∈ Λn such that Rαn ∩ {xk : k ∈ N} is infinite
(otherwise, we can pick αn ∈ Λn such that x ∈ Rαn ). Then x ∈ Rαn is always
true. By Lemma 2.5.6 and (8.1), {Rαn }n∈N is a net at x. Let β = (αn ). Then
β ∈ M and f (β) = x.
(8.3) f is continuous.
Let U ∈ τ (X) and β ∈ f −1 (U). Denote β = (αn ), then there is m ∈ N such
that f (β) ∈ Rαm ⊂ U, so β ∈ {γ ∈ M : πm (γ ) = αm } ⊂ f −1 (U), and hence
f −1 (U) ∈ τ (M).
(8.4) f is a closed mapping.
Let F ∈ τ c (M). If x ∈ f (F) − f (F), then there is a sequence {xi } in f (F)
converging to x. To each i ∈ N, take βi = (αi,n ) ∈ F ∩ f −1 (xi ). Then for
each n ∈ N, xi ∈ Rαi,n ∈ Rn∗ . By Lemma 2.5.4, there is m ∈ N such that
{R ∈ R1∗ : R ∩ {xi : i  m} = ∅} is finite, so there exist an infinite set
I1 in N and α1 ∈ Λ1 such that for each i ∈ I1 , αi,1 = α1 . By the inductive
method, we can  choose a decreasing sequence {In } of infinite sets in N and
β = (αn ) ∈ n∈N Λn , such that, αi,n = αn whenever n ∈ N and i ∈ In . For
each n ∈ N, pick k(n) ∈ In such that k(n) < k(n + 1). Then βk(n) → β, and
hence β ∈ F. For every n ∈ N and i ∈ In , we have xi ∈ Rαn , so x ∈ Rαn . Let
x ∈ U ∈ τ (X). Since xk(n) → x, by Lemma 2.5.4 and (8.1), there is m ∈ N such
that {xk(n) } is eventually in ∪{R ∈ Rm : R ∩ {xk(n) : n ∈ N} is infinite } ⊂ U.
If i  m, then k(i) ∈ Ii ⊂ Im , so Rαm = Rαk(i),m , and hence {Rαn }n∈N is a net at
x. Thus x = f (β) ∈ f (F), a contradiction. Therefore f is a closed mapping.

In summary, X is
a Lašnev space.
(3) ⇒ (4). Let n∈N Pn be a k-network for a Fréchet–Urysohn space X, where
each Pn is HCP and Pn ⊂ Pn+1 . For each n ∈ N, put

Dn = {x ∈ X : (Pn )x is not finite},

Fn = {P − Dn : P ∈ Pn } ∪ {{x} : x ∈ Dn }.

Then, by Lemma 2.5.4, to prove that n∈N Fn is a σ -compact-finite k-network for
X, we only need to show it is a k-network. For each compact set K ⊂ U ∈ τ in X,
there exist m ∈ N and P ∈ Pm<ω such that K ⊂ ∪P ⊂ U. Let

F = {P − Dm : P ∈ P} ∪ {{x} : x ∈ K ∩ Dm }.

Then F ∈ Fm<ω and K ⊂ ∪F ⊂ U. 

(4) ⇒ (5) is obviously. Next, we prove (5) ⇒ (3). Let n∈N Pn be a k-network
for a Fréchet–Urysohn space X such that each Pn is CF and closed under finite
unions in X. We assume that each Pn covers X. Put Pn = {Pλ : λ ∈ Λn }. We define
an equivalence relation ∼ in X by
2.5 Closed Mappings 91

x ∼ y if and only if {λ ∈ Λn : x ∈ Pλ } = {λ ∈ Λn : y ∈ Pλ }.

Then X is decomposed by the family Hn of all equivalence classes. Since Pn is CF,

Hn is a disjoint and compact-finite refinement of Pn . Hn is weakly HCP because
X is a k-space, thus it is HCP by Lemma 2.5.5. Let K ⊂ U with K compact and
U open in X. There exist m ∈ N and P ∈ Pm such that K ⊂ P ⊂ U. Then Hm|K
is finite, thus K ⊂ ∪H ⊂ P for some finite H ⊂ Hm . Therefore n∈N Hn is a
σ -HCP k-network for X. This completes the proof. 
In the second part of this section, we establish several metrization theorems by
using the above characterizations of Lašnev spaces.

Lemma 2.5.9 ([87]) Suppose P is a countably weakly HCP family of open sets in
a space X and A ⊂ X. If x ∈ Ad and there is a G δ -set G in X containing x such that
G ∩ (A − {x}) = ∅, then (P)x is finite.

Proof Otherwise, there is a countable family {Pn }n∈N in (P)x . Let G = n∈N G n ,
where G n ∈ τ . Put

H1 = A ∩ P1 ∩ G 1 , Hn+1 = Hn ∩ Pn+1 ∩ G n+1 , n ∈ N.

Then
 
x ∈ P1 ∩ G 1 ∩ A − {x} ⊂ H1 − {x} = Hn − Hn+1 = Hn − Hn+1 ,
n∈N n∈N

so there is m ∈ N such that x ∈ Hm − Hm+1 , and hence Pm+1 ∩G m+1 ∩(Hm −Hm+1 ) =
∅, a contradiction. 

Corollary 2.5.10 Let P be a countably HCP family of open sets in a space X. If x

is an accumulation point in X and {x} is a G δ -set, then (P)x is finite.

Corollary 2.5.11 Let n∈N Pn be a local base of a point x in a space X such that
each Pn is an HCP family. If x is an accumulation point in X, then Pn is finite.

Proof For every n ∈ N and P ∈ Pn , pick x(P) ∈ P − {x} and let Fn = {x(P) : P ∈
Pn }. Then Fn ∈ τ c . Put A = n∈N Fn and G = X − A. Then x ∈ Ad ∩ G, G is a
G δ -set in X and G ∩ (A − {x}) = ∅. By Lemma 2.5.9, Pn is finite. 

Lemma 2.5.12 ([87]) Let P be a countably weakly HCP family of open sets in a
space X. If X is a k-space, then ∩P ∈ τ .

Proof For each K ∈ K (X), by Lemma 2.5.4, there is F ∈ K <ω such that (P)K−F is
a finite set. Then P∩K ⊂ F for each P ∈ P −(P)K−F , and hence K ∩(∩P) ∈ τ (K).
Since X is a k-space, ∩P ∈ τ (X). 

Lemma 2.5.13 ([418]) Let X be a Fréchet–Urysohn space with a CF family P.

Then ∩P ◦ ⊂ [∩P]◦ .
92 2 Mappings on Metric Spaces

Proof Suppose there is a point x ∈ ∩P ◦ − [∩P]◦ . Then x ∈ X − ∩P. Take a

sequence {xn } in X − ∩P converging to x. Let K = {x} ∪ {xn : n ∈ N}. Then for
each P ∈ P, P ∩ K is an open set in K containing x and the family {P ∩ K : P ∈ P}
is finite. Hence ∩{P ∩ K : P ∈ P} is an open set in K containing x. But

∩{P ∩ K : P ∈ P} = (∩P) ∩ K = {x},

a contradiction. 

Definition 2.5.14 Suppose P is a topological property and B is a family of sets in

a space X. We say that B has the property P at non-isolated points if B has the
property P at each non-isolated point of X.
For example, suppose B is a base for a space X. We say that B is a σ -locally
finite base at non-isolated points [252] if, B = n∈N Bn and for each n ∈ N, Bn is
locally finite at each non-isolated point of X.
Let B be a family of subsets of a space X. For every x ∈ X, B is called HCP at
x if, for any H(B) ⊂ B ∈ B, x ∈ ∪{H(B) : B ∈ B}, then x ∈ ∪{H(B) : B ∈ B}.
Obviously, B is HCP for X if and only if for every x ∈ X, B is HCP at x.

Theorem 2.5.15 For every regular space X, the following are equivalent:
(1) X is a metrizable space.
(2) X has a base which is σ -discrete at non-isolated points [252].
(3) X has a base which is σ -locally finite at non-isolated points [252].
(4) X has a σ -HCP base [87]. (The Burke–Engelking–Lutzer metrization theorem)
(5) X is a k-space with a σ -countably weakly HCP base [87].
(6) X is of the point G δ -property and has a σ -countably HCP base [211].
(7) X is a strongly Fréchet–Urysohn space with a σ -countably weakly HCP base.
(8) X is a strongly Fréchet–Urysohn space with a σ -compact-finite k-network [299].
(9) X is a k-space with a σ -CF quasi-base [418].

Proof By Theorem 1.3.2, we obtain (1) ⇒ (2) and (9). Obviously (2) ⇒ (3). (8) ⇒
(1) is obtained by Theorems 2.5.8 and 2.2.2. By Corollary 2.5.11, we obtain (4) ⇒
(5) and (6). (6) ⇒ (7) is obtained by Corollary 2.5.10. By Lemma 2.5.5, Theorems
2.5.8 and 2.2.2, we obtain (7) ⇒ (1).
(3) ⇒ (4). It is sufficient to prove that if B is locally finite at non-isolated
points for X, then B is HCP for X. Let B = {Bα : α ∈ Γ }. For every α ∈ Γ ,
choose Hα ⊂ Bα . Put H = {Hα }α∈Γ . If x ∈ ∪H , we can assume that x is a non-
isolated point of X, then there exists an open neighborhood U(x) of x such that the
subfamily (H )U(x) is finite because H is locally finite at non-isolated points. Since
 
∪H = ∪(H − (H )U(x) ) ∪(H )U(x) and U(x) ∪(H − (H )U(x) ) = ∅, we
have x ∈ ∪(H )U(x)⊂ ∪H .
(5) ⇒ (7). Let n∈N Pn be a base for the k-space X such that each Pn is a
countably weakly HCP family. For each x ∈ X, by Lemma 2.5.12, {∩(Pn )x }n∈N is
a local base of x, thus X is a strongly Fréchet–Urysohn space.
2.5 Closed Mappings 93


(9) ⇒ (8). Let X be a k-space with a σ -CF quasi-base B = n∈N Bn . Since
every quasi-base is a k-network, we only need to show X is first countable by Theorem
2.5.8.
Since X is a k-space that each compact subset is metrizable, by Corollary 2.3.5,
X is sequential.
Now we prove X is Fréchet–Urysohn. Suppose X is not Fréchet–Urysohn. Then
there is a subset A ⊂ X such that the set

à = {x ∈ X : there is a sequence {xn } ⊂ A such that xn → x}

is not closed in X. Since X is sequential, and there are a point x ∈ X − Ã and a sequence
{xn } ⊂ Ã − A converging to x. For each n ∈ N, take a sequence {xn,m }m∈N ⊂ A
converging to xn . For each n, m ∈ N, let

Bn,m = {B ∈ Bn : {x} ∪ {xn : n  m} ⊂ B◦ }.

Using the same argument as in Lemma 2.5.13, for each n, m ∈ N we can take a
function fn,m : {k ∈ N : k  m} → N satisfying

C(fn,m ) = {xi,j : i  m, j  fn,m (i)} ⊂ ∩Bn,m .

Note that every neighborhood of x contains some C(fn,m ). Indeed, let U be a neigh-
borhood of x. Since B is a quasi-base for X, there is B ∈ B with x ∈ B◦ ⊂ B ⊂ U.
Let B ∈ Bn,m for some n, m ∈ N. Then C(fn,m ) ⊂ B. For each k ∈ N, let
Nk = ∩{C(fn,m ) : n, m  k}. Obviously, each Nk is nonempty and Nk+1 ⊂ Nk .
Take an arbitrary point yk ∈ Nk for each k ∈ N. Then {yk } is a sequence in A
converging to x. This is a contradiction.
Finally, X is first countable. Let x ∈ X. For each n ∈ N, let Bn = {B ∈ Bn :
x ∈ B◦ }. By Lemma 2.5.13, x ∈ [∩Bn ]◦ . Hence {[∩Bn ]◦ : n ∈ N} is a countable
neighborhood base at x, thus X is first countable. 
We give some explanations for the conditions of Theorem 2.5.15 by using the
following examples.
Example 2.5.16 ([87]) There is a non-metrizable regular space which has a σ -
weakly HCP base.
Let A be the set of ordinal numbers with the cardinality smaller than ℵω . Let
Z = {0, 1}A . For every z ∈ Z and α ∈ A, let z(α) = πα (z). Denote the element of Z
all coordinates of which are 0 by s. Suppose B is an open neighborhood base of s in
the product space Z. Let

X = {s} ∪ {z ∈ Z : {α ∈ A : z(α) = 0} ∈ A<ω }.

Define the following topology for X: B|X is a neighborhood base of the only accu-
mulation point s of X. Then X is a regular space which is not first countable, and
hence X is not a metrizable space. For each n ∈ N, define
94 2 Mappings on Metric Spaces

P1,n = {{z} : z ∈ X − {s} and |{α ∈ A : z(α) = 0}| = n}.

Then P1,n is a discrete family of open sets in X. For each B ∈ B, let

ΛB = {α ∈ A : πα (B) = {0}}.

Then ΛB ∈ A<ω and B is determined by ΛB uniquely. Furthermore, let

P2,n = {B ∩ X : B ∈ B and ΛB ⊂ [0, ωn ]},

where ωn is the smallest ordinal number of cardinality ℵn . Then |P2,n |  ℵn . To

prove that P2,n is a weakly HCP family, we only need to show that for every B ∩ X ∈
P2,n and p(B) ∈ B ∩ X − {s}, we have s ∈ / {p(B) : B ∩ X ∈ P2,n }. Let

ΓB = {α ∈ A : p(B)(α) = 0},
Γ = ∪{ΓB : B ∩ X ∈ P2,n }.

Then ΓB ∈ A<ω , and hence |Γ |  ℵn . Pick β ∈ A − Γ . Let

V = {z ∈ Z : z(β) = 0}.

Then s ∈ V ∩ X ∈ τ (X) and V ∩ {p(B) : B ∩ X ∈ P2,n } = ∅, so

/ {p(B) : B ∩ X ∈ P2,n }.
s∈

Thus X has a σ -weakly HCP base n∈N (P1,n ∪ P2,n ). 
By Theorem 2.5.15, X is not a k-space. Since X = {s} ∪ n∈N (∪P1,n ), X is an
Fσ -discrete space (a space X is said to be an Fσ -discrete space if X is the union of
countably many discrete closed subsets of it), and hence X is of the point G δ -property.

Example 2.5.17 ([85]) The property of having a σ -weakly HCP base is not preserved
by perfect mappings.
Let {Zn }n∈N be a family of disjoint regular
 spaces having σ -weakly
 HCP bases
(see Example 2.5.16). Take a point a ∈ / n∈N Zn . Let X = {a} ∪ n∈N Zn and give X
the following topology: each Zn is an open subspace of X; each basic neighborhood
of a has the form {a} ∪ nk Zn , k ∈ N. Since each Zn is a clopen subspace of X, it
is easy to verify that X is a non-metrizable regular space with a σ -weakly HCP base.
For each n ∈ N, Zn is not first countable. Let zn be a point which is not first
countable in Zn . Define Z = {a} ∪ {zn : n ∈ N}. Since the sequence {zn } converges to
a in X, Z is a compact subset of X. It is easy to verify that the set Z has no countable
neighborhood base in X. Take Y = X/Z and let f : X → Y be the quotient mapping.
Then f is a perfect mapping and f (a) has no countable neighborhood  base in Y . We
prove that the space Y has no σ -weakly HCP base. Otherwise, let n∈N Pn be a base
for Y such that each Pn is weakly HCP. Denote b = f (a). Since b has no countable
neighborhood base in Y , there is m ∈ N such that Pm is not point-countable at b,
2.5 Closed Mappings 95

so there is an uncountable family {Vα : α < ω1 } ⊂ (Pm )b . For each i < ω, f −1 (Vi )
is an open neighborhood of a, so there exist a strictly increasing sequence {ni } in N
and a sequence {xi } in X, such that, xi ∈ (Zni − {zni }) ∩ f −1 (Vi ) and the sequence
{xi } converges to a in X. Let A = {f (xi ) : i ∈ N}. Then b ∈ A, and hence there is a
subset {yα : α < ω1 } of Y such that each yα ∈ Vα ∩ A − {yβ : β < α} ⊂ A, which
contradicts the countability of A. So Y has no σ -weakly HCP base, which shows the
property of having a σ -weakly HCP base is not preserved by perfect mappings.

Example 2.5.18 ([85]) A regular space with a σ -weakly HCP base may not be a
meta-Lindelöf space.
Take an uncountable regular cardinal number κ and expand it to a strictly increas-
ing sequence {δn }n∈ω of regular cardinal numbers such that δ0 = κ. Let Z = {0, 1}λ ,
where λ = supn∈ω δn . For each z ∈ Z, denote

supp z = {t ∈ λ : z(t) = 1}.

Then for each t ∈ λ, there is only one et ∈ Z such that supp et = {t}. Let E = {et :
t ∈ δ0 }. For each n ∈ ω, denote

Jn = {z ∈ Z : |λ − supp z|  κ, λ − supp z ⊂ δn }.

Obviously, E ∩ Jn = ∅. Let X = E ∪ n∈ω Jn . 
Give X the following topology: every point in n∈ω Jn is isolated; a neighborhood
base of each point et in E is taken as the restriction of a neighborhood base of this
point in the product space Z to X. It is obvious that X is a completely regular space.
For every t ∈ δ0 and F ∈ [λ − {t}]<ω , let

U(t, F) = {x ∈ X : t ∈ supp x = λ − F}.

Then U(t, F) is a basic neighborhood of et in X. For each n ∈ ω, let

Bn = {U(t, F) : t ∈ δ0 , F ∈ [δn − {t}]<ω }.

(1) For every t ∈ δ0 and β ∈ λ − δn , since U(t, {β}) ∩ Jn = ∅, et is not an

accumulation point of Jn , so Jn is a discrete closed subset of X.
(2) Since |Bn | = δn , we can take Bn = {Bα : α ∈ δn }. Let P = {pα : α ∈ δn }, where
pα ∈ Bα . We prove P is a discrete closed subset of X. For each t ∈ δ0 , since
U(t, ∅)∩E = {et }, E is a discrete closed subset of X, so wemay assume E ∩P =
∅. Since | α∈δn (λ−supp pα )|  δn , there is γ ∈ λ−(δ0 ∪ α∈δn (λ−supp pα )).
Then U(t, {γ }) ∩ P = ∅, ∀t ∈ δ0 . Thus, P is a discrete closed subset of X. As
a consequence, Bn is weakly HCP.
Define  
B= Bn ∪ {{x} : x ∈ Jn }.
n∈ω n∈ω
96 2 Mappings on Metric Spaces

Obviously, B is a base for X. By (1) and (2), B is σ -weakly HCP.

(3) For each family of basic open sets in X which has the form W = {U(t, Ft ) : t ∈
δ0 }, there is p ∈ X such that ord(p, W ) = κ.
In fact, recall the Δ-system lemma. A family D is said to be a Δ-system if there
is a set R such that ∀A, B ∈ D[(A = B) ⇒ A ∩ B = R]. The Δ-system lemma
[237]: Suppose κ is an uncountable regular cardinal number and X is an infinite
set. If D is a subset of [X]<ω with the cardinality κ, then D contains a Δ-system
with the cardinality κ.
By the Δ-system lemma, the family {{t} ∪ Ft : t ∈ δ0 } of finite sets in λ contains
a subfamily {{t} ∪ Ft : t ∈ Λ} with the cardinality κ such that for some R ⊂ X
and every different t, t ∈ Λ, ({t} ∪ Ft ) ∩ ({t } ∪ Ft ) = R, so U(t, F
t ) = U(t , Ft ).
We may assume that Λ ∩ R = ∅ and there is m  ∈ ω such that t∈Λ Ft ⊂ δm .
Take p ∈ Z such that p(t) = 1 and supp p = λ − t∈Λ Ft . Since | t∈Λ Ft |  κ,
p ∈ Jm ∩ U(t, Ft ), ∀t ∈ Λ. Therefore, ord(p, W ) = κ.
By (3), the open cover {U(t, ∅) : t ∈ δ0 } of X has no point-countable open
refinement, so X is not a meta-Lindelöf space.

Example 2.5.19 There is a non-metrizable regular space which has a σ -countably

HCP base.
In an uncountable set X, fix a point p as the special point of X. Define the following
topology on X: for each F ⊂ X, F ∈ τ c (X) if and only if either p ∈ F or F is
countable.
The topological space is called a Fortissimo space [264, 437] and denoted by Xp .
Every Fortissimo space Xp has the following properties:
(1) Xp is a regular Lindelöf space.
(2) Xp does not have the point G δ -property.
(3) Any family of subsets of Xp is a countably HCP family.
(4) For any uncountable subset A of Xp , {{x} : x ∈ A} is not a closure-preserving
family of sets.
(5) Each compact set in Xp is a finite set.
We verify the above items one by one as follows:

(1) Obviously, each singleton of Xp is a closed set, and

{{x} : x ∈ Xp − {p}} ∪ {B ⊂ Xp : p ∈ B and |Xp − B|  ℵ0 }

is a closed base for Xp , so Xp is a regular space. For any open cover U of Xp ,

pick U ∈ (U )p . Then |Xp − U|  ℵ0 , so U has a countablesubcover.
(2) For anysequence {Un } of open neighborhoods of p, |Xp − n∈N Un |  ℵ0 , so
{p} = n∈N Un , and hence Xp does not have the point G δ -property.
(3) Suppose P is a family of sets in Xp . For any countable subfamily {Pn }n∈N of P
and Qn ⊂ Pn , let F = n∈N Qn . If p ∈ F, then F is a closed set in Xp ; if p ∈ / F,
then |Qn |  ℵ0 , so |F|  ℵ0 , and hence F is also a closed set in Xp . Thus P is
a countably HCP family.
2.5 Closed Mappings 97

(4) Since A is uncountable, p ∈ A − {p}, so A − {p} is not a closed set in Xp , and

hence {{x} : x ∈ A} is not a closure-preserving family of sets.
(5) We may assume p ∈ K ∈ K (X). If K is infinite, take an infinite set {xn : n ∈ N}
in K − {p}, then the open cover {Xp − {xn : n ∈ N}} ∪ {{xn } : n ∈ N} of K has
no finite subcover, a contradiction. Thus K is a finite set.
Question 2.5.20 ([310]) Does every space with a σ -weakly HCP base have the
point G δ -property?

Question 2.5.21 ([312]) Is every regular k-space with a σ -compact-finite k-network

a meta-Lindelöf space?
In the last part of this section, we discuss properties of countable products and
perfect preimages of Lašnev spaces. The following example shows that the property
of having a σ -HCP k-network is not finitely productive.

 ([221]) Sω1 × S1 has no σ -HCP k-network.

Example 2.5.22
Let X = α<ω1 Xα , where the common point s of all Xα = {s} ∪ {xα,n : n ∈ N} is
the only accumulation point in X, and U is an open neighborhood of s if and only if
for each α < ω1 , there is mα ∈ N such that

{s} ∪ {xα,n : n  mα , α < ω1 } ⊂ U.

Then X is a copy of Sω1 .

Let F be a σ -HCP closed k-network of Sω1 × S1 . By transfinite induction, we can
choose an uncountable set {Fα }α<ω1 in F such that |Fα ∩ ({s} × S1 )| = ℵ0 . In fact,
we first pick F ∈ F <ω such that X0 × S1 ⊂ ∪F . Then there exist F0 ∈ F and
x0 ∈ F0 ∩ ((X0 − {s}) × S1 ) such that |F0 ∩ ({s} × S1 )| = ℵ0 . Let α < ω1 and assume
that for each β < α, we have chosen Fβ ∈ F and xβ ∈ Fβ ∩ ((Xβ − {s}) × S1 ) such
that |Fβ ∩ ({s} × S1 )| = ℵ0 . Since the closed set {xβ : β < α} does not meet the
compact set Xα × S1 in Sω1 × S1 , there is F ∈ F <ω such that

Xα × S1 ⊂ ∪F ⊂ Sω1 × S1 − {xβ : β < α},

so there exist Fα ∈ F and xα ∈ Fα ∩((Xα −{s})×S1 ) such that |Fα ∩({s}×S1 )| = ℵ0 .

Since Fα ∩ {xβ : β < α} = ∅, Fβ = Fα whenever β < α. This completes the
inductive construction.
Because of the compactness of {s} × S1 and Lemma 2.5.4, {F ∈ F : |F ∩ ({s} ×
S1 )| = ℵ0 } is countable, a contradiction. So Sω1 × S1 has no σ -HCP k-network.

Lemma 2.5.23 ([199]) Suppose neither X nor Y is a discrete space. If X × Y is a

Lašnev space, then it is a metrizable space.

Proof Since X is a non-discrete Fréchet–Urysohn space, X contains a closed copy

of S1 , so S1 × Y is a Fréchet–Urysohn space. By Proposition 2.3.20 and Theorem
2.5.15, Y is a metrizable space. With the same reason, X is a metrizable space. 

Theorem 2.5.24 Suppose {Xn } is a sequence of Lašnev spaces.

98 2 Mappings on Metric Spaces


(1) If a subspace Z of n∈N Xn is a Fréchet–Urysohn space, then Z is a Lašnev space
[448].  
(2) If each |Xn |  2 and n∈N Xn is a Fréchet–Urysohn space, then n∈N Xn is a
metrizable space.

Proof (1) By Theorem 2.5.8, Xn has a k-network m∈N Pn,m , where Pn,m is a
compact-finite family of sets in Xn and Pn,m ⊂ Pn,m+1 . For each i ∈ N, take
⎛ ⎞  
 
Pi = ⎝ Pn,i ⎠ × Xn .
ni n>i


Then ( i∈N Pi )|Z is a σ -compact-finite k-network for the regular Fréchet–
Urysohn space Z, so Z is a Lašnev space.
(2) Let  
X= X2n , Y = X2n−1 .
n∈N n∈N

 neither X nor Y is a discrete space. Since X × Y is homeomorphic to

Then
n∈N Xn , itis a Lašnev space. By Lemma 2.5.23, X × Y is a metrizable space,
and hence n∈N Xn is a metrizable space. 

Question 2.5.25 ([448]) Find characterizations for subspaces of countable products

of Lašnev spaces.
The mapping π1 : Sω × S1 → Sω is a perfect mapping. Although Sω × S1 has a
G δ -diagonal, Sω × S1 has no σ -HCP k-network (and hence is not a Lašnev space).
Thus, not every space with a σ -HCP k-network (or even a Lašnev space) satisfies
the perfect preimage G δ -diagonal theorem. However, we have the following result.

Corollary 2.5.26 ([260]) Suppose f : X → Y is a perfect mapping, and Y is a

Lašnev space. If X is a Fréchet–Urysohn space with a G δ -diagonal, then X is a
Lašnev space.

Proof As a perfect preimage of a paracompact space, X is a paracompact space. By

Corollary 2.2.11, there is a metric space M and a one-to-one mapping g : X → M.
By using Corollary 2.1.9 for the property of “have a σ -compact-finite k-network”,
we obtain that X has a σ -compact-finite k-network. Thus X is a Lašnev space. 

2.6 Compact-Covering Mappings

The purpose of this section is to give a preliminary introduction of characterizations of

various compact-covering images of metric spaces obtained by Michael and Nagami
[340]. In the next section we will continue to introduce the characterizations of
compact-covering s-images of metric spaces.
2.6 Compact-Covering Mappings 99

A mapping f : X → Y is said to be almost open [24] if, for each y ∈ Y , there is

x ∈ f −1 (y) such that f (U) is a neighborhood of y in Y whenever U is a neighborhood
of x in X. Obviously, each open mapping is an almost open mapping and each almost
open mapping is a countably bi-quotient mapping.
Theorem 2.6.1 (1) A space X is a compact-covering image of a metric space if and
only if every compact set in X is metrizable [340].
(2) A space X is a compact-covering quotient (resp. pseudo-open, countably bi-
quotient) image of a metric space if and only if X is a k-space (resp. Fréchet–
Urysohn space, strongly Fréchet–Urysohn space) such that every compact set
in X is metrizable [340].
(3) X is a compact-covering and almost open image of a metrizable space if and
only if X is a first countable space and each compact subset of X is metrizable
[496].
Proof (1) Suppose X is the image of a metric space M under a compact-covering
mapping f . For each K ∈ K (X), there is L ∈ K (M) such that f (L) = K. Since
L is a compact metric space, K is a metrizable space. Conversely, suppose every
compact set in X is metrizable. Let M = ⊕K (X). Then M is a metrizable space
and the natural mapping from M onto X is a compact-covering mapping.
(2) The necessity is obtained by Proposition 2.3.1, and the sufficiency is followed
from the mapping lemma (see Proposition 2.1.12).
Note that, since a space X is a sequential space if and only if X has the weak
topology with respect to {K ∈ K (X) : K is metrizable subspace}, the condition
of “a k-space” in (2) is equivalent to that of “a sequential space”.
(3) It is easy to verify that the first countability is preserved by almost open mappings.
So by (1), we can get the necessity. Assume that X is a first countable space and
each compact subset of X is metrizable. By (1) and Theorem 2.4.5, there exist
metrizable spaces M1 , M2 , a compact-covering mapping f : M1 → X and an
open mapping g : M2 → X. Let M = M1 ⊕ M2 and define h : M → X such
that h|M1 = f and h|M2 = g. Then M is a metrizable space and h is a compact-
covering and almost open mapping. Thus X is a compact-covering and almost
open image of a metrizable space. 
The following characterization of compact-covering open images of metric spaces
relies on the next lemma.
Lemma 2.6.2 ([340]) Let K be a metrizable compact set in a space X. If there is a
countable neighborhood base of K in X, then K has a countable outer base in X.
Proof Let {Un }n∈N and {Vn }n∈N be a countable base of the subspace K and a countable
open neighborhood base of the set K in X respectively. Take

A = {(n, m) ∈ N × N : Un ⊃ U m }.

Then for every (n, m, k) ∈ A × N, there is Un,m ∈ τ (X) such that U m ⊂ Un,m ⊂
U n,m ⊂ X − (K − Un ). Let W (n, m, k) = Un,m ∩ Vk . We denote the family of
100 2 Mappings on Metric Spaces

all the finite intersections of the sets with the form of W (n, m, k) by H . Then H
is countable. We prove that H is an outer base of K in X. For every p ∈ K and
p ∈ U ∈ τ (X), define

B = {α ∈ A × N : p ∈ W (α)};

H(F) = W (α), F ∈ B<ω .
α∈F

Assume that there is no F ∈ B<ω such that H(F) ⊂ U. Pick p(F) ∈ H(F) − U. Let

Q(F) = {p(F ) : F ⊂ F ∈ B<ω }.

Then K ∩ Q(F) = ∅. Because otherwise, there exist k, n, m ∈ N such that Vk ∩

Q(F) = ∅ and p ∈ U m ⊂ Un . Denote α = (n, m, k) and F = F ∪ {α}. Then
α ∈ B and p(F ) ∈ W (α) ∩ Q(F) ⊂ Vk ∩ Q(F) = ∅, a contradiction. Thus,
{K ∩ Q(F) : F ∈ B<ω } has the finite intersection property, and hence K ∩ (∩{Q(F) :
F ∈ B<ω }) = ∅. On the other hand, for each x ∈ K − {p}, pick α ∈ B such that
x∈ / Q({α}), so (K − {p}) ∩ (∩{Q(F) : F ∈ B<ω }) = ∅. It follows
/ W ({α}). Then x ∈
that ∩{K ∩ Q(F) : F ∈ B<ω } = {p} ⊂ U. By the compactness of K, there is F ∈ B<ω
such that U ∩ Q(F) = ∅, a contradiction. Consequently, K has a countable outer
base in X. 

By Lemma 2.6.2 and Proposition 2.4.4, we obtain the following theorem.

Theorem 2.6.3 ([340]) A space X is a compact-covering open image of a metric

space if and only if every compact set in X is metrizable and has a countable neigh-
borhood base in X.

The above theorem can be extended by using weak neighborhood bases instead
of neighborhood bases, i.e., we can characterize spaces in which every compact set
is metrizable and has a countable neighborhood wake base in X by certain images
of metrizable spaces [295].

Example 2.6.4 ([233]) There is a closed mapping which is not a sequence-covering

mapping.
Denote the Mrówka space by ψ(N) (see Example 1.8.4). Define f : ψ(N) → S1
by f (ψ(N) − N) = {0} and f (n) = 1/n, n ∈ N. Then f is a closed mapping.
In fact, if U is an open set in ψ(N) and f −1 (0) ⊂ U, then N − U is a finite set.
Because otherwise, let {ni } be a sequence consisting infinite points in N − U. By
the maximality of A , there is A ∈ A such that A contains infinitely many points of
{ni }, and hence for any finite subset F of A, A − F ⊂ U, which contradicts the fact
that U is a neighborhood of the point A in ψ(N). Let V = {0} ∪ {1/n : n ∈ N ∩ U}.
Then V is an open neighborhood of 0 in S1 and f −1 (V ) = U. However, f is not a
sequence-covering mapping. Because otherwise, there is a compact set L in ψ(N)
such that f (L) = S1 . So N is a dense subset in ψ(N) and N ⊂ L, thus ψ(N) = L, a
contradiction.
2.6 Compact-Covering Mappings 101

In the second part of this section, we investigate two classes of special compact-
covering mappings and images of metrizable spaces under these mappings.

Definition 2.6.5 Suppose f : X → Y is a mapping.

(1) f is called a proper mapping [71] (or k-mapping [172]) if f −1 (K) is a compact
subset of X whenever K is a compact subset of Y .
(2) f is called a subproper mapping [54] if, there is a subset Z of X such that f (Z) = Y
and Z ∩ f −1 (K) is a compact subset of X whenever K is a compact subset of Y .
(3) The proper images of metrizable spaces are called k-metrizable spaces [54]. The
subproper images of metrizable spaces are called k ∗ -metrizable spaces [54].
Obviously, each perfect mapping is a proper mapping, each proper mapping is
a subproper mapping and each subproper mapping is a compact-covering mapping.
Thus, each metrizable space is a k-metrizable space, and each k-metrizable space is
a k ∗ -metrizable space.
A subset A of a space X is called a k-closed set in X if for each K ∈ K (X), A ∩ K
is closed in K. Obviously, each closed set is a k-closed set, and a space X is a k-space
if and only if every k-closed subset of X is closed in X.
The k-coreflection kX of a space X is the set X with the following topology: a
subset A of X is a closed set in kX if and only if A is k-closed in X. It is easy to verify:
(1) The identity idX : kX → X is a mapping.
(2) K (X) = K (kX).
(3) kX is a k-space.
(4) X is a k-space if and only if kX = X.

Definition 2.6.6 ([293]) A sequence {Un } of covers of a space X is called a point-

star network for X if {st(x, Un )}n∈N is a net at x in X for each x ∈ X. Let Φ be a
family property. If every Un in a point-star network {Un } for X has Φ, then {Un } is
called a Φ point-star network.
The concept of point-star networks is a generalization of the concept of weak
developments (see Definition 1.6.13). A weak development may be called a point-
star weak base. We can define a point-star sequential neighborhood network in a
similar way.

Theorem 2.6.7 For every space X, the following are equivalent:

(1) X is a k-metrizable space.
(2) kX is a metrizable space [54].
(3) X has a compact-finite k-closed point-star network [302].

Proof (1) ⇒ (3). Let f : M → X be a proper mapping, where M is a metrizable

space. By Theorem 1.3.5, there is a sequence {Un } of open covers of M such that
for each K ∈ K (M), {st(K, Un )}n∈N is a neighborhood base of K in M. For each
n ∈ N, let Bn be a locally finite closed refinement of Un , because M is paracompact.
Let Pn = f (Bn ). Then Pn is a compact-finite k-closed cover of X. Because in
fact, suppose H ∈ K (X). Then f −1 (H) ∈ K (M), so (Bn )f −1 (H) is finite, and hence
102 2 Mappings on Metric Spaces

(Pn )H is also finite. If P ∈ Pn , then there is B ∈ Bn such that P = f (B), thus

P ∩ H = f (B ∩ f −1 (H)) is a closed subset of H.
For each x ∈ U ∈ τ (X), since the compact set f −1 (x) ⊂ f −1 (U) ∈ τ (M), there is
n ∈ N such that st(f −1 (x), Un ) ⊂ f −1 (U), so st(f −1 (x), Bn ) ⊂ f −1 (U), and hence
st(x, Pn ) = f (st(f −1 (x), Bn )) ⊂ U. Therefore, {Pn } is a compact-finite k-closed
point-star network for X.
(3) ⇒ (2). Suppose {Pn } is a compact-finite k-closed point-star network for X.
Let P = n∈N Pn . If K is a compact subset of X, then P|K is a countable network
for K, so K is metrizable in X, and hence K is also metrizable in kX. For each n ∈ N,
it is obvious that Pn is a compact-finite closed cover of kX. Since kX is a k-space,
Pn is locally finite. For each x ∈ X, since x ∈ X − ∪{P ∈ Pn : x ∈ / P} ⊂ st(x, Pn ),
st(x, Pn ) is a closed neighborhood of x in kX.
We first prove kX is a first countable regular space. We may assume that each
Pn+1 is a refinement of Pn . Let x ∈ U ∈ τ (kX). Then there is m ∈ N such that
st(x, Pm ) ⊂ U. Because otherwise, for each n ∈ N, st(x, Pn ) ⊂ U, and hence there
exist Pn ∈ (Pn )x and pn ∈ Pn − U. Since {Pn } is a point-star network for X, the
sequence {pn } converges to x in X. Since ({x} ∪ {pn : n ∈ N}) ∩ U is an open subset
of the subspace {x} ∪ {pn : n ∈ N} of X, {pn } is eventually in U, a contradiction. That
shows x ∈ intkX [st(x, Pm )] ⊂ intkX [st(x, Pm )] ⊂ st(x, Pm ) ⊂ U. So kX is a first
countable regular space.
Now we prove P is a k-network for kX. Suppose a sequence {xn } in kX converges
to x ∈ U ∈ τ (kX). Then there is m ∈ N such that st(x, Pm ) ⊂ U. Since st(x, Pm )
is a neighborhood of x in kX, the sequence {xn } is eventually in st(x, Pm ). Since
Pm is point-finite, there exist P ∈ Pm and a subsequence {xni } of {xn } such that
{x} ∪ {xni : i ∈ N} ⊂ P ⊂ U. So P is a cs∗ -network for kX. By Proposition 1.6.7,
P is a k-network for kX.
By Theorem 2.5.15, kX is a metrizable space.
(2) ⇒ (1). Suppose kX is a metrizable space. Since idX : kX → X is a proper
mapping, X is a k-metrizable space. 
In the following, we give a characterization of k ∗ -metrizable spaces. Suppose X
is a space, denote
P ∗ (X) = {A ⊂ X : A = ∅}.

For each x ∈ X, let

Nw(x) = {(An )n∈ω ∈ P ∗ (X)ω : An+1 ⊂ An , and if U is a neighborhood

of x in X, then there is n ∈ ω such that An ⊂ U}.

If (An )n∈ω ∈ Nw(x), foreach n ∈ ω, take xn ∈ An , then the sequence {xn }

converges to x in X, so x ∈ n∈ω An . That shows {An }n∈ω is a decreasing net at x in
X, and hence {{x}∪ An }n∈ω is also a decreasing net at x in X.
Let Nw(X) = x∈X Nw(x). Give the power set P(X) the discrete topology. The
set Nw(X) with the subspace topology of the product space P(X)ω is called the
network hyperspace on X [54]. Obviously, each network hyperspace is a metrizable
2.6 Compact-Covering Mappings 103

space. Since X is a T2 space, for different points x, y in X, Nw(x) ∩ Nw(y) = ∅.

Define a function π : Nw(X) → X as follows: for every x ∈ X and (An )n∈ω ∈ Nw(x),

π((An )n∈ω ) = x, i.e., π −1 (x) = Nw(x).


Obviously, π((An )n∈ω ) = n∈ω An .

Lemma 2.6.8 ([54]) π : Nw(X) → X is a mapping.

Proof Obviously, π is an onto function. Let x ∈ X and (An )n∈ω ∈ Nw(x). If U is a

neighborhood of x in X, then there is m ∈ ω such that Am ⊂ U. Let

V = {(Bn )n∈ω ∈ Nw(X) : Bm = Am }.

Then V is an open neighborhood of (An )n∈ω in the metrizable space

 Nw(X) and
π(V ) ⊂ U. Because in fact, if (Bn )n∈ω ∈ V , then π((Bn )n∈ω ) = n∈ω Bn ⊂ Bm =
Am ⊂ U, and hence π(V ) ⊂ U. That proves π is continuous. 

Theorem 2.6.9 ([54]) A regular space X is k ∗ -metrizable if and only if X has a

σ -compact-finite k-network.

Proof The necessity. Suppose X is a k ∗ -metrizable regular space. Then there exist a
metrizable space M and a subproper mapping f : M → X, that is, there is a subset Z
of M such that f (Z) = X and Z ∩ f −1 (K) is a compact subset of M whenever K is a
compact subset of X. Let B = n∈N Bn be a σ -locally finite base for the metrizable
space M, where each Bn is locally finite. We prove that f (B|Z ) is a σ -compact-finite
k-network for X.
First, for each n ∈ N, f (Bn|Z ) is compact-finite in X. In fact, let K ∈ K (X). Then
Z ∩ f −1 (K) ∈ K (M), so only finitely many elements of Bn meet with Z ∩ f −1 (K),
and hence only finitely many elements of Bn|Z meet with f −1 (K). It follows K meet
with only finitely many elements of f (Bn|Z ). Therefore, f (B|Z ) is σ -compact-finite
in X.
Suppose K is a compact set, U is an open set in X and K ⊂ U. Since B is a
base for the space M and the compact set Z ∩ f −1 (K) ⊂ f −1 (K) ⊂ f −1 (U), there is

B ∈ B <ω such that Z ∩ f −1 (K) ⊂ ∪B ⊂ f −1 (U), so f (B|Z ) ∈ [f (B|Z )]<ω and
−1
K = f (Z ∩ f (K)) ⊂ ∪f (B|Z ) ⊂ U. Thus, f (B|Z ) is a k-network for X.
 The sufficiency. Suppose X has a σ -compact-finite k-network P. Denote P =
k∈ω Pk , where each Pk is compact-finite. We may assume that P is closed under
finite intersections and for each m ∈ ω, km Pk is also a k-network for X. Define
a sequence {βk }k∈ω of ordinal numbers as follows: β0 = 0; let αk be the first ordinal
number with cardinality |Pk | and let βk+1 = βk + αk . Denote

Pk = {Pα : βk  α < βk+1 }, k ∈ ω.

Let βω = supk∈ω βk . Then P = {Pα : α < βω }.

104 2 Mappings on Metric Spaces

For each x ∈ X, denote Λ(x) = {α < βω : x ∈ Pα }. For each k ∈ ω, since

Pk is point-finite, Λ(x) ∩ [βk , βk+1 ) is finite, so Λ(x) has the order type of ω, and
hence there is an order-preserving bijection αx : ω → Λ(x). For each k ∈ ω, let
Ak = ik Pαx (i) . If U is a neighborhood of x in X, then by the regularity of X, there
is P ∈ P such that x ∈ P ⊂ P ⊂ U. Since x ∈ P, there is k ∈ ω such that P = Pαx (k) ,
and then Ak ⊂ P ⊂ U. So (Ak )k∈ω ∈ Nw(x). Let s(x) = (Ak )k∈ω and hence we have
defined a function s : X → Nw(X) such that π ◦ s(x) = x.
Claim. If K is a compact subset of X, then s(K) is a compact subset of Nw(X).
In fact, since every compact subset of X is metrizable and Nw(X) is a metrizable
space, we only need to prove that if {xn } is a convergent sequence in X, then the
sequence {s(xn )} has a convergent subsequence in Nw(X). Let K = {x}∪{xn : n ∈ ω},
where xn → x ∈ X. Define Λ(K) = {α < βω : K ∩ Pα = ∅}. For each k ∈ ω,
since Pk is compact-finite, Λ(K) ∩ [0, βk+1 ) is finite, so Λ(K) is countable. For
each m ∈ ω, since {Pα : βm  α < βω } is a k-network  for X, there is an increasing
sequence {ki }i∈ω of natural numbers such that K ⊂ βk α<βk Pα .
i i+1
For each n ∈ ω, the index set Λ(xn ) = {α < βω : xn ∈ Pα } has the order type
of ω and we may suppose αn = αxn : ω → Λ(xn ) is an order-preserving bijection.
For each i ∈ ω, since xn ∈ K, αn (i) <  βki+1 , so αn (i) ∈ Λ(K) ∩ [0, βki+1 ). Denote
Λi (K) = Λ(K) ∩ [0, βki+1 ). Let  = i∈ω Λi (K) be the product space, where each
finite set Λi (K) has the discrete topology. Then  is a compact metrizable space and
each αn ∈ . Thus, the sequence {αn }n∈ω in  has a subsequence {αn }n∈J converging
to some point  α . For each i ∈ ω, since lim αn (i) =  α (i), αn (i) = 
α (i) when n is
Jn→∞
sufficient large in J, so xn ∈ Pαn (i) = Pα(i) , and hence x = lim xn ∈ Pα(i) .
Jn→∞
In the above, we have proved that for each i ∈
ω, αn (i) = 
α (i) when n is sufficient
large in J, so for each k ∈ ω, ik Pαn (i) = ik Pα(i) when n is sufficient large
  ω
 
in J. Let p = α (i) k∈ω ∈ P(X) . Since every s(xn ) =
ik P ik Pαn (i) k∈ω ,
the sequence {s(xn )}n∈J converges to p in the product space P(X)ω . Below we
prove p ∈ Nw(x). By the regularity of X, we only need to prove that if U is a
neighborhood of x in X, then there is m ∈ ω such that Pα(m) ⊂ U. Since P is
a k-network for X, there is γ < βω such that Pγ contains infinitely many terms
of the convergent sequence {xn }n∈J and Pγ ⊂ U. Let J1 = {n ∈ J : xn ∈ Pγ }.
Then J1 is infinite. For each n ∈ J1 , since γ ∈ Λ(xn ), there is only one natural
number kn  |Λ(K) ∩ [0, γ ]| such that γ = αxn (kn ). Since Λ(K) ∩ [0, γ ] is a finite
set, there is m ∈ ω such that J2 = {n ∈ J1 : kn = m} is infinite. Since {αn }n∈J2
converges to α, 
α (m) = αxn (m) = αxn (kn ) = γ for sufficient large n in J2 , and hence
Pα(m) = Pγ ⊂ U.
To sum up, there is a subsequence {s(xn )}n∈J of {s(xn )}n∈ω converging to p in
Nw(X). Thus, if K is a compact subset of X, then s(K) is a compact subset of
Nw(X). We have proved Claim.
To complete the proof, it remains to prove π : Nw(X) → X is a subproper
mapping. Let Z = s(X). Then Z ⊂ Nw(X) and π(Z) = π ◦ s(X) = X. If K is a
compact subset of X, then Z ∩ π −1 (K) = s(K) is a compact subset of Nw(X), so π
is subproper. Therefore, X is k ∗ -metrizable. 
2.6 Compact-Covering Mappings 105

Example 2.8.14 indicates that a space having a σ -discrete k-network may not be
a k ∗ -metrizable space, and hence the regularity in Theorem 2.6.9 is important.

Corollary 2.6.10 For every space X, the following are equivalent:

(1) X is a Lašnev space.
(2) X is a k ∗ -metrizable Fréchet–Urysohn space.
(3) X is a Fréchet–Urysohn space with a σ -compact-finite family P such that P is
a k-network for X.

Proof (1) ⇒ (2) is obtained by Theorems 2.5.8 and 2.6.9.

(2) ⇒ (3). Suppose X is a k ∗ -metrizable Fréchet–Urysohn space. By using the
notations in the proof of the necessity of Theorem 2.6.9. we first prove f (Z ∩ B) ⊂
f (B) for every B ∈ B. Suppose x ∈ f (Z ∩ B). Since X is a Fréchet–Urysohn space,
there is a sequence {bn } in Z ∩ B such that {f (bn )} converges to x. Since each bn ∈
Z ∩ f −1 ({x} ∪ {f (bn ) : n ∈ N}) and Z ∩ f −1 ({x} ∪ {f (bn ) : n ∈ N}) is a compact
subset of M, {bn } has an accumulation point b ∈ B. By the continuity of f , we have
x = f (b) ∈ f (B). So f (Z ∩ B) ⊂ f (B).
We have proved f (B|Z ) is σ -compact-finite in the proof of the necessity of The-
orem 2.6.9. It remains to prove f (B|Z ) is a k-network for X. Suppose K is compact
and U is open in X such that K ⊂ U. Then there is B ∈ B <ω such that

Z ∩ f −1 (K) ⊂ ∪B ⊂ ∪B ⊂ f −1 (U),

so f (B|Z ) ∈ [f (B|Z )]<ω and

K = f (Z ∩ f −1 (K)) ⊂ ∪f (B|Z

) ⊂ ∪f (B|Z ) ⊂ ∪f (B ) ⊂ U.

Thus, f (B|Z ) is a k-network for X.

(3) ⇒ (1). Suppose the Fréchet–Urysohn space X has a σ -compact-finite family
P such that P is a k-network for X. By Lemma 2.5.5, P is a σ -HCP closed k-
network for X. Since the proof of (3) ⇒ (1) in Theorem 2.5.8 dedicates that every
Fréchet–Urysohn space with a σ -HCP closed k-network is a Lašnev space, X is a
Lašnev space. 

Example 2.6.11 On the property of every compact set being metrizable and the
generalized sequentiality property.
(1) A space in which every compact set is metrizable may not be a k-space. The
Michael space (see Example 1.8.8) is such a space.
(2) Suppose A is the class of spaces in which every compact set is metrizable. We
have the following conclusions in the class A :
(i) There is a k ∗ -metrizable k-space which is neither a k-metrizable space nor
a Fréchet–Urysohn space. The Arens space S2 (see Example 1.8.6) is such
a space.
106 2 Mappings on Metric Spaces

(ii) There is a Fréchet–Urysohn space which is not a strongly Fréchet–Urysohn

space. The sequential fan Sω (see Example 1.8.7) is such a space.
(iii) There is a strongly Fréchet–Urysohn space which is not a first countable
space. The space X/K in Example 2.4.14 is such a space.
(iv) There is a first countable space which is not k ∗ -metrizable and does not
satisfying the condition that every compact set has a countable neighborhood
base. The butterfly space (see Example 1.8.3) is such a space. The butterfly
space X illustrates a noteworthy fact that a space X may not be a compact-
covering open image of a metric space, even if X is both a compact-covering
image of a metric space and an open image of a metric space.
(3) A space in which every compact subset has a countable neighborhood base does
not imply a space in which every compact set is metrizable. The Alexandroff
double-arrow space (see Example 1.8.9) offers such an example.

2.7 s-Mappings

The purpose of this section is to give characterizations of various classes of s-images

of metric spaces. The problem of seeking the intrinsic characterizations of quotient s-
images of metric spaces was regarded as an “important and difficult” problem posed
by Arhangel’skiı̌ [31]. Looking for the solutions of this problem and to improve them
gradually has experienced about 20 years. Until 1987, some satisfactory answers were
given by using the concept of cs∗ -networks. In this section, starting from the point-
countable families, we describe the intrinsic characterizations of quotient s-images,
pseudo-open s-images, open s-images and closed s-images of metric spaces by using
the concepts of cs∗ -networks, k-networks, bases and so on.
A cover U of a set X is said to be precisely refining V if, U = {Uα }α∈A , V =
{Vα }α∈A and Uα ⊂ Vα for every α ∈ A.
Proposition 2.7.1 Let K and P be covers of a space X, where K ⊂ K (X). If
K and P satisfy that for each K ∈ K , there is a sequence {Pn } of finite subsets
of P such that
(i) for each n ∈ N, Pn is precisely refined by a finite closed cover of K;
(ii) if {Pn }n∈N is a net at x in X for every x ∈ K and Pn ∈ (Pn )x , ∀n ∈ N,
then there are a metrizable space M and a mapping f : M → X such that
(1) for each K ∈ K , there is L ∈ K (M) such that f (L) = K;
(2) for each E ⊂ X, if (P)E is countable, then f −1 (E) has a countable base.
Proof Denote P = {Pα }α∈Λ . For each i ∈ N, let Λi be the space obtained by giving
the discrete topology to the set Λ. Define
  
M = β = (αi ) ∈ Λi : {Pαi } is a net at some point x(β) in X .
i∈N
2.7 s-Mappings 107

Then M is a metrizable space and for each β ∈ M, a point x(β) in X can be determined
by β uniquely. So we can define a function f : M → X by f (β) = x(β). By (ii), f is
a mapping. By the definition of f , (2) holds. We prove (1) holds below.
For each K ∈ K , assume that a sequence {Pn } of finite sets of P satisfies (i)
and (ii). For each n ∈ N, there exist Γn ∈ Λ<ωn and a closed cover {Kα : α ∈ Γn } of
K such that Pn = {Pα : α ∈ Γn } and Kα ⊂ Pα . Let
  
L = (αi ) ∈ Γi : Kαi = ∅ .
i∈N i∈N

  
Then L ∈ τ c ( i∈N Γi ). Because in fact,
 if (αi ) ∈ i∈N Γi − L, then i∈N Kαi = ∅,
and hence there is i0 ∈ N such that ii0 Kαi = ∅. Let
  
W = (βi ) ∈ Γi : βi = αi , i  i0 .
i∈N

 
Then (αi ) ∈ W ∈ τ ( i∈N Γi ) and  W ∩ L = ∅.  Therefore, L ∈ K ( i∈N Γi ). For
each α = (αi ) ∈ L, there is x ∈ i∈N Kαi ⊂ K ∩ i∈N Pαi , so α ∈ M and f (α) = x,
and hence L ⊂ M and f (L) ⊂ K. On the other hand, for every x ∈ K and i ∈ N, there
is αi ∈ Γi such that x ∈ Kαi . Let α = (αi ). Then α ∈ L and f (α) = x, so f (L) ⊃ K,
thus f (L) = K. 
Proposition 2.7.2 Let f : X → Y be a sequentially quotient mapping. If P is a
cs∗ -network for X, then f (P) is a cs∗ -network for Y .
Lemma 2.7.3 ([269]) If P is a point-countable cs∗ -network for X and K = S (X),
then K and P satisfy the assumptions of Proposition 2.7.1.
Proof For each K ∈ K , denote K = {x} ∪ {xn : n ∈ N}, where {xn } is a nontrivial
sequence converging to x. We first prove that for every K ⊂ U ∈ τ , there is F ∈ P <ω
with the following property, briefly referred as Φ(K, U):

K ⊂ ∪F ⊂ U and F|K ⊂ τ c − {∅}.

Let P = {P  ∈ P : x ∈ P ⊂ U} = {Pi }i∈N . If for each k ∈ N, {xn } is

not eventually
 in ik Pi , then we can take nk ∈ N for each k ∈ N such that
xnk ∈ X − ik Pi . Since P is a cs∗ -network, there exist a subsequence {yj } of {xnk }
and m ∈ N such that the sequence {yj } ⊂ Pm , a contradiction.
 As a consequence,
there is k ∈ N such that the sequence {xn } is eventually in ik Pi . Now it is easy to
see that there exists F ∈ P <ω with the property Φ(K, U).
The family {F ∈ P <ω : F has the property Φ(K, X)} is countable, and we
denote it as {Pn }n∈N . Obviously, {Pn } satisfies (i) of Proposition 2.7.1. Below we
prove {Pn } satisfies (ii) of Proposition 2.7.1.
Suppose y ∈ K and Pn ∈ (Pn )y , ∀n ∈ N. Let y ∈ V ∈ τ . If y = x, take
K1 = V ∩ K, then there is F ∈ P <ω with the property Φ(K1 , V ) and there
is F ∈ P <ω , such that, K − K1 ⊂ ∪F ⊂ X − K1 , so F ∪ F also has
108 2 Mappings on Metric Spaces

the property Φ(K, X), and hence there is i ∈ N such that F ∪ F = Pi , thus
y ∈ Pi ⊂ ∪F ⊂ V . If y = x, then there is P ∈ P such that y ∈ P ⊂ V − (K − {y}),
so there is F ∈ P <ω with the property Φ(K − {y}, X − {y}), and hence F ∪ {P}
has the property Φ(K, X), it follows that there is j ∈ N such that F ∪ {P} = Pj ,
therefore y ∈ Pj = P ⊂ V . Thus {Pn }n∈N is a net at y. 

Theorem 2.7.4 ([269]) For every space X, the following are equivalent:
(1) X has a point-countable cs∗ -network.
(2) X is a sequentially quotient s-image of a metric space.
(3) X is a sequence-covering s-image of a metric space.

Proof By Proposition 2.7.1 and Lemma 2.7.3, we get (1) ⇒ (3). (3) ⇒ (2) is obtained
by Proposition 2.1.13. (2) ⇒ (1) follows from the Nagata–Smirnov metrization
theorem and Proposition 2.7.2. 

Corollary 2.7.5 ([167, 455]) For every space X, the following are equivalent:
(1) X is a sequential space with a point-countable cs∗ -network.
(2) X is a sequentially quotient s-image of a metric space.
(3) X is a sequence-covering quotient s-image of a metric space.
(4) X is a quotient s-image of a metric space.

Proof By Theorem 2.7.4, the mapping lemma (see Proposition 2.1.12) and Propo-
sition 2.3.1 we get (1) ⇔ (2) ⇔ (3) and (2) ⇔ (4). 
The following facts should be mentioned:
(1) In (1), (3) and (4) of Corollary 2.7.5, if we replace “sequential space” and
“quotient mapping” with “Fréchet–Urysohn space” and “pseudo-open mapping”
respectively, then the conclusion still holds.
(2) There is a k-space with a point-countable cs∗ -network which is not a sequential
space. The compactification βN is such a space.

Corollary 2.7.6 ([194]) Every space with a point-countable weak base is a quo-
tient s-image of a metric space.

Now we turn to discuss relationships among spaces with a point-countable cs∗ -

network, spaces with a point-countable k-network and compact-covering quotient s-
images of metric spaces.

Definition 2.7.7 Let P be a family of sets in a space X. For A ⊂ X and F ⊂ P,

we have the following definitions:
(1) F is called a minimal cover [344] (or an irreducible cover [119]) of A if, F
covers A and A cannot be covered by any proper subset of F .
(2) F is called a minimal interior cover [89] of A if, A ⊂ (∪F )◦ and for any proper
subset H of F , A ⊂ (∪H )◦ .
2.7 s-Mappings 109

Lemma 2.7.8 [344] (The Miščenko lemma) If P is a point-countable cover of

a space X, then any subset of X has only countably many minimal finite covers
consisting of elements of P.

Proof Suppose A is a subset of X and {Pα }α∈Λ is the set of all minimal finite covers
of A consisting of elements of P. If the lemma is not true, then there is n ∈ N
such that Ψ = {Pα : α ∈ Λ, |Pα | = n} is uncountable. For each P ∈ P, take
Ψ (P) = {Pα ∈ Ψ : P ∈ Pα }. Let x1 ∈ A. Then Ψ = ∪{Ψ (P) : x1 ∈ P ∈ P}.
Since P is point-countable, there is P1 ∈ P such that x1 ∈ P1 and |Ψ (P1 )| > ℵ0 .
Then n > 1 and there is x2 ∈ A − P1 . Let Ψ (P1 , P) = {Pα ∈ Ψ (P1 ) : P ∈ Pα }.
Then Ψ (P1 ) = ∪{Ψ (P1 , P) : x2 ∈ P ∈ P}, and hence there is P2 ∈ P such
that x2 ∈ P2 = P1 and |Ψ (P1 , P2 )| > ℵ0 . Repeating the above process, we can
get a set {xi : i  n} and a family {Pi }in such that xi ∈ Pi ∈ P and when
i = j  n, Pi = Pj and |Ψ (P1 , . . . , Pn )| > ℵ0 . However |Ψ (P1 , . . . , Pn )| = 1, a
contradiction. 

Miščenko [344] proved that a compact space with a point-countable base has a
countable base. Although this result has a lot of improved forms (for example Lemma
3.1.6), the following Miščenko’s proof is worth admiring.
Suppose P is a point-countable base of a compact space X. Let

H = ∪{F ∈ P <ω : F is a minimal cover of X}.

By the Miščenko lemma, H is countable. We prove H is a base for X. Suppose

x ∈ U ∈ τ . Take P ∈ P such that x ∈ P ⊂ U. If y ∈ X−P, then there is Py ∈ P such
that y ∈ Py ⊂ X − {x}. Since X − P is compact, the cover {Py : y ∈ X − P} of X − P
has a finite subcover P . Thus, {P} ∪ P is a finite cover of X consisting of elements
of P, and hence there exists a minimal cover F of X such that F ⊂ {P} ∪ P . It
follows that P ∈ F ⊂ H and x ∈ P ⊂ U. So H is a countable base for X.

Lemma 2.7.9 Let P be a point-countable k-network in a space X. If K ∈ K (X),

then there is a sequence {Pn } of finite subfamilies of P covering K such that {Pn }
satisfies the condition (ii) of Proposition 2.7.1.

Proof By the Miščenko lemma, denote

{F ∈ P <ω : F is a minimal cover of K} by {Pn }n∈N .

For every x ∈ K and Pn ∈ (Pn )x , ∀n ∈ N, let x ∈ V ∈ τ (X). Take W ∈ τ (K)

such that x ∈ W and clK (W ) ⊂ V . Then there exist H1 , H2 ∈ P <ω such that
clK (W ) ⊂ ∪H1 ⊂ V and K − W ⊂ ∪H2 ⊂ X − {x}. So K ⊂ ∪(H1 ∪ H2 ), and
hence there is n ∈ N such that Pn ⊂ H1 ∪H2 , thus x ∈ Pn ⊂ ∪H1 ⊂ V . Therefore,
{Pn }n∈N is a net at x. 

By Lemma 2.7.9, Propositions 2.7.1 and 2.1.12 (the mapping lemma), we have
the following corollary.
110 2 Mappings on Metric Spaces

Corollary 2.7.10 ([337]) If a space X has a point-countable closed k-network, then

X is a compact-covering s-image of a metric space. Suppose further that X is a
k-space, then X is a compact-covering quotient s-image of a metric space.
Lemma 2.7.11 ([262]) Suppose f : X → Y is a quotient mapping and X is a k-
space. If B is a k-network for X and f (B) is a point-countable family of sets in Y ,
then f (B) is a k-network for Y .
Proof Let P = f (B). For a compact set K ⊂ U ∈ τ (Y ), take

H = {P ∈ P : P ∩ K = ∅, P ⊂ U}.

Then there is F ∈ H <ω such that K ⊂ ∪F . Because otherwise, for each y ∈ K,

denote (H )y = {Pi (y)}i∈N . Then there is a subset A = {yn : n ∈ N} of K such that
yn ∈/ Pi (yj ) whenever i, j < n. Take a ∈ Ad , and let L = A − {a}. Then L ∈ / τ c (Y ),
so there is C ∈ K (X) such that L ∩ f (C) ∈ / τ (Y ) (see the proof of Proposition
c

2.3.1(1)), and hence A ∩ f (C) is an infinite set. Take D = f −1 (K) ∩ C. Then there is
B ∈ B <ω such that D ⊂ ∪B ⊂ f −1 (U), and hence f (D) ⊂ ∪f (B ) ⊂ U. Since
f (D) ∩ A = f (C) ∩ A, (∪f (B )) ∩ A is an infinite set, so there is P ∈ f (B ) ∈ H <ω
containing an infinite subset of A. Let P = Pi (yj ) for some i, j ∈ N. Then there is
n > i, j such that yn ∈ Pi (yj ), a contradiction. Thus P is a k-network for Y . 
Theorem 2.7.12 Suppose X is a locally compact metric space. If f : X → Y is
a quotient s-mapping, then f is a compact-covering mapping and there is a point-
countable closed k-network for Y .
Proof Let B be a σ -locally finite base for X satisfying B ⊂ K (X) and let P =
f (B). Then P is a point-countable family of closed sets in Y . By Lemma 2.7.11, P
is a k-network for Y . If K ∈ K (Y ), then there is P ∈ P <ω such that K ⊂ ∪P ,
<ω
so there is F ∈ B such that f (F ) = P , and hence there is L ∈ K (X) such that
f (L) = K. Thus f is a compact-covering mapping. 
Nagami [369] proved that every quotient L-mapping on a locally compact para-
compact space is a compact-covering mapping. Below we construct an example to
show Corollary 2.7.10 is not reversible.

Lemma 2.7.13 Let P be a point-countable family of sets in a strongly Fréchet–

Urysohn space X. If K (X) refines P F , then (P F )◦ covers X.
Proof Denote (P)C = {Pi (C)}i∈N for each countable set C in X. If there is x ∈
X − ∪(P F )◦ , then by the Fréchet–Urysohn property,we can choose a sequence {Cn }
of countable subsets of X such that C1 = {x}, x ∈ n∈N C n and Cn ∩ Pi (Cj ) = ∅
whenever i, j < n. So each element of P only meets finitely many Cn . By the strongly
Fréchet–Urysohn property, there exist a subsequence {Cnk } of {Cn } and xk ∈ Cnk
such that xk → x. Since {x} ∪ {xk : k ∈ N} ∈ K (X), there is an element P in P
containing infinitely many terms of {xk }, and hence P meets infinitely many Cn , a
contradiction. 
2.7 s-Mappings 111

Example 2.7.14 ([437, 492]) A compact-covering open image of a separable metric

space which is a non-regular space with a countable base and has no point-countable
closed k-network.
Let X = R. Define the pointed irrational extension topology on X as follows: for
each x ∈ X, an element of a neighborhood base of x has the form {x} ∪ (P ∩ U),
where U is a Euclidean neighborhood of x in R. The space X is called the pointed
irrational extension topological space X of R. Obviously, X has a countable base.
Since Q is a closed subspace of X and V = X for any V with Q ⊂ V ∈ τ (X), X is
not a regular space. By Proposition 2.4.4, X is a compact-covering open image of a
separable metric space.
Suppose there is a point-countable closed k-network P for X. Define H =
{P ∈ P : P◦ = ∅}. Then K (X) refines H F . In fact, for each K ∈ K (X), let
V = (K ∩ Q) ∪ P. Since Q is a discrete closed subspace of X, K ⊂ V ∈ τ , so there
is F ∈ P <ω such that K ⊂ ∪F ⊂ V . If (∪F )◦ = ∅, then there is an open interval
J in R such that J ∩ P ⊂ ∪F , so Q ∩ J ⊂ Q ∩ J ∩ P ⊂ Q ∩ V = Q ∩ K, and hence
Q ∩ J is a finite set, a contradiction. Thus (∪F )◦ = ∅, so F ∈ H <ω and K ⊂ ∪F .
Hence K (X) refines H F . By Lemma 2.7.13, X = ∪(H F )◦ , which contradicts the
definition of H . Consequently, X has no point-countable closed k-network.
In the second part of this section, we search for intrinsic characterizations of open
s-images of metric spaces.

Lemma 2.7.15 ([89]) If P is a point-countable family of sets in a Fréchet–Urysohn

space X, then any subset of X has only countably many finite minimal interior covers
consisting of elements of P.

Proof For each F ∈ P <ω , define

H (F ) = {H ⊂ X : F is a finite minimal interior cover of H}.

Let A ⊂ X. If there are uncountably many F ∈ P <ω such that A ∈ H (F ), then

there exist m ∈ N and an uncountable subset Ψ of P <ω such that |F | = m and
A ∈ H (F ) whenever F ∈ Ψ . Suppose R is a maximal subset of P such that
R ⊂ F for uncountably many F ∈ Ψ . Then 0  |R| < m and A ⊂ (∪R)◦ . Pick
x ∈ A − (∪R)◦ . Then x ∈ X − ∪R. Since X is a Fréchet–Urysohn space, there is a
countable subset L of X − ∪R such that x ∈ L. Let Ω = {F ∈ Ψ : R ⊂ F }. If
F ∈ Ω, then x ∈ (∪F )◦ , so L meets some element of F . By the point-countability
of P and the uncountability of Ω, there is P ∈ P such that L ∩ P = ∅ and
there are uncountably many elements of Ω containing P. Then P ∈/ R and there are
uncountably many F in Ω (hence in Ψ ) containing R ∪ {P}, which contradicts the
maximality of R. 

The assumption “Fréchet–Urysohn space” in Lemma 2.7.15 can be replaced by

the “space of countable tightness” weaker than the “Fréchet–Urysohn space” [89].
A space X is said to be of countable tightness if, for every A ⊂ X and x ∈ A ⊂ X,
there is a countable subset C of A such that x ∈ C.
112 2 Mappings on Metric Spaces

Proposition 2.7.16 ([89]) Suppose there is a point-countable family P of sets in

a space X satisfying that for every x ∈ U ∈ τ , there is F ∈ (P)<ω x such that
x ∈ (∪F )◦ ⊂ ∪F ⊂ U. Then X has a point-countable base.

Proof For each F ∈ P <ω , let

H (F ) = {H ⊂ X : F is a finite minimal interior cover of H},

V (F ) = (∪(H (F ) ∩ P))◦ , and V = {V (F ) : F ∈ P <ω }.

First, V is a point-countable family of sets in X. Because if x ∈ V (F ), then there

is A ∈ H (F ) ∩ P such that x ∈ A. Since X is a first countable space, by Lemma
2.7.15, A ∈ H (F ) for only countably many F ∈ P <ω . Since (P)x is countable,
(V )x is also countable. Second, V is a base of X. Because for every x ∈ U ∈ τ ,
there is F ∈ P <ω such that x ∈ (∪F )◦ ⊂ U. We may assume that F is a minimal
interior cover of {x}. Take B ∈ (P)<ω ◦
x such that x ∈ (∪B) ⊂ ∪B ⊂ (∪F ) . If
◦
◦
B ∈ B, then B ∈ H (F ), so (∪B) ⊂ V (F ), and hence x ∈ V (F ) ⊂ U. Thus V
is a point-countable base for X. 

A mapping f : X → Y is said to be countable-to-one [167] if every f −1 (y) is a

countable subset of X.

Theorem 2.7.17 For every space X, the following are equivalent:

(1) X has a point-countable base.
(2) X is a compact-covering open s-image of a metric space [340]. (The Michael-
Nagami theorem.)
(3) X is an open s-image of a metric space [401].
(4) X is a countable-to-one open image of a metric space [309].
(5) X is a countably bi-quotient s-image of a metric space [123].

Proof (1) ⇒ (2). Suppose X has a point-countable base. By the Miščenko lemma,
every compact set in X has a countable outer base. Further by Proposition 2.4.4, X
is a compact-covering open s-image of a metric space.
(2) ⇒ (3) and (4) ⇒ (5) are obvious.
(3) ⇒ (4). Let f : M → X be an open s-mapping, where M is a metric  space. For
each x ∈ X, let Dx be a countable dense subset of f −1 (x). Define D = x∈X Dx and
g = f|D : D → X. Then g is a countable-to-one mapping. Below we prove g is an
open mapping. Let U be an open subset of D. Then there is an open subset V of M
such that U = V ∩ D. If g(U) is not open in X, then there is x ∈ g(U) ∩ X − g(U).
Since X is first countable, there is a sequence {xn } in X − g(U) converging to x.
Since x ∈ f (V ) and f (V ) is open in X, there is xm ∈ f (V ), so f −1 (xm ) ∩ V = ∅,
and hence Dxm ∩ V = ∅. Pick zm ∈ Dxm ∩ V ⊂ U. Then xm = g(zm ) ∈ g(U),
a contradiction.Thus g is an open mapping, it follows that X is a countable-to-one
open image of a metric space.
(5) ⇒ (1). Suppose M is a metric space and f : M → X is a countably bi-quotient
s-mapping. Let B be a point-countable base for M. Then f (B) is a point-countable
2.7 s-Mappings 113

family of sets in X satisfying the assumption of Proposition 2.7.16, and hence X has
a point-countable base. 

Corollary 2.7.18 For every space X, the following are equivalent:

(1) X has a point-countable base.
(2) X is a Fréchet–Urysohn space with a point-countable weak base.
(3) X is a strongly Fréchet–Urysohn space with a point-countable cs∗ -network.

Proof (1) ⇒ (2) is obvious, (2) ⇒ (3) can be obtained by Corollaries 1.6.18 and
1.6.20, and (3) ⇒ (1) follows from Corollary 2.7.5, the mapping lemma (see Propo-
sition 2.1.12) and Theorem 2.7.17. 

Corollary 2.7.19 [122, 123] (The Filippov theorem) The point-countable base prop-
erty is invariant under perfect mappings or countably bi-quotient s-mappings.

Proof Suppose f : X → Y is a mapping and X has a point-countable base. By

Theorems 2.7.17 and 2.6.1, every compact space with a point-countable base is
metrizable. Therefore, if f is a perfect mapping, then f is a countably bi-quotient s-
mapping. It was proved by (4) ⇒ (1) of Theorem 2.7.17 that the point-countable
base property is invariant under countably bi-quotient s-mapping, so Y has a point-
countable base. 

Let f : X → Y be a mapping. f is called a bi-quotient mapping [123, 171, 332] (or

limit lifting mapping [171]) if, for each y ∈ Y and each open family U in X covering
f −1 (y), there is P ∈ U F such that y ∈ f (P)◦ . Filippov [123] proved the bi-quotient s-
mapping case for Corollary 2.7.19. Obviously, every open mapping is a bi-quotient
mapping, and every bi-quotient mapping is a countably bi-quotient mapping.
In 1973, Michael and Nagami [340] asked the following question: Is every quo-
tient s-image of a metric space a compact-covering quotient s-image of a metric
space? In 1999, H. Chen [101] answered this question negatively. In 2003, Chen
[102] answer this question negatively once again by constructing a regular coun-
terexample under the assumption that there exists a σ -set. These two examples are
quite complex and readers can refer to the reference papers listed.
In the third part of this section, we investigate intrinsic characterizations of closed
s-images of metric spaces. We first prove a property of closed mappings of metric
spaces, and Theorem 3.4.16 is a more general form of this result.

Lemma 2.7.20 ([284]) Suppose f : X → Y is a quotient mapping, where X is a

sequential space and Y contains a closed copy of Sω1 (resp. Sω ). If f is a peripheral
L-mapping (resp. peripherally compact mapping), then X contains a closed copy of
S2 or Sω1 (resp. Sω ).

Proof Suppose Y contains a closed copy of Sω1 and f is a peripheral L-mapping. We

may assume that Y = {b} ∪ α<ω1 Yα is a copy of Sω1 , where every sequence Yα in
Y − {b} converges to b and {Yα : α < ω1 } is a family of pairwise disjoint sets. Since
114 2 Mappings on Metric Spaces

each Yα is not closed in Y , f −1 (Yα ) is not closed in X. Since X is a sequential space,

there is a sequence Tα in f −1 (Yα ) converging to some point xα ∈ X − f −1 (Yα ), so
xα ∈ ∂f −1 (b).
 Let L = {xα : α < ω1 }.
Claim. α<ω1 Fα is a closed discrete subset of X, where each Fα is a finite subset
of Tα .  −1
 In fact, since α<ω1 f (Fα ) is a closed discrete subset of  Y , {f (y) : y ∈
α<ω1 f (Fα )} is a closed discrete family in X, and hence α<ω1 Fα is a closed
discrete subset of X.
If L is not a closed discrete subset of X, then since X is a sequential space, there
is a sequence {xαn } consisting of different elements of L which converges to some
point a ∈ X − L. Let

M = {a} ∪ {xαn : n ∈ N} ∪ (∪{Tαn : n ∈ N}).

By Claim, M is a sequentially closed subset of X homeomorphic to S2 , so M is closed

in X. Thus, X contains a closed copy of S2 .
If L is a closed discrete subset of X, then since ∂f −1 (b) is a Lindelöf space, L is
also a Lindelöf space, so L is countable, and hence there exist ω1 sequences Tα such
that all of them converge to xα . We may assume that L = {a}. Let

S = {a} ∪ (∪{Tα : α < ω1 }).

By Claim and the assumption that X is a sequential space, S is a closed copy of Sω1
in X.
For the case of Sω , by using the fact that f is a peripherally compact mapping,
similarly we can prove that X contains a closed copy of S2 or Sω . 
Lemma 2.7.21 ([454]) If f : X → Y is a closed mapping and X is a metric space,
then f is a peripherally compact mapping (resp. peripheral L-mapping) if and only
if Y contains no closed copy of Sω (resp. Sω1 ).
Proof Our proof is only for the peripheral L-mappings case. Since any metrizable
space does not contain a closed copy of S2 or Sω1 , we get the necessity by Lemma
2.7.20.
Conversely, if f is not a peripheral L-mapping, then there is s ∈ Y such that
∂f −1 (s) is not a Lindelöf subspace of X, and hence there exist a discrete closed
set {xα : α < ω1 } in ∂f −1 (s) and a discrete family {Dα }α<ω1 of open sets in X such
that xα ∈ Dα . If s ∈ V ∈ τ (Y ), then f −1 (V ) ∩ (Dα − f −1 (s)) = ∅ for any α < ω1 ,
it follows that V ∩ (f (Dα ) − {s}) = ∅, and hence s ∈ f (Dα ) − {s}. Since Y is a
Fréchet–Urysohn space, there is a subset Eα = {yα,n : n ∈ N} ⊂ f (Dα ) − {s} such
that the sequence {yα,n }n∈N converges to the point s. Since {f (Dα )}α<ω1 is an HCP
family of sets in Y , {Eα ∪ {s}}α<ω1 is also an HCP family in Y . For each α < ω1 , by
Lemma 2.5.4, there exist Fα ∈ Eα<ω and Λα ∈ ω1<ω such that (Eα − Fα ) ∩ Eβ = ∅
whenever β ∈ ω1 − Λα . Let Kα = Eα − Fα . By the transfinite induction method,
 subset Γ of ω1 such that {Kα }α∈Γ is a disjoint family.
one can choose an uncountable
Then the subspace {s} ∪ ( α∈Γ Kα ) of Y is a closed copy of Sω1 . 
2.7 s-Mappings 115

A family P of subsets in a space X is said to be locally countable if every point

of X has a neighborhood which only meets at most countably many elements of P.

Theorem 2.7.22 For every space X, the following are equivalent:

(1) X is a closed s-image of a metric space.
(2) X is a Fréchet ℵ-space [150].
(3) X is a Lašnev space containing no closed copy of Sω1 [454].

Proof By Theorem 2.5.8 and Example 1.8.7 we obtained (2) ⇒ (3), and (3) ⇒ (1)
is followed from Lemmas 2.7.20 and 2.1.15. Below we prove (1) ⇒ (2).
Suppose M is a metric space and f : M → X is a closed s-mapping. Obviously,
X is a paracompact Fréchet–Urysohn space. Let B be a σ -locally finite base for M
and let P = f (B). By Propositions
 2.1.16 and 2.5.7, P is a σ -locally countable
k-network for X. Denote P = i∈N Pi , where Pi is a locally countable family
of sets and Pi ⊂ Pi+1 . For each i ∈ N, since Pi is locally countable and X is
paracompact, there is a locally finite open cover Ui of X such that each element
of Ui only meets countably many sets  in Pi , and hence Pi ∧ Ui is a σ -locally
finite family of sets. We prove that i∈N (Pi ∧ Ui ) is a k-network for X. Let K
be a compact set and K ⊂ U ∈ τ (X). Then there exist m ∈ N, P ∈ Pm<ω and
U ∈ Um<ω such that K ⊂ ∪P ⊂  U and K ⊂ ∪U , so P ∧ U ∈ (Pm ∧ Um )<ω
and K ⊂ ∪(P ∧ U ) ⊂ U. Thus i∈N (Pi ∧ Ui ) is a k-network for X, and hence
X is an ℵ-space. 

A space X is said to be bi-sequential [336] if, whenever a filter base F accumulates

at x in X, then there is a decreasing sequence {An } in X converges to x and every An
intersects every F ∈ F . An ω-filter base is defined to be a family F of nonempty
subsets of X such that for any countable subfamily F of F , there is F ∈ F such that
F ⊂ ∩F . A space X is said to be weakly bi-sequential [304] if, for any ω-filter base
B in X accumulating at some x ∈ X there is a countable filter base C in X such that C
converges to x and every B ∈ B intersects every C ∈ C . Every be-sequential space
is weakly be-sequential, and every weakly bi-sequential space is Fréchet–Urysohn.
C. Liu [304] proved that a Lašnev space X is a closed s-image of a metric space if
and only if X is weakly bi-sequential.

2.8 ss-Mappings

As a special case of the s-images of metric spaces, we introduce in this section

the ss-images of metric spaces. In 1996, C. Liu and M. Dai [308] found for the
first time a characterization of quotient s-images of locally separable metric spaces.
Seeking more concise characterizations of this kind images is still an open problem
[311]. The class of ss-mappings plays an exclusive role in characterizing images of
locally separable metric spaces. This section consists mainly four parts. The first
two parts are about images of separable metric spaces and quotient ss-images of
metric spaces respectively. The third part is about pseudo-open ss-images or closed
ss-images of metric spaces and about pseudo-open s-images or closed s-images of
116 2 Mappings on Metric Spaces

locally separable metric spaces. In the last part, we describe some kinds of additional
conditions associated with s-images and ss-images by means of several examples.

Definition 2.8.1 ([255]) Let f : X → Y be a mapping. Then f is called an ss-

mapping if for each y ∈ Y , there is an open neighborhood V of y in Y such that
f −1 (V ) is a separable subspace of X.
First, we give a characterization of ss-images of metric spaces. Note that if a
mapping f : X → Y is an ss-mapping, then X is a locally separable space.

Proposition 2.8.2 ([21, 22]) Let f : X → Y be a mapping. If P is a network for X,

then f (P) is a network for Y .

Thus, every regular image of a cosmic space is a cosmic space [21, 22].

Theorem 2.8.3 ([288]) A space X has a locally countable network if and only if X
is an ss-image of a metric space.

Proof Let K = {{x} : x ∈ X} and let P be a locally countable network for X. Then
by Proposition 2.7.1 we can get the necessity. Conversely, let X be the image of a
metric space M under an ss-mapping f . If B is a σ -locally finite base in M, then
f (B) is a network in X. For each x ∈ X, there is an open neighborhood V of x such
that f −1 (V ) is a separable subspace of M, so f −1 (V ) only meets countably many
elements of B, and hence V only meets countably many elements of f (B). Thus
f (B) is a locally countable network for X. 

Corollary 2.8.4 ([331]) A regular space X is a cosmic space if and only if X is an

image of a separable metric space.

Next, we give characterizations of quotient ss-images of metric spaces. A family

U of sets is said to be star-countable if, for each U ∈ U , (U )U is countable.

Lemma  2.8.5 ([177]) If U is a star-countable family of sets in a space X, then

U = α∈Λ Uα , where each Uα is countable and (∪Uα ) ∩ (∪Uβ ) = ∅ whenever
α, β ∈ Λ and α = β.

Proof For every A, B ∈ U , we said that A ∼ B if there are finitely many sets {Ui }in
in U such that A = U1 , B = Un , and Ui ∩ Ui+1 = ∅ when i < n. For each A ∈ U ,
let UA = {B ∈ U : A ∼ B}. Then UA is countable, U = ∪{UA : A ∈ U } and for
each pair A, B of U , (∪UA ) ∩ (∪UB ) = ∅ if and only if UA = UB . 

Theorem 2.8.6 ([255]) For every space X, the following are equivalent:
(1) X has a locally countable cs∗ -network.
(2) X has a locally countable cs-network.
(3) X is a sequentially quotient ss-image of a metric space.
(4) X is a sequence-covering ss-image of a metric space.
(5) X is a compact-covering ss-image of a metric space.
If further assume that X is a regular space, then each item above is also equivalent
to the following:
2.8 ss-Mappings 117

(6) X has a locally countable k-network.

Proof (1) ⇒ (2). Let P be a locally countable cs∗ -network for X. For each x ∈ X,
there is an open neighborhood Vx of x such that Vx only meets countably many
elements of P. Let U = {P ∈ P : there is some Vx ⊃ P}. Then U is star-countable.
By Lemma 2.8.5, we can suppose U = α∈Λ Uα , where each Uα  is countable and
(∪Uα ) ∩ (∪Uβ ) = ∅ whenever α, β ∈ Λ and α = β. Let F = α∈Λ UαF . Then
F is also locally countable. Below we prove that F is a cs-network for X. Suppose
{xn } is a sequence in X converging to a point x ∈ V ∈ τ . By Lemma 2.7.3, there
exist m ∈ N and P ∈ (P)<ω x such that {x} ∪ {xn : n  m} ⊂ ∪P ⊂ V ∩ Vx .
Then there is only one α ∈ Λ such that x ∈ ∪Uα , so ∪P ∈ UαF , and hence F is a
cs-network for X.
(2) ⇒ (5). Let K = K (X) and let P be a locally countable cs-network for
X. We only need to prove K and P satisfy the conditions of Proposition 2.7.1.
For each K ∈ K , since P|K is a countable cs-network for K, K is metrizable. Let
P = (P)K . Since

{F ∈ P <ω : F is precisely refined by an finite closed cover of K}

is countable, we can denote it by {Pn }n∈N . We prove {Pn } satisfies Proposition

2.7.1(ii). Suppose x ∈ K and Pn ∈ (Pn )x , ∀n ∈ N. If x ∈ V ∈ τ (X), then there is an
open neighborhood W of x in K such that clK (W ) ⊂ V . Let {Vi }i∈N be a decreasing
local base of x in K. Put

Px = {P ∩ K : P ∈ P , P ⊂ V and there is i ∈ N such that Vi ⊂ P ∩ K}.

By the proof of Proposition 1.6.21, Px is a neighborhood base of x in K. So there

is Px ∈ P such that x ∈ intK (Px ∩ K) ⊂ Px ⊂ V , and hence there is Vx ∈ τ (K)
such that x ∈ Vx ⊂ clK (Vx ) ⊂ intK (Px ∩ K). Obviously, the compact set K − Vx ⊂
X − {x} ∈ τ (X). As proved above, for each y ∈ K − Vx , there exist Py ∈ P and
Vy ∈ τ (K) such that y ∈ Vy ⊂ clK (Vy ) ⊂ intK (Py ∩ K) ⊂ Py ⊂ X − {x}. Now
the open cover {Vy : y ∈ K − Vx } of K − Vx has a finite subcover {Vyk }km . Let
F = {Px } ∪ {Pyk : k  m}. Then F is precisely refined by a finite closed cover of K,
so there is j ∈ N such that F = Pj , thus x ∈ Pj = Px ⊂ V . Consequently, {Pn }n∈N
is a net at x.
(5) ⇒ (4) ⇒ (3) is obvious. By Theorem 1.3.2 and Proposition 2.7.2, we get
(3) ⇒ (1). (1) ⇒ (6) follows from Proposition 1.6.7 and Corollary 2.2.20. Finally, if
a regular space X has a locally countable k-network, then X has a locally countable
closed k-network, and hence X has a locally countable cs∗ -network. 

Corollary 2.8.7 ([331]) A regular space X is an ℵ0 -space if and only if X is a

compact-covering (resp. sequence-covering, sequentially quotient) image of a sep-
arable metric space.

Example 2.8.8 ([465]) There is a connected ℵ0 -space which is not an image of a

connected metric space.
118 2 Mappings on Metric Spaces

Fix p ∈ βR − R and let X = R ∪ {p}. Take X as a subspace of the compactifica-

tion βR. Then X is a connected space. By Example 2.3.19, there is no sequence in R
converging to p, so X has a countable cs-network, and hence X is an ℵ0 -space. Since
X is not an s-connected space, by Theorem 2.3.17, X is not an image of a connected
metric space.

Lemma 2.8.9 ([259]) Every k-space with a σ -locally countable cs∗ -network is a
meta-Lindelöf sequential space.

Proof Suppose X is a k-space and P is a σ -locally countable cs∗ -network for X.

Then for each K ∈ K (X), P|K is a countable network for K, and hence K is
metrizable. By Corollary
 2.3.5, X is a sequential space.
Suppose P = n∈N Pn is closed under finite intersections, where Pn is locally
countable and Pn ⊂ Pn+1 . We first prove that for any locally countable family
F = {Fα }α∈Γ in X, there is a point-countable family W = {Wα }α∈Γ of open sets
of X such that Fα ⊂ Wα (for simplicity, we say that W is a point-countable open
expansion of F ). Let

Λ = {λ : λ is a finite sequence consisting of elements of N}.

For each λ ∈ Λ, we inductively define a locally countable family F (λ) =

{Fα (λ)}α∈Γ of sets in X as follows.
Let F (∅) = F and Fα (∅) = Fα . Suppose a locally countable family F (λ) has
been defined. For each n ∈ N, denote the finite sequence consisting of members of
N obtained by adding one term n behind λ by λn. Take

P(λn) = {P ∈ Pn : Fα (λ) ∩ P = ∅ only for countably many α ∈ Γ };

Fα (λn) = ∪{P ∈ P(λn) : Fα (λ) ∩ P = ∅}, α ∈ Γ ;
F (λn) = {Fα (λn) : α ∈ Γ }.

For each x ∈ X, there is an open neighborhood V of x such that V only meets

countably many elements of Pn . Take

(Pn )V ∩ P(λn) = {Pi }i∈N .

For each i ∈ N, there is a countable set Γi in Γ such that Pi ∩ Fα (λ) = ∅ whenever

α ∈ Γ − Γi . Hence

|{α ∈ Γ : there is i ∈ N such that Pi ∩ Fα (λ) = ∅}|  ℵ0 .

It means that V ∩ Fα (λn) = ∅ holds only for countably many α in Γ , so F (λn) is

a locally countable
 family.
Let Wα = λ∈Λ Fα (λ) for each α ∈ Γ . Then Fα ⊂ Wα and Wα is a sequentially
open set in X. Because in fact, if a sequence {xn } converges to x ∈ Wα , then there
is λ ∈ Λ such that x ∈ Fα (λ), and hence there is an open neighborhood W of x
2.8 ss-Mappings 119

such that (F (λ))W is countable. By the proof of Lemma 2.7.3, there exist k ∈ N
and H ∈ (Pk )<ω x such that the sequence {xn } is eventually in ∪H ⊂ W . Since
∪H ⊂ Fα (λk) ⊂ Wα , {xn } is eventually in Wα , so Wα is a sequentially open set in
X. Thus Wα is an open set in X, because X is a sequential space.
Furthermore, take W = {Wα }α∈Γ . Then W is point-countable. Because otherwise,
there exist x ∈ X and an uncountable subset Γ of Γ such that x ∈ Wα whenever
α ∈ Γ . For each α ∈ Γ , there is λα ∈ Λ such that x ∈ Fα (λα ), and hence there
exist an uncountable subset Γ of Γ and λ ∈ Λ such that λα = λ for each α ∈ Γ ,
and it means x ∈ Fα (λ), which contradicts the point-countability of F (λ).
Finally, we prove that X isa meta-Lindelöf space. Let U be any open cover of
X. Then U has a refinement i∈N Fi , where each Fi = {Fα : α ∈ Γi } is a locally
countable family of sets of X. For each i ∈ N, suppose Wi = {Wα : α ∈ Γi } is a
 expansion of Fi . For each α ∈ Γi , choose Uα ∈ U such that
point-countable open
Fα ⊂ Uα . Then i∈N {Uα ∩ Wα : α ∈ Γi } is a point-countable open refinement of
U . Hence X is a meta-Lindelöf space. 

Lemma 2.8.10 Every locally separable meta-Lindelöf space is the topological sum
of separable Lindelöf spaces.

Proof If X is a locally separable meta-Lindelöf space, then X has a point-countable

open cover U consisting of separable subspaces. Since every point-countable family
of open sets in a separable space is countable, U is a star-countable family. By
Lemma 2.8.5, X has a cover {Xα }α∈Λ consisting of 
disjoints open sets such that each
Xα is a separable subspace of X, and hence X = α∈Λ Xα . Since every separable
meta-Lindelöf space is a Lindelöf space, each Xα is a separable Lindelöf space. 

Corollary 2.8.11 ([255, 259]) For every space X, the following are equivalent:
(1) X is a k-space with a locally countable cs∗ -network.
(2) X is the topological sum of sequential spaces with a countable cs-network.
(3) X is a compact-covering quotient (sequentially quotient or sequence-covering)
ss-image of a metric space.
(4) X is a quotient ss-image of a metric space.
If we assume further that X is a regular space, then each term of the above is
equivalent the following:
(5) X is a k-space with a locally countable k-network.

Proof By Theorem 2.8.6, Lemmas 2.8.9 and 2.8.10 we obtain (1) ⇔ (2). (1) ⇒
(3) follows from Theorem 2.8.6 and the mapping lemma. (3) ⇒ (4) is obvious. By
Propositions 2.3.1, 2.1.12 (the mapping lemma) and 2.7.2, (4) ⇒ (1) holds. Under
the assumption of regular spaces, (1) ⇔ (5) follows from Theorem 2.8.6. 

In the third part of this section, we give characterizations of pseudo-open ss-images

or closed ss-images of metric spaces. We also give characterizations of pseudo-
open s-images or closed s-images of locally separable metric spaces.

Lemma 2.8.12 For every space X, the following are equivalent [167]:
(1) X is a Fréchet–Urysohn space with a countable cs∗ -network.
120 2 Mappings on Metric Spaces

(2) X is a pseudo-open image of a separable metric space.

(3) X is both a separable space and a pseudo-open s-image of a metric space.
If further assume that X is a regular space, then each item above and each of
the following are equivalent [130]:
(4) X is a Fréchet–Urysohn ℵ0 -space.
(5) X is a closed image of a separable metric space.
Proof By Proposition 2.7.1, Lemma 2.7.3 and the mapping lemma (see Proposition
2.1.12), we get (1) ⇒ (2). (5) ⇒ (2) ⇒ (3) is obvious. By Lemma 2.7.11, (2) ⇒ (4)
holds. Below we prove (3) ⇒ (1) and (4) ⇒ (5).
(3) ⇒ (1). By Proposition 2.3.1, X is a Fréchet–Urysohn space. By Corollary
2.7.5, X has a point-countable cs∗ -network P. Let D be a countable dense subset of
X and let H = (P)D . Then H is countable. We prove that H is a cs∗ -network for
X. Let {xn } be a sequence converging to x ∈ U ∈ τ in X. Define

S = {x} ∪ {xn : n ∈ N},

H = {H ∈ H : x ∈ H ⊂ U}.

Below we show H = ∅. Denote H = {Hi }i∈N . 

We first prove there is m ∈ N such that x ∈ intS (( im Hi ) ∩ S). Otherwise,
there is a sequence {zm } satisfying zm → x and zm ∈ S − im Hi . Note that,
 
( im Hi ) ∩ S is a closed set in X, we get zm ∈ D − ( im Hi ) ∩ S, and hence there

is a sequence {zm,k }k converging to zm in D − ( im Hi ). Thus x ∈ {zm,k : m, k ∈ N},
it follows that there is a sequence in {zm,k : m, k ∈ N}, we denote it as {zmj ,kj }j∈N ,
converging to x when mj → +∞. We may assume there is P ∈ P such that
{x} ∪ {zmj ,kj : j ∈ N} ⊂ P ⊂ U, it means P ∈ H , and hence H = ∅. Take i ∈ N
such that P = Hi . Pick j ∈ N such that mj   i. Then zmj ,kj ∈
/ P, a contradiction.
Thus, there is m ∈ N such that x ∈ intS (( im Hi ) ∩ S), and hence there exist a
subsequence {xnj } of S and i ∈ N such that {x} ∪ {xnj : j ∈ N} ⊂ Hi ⊂ U. So H is a
countable cs∗ -network for X.
(4) ⇒ (5). By Theorem 2.7.22, there exist a metric space M and a closed s-mapping
f : M → X. Since X is a Lindelöf space and f is a closed L-mapping, M is a Lindelöf
space. Thus X is a closed image of a separable metric space. 
Note that the meta-Lindelöf property is invariant under pseudo-open s-mappings
(see Corollary A.5.3 in Appendix A). By Corollary 2.8.11, Lemmas 2.8.10 and
2.8.12, we have the following theorem.
Theorem 2.8.13 ([255, 259]) For every space X, the following are equivalent:
(1) X is a Fréchet–Urysohn space with a locally countable cs∗ -network.
(2) X is a pseudo-open s-image of a locally separable metric space.
(3) X is a pseudo-open ss-image of a metric space.
(4) X is both a locally separable space and a pseudo-open s-image of a metric space.
If further assume that X is a regular space, then each item above and each of
the following are equivalent:
2.8 ss-Mappings 121

(5) X is a closed s-image of a locally separable metric space.

(6) X is both a locally separable space and a closed s-image of a metric space.

Example 2.7.14 shows that the assumption of regularity in Lemma 2.8.12 and The-
orem 2.8.13 is important. The next example shows that the assumption of regularity
in Theorem 2.8.6 and Corollary 2.8.11 is also necessary.

Example 2.8.14 There is a space X with a locally countable and σ -discrete k-

network, such that, X is neither a meta-Lindelöf space nor a k ∗ -metrizable space,
and X has no point-countable cs∗ -network.
Let X be the half-disc topological space (see Example 2.2.22(3)). It is easy to verify
that X is a separable first countable space but X is not a Lindelöf space. That shows X
is not a meta-Lindelöf space, so X has no point-countable base. By Corollary 2.7.18,
X has no point-countable cs∗ -network. Since X is not a regular space, by Corollary
2.6.10, X is not a k ∗ -metrizable space.
For every x ∈ R2 , r > 0, let B(x, r) be a spherical neighborhood of x in R2 .
Define
P = {{p} : p ∈ L} ∪ {B(q, 1/n) ∩ S : q ∈ Q × Q, n ∈ N}.

Since L is a discrete closed set in X, P is a locally countable and σ -discrete family

of sets in X. We prove P is a k-network for X. Assume that K ⊂ U ∈ τ , where
K is a compact set in X and τ is the half-disc topology. For each x ∈ X, denote
{P ∈ P : x ∈ P ⊂ U} = {Pi (x)}i∈N . Then K is covered by some finite subfamily of
{Pi (x) : x ∈ K, i ∈ N}. Because otherwise, there is a sequence {pn } in K such that
pn ∈/ Pi (pj ) whenever i, j < n. So there is a subsequence {pnk } of {pn } converging
to p ∈ K. Since L is discrete, we may assume all pnk ∈ S, and hence {pnk } also
converges to p in the Euclidean subspace topology τ ∗ on X. Since {B(q, 1/n) ∩ X :
q ∈ Q × Q, n ∈ N} is a countable base of τ ∗ , there exist q ∈ Q × Q and h, m ∈ N
such that {p} ∪ {pnk : k  h} ⊂ B(q, 1/m) ∩ X ⊂ U, and hence {pnk : k  h} ⊂
B(q, 1/m) ∩ S ⊂ U. Thus, there exist i, j ∈ N such that B(q, 1/m) ∩ S = Pi (pj ), and
hence we can pick n > i, j such that pn ∈ Pi (pj ), a contradiction. Therefore, P is a
k-network for X.

Example 2.8.15 There is a compact-covering open s-image of a metric space which

is not metrizable.
It was proved in Example 1.8.5 that the Michael line X is a Lindelöf regular space
with a point-countable base. By Theorem 2.7.17, X is a compact-covering open s-
image of a metric space. But X is not a β-space, and hence it is neither a metrizable
space nor an ℵ0 -space. This example shows that the assumption of the separability
in Lemma 2.8.12(3) and Theorem 2.8.13(4) can not be replaced with the Lindelöf
property.

Example 2.8.16 ([167]) There is a compact-covering finite-to-one quotient mapping

f : M → X such that M is a locally compact metrizable space M, X is a separable
regular space but X is not a meta-Lindelöf space.
122 2 Mappings on Metric Spaces

Let
X = I × S1 , Y = I × (S1 − {0}).

Define a topology for X as follows: Y is a Euclidean subspace of X and each element

of a neighborhood base of (t, 0) ∈ X has the form

{(t, 0)} ∪ (∪{V (t, k) : k  n}), n ∈ N,

where V (t, k) is an open neighborhood of (t, 1/k) in the subspace I × {1/k}. Let

M = (⊕{I × {1/n} : n ∈ N}) ⊕ (⊕{{t} × S1 : t ∈ I}).

Then M is a locally compact metrizable space. Let f : M → X be the natural mapping.

Since X has the weak topology with respect to the point-finite cover {I × {1/n} : n ∈
N} ∪ {{t} × S1 : t ∈ I}, by Proposition 2.3.3, f is a finite-to-one quotient mapping.
By Theorem 2.7.12, f is a compact-covering mapping.
Obviously, X is a separable regular space. Since I × {0} is an uncountable discrete
closed subspace of X, X is not a Lindelöf space, and hence X is not a meta-Lindelöf
space. By Lemma 2.8.9, X has no σ -locally countable cs∗ -network.
Example 2.8.16 shows the following facts:
(1) The assumption “pseudo-open mapping” in Lemma 2.8.12(3) and Theorem
2.8.13(4) cannot be reduced to “quotient mapping”.
(2) A finite-to-one compact-covering quotient image of a locally compact metric
space may not have a σ -locally countable cs∗ -network.
(3) Not every quotient s-image of a locally separable metric space is a quotient ss-
image of a metric space.
Example 2.8.17 ([259]) There is a regular space with a locally countable k-network
which is not an ℵ-space.
Let X = ω1 × S1 . Define a topology for X as follows: each point of X − (ω1 × {0})
 element of a neighborhood base of a point (α, 0) ∈ X has
is isolated in X and each
the form {(α, 0)} ∪ ( nm (V (α, n) × {1/n})), where m ∈ N and V (α, n) is an open
neighborhood of α in ω1 with the ordered topology. Then X is a regular space. Let

P = {{x} : x ∈ X} ∪ {{α} × ({0} ∪ {1/n : n  m}) : α < ω1 , m ∈ N}.

Then P is a locally countable family of sets in X. Suppose K ∈ K (X). Then

(17.1) for each s ∈ S1 , K ∩ (ω1 × {s}) is a finite set;
(17.2) if (α, 0) ∈ ω1 × {0} − K, then (α, 1/n) ∈ K only for finitely many n ∈ N;
(17.3) the set K − ∪{{α} × S1 : (α, 0) ∈ K ∩ (ω1 × {0})} is a finite set.
Consequently, P is a locally countable k-network for X.
Take f = π1 : X → ω1 and give ω1 the ordered topology. Then f is a pseudo-
open compact mapping. If X is an ℵ-space, then ω1 is a subparacompact space (see
Proposition 3.4.14), a contradiction. Thus X is not an ℵ-space.
2.8 ss-Mappings 123

Example 2.8.17 shows that in Lemma 2.8.9 and Corollary 2.8.11 the assumption
“k-space” cannot be omitted. Sakai [416] constructed a locally countable regular
space X in which every compact set is a finite set and X is not a countably metacompact
space. This space has a locally countable k-network and it is not a perfect space.

2.9 π-Mappings

In the above two sections, we introduced some mappings with separable fibers. In
this section, we turn to discuss π -mappings which is closely relevant to mappings
with compact fibers. Our purpose is to extend the concept of weak developments,
to explore the characterizations of quotient π -images, pseudo-open π -images and
open π -images of metric spaces and to characterize symmetrizable spaces and semi-
metrizable spaces satisfying the weak Cauchy condition by means of such images
of metric spaces.

Definition 2.9.1 ([401]) Let (X, d) be a metric space. A mapping f : X → Y is

called a π -mapping if d(f −1 (y), X − f −1 (U)) > 0 whenever y ∈ U ∈ τ (Y ).
Obviously, every compact mapping on a metric space is a π -mapping. For each
space X, let M be a space obtained by giving X the discrete topology. Then idM :
M → X is a compact mapping, hence a π -image on the metric space X. Thus it is
necessary for us to add certain additional conditions when discussing the π -mappings
on metric spaces.
Definition 2.9.2 Let P be a cover of a space X.
(1) P is called a cfp-cover of X [486] if, for each K ∈ K (X), there is F ∈ P <ω
such that F is precisely refined by a finite closed cover of K.
(2) P is called an fcs-cover of X [153] if, for any sequence S converging to a point
x in X, there is P ∈ (P)<ω
x such that S is eventually in ∪P .

∗
(3) P is called a cs -cover of X [243] if, for any convergent sequence S in X, there
is P ∈ P such that some subsequence of S is eventually in P.
fcs-covers were called wcs-covers by Y. Ge [152]. It is easy to verify that for every
space X, open cover ⇒ cfp-cover ⇒ fcs-cover ⇒ cs∗ -cover.
Lemma 2.9.3 Let P be an fcs-cover of a space X. If S ∈ S (X), then there is F ∈
P <ω such that F is precisely refined by a finite closed cover of S.
Proof Suppose S is a sequence converging to x ∈ X. Then there is P ∈ (P)<ω x
such that S is eventually in ∪P . Since S − ∪P is a finite set and P ∩ S is a closed
set for each P ∈ P , there is F ∈ P <ω such that F is precisely refined by a finite
closed cover of S. 
Proposition 2.9.4 Suppose f : X → Y is a mapping and (X, d) is a metric space.
For each n ∈ N, define Un = {f (B(x, 1/n)) : x ∈ X}.
124 2 Mappings on Metric Spaces

(1) If f is a π -mapping, then {Un } is a point-star network for Y [297].

(2) If f is a compact-covering mapping, then Un is a cfp-cover of Y [293].
(3) If f is a sequence-covering mapping, then Un is an fcs-cover of Y [152].
(4) If f is a sequentially quotient mapping, then Un is a cs∗ -cover of Y [297].

Proof (1) For every y ∈ U ∈ τ (Y ), there is n ∈ N such that d(f −1 (y), X −

f −1 (U))  1/n. Take m = 2n. If y ∈ f (B(x, 1/m)), then f −1 (y) ∩ B(x, 1/m) =
∅. If B(x, 1/m) ⊂ f −1 (U), then d(f −1 (y), X − f −1 (U)) < 2/m = 1/n, a
contradiction. So B(x, 1/m) ⊂ f −1 (U), and hence f (B(x, 1/m)) ⊂ U, it follows
that st(y, Um ) ⊂ U. Thus {Un } is a point-star network for Y .
(2) For each K ∈ K (Y ), there is L ∈ K (X) such that f (L) = K, and hence there
is a finite subfamily F of {B(x, 1/n)}x∈X such that F is precisely refined by a
finite closed cover L of L. Then the finite subfamily f (F ) of Un is precisely
refined by the finite closed cover f (L ) of K, and hence Un is a cfp-cover of Y .
(3) Suppose S is a sequence converging to some point y in Y . Then there is L ∈ K (X)
such that f (L) = S ∪ {y}, and hence there is a finite set F in X such that
f −1 (y)∩L ⊂ x∈F B(x, 1/n). Then U = {f (B(x, 1/n))}x∈F is a finite subfamily
of Un and S is eventually in ∪U . Because otherwise, there is a subsequence {yk }
of S with {yk } ⊂ Y −∪U . For each k ∈ N, there is xk ∈ L − x∈FB(x, 1/n) such
that f (xk ) = yk . If a is an accumulation point of {xk }, then a ∈
/ x∈F B(x, 1/n),
and hence f (a) = y, a contradiction.
(4) When f is a sequentially quotient mapping, it is easy to verify that Un is a
cs∗ -cover of Y . 

Proposition 2.9.5 Let {Un } be a point-star network for a space X. Then there are a
metric space (M, d) and a π -mapping f : M → X such that
(1) if {Un } is a sequence of open covers of X, then f is an open mapping [179];
(2) if {Un } is a sequence of cfp-covers of X, then f is a compact-covering mapping
[294];
(3) if {Un } is a sequence of fcs-covers of X, then f is a sequence-covering mapping
[152];
(4) if {Un } is a sequence of cs∗ -covers of X, then f is a sequentially quotient mapping
[297];
(5) if E is a subset of X and each (Un )E is countable, then f −1 (E) has a countable
base.

Proof It is easy to verify that for each x ∈ X with x ∈ Ui ∈ Ui (∀i ∈ N), {Ui }i∈N
is a net at x. For each i ∈ N, let Ui = {Uα }α∈Λi and let each Λi be a discrete space.
Define
  
M = α = (αi ) ∈ Λi : {Uαi } is a net at some point x(α) ∈ X .
i∈N

Then M is a metrizable space, and the metric d defined in the following way is a
compatible metric on M: for every pair α, β ∈ M,
2.9 π -Mappings 125

0, α = β,
d(α, β) =
max{1/k : πk (α) = πk (β), k ∈ N}, α = β.

Define f : M → X by f (α) = x(α). Then f is a mapping. For every x ∈ U ∈ τ (X),

there is n ∈ N such that st(x, Un ) ⊂ U. For every α ∈ f −1 (x) and β ∈ M with
d(α, β) < 1/n, we have πi (α) = πi (β) whenever i  n, so x ∈ Uπn (α) = Uπn (β) , and
hence
f (β) ∈ Uπi (β) ⊂ Uπn (β) ⊂ U.
i∈N

As a consequence, d(f −1 (x), M − f −1 (U))  1/n, thus f is a π -mapping.

(1) If {Un } is a sequence of open covers of X, then for every n ∈ N and αi ∈ Λi (∀i 
n), we have
f (B(α1 , . . . , αn )) = ∩{Uαi : i  n},

where B(α1 , . . . , αn ) = {β ∈ M : πi (β) = αi , i  n}. So f is an open mapping.

(2) If {Un } is a sequence of cfp-covers of X, then for every K ∈ K (X) and n ∈ N,
there is Pn ∈ Pn<ω such that Pn = {Pα }α∈Γn is precisely refined by the closed
cover {Kα }α∈Γn of K, where Γn ⊂ Λn . We may assume each Kα is not empty.
Take   
L = (αn ) ∈ Γn : Kαn = ∅ .
n∈N n∈N

Similar to the proof of Proposition 2.7.1(1), we can prove that L is a compact

set in M and f (L) = K. So f is a compact-covering mapping.
(3) If {Un } is a sequence of fcs-covers of X, then for each S ∈ S (X), by Lemma
2.9.3 and the proof of (2), there is a compact set L in M such that f (L) = S. So
f is a sequence-covering mapping.
(4) Suppose {Un } is a sequence of cs∗ -covers of X and {xn } is a nontrivial sequence
converging to x0 ∈ X. Since U1 is a cs∗ -cover of X, there exist a subsequence T1
of {xn } and α1 ∈ Λ1 such that T1 is eventually in Uα1 . By the inductive method,
for each i ∈ N, we can choose Ti and αi ∈ Λi such that Ti+1 is a subsequence
of Ti and Ti is eventually in Uαi , and hence Ti ⊂ ki Uαk . Pick xni ∈ Ti and
βi ∈ f −1 (xni ) such that ni < ni+1 and πk (βi ) = αk whenever k  i. Then
lim πk (βi ) = αk . Let β0 = (αi ). Then the sequence {βi } converges to β0 in M,
i→∞
and hence f is a sequentially quotient mapping.
By the definition of f , we can obtain (5) easily. 

Theorem 2.9.6 For every space X, the following are equivalent:

(1) X is a developable space.
(2) X is a compact-covering open π -image of a metric space [300].
(3) X is an open π -image of a metric space [179].
126 2 Mappings on Metric Spaces

Definition 2.9.7 ([29]) A symmetrizable space (X, d) is said to satisfy the weak
Cauchy condition if each convergent sequence {xn } in X has a Cauchy subsequence
{xni }, i.e. for any ε > 0, there is k ∈ N such that d(xni , xnj ) < ε whenever i, j > k.
It is easy to verify that for any symmetrizable space (X, d) and any convergent
sequence {xn } in X, {xn } has a Cauchy subsequence if and only if for any ε > 0, there
is a subsequence {xni } of {xn } such that d(xni , xnj ) < ε for every i, j ∈ N.

Lemma 2.9.8 ([78, 456]) Let (X, d) be a symmetrizable space. Then d satisfies the
weak Cauchy condition if and only if for any F ⊂ X, if F is not closed in X, then for
any ε > 0, there is x, y ∈ F with x = y such that d(x, y) < ε. The above condition
is equivalent to that X is a sequential space with a cs∗ -cover point-star network.

Proof (8.1) Suppose (X, d) is a symmetrizable space satisfying the weak Cauchy
condition. By Proposition 1.6.16, X is a sequential space. For each n ∈ N, take
Un = {A ⊂ X : diamA < 1/n}. Then for each x ∈ X, st(x, Un ) = B(x, 1/n). So
{Un } is a point-star network in X. For each n ∈ N and any sequence {xk } converging to
some point x in X, there is a Cauchy subsequence {xki } such that d(x, xki ) < 1/(n+1)
for every i ∈ N, and hence there is m ∈ N such that d(xki , xkj ) < 1/(n + 1) whenever
i, j  m. Let An = {x} ∪ {xki : i  m}. Then An ∈ Un . Thus Un is a cs∗ -cover of X.
(8.2) Let {Un } be a cs∗ -cover point-star network in the sequential space X, and
we may assume Un+1 refines Un . We first prove that st(x, Un ) is a sequential neigh-
borhood of x for every x ∈ X and n ∈ N. Otherwise, there is a sequence {xm } in
X − st(x, Un ) converging to x, and hence there is a subsequence {xmi } eventually
in some U ∈ Un , so xmi ∈ U ⊂ st(x, Un ), a contradiction. Since X is a sequential
space, {st(x, Un )}n∈N is a weak base of x. By the sufficiency of Proposition 1.6.14,
there is a symmetric d on X such that st(x, Un ) = B(x, 1/2n ) for every x ∈ X and
n ∈ N. Let F ⊂ X and ε > 0. Take m ∈ N such that 1/2m < ε. If F is not a closed
set in X, then there is a sequence {xn } in F converging to x ∈/ F, so there is U ∈ Um
such that some subsequence of {xn } is eventually in U, and hence there are x, y ∈ F
with x = y such that d(x, y) < 1/2m < ε.
(8.3) Let (X, d) be a symmetrizable space satisfying that if F ⊂ X and F is
not closed in X, then for any ε > 0, there is a pair x, y of F with x = y such that
d(x, y) < ε. Suppose ε > 0 and {xn } is a nontrivial convergent sequence in X. If every
subsequence {yn } of {xn } has a subsequence {zn } such that d(z1 , zn )  ε whenever
n > 1, then there is a subsequence {an } of {yn } such that d(an , am )  ε whenever
n = m, a contradiction. So there is a subsequence {yn } of {xn } such that d(z1 , zn ) < ε
for every subsequence {zn } of {yn }. Thus, we can choose a subsequence {an } of {xn }
such that d(an , am ) < ε, which shows that {xn } is a Cauchy subsequence. 

Theorem 2.9.9 ([232]) A space X is a quotient π -image of a metric space if and

only if X is a symmetrizable space satisfying the weak Cauchy condition.

Proof By Lemma 2.9.8, Propositions 2.1.12 (the mapping lemma) and 2.9.5, we get
the sufficiency. The necessity is obtained by Proposition 2.9.4. 
2.9 π -Mappings 127

Theorem 2.9.10 ([485]) A space X is a compact-covering quotient π -image of a

metric space if and only if X has a weak development consisting of cfp-covers.

Proof By Propositions 2.9.5 and 2.1.12 (the mapping lemma), we get the sufficiency.
The necessity follows from Proposition 2.9.4. 

Example 2.9.11 ([232]) There is a symmetrizable space which does not satisfy the
weak Cauchy condition.
Take X = R. Define a symmetric d on X as follows: for every x, y ∈ X,

1, x, y ∈ P and x = y,
d(x, y) =
|x − y|, otherwise.

Give X the symmetric topology generated by d. Then for every x ∈ X and n ∈ N,


(x − 1/n, x + 1/n), x ∈ Q,
B(x, 1/n) =
{x} ∪ ((x − 1/n, x + 1/n) − P), x ∈ P.

Hence,
(11.1) any convergent sequence in the subspace P is trivial;
(11.2) if x ∈ Q, A ⊂ X and x is an accumulation point of A with respect to the
Euclidean topology τ ∗ on R, then x is also an accumulation point of A in X.
For each n ∈ N and any symmetric ρ on X, define

Dn = {x ∈ P : ρ(x, P − {x})  1/n}.


If there is x ∈ P − n∈N Dn , then for each n ∈ N, there is xn ∈ P − {x} such
that ρ(x, xn ) < 2/n, and hence xn → x ∈ P, which contradicts (11.1). Thus
P = n∈N Dn . Since (R, τ ∗ ) is the second category, there is m ∈ N such that
intτ ∗ (clτ ∗ (Dm )) = ∅. Pick x ∈ Q ∩ clτ ∗ (Dm ). By (11.2), x ∈ Q ∩ Dm , so Dm is not a
closed set in X. By Lemma 2.9.8, (X, ρ) does not satisfy the weak Cauchy condition.
For the above X, by Theorem 2.3.6, there exist a metric space M and a quotient
mapping f : M → X. By Theorem 2.9.9, f is not a π -mapping with respect to any
compatible metric on M. However, this is not the same case for semi-metrizable
spaces.

Lemma 2.9.12 ([78]) Every semi-metrizable space has a compatible semi-metric

satisfying the weak Cauchy condition.

Proof Let (X, d) be a semi-metrizable space. For each r ∈ R+ , define

B(r) = {B ⊂ X : d(x, y)  r for every x, y ∈ B with x = y}.

128 2 Mappings on Metric Spaces

For every x, y ∈ X, let

A(x, y) = {z ∈ X : there is B ∈ B(d(x, y)/2) such that x, y ∈ B and z ∈ B},

ρ(x, y) = inf{d(x, z) + d(z, y) : z ∈ A(x, y)}.

(12.1) ρ is a semi-metric on X.
Obviously, ρ is a symmetric on X and for every x, y ∈ X, ρ(x, y)  d(x, y),
and hence Bd (x, ε) ⊂ Bρ (x, ε) for any x ∈ X and ε > 0. If there is a sequence
{xn } in X such that xn ∈ Bρ (x, 1/n)−Bd (x, ε), take εn = d(x, xn ), then εn  ε.
Since ρ(x, xn ) < 1/n, there is yn ∈ A(x, xn ) such that d(x, yn ) + d(yn , xn ) <
1/n, and hence there is Bn ∈ B(εn /2) such that x, xn ∈ Bn , yn ∈ Bn and
d(x, yn ) < 1/n. Choose a sequence {yn,i }i converging to yn in Bn . Then there
is i ∈ N such that yn,i ∈ Bd (x, ε/2), which means d(x, yn,i ) < ε/2. Since
yn,i ∈ Bn ∈ B(εn /2) and x ∈ Bn , d(x, yn,i )  εn /2  ε/2, a contradiction.
Thus, there is n ∈ N such that Bρ (x, 1/n) ⊂ Bd (x, ε). So ρ is a semi-metric
on X.
(12.2) (X, ρ) satisfies the weak Cauchy condition.
For any ε > 0, F ⊂ X, if F ∈ / τ c (X), then pick z ∈ F − F and take

B = F ∩ Bd (z, ε/2), ε1 = inf{d(x, y) : x, y ∈ B with x = y}.

Then B ∈ B(ε1 ) and ε1  ε. Choose x, y ∈ B such that 0 < d(x, y) < 2ε1 .
Then B ∈ B(d(x, y)/2) and z ∈ B, and hence z ∈ A(x, y). So ρ(x, y) 
d(x, z) + d(z, y) < ε. By Lemma 2.9.8, (X, ρ) satisfies the weak Cauchy
condition. 
Example 2.9.13 ([78]) There is a semi-metrizable space (X, d) such that d does not
satisfy the weak Cauchy condition.
Let X = A ∪ B, where

A = {(0, a) ∈ R2 : |a|  1}, B = {(b, sin(1/b)) : 0 < b  1}.

Give X the Euclidean topology. Define a symmetric d on X as follows: let x =

(x1 , x2 ), y = (y1 , y2 ) ∈ X; if x ∈ A or y ∈ A, take

d(x, y) = (x1 − y1 )2 + (x2 − y2 )2 ;

if x, y ∈ B, then d(x, y) is the arc length in B between x and y. Then d is a semi-metric

compatible to the topology of X. Since no Cauchy sequence in B converging to a
point of A, any sequence in B converging to a point of A has no Cauchy subsequence
with respect to d, and hence d does not satisfy the weak Cauchy condition. 
Theorem 2.9.14 ([1, 78]) For every space X, the following are equivalent:
(1) X is a semi-metrizable space.
(2) X is a countably bi-quotient π -image of a metric space.
2.9 π -Mappings 129

(3) X is a pseudo-open π -image of a metric space.

Proof By Lemma 2.9.12, Theorem 2.9.9 and the mapping lemma (see Proposition
2.1.12), we get (1) ⇒ (2). (2) ⇒ (3) is obvious. By Theorems 2.9.9 and 1.2.8, (3) ⇒
(1) holds. 

Corollary 2.9.15 Metrizability is invariant under closed π -mappings.

Question 2.9.16 (1) Is every semi-metrizable space a sequence-covering π -image

of a metric space?
(2) Is every semi-metrizable space a compact-covering π -image of a metric space?

Example 2.9.17 ([277]) There exist a metric space (X, d) and a countably bi-
quotient π -mapping f : (X, d) → S1 such that f is not a sequence-covering mapping.
Let A be a maximal almost disjoint family of N (see Example 1.8.4) and denote
the uncountable family A as {Aα }α∈Γ . Let Bα = {α} ∪ Aα for each α ∈ Γ . Define a
symmetric dα on Bα as follows: for every x, y ∈ Bα ,
⎧
⎨ 0, x = y,
dα (x, y) = 1/y, x = y, x = α,
⎩
|1/x − 1/y|, x = y, x = α, y = α.

Then (Bα , dα ) is a metric space. Suppose X = α∈Γ Bα and d is a standard topo-
logical sum metric on X. Define a function f : X → S1 by

0, x ∈ Γ,
f (x) =
1/x, x ∈
/ Γ.

(17.1) f is a mapping.
It is obvious that f −1 (y) = ⊕{1/y : 1/y ∈ Aα } is a clopen set in X for
each y ∈ S1 − {0}. If U is a neighborhood of 0 in S1 , then for each α ∈ Γ ,
f −1 (U) ∩ Bα is an open set in Bα , so f −1 (U) ∈ τ (X).
(17.2) f is a countably bi-quotient mapping.
By the mapping lemma (see Proposition 2.1.12), we only need to prove that
f is a quotient mapping. Suppose U ⊂ S1 and f −1 (U) ∈ τ (X). For each
y ∈ U, we may assume y = 0. If U is not a neighborhood of y, then there
is an infinite subset M of N such that 1/n ∈ / U for each n ∈ M. If M ∈ A ,
then there is α ∈ Γ such that Bα = {α} ∪ M. Since f −1 (U) is a neighborhood
of α, the sequence Bα is eventually in f −1 (U), so the sequence {1/n}n∈M is
eventually in U, a contradiction. Therefore, M ∈ / A , so there is α ∈ Γ such
that M ∩ Aα is an infinite set, and hence the sequence {x : x ∈ M ∩ Aα } is
eventually in f −1 (U), a contradiction. Thus, U is a neighborhood of y, and
hence f is a quotient mapping.
(17.3) f is a π -mapping.
Otherwise, there exist z ∈ S1 and an open neighborhood U of z such that
d(f −1 (z), X − f −1 (U)) = 0, and hence there exist sequences {zn } and {xn }
130 2 Mappings on Metric Spaces

in X such that zn ∈ f −1 (z), xn ∈ X − f −1 (U) and d(zn , xn ) < 1/n. Then

f (zn ) = z and f (xn ) ∈
/ U for every n ∈ N. So there is αn ∈ Γ such that
xn , zn ∈ Bαn and dαn (zn , xn ) < 1/n. Hence |f (zn ) − f (xn )| < 1/n, thus
f (xn ) → z, a contradiction.
(17.4) f is not a sequence-covering mapping.
Otherwise, there is a compact set K in X such that f (K) = S1 . By the
compactness of K, there is Γ ∈ Γ  <ω
such that K ⊂ α∈Γ Bα . Pick β ∈
Γ − Γ . Then there is n0 ∈ Aβ − ( α∈Γ Aα ) ⊂ Aβ − K, and hence there
does not exist x0 ∈ K such that f (x0 ) = 1/n0 , a contradiction.

2.10 Compact Mappings

In this section, we investigate characterizations of quotient compact images, pseudo-

open compact images and open compact images of metric spaces on the basis of
π -mappings of metric spaces. We also characterize developable spaces and meta-
compact developable spaces by means of such images.

Theorem 2.10.1 ([277]) Every sequentially quotient compact mapping on a metric

space is a sequence-covering mapping.

Proof Suppose f : X → Y is a sequentially quotient compact mapping and X is a

metric space. Let {yn } be a nontrivial sequence converging to a point y0 in Y . Take

S1 = {y0 } ∪ {yn : n ∈ N}, X1 = f −1 (S1 ) and g = f|X1 .

Then g is also a sequentially quotient compact mapping. By the mapping lemma (see
Proposition 2.1.12), g is a pseudo-open mapping. Let {Un }n∈N be a decreasing neigh-
borhood base of the compact set g −1 (y0 ) in the metric space X1 . Then for each n ∈ N,
y0 ∈ g(Un )◦ , so there is in ∈ N such that for each i  in , yi ∈ g(Un ), and hence
g −1 (yi ) ∩ Un = ∅. We may assume 1 < in < in+1 . For each j ∈ N, take

f −1 (yj ), j < i1 ,
xj ∈
f −1 (yj ) ∩ Un , in  j < in+1 .

Let K = g −1 (y0 ) ∪ {xj : j ∈ N}. Then K is a compact set in X1 and g(K) = S1 , so

f (K) = S1 . Thus, f is a sequence-covering mapping. 

Corollary 2.10.2 ([277]) Every quotient compact mapping on a metric space is a

sequence-covering mapping.

Example 2.10.3 ([339]) There is a countably bi-quotient compact mapping from a

separable metric space onto a metric space which is not a compact-covering mapping.
2.10 Compact Mappings 131

Let Z = I × I and A = {A ∈ K (Z) : π1 (A) = I}. Then |A | = 2ℵ0 , and hence

there is a one-to-one onto function μ : I → A . If s ∈ I, then s ∈ π1 (μ(s)), so
μ(s) ∩ π1−1 (s) = ∅. Pick xs ∈ μ(s) ∩ π1−1 (s). Let Os be an open interval in {s} × I of
length 1/4 containing xs , and let X = {(s, t) ∈ Z : (s, t) ∈
/ Os }. Then X is a separable
metric space. Define f = π1|X . Then f : X → I is a compact mapping.
(3.1) f is not a compact-covering mapping.
Otherwise, there is a compact set B in X such that f (B) = I. Then B ∈ A , so
there is s ∈ I such that B = μ(s), and hence xs ∈ μ(s) − X, thus B ⊂ X, a
contradiction.
(3.2) f is a countably bi-quotient mapping.
By the mapping lemma (see Proposition 2.1.12), we only need to prove that f is
a sequence-covering mapping. Suppose {sn } is a sequence converging to some
point s0 in I. Fix two points (s0 , t1 ), (s0 , t2 ) in f −1 (s0 ) such that |t1 − t2 | = 1/2.
Let K = {si : i ∈ ω} and C = f −1 (K) ∩ (K × {t1 , t2 }). Then C is a compact
set in X and f (C) = K. So f is a sequence-covering mapping.

Definition 2.10.4 A sequence {Un } of covers of a space X is called a point-finite

weak development of X [209] (resp. point-finite semi-development [1], point-finite
development) if, {Un } is a weak development (resp. semi-development, development)
of X and each Un is a point-finite cover of X.

Lemma 2.10.5 ([276]) Suppose {Un } is a sequence of point-finite covers of a space

X such that Un+1 refines Un . Then {Un } is a cs∗ -cover point-star network for X if
and only if {Un } is a point-star sequential neighborhood network for X.

Proof By (8.2) of Lemma 2.9.8, if {Un } is a cs∗ -cover point-star network for X, then
{Un } is a point-star sequential neighborhood network for X. Conversely, suppose
{Un } is a point-star sequential neighborhood network for X. For each n ∈ N and any
nontrivial sequence {xk } converging to x in X, if m  n and st(x, Um ) = {x}, pick
zm ∈ st(x, Um ) − {x}, then there is i ∈ N such that

st(x, Ui ) ⊂ X − {zm : m  n and st(x, Um ) = {x}}.

Since {xk } is eventually in st(x, Ui ), i > n and {xk } is eventually in st(x, Un ). Since
Un is point-finite, there is a subsequence of {xk } eventually in some element of Un .
So Un is a cs∗ -cover of X. 

Theorem 2.10.6 For every space X, the following are equivalent:

(1) X has a point-finite weak development.
(2) X is a sequentially quotient and quotient compact image of a metric space [484].
(3) X is a quotient compact image of a metric space [209, 263].
(4) X is a sequence-covering quotient compact image of a metric space [484].

Proof (1) ⇒ (2). Let {Un } be a point-finite weak development for X. For each i ∈ N,
let Ui = {Uα }α∈Λi . By using the notations of Proposition 2.9.5, we can define a
132 2 Mappings on Metric Spaces

metric space M and a mapping  f : (M, d) → X. For every x ∈ X and i ∈ N, take

Γ
i = {α ∈ Λ i : x ∈
 αU }. Then Γ
i∈N i is a compact set in  Λi . If α =−1(αi ) ∈
i∈N
Γ
i∈N i , then x ∈ U
i∈N αi , so α ∈ M and f (α)
 = x, and hence i∈N Γi ⊂ f (x). If
α = (αi ) ∈ f −1 (x),
 then x ∈ U
i∈N αi , so α ∈ Γ
i∈N i , and hence f −1
(x) ⊂ i∈N Γi .
−1
Thus f (x) = i∈N Γi , it follows that f is a compact mapping. By Lemma 2.10.5
and Proposition 2.9.5, f is a sequentially quotient mapping.
(2) ⇔ (3) ⇔ (4) follows from the mapping lemma (see Proposition 2.1.12) and
Corollary 2.10.2. We prove (3) ⇒ (1). Let X be an image of a metric space M
under a quotient compact mapping f . By Theorem 1.3.5, there is a sequence {Bi }
of locally finite open covers of M, such that, Bi+1 refines Bi and {st(K, Bi )}i∈N is
a neighborhood base of K in M for each K ∈ K (M). Let Ui = f (Bi ). Then Ui
is a point-finite cover of X. Below we prove {Ui } is a weak development for X. If
x ∈ U ∈ τ (X), then there is n ∈ N such that st(f −1 (x), Bn ) ⊂ f −1 (U), and hence
st(x, Un ) ⊂ U. On the other hand, if a subset U of X satisfying that for each x ∈ U,
there is n ∈ N such that st(x, Un ) ⊂ U, then for any z ∈ f −1 (U), there is n ∈ N
such that st(f (z), Un ) ⊂ U, so st(z, Bn ) ⊂ f −1 (U), and hence f −1 (U) ∈ τ (M), thus
U ∈ τ (X). Therefore, {Ui } is a point-finite weak development for X. 

Theorem 2.10.7 ([15]) For every space X, the following are equivalent:
(1) X is a sequence-covering quotient compact image of a separable metric space.
(2) X is a quotient π -image of a separable metric space.
(3) X is a g-second countable symmetric space.

Proof (1) ⇒ (2) is obvious. By Corollary 2.8.11, Theorems 2.9.9 and 1.6.22 we get
(2) ⇒ (3).
Now suppose X is a g-second countable symmetric space. Let

P= Px = {Pn : n ∈ N}
x∈X

is a weak base for X, where each Px is a weak base of x in X. Let d be a symmetric

on X. For every m, n ∈ N, define

Am,n = {x ∈ X : B(x, 1/n) ⊂ Pm }, Bm,n = X − Am,n and Fm,n = {Pm , Bm,n }.

(7.1) Fm,n is an fcs-cover of X.

Suppose L = {xi } is a sequence in X converging to a point x ∈ X. We may
assume x ∈/ Pm ∩ Bm,n . If x ∈ Am,n , then B(x, 1/n) ⊂ Pm , so L is eventually in
Pm ∈ Fm,n . If x ∈ / Am,n , then x ∈ Bm,n , so x ∈
/ Pm , and hence L is eventually
in Bm,n ∈ Fm,n . Because otherwise, there is a subsequence {xik } of L such that
each xik ∈ Am,n , so xik ∈ B(xik , 1/n) ⊂ Pm . Since {xik } converges to x, we may
assume d(x, xik ) < 1/n for every k ∈ N, and hence x ∈ B(xik , 1/n) ⊂ Pm , a
contradiction.
(7.2) {Fm,n }m,n∈N is a point-star network for X.
Suppose x ∈ U ∈ τ (X). Since Px is a weak base of x in X, there is m0 ∈ N such
2.10 Compact Mappings 133

that Pm0 ∈ Px and Pm0 ⊂ U. If B(x, 1/n) ⊂ Pm0 for each n ∈ N, then there is
a sequence {xn } in X such that xn ∈ B(x, 1/n) − Pm0 for each n ∈ N, so {xn }
converges to x, which contradicts the fact that Pm0 is a sequential neighborhood
of x. Thus, there is n0 ∈ N such that B(x, 1/n0 ) ⊂ Pm0 , so x ∈ Am0 ,n0 , and
hence st(x, Fm0 ,n0 ) = Pm0 ⊂ U. Consequently, {Fm,n }m,n∈N is a point-star
network for X.
By using the notations in Proposition 2.9.5, from (7.1) and (7.2), we know there
exist a metrizable space M and a sequence-covering mapping f : (M, ρ) → X.
Since each Fm,n is finite, M is a separable metrizable space. By the mapping
lemma (see Proposition 2.1.12) and the proof of Theorem 2.10.6, f is a compact
quotient mapping. 

Corollary 2.10.8 For every regular space X, the following are equivalent:
(1) X is a (compact-covering and) quotient compact image of a separable metric
space.
(2) X is a quotient π -image of a separable metric space.
(3) X is a g-second countable space.

Proof (1) ⇒ (2) is obvious. (2) ⇒ (3) is obtained by Theorem 2.10.7.

(3) ⇒ (1). Suppose B is a countable
 weak base for X and we may assume
B ⊂ τ c (X). Let B = {Bi }i∈N = x∈X Bx , where Bx is a weak base of x. For each
i ∈ N, define
Ci = {x ∈ X : Bi ∈/ Bx }, Ui = {Bi , Ci }.

Then Ui is a cfp-cover of X. Because in fact, for any K ∈ K (X), take

K1 = Bi ∩ K and K2 = K − Bi .

Then K1 ⊂ Bi and K = K1 ∪ K2 . Since K is metrizable, for each x ∈ K2 , there is a

sequence {xn } converging to x ∈ K in K − Bi , so Bi ∈ / Bx , and hence x ∈ Ci . Thus
K2 ⊂ Ci . If x ∈ X, then
⎧
⎨ Bi , Bi ∈ Bx ,
st(x, Ui ) = X, Bi ∈ / Bx , x ∈ Bi ,
⎩
Ci , Bi ∈
/ Bx , x ∈
/ Bi .

So {st(x, Ui )}i∈N is a weak base of x, which shows {Ui } is a weak development for X.
By Proposition 2.9.5 and Theorem 2.10.6, X is a compact-covering quotient compact
image of a separable metric space. 

Corollary 2.10.9 Consider the following conditions:

(1) X is a (compact-covering and) quotient compact ss-image of a metric space.
(2) X is a quotient π -ss-image of a metric space.
(3) X has a locally countable weak base.
(4) X is a topological sum of g-second countable spaces.
134 2 Mappings on Metric Spaces

(5) X is a g-first countable space with a locally countable cs∗ -network.

(6) X is a g-first countable space with a locally countable k-network.
Then (1) ⇒ (2) ⇒ (3) ⇔ (4) ⇔ (5) ⇒ (6). If further assume that X is a regular
space, then (6) ⇒ (1).

Proof (1) ⇒ (2) is obvious. By Corollary 2.8.11 and Theorem 2.9.9, we get
(2) ⇒ (5). (5) ⇒ (3) follows from Corollary 2.8.11 and Proposition 1.6.21. By
Corollary 2.8.11, (3) ⇒ (4) holds. (4) ⇒ (5) ⇒ (6) is obvious. When X is a regular
space, (6) ⇒ (1) can be obtained by Corollaries 2.8.11 and 2.10.8. 

Example 2.8.14 shows the assumption of regularity in the above corollary is

necessary.

Example 2.10.10 There is a g-first countable space with a locally countable k-

network which is not a quotient π -image of a metric space [276].
Let X be the pointed irrational extension topological space of R (see Example
2.7.14). We use the notations in Example 2.7.14 and first prove (X, τ ) is not a
countably metacompact space. Denote the Euclidean topology for X by τ ∗ and let
Q = {rn : n ∈ N}. Take Fn = {ri : i  n} for each n ∈ N. Since Q is a discrete closed
set in X, {Fn }n∈N is a decreasing sequence of closed sets in X and n∈N Fn = ∅. If
X is a countably metacompact space, then there is a sequence {G n }n∈N of open sets
in X such that Fn ⊂ G n and n∈N G n = ∅ (see Proposition A.2.8 in Appendix A).
For every n ∈ N and x ∈ Fn , there is Un,x ∈ τ ∗ such that

x ∈ Un,x , Un,x ∩ {ri : i < n} = ∅ and {x} ∪ (P ∩ Un,x ) ⊂ G n ,

so Un,x = (Un,x ∩ Q) ∪ (Un,x ∩ P) ⊂ G n , and hence there  is On ∈ τ ∗ such that

∗
Fn ⊂ On ⊂ G n . Thus On is a dense open subset of (R, τ ) and n∈N On = ∅, which
contradicts the fact that τ ∗ is a Baire topology. Consequently, X is not a countably
metacompact space, so X is not a subparacompact space either (see Proposition
A.4.11 in Appendix A).
If X is a quotient π -image of a metric space, then by the mapping lemma (see
Proposition 2.1.12) and Theorem 2.9.14, X is a semi-metrizable space, and hence X
is a subparacompact space, a contradiction.

Question 2.10.11 ([200]) Is every symmetrizable space with a σ -point-finite cs-

network a quotient compact image of a metric space?

Definition 2.10.12 Let P be a base of a space X.

(1) P is a uniform base [2] (resp. uniform base at non-isolated points [247]) for
X, if for each (resp. each non-isolated) point x ∈ X and each countably infinite
subset P of (P)x , P is a neighborhood base at x.
(2) P is a point-regular base [2] (resp. point-regular base at non-isolated points
[247]) for X, if for each (resp. each non-isolated) point x ∈ X and x ∈ U with U
open in X, {P ∈ (P)x : P ⊂ U} is finite.
2.10 Compact Mappings 135

Definition 2.10.13 ([247]) A space X is called developable at non-isolated points

if X has a sequence {Pn } of open subsets such that {st(x, Pn )}n∈N is a neighborhood
base at each non-isolated point x ∈ X. The sequence {Pn } is called a development
at non-isolated points for X.
In the definitions, “at non-isolated points” means “at each non-isolated point of
X”. In the following, let I(X) be the set of all isolated points in a space X, and
I (X) = {{x} : x ∈ I(X)}.

Theorem 2.10.14 ([247]) The following are equivalent for a space X:

(1) X is an open, peripherally compact image of a metric space.
(2) X has a uniform base at non-isolated points.
(3) X has a point-regular base at non-isolated points.
(4) X has a point-finite development at non-isolated points.

Proof (1) ⇒ (2). Let M be a metric space and f : M → X be an open, peripherally

compact mapping. By Theorem 1.3.5, we can choose a sequence {Bi } of open covers
 {st(K, Bi )}i∈N is a neighborhood base of K in M for each K ∈ K (M).
of M such that
Let P = i∈N Pi , where each Pi = f (Bi ). Then P is a base for X because f is
open. For each i ∈ N, we can assume that Bi+1 is a locally finite open refinement of
Bi . Let x ∈ U − I(X) with U open in X, and let P be a countably infinite subset
of (P)x . Then intf −1 (x) = ∅, thus f −1 (x) = ∂f −1 (x) is compact in M. There exists
m ∈ N such that st(f −1 (x), Bm ) ⊂ f −1 (U), so st(x, Pm ) ⊂ U. For each i ∈ N,
{B ∈ Bi : B ∩ f −1 (x) = ∅} is finite by the local finiteness of Bi , i.e., (Pi )x is finite.
There exists P ∈ P ∩ Pi for some i  m. Thus P ⊂ st(x, Pi ) ⊂ st(x, Pm ) ⊂ U.
Hence, P is a uniform base at non-isolated points.
(2) ⇒ (3). Let P be a uniform base at non-isolated points for X. If there exist
a non-isolated point x ∈ X and an open subset U in X with x ∈ U such that
{P ∈ (P)x : P ⊂ U} is infinite, take {Pn : n ∈ N} ⊂ {P ∈ (P)x : P ⊂ U}.
Then {Pn : n ∈ N} is a neighborhood base at x, thus Pm ⊂ U for some m ∈ N, a
contradiction. Therefore, P is a point-regular base at non-isolated points for X.
(3) ⇒ (4). Let P be a point-regular base at non-isolated points for X. Obviously,
P is also a uniform base at non-isolated points for X. We can assume that |P| = 1
if P ∈ P and P ⊂ I(X).
Claim. Let x be a non-isolated point of X and x = y ∈ X. Then {H ∈ P :
{x, y} ⊂ H} is finite.
In fact, {H ∈ P : {x, y} ⊂ H} ⊂ (P)x . If {H ∈ P : {x, y} ⊂ H} is infinite, then
it is a local base at x, hence y = x, a contradiction.

(14.1) P is point-countable at non-isolated points in X.

Let x ∈ X be a non-isolated point. There is a nontrivial sequence {xn } con-
x. By Claim, {P ∈ (P)x : xn ∈ P} is finite for each n ∈ N, then
verging to
(P)x = n∈N {P ∈ P : {x, xn } ⊂ P} is countable.
A family F of subsets of X is said to have the property () if, for any
F ∈ F − I (X), then {H ∈ F : F ⊂ H} is finite.
136 2 Mappings on Metric Spaces

(14.2) P has the property ().

Since F ∈ P − I (X), then F contains a non-isolated point and |F| > 1. By
Claim, P has the property ().
Put

P m = {H ∈ P : if H ⊂ P ∈ P, then P = H} ∪ I (X), and

P = (P − P m ) ∪ I (X).

(14.3) P m is an open cover and is point-finite at non-isolated points of X.

There exists HP ∈ P m such that P ⊂ HP for each P ∈ P − I (X) by (14.2).
Thus, Pm is an open cover of X. If P m is not point-finite at some non-isolated
point x ∈ X, then there exists an infinite subset {Hn : n ∈ N} of (P m )x . For
each n ∈ N, Hn+1 ⊂ H1 , there exists xn ∈ Hn+1 − H1 . Then the sequence {xn }
converges to x ∈ H1 , a contradiction.
(14.4) P is a point-regular base at non-isolated points for X.
Let x ∈ U − I(X) with U open in X. There exist V, W ∈ P and y ∈ V − {x}
such that x ∈ W ⊂ V − {y} ⊂ V ⊂ U. Thus, W ∈ P . Then P is a base
for X, and it is a point-regular base at non-isolated points for X.
Put
⎡⎛ ⎞ ⎤m

P1 = P m and Pn+1 = ⎣⎝P − Pi ⎠ ∪ I (X)⎦ , n ∈ N.
in


Then P = n∈N Pn by (14.2).
(14.5) {Pn } is a point-finite development at non-isolated points for X.
Each Pn is point-finite at non-isolated points by (14.3) and (14.4). If x ∈
U − I(X) with U open in X, then {P ∈ (P)x : P ⊂ U} is finite, thus there is
n ∈ N such that P ⊂ U whenever x ∈ P ∈ Pn , i.e., st(x, Pn ) ⊂ U. So {Pn }
is a development at non-isolated points.

(4) ⇒ (1). First, a metric space M and a function f : M → X are defined as

follows. Let {Pn } be a point-finite development at non-isolated points for X. For
each n ∈ N, assume that I (X) ⊂ Pn , put Pn = {Pα : α ∈ Λn }, and endow Λn
with the discrete topology. Put
 
M = α = (αn ) ∈ Λn : {Pαn }n∈N is a neighborhood
n∈N

base at some point x(α) ∈ X .

Then M, which is a subspace of the product space n∈N Λn ,  is a metric space. Define
a function f : M → X by f ((αn )) = xα . Then f ((αn )) = n∈N Pαn , and f is well
defined. It is easy to see that f is an open mapping by Proposition 2.9.5. The following
will prove that f is a peripherally compact map.
2.10 Compact Mappings 137

Let x ∈ X. If x ∈ I(X), then ∂f −1 (x) = ∅. If x ∈ / I(X), for eachi ∈ N, let

Γi = {α ∈ Λi : x ∈ Pα }, then Γi is finite. Thus, ∂f −1 (x) = f −1 (x) = n∈N Γn by
the proof of Theorem 2.10.6, so ∂f −1 (x) is compact in M. 

Corollary 2.10.15 ([247]) Let X be a space having a uniform base at non-isolated

points. Then,
(1) X is a quasi-developable space with a σ -interior-preserving base.
(2) X is a developable space if I(X) is an Fσ -set in X.

Proof By Theorem 2.10.14, we may assume that {P}n∈N is a point-finite develop-

ment at non-isolated points for X.
(1) Put P0 = I (X). It is easy to check that {Pn }n∈ω is a quasi-development for X.
Let P = n∈ω Pn . Then P is a base for X and each Pn is interior-preserving.
Indeed, for each F ⊂ Pn , if x ∈ ∩F − I(X), then (Pn )x is finite, thus ∩F is
a neighborhood of x in X. Thus P is a σ -interior-preserving base for X.
σ -set, there exists a sequence {G n } of open subsets of X such that
(2) If I(X) is an F
X − I(X) = n∈N G n . For each n ∈ N, let Un = {G n } ∪ {{x} : x ∈ X − G n }.
Then {Pn , Un }n∈N is a development for X. Hence, X is a developable space. 

Lemma 2.10.16 ([1]) For every space X, the following are equivalent:
(1) X has a point-finite development.
(2) X has a point-finite semi-development.
(3) X is a Fréchet–Urysohn space with a point-finite weak development.
(4) X is a metacompact developable space.

Proof (4) ⇒ (1) ⇒ (3) is obvious. By Proposition 1.6.17, we get (3) ⇒ (2). Below
we prove (2) ⇒ (4). Suppose X has a point-finite semi-development. By Theorem
2.10.6 and the mapping lemma (see Proposition 2.1.12), X is a countably bi-quotient
compact image of a metric space, so X is a metacompact space (see Corollary A.2.7
in Appendix A). Further by Theorems 2.7.17, 2.9.14 and 1.2.14, X is a developable
space. 

Thus, by Proposition 2.9.5 and Theorem 2.10.6, we get the following characteri-
zations of open compact images of metric spaces.

Theorem 2.10.17 ([23, 33]) For every space X, the following are equivalent:
(1) X is an open compact (and compact-covering) image of a metric space.
(2) X is a pseudo-open compact image of a metric space.
(3) X is a metacompact developable space.
(4) X is a perfect, metacompact space which is an open peripherally compact image
of a metric space [247].
(5) X has a uniform base.
(6) X has a point-regular base.
138 2 Mappings on Metric Spaces

Proof (1) ⇔ (2) ⇔ (3) by Lemma 2.10.16, Proposition 2.9.5 and Theorem 2.10.6.
(1) ⇔ (4) by Theorem 2.10.14 and Corollary 2.10.15. (1) ⇔ (5) ⇔ (6) by the
proof of Theorem 2.10.14. 

Michael [328] proved that every open compact mapping on a metric space is a
compact-covering mapping. Example 2.10.3 shows the assumption “open mapping”
cannot be weaken to “countably bi-quotient mapping”. P. Yan [487] and F. Lin [249]
etc. discussed the characterizations of quotient images of submetrizable spaces. The
characterizations of quotient compact images of metric spaces in Theorem 2.10.6
and open compact images of metric spaces in Theorem 2.10.17 have evoked a lot of
interesting work. For example, Theorem 2.10.6 has an improved form [14] that for
every space X, the following are equivalent:
(1) X is a sequentially quotient compact image of a metric space.
(2) X is a sequence-covering compact image of a metric space.
(3) X has a point-star network consisting of point-finite cs∗ -covers.
(4) X has a uniform cs∗ -network.
(5) X has a point-regular cs∗ -network.

Theorem 2.10.18 ([252]) The following are equivalent for a space X:

(1) X has a point-countable base, and a point-regular base at non-isolated points.
(2) X has a point-countable base which is point-regular at non-isolated points.
(3) X is an open peripherally compact, s-image of a metric space.
(4) X is an open s-image of a metric space, and is an open peripherally compact
image of a metric space.
(5) X is a meta-Lindelöf space with a point-regular base at non-isolated points.

Proof (1) ⇒ (2). Suppose that X has a point-countable base B, and a point-regular
 at non-isolated points P. By Theorem 2.10.14, we can assume that P =
base
n∈N Pn satisfies the following conditions:

(i) Pn is an open cover and is point-finite at non-isolated points for X;

(ii) {Pn } is a development at non-isolated points for X.
For each n ∈ N, put

B = {B ∈ B : B ⊂ I(X)};
Vn (B) = {P ∈ Pn : B ⊂ P}, B ∈ B ;
P̂ = ∪{B ∈ B : P ∈ Vn (B)}, P ∈ Pn ;
P̂n = {P̂ : P ∈ Pn }.

Then P̂n is point-countable. In fact, if x ∈ P̂ ∈ P̂n , then there is B ∈ B such that

x ∈ B and P ∈ Vn (B ). Since {B ∈ B : x ∈ B} is countable, and each Vn (B) is finite
for each B ∈ B by the condition (i), it follows that {P ∈ Vn (B) : x ∈ B ∈ B } is
countable.
2.10 Compact Mappings 139

Put ! "

P̂ = P̂n ∪ I (X).
n∈N

Then P̂ is point-countable. If x ∈ U − I(X) with U open in X, then there is m ∈ N

such that x ∈ st(x, Pm ) ⊂ U by the condition (ii). Take P ∈ Pm with x ∈ P, then
there is B ∈ B such that x ∈ B ⊂ P, thus P ∈ Vm (B), and x ∈ B ⊂ P̂ ⊂ P ⊂ U. So
P̂ is a base for X. Finally, it is easy to see that P̂ is point-regular at non-isolated
points by P̂ ⊂ P for each P ∈ P.
(2) ⇒ (3). Let P be a point-countable base which is point-regular
 at non-isolated
points for X. By Theorem 2.10.14, we can express P by n∈N Pn , where each Pn
is a point-countable open cover of X, and {Pn } is a point-finite development at non-
isolated points for X. Put Pn = {Pα : α ∈ Λn } for each n  ∈ N. By the proof of
(4) ⇒ (1) of Theorem 2.10.14, there are a metric space M ⊂  n∈N Λn and an open
peripherally compact mapping f : M → X such that f −1 (x) ⊂ n∈N {α ∈ Λn : x ∈
Pα }, thus f is an s-mapping.
(3) ⇒ (4) is obvious. (4) ⇒ (5) by Theorem 2.10.14.
(5) ⇒ (1). Let X be a meta-Lindelöf space with a point-regular base at non-
isolated points. By Theorem 2.10.14, there is a sequence {Pn } of open covers of X
such that {Pn } is a point-finite development at non-isolated points for X. For each
n ∈ N, let Bn be a point-countable open refinement of Pn . And put
! "

B= Bn ∪ I (X).
n∈N

Then B is a point-countable base for X. In fact, if a non-isolated point x ∈ U with

U open in X, then there is n ∈ N such that st(x, Pn ) ⊂ U. Take B ∈ Bn with x ∈ B,
then x ∈ B ⊂ st(x, Bn ) ⊂ st(x, Pn ) ⊂ U. 

Example 2.10.19 The Michael line (see Example 1.8.5) is an open compact image
of a metacompact developable space [56].
We use the notations of Example 1.8.5, and the Michael line and the Bernstein
set are denoted by X and B respectively. Let

H = (I × {0}) ∪ (B × N);
V (x, m) = {x} × ({0} ∪ {n : n  m}), x ∈ I, m ∈ N;
W (J, m) = ((J ∩ (I − B)) × {0})∪
((J ∩ B) × {n : n  m}), J ⊂ I, m ∈ N.

Give H the following topology: each element of a base has the forms V (x, m) for
every x ∈ B and m ∈ N, W (J, m) for every open interval J ⊂ I and m ∈ N, and {h}
for each h ∈ B × N. There is a development {Um } for I with respect to Euclidean
topology such that Um is a finite set and Um+1 refines Um . For each m ∈ N, let Pm
140 2 Mappings on Metric Spaces

be the family of all sets of the following forms: V (x, m) (for each x ∈ B), W (U, m)
(for each U ∈ Um ) and {h} (for each h ∈ B × {1, 2, . . . , m − 1}). Then {Pm }m2 is
a point-finite development for H, so H is a metacompact developable space.
It is easy to verify that π1|H : H → X is an open compact mapping and X has a
point-countable base which is uniform at non-isolated points. Since the set B is not
an Fσ -set in X, X is not an open compact image of a metric space. Let X ∗ be a copy
of the Michael line X and f : X → X ∗ be a homeomorphism. Put Z = X ⊕ X ∗ ,
and let Y be the quotient space obtained from Z by identifying {x, f (x)} to a point
for each x ∈ X − B. Then Y has a point-countable base which is uniform at non-
isolated points. Hence, Y is an open peripherally compact, s-image of a metric space
by Theorem 2.10.18.
Because the Michael line is not a β-space, it is not a metacompact developable
space either. From this we can know that the class of open compact images of
metric spaces is not closed under open compact mappings. Hence, the class of spaces
containing metric spaces and closed under open compact mappings is of special
significance.

Definition 2.10.20 ([31]) The class MOBI is the smallest class of spaces such that
(1) every metric space is in this class;
(2) this class is closed under open compact mappings.
Obviously, the class MOBI is preserved by open compact mappings.

Theorem 2.10.21 ([56]) A space Y is in the class MOBI if and only if there is
a metric space M and finitely many open compact mappings f1 , . . . , fn such that
(fn ◦ · · · ◦ f1 )(M) = Y .

Proof Denote the set of classes of spaces

 satisfying the two conditions of Definition
2.10.20 by {Hα }α∈Λ . Then MOBI = α∈Λ Hα . Let

B ={X : there exist a metric space Z and finitely many open

compact mappings f1 , . . . , fn , such that (fn ◦ · · · ◦ f1 )(Z) = X}.

Then there is α ∈ Λ such that B = Hα , and hence B ⊃ MOBI. If X ∈ B, then

there is a metric space Z and finitely many open compact mappings f1 , . . . , fn such
that (fn ◦ · · · ◦ f1 )(Z) = X. By Definition 2.10.20, X ∈ Hα for each α ∈ Λ, so
B ⊂ Hα , and hence B ⊂ MOBI. Thus B = MOBI. 

By Corollary 2.7.19 and Theorem 2.10.21, we have the following corollary.

Corollary 2.10.22 ([31]) If Y ∈ MOBI, then Y has a point-countable base.

The problem whether the inverse proposition of Corollary 2.10.22 is set up affords
much food for thought. Chaber [98] proved that every T1 space with a point-countable
base is an open compact image of a metacompact developable T1 space. So in the
class of T1 spaces, the class MOBI can be characterized by the class of spaces with a
2.10 Compact Mappings 141

point-countable base. If we denote the subclass of class MOBI consisting of spaces

every open compact image of which is a Ti space, as we described in Theorem
2.10.21, by MOBIi , where i = 2, 3, 4, then a major problem is whether every Ti
space with a point-countable base belongs to the class MOBIi . Example 2.10.19
shows the Michael line belongs to the class MOBI2 .
Question 2.10.23 ([31]) Is the class MOBI preserved by perfect mappings?
Question 2.10.24 ([98]) Does every space with a point-countable base belong to
MOBI2 ?

2.11 σ -Locally Finite Mappings

The way of investigating the relationships between spaces and mappings in the pre-
vious sections is to study characterizations of images of metric spaces under several
well-known mappings. For the classes of generalized metric spaces, after defining
suitable mappings, we may characterize these classes by means of images of metric
spaces under such mappings. Following this train of thought, by σ -locally finite map-
pings and σ -mappings defined in this section, we characterize σ -spaces and ℵ-spaces
in terms of images of metric spaces under these two mappings.

Definition 2.11.1 ([338]) A mapping f : X → Y is called a σ -locally finite mapping

if, for every σ -locally finite cover P of X, there is a refinement F of P such that
f (F ) is a σ -locally finite cover of Y .
Let P and F be families of sets in X. Then F is called a B-refinement of P if,
F partially refines P and each element of P is the union of some sets in F .
Lemma 2.11.2 ([338]) If P is a locally finite family of set in a space X, then P
has a disjoint locally finite B-refinement F such that (P)F is finite for any F ∈ F .
Proof Let

F(B) = ∩B − ∪(P − B), B ⊂ P;

F = {F(B) : ∅ = B ⊂ P}.

Then P is a locally finite B-refinement of F consisting of disjoint sets and (P)F

is finite for any F ∈ F . 
Proposition 2.11.3 ([338]) If f : X → Y is a mapping, then (1) ⇒ (2) ⇔ (3) for
the following conclusions:
(1) Each open cover U of X has a refinement B such that f (B) is a σ -locally finite
family of sets in Y .
(2) Each σ -locally finite family of sets in X has a B-refinement B such that f (B) is
a σ -locally finite family of sets in Y .
142 2 Mappings on Metric Spaces

(3) f is a σ -locally finite mapping.


Proof (1) ⇒ (3). Suppose P = n∈N Pn is a cover of X such that each Pn is
locally finite. For each n ∈ N, there is an open cover Un of X such that each element
of Un only meets finitely many elements of P n , and hence Un has a refinement Bn
such that
 f (B n ) is σ -locally finite in Y . Then n∈N (Pn ∧ Bn ) is a refinement of P
and f ( n∈N (Pn ∧ Bn )) is a σ -locally finite cover of Y . Thus f is a σ -locally finite
mapping.
(2) ⇒ (3) is obvious, we prove (3) ⇒ (2). We only need to prove that each locally
finite family P of sets in X has a B-refinement B such that f (B) is σ -locally finite
in Y . By Lemma 2.11.2, P has a locally finite B-refinement F consisting of disjoint
sets such that each element of F meets only finitely many elements of P. Let

H = F ∪ {X − ∪F }.

Then H is a locally finite cove of X, and hence H has a refinement R such that
f (R) is a σ -locally finite family of sets in Y . Since H is a family of disjoint sets,
R is a B-refinement of H . Define

B = {B ∈ R : B ⊂ ∪F }.

Then B is B-refinement of F , so B is a B-refinement of P and f (B) is a σ -locally

finite family of sets in Y . 
Corollary 2.11.4 ([338]) Let f : X → Y be a mapping. If one of the following is
true, then f is a σ -locally finite mapping:
(1) There is an almost (mod k)-network P for X such that f (P) is a σ -locally finite
family of sets in Y .
(2) f is a closed L-mapping and Y is a subparacompact space.
Proof Suppose U is an open cover of X.
(1) Let
H = {P ∈ P : P ⊂ ∪F , F ∈ U <ω }.

Then H is a refinement of U F and f (H ) is a σ -locally finite family of sets in

Y , so U has a refinement B such that f (B) is a σ -locally finite family of sets
in Y . By Proposition 2.11.3, f is a σ -locally finite mapping.
(2) For each y ∈ Y , there exist a countable subfamily Uy of U covering f −1 (y) and
an open neighborhood Vy of y such that f −1 (Vy ) ⊂ ∪Uy . Let V = {Vy }y∈Y . Then
the open cover V of Y has a σ -discrete closed refinement F . For each F ∈ F ,
there is y(F) ∈ Y such that F ⊂ Vy(F) . Let

B = {f −1 (F) ∩ U : F ∈ F , U ∈ Uy(F) }.

Then B refines U and f (B) is a σ -locally finite family of sets in Y . Thus f is a

σ -locally finite mapping. 
2.11 σ -Locally Finite Mappings 143

Corollary 2.11.5 ([338]) The following topological properties are invariant under
σ -locally finite mappings:
(1) having a σ -locally finite network;
(2) having a σ -locally finite almost (mod k)-network.

Proof Suppose f : X → Y is a σ -locally finite mapping. Let P be a σ -locally

finite network (resp. σ -locally finite almost (mod k)-network) for X. By Proposition
2.11.3, P has a B-refinement B such that f (B) is a σ -locally finite family of sets
in Y . It is easy to verify that f (B) is a network (resp. an almost (mod k)-network)
for Y . 

Let X be the Michael line (see Example 1.8.5). Since X has a point-countable
base, there exist a metric space M and an open s-mapping f : M → X. Because X
is not a σ -space, by Corollary 2.11.5, f is not a σ -locally finite mapping. Hence, the
condition “closed mapping” cannot be replaced with “open mapping” in Corollary
2.11.4.
The following lemma is obtained by Lemma 1.5.13 and Proposition 2.4.8.

Lemma 2.11.6 ([338]) If P is a σ -locally finite closed almost (mod k)-network

which is closed under finite intersections in a space Y , then there exist a metrizable
space M, a σ -discrete base B for M and a subspace X of Y × M satisfying the
following conditions: let f = π1|X and g = π2|X , then
(1) P = f (g −1 (B));
(2) g : X → M is a perfect mapping.

Theorem 2.11.7 ([338]) For every regular space X, the following are equivalent:
(1) X is a σ -space (resp. strong Σ-space).
(2) X is a σ -locally finite image of a metric space (resp. paracompact M-space).

Proof By Corollary 2.11.5, we get (2) ⇒ (1) (for the paracompact M-space case,
we should further use Corollary 2.2.8 and Proposition 1.5.14).
(1) ⇒ (2). We first assume that P is a σ -locally finite closed network which
is closed under finite intersections for the σ -space X. Let M = X and give M the
topology generated by taking P as a base. Then P is a σ -locally finite clopen base
for M, and hence M is a metric space. Let f = idM : M → X. Since P is a network
for X, f is a mapping. Further by Corollary 2.11.4, f is a σ -locally finite mapping.
Now suppose P is a σ -locally finite closed almost (mod k)-network which is
closed under finite intersections for the strong Σ-space X. By Lemma 2.11.6, there
exist a metrizable space M, a σ -discrete base B for M and a subspace Z of X × M
satisfying the conditions that P = f (g −1 (B)) and g : Z → M is a perfect mapping,
where f = π1|Z and g = π2|Z . Then Z is a paracompact M-space and g −1 (B) is a
(mod k)-network for Z. By Corollary 2.11.4, f is a σ -locally finite mapping. 

The assumption of regularity in the above theorem is necessary. The pointed irra-
tional extension topological space X of R is a σ -locally finite image of a metric space
144 2 Mappings on Metric Spaces

(see Examples 2.7.14 and 2.10.10). Since every strong Σ-space is a subparacompact
space (see Theorem 3.2.11), X is not a strong Σ-space.
For spaces with a (mod k)-network, we have the following result parallel to Corol-
lary 2.8.4.

Corollary 2.11.8 ([338]) A regular space X has a countable (mod k)-network if and
only if there exist a separable metric space M and a perfect mapping g : Z → M
such that X is an image of Z under a mapping.

Example 2.11.9 ([333]) There exist a locally compact paracompact space Z and a
closed mapping f : Z → X such that
(1) X is not a strong Σ-space;
(2) f is not a σ -locally finite mapping.
Suppose
 hα : ω1 + 1 → Tα is a homeomorphic mapping for each α < ω1 . Let
Z = α<ω1 Tα . Then Z is a locally compact paracompact space. Let A = {hα (ω1 ) :
α < ω1 }. Denote the quotient space Z/A by X and let f : Z → X be the quotient
mapping. Then f is a closed mapping. Assume X is a strong Σ-space. Let F be a σ -
locally finite closed (mod k)-network with respect to K in X. Define  ω1 = f (h0 (ω1 )).
For every α, β < ω1 , let [α(β), ω1 ) = f (hβ ([α, ω1 ))). Since F is σ -locally finite,
for each β < ω1 , {F ∈ F :  ω1 ∈ F and F ∩ [0(β),  ω1 ) = ∅} is countable, and hence
there is αβ < ω1 such that ω1 ∈ F whenever F ∈ F and F ∩ [αβ (β),  ω1 ) = ∅. For
each F ∈ F , define

α(F) = {γ < ω1 : F ∩ [0(γ ), 

ω1 ) = ∅},
F0 = {F ∈ F : 
ω1 ∈ F, α(F) < ω1 }.

Then F0 is countable. So there is β < ω1 such that β > sup{α(F) : F ∈ F0 }. Thus,

if F ∈ F and F ∩ [αβ (β), 
ω1 ) = ∅, then F ∩ [0(γ ), 
ω1 ) = ∅ for uncountably many
γ < ω1 . Let
{F ∈ F : F ∩ [αβ (β),  ω1 ) = ∅} = {Fn }n∈N .

Pick K ∈ K such that K ∩ [αβ (β),  ω1 ) = ∅. Since {γ < ω1 : K ∩ [0(γ ), 

ω1 ) = ∅}
is a finite set, we can choose a sequence {xn } in X and a nontrivial sequence {γn } in
ω1 such that xn ∈ (Fn − K) ∩ [0(γn ),  ω1 ). So there is F ∈ F such that K ⊂ F ⊂
X − {xn : n ∈ N}, and hence there is m ∈ N such that F = Fm , it follows xm ∈ F, a
contradiction. Thus X is not a strong Σ-space. Moreover, by Corollary 2.11.5, f is
not a σ -locally finite mapping.
In the second part of this section, we characterize ℵ-spaces in terms of specific
images of metric spaces.

Definition 2.11.10 ([266]) A mapping f : X → Y is called a σ -mapping if there is

a base B for X such that f (B) is a σ -locally finite family of sets in Y .
Every mapping on a separable metric space is a σ -mapping. By Proposition 2.11.3,
every σ -mapping is a σ -locally finite mapping.
2.11 σ -Locally Finite Mappings 145

Theorem 2.11.11 ([266]) A regular space X is a σ -space if and only if X is a σ -

image of a metrizable space.

Proof Suppose X is a σ -space. Then by the proof of (1) ⇒ (2) in Theorem 2.11.7,
X is a σ -image of a metrizable space. Conversely, if a σ -image of a metric space is
a regular space, then it is a σ -space, because every σ -mapping is a σ -locally finite
mapping. 

Theorem 2.11.12 For every regular space X, the following are equivalent:
(1) X is an ℵ-space.
(2) X is a sequentially quotient σ -image of a metric space [459].
(3) X is a sequence-covering σ -image of a metric space [459].
(4) X is a compact-covering σ -image of a metric space [266].

Proof (1) ⇒ (4). Suppose P is a σ -locally finite closed

 k-network which is closed
under finite intersections for the ℵ-space X. Let P = n∈N Pn , where each Pn =
{Pα }α∈Λn is locally finite and we may assume X ∈ Pn ⊂ Pn+1 . Give Λn the discrete
topology. Let
 

M = α = (αi ) ∈ Λi : {Pαi } is a net at a point x(α) in X .
i∈N

Then M is a metrizable space and for each α ∈ M, x(α) is uniquely determined. So

we can define a function f : M → X by f (α) = x(α). It is easy to verify that f is a
mapping. Below we prove f is a compact-covering σ -mapping.

(12.1) f is a σ -mapping.
For every n ∈ N and αn ∈ Λn , define

B(α1 , . . . , αn ) = {α ∈ M : πi (α) = αi , i  n},

B = {B(α1 , . . . , αn ) : αi ∈ Λi , i  n}.

Then B is a base for M. To prove f is a σ -mapping, we only need  to verify

f (B(α1 , . . . , α n )) = P
in αi . Obviously, f (B(α1 , . . . , α n )) ⊂ in Pαi . If
x ∈ in Pαi , then there is β = (βk ) ∈ M such that f (β) = x. For each
k ∈ N, there is αk+n ∈ Λk+n such that  Pαk+n = Pβk . Let α = (αk ). Then
α ∈ B(α1 , . . . , αn ) and f (α) = x. So in Pαi ⊂ f (B(α1 , . . . , αn )).
(12.2) f is a compact-covering mapping.
For each K ∈ K (X), by Lemma 2.7.9, there is a sequence {Fi } of finite
subsets of P satisfying the conditions (i) and (ii) of Proposition 2.7.1. Take
a strictly increasing sequence {ni } in N such that Fi ⊂ Pni . We may assume
n1 = 1. For every i ∈ N and ni  n < ni+1 , let Pn = Fi . Then Pn ⊂ Pn ,
so there is Γn ∈ Λ<ω
n such that Pn = {Pα }α∈Γn . Let
146 2 Mappings on Metric Spaces
 

L = α = (αn ) ∈ Γn : K ∩ Pαn = ∅ .
n∈N n∈N

By the proof of Proposition 2.7.1, L is a compact set in M and f (L) = K.

Thus f is a compact-covering mapping.

(4) ⇒ (3) ⇒ (2) is obvious. It remains to prove (2) ⇒ (1). Suppose f : M → X

is a sequentially quotient σ -mapping and M is a metric space. Then there is a base B
for M such that f (B) is a σ -locally finite family of sets in X. By Propositions 2.7.2
and 1.6.7, f (B) is a σ -locally finite k-network for X, and hence X is an ℵ-space. 

Example 2.11.13 There exist a metric space M and a σ -locally finite mapping f :
M → X such that
(1) f is a compact-covering mapping and X is not an ℵ-space;
(2) f is a compact mapping and X is not a g-metrizable space;
(3) f is not a σ -mapping.
Let X be the V -space in Examples 1.8.1 and 2.4.16. By Example 2.4.16, there exist
a metric space M and a finite-to-one compact-covering open mapping f : M → X.
By Corollary 2.11.4, f is a σ -locally finite mapping. By Theorem 2.5.15, X is not
an ℵ-space. Further by Corollary 1.6.8 and Theorem 1.6.22, X is not a g-metrizable
space either. By Theorem 2.11.12, f is not a σ -mapping.

The relevant results on characterizations of g-metrizable spaces by means of

mappings are in Theorem 3.9.13.
Chapter 3
Generalized Metric Spaces

In the second chapter, we described intrinsic characterizations of images or preimages

of metric spaces under certain mappings. Thus, by using the classification principle
of mappings and spaces, we can deal with several important concepts in general
topology, such as bases, sequences of covers, networks, k-networks, cs∗ -networks,
(mod k)-networks etc. and the classes of generalized metric spaces derived by them,
such as developable spaces, p-spaces, semi-metrizable spaces, spaces with general-
ized sequentiality property, M-spaces, σ -spaces, strong Σ-spaces etc. in a unified
way by means of metric spaces and closed mappings, open mappings, countably
bi-quotient mappings, pseudo-open mappings, quotient mappings, s-mapping, com-
pact mappings, π -mappings and so on. This not only shows the effect and power of
investigating spaces by mappings, but also shows that the generalized metric spaces
appeared in the second chapter occupy the core position in the theory of topological
spaces.
The basic problem of the principle of mutual classifications of spaces and map-
pings is that, under which classes of mappings certain topological properties remain
unchanged (see Question 2.0.6). One of the central topics of this chapter is to deter-
mine the given classes of generalized metric spaces are preserved by which kinds
of mappings. Therefore, we must fully investigate various properties of generalized
metric spaces. One of the most wonderful content of this chapter i.e. seeking the
characterizations of these generalized metric spaces is particularly important.
A considerable amount of generalized metric spaces involved in the last two
chapters are defined based on the Bing–Nagata–Smirnov metrization theorem (see
Theorem 1.3.2), the characterizations of paracompact spaces obtained by Michael
(see Theorem A.1.2 in Appendix A) or Burke–Junnila’s characterization of sub-
paracompact spaces (see Theorem A.3.3 in Appendix A), and by using σ -discrete,
σ -locally finite, σ -closure-preserving or σ -cushioned families of sets. This is a valid
method to generate many classes of generalized metric spaces, which leads to a kind
of interesting questions: under which conditions the classes of spaces defined by
using these four types of families are equivalent?
For this kind of question, it is particularly important to consider whether these
classes of spaces are invariant under closed mappings. As will be seen in the sequel
© Atlantis Press and the author(s) 2016 147
S. Lin and Z. Yun, Generalized Metric Spaces and Mappings,
Atlantis Studies in Mathematics 6, DOI 10.2991/978-94-6239-216-8_3
148 3 Generalized Metric Spaces

that the successful answer of this question for σ -spaces has become a model in
general topology. Generally, for σ -discrete families and σ -locally finite families, this
question mostly has a positive answer; but for σ -locally finite families and σ -closure-
preserving families, the answer is complicated; for σ -closure-preserving families and
σ -cushioned families, the problem seems to be very difficult. One of the features of
this chapter is a full description of spaces defined by σ -HCP families. From the
Burke–Engelking–Lutzer metrization theorem (see Theorem 2.5.15), one can see
the dawn of solving the problem that under which conditions σ -HCP families are
σ -locally finite families. Moreover, in seeking suitable classes of spaces invariant
under closed mappings, this problem occupy the equally important position as in
studying σ -closure-preserving family of sets.

3.1 Spaces with Point-Countable Covers

In the previous chapter, to answer the question of characterizing quotient s-images

of metric spaces raised by Arhangel’skiı̌ [31], and to study the relevant problems
on open s-images, open compact images and so on, we investigated numbers of
point-countable covers with specific properties and preliminarily discussed the rela-
tionships among them. In this section, we focus on the properties of spaces with
a point-countable cover, including metrization theorems and mapping properties of
these spaces.

Definition 3.1.1 Let P be a family of sets in a space (X, τ ).

(1) P is called a pseudo-base for X [41] if P ⊂ τ and P is a pseudo-network for
X.
(2) P is called a pseudo-F-quasi-base (briefly pseudo-Fq-base) for X if, there is
F ∈ P <ω , such that, x ∈ (∪F )◦ ⊂ ∪F ⊂ X − {y} for each pair x, y of X with
x = y.
(3) P is called a pseudo-k-network for X if, for every K ∈ K (X) and y ∈ X − K,
there is F ∈ P <ω such that K ⊂ ∪F ⊂ X − {y}.

Proposition 3.1.2 The following spaces have a point-countable k-network:

(1) spaces having a σ -countably weakly HCP k-network [130];
(2) quotient s-images of metric spaces [167].

Proof By Lemma 2.5.4 and the proof of (3) ⇒ (4) in Theorem 2.5.8, every space
with a σ -countably weakly HCP k-network has a σ -compact-finite k-network, and
hence has a point-countable k-network. By Lemma 2.7.11, every quotient s-image
of a metric space has a point-countable k-network. 

Proposition 3.1.3 ([90]) Every subparacompact space with a Gδ -diagonal has a

point-countable pseudo-Fq-base.
 3.1 Spaces with Point-Countable Covers 149

Proof Suppose Xis a subparacompact space with a Gδ -diagonal. Then X has a closed
pseudo-network n∈N Pn such that each Pn is locally finite. For each n ∈ N, define

Fn (B) = ∩B − ∪(Pn − B), B ⊂ Pn ;

Fn = {Fn (B) : B ⊂ Pn }.

Then Fn is a disjoint cover of X. For every x, y ∈ X with x = y, there exist m ∈ N

and B ∈ Pm such that y ∈ B ⊂ X − {x}. Let

H = {Fm (B) : B ⊂ (Pm )x }.

Then H ∈ Fm<ω . If B ⊂ (Pm )x , then B ∈ Pm − B, so Fm (B) ⊂ X − B, and hence

∪H ⊂ X − {y}. Let U = X − ∪(Pm − (Pm )x ). Then x ∈ U ∈ τ . If z ∈ U, then
(Pm )z ⊂ (Pm )x , so z ∈ Fm ((Pm )z )∈ H , and hence U ⊂ ∪H . Consequently,
x ∈ (∪H )◦ ⊂ ∪H ⊂ X − {y}. Thus n∈N Fn is a point-countable pseudo-Fq-base
for X. 

Corollary 3.1.4 Every semi-stratifiable space has a point-countable pseudo-Fq-

base.

Question 3.1.5 (1) Is every β-space with a point-countable pseudo-Fq-base a semi-

stratifiable space?
(2) Is every β-space with a point-countable base a semi-stratifiable space?
(3) Is every collectionwise normal β-space with a point-countable base metrizable
[45]?

Lemma 3.1.6 Every countably compact space belonging to one of the following
classes is a compact metrizable space:

(1) the class of spaces with a point-countable pseudo-Fq-base [90];

(2) the class of k-spaces with a point-countable pseudo-k-network [167].

Proof Suppose X is a countably compact space.

(1) Let P be a point-countable pseudo-Fq-base for X. Then P has the following
property: If A ∈ X <ω and x ∈ X − A, then there is F ∈ P <ω such that x ∈ (∪F )◦ ⊂
∪F ⊂ X − A.
Let
H = ∪{F ∈ P <ω : F is a minimal cover of X}.

By the Miščenko lemma, H is countable.

(6.1) H is a pseudo-Fq-base for X.

For every x, y ∈ X with x = y, there is F ∈ P <ω such that x ∈ (∪F )◦ ⊂ ∪F ⊂

X − {y}. We may assume F is a minimal interior cover of {x}. For each F ∈ F , there
is x(F) ∈ (∪F )◦ ∩ (X − ∪(F − {F})), and hence x(F) ∈ F ∩ (∪F )◦ − ∪(F − {F}).
 150 3 Generalized Metric Spaces

Let A = {x(F) : F ∈ F }. For each z ∈ X − A, there is Fz ∈ P <ω such that

z ∈ (∪Fz )◦ ⊂ ∪Fz ⊂ X − A. Define

R = F ∪ (∪{Fz : z ∈ X − A}) .

For each p ∈ X, let (R)p = {Pi (p)}i∈N . We prove that there is a finite subfamily
of R covering X. Otherwise, there is an infinite subset {xn : n ∈ N} of X such that
xn ∈
/ Pi (xj ) whenever i, j < n. Let a be an accumulation point of {xn : n ∈ N}. Since

X = (∪F )◦ ∪ ( (∪Fz )◦ ),
z∈X−A

there is P ∈ R such that P contains a subsequence {xnk } of the sequence {xn }. Pick
i, j ∈ N such that P = Pi (xj ). Take m ∈ N such that nm > i, j. Then xnm ∈ Pi (xj ), a
contradiction. So there is B ∈ R <ω such that X = ∪B. Suppose the family B is a
minimal cover of X. Then B ⊂ H . For each F ∈ F , since (R)x(F) = {F}, F ∈ B,
so F ⊂ H . Thus H is a pseudo-Fq-base.

(6.2) X has a countable closed network.

Define

E = {X − (∪F )◦ : F ∈ H <ω },
C = {∩E
: E
∈ E <ω }.

If x ∈ V ∈ τ , then by (6.1), {x} = ∩(E )x , so {V } ∪ {X − E : E ∈ (E )x } is an open

cover of X, and hence there is E
∈ E <ω such that x ∈ ∩E
⊂ V . Therefore C is a
countable closed network for X.
Thus, X is a semi-stratifiable space, and by Corollary 1.4.12, X is a compact
metrizable space.
(2) Let P be a point-countable pseudo-k-network for the k-space X. Then P is
a pseudo-Fq-base.
In fact, if K ∈ K (X), then P|K is a point-countable pseudo-Fq-base for K, so K
is metrizable. By Corollary 2.3.5, X is a sequential space. For every x, y ∈ X with
x = y, there is a closed neighborhood L of x such that L ⊂ X − {y}. By a proof
similar to that of Proposition 1.6.7, we can prove that there is F ∈ P <ω such that
L ⊂ ∪F ⊂ X −{y}, so x ∈ (∪F )◦ ⊂ ∪F ⊂ X −{y}. Hence P is a pseudo-Fq-base
for X, thus X is a compact metrizable space. 

Theorem 3.1.7 If a strongly Fréchet–Urysohn space X has a point-countable k-

network, then X is a first countable space [276]. If further assume X is a regular
space, then X has a point-countable base [167].
3.1 Spaces with Point-Countable Covers 151

Proof Let P be a point-countable k-network for the strongly Fréchet–Urysohn

space X.
(7.1) For every x ∈ W ∈ τ , there is F ∈ P <ω such that x ∈ (∪F )◦ ⊂ ∪F ⊂ W .
Let P
= {P ∈ P : P ⊂ W }. Then P
is a point-countable k-network for W . By
Lemma 2.7.13, there is F ∈ P
<ω such that x ∈ (∪F )◦ ⊂ ∪F ⊂ W .

(1) Suppose X is a regular space. For each F ∈ P <ω , define

H(F ) = {x ∈ X : F is a minimal interior cover of {x}}.

Let P = {Pα }α∈Λ . For each α ∈ Λ, define

Bα = Pα ∪ (∪{H(F ) : Pα ∈ F ∈ P <ω }).

If x ∈ H(F ), where Pα ∈ F ∈ P <ω , then x ∈ (∪F )◦ − (∪(F − {Pα }))◦ ,

and hence x ∈ Pα . Thus Bα ⊂ Pα . Let B = {Bα }α∈Λ . By Lemma 2.7.15, for each
x ∈ X, {F ∈ P <ω : x ∈ H(F )} is countable, and hence (B)x is countable. Below
we prove B satisfies the assumption of Proposition 2.7.16. For every x ∈ U ∈ τ ,
by the regularity of X, there is W ∈ τ such that x ∈ W ⊂ W ⊂ U. By (7.1), there
is F ∈ P <ω such that x ∈ (∪F )◦ ⊂ ∪F ⊂ W . We may assume x ∈ H(F ). Let
R = {Bα : Pα ∈ F }. Then x ∈ ∩R and x ∈ (∪R)◦ ⊂ ∪F ⊂ U. Thus X has a
point-countable base.
(2) Suppose X is not a regular space. Fix a point x0 ∈ X, define a new topology
τ ∗ on X as follows: for each x = x0 , {x} ∈ τ ∗ ; a neighborhood base of x0 in τ ∗ is the
same as that in τ . Then τ ∗ is a regular space. 
Suppose {A n } is a decreasing sequence of sets in X and x ∈ n∈N clτ (An ). If x =∗ x0 ,
∗

then x ∈ n∈N A n , and we take xn = x ∈ An . Then {xn } converges to x in τ . If

x = x0 , then x ∈ n∈N clτ (An ). So there is xn ∈ An such that {xn } converges to x in τ ,
and hence {xn } also converges to x in τ ∗ . Thus (X, τ ∗ ) is a strongly Fréchet–Urysohn
space.
Let P ∗ = {{x} : x = x0 }∪P. Then P ∗ is point-countable. Suppose x ∈ W ∈ τ ∗ .
If x = x0 , let F = {x} ∈ P ∗ , then x ∈ intτ ∗ (F) = F ⊂ W . If x = x0 , then
there is V ∈ τ such that x ∈ V ⊂ W . By (7.1), there is F ∈ P <ω such that
x ∈ intτ (∪F ) ⊂ ∪F ⊂ V . Then x ∈ intτ ∗ (∪F ) ⊂ ∪F ⊂ W . So P ∗ satisfies (7.1)
in (X, τ ∗ ). By (1), (X, τ ∗ ) has a point-countable base, and hence there is a countable
local base of x0 in (X, τ ). Consequently, X is a first countable space. 

Theorem 3.1.8 ([167]) For every space X, the following are equivalent:

(1) X is a metrizable space.

(2) X is an M-space with a point-countable pseudo-Fq-base.
(3) X is a paracompact M-space with a point-countable pseudo-k-network.
(4) X is a k-and M-space with a point-countable pseudo-k-network.
152 3 Generalized Metric Spaces

Proof (1) ⇒ (2) is obvious. By Lemma 3.1.6, we get (2) ⇒ (3). By Propositions
2.4.10 and 2.4.11, we know (3) ⇒ (4). Below we prove (4) ⇒ (1). By Lemma 3.1.6,
every k-and M-space with a point-countable pseudo-k-network is a paracompact
space. By Theorem 2.2.12, we only need to prove (∗): Every submetacompact space
with a point-countable pseudo-k-network and a strict p-sequence has a Gδ -diagonal.
Suppose {Un } is a strict p-sequence in the submetacompact space X and X has a
point-countable pseudo-k-network. For each x ∈ X, define

Cx = ∩{st(x, Un ) : n ∈ N}.

Then Cx ∈ K (X). By Lemma 3.1.6, Cx is a compact metrizable subspace of X,

and by Lemma 2.6.2, Cx has a countable outer base in X, so there is a countable
neighborhood base of x, and hence X is a first countable space. By the proof of
(7.1) in Theorem 3.1.7, X has a point-countable pseudo-Fq-base, so X 2 also has a
point-countable pseudo-Fq-base, we denoted it by P.
For each i ∈ N, assume {H (i, k)} is an open θ -refinable sequence of Ui , and
let H (i, j, k, l) = H (i, k) × H (j, l). For the diagonal Δ in X 2 and each H ∈
H (i, j, k, l), by Lemma 2.7.15, the set of all finite minimal interior covers of Δ ∩ H
consisting of elements of P is countable, and we denote it by H . Let

H = {H (i, j, k, l, m) : m ∈ N},
H(i, j, k, l, m) = (∪H (i, j, k, l, m))◦ ,
W (i, j, k, l, m) = ∪{H ∩ (∪{H(i, j, k, l, n) : n  m}) : H ∈ H (i, j, k, l)}.

Then Δ ⊂ ∩{W (i, j, k, l, m) : i, j, k, l, m ∈ N}. On the other hand, since {Ui ×

Uj }i,j∈Nis a strict p-sequence in X 2 , for each p = (x, y) ∈ X 2 − Δ, we define
Dp = i,j∈N st(p, Ui × Uj ). Then Dp ∩ Δ is a compact set in X 2 , and hence there
is F ∈ P <ω such that Dp ∩ Δ ⊂ (∪F )◦ ⊂ ∪F ⊂ X 2 − {p}, so there exist i, j ∈ N
such that Dp ⊂ st(p, Ui × Uj ) ⊂ (∪F )◦ ∪ (X 2 − Δ). Thus there exist k, l, n ∈ N
such that (H (i, j, k, l))p = {Ht }tn . For each t  n, since Δ ∩ Ht ⊂ (∪F )◦ , there is
mt ∈ N such that H (i, j, k, l, mt ) is a finite minimal interior cover of Δ ∩ Ht . Take
m = max{mt : t  n}. Then p ∈ / W (i, j, k, l, m). So

Δ = ∩{W (i, j, k, l, m) : i, j, k, l, m ∈ N},

and hence X has a Gδ -diagonal. 

By Proposition 3.1.2 and Theorem 3.1.8, we have the following corollary.

Corollary 3.1.9 ([167]) A quotient s-image of a metric space is a metrizable space

if and only if it is an M-space.
To illustrate more properties of point-countable covers, below we introduce the
concepts of monotonically monolithic spaces and weakly monotonically monolithic
spaces.
3.1 Spaces with Point-Countable Covers 153

Definition 3.1.10 A space X is monotonically monolithic [466] if one can assign to

each A ⊂ X a family P(A) of subsets of X such that
(1) |P(A)|  max{|A|, ω};
(2) A ⊂ B ⇒ P(A) ⊂ P(B); 
(3) if {Aα : α
< δ} is an increasing family of subsets of X, and A = α<δ Aα , then
P(A) = α<δ P(Aα );
(4) P(A) contains a net at every point of A.

A space X is weakly monotonically monolithic [398] if it has an operator satisfying

the above conditions but with condition (4) replaced by
(w4) if A is not closed, then P(A) contains a net at some point x ∈ A − A.
Obviously, every monotonically monolithic space is weakly monotonically mono-
lithic. One of the functions of weakly monotonically monolithic spaces is the fol-
lowing result.

Theorem 3.1.11 ([398]) Weakly monotonically monolithic spaces are D-spaces.

Corollary 3.1.12 All the following spaces are D-space:

(1) regular k-spaces with a point-countable k-network [398];

(2) sequential spaces with a point-countable cs∗ -network [396];
(3) spaces with a point-countable weak base [84, 395].

Proof We only prove that every regular k-space with a point-countable k-network is
a D-space. Suppose P is a point-countable k-network for a regular k-space X. For
each A ⊂ X, let
P(A) = {P : P ∈ (P)A }.

Obviously, the family P(A) satisfies (1) − (3) of Definition 3.1.10. Below we verify
it also satisfies (w4) in Definition 3.1.10. If A is not closed in X, then by Corollary
2.3.5 and Lemma 3.1.6, X is a sequential space, so there is a sequence {xn } in A
converging to some point x ∈ X − A, and thus x ∈ A. If x ∈ U ∈ τ , then by the
regularity of X, there is an open set V in X such that x ∈ V ⊂ V ⊂ U. Since P
is a k-network for X, there exist P ∈ P and a subsequence {xni } of {xn } such that
{xni : i ∈ N} ⊂ P ⊂ V , so P ∈ P(A) and x ∈ P ⊂ U. Thus, X is a weakly
monotonically monolithic space. By Theorem 3.1.11, X is a D-space. 

Example 3.1.13 The relationships among spaces with different point-countable cov-
ers.
(1) There is a space with a point-countable k-network and a point-countable pseudo-
base which has no point-countable cs∗ -network. The space Sω1 (see Example
1.8.7) is such a space.
(2) There is a space with a point-countable cs-network which has no point-countable
pseudo-k-network. The compactification βN is such an example.
154 3 Generalized Metric Spaces

(3) There is a space with a point-countable base which has no point-countable closed
k-network. The pointed irrational extension topological space of R (see Example
2.7.14) is such an example. Furthermore, Foged [132] constructed a regular
space with a point-countable base which does not have a point-countable closed
k-network.
(4) There is a space with a point-countable pseudo-Fq-base which has no point-
countable pseudo-base.
Let X be the set ω1 + 1 with the topology such that ω1 is the only accumulation
point of X and the neighborhoods of it are the same as those in the ordered
topology. Since X does not have the point Gδ -property, X has no point-countable
pseudo-base. Define

P = {{α} : α  ω1 } ∪ {(α, ω1 ) : α < ω1 }.

Then P is a point-countable pseudo-Fq-base for X.

(5) There is a countably compact space with a point-countable closed k-network
which has no point-countable pseudo-Fq-base, the Frolík space in Example
2.4.15(4) is such an example.
The following example shows that having a point-countable closed k-network
does not imply having a point-countable cs-network.
Example 3.1.14 ([167]) There exist a locally compact metric space Z, a compact-
covering quotient compact mapping f : Z → X and a perfect mapping g : X → Y
such that
(1) X has a point-countable closed k-network;
(2) X has no point-countable cs-network [291];
(3) Y has no point-countable cs∗ -network.
Suppose S(x) is a copy of S1 for each x ∈ I. Take

Z = I ⊕ (⊕{S(x) : x ∈ I}).

Then Z is a locally compact metric space. Let X be the quotient space of Z obtained
by identifying each x ∈ I and the limit point of S(x) to one point, and we denote
the quotient mapping by f . Then f is a compact mapping. By Theorem 2.7.12, f
is a compact-covering mapping and X has a point-countable closed k-network. By
Theorem 2.10.6, X is a g-first countable space. We denote the quotient space of X
by identifying the compact set I to one point as Y and denote the quotient mapping
by g. Then g is a perfect mapping and Y contains a closed copy of Sω1 . By Example
1.8.7, Y has no point-countable cs∗ -network.
If X has a point-countable cs-network, then by Proposition 1.6.21, X has a point-
countable weak base. Suppose B is a point-countable weak base of X which is closed
under finite intersections. For each x ∈ X, note that each point of X − I is isolated
in X and the subspace I of X is an Euclidean subspace, we can choose a countable
weak base Bx = {Bx,n }n∈N of x such that Bx ⊂ B and
3.1 Spaces with Point-Countable Covers 155

(14.1) for each n ∈ N, Bx,n+1 ⊂ Bx,n ;

(14.2) for each x ∈ X − I, Bx,n = {x};
(14.3) for each x ∈ I, there is a sequence {Vx,n } of open intervals containing x in I
with rational ends and Vx,n ⊂ Bx,n .
Let
P = ∪{Bx : x ∈ X}, P
= ∪{Bx : x ∈ I}.

Then P is a weak base of X. On the one hand, P − P

is a σ -discrete family of
sets in X. On the other hand, for each fixed Vx,n , by the point-countability of B,
{B ∈ P
: Vx,n ⊂ B} is countable, and by

P
= {B ∈ P
: there are n ∈ N, x ∈ I such that Vx,n ⊂ B},

P
is also countable. So X has a σ -discrete weak base, and hence X is an ℵ-space.
Assume that R is a σ -locally finite closed k-network for X. Then g(R) is a σ -locally
finite closed k-network for Y , a contradiction. Thus X has no point-countable cs-
network.
In the second part of this section, we discuss mapping properties of spaces with
a point-countable cover.
Proposition 3.1.15 ([167]) The following properties are invariant under perfect
mappings:
(1) having a point-countable pseudo-Fq-base;
(2) having a point-countable pseudo-k-network.
Proof We only prove (1), because the proof of (2) is similar. Suppose f : X → Y is
a perfect mapping and P is a point-countable pseudo-Fq-base for X. Define

H(F ) = {y ∈ Y : F is a minimal cover of f −1 (y)}, F ∈ P <ω ;

H = {H(F ) : F ∈ P <ω }.

By the Miščenko lemma, H is a point-countable family of sets in Y . For every y, z ∈

Y with y = z, if x ∈ f −1 (z), then f −1 (y) ⊂ X − {x}, and hence there is F ∈ P <ω
such that f −1 (y) ⊂ (∪F )◦ ⊂ ∪F ⊂ X − {x}. Let R = {H(F
) : F
⊂ F }. Then
R ∈ H <ω and ∪R = {p ∈ Y : f −1 (p) ⊂ ∪F }, so y ∈ Y − f (X − (∪F )◦ ) ⊂ ∪R,
and hence y ∈ (∪R)◦ ⊂ ∪R ⊂ Y − {z}. Thus Y has a point-countable pseudo-Fq-
base. 
Proposition 3.1.16 Suppose the restriction of a mapping f : X → Y to every closed
subspace of X is a compact-covering mapping. If X has a point-countable k-network,
then Y also has a point-countable k-network.
Proof By Lemma 3.1.6 and the fact that f is a compact-covering mapping, every
compact subspace of Y is a sequential space. Suppose P is a point-countable k-
network for X. For each y ∈ Y , pick xy ∈ f −1 (y). Let D = {xy : y ∈ Y } and
156 3 Generalized Metric Spaces

Q = {f (D ∩ P) : P ∈ P}. Then Q is point-countable cover of Y . Suppose {yn }

is a sequence in Y converging to y ∈ U ∈ τ (Y ). We may assume that all yn ∈ U.
Let Z = {xn : n ∈ N} ∪ f −1 (y), where xn = xyn . Then Z is a closed set in X and
f|Z : Z → {y} ∪ {yn : n ∈ N} is a compact-covering mapping. Take a compact set
L in Z such that f (L) = {y} ∪ {yn : n ∈ N}. Then xn ∈ L ⊂ f −1 (U), so there is
P ∈ P such that P contains infinitely many terms of {xn } and P ⊂ f −1 (U). Hence,
f (D ∩ P) contains infinitely many terms of {yn } and f (D ∩ P) ⊂ U. By the remark
after Proposition 1.6.7, Q is a point-countable k-network for Y . 

By Proposition 2.1.16, we have the following corollary.

Corollary 3.1.17 ([167, 291]) The property of having a point-countable k-network

is invariant under closed L-mappings on regular spaces or under perfect mappings.
The following examples show that some classes of spaces with certain point-
countable covers are preserved by neither perfect mappings nor closed or open s-
mappings.

Lemma 3.1.18 (Tanaka, 2000) Every completely regular space is a perfect image
of a completely regular space with a point-countable cs-network.

Proof Suppose X is a completely regular space. Let D be the space obtained by giving
X the discrete topology and let f : D → X be the identical mapping. By the Tychonoff
compact extension theorem, f has a continuous extension h = βf : βD → βX. Then
h is a perfect mapping. Let E = {z ∈ βD : h(z) ∈ X} and g = h|E : E → X. Then
E is a completely regular space and g is also a perfect mapping. Since there is no
nontrivial convergent sequence in βD, there does not exist any nontrivial convergent
sequence in E either. Let P = {{z} : z ∈ E}. Then P is a point-countable cs-
network for E, and hence X is a perfect image of a completely regular space with a
point-countable cs-network. 

The following Example 3.1.19 is a preparing of Example 3.1.20.

Example 3.1.19 Denote the Arens space as S2 (see Example 1.8.6). Then
(1) S2 is a finite-to-one quotient image of a separable metric space;
(2) the sequential fan Sω is a perfect image of S2 .
Denote the Arens space as X = ω ∪ (N × N). If we use the notations in Example
1.8.6, then X has the weak topology with respect to the cover

H = {ω} ∪ {V (n, 1) : n ∈ N}.

Let f : ⊕H → X be the natural mapping. Then ⊕H is a separable metric space

and f is a finite-to-one quotient mapping.
3.1 Spaces with Point-Countable Covers 157

For the sequential fan Y = {0} ∪ (N × N) (see Example 1.8.7), define g : X → Y

as follows: 
0, x ∈ ω,
g(x) =
x, x ∈ N × N.

Then g is a perfect mapping.

Example 3.1.20 The following spaces or properties are not preserved by perfect
mappings:

(1) quotient π -images of metric spaces, Example 3.1.19 is such an example, because
Sω is not a g-first countable space;
(2) quotient compact images of metric spaces, Example 3.1.19 is such an example,
because Sω is not a g-first countable space;
(3) pseudo-open π -images of metric spaces, Example 2.4.14 is such an example,
because by Theorem 2.9.14, the butterfly space is a pseudo-open π -image of a
metric space;
(4) quotient s-images of metric spaces, because by Corollary 2.7.5, the space Y in
Example 3.1.14 is not a quotient s-image of a metric space;
(5) having a point-countable weak base, Example 3.1.19 is such an example;
(6) having a point-countable cs-network, the fan space Sω1 is such an example by
Lemma 3.1.18;
(7) having a point-countable closed k-network, Example 3.1.14 is such an example;
(8) having a point-countable cs∗ -network, Example 3.1.14 is such an example.

Question 3.1.21 ([82]) Are spaces with a point-countable pseudo-base preserved

by perfect mappings?

Question 3.1.22 ([167]) Are pseudo-open s-images of metric spaces preserved by

perfect mappings or pseudo-open s-mappings?
A point x in a space X is called a P-point if, the intersection of countably many
neighborhoods of x is again a neighborhood of x in X. A point x in a space X is called
a weak P-point if, for each countable set C in X − {x}, x ∈ / C. Obviously, every
P-point is a weak P-point in a space. It is well-known that the space N∗ = βN − N
contains a P-point under CH [414] and many weak P-points [236].

Example 3.1.23 ([416]) There is a space with a point-countable k-network which is

not preserved by a closed mapping.
Let D be a set of cardinality ℵ1 . Give D the discrete topology. Define

D∗ = βD − D,
E = {p ∈ D∗ : p is a weak P-point of D∗ }.

Let X = D ∪ E. Give X the subspace topology of βD. Then E is a closed subspace

of X.
158 3 Generalized Metric Spaces

(23.1) E is a dense subset of D∗ .

Suppose V ∈ τ (βD) and V ∩ D∗ = ∅. Then there is W ∈ τ (βD) such that
clβD (W ) ⊂ V and W ∩ D∗ = ∅. Since D is dense in βD, W ∩ D contains a
countably infinite subset C. Because clβD (C) is homeomorphic to βN,

∅ = clβD (C) ∩ E ⊂ clβD (W ) ∩ D∗ ⊂ V ∩ D∗ ,

so E ∩ V ∩ D∗ = ∅. Hence E is a dense subset of D∗ .

(23.2) If K ∈ K (X), then K is a finite set.
Since each countable subset of a space consisting of weak P-points is discrete
and closed, so each compact space consisting of weak P-points is finite, and
hence K ∩ E is a finite set. If K is an infinite set, then K ∩ D is also an infinite
set. Fix a nontrivial sequence {xn } in K ∩ D. Then there is p ∈ (clβD {xn : n ∈
N}) ∩ E ⊂ K. Because any subsequence of {xn } has an accumulation point
and this accumulation point only can be in K ∩ D∗ , and K ∩ E is a finite set,
there is a subsequence of {xn } converging to p, which contradicts the fact that
there is no nontrivial convergent sequence in βD.
Let Y = X/E and let f : X → Y be the quotient mapping. Then f is a closed
mapping.
(23.3) Y is homeomorphic to the Alexandroff one point compactification ωD of D.
Let Y = D ∪ {e}, where {e} = f (E). Suppose e ∈ U ⊂ Y . If Y − U is a finite
set, then U ∈ τ (Y ). If Y − U is an infinite set, then E ∩ f −1 (Y − U) = ∅.
But
E ∩ clX (f −1 (Y − U)) = E ∩ clβD (f −1 (Y − U)) = ∅,

so f −1 (Y − U) ∈
/ τ c (X), and hence U ∈/ τ (Y ), which means U ∈ τ (Y ) if and
only if Y − U is a finite set. Thus Y is homeomorphic to ωD.

By (23.2), X has a point-countable k-network. Since the singleton ωD − D is not

a Gδ -set in ωD, ωD is not a metrizable space. By Lemma 3.1.6, Y has no point-
countable k-network.

Question 3.1.24 ([416]) Can every space be represented as a closed image of a

space with a point-countable k-network?
Gruenhage, Michael and Tanaka [167] raised the question whether quotient s-
images of metric spaces are invariant under open s-mappings. The answer of this
question is negative.

Example 3.1.25 ([309]) There is a quotient compact image X of a metrizable space

such that Sω1 is an open compact image of X.
(25.1) R has ω1 many disjoint dense sets with respect to the Euclidean topology.
For each r ∈ R, let r + Q = {r + q : q ∈ Q}. Pick p0 ∈ P. Then p0 + Q is
a dense set in R that is disjoint with Q. For each α < ω1 , suppose we have
chosen a family {pβ + Q : β < α} of disjoint dense sets. Let A = R − ∪{pβ +
Q : β < α} and take pα ∈ A ∩ P. Then (pα + Q) ∩ (pβ + Q) = ∅ for every
3.1 Spaces with Point-Countable Covers 159

β < α. Because otherwise, there exist r1 , r2 ∈ Q such that pα + r1 = pβ + r2 ,

so pα = pβ + r2 − r1 ∈ pβ + Q, a contradiction. Thus, there are ω1 many
disjoint dense sets in R, and we denote them by {pα + Q : α < ω1 }.
(25.2) Construct a quotient compact image X of a metrizable space.
For each α < ω1 , let (pα + Q) ∩ I = {pα + rn : n ∈ N}. For every n, j ∈ N, let
xj (pα , rn ) = (pα + rn , 1/j) and x(pα , rn ) = (pα + rn , 0). Then xj (pα , rn ) →
x(pα , rn ) in R2 . Define
 

Mα = ({xj (pα , rn ) : j ∈ N} ∪ {x(pα , rn )}) ∪ {xα (j) : α < ω1 , j ∈ N},
n∈N

where xα (j) ∈ R2 . Define a topology on Mα as follows: each xj (pα , rn ) is

isolated; each element of a neighborhood base of xα (j) has the form {xα (j)} ∪
{xj (pα , rn ) : n  m}, m ∈ N; each element of a neighborhood base of
x(pα , rn ) has the form {x(pα , rn )} ∪ {xj (pα , rn ) : j  m}, m ∈ N. It is easy to
verify that Mα is a countable first countable regular space, hence a metrizable
space. Let M = α<ω1 Mα and let X be the quotient space of I ⊕ M obtained
by identifying each x(pα , rn ) and pα + rn to a point. Then X is a quotient
compact image of a metrizable space.
(25.3) Construct an open compact mapping f : X → Sω1 .
Let Sω1 = {∞} ∪ {xj (α) : j ∈ N, α < ω1 }, where xj (α) → ∞ for each
α < ω1 . Define f : X → Sω1 as follows:

f (I) = {∞} , and f ({xj (pα , rn ) : n ∈ N} ∪ {xα (j)}) = {xj (α)}.

Then f is an open compact mapping, hence an s-mapping.

By the fact that Sω1 is not a quotient s-image of a metric space, we have
(25.4) a quotient s-image of a metric space may not be preserved by an open s-
mapping;
(25.5) a g-first countable space may not be preserved by an open compact mapping.
Furthermore, we can ask whether k-spaces with a point-countable k-network are
preserved by quotient compact mappings [245]. C. Liu [303] and Shibakov [421]
answered this question negatively by constructing a regular space X, such that, X is
a quotient compact image of a metric space and an open compact image of X has no
point-countable k-network.
A family P of sets in a space X is called compact-countable if (P)K is count-
able for each K ∈ K (X). Obviously, every locally countable family is a compact-
countable family, hence a point-countable family. On the other hand, it is easy to
verify that every point-countable base is a compact-countable base.

Question 3.1.26 ([311]) Does every regular Fréchet–Urysohn space with a point-
countable k-network have a compact-countable k-network?
160 3 Generalized Metric Spaces

3.2 Σ-Spaces

In the previous two chapters, we made a preliminary discussion on the properties

of strong Σ-spaces. In particular, we have found the prominent role of this class
of spaces on studying paracompactness of product spaces and images of metric
spaces. The purpose of this section is to investigate properties of spaces determined
by (mod k)-network systematically following Nagami [368], Michael [333] and
Okuyamas’ [386] way of thinking, especially, to prove the class of strong Σ-spaces
is equivalent to class of spaces with a σ -discrete closed (mod k)-network.
This section mainly consists of three parts. In the first part, we give the definition
and characterizations of Σ-spaces. In the second part, we establish the relationships
between spaces determined by different (mod k)-networks, including a result proved
by L. Peng [394] in 2005 that every strong Σ ∗ -space of the point Gδ -property is
a strong Σ-space. In the third part, we consider several mapping theorems, partic-
ularly a decomposition theorem of strong Σ-spaces. Finally, we show that some
implications do not hold by examples.

Definition 3.2.1 (1) A family P of set in a space X is called a quasi-(mod k)-

network for X [386] if, there is a cover H of X by closed countably compact
sets, such that, for each C ∈ H , {P ∈ P : C ⊂ P} is a network of C in X. P
is also called a quasi-(mod k)-network w.r.t. H .
(2) A space with a σ -locally finite (resp. σ -HCP, σ -closure-preserving) closed
quasi-(mod k)-network is called a Σ-space [368] (resp. Σ ∗ -space [386], Σ -
space [386]).
(3) A space with a σ -HCP (resp. σ -closure-preserving) closed (mod k)-network is
called a strong Σ ∗ -space [386] (resp. strong Σ -space [333]).
(4) We can define pair-quasi-(mod k)-networks and pair-(mod k)-networks similar
to Definition 1.5.3.
Obviously, every Σ-space is a Σ ∗ -space, every Σ ∗ -space is a Σ -space, and every

Σ -space is a space with a σ -cushioned quasi-(mod k)-network. By Proposition
1.5.4, every semi-stratifiable space is a space with a σ -cushioned (mod k)-network.
If X has a σ -locally finite cover, a σ -HCP cover or a σ -closure-preserving cover by
countably compact closed sets (resp. compact sets), then X is a Σ-space (resp. strong
Σ-space), a Σ ∗ -space (resp. strong Σ ∗ -space) or a Σ -space (resp. strong Σ -space)
respectively. It is easy to show that the following topological spaces are all additive
and closed hereditary: Σ-spaces (strong Σ-spaces), Σ ∗ -spaces (strong Σ ∗ -spaces)
and Σ -spaces (strong Σ -spaces).
In the following, we give some characterizations of spaces defined in the (2) and
(3) of Definition 3.2.1. For the convenience of narration, in this section, we denote
∩(P)x as C(P, x) for each point x ∈ X and any cover P of a space X.

Proposition 3.2.2 ([386]) A space X is a Σ-space if and only if X has a sequence

{Fn } of locally finite closed covers, such that, if xn ∈ C(Fn , x) for every x and {xn }
in X, then {xn } has an accumulation point.
3.2 Σ-Spaces 161


Proof Suppose n∈N Pn is a closed quasi-(mod k)-network w.r.t. H in the Σ-space
X, such that, each Pn is locally finite and Pn ⊂ Pn+1 . Define Fn = Pn ∪ {X}.
Then Fn is a locally finite closed cover of X. If there exist x ∈ X and a sequence
{xn } without any accumulation point in X such that xn ∈ C(Fn , x), take H ∈ (H )x ,
then there is m ∈ N such that H ⊂ X − {xn : n  m}, and hence there exist k ∈ N
and P ∈ Pk such that H ⊂ P ⊂ X − {xn : n  m}. Pick j  max{m, k}. Then
xj ∈ C(Fj , x) ⊂ P ⊂ X − {xj }, a contradiction.
Conversely, we may assume that Fn ⊂ Fn+1 and  each Fn is closed under finite
intersections. Then C(Fn , x) ∈ Fn . Let H = { n∈N C(Fn , x) : x ∈ X}. By the
 lemma, H is a cover of X consisting of countably compact closed
convergence
sets and n∈N Fn is a closed quasi-(mod k)-network for X w.r.t. H . Thus X is a
Σ-space. 
Proposition 3.2.3 ([372]) A space X is a Σ -space if and only if there is a Σ -
function on X, i.e. there is a g-function on X such that
(1) for every x, y ∈ X and n ∈ N, if x ∈ g(n, y), then g(n, x) ⊂ g(n, y);
(2) if x ∈ X and {xn } is a sequence in X with x ∈ g(n, xn ) for each n ∈ N, then {xn }
has an accumulation point.

Proof The necessity. Suppose n∈N Pn is a closed quasi-(mod k)-network w.r.t. H
for the Σ -space X such that each Pn is closure-preserving. Define g : N × X → τ
by
g(n, x) = X − ∪{P ∈ Pi : i  n, x ∈ / P}.

Obviously, g is a g-function satisfying condition (1). If there exist x ∈ X and a

sequence {xn } such that x ∈ g(n, xn ) and {xn } has no accumulation point, then there
exist H ∈ (H )x , m, k ∈ N and P ∈ Pk such that H ⊂ P ⊂ X − {xn : n  m}. Pick
j  max{m, k}. Then g(j, xj ) ⊂ X − P ⊂ X − {x}, a contradiction. So g also satisfies
condition (2).
The sufficiency. Suppose g is a Σ -function on X. For each n ∈ N, let

Pn = {X − g(n, A) : A ⊂ X}.

By (1), Pn is a closure-preserving family of closed sets and C(Pn , x) ∈ Pn for

each x ∈ X. If x ∈ X and {xn } is a sequence in X with xn ∈ C(Pn , x) for each n ∈ N,
then x ∈ g(n, xn ), and hence {xn } has an accumulation point. Let

H ={ C(Pn , x) : x ∈ X}.
n∈N


By the convergence lemma, n∈N Pn is a closed quasi-(mod k)-network for X w.r.t.
H . Thus X is a Σ -space. 
Proposition 3.2.4 ([290, 386]) A space X a Σ -space if and only if there is a
sequence {Fn } of closure-preserving closed covers of X, such that, if xn ∈ C(Fn , x)
for some x ∈ X and each n ∈ N, then the sequence {xn } has an accumulation point.
162 3 Generalized Metric Spaces

Proof Proof of the necessity is similar to that in Proposition 3.2.2. In the following,
we prove the sufficiency. Let {Fn } be a sequence of closure-preserving closed covers
of X such that any sequence {xn } with xn ∈ C(Fn , x) for some x ∈ X has an
accumulation point. Without loss of generality, we can assume that Fn ⊂ Fn+1 for
each n ∈ N. For each n ∈ N, put Pn = {∩E : ∅ = E ⊂ Fn }. Then Pn is a closure-
preserving closed cover of X. In fact, it is easy to see Pn is a closed cover of X. Let
{∩Eα : α ∈ Γ } ⊂ Pn , where Eα ⊂ Fn for each α ∈ Γ . If x ∈ / ∪{∩Eα : α ∈ Γ }, then
there is an Eα ∈ Eα such that x ∈
/ Eα for each α ∈ Γ . Since Fn is closure-preserving,

x ∈ X − ∪{Eα : α ∈ Γ }
= X − ∪{Eα : α ∈ Γ } ⊂ X − ∪{∩Eα : α ∈ Γ },

which shows that Pn is closure-preserving. Moreover, we have C(Pn , x) ⊂

C(Fn , x) and C(Pn , x) ⊂ Pn for each x ∈ X and n ∈ N.
Put
H = { C(Pn , x) : x ∈ X}
n∈N

Then H is a closedcover of X by countably compact sets. For each x ∈ X and

an open set U with  n∈N C(Pn , x) ⊂ U, we prove that there is m ∈ N such that
C(Pm , x) ⊂ U; thus, n∈N Pn is a quasi-(modk)-network for X. Suppose not, we
can choose a sequence {xn } with xn ∈ C(Pn , x) − U for each n ∈ N, then {xn }
has anaccumulation point y. Since y ∈ {xm : m  n} ⊂ C(Pn , x) for each n ∈ N,
y ∈ n∈N C(Pn , x) ⊂ U. However, y ∈ {xn : n ∈ N} ⊂ X − U = X − U, a
contradiction. Hence, X is a Σ -space. 
Question 3.2.5 ([290]) If X has a sequence {Fn } of HCP closed covers such that
any sequence {xn } with xn ∈ C(Fn , x) for some x ∈ X has an accumulation point, is
X a Σ ∗ -space?
Proposition 3.2.6 ([290]) A space X has a σ -cushioned quasi-(mod k)-network if
and only if there exist a cover H of X consisting of countably compact closed sets
and a g-function on X, such that, for each H ∈ H and any sequence{xn } in X with
H ∩ g(n, xn ) = ∅, {xn } has an accumulation point in H.

Proof The necessity. Suppose n∈N Pn is a σ -cushioned quasi-(mod k)-network
for X w.r.t. H such that each Pn is a cushioned family. Define a g-function g :
N × X → τ on X as follows:

g(n, x) = X − ∪{P1 : (P1 , P2 ) ∈ Pi , i  n, x ∈

/ P2 }.

If H ∈ H and {xn } is a sequence in X with H ∩ g(n, xn ) = ∅ and {xn } has no

accumulation point in H, then there is m ∈ N such that H ∩ {xn : n  m} = ∅.
Because otherwise, let Fk = {xn : n  k} for each k ∈ N. Then
 {H ∩Fk } is a sequence
of nonempty decreasing closed sets in H, so there is h ∈ k∈N (H ∩ Fk ), it follows
that h is an accumulation point of {xn } in H, a contradiction. Thus there is m ∈ N
3.2 Σ-Spaces 163

such that H ⊂ X − {xn : n  m}, and hence there exist i ∈ N and (P1 , P2 ) ∈ Pi
such that H ⊂ P1 ⊂ P2 ⊂ X − {xn : n  m}. Pick j  max{m, i}. Then g(j, xj ) ⊂
X − P1 ⊂ X − H, a contradiction.
The sufficiency. For each n ∈ N, define

P(n, U) = X − g(n, X − U), U ∈ τ ;

Pn = {(P(n, U), U) : U ∈ τ }.

Then Pn is a cushioned family in X. Suppose H ∈ H , U ∈ τ and H ⊂ U. If

P(n, U) ⊃ H for each n ∈ N, then there is a sequence {xn } in X − U such that
H ∩ g(n, xn ) = ∅, so {xn } has an accumulation point in H − U, a contradiction. Thus
X has a σ -cushioned quasi-(mod k)-network. 

Consequently, every space with a σ -cushioned quasi-(mod k)-network is a β-

space.

Theorem 3.2.7 ([217]) For every space X, the following are equivalent:
(1) X is a strong Σ -space.
(2) X is a submetacompact Σ -space.
(3) X is an isocompact Σ -space.

Proof By a characterization of submetacompactness obtained by Junnila (see The-

orem A.4.8 in Appendix A), every directed open cover of a submetacompact space
has a σ -closure-preserving closed refinement, and hence every strong Σ -space is a
submetacompact space. Other conclusions are obvious. 

Question 3.2.8 Is every space with a σ -cushioned (mod k)-network a submetacom-

pact space?
This question is relevant to the following Junnila–Katuta question [217, 227]: if
every open cover of a space has a σ -cushioned refinement, then is it a submetacompact
space? If the answer of the Junnila–Katuta question is positive, then so is that of
Question 3.2.8.
For strong Σ ∗ -spaces, we have a conclusion parallel to Theorem 3.2.7.

Lemma 3.2.9 ([74]) If a family U of open sets in a space X covers K ∈ K (X),

then there is O ∈ τ such that K ⊂ O and U|O has a finite closed refinement in O.

Proof There is a finite subfamily {Ui }in of U covering K. For each x ∈ K, there is
i(x)  n such that x ∈ Ui(x) . Let Fx = K − Ui(x) . Then there exist Vx , Wx ∈ τ such
that x ∈ Vx , Fx ⊂ Wx and Vx ∩ Wx = ∅. Let G x = Wx ∪ Ui(x) . Then K ⊂ Gx ∈ τ .
Take a finite subset {xj }jm of K such that K ⊂ jm Vxj . Let
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
 
O=⎝ Gxj ⎠ ∩ ⎝ Vxj ⎠ ∩ ⎝ Ui ⎠ .
jm jm in
164 3 Generalized Metric Spaces

⊂ O ∈ τ . For each j  m, let Pj = clO (Vxj ∩ O). Then Pj is a closed set in

Then K 
O, O = jm Pj and Pj − Ui(xj ) ⊂ (Gxj − Ui(xj ) ) ∩ clX (Vxj ) ⊂ Wxj ∩ clX (Vxj ) = ∅.
So Pj ⊂ Ui(xj ) , and hence {Pj }jm is a closed refinement of U|O in O. 

Theorem 3.2.10 ([74]) For every space X, the following are equivalent:
(1) X is a strong Σ ∗ -space.
(2) X is a subparacompact Σ ∗ -space.
(3) X is an isocompact Σ ∗ -space.

Proof We only need to prove that every strong Σ ∗ -space X is a subparacompact

space. Let P be a σ -HCP closed (mod k)-network in X w.r.t. K . Suppose U is an
open cover of X. For each x ∈ X, pick Kx ∈ (K )x . By Lemma 3.2.9, there is Ox ∈ τ
such that Kx ⊂ Ox and U|Ox has a finite closed refinement Fx in Ox , and hence there
is Px ∈ P such that Kx ⊂ Px ⊂ Ox . Let F = {Px ∩ F : x ∈ X, F ∈ Fx }. Then F
is a σ -closure-preserving refinement of U . For every x ∈ X and F ∈ Fx , suppose
z ∈ X − (Px ∩ F), define

X − Px , z ∈ X − Px ,
Lz =
Ox − F, z ∈ Px − F.

Then z ∈ Lz ∈ τ and Lz ∩ Px ∩ F = ∅, so Px ∩ F is a closed set in X. Thus U has a

σ -closure-preserving closed refinement, and hence X is a subparacompact space.


The following theorem realizes our desires stated at the beginning of this chapter.

Theorem 3.2.11 ([74, 220]) For every space X, the following are equivalent:
(1) X has a σ -discrete closed (mod k)-network.
(2) X is a strong Σ-space.
(3) X is a subparacompact Σ-space.
(4) X is an isocompact Σ-space.

Proof We only prove (2)  ⇒ (1). The proofs of other parts are similar to those of
Theorem 3.2.10. Suppose n∈N Pn is a σ -locally finite closed (mod k)-network in
X w.r.t. K such that each Pn is locally finite. Since X is a subparacompact space,
there is a σ -discrete closed cover Fn of X for each n ∈ N, such that, every element
of Fn only meets finitely many elements of Pn , so Pn ∧ Fn is a σ -discrete family
of closed sets. Let
  <ω 

P = ∩F : F ∈ Pn ∧ Fn .
n∈N

Then P is a σ -discrete family of closed sets in X which is closed under finite

intersections. If x ∈ K ∈ K and K ⊂ U ∈ τ , then there exist n ∈ N, P ∈ Pn and
F ∈ Fn such that x ∈ F and K ⊂ P ⊂ U, so P ∩ F ∈ P and x ∈ P ∩ F ⊂ U, and
3.2 Σ-Spaces 165

hence P is an almost (mod k)-network for X. By Lemma 1.5.13, P is a σ -discrete

closed (mod k)-network for X. 

Question 3.2.12 ([450]) Does every Σ-space have a σ -discrete closed quasi-
(mod k)-network?
In the second part of this section, we investigate relationships among spaces having
different (mod k)-networks.
Corollary 2.11.8 established the relationship between spaces with a countable
(mod k)-network and perfect preimages of separable metric spaces. We shall further
investigate characterizations of spaces with a countable (mod k)-network. Note that
first, if the assumption of almost (mod k)-network in Lemma 1.5.13 is replaced
with that of almost quasi-(mod k)-network, then there is a closed cover H of X by
countably compact sets satisfying the relevant conditions, i.e. we have the following
lemma.

Lemma 3.2.13 Let P be a point-countable family of closed sets in a space X which

is closed under finite intersections. If P is an almost quasi-(mod k)-network for X,
then there is a closed cover H of X by countably compact sets such that
(1) P is a quasi-(mod k)-network w.r.t. H ;
(2) if x ∈ P ∈ P, then there is H ∈ H such that x ∈ H ⊂ P.

Lemma 3.2.14 ([256]) Suppose P is a weakly HCP family of sets in an ℵ1 -compact

space X. Let
E = {x ∈ X : |(P)x | > ℵ0 }.

Then
(1) {P − E : P ∈ P} is countable.
(2) E is a discrete countable closed subspace of X.

Proof (1) If {P − E : P ∈ P} is uncountable, then there exist an uncountable

subfamily {Pα }α<ω1 of P and a subset {xα : α < ω1 } of X such that xα ∈ Pα − E.
Since P is weakly HCP and X is ℵ1 -compact, we may assume that there is x ∈ X −E
such that xα = x for every α < ω1 , which contradicts the definition of E.
(2) Suppose Z ⊂ E with |Z|  ℵ1 . Denote Z = {xα : α ∈ Λ}. By the definition of
E and the Zermelo theorem on well-ordering, we can choose by transfinite induction
a subset {Pα }α∈Λ of P such that all Pα are different and xα ∈ Pα , so Z is a discrete
countable closed subspace of X, and hence E is a discrete countable closed subspace
of X. 

Theorem 3.2.15 A space X has a countable closed quasi-(mod k)-network if and

only if X is an ℵ1 -compact Σ ∗ -space.

Proof The necessity [462]. Let P be a countable closed quasi-(mod k)-network

for X w.r.t. H . Suppose X has an uncountable discrete closed subspace Z, we may
assume Z = {xα : α < ω1 }. For each α < ω1 , take Cα ∈ H such that xα ∈ Cα .
166 3 Generalized Metric Spaces

Then |Z ∩ Cα | < ℵ0 . So there exist β(α) < ω1 and Pα ∈ P such that Cα ⊂ Pα ⊂

X − {xβ : β(α) < β < ω1 }. Since the family {Pα }α<ω1 is actually a countable
family, we can rearrange it as {Pn }n∈N . For each n ∈ N, there is βn < ω1 such that
Pn ⊂ X − {xβ : βn < β < ω1 }. Pick γ < ω1 such that γ > sup{βn : n ∈ N}. Then
xγ ∈ n∈N Pn ⊂ X − {xγ }, a contradiction. Thus X is an ℵ1 -compact space.
The sufficiency. Suppose n∈N Pn is a closed quasi-(mod k)-network in an ℵ1 -
compact space X w.r.t. H such that each Pn is an HCP family. For each n ∈ N,
define

En = {x ∈ X : |(Pn )x | > ℵ0 },
Rn = {P − En : P ∈ Pn } ∪ {{x} : x ∈ En }.

By Lemma 3.2.14, |Rn |  ℵ0 . Let

  <ω 

F = ∩R : R ∈ Rn .
n∈N

Then |F |  ℵ0 and F is a family of closed sets which is closed under finite

intersections. For every x ∈ C ∈ H and C ⊂ U ∈ τ , there exist n ∈ N and P ∈ Pn
such that C ⊂ P ⊂ U. Define

{x}, x ∈ En ,
F=
P − En , x ∈ X − En .

Then F ∈ F and x ∈ F ⊂ U. So F is an almost quasi-(mod k)-network for X. By

Lemma 3.2.13, F is a closed quasi-(mod k)-network for X. 

Corollary 3.2.16 ([386]) For every space X, the following are equivalent:
(1) X has a countable closed (mod k)-network.
(2) X is a Lindelöf Σ ∗ -space.
(3) X is an ℵ1 -compact strong Σ ∗ -space.

Proof (1) ⇒ (2). Suppose X has a countable closed (mod k)-network P. For any
open cover U of X, denote {P ∈ P : there is V ∈ U <ω such that  P ⊂ ∪V } as
{Pn }n∈N . For each n ∈ N, fix Un ∈ U <ω such that Pn ⊂ ∪Un . Then n∈N Un is a
countable subcover of U . So X is a Lindelöf space.
(2) ⇒ (3) is obvious. (3) ⇒ (1) is obtained by Theorems 3.2.15 and 3.2.7. 

In the class of regular spaces, every first countable space is equivalent to a q-space
of the point Gδ -property (see Theorem 1.7.7). When discussing the problem what
class of Σ ∗ -spaces are Σ-spaces, we may attach appropriate generalized sequentiality
properties. Below we prove that every strong Σ ∗ -space having the q-space property
or every strong Σ ∗ -space of the point Gδ -property is a Σ-space.
3.2 Σ-Spaces 167

Lemma 3.2.17 ([261]) Let P be an HCP family of closed sets in a q-space X.

Define

D(P) = {x ∈ X : |(P)x |  ℵ0 },
R(P) = {P − D(P) : P ∈ P} ∪ {{x} : x ∈ D(P)}.

Then R(P) is a locally finite family in X.

Proof Let g be a q-function on X. We first prove that D(P) is a discrete closed

subspace of X. Otherwise, there exist A ⊂ D(P) and x ∈ A − A. Take a nontrivial
sequence {xn } in A such that xn ∈ g(n, x). Choose a subfamily {Pn }n∈N of P con-
sisting of different sets such that xn ∈ Pn . Then {xn : n ∈ N} is a discrete closed set
in X, a contradiction. To prove that R(P) is a locally finite family of sets in X, we
only need to prove {P − D(P) : P ∈ P} is a point-finite family in X. If there exist
y ∈ X and  a countable subfamily {Pn } n∈N of P such that y ∈ n∈N Pn − D(P),
then y ∈ n∈N Pn − {y}, and hence we can choose a nontrivial sequence {yn } in X
such that yn ∈ g(n, y) ∩ (Pn − {y}), a contradiction. Thus R(P) is a locally finite
family in X. 

Proposition 3.2.18 If a Σ ∗ -space X is a q-space, then X is a Σ-space.


Proof Suppose n∈N Pn is a closed quasi-(mod k)-network in the  q-space X w.r.t.
H such that each Pn is an HCP family. By Lemma 3.2.17, n∈N R(Pn ) is a
σ -locally finite family of closed sets in X. Define
  <ω 

F = ∩R : ∅ = R ∈ R(Pn ) .
n∈N

Then F is a σ -locally finite family of closed sets which is closed under finite inter-
sections in X. By the proof of sufficiency of Theorem 3.2.15, F is an almost quasi-
(mod k)-network for X. Further by Lemma 3.2.13, F is a closed quasi-(mod k)-
network for X. So X is a Σ-space. 

Lemma 3.2.19 ([393]) Every discrete closed set in a submetacompact space of the
point Gδ -property is a Gδ -set.

Proof Suppose X is a submetacompact space X of the point Gδ -property. Let F be

 each x ∈ F, there is a decreasing sequence {Vx,n }
a discrete closed set of X. For
of open sets in X such that n∈N Vx,n = {x} and F ∩ Vx,n = {x}. For each n ∈ N,
let Un = {Vx,n : x ∈ F} ∪ {X − F}. Then Un is an open cover of X, so there is a
sequence {Un,m }m∈N of open covers of X such that {Un,m }m∈N is an open θ -refinable
sequence of Vn = Un ∧ ( i+jn Ui,j ). For every n, m ∈ N, let On,m = ∪(Un,m )F .
Then F ⊂ On,m ∈τ .
Suppose x ∈ n,m∈N On,m . Let Fn,m = {U ∈ Un,m : x ∈ U ⊂ X − F}. By the
construction of Un,m , there is a sequence {Fnk ,mk }k∈N of finite families, such that,
168 3 Generalized Metric Spaces

Fnk+1 ,mk+1 partially refines Fnk ,mk ∧ Uk+1 and nk < nk+1 . By the König lemma, there
is a decreasing sequence {Uk } of sets  such that Uk∈ Fnk ,mk . Fix y ∈ U1 ∩ F. Then
Uk ∩ F = {y} and Uk ⊂ Vy,nk . So x ∈ k∈N Uk ⊂ k∈N Vy,nk = {y} ⊂ F, and hence
F = n,m∈N On,m . 

A subset F in a space X is said to be an Fσ -discrete set if F is the union of

countably many discrete closed sets in X.

Lemma 3.2.20 ([393]) Suppose P is a closure-preserving closed cover of a space

X. Let F = {x ∈ X : |C(P, x)| < ℵ0 }. If every discrete closed set in X is a Gδ -set,
then F is an Fσ -discrete set in X.

Proof Let Fn = {x ∈ X : |C(P, x)|  n} for each n ∈ N. We only need to prove

each Fn is an Fσ -discrete set in X.
(20.1) Fn is a closed set and F1 is a discrete set in X.
For each y ∈ X, let Uy = X − ∪{P ∈ P : y ∈ / P}. Then y ∈ Uy ∈ τ and
C(P, y) ⊂ C(P, x) whenever x ∈ Uy . So if y ∈ X − Fn , then Uy ∩ Fn = ∅,
and hence Fn is a closed set. Since Uy ∩ F1 ⊂ {y}, F1 is also a discrete set in
X.
(20.2) Fn+1 − Fn is a discrete set in X.
Let x ∈ Fn+1 − Fn . If y ∈ (Fn+1 − Fn ) − C(P, x), then C(P, x) ⊂ C(P, y),
and hence there is Py ∈ P such that y ∈ Py ⊂ X − {x}. Define

Vx = X − ∪{Py : y ∈ (Fn+1 − Fn ) − C(P, x)}.

Then x ∈ Vx ∈ τ and Vx ∩ (Fn+1 − Fn ) ⊂ C(P, x). Thus Fn+1 − Fn is a

discrete set in X.
(20.3) Each Fn is an Fσ -discrete set.
Since F1 is a discrete closed set in X, F1 is a Gδ -set in X, and
 hence there is
a sequence {Hm } of closed sets in X such that F2 − F1 = m∈N Hm . Since
Hm ∩ (F2 − F1 ) = Hm ∩ F2 is a discrete closed set in X, F2 − F1 is an
Fσ -discrete set in X, so F2 is an Fσ -discrete set in X.
For each m ∈ N, Hm ∩ F2 is a Gδ -set in X, so there is a sequence {Hm,k } of
closed sets in X such that F3 − (Hm ∩ F2 ) = k∈N Hm,k . Since  Hm,k ∩ (F3 −
F2 ) = Hm,k ∩ F3 is a discrete closed set in X, F3 − F2 = m,k∈N Hm,k is an
Fσ -discrete set in X. Thus F3 is an Fσ -discrete set in X.
By the inductive method, we can prove that each Fn is an Fσ -discrete set. 

Lemma 3.2.21 ([462]) Let P be an HCP family in a space X. For each n ∈ N,

define
Pn = {P1 ∩ P2 · · · ∩ Pn : Pi ∈ P, i  n}.

Then Pn is also an HCP family in X.

Proof Obviously, the conclusion holds when n = 1. Assume the conclusion holds
when some n = k ∈ N. For n = k + 1, let
3.2 Σ-Spaces 169

P = {Pα : α ∈ Λ}, Pk = {Qβ : β ∈ Γ }.

Then Pn = P ∧ Pk . Thus,

∪{Hα,β : (α, β) ∈ Λ × Γ } = ∪ {∪{Hα,β : β ∈ Γ } : α ∈ Λ}

= ∪ {H α,β : (α, β) ∈ Λ × Γ }

for every Hα,β ⊂ Pα ∩ Qβ ∈ Pn , and hence Pn is an HCP family in X. 

Lemma 3.2.22 ([394]) Suppose X is a Σ ∗ -space. If every discrete closed set in X

is a Gδ -set, then X is a Σ-space.

Proof Suppose P = n∈N Pn is a closed quasi-(mod k)-network for X w.r.t. H ,
such that, each Pn is an HCP cover of X and Pn ⊂ Pn+1 . By the proof of necessity
of Proposition 3.2.2, we have

(22.1) If xn ∈ C(Pn , x) for a point x and a sequence {xn } in X, then {xn } has an
accumulation point.
For each n ∈ N, let

Dn = {x ∈ X : |(Pn )x |  ℵ0 }, D = Dn ;
n∈N

Gn = {x ∈ X : |C(Pn , x)| < ℵ0 }, G = Gn ;
n∈N

C(x) = C(Pn , x), x ∈ X.
n∈N

(22.2) G is an Fσ -discrete set in X and Dn ⊂ G.

By Lemma 3.2.20, G is an Fσ -discrete set. If x ∈ Dn −G, then (Pn )x is infinite
and for every m  n, C(Pm , x) is also infinite. So one can inductively choose

xm ∈ C(Pm , x) − {xi : n  i < m},

Pm ∈ (Pn )x − {Pi : n  i < m}.

Then {xi : i  n} is a discrete closed set in X, which contradicts (22.1). Hence

Dn ⊂ G.
(22.3) Dn is a Gδ -set in X.
For each x∈ X, there is a decreasing sequence {g(n, x)} of open sets in X
such that n∈N g(n, x) = {x}. For every n, m, i, j ∈ N, define
170 3 Generalized Metric Spaces
⎧ ⎫
⎨ ⎬
Qn,m = Pk : Pk ∈ Pn , k  m ,
⎩ ⎭
km
⎧ ⎫
⎨  ⎬
Rn,m,i,j = Q− g(j, d) : Q ∈ Qn,m .
⎩ ⎭
d∈Q∩Di

Then both Qn,m and Rn,m,i,j are HCP families of closed sets in X. We prove
for each i ∈ N,

Di ⊂ (X − ∪{R ∈ Rn,m,i,j : R ∩ Di = ∅}) ⊂ G.
n,m,j∈N

In fact, suppose x ∈ X − G. Since Pn is not point-finite at any point of

Dn , every countable subset of Dn is a discrete closed set in X, and hence
there is n ∈ N such that C(Pn , x) ∩ Di is a finite set. Pick m ∈ N such that
|(Pn )x | = m. Then 
C(Pn , x) ∈ Qn,m . Let F = C(Pn , x) ∩ Di . Then there is
/ d∈F g(j, d), so x ∈ C(Pn , x) − d∈F g(j, d) ∈ Rn,m,i,j
j ∈ N such that x ∈
and (C(Pn , x) − d∈F g(j, d)) ∩ Di = ∅.
 (22.2), there is a sequence {Ak } of closed sets in X such that G − Di =
By
k∈N Ak . So
 

Di = (X − ∪{R ∈ Rn,m,i,j : R ∩ Di = ∅}) ∩ (X − Ak )
n,m,j∈N k∈N

is a Gδ -set in X.
(22.4) X is a Σ-space.
 ∗
Let D = n∈N Dn∗ , where Dn∗ is a discrete closed set in X and Dn∗ ⊂ Dn+1 .
Let Dn = m∈N On,m , where On,m is an open set in X. Define

Pn,m = {P − On,m : P ∈ Pn } ∪ {X}.

Then Pn,m is a locally finite closed cover of X. Because in fact, for each
x ∈ X, one may assume x ∈ X − On,m ⊂ X − Dn , it follows that Pn,m is
locally finite in X.
Define Gn = {{x} : x ∈ Dn∗ } ∪ {X}. Then Gn is a locally finite closed cover of
X. Let ⎛ ⎞  
  
P =⎝
∗
Pn,m ⎠ ∪ Gn = Fn ,
n,m∈N n∈N n∈N

where each Fn is a locally finite closed cover of X.

Suppose x is a point and {xi } is a sequence in X with xi ∈ C(Fi , x). If x ∈ D,
then there is n ∈ N such that C(Gn , x) = {x}, so {xi } converges to x. If
3.2 Σ-Spaces 171


x ∈ X − D, then x ∈ / n∈N Dn , so for each n ∈ N, there is mn ∈ N such
/ On,mn , and hence C(Pn,mn , x) ⊂ C(Pn , x). By (22.1), {xi } has an
that x ∈
accumulation point. By Proposition 3.2.2, X is a Σ-space. 

By Theorems 3.2.10 and 3.2.11, Lemmas 3.2.19 and 3.2.22, we have the following
result.

Theorem 3.2.23 ([394]) Every strong Σ ∗ -space of the point Gδ -property is a strong
Σ-space.

Question 3.2.24 ([386]) Is every perfect Σ -space a Σ-space?

Theorem 3.2.25 ([280]) Every space with a σ -cushioned (mod k)-network is a D-

space.

Proof Suppose n∈N Pn is a σ -cushioned (mod k)-network in a space X w.r.t. K ,
where each Pn is a well-ordered cushioned family and Pn ⊂ Pn+1 . Let φ be a
neighborhood assignment for X. We first define Dn ⊂ X inductively for each n ∈ N
as follows:
Let D0 = ∅. Assume for each m ∈ {0, 1, . . . , n − 1}, we have defined Dm ⊂ X.
To define Dn , we further define a finite set Dαn ⊂ X inductively for an ordinal α
in the following way:
Let D0n = ∅. Assume that for all 0  β < α, we have defined Dβn . Let
    
U = ∪ φ(d) : d ∈ ∪ Dβn : β < α ∪ Dn−1 .

We denote the following claim by Rαn : There exist K ∈ K , (P1 , P2 ) ∈ Pn and

{x1 , . . . , xk } ⊂ K − U, such that,

K ⊂ P1 ⊂ P2 ⊂ U ∪ φ(x1 ) ∪ · · · ∪ φ(xk ).

If the claim Rαn does not hold, then our inductive process on α is complete. Other-
wise, let (P1 , P2 ) be the first element of Pn satisfying Rαn , and let Dαn = {x1 , . . . , xk }.
Suppose γn is the first ordinal α such that Rαn is not true. Define
 

Dn = Dαn ∪ Dn−1 .
α<γn


Let D = n∈N Dn . Below we prove X is a D-space.

(25.1) X = d∈D φ(d). 
Otherwise,
 there is K ∈ K such that K − d∈D φ(d) = ∅. Let L = K −
d∈D φ(d). Then L is a nonempty compact set in X. Since φ is a neighborhood
assignment for X, there is a finite subset {x1 , . . . , xk } of L such that L ⊂
φ(x1 ) ∪ · · · ∪ φ(xk ). Let
172 3 Generalized Metric Spaces

M = K − (φ(x1 ) ∪ · · · ∪ φ(xk )).


Then M ⊂ K −L ⊂ d∈D φ(d). By the compactness of M, there is j ∈ N
such that M ⊂ d∈Dj φ(d), so
⎛ ⎞

K ⊂ φ(x1 ) ∪ · · · ∪ φ(xk ) ∪ ⎝ φ(d)⎠ .
d∈Dj

Since P is a (mod k)-network in X w.r.t. K , there exist m ∈ N and (P1 , P2 ) ∈

Pm such that
⎛ ⎞

K ⊂ P1 ⊂ P2 ⊂ φ(x1 ) ∪ · · · ∪ φ(xk ) ∪ ⎝ φ(d)⎠ .
d∈Dj


Let n = max{j + 1, m} and U = ∪{φ(d) : d ∈ Dn−1 }. Then d∈Dj φ(d) ⊂
U, (P1 , P2 ) ∈ Pn and {x1 , . . . , xk } ⊂ K −U. So the claim R1n holds, it follows
that D1 = {x1 , . . . , xk } and K ⊂ P1 ⊂ P2 ⊂ d∈D φ(d), a contradiction.
n

(25.2) Dn is a discrete closed set in X.

Obviously, (25.2) is true for n = 0. Suppose Dn−1 is a discrete closed  set in X.
To prove Dn is a discrete closed set in X, we only need to prove α<γn Dαn is
discrete. For each α < γn , let (P1nα̃ , P2nα̃ ) be the first element of Pn satisfying
  
the claim Rαn . Suppose x ∈ α<γn Dαn . Since α<γn Dαn ⊂ α<γn P1nα̃ ,

x∈ P2nα̃ ⊂ ∪{φ(d) : d ∈ (∪{Dαn : α < γn }) ∪ Dn−1 }.
α<γn

 
Since Dαn ∩ d∈Dn−1 φ(d) = ∅, x ∈ ∪{φ(d) : d ∈ α<γn Dαn }. Let αx be the
smallest ordinal α in γn satisfying x ∈ ∪{φ(d) : d ∈ Dαn }. Define

V = ∪{φ(d) : d ∈ Dαn x } − ∪{P1nα̃ : α < αx }.


Then V  is an open neighborhood of x and V ∩ ( α<γn Dαn ) ⊂ Dαn x is finite.
Hence, α<γn Dαn is a discrete closed set.
(25.3) D is a discrete closed set in X. 
For any x ∈ X, by (25.1), there is n ∈ N such that x ∈ d∈Dn φ(d). Since
Dn is 
a discrete closed set, there is an open neighborhood W of x, such that,
W ⊂ d∈Dn φ(d) and W contains at most one point in Dn . So
⎛⎛ ⎞ ⎞

W ∩ D ⊂ ⎝⎝ φ(d)⎠ ∩ (D − Dn )⎠ ∪ (W ∩ Dn ).
d∈Dn
3.2 Σ-Spaces 173

For every y ∈ D − Dn , there exist m > n and α < γm such that y ∈ Dαm . Let
    
U = ∪ φ(d) : d ∈ ∪ Dβm : β < α ∪ Dm−1 .
 
Then d∈Dn φ(d)
 ⊂ U and Dα ∩ U = ∅. So y ∈
m
/ d∈Dn φ(d), and hence
(D − Dn ) ∩ ( d∈Dn φ(d)) = ∅, which shows W contains at most one point
of D, thus D is a discrete closed set. 

Theorem 3.2.26 ([397, 490]) The union of finitely many strong Σ-spaces is a D-
space.

Proof Suppose that a space X is the union of a finite collection {Xi }ik of strong
Σ-spaces. We argue by induction. If k = 1, the statement is obviously true. Assume
now that the statement holds for k = n, for some n ∈ N, and let us show that it is
true for k = n + 1.
Let φ be any neighborhood assignment of X. For the sake of simplicity,  for each
Y ⊂ X, denote φ(Y ) = y∈Y φ(y). For each i  n + 1, let Pi = j∈N Pi,j be
a closed (mod k)-network for Xi w.r.t. Ki , where Pi,j is locally finite in Xi and
Pi,j ⊂ Pi,j+1 for each j ∈ N. For each P ∈ Pi,j , let BP be the collection of all finite
subsets F ⊂ Xi for which there is an open subset U in Xi such that F ⊂ P−U ⊂ φ(F).
For each i  n + 1, put Ai,1 = {x ∈ X : Pi,1 is not locally finite at x}. We can
see that Ai,1 is a closed subset of X, and Ai,1 ∩ Xi = ∅. It follows by the inductive
assumption that Ai,1 is a D-space, so there is a locally finite subset Di,1 of Ai,1 , hence,
a locally finite subset of X such that Ai,1 ⊂ φ(Di,1 ). For each P ∈ Pi,1 , if there
exists some BP ∈ BP such that BP ⊂ P − φ(Di,1 ) ⊂ φ(BP ), choose just one BP

for P and denote it by  BP , else we put BP = ∅. Put Bi,1 = {BP : P ∈ Pi,1 },

D1 = in+1 (Di,1 ∪ Bi,1 ), then it is easy to see that D1 is a locally finite subset
of X. Put G1 = φ(D1 ).
Assume that we have already defined an open subset Gj and a locally finite subset
Dj of X for some j ∈ N. For each i  n + 1, put

Pi,j+1 = {P − Gj : P ∈ Pi,j+1 },

Ai,j+1 = {x ∈ X : Pi,j+1 is not locally finite at x}.

We can see that Ai,j+1 is a closed subset of X, and Ai,j+1 ∩ Xi = ∅. It follows by

the inductive assumption that Ai,j+1 is a D-space since strong Σ-spaces are closed
hereditary. So there is a locally finite subset Di,j+1 of Ai,j+1 , hence, a locally finite
subset of X, such that Ai,j+1 ⊂  φ(Di,j+1 ). For each P ∈ Pi,j+1 , if there exists some
BP ∈ BP such that BP ⊂ P − (φ(Di,j+1 ) ∪ Gj ) ⊂ φ(BP ), choose just one set BP
for P and denote it by BP
, else we put BP
= ∅. Let

Bi,j+1 = {BP
: P ∈ Pi,j+1 },
 
Dj+1 = (Di,j+1 ∪ Bi,j+1 ) ∪ Dj .
in+1
174 3 Generalized Metric Spaces

to see that Dj+1 is a locally finite subset of X. Put Gj+1 = φ(Dj+1 )∪Gj .
Then it is easy
Put D = j∈N  Dj .
Claim 1. X = d∈D φ(d).
Assume the contrary. Then there exist i  n + 1 and K ∈ Ki such that

K
= K − φ(D) = ∅.

Since K
is compact in Xi , we can find a finite subset F of K
such that K
⊂ φ(F).
Put K

= K − φ(F). We can find the smallest l ∈ N such that K

⊂ φ(Dl ). We take
the first Pi,j containing a P
such that

K ⊂ P
⊂ φ(F) ∪ (φ(Dl ) ∩ Xi ) ⊂ φ(F) ∪ φ(Dl ).

Let m = max{l, j}. Then P

∈ Pi,m , so F ∈ BP
, and we can get a nonempty subset
BP
of Xi in BP
with respect to P
. Hence P
must be covered by Gm , a contradiction.
Claim 2. D is closed and discrete in X.
For each x ∈ X, x ∈ φ(Dl ) for some l ∈ N by Claim 1. It means that x is separated
from D − Dl by the construction of Dl . Meanwhile, Dk is locally finite in X for each
k  l, hence, D is locally finite in X. 

Under ♦, there is a hereditarily Lindelöf T2 space X which is not a D-space but

is the union of two D-spaces [435, 436].
In the third part of this section, we introduce the mapping theorems of Σ-spaces.
One can verify the following proposition directly.

Proposition 3.2.27 Let f : X → Y be a mapping.

(1) If P is a (mod k)-network for X w.r.t. K , then f (P) is also a (mod k)-network
for Y w.r.t. f (K ).
(2) If f is a closed mapping and P is a quasi-(mod k)-network for X w.r.t. H , then
f (P) is also a quasi-(mod k)-network for Y w.r.t. f (H ).
Since every locally finite family in a countably compact space is finite, every
locally finite family is invariant under quasi-perfect mappings.

Corollary 3.2.28 Let f : X → Y be a mapping.

(1) If f is a quasi-perfect mapping and X is a Σ-space, then Y is a Σ-space.
(2) If f is a closed mapping and X is a Σ ∗ -space (resp. strong Σ ∗ -space, Σ -space
or strong Σ -space), then Y is a Σ ∗ -space (resp. strong Σ ∗ -space, Σ -space or
strong Σ -space).

Proposition 3.2.29 ([74]) Let f : X → Y be a closed L-mapping. If X is a strong

Σ-space, then Y is a strong Σ-space.

Proof Since X is a strong Σ-space, by Theorem 3.2.11, X is a subparacompact space,

 hence Y is also a subparacompact space (see Corollary A.3.4 in Appendix A). Let
and
n∈N Pn be a closed (mod k)-network for X w.r.t. K such that each Pn is discrete.
3.2 Σ-Spaces 175

For each n ∈ N, there is an open cover Un of X such that every element of Un only
meets at most one element of Pn . For each y ∈ Y , there is a countable subfamily Un,y
of Un such that f −1 (y) ⊂ ∪Un,y . Then there is an open neighborhood Vn,y of y such
that f −1 (Vn,y ) ⊂ ∪Un,y . Let Vn = {Vn,y : y ∈ Y }. Then Vn is an open cover of Y , and
hence there is a σ -discrete closed refinement Fn of Vn . Let Fn = {Fα : α ∈ Λn }.
Then for each α ∈ Λn , there is yα ∈ Y such that Fα ⊂ Vn,yα . Define

Bn = {f −1 (Fα ) ∩ U : α ∈ Λn , U ∈ Un,yα }.

Then Bn is a cover of X. Let Un,yα = {Uα,j }j∈N . Then Bn = j∈N Bn,j , where
Bn,j = {f −1 (Fα ) ∩ Uα,j : α ∈ Λn }. Let Cn = (Pn ∧ Bn )− . Since Pn is a discrete
family of closed sets in X, each element ofPn is the union of some subfamilies of
Cn . Let Pn = {Pβ : β ∈ Γn }. Then Cn = j∈N Cn,j , where

Cn,j = {Pβ ∩ f −1 (Fα ) ∩ Uα,j : α ∈ Λn , β ∈ Γn }.

Then
f (Cn,j ) = {f (Pβ ∩ Uα,j ) ∩ Fα : α ∈ Λn , β ∈ Γn }.

Since every Uα,j only meets at most one element of Pn and Fn is a σ -discrete
family of closed sets in X, f (Cn,j ) is also a σ -discrete family of closed sets in Y . We
prove n,j∈N f (Cn,j ) is an almost (mod k)-network for Y . For each y ∈ Y , there is
x ∈ K ∈ K such that y = f (x). For every open set W ⊃ f (K) in Y , there is β ∈ Γn
such that K ⊂ Pβ ⊂ f −1 (W ). Since Pβ is the union of some sets in Cn , there exist
α ∈ Λn and j ∈ N such that x ∈ Pβ ∩ f −1 (Fα ) ∩ Uα,j ⊂ Pβ ⊂ f −1 (W ), and hence

y ∈ f (Pβ ∩ Uα,j ) ∩ Fα ⊂ f (Pβ ) ⊂ W . By Lemma 1.5.13, n,j∈N f (Cn,j ) is an almost
(mod k)-network for Y . Thus Y is a strong Σ-space. 

Theorem 3.2.30 ([462]) Suppose f : X → Y is a closed mapping. If X is a Σ-space,

then there is an Fσ -discrete subspace Z of Y such that for each y ∈ Y − Z, f −1 (y) is
an ℵ1 -compact subspace of X.

Proof Let n∈N Pn be a closed quasi-(mod k)-network in X w.r.t. H , where Pn is a
locally finite family which is closed under finite intersections and X ∈ Pn ⊂ Pn+1 .
For each n ∈ N, define

Zn = ∪{f (P) ∩ f (Q) : P, Q ∈ Pn and |f (P) ∩ f (Q)| < ℵ0 }.


Let Z = n∈N Zn . By Lemma 3.2.20, Z is an Fσ -discrete subspace of Y .
Let y ∈ Y − Z. We prove that for each n ∈ N, f −1 (y) only meets finitely many
elements of Pn . In fact, fix x ∈ f −1 (y) first. Suppose there is {Pi }i∈N ⊂ Pn such
that Pi ∩ f −1 (y) = ∅. For each m  n, let Qm = C(Pm , x). Then Qm ∈ Pm . Since
y ∈ f (Pm ) ∩ f (Qm ) − Z, f (Pm ) ∩ f (Qm ) is an infinite set. So there is a nontrivial
sequence {ym }mn in Y such that ym ∈ f (Pm ) ∩ f (Qm ). Pick pm ∈ Pm ∩ f −1 (ym ) and
qm ∈ Qm ∩ f −1 (ym ). Then {pm } has no accumulation point in X, and hence {ym } has
176 3 Generalized Metric Spaces

no accumulation point in Y either, it follows that {qm } has no accumulation point in

X, which contradicts Proposition 3.2.2.
Thus, there is a countable closed quasi-(mod k)-network for f −1 (y). Further by
Theorem 3.2.15, f −1 (y) is an ℵ1 -compact subspace of X. 

By Corollary 3.2.16, we have the following corollary.

Corollary 3.2.31 ([462]) Suppose f : X → Y is a closed mapping. If X is a strong

Σ-space, then there is an Fσ -discrete subspace Z of Y such that for each y ∈ Y − Z,
f −1 (y) is a Lindelöf subspace of X.

Theorem 3.2.30 and Corollary 3.2.31 decompose a closed image of a space with
some generalized metric properties as the union of an Fσ -discrete subspace and a sub-
space in which the fiber of each point has better properties. The research on theorems
like that was started by Lašnev [239], and was developed further by Arhangel’skiı̌
[31], Chaber [97], Yajima, Tanaka [462] and so on later. These theorems are called
the decomposition theorems for closed mappings or decomposition theorems for the
range.

Definition 3.2.32 Let Φ and  be two topological properties. The property Φ is

said to satisfy the decomposition theorem of type  if, for every space X with the
property Φ and any closed mapping f : X → Y , there is an Fσ -discrete subspace Z
of Y such that for each y ∈ Y − Z, f −1 (y) is a subspace of X with the property .
By the above definition, Corollary 3.2.31 can be briefly described as that the strong
Σ-property satisfies the decomposition theorem of type Lindelöfness. It is easy to
verify that for two topological properties Φ and , where property  is invariant
under a closed mapping, if Φ satisfies the decomposition theorem of type , then
each closed image of Φ also satisfies the decomposition theorem of type .

Question 3.2.33 (1) Does every strong Σ ∗ -space satisfy the decomposition theo-
rem of type Lindelöfness [399]?
(2) Does every perfect strong Σ-space satisfy the decomposition theorem of type
compactness [97]?
In Proposition 1.5.10, we proved that the strong Σ-property is countably produc-
tive. By the same argument it is easy to show that the strong Σ -property is also
countably productive [162]. As an application of mapping theorems, we show the
strong Σ ∗ -property is even not finitely productive (see Example 3.2.37(3)).

Theorem 3.2.34 ([386]) For every paracompact space X, the following are equiv-
alent:
(1) X is a Σ-space.
(2) X × I is a Σ-space.
(3) X × I is a Σ ∗ -space.
3.2 Σ-Spaces 177

Proof We only need to prove (3) ⇒ (1). Since π1 : X × I → X is a perfect  mapping,

X × I is a paracompact space, so X × I is a strong Σ ∗ -space. Let n∈N Pn be a
closed (mod k)-network for X × I w.r.t. K , where each Pn is an HCP family in
X × I and Pn ⊂ Pn+1 . First, for each n ∈ N, we construct inductively a family
{V (α1 , . . . , αn ) : αi ∈ Λi , i  n} of open sets in X and a family {I(α1 , . . . , αn ) :
αi ∈ Λi , i  n} in I such that
(34.1) {V (α1 , . . . , αn ) : αi ∈ Λi , i  n} is a locally finite open cover of X;
(34.2) V (α1 , . . . , αn+1 ) ⊂ V (α1 , . . . , αn ), I(α1 , . . . , αn+1 ) ⊂ I(α1 , . . . , αn );
(34.3) if V (α1 , . . . , αn ) = ∅, then I(α1 , . . . , αn ) is a nontrivial closed interval in I;
(34.4) V (α1 , . . . , αn ) × I(α1 , . . . , αn ) only meets finitely many elements of Pn .
Let V (∅) = X and I(∅) = I. Suppose for each n  k, the families satisfying
the above conditions have been constructed. Consider n = k + 1. For each i  k,
fix αi ∈ Λi such that V (α1 , . . . , αk ) = ∅. For every x ∈ V (α1 , . . . , αk ), by Lemma
2.5.4, there is a finite set Kx in X × I such that Pk+1 is locally finite in {x} ×
I(α1 , . . . , αk ) − Kx , and hence there is a nontrivial closed interval Ix in I(α1 , . . . , αk )
such that ({x} × Ix ) ∩ Kx = ∅. Pick an open neighborhood Ux of x in V (α1 , . . . , αk )
such that
U x × Ix ⊂ X × I − ∪{P ∈ Pk+1 : P ∩ ({x} × Ix ) = ∅}.

Then there is a locally finite open cover {Vγ : γ ∈ Γ (α1 , . . . , αk )} of V (α1 , . . . , αk )

refining {Ux : x ∈ V (α1 , . . . , αk )}. Let ϕ be a function from Γ (α1 , . . . , αk ) to
V (α1 , . . . , αk ) such that Vγ ⊂ Uϕ(γ ) whenever γ ∈ Γ (α1 , . . . , αk ). Define

Λk+1 = ∪{Γ (α1 , . . . , αk ) : αi ∈ Λi , i  k, V (α1 , . . . , αk ) = ∅};

⎧
⎨ V (α1 , . . . , αk+1 ) ∩ Vαk+1 , V (α1 , . . . , αk ) = ∅ and
V (α1 , . . . , αk+1 ) = αk+1 ∈ Γ (α1 , . . . , αk ),
⎩
∅, otherwise;

I , V (α1 , . . . , αk+1 ) = ∅,
I(α1 , . . . , αk+1 ) = ϕ(αk+1 )
∅, V (α1 , . . . , αk+1 ) = ∅.

Then the families {V (α1 , . . . , αk+1 ) : αi ∈ Λi , i  k + 1} and {I(α1 , . . . , αk+1 ) :

αi ∈ Λi , i  k + 1} satisfy all the conditions required.
Define

Z= (∪{V (α1 , . . . , αn ) × I(α1 , . . . , αn ) : αi ∈ Λi , i  n}).
n∈N


Then Z is a closed set in X × I and n∈N Pn|Z is a σ -locally finite closed (mod k)-
 K|Z . So Z is a strong Σ-space. Let f = π
network for Z w.r.t. 1|Z . For each x ∈ X,
there is (αn ) ∈ n∈N Λn such that x ∈ V (α1 , . . . , αn ). Take y ∈ n∈N I(α1 , . . . , αn ).
Then (x, y) ∈ Z and f (x, y) = x, so f : Z → X is a perfect mapping. Thus X is a
strong Σ-space. 
Question 3.2.35 (1) If X × I is a strong Σ ∗ -space, then is X a strong Σ-space?
178 3 Generalized Metric Spaces

(2) Are Σ-spaces finitely productive [158]?

In the last part of this section, by the following examples, we not only show that
some implications and topological operations among the classes of Σ-spaces and
relevant spaces do not hold, but also show some mapping properties of these spaces.
We prove the following lemma first.
Lemma 3.2.36 If every countably compact closed set in a Σ-space X is a finite set,
then X is an Fσ -discrete space.
Proof Let P be a σ -locally finite closed quasi-(mod k)-network for X w.r.t. H .
One may assume P is closed under finite intersections. Let H = {Hα }α∈Λ . For
each α ∈ Λ, denote {P ∈ P : Hα ⊂ P} = {Pα,i }i∈N and Gα,n = in Pα,i , n ∈ N.
Then there is k ∈ N such that |Gα,k | < ℵ0 . Because otherwise, there is a sequence
{xn } in X such that x1 ∈ Gα,1 and xn+1 ∈ Gα,n+1 − {xi : i  n}. Since {Gα,n }n∈N is
a decreasing network for Hα in X, Hα ∪ {xn : n ∈ N} is a countably compact closed
 That shows there is k(α) ∈ N such that |Gα,k(α) | < ℵ0 and
set in X, a contradiction.
Gα,k(α) ∈ P. So X = α∈Λ Gα,k(α) is an Fσ -discrete space. 
Example 3.2.37 Some implications and topological operations among the classes of
Σ-spaces and relevant spaces do not hold.
(1) There is a strong Σ -space which is not subparacompact, the space X in Example
1.8.10 is such a space.
(2) There is a regular Lindelöf Σ -space which is not a Σ ∗ -space, the Fortissimo
space Xp in Example 2.5.19 is such a space. It is obvious that Xp is a regular
Lindelöf space. Since {{p, x} : x ∈ Xp − {p}} is a closure-preserving cover of Xp ,
Xp is a Σ -space. By Corollary 3.2.16 and Lemma 3.2.36, Xp is not a Σ ∗ -space.
(3) There is a paracompact Σ ∗ -space which is not a Σ-space, for example the space
X in Example 2.11.9. Because by Corollary 3.2.28, X is a Σ ∗ -space; but by
Theorem 3.2.34, X × I is not a Σ ∗ -space. This example also shows that strong
Σ ∗ -spaces are not finitely productive.
(4) There is a strong Σ-space which does not have the point Gδ -property, the space
ω1 + 1 with the ordered topology is such a space.
(5) There is a regular space with a countable quasi-(mod k)-network which is not a
Σ-space [450].
Let X = [0, ω1 ] × R − {ω1 } × P. Then X is a regular space. For r ∈ R, define a
countably compact closed set lr in X as follows: if r ∈ Q, then lr = [0, ω1 ] × {r};
if r ∈ P, then lr = [0, ω1 ) × {r}. Suppose B is the set of all the open intervals with
two rational ends in R. Let

C = {lr : r ∈ R},
P = {lr : r ∈ Q} ∪ {[0, ω1 ) × B : B ∈ B}.

Then P is a countable quasi-(mod k)-network for X w.r.t. C .

Let F be a σ -locally finite closed quasi-(mod k)-network for X w.r.t. H . We
prove that for each r ∈ P, there is H ∈ H such that π1 (lr ∩ H) is an uncountable
3.2 Σ-Spaces 179

set in [0, ω1 ). In fact, since lr only meets countably many elements of F , the set
{F ∈ F : π1 (lr ∩ F) is a nonempty countable set in [0, ω1 )} is countable, so there
is α < ω1 , such that, if F ∈ F and [α, ω1 ) ∩ π1 (lr ∩ F) = ∅, then π1 (lr ∩ F) is
uncountable. Pick H ∈ H such that (α, r) ∈ H. If π1 (lr ∩ H) is countable, then
there is β < ω1 such that sup(π1 (lr ∩ H)) < β. Let U = X − {(γ , r) : γ  β}. Then
H ⊂ U ∈ τ , and there is no F ∈ F such that H ⊂ F ⊂ U, a contradiction.
Thus, for each r ∈ P, there is Hr ∈ H such that π1 (lr ∩ Hr ) is an uncountable set
1 ), and hence there is Br ∈ (B)r such that Hr ∩ ({ω1 } × clR (Br )) = ∅. Let
in [0, ω
F = n∈N Fn , where each Fn is locally finite in X. Since F is a quasi-(mod k)-
network for X w.r.t. H , there exist nr ∈ N and Fr ∈ Fnr such that

Hr ⊂ Fr ⊂ X − ({ω1 } × Br ).

By the Baire category theorem, there exist m ∈ N, B0 ∈ B and a sequence

{ri } in B0 ∩ P such that {ri } converges to a point q ∈ Q ∩ clR (B0 ), and every
nri = m, Bri = B0 . For each i ∈ N, since (ω1 , q) ∈ / Fri , there exist α < ω1
and q ∈ U ∈ τ (R) such that Fri ∩ ((α, ω1 ] × U) = ∅. Since the uncountable set
π1 (lr ∩ Hr ) ⊂ π1 (lr ∩ Fr ), Fr ∩ ((α, ω1 ] × U) = ∅ whenever r ∈ U ∩ P, so Fr = Fri .
Hence there is a subsequence {rij } of {ri } such that the Frij are different for different
j, which contradicts the local finiteness of Fm in (ω1 , q). Thus X is not a Σ-space.
Example 3.2.38 Some mapping properties of Σ-spaces and relevant spaces.
(1) Neither Σ-spaces nor strong Σ-spaces are preserved by closed mappings, Exam-
ple 2.11.9 is such a counterexample.
(2) Strong Σ-spaces do not satisfy the decomposition theorem of type compactness
[97]: there exist a σ -compact regular space X and a closed mapping f : X → I
such that no fiber of f is compact in X.
Let X = I × I. Give X the following topology: for each s ∈ I, {s} × (0, 1]
is an open subspace of X and has the Euclidean topology; each element of a
neighborhood base of (s, 0) ∈ X has the form V × I − ({s} × (0, 1]), where V
is a Euclidean neighborhood of s in I. Obviously, X is a regular space. For each
n ∈ N, I × ({0} ∪ [1/n, 1]) is a compact set in X, and hence X is a σ -compact
space. Let f = π1 : X → I. By Proposition 2.1.7, f is a closed mapping. For
any s ∈ I, f −1 (s) = {s} × I is not a compact set in X.
(3) Strong Σ -spaces do not satisfy the decomposition theorem of type Lindelöf-
ness [97]: there exist a paracompact space X with a closure-preserving cover
consisting of compact sets and a closed mapping f : X → I such that no fiber
of f is a Lindelöf subspace of X.
Let X = I × I. Give X the following topology: each point in I × (0, 1] is isolated
in X; for each s ∈ I, each element of a neighborhood base of (s, 0) ∈ X has
the form V × I − ({s} × (0, 1]), where V is a Euclidean neighborhood of s in
I. Obviously, X is a paracompact space. For each t ∈ (0, 1], let Kt = I × {0, t}.
Then Kt is a compact set in X and {Kt : t ∈ (0, 1]} is a closure-preserving cover
of X. Let f = π1 : X → I. Then f is a closed mapping. For any s ∈ I, f −1 (s) is
not a Lindelöf subspace of X.
180 3 Generalized Metric Spaces

Since X is a first countable strong Σ -space, the assumption “Σ ∗ -space” in Propo-

sition 3.2.18 and Theorem 3.2.23 cannot be weaken to “Σ -space”.

3.3 σ -Spaces and Semi-Stratifiable Spaces

Both σ -spaces and semi-stratifiable spaces are regarded as the most successful gen-
eralizations of metric spaces, which is mainly reflected in the beautiful nature of
these spaces. Every semi-stratifiable space can be characterized as a space with a
σ -cushioned network. In this section, we shall proved that every σ -space is equiv-
alent to a regular space with a σ -discrete or σ -closure-preserving network. These
elegant results constitute the essence of general topology. In this section, we also
investigate the characterizations and mapping properties of these spaces, and intro-
duce the (G)-covering property which is closely relevant to σ -spaces.
We first introduce several important characterizations of σ -spaces.

Theorem 3.3.1 For every space X, the following are equivalent:

(1) X has a σ -discrete closed network.
(2) X has a σ -locally finite closed network.
(3) X has a σ -closure-preserving closed network.
(4) X has a σ -function, i.e. there is a g-function on X such that
(i) if x, y ∈ X, n ∈ N and x ∈ g(n, y), then g(n, x) ⊂ g(n, y);
(ii) if x ∈ X, {xn } ⊂ X and x ∈ g(n, xn ) for each n ∈ N, then xn → x.
If further assume X is a regular space, then each item of (1) – (4) is equivalent
to the following (5) or (6):
(5) There is a g-function on X such that if x ∈ g(n, xn ) and xn ∈ g(n, yn ), then
yn → x.
(6) X is a σ -space.

Proof Obviously, (1) ⇒ (2) ⇒ (3), and (2) ⇔ (6) when X is a regular space. By the
 and 1.5.4 we get (3) ⇔ (4).
proofs of Propositions 1.5.16
(3) ⇒ (1). Let P = n∈N Pn be a closed network for X, where each Pn =
{Pα : α ∈ Γn } is closure-preserving. For every n, m ∈ N and α ∈ Γn , define
Fα,m = ∪{P ∈ Pm : P ∩ Pα = ∅}. Then Pα and Fα,m are disjoint closed sets. Let
Hn,m = {{Pα , Fα,m } : α ∈ Γn }. Then each element of Hn,m has the following
form:

HΓ
= (∩{Pα : α ∈ Γ
}) ∩ (∩{Fα,m : α ∈ Γn − Γ
}), where Γ
⊂ Γn .

(1.1) Hn,m is a discrete family of closed sets in X.

It is easy to verify that Hn,m is a family of disjoint closed sets. Below we
prove it is closure-preserving. Suppose {Γλ
: λ ∈ Λ} is a subfamily of Γn
and x ∈ / ∪{HΓλ
: λ ∈ Λ}. If λ ∈ Λ, then either x ∈ / ∩{Pα : α ∈ Γλ
}, or
3.3 σ -Spaces and Semi-Stratifiable Spaces 181

x∈/ ∩{Fα,m : α ∈ Γn − Γλ
}, so either there is αλ ∈ Γλ
such that x ∈
/ Pαλ , or
there is αλ ∈ Γn − Γλ
such that x ∈
/ Fαλ ,m . Define

Λ
= {λ ∈ Λ : x ∈
/ ∩{Pα : α ∈ Γλ
}},
Λ

= {λ ∈ Λ : x ∈/ ∩{Fα,m : α ∈ Γn − Γλ
}},
F
= ∪{Pαλ : λ ∈ Λ
}, F

= ∪{Fαλ ,m : λ ∈ Λ

}.

Then Λ = Λ
∪ Λ

and F
, F

∈ τ c . Let V = X − (F
∪ F

). Then the open

neighborhood V of x is not disjoint with ∪{HΓλ
: λ ∈ Λ}. So ∪{HΓλ
: λ ∈ Λ}
 set. Thus Hn,m is closure-preserving, and hence it is discrete.
is a closed
(1.2) H = n,m∈N Hn,m is a network for X.
For every x ∈ U ∈ τ , there exist n ∈ N and β ∈ Γn such that x ∈ Pβ ⊂ U.
Define Γ
= {α ∈ Γn : x ∈ Pα }. Then x ∈ ∩{Pα : α ∈ Γ
} ⊂ Pβ ⊂ U.
Since x ∈ / ∪{Pα : α ∈ Γn − Γ
}, there exist m ∈ N and δ ∈ Γm such that
x ∈ Pδ ⊂ X − ∪{Pα : α ∈ Γn − Γ
}. So Pδ ∩ (∪{Pα : α ∈ Γn − Γ
}) = ∅,
and hence Pδ ⊂ Fα,m for each α ∈ Γn − Γ
. It follows that x ∈ Pδ ⊂ ∩{Fα,m :
α ∈ Γn − Γ
}, so x ∈ HΓ
⊂ Pβ ⊂ U. Thus H is a network for X.
(4) ⇒ (5). Suppose g is a σ -function on X. If x ∈ g(n, xn ) and xn ∈ g(n, yn )
for a point x and sequences {xn }, {yn } in X, then x ∈ g(n, yn ), so yn → x.
(5) ⇒ (1). Suppose there is a g-function on the regular space X satisfying (5).
Let < be a well-ordering on X. Define a function H : N × N × X → τ c by
⎛  ⎛ ⎞⎞
 
H(i, n, x) = X − ⎝ g(i, y) ∪ ⎝ g(n, y)⎠⎠ .
y<x y∈g(i,x)
/

Then H(i, n, x) ⊂ g(i, x). For every i, n ∈ N, let

H (i, n) = {H(i, n, x) : x ∈ X}.

Then H (i, n) is a discrete family in X. Because in fact, let z ∈ X, and y =

min{h ∈ X : z ∈ g(i, h)}. Then g(i, y) ∩ H(i, n, x) = ∅ if y < x, and
g(n, z) ∩ H(i, n, x) = ∅ if x < y. Thus g(i, y) ∩ g(n, z) is an open set in X
containing z which meets at most one element of H (i, n). For every i, n, m ∈
N, Define
F (i, n, m) = {F(i, n, m, x) : x ∈ X},

where F(i, n, m, x) = {y ∈ H(i, n, x) : x ∈ g(m, y)}. Then F (i, n, m) is a

discrete family in X. For every p ∈ U ∈ τ and i ∈ N, let

xi = min{h ∈ X : p ∈ g(i, h)}.

Then there is ni ∈ N such that p ∈ / ∪{g(ni , y) : y ∈ / g(i, xi )}, so p ∈

H(i, ni , xi ) ⊂ g(i, xi ), and hence xi → p. Thus for each m ∈ N, there is im  m
182 3 Generalized Metric Spaces

such that xim ∈ g(m, p). Let Fm = F(im , nim , m, xim ). Then p ∈ Fm . If there is
a sequence {ym } in X such that ym ∈ Fm − U, then p ∈ g(im , xim ) ⊂ g(m, xim )
and xim ∈ g(m, ym ), and hence ym → p, a contradiction. Consequently, there
is m ∈ N such that Fm ⊂ U. Thus i,n,m∈N F (i, n, m) is a σ -discrete network
for X. 
(1) ⇔ (2) ⇔ (3) of Theorem 3.3.1 is called the Nagata–Siwiec theorem [426],
and the characterizations of σ -spaces in terms of g-functions are the masterpieces of
Heath and Hodel [184].
Corollary 3.3.2 For every space X, the following are equivalent:
(1) X has a σ -locally finite closed network.
(2) X is a strong Σ-space with a point-countable pseudo-k-network [167].
(3) X is a Σ -space with a Gδ -diagonal [217].
(4) X is both a σ -space and a Σ -space [422].
Proof (1) ⇒ (2), (3) is obvious. (3) ⇒ (4) follows from Theorems 1.4.10, 3.2.7 and
Proposition 1.5.20.
(2) ⇒ (4). Let P be a σ -locally finite closed (mod k)-network for X which is
closed under finite intersections. Since P is an almost (mod k)-network for X, by
Lemma 2.11.6, there exist a metrizable space M, a σ -discrete base B for M and a
subspace Z of X × M such that
(2.1) P = f (g−1 (B));
(2.2) g : Z → M is a perfect mapping,
where f = π1|Z , g = π2|Z . By (2.2), Z is a paracompact M-space with a point-
countable pseudo-k-network. By Theorem 3.1.8, Z is a metrizable space. Since
B is a base for M, g−1 (B) is a (mod k)-network for Z. Further by (2.1) and
Corollary 2.11.4, f is a σ -locally finite mapping. By Corollary 2.11.5, X has a
σ -locally finite network.
Suppose Q is a σ -locally finite network for X. Then Q is a σ -locally finite
family of closed sets for X. For every x, y ∈ X with x = y, take two disjoint
open sets U and V such that x ∈ U and y ∈ V . Then there is Q ∈ Q such that
x ∈ Q ⊂ U. Since U ∩ V = ∅, x ∈ Q ⊂ X − {y}. So Q is a pseudo-network
for X, and hence X is a σ -space.
(4) ⇒ (1). Let g and h be a σ -function and a Σ -function on X respectively.
Define q : N × X → τ by

q(n, x) = g(n, x) ∩ h(n, x).

Then q is a g-function on X and for any x, y ∈ X and n ∈ N, if x ∈ q(n, y), then

q(n, x) ⊂ q(n, y). Suppose x ∈ q(n, xn ) for a point x and a sequence {xn } in X.
Since h is a Σ -function on X, any subsequence of {xn } has an accumulation
point. Let p be an accumulation point of {xn }. If p = x, then there is n  k
such that x ∈/ g(k, p) and xn ∈ g(k, p), so x ∈ g(n, xn ) ⊂ g(k, xn ) ⊂ g(k, p),
a contradiction. Thus x is the only accumulation point of all subsequences
3.3 σ -Spaces and Semi-Stratifiable Spaces 183

of {xn }, it follows xn → x. Thus q is a σ -function on X, and hence X has a

σ -locally finite closed network. 

Definition 3.3.3 A space X is said to satisfy (G) [106] if, there is a family W =
{Wx : x ∈ X} in X with the following properties (1) and (2):
(1) for each x ∈ X, x ∈ ∩Wx and |Wx |  ℵ0 ;
(2) for every x ∈ U ∈ τ , there is an open set V (x, U) containing x, such that, there
is W ∈ Wy with x ∈ W ⊂ U whenever y ∈ V (x, U).
The family W is called a point-network for X [51]. If further assume W has the
following property (3), then X is said to satisfy uniform (G) [354]:
(3) for each x ∈ X, any infinite subfamily of Wx is a net at x.
When referring to the spaces satisfying (G) in the following discussion, V (x, U)
is always the corresponding set relative to x and U in the property (2) above.

Lemma 3.3.4 ([355]) (1) Every stratifiable space satisfies uniform (G).
(2) Every space satisfying (G) is a hereditarily meta-Lindelöf space.
(3) Every space satisfying uniform (G) is a hereditarily metacompact space.

Proof (1) Suppose (X, τ ) is a stratifiable space. By Proposition 1.3.13 there exists
a decreasing g-function on X satisfying the condition: if y ∈ X − H ∈ τ , then
there is m ∈ N such that y ∈ / g(m, H). It follows from Theorem 1.4.14 that X is a
monotonically normal space. Let D be a monotone normality operator for X such that
D(F, K) ∩ D(K, F) = ∅ for disjoint closed subsets F, K of X. For every x ∈ U ∈ τ ,
put O(x, U) = D({x}, X − U). Then the function O satisfies as follows: for any
xi ∈ Ui ∈ τ with i = 1, 2, O(x1 , U1 ) ∩ O(x2 , U2 ) = ∅ implies x1 ∈ U2 or x2 ∈ U1 .
For every x ∈ X and n ∈ N, let V (n, x) = O(x, g(n, x)), and W (n, x) = {y ∈
X : x ∈ V (n, y)}. Let Wx = {W (n, x) : n ∈ N} for each x ∈ X. We will show that
W = {Wx : x ∈ X} satisfies uniform (G).
Take any x ∈ U ∈ τ . There exists n ∈ N such that x ∈ / g(m, X − U) for each
m > n. Let U1 = U − g(m, X − U). And we define V = O(x, U1 ∩ V (m, x)). Then
V ⊂ V (m, x), so x ∈ W (m, y) for any y ∈ V .
Finally, we need to show that W (m, y) ⊂ U. Take z ∈ W (m, y). Then y ∈
V (m, z). So V ∩ V (m, z) = ∅, hence either z ∈ U1 ∩ V (m, x) or x ∈ g(m, z). If
z ∈ U1 ∩ V (m, x), trivially z ∈ U. If x ∈ g(m, z), then z ∈
/ X − U, hence z ∈ U. This
finishes the proof of (1).
Since (G) (resp. uniform (G)) property is hereditary, we only need to prove that
every space satisfying (G) (resp. uniform (G)) is a meta-Lindelöf space (resp. meta-
compact space).
(2) Let W = {Wx : x ∈ X} be a family in X satisfying (G). If U is an open cover
of X, then defineU = {Uα }α<κ , where κ is a cardinal number. For each α < κ,
let Pα = Uα − β<α Uβ and Oα = ∪{V (x, Uα ) : x ∈ Pα }. Then Oα ∈ τ and
Pα ⊂ Oα ⊂ Uα . Let O = {Oα }α<κ . Then O is an open refinement of U . O is point-
countable. Otherwise, there exist z ∈ X and a sequence {αγ }γ <ω1 of strictly increasing
cardinal numbers in [0, κ) such that z ∈ Oαγ − Pα0 . Then there is xγ ∈ Pαγ such
that z ∈ V (xγ , Uαγ ), and hence there is Wγ ∈ Wz such that xγ ∈ Wγ ⊂ Uαγ . Since
184 3 Generalized Metric Spaces

/ Uαγ whenever γ < γ

< ω1 , xγ
∈
xγ
∈ / Wγ , which contradicts the countability of
Wz . Thus X is a meta-Lindelöf space.
(3) Suppose W = {Wx : x ∈ X} satisfies uniform (G). Then the family O above is
point-finite. Otherwise, by the same argument as that in the proof of (1), there exist
a point z, a sequence {xn } in X and a subfamily {Wn }n∈N of Wz such that xn ∈ Wn
and xn ∈ / Wm whenever n > m. Thus {Wn }n∈N is a net at z, so there is n ∈ N with
Wn ⊂ W1 − {x1 }, and hence n > 1 and xn ∈ W1 , a contradiction. Consequently, X is
a metacompact space. 

Suppose afamily W = {Wx : x ∈ X} satisfies (G) in a space X. For each A ⊂ X,

let W (A) = x∈A Wx . Obviously, W (A) satisfies (1)–(4) in Definition 3.1.10, so X is
a monotonically monolithic space. By Theorem 3.1.11, X is a D-space. Thus, a space
satisfying (G) is a hereditarily D-space [434]. (G) and monotonically monolithic are
distinct properties [467].

Lemma 3.3.5 ([143]) Every semi-stratifiable space satisfying (G) has a σ -discrete
network.

Proof Let W = {Wx : x ∈ X} be a family satisfying (G) in a semi-stratifiable

space X, where Wx = {W (n, x)}n∈N . Suppose g is a semi-stratifiable function on X.
Let < be a well-ordering on X. For every i, n, k ∈ N, define
⎛  ⎛ ⎞⎞
 
H(i, n, x) = X − ⎝ g(i, y) ∪ ⎝ g(n, y)⎠⎠ , x ∈ X;
y<x y∈g(i,x)
/

F(i, n, k, x) = H(i, n, x) ∩ W (k, x), x ∈ X;

F (i, n, k) = {F(i, n, k, x) : x ∈ X}.

Then H(i, n, x) ⊂ g(i, x), and by the proof of Theorem 3.3.1, F (i, n, k) is a discrete
family in X. For every p ∈ U ∈ τ and i ∈ N, let

xi = min{h ∈ X : p ∈ g(i, h)}.

Then there is ni ∈ N such that p ∈ / ∪{g(ni , y) : y ∈ / g(i, xi )}, so p ∈ H(i, ni , xi ) ⊂

g(i, xi ), and hence xi → p. It follows that there exist i0 , k ∈ N such that xi0 ∈ V (p, U)
and p ∈ W (k, xi0 ) ⊂ U. Thus p ∈ F(i0 , ni0 , k, xi0 ) ⊂ U. Therefore, X has a σ -discrete
network. 

Lemma 3.3.6 ([354]) A space satisfying both uniform (G) and the point Gδ -property
is a semi-stratifiable space.

Proof Suppose a space X satisfies the point Gδ -property and W = {Wx : x ∈ X} is

a family satisfying uniform (G).
We construct a sequence {Hn } of point-finite covers of X such that {Hn } is a
Gδ -diagonal sequence in X.
3.3 σ -Spaces and Semi-Stratifiable Spaces 185

For each x ∈ X, let {G(n, x)} be a decreasing sequence of open sets in X with the
intersection {x}. For each n ∈ N, define a point-finite open cover Hn of X, a function
hn : Hn → X and an open neighborhood O(n, x) of x inductively as follows:
First, let H0 = {X}. Fix z ∈ X and define h0 : H0 → X by h0 (X) = z. Let

G(1, x), x = z,
O(1, x) =
G(1, x) − {z}, x = z.

Suppose for each m  n and x ∈ X, Hm−1 , hm−1 and O(m, x) have been defined.
Let < be a well-ordering on Hn−1 . For each H ∈ Hn−1 , by Lemma 3.3.4, the open
cover {H ∩ V (x, O(n, x))}x∈H of H has a point-finite open refinement Fn (H). Define

Hn (H) = Fn (H) − ∪{Fn (H

) : H
< H}, H ∈ Hn−1 ;
Hn = ∪{Hn (H) : H ∈ Hn−1 }.

Then Hn is a point-finite open cover of X. For each H ∈ Hn , there is a unique

set H
∈ Hn−1 such that H ∈ Hn (H
) ⊂ Fn (H
). Pick xH ∈ H
such that H ⊂
H
∩ V (xH , O(n, xH )). Let hn (H) = xH . Define

O(n + 1, x) = G(n + 1, x) − {hm (H) : m  n, H ∈ (Hm )x and x = hm (H)}.

Then x ∈ O(n + 1, x) ∈ τ .
Now we prove {Hn } is a Gδ -diagonal sequence in X. Otherwise, there are x, y ∈ X
with x = y and a sequence {Hn } of sets in X such that for each n ∈ N, x, y ∈ Hn ∈ Hn .
For each n ∈ N, there is a unique family {Hnm }mn such that

Hnn = Hn , Hnm ∈ Hm and Hnm ∈ Hm (Hnm−1 ), ∀m > 1.

Then x ∈ Hn ⊂ Hnm ⊂ Hnm−1 . By the point-finiteness of Hn , for each n ∈ N, we can

define an infinite subset In of N and in ∈ N, such that,

(i) in = min In ;
(ii) In+1 ⊂ In − {in };
(iii) m, k ∈ In ⇒ Hmn = Hkn .

Let Kn = Hinn , qn = hn (Kn ). Then Kn = Hmn (∀m ∈ In ), and qn ∈ Km ⊃ Kn

whenever n > m. Since x ∈ Kn ⊂ V (qn , O(n, qn )), by (G), there is Wn ∈ Wx such
that qn ∈ Wn ⊂ O(n, qn ).
Take disjoint open sets Ux and Uy in X containing x and y respectively. We
may assume that there is an infinite subset J of N such that qn ∈ / Ux for each
n ∈ J. Then Wn ⊂ Ux . By uniform (G), {Wn : n ∈ J} is a finite set. We may
further assume Wn = Wm for every n, m ∈ J. Then qm ∈ O(in , qn ), and hence
by the definition of O(in , qn ), qn = qm . Denote allthe qn by q. For each n ∈ J,
x ∈ V (q, O(n, q)) ⊂ O(n, q) ⊂ G(n, q). So x ∈ n∈J G(n, q) = {q}, and hence
q ∈ Ux , a contradiction.
186 3 Generalized Metric Spaces

Finally, we prove X is a semi-stratifiable space. Let {Un } be a Gδ -diagonal

sequence in X such that Un+1 refines Un . For every n ∈ N and x ∈ X, take U(n, x) ∈
(Un )x and define g(n, x) = V (x, U(n, x)). If there exist a point x and a sequence {xn }
in X such that x ∈ g(n, xn ) and xn  x, then there is an open neighborhood U of x
such that {xn } is not eventually in U. Since x ∈ V (xn , U(n, xn )) ⊂ U(n, xn ), there is
Wn ∈ Wx such that xn ∈ Wn ⊂ U(n, xn ) ⊂ st(x, Un ). Let I = {n ∈ N : xn ∈ / U}.
Then I is an infinite set, so {Wn : n ∈ I} is a finite set.
 We may assume W n = Wm
whenever n, m ∈ I. Then xn ∈ st(x, Um ), so xn ∈ m∈I st(x, Um ) = {x} ⊂ U, a
contradiction. Then X is a semi-stratifiable space. 
Theorem 3.3.7 ([143, 354]) Every regular space satisfying both uniform (G) and
the point Gδ -property is a metacompact σ -space.
Question 3.3.8 ([354]) Does every metacompact σ -space satisfy uniform (G)?
Below we discuss the mapping properties of σ -spaces.
Proposition 3.3.9 σ -spaces are preserved by quasi-perfect mappings.

Proof Suppose f : X → Y is a quasi-perfect mapping and n∈N Fn is a closed
pseudo-network in X, where each Fn is a closure-preserving family which is closed
under finite intersections and Fn ⊂ Fn+1 . Define a function g : N × X → τ by

g(n, x) = X − ∪{F ∈ Fn : x ∈
/ F}.

Then g is a σ -function on X. We prove that for each y ∈ Y ,

f −1 (y) = ∩{g(n, f −1 (y)) : n ∈ N}.

Otherwise, there exist a point x and a sequence {xn } in X such that xn ∈ f −1 (y) and
x ∈ g(n, xn ) − f −1 (y). Let a be an accumulation point of {xn } in  f −1 (y). Then there
 a subsequence {xnk } of {xn } such that xnk ∈ g(k, a), so x ∈ k∈N g(nk , xnk ) ⊂
is
k∈N g(k, a) = {a}, a contradiction.
For each n ∈ N, define Pn = f (Fn ). Then Pn is a closure-preserving family of
closed sets in Y . For every y, z ∈ Y with y = z, pick xz ∈ f −1 (z). Then there is m ∈ N
with xz ∈/ g(m, f −1 (y)), so there is F ∈ Fm such that xz ∈ F and F ∩ f −1 (y) = ∅,
and hence z ∈ f (F) ⊂ Y − {y}. Thus n∈N Pn is a σ -closure-preserving closed
pseudo-network for Y . 

By Theorem 3.3.1 and Corollary 3.3.2, we have the following results on images
and preimages of σ -spaces respectively.
Proposition 3.3.10 ([426]) Spaces with a σ -locally finite closed network are pre-
served by closed mappings.
Proposition 3.3.11 ([43]) Suppose f : X → Y is a perfect mapping. If Y has a
σ -locally finite closed network and X has a Gδ -diagonal, then X has a σ -locally
finite closed network.
3.3 σ -Spaces and Semi-Stratifiable Spaces 187

Theorem 3.3.12 ([97]) Every σ -space satisfies the decomposition theorem of type
compactness.

Proof Suppose n∈N Pn is a closed network in a σ -space X, such that, each Pn is a
locally finite family which is closed under finite intersections and X ∈ Pn ⊂ Pn+1 .
Let f : X → Y be a closed mapping. For every n ∈ N and y ∈ Y , define

En = f (Pn ), C(n, y) = ∩(En )y ,

and let
En = {y ∈ Y : |C(n, y)| < ℵ0 }, Z = ∪{En : n ∈ N}.

By Lemma 3.2.20, Z is an Fσ -discrete subspace of Y . For each y ∈ Y − Z, C(n, y)

is an infinite set, so there is a nontrivial sequence {yn } in Y such that yn ∈ C(n, y).
Thus for each n  m, P ∩ f −1 (yn ) = ∅ whenever P ∈ (Pm )f −1 (y) . Since

∪{E ∩ E
: there is n ∈ N such that E, E
∈ En and |E ∩ E
| < ℵ0 } ⊂ Z,

by the proof of Theorem 3.2.30, f −1 (y) is a Lindelöf subspace of X.

Let A = f −1 (y). If A is not a compact set in X, then A is not a countably compact
set in X, and hence there is a countable family V of open sets in X such that V covers
A, but any finite subfamily of V can not cover A. For each x ∈ A, there exist Vx ∈ V
and Wx ∈ τ (X) such that x ∈ Wx ⊂ W x ⊂ Vx . Then the cover {Wx }x∈A of A has
a countable subcover {Wi }i∈N and any finite subfamily of {W i }i∈N can not cover A.
Take x1 ∈ A ∩ W1 and  let U1 = W1 . There exist x2 ∈ A − U 1 and n1 ∈ N such that
x2 ∈ Wn1 . Let U2 = in1 Wi . Then U1 ⊂ U2 and x2 ∈ A ∩ (U2 − U 1 ). Continue in
this way, we can construct an increasing sequence {Uk } of open sets in X covering A
and A ∩ (Uk+1 − U k ) = ∅.
Pick xk ∈ A ∩ (Uk+1 − U k ) for each k ∈ N. Then there is Fk ∈ Pnk such that
xk ∈ Fk ⊂ X − U k and nk+1 > nk , and hence there is zk ∈ Fk ∩ f −1 (ynk ). Since
yn → y, {zk } has an accumulation point. Let z be this accumulation point. Then z ∈ A
and there is m ∈ N such that z ∈ Um , so there is k  m such that zk ∈ Um ∩ Fk = ∅,
a contradiction. It follows that f −1 (y) is compact in X, thus every σ -space satisfies
the decomposition theorem of type compactness. 

Question 3.3.13 ([97]) Does every perfect preimage of a σ -space satisfy the decom-
position theorem of type compactness?

Theorem 3.3.14 ([445]) Every quasi-perfect preimage of a countably paracom-

pact σ -space satisfies the decomposition theorem of type countable compactness.

Proof Suppose g : X → Z is a quasi-perfect mapping, where Z isa countably

paracompact σ -space. Let f : X → Y be a closed mapping. Assume n∈N Fn is a
network for Z, such that, each Fn is a locally finite closed cover of Z which is closed
under finite intersections and Fn ⊂ Fn+1 . For every n, i ∈ N, define
188 3 Generalized Metric Spaces

Bn = g−1 (Fn ), Pn = f (Bn ),

Yn,i = ∪{Pn,1 ∩ · · · ∩ Pn,i : for each k  i, Pn,k ∈ Pn and |Pn,1 ∩ · · · ∩ Pn,i | < ℵ0 }.

By Lemma 3.2.21, each Yn,i is a discrete closed set in Y . Let Y0 = Y − n,i∈N Yn,i .
Suppose y0 ∈ Y0 . If f −1 (y0 ) is not countably compact, then since g is quasi-
perfect, there is a sequence {xi } in f −1 (y0 ), such that, g(xi ) = g(xj ) whenever i = j
and {g(xi ) : i ∈ N} is a discrete closed set in Z. By the countable paracompactness
of Z, there is a locally finite family {Vi }i∈N of open sets in Z such that g(xi ) ∈ Vi (see
Theorem A.1.11 in Appendix A). For every i, n ∈ N, there is Fn,i ∈ Fn such that
g(xi ) ∈ Fn,i and Fn,i = ∩(Fn )g(xi ) . Then there is ji ∈ N such that Fn,i ⊂ Vi whenever
n  ji . We may assume ji  < ji+1 . Let Bn,i = g−1 (Fn,i ). Then xi ∈ Bn,i ∈ Bn ,
so y0 ∈ f (Bji ,k ), and hence ki f (Bji ,k ) is an  infinite set. It follows that there is a
nontrivial sequence {yi } in Y such that yi ∈ ki f (Bji ,k ) − {y0 }, as a consequence
yi ∈ f (g−1 (Vi )). Thus {yi : i ∈ N} is a discrete closed set in Y . On the other hand, yi ∈
f (Bji ,1 ), so there is xi
∈ f −1 (yi )∩Bji ,1 . By the structure of {Fn,i }, {Bji ,1 } is a network of
the countably compact set g−1 (g(x1 )) in X, so {xi
} has an accumulation point x0 , and
hence {f (xi
)} has an accumulation point f (x0 ), i.e. {yi } has an accumulation point, a
contradiction. Consequently, X satisfies the decomposition theorem of type countable
compactness. 

In the third part of this section, we investigate mapping properties of semi-

stratifiable spaces and relevant spaces.
Proposition 3.3.15 Let f : X → Y be a mapping.
(1) If f is a closed mapping and X is a β-space, then Y is a β-space [464].
(2) If f is a quasi-perfect mapping and Y is a β-space, then X is a β-space [43].
Proof (1) follows from Proposition 1.7.2. For (2), suppose g is a β-function on Y .
Define a g-function h : N × X → τ (X) by h(n, x) = f −1 (g(n, f (x))). If x is a
point and {xn } is a sequence in X such that x ∈ h(n, xn ), then f (x) ∈ g(n, f (xn )), so
{f (xn )} has an accumulation point in Y . Since f is a quasi-perfect mapping, {xn } has
an accumulation point in X. Thus X is a β-space. 
Proposition 3.3.16 Let f : X → Y be a mapping.
(1) If f is a closed mapping and X is a semi-stratifiable space, then Y is also a
semi-stratifiable space [110].
(2) If f is a perfect mapping, Y is a semi-stratifiable space and X has a Gδ -diagonal,
then X is a semi-stratifiable space [74].
Proof (1) can be verified directly by Definition 1.4.3. For (2), since Y is a subpara-
compact space, X is also a subparacompact space (see Proposition A.3.2 in Appendix
A). By Propositions 1.5.20, 3.3.15 and Theorem 1.7.7, X is a semi-stratifiable space.

Lemma 3.3.17 ([439]) Let X be a countably paracompact space. If there exist a
closed set B and a sequence {Bn } of closed sets in X, such that,
3.3 σ -Spaces and Semi-Stratifiable Spaces 189

(1) B ∩ Bn = ∅, n ∈ N;
∈ τ , then there is m ∈ N such that Bn ⊂ V whenever n  m,
(2) if B ⊂ V 
then B ∩ n∈N Bn is a countably compact set in X.

Proof Otherwise, there is a discrete closed set {xk : k ∈ N} in B ∩ n∈N Bn , and
 locally finite family {Uk }k∈N of open sets in X such that
hence there is a countable
xk ∈ Uk . Since xk ∈ n∈N Bn , Uk ∩ Bn = ∅ for infinitely many n, and hence there
exist sequences {nk } ⊂ N and {yk } ⊂ X such that yk ∈ Uk ∩Bnk . Let S = {yk : k ∈ N}.
Then S is a closed set in X, so there is m ∈ N such that Bn ⊂ X − S whenever n  m,
it follows ym ∈ Bnm ⊂ X − {ym }, a contradiction. 

Theorem 3.3.18 ([439]) Every countably paracompact semi-stratifiable space sat-

isfies the decomposition theorem of type compactness.

Proof Suppose F : N × τ → τ c is a semi-stratification on a countably paracompact

space X. Let f : X → Y be a closed mapping. For every y ∈ Y and n ∈ N, define

Vy = X − f −1 (y), U(n, y) = X − F(n, Vy ).

Then f −1 (y) ⊂ U(n, y), and hence there is an open neighborhood O(n, y) of y such
that f −1 (O(n, y)) ⊂ U(n, y). For each n ∈ N, let

Yn = {y ∈ Y : ∀y
∈ Y − {y}, y ∈
/ O(n, y
)}.

Then Yn ∩ O(n, y ) ⊂ {y
}. Thus Yn is a discrete closed set in Y .
Let y ∈ Y − n∈N Yn . For each n ∈ N, there is yn = y such that y ∈ O(n, yn ).
 prove
We first that {yn } satisfying (∗): For any subsequence {nk } of {n}, f −1 (y) ⊂
−1 (ynk ).
k∈N f  
Otherwise, there is x ∈ f −1 (y) − k∈N f −1 (ynk ). Let U = X − k∈N f −1 (ynk ).
Then there is k ∈ N such that x ∈ F(k, U), so U ⊂ X − f −1 (ynk ) = Vynk , and hence

f −1 (O(nk , ynk )) ⊂ U(nk , ynk ) = X − F(nk , Vynk ) ⊂ X − F(nk , U).

On the other hand, y ∈ O(nk , ynk ) and f −1 (y) ⊂ U(nk , ynk ), it follows that x ∈
F(k, U) ∩ f −1 (y) ⊂ F(nk , U) ∩ U(nk , ynk ) = ∅, a contradiction.
Let f −1 (y) ⊂ V ∈ τ . Then there is an open neighborhood Oy of y such that
f (Oy ) ⊂ V . If there is a subsequence {nk } of {n} such that f −1 (ynk ) ⊂ f −1 (Oy ),
−1
 
then f −1 (Oy )∩( k∈N f −1 (ynk )) = ∅, and hence f −1 (y)∩ k∈N f −1 (ynk ) = ∅, which
contradicts (∗). So there is m ∈ N such that f −1 (yn ) ⊂ f −1 (Oy ) ⊂ V whenever

n  m. By (∗) and Lemma 3.3.17, f −1 (y) = f −1 (y) ∩ n∈N f −1 (yn ) is a compact set
in X. 

Question 3.3.19 (1) Does every quasi-perfect preimage of a countably paracom-

pact semi-stratifiable space satisfy the decomposition theorem of type countable
compactness [162]?
190 3 Generalized Metric Spaces

(2) Does every semi-stratifiable regular space satisfy the decomposition theorem of
type compactness?
Below we discuss the mapping properties of semi-metrizable spaces relevant to
semi-stratifiable spaces.

Lemma 3.3.20 ([119]) Suppose f : X → Y is a perfect mapping and Y is a first

countable space. If each f −1 (y) is a first countable subspace of X, then X is a first
countable space.

Proof Let y = f (x) for any x ∈ X. Suppose {Ui }i∈N is a decreasing local base of y.
Let {Vi }i∈N be a sequence of open sets in X such that {Vi ∩ f −1 (y)}i∈N is a local base
of x in f −1 (y). For each i ∈ N, define Fi (x) = f −1 (y) − Vi . Then Fi (x) ∈ K (X), and
hence there exist disjoint open sets Wi and Gi in X such that x ∈ Wi and Fi (x) ⊂ Gi .
We may assume Wi ⊂ Vi and Wi+1 ⊂ Wi . We prove {Wi ∩ f −1 (Ui )}i∈N is a local base
of x in X. Let xi ∈ Wi ∩ f −1 (Ui ). If x0 is an accumulation point of {xi }, then f (xi ) → y
because f (xi ) ∈ Ui , and hence x0 ∈ f −1 (y). If x0 = x, then there is n ∈ N such that
x0 ∈ X −Vn , so x0 ∈ Gn , it follows that there exist infinitely many xi ∈ Wn ∩Gn = ∅,
a contradiction. Consequently, x0 = x, which means {xi } at most has an accumulation
point x. If xi  x, then there is a discrete subsequence {xin } of {xi }. We may assume
/ f −1 (y) because of the compactness of f −1 (y), so {f (xin )} is discrete in Y , a
xin ∈
contradiction. Thus X is a first countable space. 

By Corollary 1.4.5, we have the following mapping theorems of semi-metrizable

spaces.

Corollary 3.3.21 Let f : X → Y be a mapping.

(1) If f is a closed mapping and X is a semi-metrizable space, then Y is a semi-
metrizable space if and only if Y is a first countable space [110].
(2) If f is a perfect mapping, Y is a semi-metrizable space and X has a Gδ -diagonal,
then X is a semi-metrizable space [74].

Question 3.3.22 ([82, 110]) Is every semi-stratifiable Fréchet–Urysohn space a

closed image of a semi-metrizable space?

Question 3.3.23 ([458]) Suppose f : X → Y is a closed mapping. If X is a sym-

metrizable space and Y is a g-first countable space, then is Y a symmetrizable space?

Example 3.3.24 (1) There is a σ -space which has no σ -locally finite closed pseudo-
network. The Mrówka space ψ(D) with |D| > c (see Example 1.8.4) is such a
space.
(2) There is a semi-metrizable space which is not a Σ -space. The bowtie space (see
Example 1.8.2) is such a space.
(3) There is a paracompact Σ-space which is not a perfect preimage of any semi-
stratifiable space. The space X in Example 3.2.38(2) is such a space, because by
Corollaries 3.2.28, 3.3.2 and Theorem 3.3.14, X is not a perfect preimage of any
semi-stratifiable space.
3.3 σ -Spaces and Semi-Stratifiable Spaces 191

(4) There is a σ -space which has no Gδ -diagonal [402].

Denote the Euclidean topology for R by τ ∗ . Suppose X is the set R with the
following topology τ (i.e. the Michael line topology about P, see Example 1.8.5):
G ∈ τ if and only if there exist U ∈ τ ∗ and A ⊂ P such that G = U ∪ A. Let
X1 = R × {1}. Define f : X → X1 by f (x) = (x, 1) and define a topology on X1 such
that f is a homeomorphic mapping. Let Z = X ⊕ X1 . Denote the quotient space of Z
obtained by identifying x and f (x) to one point for each x ∈ Q by Y . Let g : Z → Y
be the quotient mapping. Then g is a finite-to-one closed mapping.
(24.1) Y is a σ -space.
By Example 1.8.5, X is a paracompact space with a σ -disjoint base, and hence
Z is also a paracompact space with a σ -disjoint base. By Propositions 1.7.12
and 3.3.9, Y is a σ -space.
(24.2) Y has no Gδ -diagonal.
Otherwise, suppose {Un } is a Gδ -diagonal sequence in Y . For each n ∈ N,
define
Pn = {x ∈ X − Q : g(f (x)) ∈/ st(g(x), Un )}.

Then X − Q = n∈N Pn . Since τ ∗ is second category, there exist m ∈ N
and an open interval J in R such that J ⊂ clτ ∗ (Pm ). Pick x ∈ J ∩ Q. There
is U ∈ Um such that g(x) ∈ U. Then g(f (x)) ∈ U, so there is V ∈ τ such
that x ∈ V ⊂ J and g(V ∪ f (V )) ⊂ U. By the definition of τ , z ∈ V ∩ Pm ,
and hence {g(z), g(f (z))} ⊂ U, it follows that g(f (z)) ∈ st(g(z), Um ), a
contradiction. Thus Y has no Gδ -diagonal.
(5) There is a subparacompact space with a Gδ -diagonal which has no G∗δ -diagonal
[281, 443].
Take Y = [0, ω1 ]. Define a topology for Y as follows: the neighborhoods of the
only accumulation point ω1 in Y are those for the ordered topology. Then {ω1 } is not a
Gδ -set in Y . Let S = [0, ω1 ) and X = Y ∪(S×N). Give X the following topology: S ×
N is a discrete open subspace of X; for each β ∈ S, each element of a neighborhood
base of β has the form {β} ∪ {(β, k) : k  n}, n ∈ N; each element of a neighborhood
base of ω1 has the form W (n, α) = {ω1 } ∪ ((α, ω1 ) × {k ∈ N : k  n}), where
n ∈ N and α is a countable ordinal number. Since W (n, α) = (α, ω1 ] ∪ W (n, α) in
X, X is not a regular space. Since [0, ω1 ] and S × {n} are discrete closed subspaces
of X, X is an Fσ -discrete space, so X is a subparacompact space with a Gδ -diagonal.
If X has a G∗δ -diagonal, then there is a sequence {Gn } of open covers of X such

that {ω1 } = n∈N st(ω1 , Gn ). For each n ∈ N, we may assume that there exist
kn ∈ Nand a countable ordinal
 number αn such
 that ω1 ∈ W (kn , αn ) ∈ Gn . Then
ω1 ∈ n∈N (αn , ω1 ] ⊂ n∈N W (kn , αn ) ⊂ n∈N st(ω1 , Gn ) = {ω1 }, and hence
{ω1 } = n∈N (αn , ω1 ], a contradiction. Thus X has no G∗δ -diagonal.
(6) There is a space satisfying uniform (G) which does not have the point Gδ -
property [51].
Suppose D is a set of cardinality ℵ1 . Let ωD = {∞} ∪ D be the one point
compactification of the discrete space D. Then {∞} is not a Gδ -set in ωD. For each
192 3 Generalized Metric Spaces

x ∈ ωD, let 
{{∞}}, x = ∞,
Wx =
{{x}, {x, ∞}}, x = ∞.

Then W = {Wx : x ∈ ωD} is a family satisfying uniform (G).

Example 3.3.25 The mappings on σ -spaces and relevant spaces.

(1) σ -spaces are not preserved by closed mappings.

The space X in Example 3.3.24(4) is a σ -space and the closed set Q is not a
Gδ -set in X. Let h : X → X/Q be the quotient mapping. Then h is a closed
mapping. Since X/Q does not have the point Gδ -property, it is not a σ -space.
(2) There is a space with a countable closed network which does not satisfy the
decomposition theorem of type compactness [97].
Let X = I × I. The topology for X is generated by a base consisting of Euclidean
open sets in I × I and sets of the form X − ({s} × (0, 1]), ∀s ∈ I. Then X
is not a regular space. For the Euclidean topology, the union of a countable
closed neighborhood base of I × I and a countable closed neighborhood base
of I × {0} is a countable closed network for X, so X is a semi-stratifiable space.
Let f = π1 : X → I. Then f is a closed mapping, and for each s ∈ I, f −1 (s)
is not compact in X, thus X does not satisfy the decomposition theorem of type
compactness.
(3) There exist an Fσ -discrete space X, a paracompact space Y and a compact open
mapping f : X → Y such that Y is not a semi-stratifiable space [281].
Spaces X and Y are those in Example 3.3.24(5). Then X is an Fσ -discrete space.
Since every open cover of Y has a disjoint open refinement, Y is a paracompact
space. Since the singleton {ω1 } in Y is not a Gδ -set in Y , Y is not a semi-stratifiable
space. Define a function f : X → Y as follows: for each x ∈ X, if x ∈ Y , then
f (x) = x; if x = (β, n) ∈ S × N, then f (x) = β. It is easy to verify that f is a
compact open mapping.

3.4 k-Semi-Stratifiable Spaces

The research on relationships between semi-stratifiable spaces and σ -spaces is actu-

ally a deepening of the research on relationships between σ -cushioned families and
σ -closure-preserving families. The discussion on these two types of families pro-
duces a brilliant page in general topology. The characterizations of paracompact-
ness obtained by Michael [327], the characterizations of subparacompact spaces
obtained by Junnila [217] and the question raised by Junnila and Katuta (see Ques-
tion 3.2.8) are sufficient to show the difficulty and importance of this research. We
have learnt that spaces with a σ -cushioned network are not equivalent to spaces with
a σ -closure-preserving network. In this section, we discuss the situation of replacing
3.4 k-Semi-Stratifiable Spaces 193

networks with k-networks, i.e. discuss k-semi-stratifiable spaces and the spaces with a
σ -closure-preserving k-network. As for the situation of bases, i.e. the famous ques-
tion on the equivalence of M3 -spaces and M1 -spaces, is the central issue of Sect. 3.5.
The spaces with a σ -closure-preserving k-network have characterizations parallel
to that of k-semi-stratifiable spaces.

Proposition 3.4.1 ([222]) For every regular space X, the following are equivalent:
(1) X has a σ -closure-preserving strict k-network.
(2) X has a σ -closure-preserving cs∗ -network.
(3) There is a k-semi-stratifiable function g on X such that for every x, y ∈ X and
n ∈ N, g(n, y) ⊂ g(n, x) whenever y ∈ g(n, x).

Proof (1) ⇒ (2) is  obvious.

(2) ⇒ (3). Let n∈N Pn be a closed cs∗ -network for X such that each Pn is
closure-preserving. Define g : N × X → τ by g(n, x) = X − ∪{P ∈ Pi : x ∈ / P, i 
n}. Then g is a k-semi-stratifiable function on X (see Proposition 1.4.15), and for
every x, y ∈ X and n ∈ N, y ∈ g(n, x) whenever y ∈ g(n, x).
(3) ⇒ (1). Suppose g is a k-semi-stratifiable function on X satisfying (3). For
each n ∈ N, define Pn = {X − g(n, A) : A ⊂ X}. Then Pn is a closure-preserving
family of closed sets in X. Let K be a compact set and K ⊂ U ∈ τ . Since g is
is m ∈ N such that K ∩ g(m, X − U) = ∅. Then
k-semi-stratifiable function, there
K ⊂ X − g(m, X − U) ⊂ U. So n∈N Pn is a σ -closure-preserving strict k-network
for X. 

Thus, regular spaces with a σ -closure-preserving k-network ⇒ k-semi-stratifiable

spaces ⇒ σ -spaces.

Question 3.4.2 ([139]) Does every k-semi-stratifiable space have a σ -closure-

preserving k-network?

Theorem 3.4.3 ([144]) Every k-semi-stratifiable Fréchet–Urysohn space is a strat-

ifiable space.

Proof By Theorem 1.4.14, we only need to prove every k-semi-stratifiable

 Fréchet–
Urysohn space X is a monotonically normal space. Let F = n∈N Fn be a pair-strict
k-network for X, such that, each Fn is a cushioned family and Fn ⊂ Fn+1 . If H and
K are disjoint closed sets in X, then for each n ∈ N, define

Un = ∪{F1 : (F1 , F2 ) ∈ Fn , F2 ∩ K = ∅} − ∪{F1 : (F1 , F2 ) ∈ Fn , F2 ∩ H = ∅},


and let D(H, K) = ( n∈N Un )◦ . To prove D is a monotone normality operator
on X, we only need to prove H ⊂ D(H, K) ⊂ D(H, K) ⊂  X − K.
If there is x ∈ H − D(H,K), then x ∈ (X − K) ∩ (X − n∈N Un ), thus there is a
sequence {xn } in X − (K ∪ ( n∈N Un )) converging to x, and hence there exist m ∈ N
194 3 Generalized Metric Spaces

and (P1 , P2 ) ∈ Fm such that {x} ∪ {xn : n ∈ N} ⊂ P1 ⊂ P2 ⊂ X − K, it follows that

P2 ∩ K = ∅. Since x ∈ H,

x ∈ X − ∪{F2 : (F1 , F2 ) ∈ Fm , F2 ∩ H = ∅}
⊂ X − ∪{F1 : (F1 , F2 ) ∈ Fm , F2 ∩ H = ∅}.

So there is k ∈ N such that xn ∈ X − ∪{F1 : (F1 , F2 ) ∈ Fm , F2 ∩ H = ∅} whenever

n  k, and hence xn ∈ P1 − ∪{F1 : (F1 , F2 ) ∈ Fm , F2 ∩ H = ∅} ⊂ Um , a
contradiction. Thus H ⊂ D(H, K).
If there is x ∈ D(H, K) ∩ K, then x ∈ D(H, K) − H, thus there is a sequence {xn }
in D(H, K) − H converging to x, so there exist m ∈ N and (P1 , P2 ) ∈ Fm such that
{x} ∪ {xn : n ∈ N} ⊂ P1 ⊂ P2  ⊂ X − H, and hence  Un ⊂ X − P1 whenever n  m.
It follows that {xn : n ∈ N} ⊂ jm Uj , thus x ∈ jm U j . As a consequence, there
is j  m such that x ∈ U j ⊂ ∪{F2 : (F1 , F2 ) ∈ Fj , F2 ∩ K = ∅} ⊂ X − K, a
contradiction. Therefore, D(H, K) ⊂ X − K. 

Corollary 3.4.4 Every k-semi-stratifiable space satisfying one of the following con-
ditions is a metrizable space:
(1) having a p-sequence [186];
(2) having a point-countable base.

Proof Suppose X is a k-semi-stratifiable space.

(1) If X has a p-sequence, then by Theorem 2.2.18, X is a wΔ-space. By Theorems
1.7.7 and 3.4.3, X is a developable stratifiable space. By the Bing metrization
criterion, X is a metrizable space.
(2) If X has a point-countable base, then X is a semi-metrizable space. By Theorems
1.2.13 and 3.4.3, X is a developable stratifiable space, so X is a metrizable space.


The condition of k-semi-stratifiable spaces is a typical metrizable factor [489].

For example, Nagata [377] proved that a space X is a metrizable space if and only
if X has a k-semi-stratifiable function g satisfying the Nagata condition: for any
Y ⊂ X, Y ⊂ ∪{g2 (n, y) : y ∈ Y }, where g2 (n, y) = ∪{g(n, x) : x ∈ g(n, y)}. Several
transformation forms of this proposition are also correct. For example, the condition
of k-semi-stratifiable function can be strengthened to that of stratifiable function g,
and g satisfies that for any Y ⊂ X, Y ⊂ ∪{g2 (n, y) : y ∈ Y } [148]; or we can replace
the condition of k-semi-stratifiable function g as follows: x ∈ g3 (n, xn ) ⇒ xn → x
[148]. There is a k-semi-stratifiable function g on each Lašnev space such that for
any Y ⊂ X, Y ⊂ ∪{g2 (n, y) : y ∈ Y } [148]. Z. Yun [491] constructed an example
to show that the condition of k-semi-stratifiable function g can not be changed into
condition (5) of Theorem 3.3.1 as follows: x ∈ g2 (n, xn ) ⇒ xn → x.
3.4 k-Semi-Stratifiable Spaces 195

Question 3.4.5 Is every k-semi-stratifiable space with a point-countable (resp. σ -

locally countable) weak base a g-metrizable space?
A space X is called an orthocompact space if each open cover of X has an interior-
preserving open refinement. Let G : X → τ (X) be a function. For each x ∈ X, define

G2 (x) = ∪{G(y) : y ∈ G(x)},

G3 (x) = ∪{G(z) : z ∈ G2 (x)}.

Lemma 3.4.6 ([216]) Suppose G : X → τ (X) is a neighborhood assignment for

a metacompact (resp. an orthocompact) semi-stratifiable space X. Then there is a
point-finite (resp. an interior-preserving) open cover V of X such that ∩(V )x ⊂
G3 (x) for each x ∈ X.

Proof Suppose g is a semi-stratifiable function on X and g(1, x) ⊂ G(x). For each

k ∈ N, let Uk = {g(k, x) : x ∈ X}. Then Uk has a point-finite (resp. an interior-
preserving) open refinement of Wk . Define

Hk = {x ∈ X : x ∈ g(k, y) ⇒ y ∈ G(x)}.

Then X = k∈N Hk . For each x ∈ X, let k(x) = min{n ∈ N : x ∈ H n }. Define
⎛ ⎞

Qn (x) = ∩ ⎝ Wi ⎠ , n ∈ N;
in
x

V (x) = Qk(x) (x) − Hi;
i<k(x)

V = {V (x) : x ∈ X}.

Then V is a point-finite (resp. an interior-preserving) open cover. Let x ∈ X and

m = k(x). Then x ∈ H m , and hence there is z ∈ Hm ∩ g(m, x) ∩ Qm (x), so z ∈ G(x)
and z ∈ Qm (x) ⊂ g(m, y) ⊂ G(y) for some y ∈ X, it follows y ∈ G(z). Thus
∩(V )x ⊂ V (x) ⊂ Qm (x) ⊂ G(y) ⊂ G3 (x). 

Proposition 3.4.7 ([254]) Every orthocompact k-semi-stratifiable space has a σ -

closure-preserving k-network.

Proof Suppose g is a k-semi-stratifiable function on an orthocompact space X. It is

easy to verify that g3 is also a k-semi-stratifiable function on X. By Lemma 3.4.6, for
each n ∈ N, there is an interior-preserving opencover Vn of X such that ∩(Vn )x ⊂
g3 (n, x). Define h : N × X → τ by h(n, x) = ∩( in Vi )x . Then h is a g-function on
X. Since h(n, x) ⊂ g3 (n, x), h is a k-semi-stratifiable function on X. By the definition
of Vn , for every x, y ∈ X and n ∈ N, h(n, y) ⊂ h(n, x) whenever y ∈ h(n, x). By
Proposition 3.4.1, X has a σ -closure-preserving k-network. 
196 3 Generalized Metric Spaces

In the following, we discuss the covering properties of k-semi-stratifiable spaces.

Theorem 3.4.8 Every open cover of a k-semi-stratifiable space has a point-

countable sequentially open refinement.

Proof Suppose F : N × τ → τ c is a k-semi-stratification for a space X.

(8.1) Every discrete family F = {Fα }α∈Λ of closed sets in X has a point-countable
sequentially open expansion.
We define inductively the expansion family as follows:

Pα∗ (∅) = Fα ,

Pα (∅) = Pα∗ (∅) − Pβ∗ (∅),
β∈Λ−{α}

P(∅) = {Pα (∅)}α∈Λ .

Suppose we have defined P(δ) for some δ ∈ N<ω . To define P(δn), n ∈ N,

let P(δ) = {Pα (δ)}α∈Λ . Define

Pα∗ (δn) = F(n, X − Pβ∗ (δ)),
β∈Λ−{α}

Pα (δn) = Pα∗ (δn) − Pβ∗ (δn),
β∈Λ−{α}

P(δn) = {Pα (δn)}α∈Λ .

Let Uα = ∪{Pα (δ) : δ ∈ N<ω }, U = {Uα }α∈Λ . Then U is the family we need.
Obviously, Fα = Pα (∅) ⊂ Uα . Now we prove each Uα is a sequentially open
X with S → x ∈ Uα . Then there is δ ∈ N<ω such
set. Let S be a sequence in 
that x ∈ Pα (δ). Let Mα = β∈Λ−{α} Pβ∗ (δ). Then x ∈/ Mα , so S is eventually

in some F(m, X − Mα ) = Pα∗ (δm). If x ∈ β∈Λ−{α} Pβ∗ (δm), then


x∈ F(m, X − Mβ )
β∈Λ−{α}
 
⊂ F(m, (X − Mβ )) ⊂ (X − Mβ ),
β∈Λ−{α} β∈Λ−{α}

thus there is β ∈ Λ − {α} such that x ∈ X − Mβ ⊂ X − Pα∗ (δ) ⊂ X − Pα (δ), a


contradiction. So x ∈ X − β∈Λ−{α} Pβ∗ (δm), thus S is eventually in Pα (δm) ⊂
Uα .
If U is not point-countable, then there exist x ∈ X and δ ∈ N<ω such that
x ∈ Pα (δ) for uncountably many α, which contradicts the assumption that
{Pα (δ)}α∈Λ is a disjoint family.
3.4 k-Semi-Stratifiable Spaces 197

(8.2) Every open cover W of X has a point-countable sequentially open refinement.

Since X is a subparacompact space, W has a closed refinement i∈N Fi such
that each Fi = {Fα }α∈Λi is discrete. Let Ui = {Uα }α∈Λi be a point-countable
sequentially open expansion
 of Fi . Take Wα ∈ W such that Fα ⊂ Wα for
each α ∈ Λi . Then i∈N,α∈Λi Wα ∩ Uα is a point-countable sequentially open
refinement of W . 

Since the property of k-semi-stratifiable k-spaces is open hereditary, we have the

following result by Theorem 3.4.8 and Corollary 2.3.5.
Corollary 3.4.9 ([279]) Every k-semi-stratifiable k-space is a hereditarily meta-
Lindelöf space.
Lemma 3.4.10 ([279]) There is a function H : N × P(X) → τ c on each k-semi-
stratifiable space X such that
(1) H(n, W ) ⊂ H(n + 1, W ) ⊂ W ;
(2) V ⊂ W ⇒ H(n, V ) ⊂ H(n, W );
(3) if W is a sequential neighborhood of x, then every sequence converging to x is
eventually in some H(m, W );
(4) if {Gα }α∈Λ is a disjoint family in X and n ∈ N, then {H(n, Gα )}α∈Λ is a discrete
family.
Proof Suppose g is a k-semi-stratifiable function on X. Define H : N × P(X) → τ c
by H(n, W ) = X − g(n, X − W ).
Let W be a sequential neighborhood of x and xn → x. If there is xni ∈ X −
H(i, W ) = g(i, X−W ) for each i ∈ N, then there is yi ∈ X−W such that xni ∈ g(i, yi ),
and hence yi → x, a contradiction. So {xn } is eventually in some H(m, W ).
Suppose {Gα }α∈Λ is a disjoint family in X. For each x ∈ X, let

V = X − H(n, ∪{Gα : α ∈ Λ, x ∈
/ Gα }).

/ Gα , then V ∩ H(n, Gα ) = ∅, which shows {H(n, Gα )}α∈Λ

Then x ∈ V ∈ τ . If x ∈
is a discrete family. 

Theorem 3.4.11 ([279]) Every k-semi-stratifiable normal k-space is a hereditarily

paracompact space.
Proof Suppose X is a k-semi-stratifiable normal k-space. Since X is a perfect subpara-
compact space, we only need to prove X is a collectionwise normal space. Suppose
H is a function satisfying the conditions in Lemma 3.4.10. Let F = {Fα }α∈Λ be a
discrete family of closed sets in X.
(11.1) F has a disjoint
 sequential neighborhood expansion.
Define Lα = β∈Λ−{α} Fβ and Gα = n∈N (H(n, X − Lα ) − H(n, X − Fα )).
Then {Gα }α∈Λ is a disjoint family. Suppose a sequence S → x ∈ Fα . Since
Lα ∩ Fα = ∅, S is eventually in H(m, X − Lα ) − H(m, X − Fα ) ⊂ Gα and
x∈/ H(m, X − Fα ), S is eventually in some H(m, X − Lα ).
198 3 Generalized Metric Spaces

(11.2) F has a discrete closed sequential neighborhood expansion.

Since H(n, X − Fα ) ∩ Fα = ∅, there is Vα (n) ∈ τ such that

Fα ⊂ Vα (n) ⊂ Vα (n) ⊂ X − H(n, X − Fα ).


Let Fα (n) = H(n, Gα ) ∩ Vα (n) and Wα = n∈N Fα (n). Then Wα is a sequen-
tial neighborhood of Fα . Because in fact, if S → x ∈ Fα , then S is eventually
in some H(m, Gα ) ∩ Vα (m) ⊂ Fα (m) ⊂ Wα .
Let W = {Wα }α∈Λ . Then W is a disjoint family. We prove itis a closure-
preserving family of closed sets. If a sequence S → x ∈ / α∈Λ
Wα for
some Λ
⊂  Λ, then x ∈
/ α∈Λ
F α , and hence S is eventually in some
H(m, X − α∈Λ
Fα ) and H(m, X − F α ) ∩ Fα (n) = ∅ whenever α ∈ Λ
and

n  m. Let E(m, Λ ) = α∈Λ

,n<m Fα (n). By Lemma 3.4.10, {Fα (n)}α∈Λ is a
discrete family of closed sets, so E(m, Λ
) is a closed set and x ∈ / E(m,

 Λ ), and

hence S is eventually
 in X − E(m, Λ ), thus S is eventually in X − α∈Λ
Wα ,
it follows that α∈Λ
Wα is a (sequentially) closed set.
(11.3) X is a collectionwise normal space.
Let H1 be a discrete family of closed sets in X. Then there is a sequence {Hn }
of discrete families of closed sets such that Hn+1  is a sequential neighborhood
expansion of Hn . Let Hn = {Hα,n }α∈Λ , Hα = n∈N Hα,n and H = {Hα }α∈Λ .
If a sequence S → x ∈ Hα , then there is i ∈ N such that x ∈ Hα,i , so S is
eventually in Hα,i+1 ⊂ Hα , and hence Hα ∈ τ . Thus H is a disjoint expansion
of H1 . 

In the following, we turn to investigate mapping properties of k-semi-stratifiable

spaces.

Theorem 3.4.12 ([275]) Every closed mapping on a k-semi-stratifiable space is a

compact-covering mapping.
Proof Suppose f : X → Y is a closed mapping and g is a k-semi-stratifiable function
on X. For each y ∈ K ∈ K (Y ), fix xy ∈ f −1 (y). Let E = {xy : y ∈ K}. Then f (E) =
f (E) = K. To complete the proof, we only need to verify E is a countably compact
space by Corollary 1.4.12. If {xn } is a sequence in E, then there is zn ∈ E ∩ g(n, xn )
for each n ∈ N. If infinitely many zn are the same point, we may assume a point
z ∈ g(n, xn ) for each n ∈ N, then {xn } has an accumulation point. If all zn are different,
then f|E is one-to-one and {f (zn )} has a convergent subsequence, and hence {zn } has
an accumulation point, it follows that {xn } also has an accumulation point. 

Corollary 3.4.13 ([146]) Suppose f : X → Y is a closed mapping and X, Y are

regular spaces. If X is a k-semi-stratifiable space, then Y is also a k-semi-stratifiable
space.
Proposition 3.4.14 ([259]) Suppose f : X → Y is a pseudo-open compact mapping.
If X is a k-semi-stratifiable space, then Y is a semi-stratifiable space.
3.4 k-Semi-Stratifiable Spaces 199

Proof Suppose G : N × τ c (X) → τ (X) is a function satisfying the condition of

Proposition 1.4.15(3). Define D : N × τ c (Y ) → τ (Y ) by

D(n, H) = int(f (G(n, f −1 (H)))).


Then H ⊂ D(n, H), and hence H ⊂ n∈N D(n, H). If y ∈ Y − H, then f −1 (y) ∩
f −1 (H) = ∅. Since f −1 (y) is a compact set in X, there is m ∈ N such that f −1 (y) ∩
G(m, f −1 (H)) = ∅, which means y ∈ / f (G(m, f −1 (H))), so y ∈/ D(m, H). Thus,
H = n∈N D(n, H). By Proposition 1.4.4, Y is a semi-stratifiable space. 

Now we generalize Lemma 2.7.21 from metric spaces to k-semi-stratifiable

spaces. By a proof same as that of Theorem 3.4.11, we have the following Lemma.

Lemma 3.4.15 Suppose X is a k-semi-stratifiable space. If D = {xα }α∈Λ is a discrete

subset of X, then there is a disjoint sequential neighborhood expansion {Uα }α∈Λ of
D, such that, for each Λ
⊂ Λ, {yα : α ∈ Λ
} ∪ D is a sequentially closed set in X,
where yα ∈ Uα .

Proof Suppose H is a function on k-semi-stratifiable space X satisfying the condi-

tions of Lemma 3.4.10. For each α ∈ Λ, define

Lα = {xβ : α = β ∈ Λ},

Gα = (H(n, X − Lα ) − H(n, X − {xα })),
n∈N

Uα = (H(n, Gα ) − H(n, X − {xα })).
n∈N

Then Uα ⊂ Gα and Uα is a sequential neighborhood of xα . Because in fact, for each

sequence S → xα , since xα ∈ / Lα , S is eventually in some H(m, X − Lα ), and hence
S is eventually in H(m, X − Lα ) − H(m, X − {xα }) ⊂ Gα . By Lemma 3.4.10, S is
eventually in some H(k, Gα ) − H(k, X − {xα }) ⊂ Uα .
It is easy to verify that {Gα }α∈Λ is a disjoint family, so {Uα }α∈Λ is also a disjoint
family. If there is Λ
⊂ Λ such that {yα : α ∈ Λ
} ∪ D is not a sequentially closed
set in X, then there is a nontrivial sequence L in {yα : α ∈ Λ
} − D converging
to x ∈ / D. We may assume there is m ∈ N such that L  ⊂ H(m, X − D). Then
L ⊂ H(n, X − {xα }) whenever α ∈ Λ and n  m, so L ⊂ α∈Λ,n<m H(n, Gα ), and
hence there exist an infinite subset L
of L and n < m such that L
⊂ α∈Λ H(n, Gα ).
By Lemma 3.4.10(4), L
is a discrete closed set in X, a contradiction. 

Theorem 3.4.16 ([284]) Suppose f : X → Y is a closed mapping and X is a k-

semi-stratifiable space. If Y is a k-space which contains no closed copy of Sω (resp.
Sω1 ), then ∂f −1 (y) is compact (resp. separable) for each y ∈ Y .
200 3 Generalized Metric Spaces

Proof Since each k-semi-stratifiable space is a σ -space, by Proposition 3.3.10 and

Theorem 3.3.1, the space Y has a σ -discrete closed network. If K is a compact subset
of Y , then K has a countable network, and hence K is metrizable. By Corollary 2.3.5,
Y is a sequential space.
Let y ∈ Y and put

A = {x ∈ f −1 (y) : there is a sequence in X − f −1 (y) converging to x}.

(16.1) A = ∂f −1 (y).
Obviously, A ⊂ ∂f −1 (y). For each x ∈ ∂f −1 (y) and any neighborhood U of
x in X, we prove A ∩ U = ∅. In fact, since X is a regular space, there is an
open set V in X such that x ∈ V ⊂ V ⊂ U. If W is a neighborhood of y in
Y , then (f −1 (W ) ∩ V ) − f −1 (y) = ∅, which means W ∩ (f (V ) − {y}) = ∅,
so y ∈ f (V ) − {y} ⊂ f (V ) = f (V ), and hence f (V ) − {y} is not closed in
Y . It follows that f (V ) − {y} is not sequentially closed in Y . But f (V ) is
closed in Y , so there is a sequence L in f (V ) − {y} converging to y. We may
assume all the terms of L are different. Since X is a k-semi-stratifiable space,
each singleton in X is a Gδ -set. By Proposition 2.1.13, there is a convergent
sequence S in V such that f (S) is a subsequence of L. By L ⊂ f (V ) − {y},
we may assume S ⊂ V − f −1 (y) and S converges to s. Then s ∈ f −1 (y), so
s ∈ A ∩ V ⊂ A ∩ U, and hence x ∈ A, which proves A = ∂f −1 (y).
(16.2) ∂f −1 (y) is separable if the space Y contains no closed copy of Sω1 .
By (16.1), we only need to prove A is separable. We first prove A is ℵ1 -
compact. Otherwise, there is an uncountable subset D of A such that D has
no accumulation point in A. We may assume D = {xα : α < ω1 }, then D is
a discrete closed subset of A. Thus D is also a discrete subset of X, so there
is a disjoint sequential neighborhood expansion {Uα }α<ω1 of D satisfying
the conclusion of Lemma 3.4.15. Since xα ∈ A for each α < ω1 , there
is a sequence Lα in X − f −1 (y) converging to xα . By the fact that Uα is a
sequential neighborhood of xα , we may assume Lα ⊂ Uα . Below we prove
{f (Lα ) : α < ω1 } is a weakly HCP family in Y .
For each α < ω1 and any yα ∈ f (Lα ), take zα ∈ Lα such that f (zα ) = yα . If
{yα : α < ω1 } is not a discrete closed subset of Y , then by the assumption that
Y is a sequential space, there is a sequence L in {yα : α < ω1 } converging
to some point l, and we may assume that all the terms of L are different. By
Proposition 2.1.13, there is a sequence S in {zα : α < ω1 } converging to some
point s such that f (S) is a subsequence of L, so f (s) = l and f (S) = f (S) ⊂ L.
If l = y, then by yα = y for each α, y ∈ / L, and hence D ∩ S ⊂ f −1 (y) ∩
−1
f (L) = ∅. Further by Lemma 3.4.15, any subspace of S is a sequentially
closed subset of X, which contradicts the assumption that S is a convergent
sequence. Thus l = y, so s ∈ f −1 (y), and hence s ∈ A. Since S ⊂ X − f −1 (y),
D ∩ S ⊂ {s}. If s ∈ D, then since D is closed in A, s ∈ D ∩ A = D, and
hence there is β < ω1 such that s = xβ ∈ Uβ . Since S converges to s, there
exist infinitely many zα ∈ Uβ ∩ Uα , which contradicts the assumption that
{Uα }α<ω1 is a family of disjoint sets. So D ∩ S = ∅. Again by Lemma 3.4.15,
3.4 k-Semi-Stratifiable Spaces 201

every subspace of S is a sequentially closed subset of X, a contradiction. Now,

we have proved {f (Lα ) : α < ω1 } is a weakly HCP family in Y .
For each α < ω1 , since f (Lα ) is a convergent sequence in Y , by Lemma 2.5.4,
there exist a finite subset Fα of f (Lα ) and a finite subset Λα of ω1 such that
(f (Lα ) − Fα ) ∩ f (Lβ ) = ∅ for every β ∈ ω1 − Λα . Let Kα = f (Lα ) − Fα .
By the transfinite induction, we can choose an uncountable  subset Γ of ω1
such that {Kα }α∈Γ consists of disjoint sets. Let W = {y} ∪ α∈Γ Kα . Then W
is a copy of Sω1 . If W is not closed in Y , because Y is a sequential space, so
there is a sequence L in W converging to some point in Y − W . Thus, every
Kα ∩ L (∀α ∈ Γ ) is finite, and hence there is a subsequence L
of L such
that every Kα ∩ L
(∀α < Γ ) is at most a singleton. Since {Kα : α ∈ Γ } is a
weakly HCP family in Y , L
is closed in Y , a contradiction. This contradiction
shows Y contains a closed copy W of Sω1 , which contradicts our assumption.
In the above, we proved A is an ℵ1 -compact subspace of X. Since A has a
σ -discrete network, A also has a countable network, so A is separable.
(16.3) ∂f −1 (y) is compact if the space Y contains no closed copy of Sω .
Similar to the proof of (16.2), we can prove that any countable infinite subset
of A has an accumulation point in A, and hence A is countably compact. Since
every countably compact subset of a σ -space is a compact metrizable space,
A is compact. By (16.1), ∂f −1 (y) = A = A is compact. 

Combined with Corollary 3.4.9, we have the following result.

Corollary 3.4.17 ([286]) Suppose f : X → Y is a closed mapping and X is a k-

semi-stratifiable k-space. If Y contains no closed copy of Sω1 , then f is a peripheral
L-mapping.

Question 3.4.18 Does every k-semi-stratifiable space satisfy the perfect preimage
Gδ -diagonal theorem?

Example 3.4.19 (1) There is a locally compact pseudo-compact developable space

which is not a meta-Lindelöf space, the Mrówka space ψ(N) (see Example 1.8.4) is
such a space.
(2) There is a k-semi-stratifiable
 space which is not a meta-Lindelöf space [253].
Take X = R ∪ ( n∈N Q × {1/n}). Give X the following topology: every point in
X − R is an isolated point of X; for each x ∈ R, each element of a neighborhood
base of x has the form
⎛ ⎞

{x} ∪ ⎝ ([ax,n , x) ∩ Q) × {1/n}⎠ , where m ∈ N and ax,n < x.
nm

Then X is a regular space.

Since R and Q × {1/n} are discrete closed subspaces of X, X is an Fσ -discrete
space. If there is an infinite compact set K in X, then there exist infinitely many n
202 3 Generalized Metric Spaces

such that K ∩ (Q × {1/n}) = ∅. Because K is a metrizable space, there exist a

subsequence {ni } of {n} and xi ∈ K ∩ (Q × {1/ni }) such that xi → x0 ∈ R. For
each i ∈ N, denote xi = (ri , 1/ni ). By the structure of neighborhoods of x0 , we may
assume ri < x0 . For each n ∈ N, let

(ri + x0 )/2, n = ni ,
qn =
x0 − 1, n = ni , i ∈ N.

Define  

U = {x0 } ∪ ([qn , x0 ) ∩ Q) × {1/n} .
n∈N

Then x0 ∈ U ∈ τ (X) and xi ∈ / U for all i ∈ N, a contradiction. Thus every compact

set in X is a finite set. Since X is an Fσ -discrete space, X has a σ -closure-preserving k-
network, so X is a k-semi-stratifiable space. By the separability of X and the fact that
R is a closed discrete subspace of X, X is neither a normal space nor a meta-Lindelöf
space.
(3) Foged [131] constructed an example to show that there is a k-semi-stratifiable
k-space which is not a normal space, and under the assumption of MA+¬CH, there
is a k-semi-stratifiable normal space which is not paracompact.
Example 3.4.20 On mapping properties of k-semi-stratifiable spaces and relevant
spaces.
(1) There is a closed L-mapping f : X → Y , where X is an ℵ0 -space and Y contains
no closed copy of Sω , such that, f is not a peripherally compact mapping.
Let X be a subspace of βN satisfying N ⊂ X and |X − N| = ℵ0 . Then X has
a countable cs-network, and hence X is an ℵ0 -space. Let C = X − N and let
f : X → X/C be the quotient mapping. Then f is a closed L-mapping, and hence
f is a compact-covering mapping. Since each compact set in X is finite, each
compact set in f (X) is also finite, so the quotient space X/C contains no closed
copy of Sω . Since ∂f −1 ([C]) = C, f is not a peripherally compact mapping.
(2) There is a closed mapping f : X → S1 such that X is a Moore space and f is not a
peripheral L-mapping (see Example 2.6.4). In this example ∂f −1 (0) = ψ(N)−N
even is not a Lindelöf space.
(3) There is a closed mapping f : X → Y , where X is an ℵ-space and Y contains no
closed copy of Sω , such that, f is not a peripheral L-mapping (C. Liu, 2005).
For the Arens space S2 = {0} ∪ N ∪ (N × N) (see Example 1.8.6) and α < ω1 ,
let Xα = S2 − N. Then Xα is a paracompact ℵ-space. Suppose X = α<ω1 Xα
and let A be the set of all accumulation points in X. Then A is a closed set in the
paracompact ℵ-space X. Suppose Y = X/A is the quotient space and f : X → Y
is the quotient mapping. Then f is a closed mapping, so f is a compact-covering
mapping. Since each compact set in X is finite, Y contains no closed copy of Sω .
However, ∂f −1 ([A]) = A is not a Lindelöf space.
Question 3.4.21 Is every normal k-and σ -space a paracompact space?
3.4 k-Semi-Stratifiable Spaces 203

Question 3.4.22 Is every regular k-space with a star-countable k-network a k-semi-

stratifiable space?

Question 3.4.23 Is every regular pseudo-open compact image of a k-semi-

stratifiable space a σ -space?

3.5 Mi -Spaces

The concept of Mi -spaces introduced by Ceder [92] opened the prelude of research
on generalized metric spaces. The questions “whether every M3 -space is an M2 -
space” and “whether every M2 -space is an M1 -space” raised by Ceder are among
the most difficult classic problems in general topology [342, 392]. Gruenhage [160]
and Junnila [216] proved that every M3 -space is an M2 -space independently. This
success inspired people pay more enthusiasm to research on the problems “whether
every M2 -space is an M1 -space”, and led to a rapid wave of research on Mi -spaces
in the late 70s to the 80s of the 20th century. Although Ceder’s questions have
not obtained the final solution, researchers created several methods to explore these
questions. A series of equivalent conditions have been given and several important
progresses have been gotten. To introduce the most important results on the research
of Mi -spaces, this section is divided into two parts. In the first part, we investigate
the characterizations of M2 -spaces, and prove that every M3 -space is an M2 -space by
Junnila’s method [216]. In the second part, mainly by using Itō’s technique [207, 208],
we give characterizations of M1 -spaces, and introduce the work done by Mizokami
[349, 351] and others. We focus on the class of hereditarily M1 -spaces and the class
of M3 -spaces in which every point has a closure-preserving local base. We establish a
discriminant rule of solving Ceder’s questions by mappings and give some mapping
theorems of M1 -spaces.
In order to understand the sequence of events of the research on Mi -spaces, in
the following, we list the main questions proposed by Ceder [92] in 1961 and the
relevant questions. For the sake of convenience, we will collectively call them Ceder’s
questions.

Question 3.5.1 Ceder’s questions:

(1) Is every M2 -space an M1 -space?
(2) Is every M3 -space an M2 -space?
(3) Does any point of an M1 -space have a closure-preserving open neighborhood
base in the space?
(4) Are M1 -spaces closed hereditary?
(5) Are M1 -spaces preserved by closed mappings or peripherally compact closed
mappings?
(6) If A is a closed subspace of an M1 -space X, then is X/A an M1 -space?
(7) Are M2 -spaces preserved by closed mappings?
204 3 Generalized Metric Spaces

(8) Are M3 -spaces preserved by closed mappings?

The relevant questions:
(9) Does every closed subset of an M1 -space have a closure-preserving open neigh-
borhood base [346]?
(10) Are M1 -spaces preserved by perfect mappings [88]?
(11) If an M3 -space X is a countable union of closed M1 -subspaces, is X an M1 -space
[161]?
We first introduce a characterization of M2 -spaces.
Proposition 3.5.2 ([68, 92]) A space X is an M2 -space if and only if X is a para-
compact σ -space such that each closed subset of X has a (σ -)closure-preserving
quasi-base in X.
Proof The necessity. Obviously, each M2 -space is a paracompact σ -space. Suppose

n∈N Bn is a quasi-base for the M2 -space X such that each Bn is a closure-preserving
family of closed sets. For each F ∈τ c , there is a decreasing sequence {Gn } of open
sets such that F ⊂ Gn and F = n∈N Gn . For each n ∈ N, define Hn = Gn and
Un = Bn|Hn . Then Un is a closure-preserving family of closed set in X. Since
{Hn : n ∈ N} is locally finite in X − F, n∈N Un is closure-preserving in X − F.
Define
 

V = ∪U : U ⊂ Un ,
n∈N
W = {V ∈ V : F ⊂ V ◦ }.

Then V is closure-preserving in X − F, so W is closure-preserving in X. Let x ∈

F ⊂ G ∈ τ . Then there exist nx ∈ N and Bx ∈ Bnx such that x ∈ Bn◦x ⊂ Bnx ⊂ G.
Let V = x∈F (Bnx ∩ Hnx ). Then V ∈ W and F ⊂ V ◦ ⊂ V ⊂ G. Thus W is a
closure-preserving quasi-base of F in X.
The sufficiency. Suppose every closed subset of a paracompact σ -space X has a σ -
closure-preserving quasi-base in X. Let n∈N Pn be a network for X such that each
Pn is a discrete family of closed sets. For each n ∈ N, denote Pn = {Pα }α∈Λn . Then
there is a discrete
 family Hn = {Hα }α∈Λn of open sets such that Pα ⊂ Hα . For each
α ∈ Λn , let k∈N Bα,k be a quasi-base of Pα in X, where Bα,k is closure-preserving
in X and ∪Bα,k ⊂ Hα . For every n, k ∈ N, define

Bn,k = ∪{Bα,k : α ∈ Λn }.

Then Bn,k is closure-preserving in X. For every x ∈ U ∈ τ , there exist n ∈ N

and α ∈ Λn such that x ∈ Pα ⊂ U, so there exist k ∈ N and B ∈ Bα,k such
that Pα ⊂ B◦ ⊂ B ⊂ U, and hence x ∈ B◦ ⊂ B ⊂ U. Thus n,k∈N Bn,k is a
σ -closure-preserving quasi-base for X, consequently X is an M2 -space. 

Theorem 3.5.3 ([160, 216]) Every M3 -space is an M2 -space.

3.5 Mi -Spaces 205

Proof Suppose g is a stratifiable function on an M3 -space X. Then g3 is also a

stratifiable function on X. Because in fact, if y ∈ X − H ∈ τ , then there is m ∈ N
such that y ∈ / g(m, H), so there is n  m such that y ∈/ g(n, g(m, H)), and hence
there is k  n such that y ∈
/ g(k, g(n, g(m, H))) ⊃ g3 (k, H). By Lemma 3.4.6, for
that ∩(Vn )x ⊂ g3 (n, x)
each n ∈ N, there is a point-finite open cover Vn of X such 
for each x ∈ X. Define h : N × X → τ by h(n, x) = ∩( in Vi )x . Then h is a
g-function on X with h(n, x) ⊂ g3 (n, x) such that for every x, y ∈ X and n ∈ N,
h(n, x) ⊂ h(n, y) whenever x ∈ h(n, y). By Proposition 1.3.12, X is an M2 -space.


Therefore, the class of M2 -spaces = the class of M3 -spaces = the class of strati-
fiable spaces.
Proposition 3.5.4 ([186]) Monotonically normal spaces are preserved by closed
mappings.
Proof Suppose f : X → Y is a closed mapping and D1 is a monotone normality
operator for X. For every pair H, K of disjoint closed sets in Y , define

D2 (H, K) = Y − f (X − D1 (f −1 (H), f −1 (K))).

Then H ⊂ D2 (H, K) ∈ τ (Y ). Obviously, if a pair H

, K
of disjoint closed sets
in Y satisfies H ⊂ H
, K ⊃ K
, then D2 (H, K) ⊂ D2 (H
, K
). If y ∈ K, then
y ∈ Y − f (D1 (f −1 (H), f −1 (K))). By D2 (H, K) ⊂ f (D1 (f −1 (H), f −1 (K))), y ∈
/
D2 (H, K), so D2 (H, K) ⊂ Y − K, and hence D2 is a monotone normality operator
for Y , thus Y is a monotonically normal space. 

Theorem 3.5.5 ([65]) (1) M3 -spaces are preserved by closed mappings.

(2) M3 -spaces satisfy the perfect preimage Gδ -diagonal theorem.
Proof (1) is obtained by Propositions 3.5.4, 3.3.16 and Theorem 1.4.14, and (2) is
obtained by Corollaries 2.2.11 and 2.1.9. 

In the second part of this section, we investigate the properties of M1 -spaces. We

first introduce several subclasses of the class of M1 -spaces, which is closely related to
Ceder’s questions. Every first countable M3 -space is called a Nagata space [92]. For
the sake of convenience of narration, in this section, the class of M3 -spaces in which
every point has a closure-preserving local base is denoted by P. The class P has
different meanings in the literature. For examples, the class of M1 -spaces in which
every closed set has a closure-preserving open neighborhood base is denoted by P
in [345]; the class of M3 -spaces in which every closed set has a closure-preserving
local base is also denoted by P in [449]. Itō [208] proved that every closed set in
an M3 -space in which every point has a closure-preserving local base has a closure-
preserving open neighborhood base (see Theorem 3.5.8), thus the three subclasses
of M1 -spaces above-mentioned coincide.
206 3 Generalized Metric Spaces

Theorem 3.5.6 ([207]) Every hereditarily M1 -space is in the class P.

Proof Suppose X is a hereditarily M1 -space. First, we prove that for each x ∈ X,

there exist regular closed sets H1 , H2 in X such that
(6.1) X = H1 ∪ H2 and if x ∈ Hi , i = 1, 2, then there
 is a sequence {Hi,n } of clopen
subsets of the subspace Hi such that {x} = n∈N Hi,n . 
In fact, take a sequence {Gn } of regular open sets in X such that {x} = n∈N Gn
and Gn+1 ⊂ Gn ⊂ G1 = X. Let

H1 = G2n−1 − G2n ,
n∈N

H2 = G2n − G2n+1 .
n∈N

Then X = H1 ∪ H2 . Since Gn is a regular open set in X, Gn − Gn+1 =

Gn − Gn+1 is a regular closed set, so H1 , H2 are regular closed sets in X.
Let
H1,n = H1 ∩ G2n−1 , H2,n = H2 ∩ G2n .

Then H1,n = H1 ∩ G2n−2 , H2,n = H2 ∩ G2n−1 . Thus {Hi,n } is a sequence of

clopen sets in the subspace Hi such that {x} = n∈N Hi,n for each x ∈ Hi .
Now, for i = 1, 2, let Bi = ∅ for each x ∈ / Hi . Suppose x ∈ Hi and n∈N Un
 base of x in Hi such that each Un is closure-preserving in Hi . Then
is a local
Vi = n∈N (Un|Gn ) is a closure-preserving local base of x in Hi , because Hi
is an M1 -space. Let Bi = V i . Then Bi is a closure-preserving regular closed
local quasi-base of x in Hi , and Bi is also a family of regular closed sets in X.
Define
B = {B1 ∪ B2 : Bi ∈ Bi , i = 1, 2}.

By (6.1), B is a closure-preserving regular closed local quasi-base of x in X,

so B ◦ is a closure-preserving local base of x in X. 

Lemma 3.5.7 ([449]) If F is a closure-preserving family of closed sets in a σ -space

X, then there is a dense Fσ -discrete subset D of X such that F ∩ D is a dense subset
of F for each F ∈ F .

Proof Suppose G : N×τ c → τ is a function satisfying the assumption of Proposition

1.4.4(3). Define
Q(F
) = ∩F
− ∪(F − F
), F
⊂ F .

Let Q(∅) = X − ∪F . Then {Q(F

) : F
⊂ F } is a cover of X. For each n ∈ N,
define
Qn (F
) = ∩F
− G(n, ∪(F − F
)), F
⊂ F
3.5 Mi -Spaces 207


and Qn (∅) = X − G(n, ∪F ). Then Q(F
) = n∈N Qn (F
) and {Qn (F
) : F
⊂
F } is a discrete family of closed sets in X. Because in fact, define

U = G(n, {x}) − ∪(F − (F )x )

for each x ∈ X, then x ∈ U ∈ τ . Suppose (F )x = F

⊂ F . If there is F ∈
F
− (F )x , then Qn (F
) ⊂ F and U ⊂ X − ∪(F − (F )x ) ⊂ X − F, and hence
Qn (F
) ∩ U = ∅; if there is F ∈ (F )x − F
, then

Qn (F
) ⊂ X − G(n, ∪(F − F
)) ⊂ X − G(n, F) ⊂ X − G(n, {x}),

and hence Qn (F
) ∩ U = ∅.
Define
Q = {Qn (F
) : F
⊂ F , n ∈ N}.

Then Q is a σ -discrete closed refinement of {Q(F

) : F
⊂ F }. Let H be a
σ -discrete closed network for X. For every H ∈ H and Q ∈ Q with H ∩ Q = ∅,
take x(H, Q) ∈ H ∩ Q. Define

D = {x(H, Q) : H ∈ H , Q ∈ Q and H ∩ Q = ∅}.

We prove that D satisfies our requirement. Obviously, D is an Fσ -discrete set in X.

For each F ∈ F and any nonempty open subset W of F, there is V ∈ τ such that
V ∩ F = W . Fix y ∈ W . Then there exist H ∈ H and Q ∈ Q such that y ∈ H ∩ Q ⊂
H ⊂ V . For every F
⊂ F and n ∈ N, let Q = Qn (F
). Since Q ∩ F = ∅, F ∈ F
,
and hence Q ⊂ F, it follows that x(H, Q) ∈ H ∩ Q ⊂ V ∩ F = W , thus F ∩ D is
dense in F. In the same way, we can prove D is dense in X. 

Theorem 3.5.8 ([208]) Every closed subset of a space in the class P has a closure-
preserving open neighborhood base.

Proof Suppose X ∈ P and F ∈ τ c (X). By Proposition 3.5.2, F has a closure-

preserving
 closed quasi-base U in X. By Lemma 3.5.7, there is a subset D =
n∈N nD of X, where Dn is a discrete closed subspace of X such that for each
U ∈ U , U ∩ D is dense in U. For each n ∈ N, there is a discrete family {Gx }x∈Dn of
open sets in X such that x ∈ Gx . Let Vx be a closure-preserving local base of x for
each x ∈ D. We may assume that ∪Vx ⊂ Gx ∩ g(n, x) for each x ∈ Dn , where g is a
stratifiable function on X. 
A function ϕ is called a selection function on U ∩ D if, ϕ : U ∩ D → x∈D Vx
satisfies ϕ(x) ∈ Vx . For U ∈ U and any selection function ϕ on U ∩ D, define

B(U, ϕ) = ∪{ϕ(x) : x ∈ U ∩ D}.

Let
B = {B(U, ϕ) : U ∈ U , ϕ is a selection function on U ∩ D}.
208 3 Generalized Metric Spaces

We prove B is a closure-preserving regular closed quasi-base of F in X.

(8.1) B is a regular closed quasi-base of F in X.

Obviously, B is a family of regular closed sets in X. For each B(U, ϕ) ∈ B,
since U ∩ D is dense in U, F ⊂ U ◦ ⊂ U = U ∩ D ⊂ B(U, ϕ). Suppose
F ⊂ W ∈ τ . Take W
∈ τ such that F ⊂ W
⊂ W
⊂ W . Then there is
U ∈ U such that F ⊂ U ◦ ⊂ U ⊂ W
, and there is φ(x) ∈ Vx such that
x ∈ φ(x) ⊂ W
for each x ∈ U ∩ D. Thus φ is a selection function on U ∩ D
and B(U, φ) ⊂ W
⊂ W .
(8.2) B is a closure-preserving family.
For every B
⊂ B and p ∈ X − ∪B
, define

U
= {U ∈ U : there is a selection function ϕ on U ∩ D such that B(U, ϕ) ∈ B
}.

Since U ⊂ B(U, ϕ), p ∈ X − ∪U

∈ τ , so there is k ∈ N such that p ∈ /
g(k, ∪U
). Let H = X − g(k, ∪U
). Then p ∈ H ∈ τ . For every U ∈ U
, n 
k and x ∈ U ∩ Dn , we have ϕ(x) ⊂ ∪Vx ⊂ g(n, x) ⊂ g(k, ∪U
) ⊂ X − H, so

B(U, ϕ) ∩ H = ∪{ϕ(x) : x ∈ U ∩ (∪{Dn : n < k})} ∩ H


for each B(U, ϕ) ∈ B
. Since {Gx : x ∈ n<k Dn } is a locally finite family, Vx
is a closure-preserving family and ∪Vx ⊂ Gx , thus

V = ∪{Vx : x ∈ Dn , n < k}

is a closure-preserving family, so {ϕ(x) : x ∈ U ∩ ( n<k Dn ), B(U, ϕ) ∈ B
}
is closure-preserving, and hence p ∈ / ∪B
. Therefore, ∪B
is a closed set in
X.
By (8.1) and (8.2), B ◦ is a closure-preserving open neighborhood base of F
in X. 

Similar to Proposition 3.5.2, we have the following theorem.

Theorem 3.5.9 ([349]) A space X is an M1 -space if and only if X is a paracompact σ -

space and each closed subset of X has a (σ -)closure-preserving open neighborhood
base in X.

The proof of the sufficiency is similar to that of Proposition 3.5.2. The necessity is
an important result obtained by Mizokami in 2004 and the proof of it is very difficult
[349]. Mizokami [349] proved that (1) every adjunction space of M1 -spaces is M1 ;
(2) a stratifiable space which is the union of countably many closed M1 -subspaces
is M1 (see Question 3.5.1 (11)). Interested readers can read his original paper.

Corollary 3.5.10 ([208]) Every Nagata space is an M1 -space.

3.5 Mi -Spaces 209

Now, we have proved that

the class of Nagata spaces ⊂ the class of hereditarily M1 -spaces

⊂ the class P = the class of M1 -spaces.

In 2000, Mizokami and Shimane [351] proved a result stronger than Corollary 3.5.10
that every k-and M3 -space is an M1 -space. Another partial result due to Mizokami,
Shimane and Kitamura [353] is the following: a stratifiable space X is M1 if it has
the following property: whenever U is dense open in X and x ∈ X − U, there is
a closure-preserving family F of closed subsets of X which is a net at x such that
F ∩ U = F for every F ∈ F . The introductions and discussions of these classes of
spaces are caused by Ceder’s questions on M1 -spaces. We believe that the mapping
theorems will be powerful means for investigating the further relationships among
these classes of spaces and for the final resolution of Ceder’s questions.

Theorem 3.5.11 [185] (The Heath-Junnila theorem) Suppose X is an M3 -space.

(1) There is an M1 -space Z such that X is homeomorphic to a closed subspace of Z.
(2) There is an M1 -space Y such that X is a perfect image of Y .
Proof Let Z = X × S1 . Give Z the following topology: each point of X × (S1 − {0})
is isolated in Z; a neighborhood base of a point in X × {0} is the neighborhood base
of this point in the product topology for X × S1 . Obviously, Z is a regular space, and
X is homeomorphic to the closed subspace X × {0} of Z. For each B ⊂ X, define

B[0] = B × {0};
B[n] = B × ({0} ∪ {1/k : k  n}), n ∈ N.

If B is a closed set in X, then B[n] = B[n] − B[0], and hence B[n] is a regular closed
set in Z.
(11.1) Z is an M1 -space. 
By Theorem 3.5.3, X has a closed quasi-base B = n∈N Bn such that each
Bn is closure-preserving. Let

Un,m = {B[m] : B ∈ Bn }, n, m ∈ N;
Vn = {{(x, 1/n)} : x ∈ X}.

Then Un,m is a closure-preserving familyof regular closedsets and Vn is a

discrete family of closed sets in Z, so ( n,m∈N Un,m ) ∪ ( n∈N Vn ) is a σ -
closure-preserving regular closed quasi-base for Z. By Proposition 1.3.11, Z
is an M1 -space. 
By Lemma 3.5.7, X has a dense subset D = n∈N Dn , where Dn is a discrete
closed subspace of X, Dn ⊂ Dn+1 and B ∩ D is dense in B for each B ∈ B.
Define
210 3 Generalized Metric Spaces

Y = X[0] ∪ (∪{Dn × {1/n} : n ∈ N}).

(11.2) Y is an M1 -space.
We first prove that clY ((B[n] − B[0]) ∩ Y ) = B[n] ∩ Y for each B ∈ B. In
fact, it is obvious that

clY ((B[n] − B[0]) ∩ Y ) ⊂ B[n] − B[0] ∩ Y = B[n] ∩ Y .

Let x ∈ B and let W be a neighborhood of x in X. For each k  n, since

B ∩ D is dense in B, there exist m  k and x
∈ B ∩ W ∩ Dm . Then
(x
, 1/m) ∈ W [k] ∩ ((B[n] − B[0]) ∩ Y ), and hence

B[n] ∩ Y ⊂ clY ((B[n] − B[0]) ∩ Y ),


 means B[n] ∩ Y is a regular closed set in Y , so ( n,m∈N Un,m|Y ) ∪
which
( n∈N Vn|Y ) is a σ -closure-preserving regular closed quasi-base for Y , thus
Y is an M1 -space.
(11.3) X is a perfect image of Y .
Take f = π1|Y : Y → X. Obviously, f is a compact mapping. Let F be a
closed set in Y . If x ∈ X − f (F), then (x, 0) ∈/ F, so there exist an open
neighborhood V of x and m ∈ N such that V [m] ∩ F = ∅. For each n < m,
since x ∈/ f (F ∩ (X × {1/n})) ⊂ Dn , there is an open neighborhood
 Vn of
x such that Vn ∩ f (F ∩ (X × {1/n})) = ∅. Let U = V ∩ ( n<m Vn ). Then
x ∈ U ∈ τ (X) and U ∩ f (F) = ∅, so f (F) ∈ τ c (X). Thus f is a closed
mapping, and hence f is a perfect mapping. 

Itō [208] proved the spaces Z and Y in Theorem 3.5.11 belong to P before
Mizokami obtained the proof of Theorem 3.5.9.

Corollary 3.5.12 ([185]) The following are equivalent:

(1) Every M3 -space is an M1 -space.
(2) M1 -spaces are closed hereditary.
(3) M1 -spaces are preserved by closed mappings.
(4) M1 -spaces are preserved by perfect mappings.

Thus, Ceder’s questions boil down to the mapping properties of M1 -spaces. In the
following, we introduce some mapping theorems of M1 -spaces.

Definition 3.5.13 Let f : X → Y be a mapping.

(1) f is called an irreducible mapping [119] if f (F) = Y for every closed proper
subset F of X.
(2) f is called a quasi-open mapping if f (U)◦ = ∅ for every nonempty open subset
U of X.
3.5 Mi -Spaces 211

Obviously, every open mapping is a quasi-open mapping.

Proposition 3.5.14 ([223]) Every irreducible pseudo-open mapping is a quasi-open

mapping.

Proof Suppose f : X → Y is an irreducible pseudo-open mapping. For each

U ∈ τ (X) − {∅}, by the irreducibility of f , there is y ∈ Y − f (X − U), it means
f −1 (y) ⊂ U, and hence y ∈ f (U)◦ , it follows that f (U)◦ = ∅. Thus f is a quasi-open
mapping. 

Lemma 3.5.15 ([223]) Suppose f : X → Y is a quasi-open closed mapping. If B

is a closure-preserving family of open sets in X, then f (B)◦ is a closure-preserving
family in Y .

Proof If y ∈ ∪f (B
)◦ for each B
⊂ B, then y ∈ f (∪B
), so there is B ∈ B

such that f −1 (y) ∩ B = ∅. If y ∈ V ∈ τ (Y ), then f −1 (V ) ∩ B = ∅, and hence

∅ = f (f −1 (V ) ∩ B)◦ = V ∩ f (B)◦ , it follows that y ∈ f (B)◦ . Thus f (B)◦ is a
closure-preserving family. 

Theorem 3.5.16 ([296]) M1 -spaces are preserved by quasi-open closed mappings.

Proof Suppose f : X → Y is a quasi-open closed mapping and X is an M1 -space. By

Theorem 3.5.5, Y is an M3 -space. For each y ∈ Y , by Theorem 3.5.9, the set f −1 (y)
has a closure-preserving open neighborhood base B in X, and by Theorem 3.5.15,
f (B)◦ is a closure-preserving local base of y. By Theorems 3.5.8 and 3.5.9, Y is an
M1 -space. 

The following corollary of Proposition 3.5.14 and Theorem 3.5.16 answers a

classic question [383, 384].

Corollary 3.5.17 ([296]) M1 -spaces are preserved by irreducible closed mappings.

Proposition 3.5.18 If A is a closed subspace of an M1 -space X, then X/A is an

M1 -space.

Proof Let q : X → X/A be the quotient mapping. Then q is a closed mapping, so

X/A is an M3 -space. By Theorem 3.5.9, each point of X/A has a closure-preserving
local base, so X/A is an M1 -space. 

Now, we know that the answers of problems (2), (3), (6), (7), (8), (9) and (11)
in Question 3.5.1 are positive, and problems (1), (4), (5) and (10) are equivalent.
The following theorem obtained by Itō [207] is an important result on the mapping
properties of M1 -spaces. We do not give his proof here for the lack of space.

Theorem 3.5.19 ([207]) Hereditary M1 -spaces are preserved by closed mappings.

212 3 Generalized Metric Spaces

Corollary 3.5.20 ([427]) Every Lašnev space is a hereditary M1 -space.

Example 3.5.21 (1) There is an M1 -space which is not a k-space, for example the
Michael space (see Example 1.8.8).
(2) There is a Nagata space which is not a Lašnev space, for example the butterfly
space (see Example 1.8.3).
(3) There is a monotonically normal space which is not an M3 -space, for example
the Sorgenfrey line (see Example 1.8.9).
(4) There is a Fréchet–Urysohn M1 -space having a point-countable k-network which
is not a Lašnev space [419].

Let X = IB ×S1 , where IB is the Michael line (see Example 1.8.5). Put A = IB ×{0}.
X is endowed by the following topology: each point in IB × {1/n : n ∈ N} is isolated;
each point in A has neighborhoods in the product space IB × S1 . Let Y = X/A
and f : X → Y be the quotient mapping. Then f is a closed L-mapping. Since X
has a point-countable base, Y is a Fréchet–Urysohn space with a point-countable
k-network by Corollary 3.1.17. Since the unique non-isolated point of Y is a Gδ -
point, Y is an M1 -space. Y has no closed subspace which is homeomorphic to Sω1
by Lemma 2.7.20. If Y is Lašnev, by Theorem 2.7.22, Y is an ℵ-space. This is a
contradiction, because Y does not have any point-countable cs∗ -network [419]. Thus
Y is not a Lašnev space.
Research on Mi -spaces is still continuing [352]. One is to discuss the Mi -space
properties of a function space Ck (X), for example Reznichenko [408] proved that for
a separable metric space X, Ck (X) is a stratifiable space if and only if X is a Polish
space. The another is to discuss local properties of Mi -spaces, such as the introduction
and research of m1 -spaces (i.e., spaces having a closure-preserving local base at every
point), m2 -spaces (i.e., spaces having a closure-preserving local quasi-base at every
point) and m3 -spaces (i.e., spaces having a cushioned local pair-base at every point)
[73, 115, 251].

3.6 Developable Spaces and p-Spaces

The class of developable spaces was introduced in general topology because of

the metrization theorems with the form of sequences of covers. We can define or
characterize numbers of generalized metric spaces, such as M-spaces, wΔ-spaces,
p-spaces, strict p-spaces and so on by means of sequences of covers. In the second
chapter, we have seen the role of these sequences in the study of metric spaces. The
purpose of this section is to introduce characterizations and mapping theorems of
generalized metric spaces relevant to developable spaces and p-spaces. Particularly,
we shall give the detailed proof of S. Jiang’s theorem [213] that every strict p-space
is a submetacompact space. This theorem solved a long standing difficult classic
problem interested by many famous general topologists.
3.6 Developable Spaces and p-Spaces 213

Theorem 3.6.1 [213] (The Jiang theorem) Every space with a strict p-sequence is
a submetacompact space.
Proof Suppose {Un } is a strict p-sequence in a space X such that Un+1 refines Un . A
cover U = {Uα : α ∈ Γ } of X is well-monotone if the index set Γ is well-ordered
by < such that Uα ⊂ Uβ if every α, β ∈ Γ with α < β. Let U = {Uα }α<γ be a
well-monotone open cover of X, where Uα ⊂ Uβ whenever α < β < γ . For each
x ∈ X, define

Cx = st(x, Un ),
n∈N
α(x) = min{α < γ : x ∈ Uα },
β(x) = min{β < γ : Cx ⊂ Uβ },
n(x) = min{n ∈ N : st(x, Un ) ⊂ Uβ(x) }.

For each n ∈ N, fix Un,x ∈ (Un )<ω

x such that Cx ⊂ ∪Un,x . Let

Un,x = (∩Un,x ) ∩ Uα(x) .

 inductively a countable family Rm of open covers of X on

Now, we construct
m ∈ N such that m∈N Rm is an open θ -refinable sequence of U . For the sake of
brevity, we make the following conventions:
(i) if f is a function from a subset of N to N, then f|n = f|{1,2,...,n−1} for each
n ∈ N − {1};
(ii) if P is a family of subsets of X, then P n = {P
⊂ P : |P
| = n} for each
n ∈ N.
Let R1 = {U }. For each m ∈ N, suppose Rm is a countable family of open covers
of X. For each n ∈ N, let

F1 = ∅;
Fn = {f : f is a function from {1, 2, . . . , n − 1} to N}, n > 1;

F= Fn .
n∈N

Fix H ∈ Rm . Below we define a family U (H , f ) of open sets in X inductively for

each f ∈ F.
First, let U (H , ∅) = ∅. Suppose we have defined a family U (H , f|n ) of open
sets in X for each f ∈ Fn+1 . For each V ∈ H n+1 , let

U(V , f ) = ∪{Uf (n),y : V = (H )y , y ∈

/ ∪U (H , f|n )},
U (H , f ) = U (H , f|n ) ∪ {U(V , f ) : V ∈ H n+1 },

and let
214 3 Generalized Metric Spaces

Am = {A ⊂ Rm × F : |A| < ℵ0 },
Bm = Am × N × N.

For each A ∈ Am , define H (A) = ∪{U (H , f ) : (H , f ) ∈ A}. For each B ∈ Bm ,

denote B = (A, i, j) and define

H (B) = {H(B, β)}β<γ ,

where
H(B, β) = ∪{Uj,y : y ∈
/ ∪H (A), n(y)  i, β(y) = β}.

For every E ∈ (Am ∪ Bm )<ω and k ∈ N, let

H (E ) = ∪{H (E) : E ∈ E },
Hk (E ) = H (E ) ∪ {Uα ∩ U : α < γ , U ∈ Uk and U ⊂ ∪H (E )},

and define
Rm+1 = Rm ∪ {Hk (E ) : k ∈ N, E ∈ (Am ∪ Bm )<ω }.

Then

(1.1) Each element of Rm+1 is an open refinement of U .

By the definition of Hk (E ), each element of Rm+1 is an open cover of X. For
each H ∈ Rm , take n ∈ N such that f ∈ Fn+1 and V ∈ H n+1 . Then there
is α < γ such that ∩V ⊂ Uα , so U(V , f ) ⊂ Uα . Since H(B, β) ⊂ Uβ for
every B ∈ Bm and β < γ , Hk (E ) refines U .
Let R = m∈N Rm .
(1.2) For every x ∈ X and H ∈ R, there is a function f : N → N such that x is
only in finitely many elements of U (H , f|n ) for each n ∈ N.
We define f inductively. Take f|1 = ∅. Suppose we have defined f|n satisfying
our requirement for some n ∈ N. For each i ∈ N, take fi ∈ Fn+1 such that fi|n =
f|n and fi (n) = i. If there is a subfamily {Vi }i∈N of different elements of H n+1
such that x ∈ U(Vi , fi ), then by the Δ-system lemma (see the proof of Example
2.5.18), there exist an infinite subset J of N and D ⊂ H such that Vi ∩ Vj = D
for every i, j ∈ J with i = j. For each i ∈ J, there is xi ∈ / ∪U (H , f|n ) such
that Vi = (H )xi and xi ∈ Ui,xi . So xi ∈ st(x, Ui ), it follows that the sequence
{xi } has an accumulation point p ∈ Cx . Let V = (H )p . If |V |  n + 2, let V

be a set with n + 2 elements of V , then there is i ∈ J such that xi ∈ ∩V

, so
V
⊂ Vi , a contradiction. Therefore, |V |  n + 1. Pick i, j ∈ J with i = j such
that xi , xj ∈ ∩V . Then V ⊂ Vi ∩ Vj , so |V |  n, and hence p ∈ ∪U (H , f|n ),
it follows that there is k ∈ J such that xk ∈ ∪U (H , f|n ), a contradiction.
For every x ∈ X and i ∈ N, define
3.6 Developable Spaces and p-Spaces 215

Sx = {y ∈ X : α(x)  α(y)  β(y) < β(x)},

Xi = {y ∈ X : n(y)  i}.

Then
(1.3) Sx ∩ Xi ⊂ Sx .
Let z ∈ Sx ∩ Xi . Then α(y)  α(x) for each y ∈ Sx , and hence α(z)  α(x).
If β(z)  β(x), take y ∈ Ui,z ∩ Sx ∩ Xi , then β(y) < β(x) and n(y)  i, so
st(y, Ui ) ⊂ Uβ(x) . By the compactness of Cz , we can pick p ∈ Cz such that
α(p) = β(z). Fix U ∈ Ui,z . Then p ∈ U ⊂ st(y, Ui ) ⊂ Uβ(y) , and hence
α(p)  β(y) < β(x)  β(z) = α(p), a contradiction. Thus β(z) < β(x), so
z ∈ Sx .
Below we turn to prove that R is a θ -sequence in X. Otherwise, define

Y = {y ∈ X : there is H ∈ R such that |(H )y | < ℵ0 }.

Then X−Y = ∅, so there is x ∈ X−Y , such that β(x) = min{β(z) : z ∈ X−Y }.

Thus for each y ∈ Sx , there exist n ∈ N and H ∈ R such that |(H )y | = n. For
these x and H , there is a function F satisfying (1.2). Then y ∈ ∪U (H , f|n )
and x is only in finitely many elements of U (H , f|n ). For every i ∈ N and
the compact set Cx ∩ Sx ∩ Xi ⊂ Sx in X, by the above argument, there exist
i , hi ∈ N, {Hk : k  hi } ⊂ Rmi and {fk : k  hi } ⊂ F, such that,
m
khi U (Hk , fk ) covers Cx ∩ Sx ∩ Xi and x is in only finitely many elements
of this cover. Let Ai = {(Hk , fk ) : k  hi }. Then
(1.4) There is ji ∈ N such that x is only in finitely many elements of H (Bi ), where
Bi = (Ai , i, ji ).
Otherwise, for each j ∈ N, there exist βj < γ and xj ∈ / ∪H (Ai ) such that
βj < βj+1 , β(xj ) = βj , x ∈ Uj,xj and n(xj )  i. So xj ∈ st(x, Uj ), and hence
{xj } has an accumulation point z in Cx . On the one hand, x ∈ Uj,xj ⊂ Uα(xj ) , and
hence α(xj )  α(x). On the other hand, if β(xj )  β(x), pick m > j+n(x), then
β(xm ) > β(xj )  β(x), so there is U ∈ Um,xm such that U ⊂ Uβ(x) . However,
by n(x) < m and x ∈ U ∈ Um , U ⊂ st(x, Um ) ⊂ Uβ(x) , a contradiction. Hence
β(xj ) < β(x). Thus, xj ∈ Sx ∩ Xi . Consequently, z ∈ Cx ∩ Sx ∩ Xi ⊂ ∪H (Ai ),
it follows that there is k ∈ N such that xk ∈ ∪H (Ai ), a contradiction.
Note that if y ∈ Cx and i > n(y), then y ∈∪(H (Ai ) ∪ H (Bi )). By the
compactness of Cx , there is I ∈ N<ω such that i∈I (H (Ai ) ∪ H (Bi )) covers
Cx and x is only in finitely many elements of this cover. Let

m = max{mi : i ∈ I},
E = {Ai : i ∈ I} ∪ {Bi : i ∈ I}.

Then H (E ) covers Cx and x is only in finitely many elements of this cover. Take
k ∈ N such that st(x, Uk ) ⊂ ∪H (E ). Then Hk (E ) ∈ Rm+1 and |(Hk (E ))x | <
ℵ0 , a contradiction. So R is a θ -sequence in X.
In summary, every well-monotone open cover of X has an open θ -refinable
216 3 Generalized Metric Spaces

sequence, and hence X is a submetacompact space (see Theorem A.4.8 in

Appendix A). 

By using the technique in the above proof, Kemoto and Yajima [229] proved that a
β-space is a submetacompact space if and only if every well-monotone open cover of
it has a σ -closure-preserving closed refinement. They also constructed an example to
show there is a normal orthocompact space which is not submetacompact, such that,
every well-monotone open cover of it has a closure-preserving closed refinement.

Corollary 3.6.2 For every space X, the following are equivalent:

(1) X is a developable space.
(2) X is a strict p-space with a point-countable pseudo-k-network [455].
(3) X is a strict p-space with a Gδ -diagonal [235].

Proof By Corollary 3.1.4, we get (1) ⇒ (2). By (∗) of Theorem 3.1.8 and Theorem
3.6.1, we have (2) ⇒ (3). (3) ⇒ (1) is obtained by Theorem 3.6.1, Proposition 1.5.20
and Theorem 1.7.7. 

Corollary 3.6.3 For every regular space X, the following are equivalent:
(1) X has a strict p-sequence.
(2) X is both a strong Σ -space and a wΔ-space [217].
(3) X is a submetacompact wΔ-space.
(4) X is a submetacompact space with a p-sequence.

Proof By Theorems 3.2.7 and 2.2.18, we only need to prove (1) ⇒ (2). Suppose
{Un } is a strict p-sequence in X. For each n ∈ N, by the submetacompactness of X,
there is a σ -closure-preserving closed cover Fn of X such that for each x ∈ X, there
is F ∈ Fn satisfying x ∈ F ⊂ st(x, Un ) (see Theorem A.4.8 in Appendix A). Let
  <ω 

F = ∩F
: F
∈ Fn .
n∈N

Then F is a σ -closure-preserving family of closed sets in X. For each x ∈ X, let

Kx = ∩(F )x , Cx = st(x, Un ).
n∈N

Then x ∈ Kx ⊂ Cx , and hence Kx ∈ K (X). If Kx ⊂ U ∈ τ , then

{U} ∪ {X − F : F ∈ (F )x }

is an open cover of X, so there is F

∈ (F )<ω

x such that Cx ⊂ U ∪ (X − ∩F ), and

hence there is n ∈ N such that st(x, Un ) ⊂ U ∪ (X − ∩F ). Take F ∈ Fn such that

3.6 Developable Spaces and p-Spaces 217

x ∈ F ⊂ st(x, Un ). Let F

= F
∪ {F}. Then ∩F

∈ F and Kx ⊂ ∩F

⊂ U. So
F is a (mod k)-network for X w.r.t. {Kx : x ∈ X}. Thus X is a strong Σ -space.


Theorem 3.6.4 ([81]) For every regular space X, the following are equivalent:
(1) X is a developable space.
(2) X is a semi-stratifiable space with a p-sequence [110].
(3) X is a symmetrizable space with a p-sequence [91].

Proof By Corollaries 3.6.3, 3.6.2 and Proposition 1.4.11 we know (2) ⇒ (1). (1) ⇒
(3) follows from Theorem 1.2.8. To complete the proof, we prove the following claim
first:
Claim. Every countably compact subspace of a symmetric space is compact
metrizable.
Suppose C is a countably compact set in X. By Theorem 1.2.11, C is a D-space.
If U is an open cover of C, then there is a neighborhood assignment φ of C such
that for each x ∈ C, φ(x) ∈ U . Since C is a D-space, there is a closed discrete
subset D of C such that C = x∈D φ(x). Since C is a countably compact space,
D is a finite set, so {φ(x) : x ∈ D} is a finite subcover of U . It follows that C is
compact, so C is a closed subspace of X, and hence C is also a symmetric space. By
Theorem 1.2.5, C is a hereditarily Lindelöf space. For each x ∈ C, the open cover
{V ∈ τ (C) : x ∈ / V } of C − {x} has acountable subcover, denoted as {Vn : n ∈ N}.
Then C − {x} = n∈N Vn , i.e. {x} = n∈N (C − Vn ). That means C has the point Gδ -
property. By Theorem 1.7.7(1), C is first countable. By Theorem 1.2.8, Corollaries
1.4.5 and 1.4.12, C is metrizable.
Now we prove (3) ⇒ (2). Suppose X is a symmetrizable space with a p-sequence.
Let {U n } be the p-sequence of X. For every x ∈ X and n ∈ N, pick Un ∈ (Un )x . Then
Dx = n∈N Un is a compact subset of X. By the above Claim, Dx is a metrizable  subset
of
 X, so there is a sequence {W n } of open sets in X such that {x} = Dx ∩ n∈N Wn =
n∈N (Un ∩ Wn ). That means X has the point Gδ -property. By Proposition 2.4.10
and Theorem 1.7.7(1), X is first countable. Further by Theorem 1.2.8 and Corollary
1.4.5, X is a semi-stratifiable space. 

Theorem 3.6.5 ([186]) A space X is metrizable if and only if X is a monotonically

normal p-space with a Gδ -diagonal.

Proof It is enough to show a monotonically normal p-space with a Gδ -diagonal is

semi-stratifiable by Theorems 1.4.14, 3.6.4 and the Bing metrization criterion (see
Theorem 1.3.5). Because X is a p-space with a G δ -diagonal, there is a sequence {Un }
of covers of X by open subsets of βX such that n∈N st(x, Un ) = {x} for each x ∈ X,
and Un+1 refines Un for each n ∈ N. For each y ∈ X and each n ∈ N, let Gn (y) be an
open subset of X such that (i) y ∈ Gn (y) and (ii) Gn (y) is a subset of some member of
Un . Define g(n, y) = D({y}, X − Gn (y)), where D is a monotone normality operator
for X in which D(H, K) ∩ D(K, H) = ∅ for each pair H, K of disjoint closed subsets
of X. Then g(n, y) ⊂ Gn (y).
218 3 Generalized Metric Spaces

Suppose {xn } is a sequence in X such that y ∈ g(n, xn ) for some y ∈ X and each
n ∈ N. To show that {xn } converges to y, it suffices to show that any subsequence
{xnk } of {xn }, no term of which is y, has y as its unique accumulation point. Fix k ∈ N.
Because xnk = y, there is jk ∈ N such that if j  jk , then xnk ∈ / st(y, Unj ) ⊃ Gnj (xnj ).
If there were an integer j  jk such that xnj ∈ / Gnk (xnk ), then it would follow that

g(nj , xnj ) = D({xnj }, X − Gnj (xnj )) ⊂ D({xnj }, {xnk })

⊂ X − D({xnk }, {xnj }) ⊂ X − D({xnk }, X − Gnk (xnk ))
= X − g(nk , xnk ),

contradicting y ∈ g(nj , xnj ) ∩ g(nk , xnk ). Therefore, xnj ∈ Gnk (xnk ) whenever j  jk .
Because βX is compact, the sequence {xnk } has an accumulation point, say z, in
βX. For each k ∈ N, z ∈ {xnj : j  jk } ⊂ Gnk (xnk ) ⊂ st(y, Unk ), the last inclusion
justified by conditions (i) and (ii) above, plus the fact that y ∈ Gnk (xnk ). Hence,
being
z ∈ k∈N st(y, Unk ) = {y}, as required. 

How to characterize p-spaces with a Gδ -diagonal? Alleche, Arhangel’skiı̌ and

Calbrix [10] introduced the following weakly developable spaces (which are different
from the weak developments defined in Definition 1.6.13): A space X is called a
weakly developable  space if there exists a sequence {Un } of open
 covers of X such
that for each x ∈ i∈N Ui with each Ui ∈ Ui , the collection { in Ui : n ∈ N} is a
local base at x. It is showed that a completely regular space is weakly developable if
and only if it is a p-space with a Gδ -diagonal [10].
In Sect. 1.5, we characterized semi-stratifiable spaces in terms of pair-networks.
In the following, we give characterizations of developable spaces by means of pair-
networks. In order to facilitate the use of the concept of pair-networks, we make the
following conventions: If P is a family of pairs of sets in X, then for each P ∈ P,
P
and P

represent its first component and second component respectively. Denote

P
= {P
: P ∈ P};
P

= {P

: P ∈ P};
st(x, P) = ∪{P

: P ∈ P, x ∈ P
}, x ∈ X;
st(A, P) = ∪{P

: P ∈ P, A ∩ P
= ∅}, A ⊂ X.

For subfamilies R1 , . . . , Rn of P, denote

R1 ∧ · · · ∧ Rn = {( Pi
, Pi

) : Pi ∈ Ri , i  n}.
in in

Theorem 3.6.6 ([81]) For every space X, the following are equivalent:

(1) X is a developable space.

(2) There is a pair-network n∈N Pn for X such that
3.6 Developable Spaces and p-Spaces 219

(i) Pn
is a locally finite family of closed sets in X and Pn

is a family of open
sets in X;
(ii) for every compact set K and U ∈ τ in X with K ⊂ U, there is m ∈ N such
that K ⊂ st(K, Pm ) ⊂ U.

(3) There is a pair-network n∈N Pn for X such that
(i) Pn
is a locally finite family of closed sets in X;
(ii) if x ∈ U ∈ τ , then there is m ∈ N such that x ∈ st(x, Pm )◦ ⊂ U.

Proof (1) ⇒ (2). Suppose {Un } is a development for X and Un+1 refines Un . For
each n ∈ N, let Un= {Uα }α∈Λn . By the subparacompactness of X, Un has a closed
refinement Fn = k∈N Fn,k , such that, Fn,k = {Fα,k }α∈Λn is a discrete family in X
and Fα,k ⊂ Uα . Let
Pn,k = {(Fα,k , Uα )}α∈Λn .

Then n,k∈N Pn,k is a pair-network for X. For every finite sequence {k1 , . . . , kn }
consisting of members of N, define

H (k1 , . . . , kn ) = ∧{Pi,ki : i  n}.

For any compact set K ⊂ U ∈ τ and x ∈ K, take a finite sequence {ki } consisting of
members of N such that x ∈ ∪Fi,ki . For each n ∈ N, define

An = ∪{H
: H ∈ H (k1 , . . . , kn ), H
∩ K = ∅ and H

⊂ U}.

Then {An } is a decreasing sequence of closed sets in X, and

 hence there is m ∈ N,
such that Am = ∅. Because otherwise, there exist y ∈ K ∩ ( n∈N An ) and j ∈ N such
that st(y, Uj ) ⊂ U, so

st(y, H (k1 , . . . , kj )) ⊂ st(y, Uj ) ⊂ U,

which contradicts the definition of Aj . Thus x ∈ st(K, H (k1 , . . . , km )) ⊂ U. By the

compactness of K, ∪{H (k1 , . . . , kn ) : ki ∈ N, i  n ∈ N} satisfies the requirement
of (2). 
(2) ⇒ (3) is obvious. We prove (3) ⇒ (1). Suppose n∈N Pn is a pair-network
in X satisfying the requirement of (3). For every n, k ∈ N, define

φn,k = {F ⊂ Pn
: |F | = k}.

For each F ∈ φn,k , let

U(F ) = int(∪{P

: P ∈ Pn , P
∈ F }) − ∪(Pn
− F ),
Un,k = {U(F ) : F ∈ φn,k }.
220 3 Generalized Metric Spaces

Then {Un,k } is a quasi-development for X. Because in fact, for every x ∈ U ∈ τ ,

there is m ∈ N such that x ∈ st(x, Pm )◦ ⊂ U. Let

F = {P
: P ∈ Pm , x ∈ P
} and k = |F |.

Then F ∈ φm,k , so x ∈ U(F ) ⊂ st(x, Pm )◦ ⊂ U. If E ∈ φm,k − {F }, then

x ∈ ∪(Pm
− E ), and hence x ∈ / U(E ). That shows U(F ) is the only element of
Um,k containing
 x, so x ∈ st(x, Um,k ) ⊂ U. Thus X is a quasi-developable space.
While n∈N Pn
is a σ -locally finite closed network for X, so X is a perfect space.
By Theorem 1.2.13, X is a developable space. 

Corollary 3.6.7
 ([289]) A space X is developable space if and only if there is a
pair-network n∈N Pn for X such that
(i) Pn
is an HCP family of closed sets in X;
(ii) if x ∈ U ∈ τ , then there is m ∈ N such that x ∈ st(x, Pm )◦ ⊂ U.

Proof We only need to prove the sufficiency. Suppose n∈N Pn is a pair-network
for X satisfying conditions (i) and (ii). By (ii), X is a first countable space. For each
n ∈ N, let

Rn = {(P
− Dn , P

) : P ∈ Pn } ∪ {({x}, st(x, Pn )) : x ∈ Dn },

where Dn = {x ∈ X : |(Pn
)x |  ℵ0 }. By Lemma 3.2.17, Rn
is a locally finite family
in X such that st(x, Rn ) = st(x, Pn ) for each x ∈ X, and hence the
of closed sets
pair-network n∈N Rn for X satisfying the assumption of Theorem 3.6.6(3), so X is
a developable space. 

Question 3.6.8 ([348]) Characterize closed images of developable spaces in terms

of σ -HCP pair-networks.
In the following, we investigate some deeper properties of metacompact devel-
opable spaces introduced in Sect. 2.10. If W is a family of open sets in X satisfying
(G) (resp. uniform (G)) in Definition 3.3.3, then we said that W satisfies open (G)
(resp. open uniform (G)).

Theorem 3.6.9 ([355]) A space is a meta-Lindelöf developable space if and only if

it is a semi-stratifiable space satisfying open (G).

Proof Suppose X is a meta-Lindelöf developable space. Take a point-countable

development {Vn } for X such that Vn+1 refines Vn . Let

Wx = (Vn )x , x ∈ X;
n∈N
W = {Wx : x ∈ X}.
3.6 Developable Spaces and p-Spaces 221

Then W satisfies open (G). Because in fact, if x ∈ U ∈ τ , then there is n ∈ N such

that st(x, Vn ) ⊂ U. Fix V ∈ (Vn )x . If y ∈ V , then V ∈ Wy and x ∈ V ⊂ U.
Conversely, suppose W = {Wx : x ∈ X} is a family in a semi-stratifiable space X
satisfying open (G). Denote Wx = {W (n, x)}n∈N . By Lemma 3.3.4, X is a meta-
Lindelöf space. Let g be a semi-stratifiable function on X. For each n ∈ N, the open
cover {g(n, x) : x ∈ X} of X has a point-countable open refinement Un . For each
U ∈ Un , take xU ∈ X such that U ⊂ g(n, xU ). Define

Bn,m = {U ∩ W (m, xU ) : U ∈ Un }, m ∈ N;

B= Bn,m .
n,m∈N

Then B is a point-countable family of open sets in X. For every x ∈ O ∈ τ , take

Un ∈ (Un )x (∀n ∈ N) and denote xn = xUn . Then x ∈ g(n, xn ), so xn → x, and
hence there exist i, m ∈ N such that xi ∈ V (x, O) and x ∈ W (m, xi ), it follows that
x ∈ Ui ∩ W (m, xi ) ⊂ O. Thus B is a point-countable base for X. By Corollary 1.4.5
and Theorem 1.2.14, X is a developable space. 

Question 3.6.10 ([107]) Does every space satisfying open (G) have a point-countable
base?
The concept of uniform bases is relevant to that of open uniform (G).

Theorem 3.6.11 For every space X, the following are equivalent:

(1) X is a metacompact developable space.
(2) X has a uniform base [177].
(3) X satisfies open uniform (G) [355].

Proof (1) ⇒ (2). Suppose X is a metacompact developable space.Take a point-finite

development {Bn } for X such that Bn+1 refines Bn . Let B = n∈N Bn . Then B
is a base for X. If x ∈ U ∈ τ and B
is an infinite subfamily of (B)x , then there is
m ∈ N such that st(x, Bm ) ⊂ U. By the point-finiteness of Bm , there exist n > m
and B ∈ B
∩ Bn . Then B ⊂ st(x, Bm ) ⊂ U. So X has a uniform base.
(2) ⇒ (3). Suppose B is a uniform base for X. If B is point-countable, then it is
easy to verify that W = {(B)x : x ∈ X} is a family satisfying open uniform (G) in X.
If (B)x is uncountable for some x ∈ X, then {B ∈ (B)x : z ∈ B} is a finite family for
every z ∈ X − {x}, so there exist an infinite subfamily {Bn }n∈N of (B)x , xn ∈ Bn − {x}
and k ∈ N such that xn is just in k elements of (B)x . Then xn → x. Since B is a base
for X, there exist a subsequence {Bi
} of (B)x and a subsequence {xni } of {xn } such
that {xnj : j  i} ⊂ Bi
⊂ X − {xnj : j < i} whenever i ∈ N, it follows that xni is in at
least i elements of (B)x , a contradiction.
(3) ⇒ (1). By Lemma 3.3.6, Theorem 3.6.9 and Lemma 3.3.4, every space satis-
fying open uniform (G) is a metacompact developable space. 
222 3 Generalized Metric Spaces

Corollary 3.6.12 The property of having a uniform base is additive, hereditary and
countably productive.
Similar to Definition 2.10.12, one can define an open uniform (G) at non-isolated
points [248], and it is true that a space X has an open uniform (G) at non-isolated
points if and only if it has a uniform base at non-isolated points [248].
In the second part of this section, we discuss the mapping properties of wΔ-spaces,
p-spaces, developable spaces and the spaces with a uniform base.
Theorem 3.6.13 ([478]) Suppose f : X → Y is a closed mapping and X is a
developable space (resp. space with a uniform base). If one of the following conditions
holds, then Y is a developable space (resp. space with a uniform base).
(1) f is a compact mapping.
(2) X is a regular space and Y is a first countable space.

Proof (1) Suppose f is a perfect mapping. Let n∈N Pn be a pair-network for X
satisfying the assumption of Theorem 3.6.6(2). For each n ∈ N, define

Rn = {(f (P
), f (P

)) : P ∈ Pn }.

Then n∈N Rn is a pair-network for Y satisfying the assumption of Theorem 3.6.6(3),
and hence Y is a developable space.
(2) Suppose X is a regular space and Y is a first countable space. By Theorem
3.3.12, there is an Fσ -discretesubspace Z of Y such that f −1 (y) is compact in X
whenever y ∈ Y − Z. Let Z = n∈N Zn , where each Zn is a discrete closed subspace
of Y . For each y ∈ Z, suppose {U(y, n)}n∈N is a countable local base of y in Y . Let

n∈N Pn be a pair-network for X satisfying the assumption of Theorem 3.6.6(2).
For every n, j ∈ N, define

Rn = {(f (P
), f (P

)) : P ∈ Pn },
Hn,j = {({y}, U(y, j)) : y ∈ Zn }.
 
Then ( n∈N Rn ) ∪ ( n,j∈N Hn,j ) is a pair-network for Y satisfying the assumption
of Corollary 3.6.7(i). Let y ∈ U ∈ τ (Y ). If y ∈ Z, then there exist n, j ∈ N such that
y ∈ Zn and U(y, j) ⊂ U, so y ∈ st(y, Hn,j )◦ ⊂ U(y, j) ⊂ U. If y ∈ Y − Z, then
f −1 (y) is a compact set in X, and hence there is m ∈ N such that

f −1 (y) ⊂ st(f −1 (y), Pm ) ⊂ f −1 (U),

it follows that y ∈ st(y, Rm )◦ ⊂ st(y, Rm ) ⊂ U. Consequently

  ⎛ ⎞
 
Rn ∪ ⎝ Hn,j ⎠
n∈N n,j∈N

satisfies the assumption of Corollary 3.6.7(ii). Thus Y is a developable space.

3.6 Developable Spaces and p-Spaces 223

By Theorem 3.6.11, for the case of uniform bases, the conclusion is also true.


Moreover, Theorem 3.6.13 is still true if we change “a developable space (resp.

space with a uniform base)” into “a developable space at non-isolated points (resp.
space with a uniform base at non-isolated points)” [250]. Here, a space X is said
to be a developable space at non-isolated points if there exists a sequence {Pn } of
open subsets of X such that {st(x, Pn )}n∈N is a neighborhood base at x in X for each
non-isolated point x ∈ X.

Question 3.6.14 ([82, 156]) Are wΔ-spaces preserved by perfect mappings?

Proposition 3.6.15 ([82]) Every quasi-perfect preimage of a wΔ-space is also a

wΔ-space.

Proof Suppose f : X → Y is quasi-perfect mapping and Y is a wΔ-space. Let

{Un } be a wΔ-sequence in Y . Suppose x ∈ X and {xn } is a sequence in X satisfying
xn ∈ st(x, f −1 (Un )). Then f (xn ) ∈ st(f (x), Un ), and hence the sequence {f (xn )}
has an accumulation point. Since f is a quasi-perfect mapping, {xn } also has an
accumulation point. Thus {f −1 (Un )} is a wΔ-sequence in X, so X is a wΔ-space.


The following corollary is obtained by Propositions 3.6.15, 3.3.16 and Theorems

1.7.7, 3.6.11.

Corollary 3.6.16 ([43, 318]) Suppose f : X → Y is a perfect mapping and X has a

Gδ -diagonal. If Y is a developable space (resp. space with a uniform base), then X
is a developable space (resp. space with a uniform base).

Every Moore space satisfies the decomposition theorem of type compactness.

However, whether every strict p-space satisfies the decomposition theorem of type
compactness is still an open problem.

Question 3.6.17 ([97]) Does every perfect preimage of a Moore space satisfy the
decomposition theorem of type compactness?
The following Example 3.6.18 shows the relationships between developable
spaces and relevant spaces.

Example 3.6.18 (1) There is a strict p-space which is not a subparacompact space,
the space X in Example 1.8.10 is such a space.
(2) There is a locally compact quasi-developable space which is not a β-space, the
space ψ(D) in Example 1.8.4(6) is such a space.
(3) Burke [79] constructed a locally compact space with a Gδ -diagonal which is not
a wΔ-space.
(4) Under CH, Alster, Burke and Davis [12] constructed a locally compact wΔ-space
with a Gδ -diagonal which is not a strict p-space.
224 3 Generalized Metric Spaces

(5) Good, Knight and Mohamad [159] constructed a completely regular pseudo-
compact space with a sharp base which has no G∗δ -diagonal, where a base B for
 if, for each x ∈ X and any sequence {Bn } of
a space X is called a sharp base
different elements of (B)x , { in Bi }n∈N is a local base of x. Every space with
a uniform base has a sharp base, and every space with a sharp base has both a
Gδ -diagonal and a point-countable base.

Example 3.6.19 On the mapping properties of developable spaces and relevant

spaces.
(1) There are a locally compact developable space X and a closed mapping f :
X → Y , such that, Y is a first countable space and f is not a peripherally compact
mapping.
For the closed mapping f : ψ(N) → S1 defined in Example 2.6.4, ∂f −1 (0) =
ψ(N) − N is not compact in ψ(N). This example shows that different of the metric
space case, Theorem 3.6.13(2) can not be obtained by peripherally compact mapping.
(2) The perfect preimages of developable spaces are not preserved by perfect
mappings [96]: There exist a locally compact developable space S, perfect mappings
g : X → S and f : X → Y , such that, Y is not a perfect preimage of any space of the
point Gδ -property (hence not a perfect preimage of a developable space).
Let S be the Mrówka space ψ(N) (see Example 1.8.4). Then N is dense in S and
A is an uncountable discrete subset of S. Take a subspace of A with cardinality
ℵ1 , and denote it by ω1 . Let Z = ω1 ∪ {z} be the one point compactification of the
discrete space ω1 . Set X = S × Z. Let g = π1 : X → S. Then g is a perfect mapping.
For each α ∈ ω1 , denote the quotient space of X obtained by identifying two points
(α, α) and (α + 1, z) to one point as Y . Let f : X → Y be the quotient mapping.
Then f is a perfect mapping.
Suppose there is a perfect mapping h : Y → T and T has the point Gδ -property. For
each n ∈ N, define fn = h◦f|{n}×Z : {n}×Z → h({n}×Z). Since {h(n, z)} is a Gδ -set in
T , {n}×Z −fn−1 (h(n, z)) is countable, so there is βn ∈ ω1 such that fn (n, α) = h(n, z)
whenever βn  α < ω1 . Hence there is β ∈ ω1 such that fn (n, α) = h(n, z) when
β  α < ω1 and n ∈ N, which means h(n, α) = h(n, z). For each k ∈ N, let
yk = f (β + k, z). For α = β + k, take a sequence {ni } in N such that ni → α. Then
(ni , α) → (α, α) in X, so h(ni , α) → h(f (α, α)) = h(f (α + 1, z)) = yk+1 in T .
On the other hand, h(ni , α) = h(ni , z), so h(ni , α) → h(f (α, z)) = yk , and hence
h−1 (h(y1 )) contains an infinite discrete closed set {yk : k ∈ N}, which contradicts
the assumption that h is a perfect mapping. Thus Y is not a perfect preimage of any
space of the point Gδ -property.
(3) There is a quasi-developable p-space X which does not satisfy the decompo-
sition theorem of type compactness [95].
Suppose Y is the Mrówka space ψ(D) with |D|  ℵ1 (see Example 1.8.4). Take
X = Y × S1 − (D × {0}). Let f = π1|X : X → Y . Then f is a closed mapping.
For each y ∈ D, f −1 (y) is not compact in X and D is not an Fσ -discrete subspace of
Y (every closed subset D of Y is a finite set), so X does not satisfy the decomposition
theorem of type compactness.
3.6 Developable Spaces and p-Spaces 225

Since both Y and S1 are quasi-developable spaces, X is also a quasi-developable

space. Using the notations in Example 1.8.4, for each n ∈ N, let

Un = {{Aα } ∪ {x(α, m) : m  n} : α ∈ Λ} ∪ {{x} : x ∈ En },

where
En = D − {x(α, m) : α ∈ Λ, m  n}.

Then {Un } is a p-sequence in Y . Let

Vn = {X ∩ (U × ({0} ∪ {1/m : m  n})) : U ∈ Un }∪

 

{U × {1/m} : U ∈ Un } .
m∈N

Then {Vn } is a p-sequence in X, and hence X is a p-space.

(4) p-spaces are not preserved by perfect mappings [95].
Continue to using the structures in (3). For each z ∈ Y − D, denote the quotient
space of X by identifying the subset {z} × S1 of X to one point as H. Let f : X → H
be the quotient mapping. Then f is a perfect mapping. If H is a p-space, let {Wn }
be a p-sequence in H. For every z ∈ Y − D and n ∈ N, there exist U(n, z) ∈ τ (Y )
and W (n, z) ∈ Wn such that {z} × S1 ⊂ X ∩ (U(n,  z) × S1 ) ⊂ f −1 (W (n, z)). Since
Y − D is not a Gδ -set in Y , there is d ∈ D ∩  ( n∈N ( z∈Y −D U(n, z))), and hence
there
 is a sequence {z n } in Y − D such that d ∈ n∈N U(n, zn ). So {d} × (S1 − {0}) ⊂
n∈N W (n, zn ), and hence {d} × (S1 − {0}) is compact in H, a contradiction. Thus
H is not a p-space.
Developable spaces are not preserved by finite-to-one pseudo-open mappings
[461]. However, the following question is still open.

Question 3.6.20 ([461]) Is every finite-to-one pseudo-open image of a Moore space

a developable space?

3.7 M-Spaces

In this section, we investigate the properties of M-spaces and paracompact M-

spaces or paracompact p-spaces in two ways. First, as perfect preimages of metric
spaces, are there any theorems for characterizing paracompact M-spaces similar to
the Nagata-Smirnov metrization theorem? By presenting the concept of (mod k)-
bases defined by Michael [333], we introduce the Michael-Lutzer theorem, discuss
several generalizations of this theorem and obtained a theorem for characterizing
paracompact M-spaces parallel to the Burke–Engelking–Lutzer metrization theo-
rem. Second, corresponding to the wΔ-conditions in M-spaces, we investigate the
spaces with a sequence of locally finite or closure-preserving closed covers satisfying
226 3 Generalized Metric Spaces

wΔ-condition, and derive the following Morita–Rishel theorem: A space is a perfect

image of an M-space if and only if it is an M ∗ -space.

Definition 3.7.1 ([333]) A family P of sets in a space X is called a (mod k)-base

(resp. quasi-(mod k)-base) for X if P is an open (mod k)-network (resp. open quasi-
(mod k)-network). A regular space with a σ -locally finite (mod k)-base (resp. quasi-
(mod k)-base) is called a (mod k)-metrizable space (resp. quasi-(mod k)-metrizable
space).
Theorem 3.7.2 ([333]) (The Michael-Lutzer theorem) A space X is a paracompact
M-space if and only if X is a (mod k)-metrizable space.
Proof The necessity is obtained by the Nagata-Smirnov metrization theorem and
Corollary 2.2.7.
The sufficiency. Suppose P is a σ -locally finite (mod k)-base for X w.r.t. K .
(1) X is a paracompact space.
Denote K = {Kα }α∈Λ . Suppose U is an open cover of X. For each α ∈ Λ, there is
<ω
Uα ∈ U  such that Kα ⊂ ∪Uα , so there is Pα ∈ P such that Kα ⊂ Pα ⊂ ∪Uα , and
hence α∈Λ {Pα } ∧ Uα is a σ -locally finite open refinement of U . By the regularity
of X, X is a paracompact space.
(2) X is a wΔ-space.

Denote P = n∈N Pn , where each Pn is locally finite and we may assume
X ∈ Pn . For every n ∈ N and P ∈ P, define Fn (P) = ∪{Q ⊂ P : Q ∈ Pn }. Then
Fn (P) is a closed set in X and Fn (P) ⊂ P. For every n, k ∈ N, let

Fn,k = {Fn (P) : P ∈ Pk },

Rn,k = {Rn,k (x) : x ∈ X},

where

Rn,k (x) = ∩{P ∈ Pk : x ∈ Fn (P)} − ∪{Fn (P) : P ∈ Pk , x ∈

/ Fn (P)}.

Since Fn,k is a closure-preserving family of closed sets in X and Pk is a point-finite

family of open sets in X, x ∈ Rn,k (x) ∈ τ for each x ∈ X. So Rn,k is an open cover
of X. Note that, for each P ∈ Pk , st(x, Rn,k ) ⊂ P whenever x ∈ Fn (P). Because in
fact, if x ∈ Rn,k (y), then y ∈ Fn (P), so Rn,k (y) ⊂ P, and hence st(x, Rn,k ) ⊂ P. For
each i ∈ N, define
Ui = ∧{Rn,k : n, k  i}.

We prove {Ui } is a wΔ-sequence in X. For each x ∈ X and any sequence {xi } in X,

if xi ∈ st(x, Ui ) and {xi } has no accumulation point, take α ∈ Λ such that x ∈ Kα ,
then there exist m, k ∈ N and P ∈ Pk such that Kα ⊂ P ⊂ X − {xi : i  m}. Further
take n ∈ N and Q ∈ Pn such that Kα ⊂ Q ⊂ Q ⊂ P. Then x ∈ Kα ⊂ Fn (P). Pick
j  max{k, m, n}. Then xj ∈ st(x, Uj ) ⊂ st(x, Rn,k ) ⊂ P ⊂ X −{xj }, a contradiction.
Thus X is a wΔ-space.
3.7 M-Spaces 227

By Corollary 2.2.19, X is a paracompact M-space. 

Now we turn to investigate the HCP family case.

Lemma 3.7.3 ([270]) Suppose P is an HCP family of open sets in a q-space X. If

x ∈ X d , then P is locally finite at x.

Proof Let
V = X − ∪{P : x ∈
/ P, P ∈ P}.

Then x ∈ V ∈ τ , and x ∈ P for each P ∈ (P)V . If P is not  locally finite at

x, then thereis an infinite subfamily {Pn }n∈N of P such that x ∈ n∈N Pn . Since
x ∈ X d , x ∈ n∈N Pn − {x}. Suppose g is a q-function on X. Then there is a nontrivial
sequence {xn } in X such that xn ∈ (Pn −{x})∩g(n, x), which contradicts the properties
of P and g. So P is locally finite at x. 

Lemma 3.7.4 ([270]) There is a σ -locally finite quasi-(mod k)-base for each space
with a σ -HCP quasi-(mod k)-base.

Proof Suppose P is a σ -HCP quasi-(mod k)-base in a space X w.r.t. H . We first

prove X is a q-space. Take H ∈ (H )x for each x ∈ X. We may assume H ∈ / τ . Let
P
= {P ∈ P : H ⊂ P}. Then P
is a neighborhood base of H in X. Suppose
f : X → X/H is the quotient mapping. Then f is a quasi-perfect mapping, and
hence f (P
) is a σ -HCP local base of f (H) in X/H. By Corollary 2.5.11, f (P
) is
countable, so P
is countable. Let P
= {Pi }i∈N . Then for any sequence {xi } in X,
if xi ∈ Pi , then {xi } has 
an accumulation point. So X is a q-space.
Now denote P = n∈N Pn , where Pn is HCP and Pn ⊂ Pn+1 . For each
n ∈ N, define
Ln = {x ∈ X : Pn is locally finite at x}.

Then Ln+1 ⊂ Ln ∈ τ . By Lemma 3.7.3, X d ⊂ Ln , so X − Ln is a discrete clopen

subspace of X. Define

Fn = Pn|Ln ∪ {{x} : x ∈ X − Ln }, n ∈ N.

Then n∈N Fn is a σ -locally finite family of open sets in X. Let

Z= Ln , H
= H|Z ∪ {{x} : x ∈ X − Z}.
n∈N


We prove n∈N Fn is a quasi-(mod k)-base for X w.r.t. H
. In fact, for each H
∈ H

with H
⊂ U

 ∈ τ , we may assume there is H ∈ H such that H = H ∩ Z. Then

H ⊂ U ∪ ( n∈N (X − Ln )), so there is m ∈ N such that H ⊂ U ∪ (X − Lm ), and
hence there exist i  m and P ∈ Pi such that H ⊂ P ⊂ U ∪ (X − Lm ). It follows
228 3 Generalized Metric Spaces

that H
⊂ P ∩ Li ⊂ U and P ∩ Li ∈ Fi . Thus X has a σ -locally finite quasi-(mod k)-
base. 

Theorem 3.7.5 ([270]) For every regular space X, the following are equivalent:
(1) X is a (mod k)-metrizable space.
(2) X has a σ -discrete (mod k)-base.
(3) X has a σ -HCP (mod k)-base.
(4) X has a σ -compact-finite (mod k)-base.
Proof By Theorems 3.7.2, 1.3.2 and Lemma 3.7.4, we only need to prove  (4) ⇒ (1).
Suppose P is a σ -compact-finite (mod k)-base for X w.r.t. K . Let P = n∈N Pn ,
where each Pn is compact-finite. By an argument similar to the proof of Corollary
1.3.4, Pn is a locally finite family in X. So P is a σ -locally finite (mod k)-base.
Thus X is a (mod k)-metrizable space. 

Because there is a non-metrizable M1 -space (see Example 1.8.8), a regular space

having a σ -closure-preserving (mod k)-base may not be a (mod k)-metrizable space.
In the second part of this section, we investigate properties of spaces with a
sequence of closed covers satisfying wΔ-condition.

Definition 3.7.6 A space X is called an M ∗ -space [201] (resp. M -space [426]) if,
there is a sequence {Fn } of locally finite (resp. closure-preserving) closed covers
of X satisfying wΔ-condition. The sequence {Fn } is called an M ∗ -sequence (resp.
M -sequence) in X.
Proposition 3.7.7 The following conclusions hold:
(1) Every M-space is an M ∗ -space [202].
(2) Every M ∗ -space is both an M -space and a Σ-space.
(3) Every M -space is both a wM-space and a Σ -space.
Proof (1) Suppose X is an M-space. Then there exist a metric space Y and a quasi-
perfect mapping f : X → Y . Since Y is developable and paracompact, there is a
sequence {Pn } of locally finite closed covers of Y such that if y ∈ Y and {yn } is any
sequence in Y with yn ∈ st(y, Pn ), then yn → y. Since f is a quasi-perfect mapping,
{f −1 (Pn )} is an M ∗ -sequence in X, thus X is an M ∗ -space.
(2) By Definition 3.7.6, every M ∗ -space is an M -space.
 {Fn } is a sequence of closed covers of X satisfying wΔ-condition. Define
Suppose
Pn = in Fi for each n ∈ N. Then {Pn } is also a sequence of closed covers of X
satisfying wΔ-condition. For eachx ∈ X, take a sequence {Pn } of sets inX such that
x ∈ Pn+1 ⊂ Pn ∈ Pn . Let Cx = n∈N Pn . By the convergence lemma, n∈N Pn is
a closed quasi-(mod k)-network for X w.r.t. {Cx : x ∈ X}. Therefore, every M ∗ -space
is a Σ-space and every M -space is a Σ -space.
(3) Suppose X is an M -space. Let {Fn } be an M -sequence in X such that Fn+1
refines Fn .
3.7 M-Spaces 229

Claim. For every k ∈ N, x ∈ X and any sequence {xn } in X satisfying xn ∈

stk (x, Fn ), {xn } has an accumulation point.
We prove Claim inductively on k ∈ N. When k = 1, Claim is just the definition
of an M -sequence. Assume Claim is true for some k ∈ N. If x ∈ X and {xn } is any
sequence in X with xn ∈ stk+1 (x, Fn ), then there exist a sequence {Fn } of closed sets
in X and a sequence {yn } such that xn ∈ Fn ∈ Fn and yn ∈ Fn ∩ stk (x, Fn ). By the
induction hypothesis, {yn } has an accumulation point y. Since {Fn } is a sequence of
closure-preserving closed covers of X, if we denote Vn = X − ∪{F ∈ Fn : y ∈ / F},
then y ∈ Vn ∈ τ . So there is a subsequence {yni } of {yn } such that yni ∈ Vi , and hence
there is a sequence {Pi } of closed sets in X such that Fni ⊂ Pi ∈ Fi . Since Vi ∩Pi = ∅,
y ∈ Pi , and hence xni ∈ st(y, Fi ), it follows that {xni } has an accumulation point.
Thus Claim is true.
Define
Un = {st(x, Fn )◦ : x ∈ X}, n ∈ N.

Then {Un } is a sequence of open covers of X. Since st2 (x, Un ) ⊂ st4 (x, Fn ), by
Claim, {Un } is a wM-sequence in X. So X is a wM-space. 

Question 3.7.8 (1) Is every M -space an M ∗ -space [365]?

(2) Is every wM-space an M-space [202]?
(3) Is every wM-space an M -space?
(4) Is every wM-space an Σ -space?
If the answer of above (3) or (4) is positive, then by Theorems 1.4.10, 3.2.7, 3.7.10
and 2.2.12, the answer of Question 1.7.10 is also positive, namely, every wM-space
with a Gδ -diagonal is metrizable. Below we give partial answers to Question 3.7.8.

Proposition 3.7.9 ([269]) Every space with a sequence of HCP closed covers sat-
isfying wΔ-condition is an M ∗ -space.

Proof Suppose {Fn } is a sequence of HCP closed covers of X satisfying wΔ-

condition. By Proposition 3.7.7, X is a q-space. By using the notations in Lemma
3.2.17, {R(Fn )} is a sequence of locally finite closed covers of X. Since R(Fn )
refines Fn , {R(Fn )} is an M ∗ -sequence in X. Thus X is an M ∗ -space. 

Theorem 3.7.10 ([219, 422]) For every space X, the following are equivalent:
(1) X is a paracompact M-space.
(2) X is an isocompact M -space.
(3) X is a submetacompact wM-space.

Proof Since every wM-space is a wΔ-space, we only need to prove that every space
satisfying the assumption of (2) or (3) is a paracompact space. Since every expandable
submetacompact space is a paracompact space (see Theorem A.4.10 in Appendix A),
every space satisfying the assumption of (3) is a paracompact space by Lemma 1.7.5.
230 3 Generalized Metric Spaces

Now let X be an isocompact M -space. Suppose {Fn } is a sequence of closure-

preserving closed covers of X satisfying wΔ-condition and Fn+1refines Fn . Let U
be a directed open cover of X. For each x ∈ X, define Kx = n∈N st(x, Fn ). By
the convergence lemma, Kx is a compact set in X, and hence there exist Ux ∈ U
and nx ∈ N such that Kx ⊂ st(x, Fnx ) ⊂ Ux . Let P = {st(x, Fnx ) : x ∈ X}. Since
x ∈ st(x, Fnx )◦ , P is a σ -closure-preserving closed refinement of U and P ◦ covers
X, so X is a paracompact space (see Theorem A.1.9 in Appendix A). 

By Theorem 2.2.12, we have the following corollary.

Corollary 3.7.11 Every M -space with a Gδ -diagonal is a metrizable space.

In the following, we discuss mapping theorems of M-spaces, M ∗ -spaces, M -
spaces and wM-spaces.

Theorem 3.7.12 [365, 372] (The Morita–Rishel theorem) A space Y is an M ∗ -space

if and only if Y is a perfect image of an M-space.

Proof The sufficiency. We only need to prove that M ∗ -spaces are preserved by quasi-
perfect mappings. Suppose f : X → Y is a quasi-perfect mapping and X is an M ∗ -
space. Let {Fn } be an M ∗ -sequence in X such that Fn+1 refines Fn . We prove that
every sequence {f (Fn )} of locally finite closed covers of Y is an M ∗ -sequence in Y .
Suppose y ∈ Y and {yn } is a sequence in Y such that yn ∈ st(y, f (Fn )). Then there is
sequence {xn } in X such that xn ∈ st(f −1 (y), Fn ) ∩ f −1 (yn ). Choose an ∈ f −1 (y) such
that xn ∈ st(an , Fn ). Suppose a is an accumulation point of {an } in X. Then there is
a subsequence {ani } of {an } such that ani ∈ st(a, Fi ), so xni ∈ st2 (a, Fi ), and hence
{xni } has an accumulation point in X (see the proof of Proposition 3.7.7(3)), it follows
that {yn } has an accumulation point in Y . Consequently, {f (Fn )} is an M ∗ -sequence
in Y . Thus Y is an M ∗ -space.
The necessity. Suppose {Fn } is an M ∗ -sequence in Y such that each Fn is a
closed  cover of Y which is closed under finite intersections. For each i ∈ N, let
Pi = ni Fn . Then {Pi } is also an M ∗ -sequence in Y . Denote Pi = {P α : α ∈ Λi }
and give iΛ the discrete topology for each i ∈ N. For each α = (α i ) ∈ i∈N Λi , let
Kα = i∈N Pαi . Then {Pαi }i∈N is a network of Kα in Y whenever {Pαi } is a decreasing
sequence of sets and Kα = ∅. Define

M = {α = (αi ) ∈ Λi : {Pαi } is a decreasing sequence of sets and Kα = ∅}.
i∈N

Then M is a metrizable space. Let

X = {(y, α) ∈ Y × M : y ∈ Kα },

and f = π1|X , g = π2|X . By Proposition 2.4.8, f : X → Y is a mapping and

g : X → M is a quasi-perfect mapping. So X is an M-space. We prove f is a perfect
mapping. For each y ∈ Y , f −1 (y) = {y} × {α ∈ M : y ∈ Kα }. By the point-finiteness
3.7 M-Spaces 231

of Pi , f −1 (y) is compact in X, so f is a compact mapping. For every n ∈ N and

αi ∈ Λi (i  n), let

B(α1 , . . . , αn ) = {α ∈ M : πi (α) = αi , i  n}.

Suppose y ∈ f (A) for some closed set A in X. Since M = ∪{B(α1 ) : α1 ∈ Λ1 },

A = ∪{A ∩ (Y × B(α1 )) : α1 ∈ Λ1 }, so f (A) = ∪{f (A ∩ (Y × B(α1 ))) : α1 ∈ Λ1 }.
Because f (A ∩ (Y × B(α1 ))) ⊂ Pα1 and P1 is locally finite,

f (A) = ∪{f (A ∩ (Y × B(α1 ))) : α1 ∈ Λ1 },

so there is β1 ∈ Λ1 such that y ∈ f (A ∩ (Y × B(β1 ))). And so on, there is βi ∈ Λi

such that y ∈ f (A ∩ (Y × B(β1 , . . . , βn ))) ⊂ Pβi . Let β = (βi ). Then β ∈ M and
y ∈ Kβ . We prove (y, β) ∈ A. Obviously, any element of a neighborhood base of
(y, β) in X has the form (V (y)×B(β1 , . . . , βn ))∩X, where V (y) is a neighborhood of
y in Y . Since V (y)∩f (A∩(Y ×B(β1 , . . . , βn ))) = ∅, A∩(V (y)×B(β1 , . . . , βn )) = ∅.
So (y, β) ∈ A = A, and hence y = f (y, β) ∈ f (A), it follows that f (A) is a closed set
in Y , thus f is a closed mapping. 

Proposition 3.7.13 ([121, 201, 203]) Paracompact M-spaces, M ∗ -spaces, M -

spaces and wM-spaces are preserved by quasi-perfect mappings.

Proof In proof of the sufficiency of Theorem 3.7.12, we have proved that M ∗ -spaces
are preserved by quasi-perfect mappings, and this proof can also be used for M -
spaces. By Theorem 3.7.10, paracompact M-spaces are preserved by quasi-perfect
mappings. Below, we prove that if f : X → Y is a quasi-perfect mapping and X is a
wM-space, then Y is a wM-space.
Suppose {Un } is a wM-sequence in X such that Un+1 refines Un . For every n ∈ N
and y ∈ Y , define
V (n, y) = Y − f (X − st(f −1 (y), Un )).

then y ∈ V (n + 1, y) ⊂ V (n, y) ∈ τ (Y ). Define

Vn = {V (n, y) : y ∈ Y }.

We prove that {Vn } is a wM-sequence in Y . First, since {Un } is a wM-sequence in X,

if {an }, {bn } are two sequences with an ∈ st(bn , Un ), then {an } has an accumulation
point if and only if {bn } has an accumulation point. Now, if yn ∈ st2 (y, Vn ) for some
y ∈ Y and a sequence {yn } in Y , then there exist sequences {y1,n }, {y2,n } and {y3,n }
in Y such that {y1,n , y} ⊂ V (n, y2,n ) and {yn , y1,n } ⊂ V (n, y3,n ). By f −1 (V (n, y)) ⊂
st(f −1 (y), Un ), we have
232 3 Generalized Metric Spaces

f −1 (y2,n ) ∩ st (f −1 (y), Un ) = ∅,
f −1 (y1,n ) ∩ st (f −1 (y2,n ), Un ) = ∅,
f −1 (y3,n ) ∩ st (f −1 (y1,n ), Un ) = ∅,
f −1 (yn ) ∩ st (f −1 (y3,n ), Un ) = ∅.

Since f is a quasi-perfect mapping, {yn } has an accumulation point, and hence Y is

a wM-space. 

By Lemmas 2.1.14, 2.1.15 and 1.7.5, we have the following corollary.

Corollary 3.7.14 Let Φ be one of the following topological properties: being an

M ∗ -space, being an M -space and being a wM-space. If f : X → Y is a closed
mapping and X has the property Φ, then Y has the property Φ if and only if Y is a
q-space.
Proposition 3.7.15 ([374]) Every wM-space satisfies the decomposition theorem of
type countable compactness.
Proof Suppose f : X → Y is a closed mapping and {Un } is a wM-sequence in X,
and we may assume Un+1 refines Un . For each n ∈ N, let
Yn = {y ∈ Y : if S is a nontrivial sequence in Y , then there exist x ∈ f −1 (y) and a
subsequence S
of S such that st(x, Un ) ∩ f −1 (S
) = ∅}.
Then Yn is a discrete closed subspace of Y . Because otherwise, since f is a closed
mapping, {f −1 (y) : y ∈ Yn } is not a locally finite family in X, so there is x ∈ X such
that st(x, Ui ) (∀i ∈ N) meets infinitely many f −1 (y) (y ∈ Yn ), and hence there is a
nontrivial sequence {yi } in Yn such that st(x, Ui ) ∩ f −1 (yi ) = ∅. Since yi ∈ Yn , by the
inductive method, there are a subsequence {yik } of {yi } and xk ∈ f −1 (yik ) such that
st(x , U ) ∩ ∪{f −1 (yim ) : m > k} = ∅. If c is an accumulation point of {xk }, then c ∈
 k −1 n
{f (yim ) : m > k} for each k ∈ N and there is xk0 ∈ st(c, Un ), so c ∈ st(xk0 , Un ),

and hence c ∈ / {f −1 (yim ) : m > k}, a contradiction. Thus, {xk : k ∈ N} is a discrete
closed set in X. Since f is a closed mapping, {yik : k ∈ N} is a discrete closed set in
Y . Suppose ck ∈ st(x, Uk ) ∩ f −1 (yik ). Then {ck : k ∈ N} is a discrete closed set in X
and ck ∈ st(x, Uk ), a contradiction.
Let Z = n∈N Yn . If y ∈ Y − Z, then f −1 (y) is a countably compact set in
X. Because otherwise, there is a discrete closed set {xn : n ∈ N} in f −1 (y). For
each n ∈ N, since y ∈ / Yn , there is a nontrivial sequence Sn = {yin }i∈N , such that,
st(x, Un )∩f −1 (Sn
) = ∅ whenever x ∈ f −1 (y) and Sn
is a subsequence of Sn . So there
is jn ∈ N such that st(xn , Un )∩f −1 (yjn ) = ∅ for each j  jn and st(x1 , Un )∩f −1 (yjnn ) =
∅. We may assume {yjnn } is a nontrivial sequence and jn < jn+1 . Let

an ∈ st(xn , Un ) ∩ f −1 (yjnn ) and bn ∈ st(x1 , Un ) ∩ f −1 (yjnn ).

Since {xn : n ∈ N} is a discrete closed set, by the proof of Lemma 1.7.5(5.1),

{st(xn , Un )}n∈N is locally finite, so {{an }}n∈N is also locally finite, and hence
 3.7 M-Spaces 233

{f −1 (yjnn )}n∈N is discrete, it follows that {{bn }}n∈N is discrete, a contradiction. Thus
f −1 (y) is countably compact in X. 

Example 3.7.16 On mapping properties of M-spaces and relevant spaces.

(1) M-spaces are not preserved by perfect mappings [361].
Let H = (ω1 + 1) × (ω1 + 1) − {(ω1 , ω1 )}. Then H is a locally compact countably
compact space. Define

P = ω1 × {ω1 }, S = {ω1 } × ω1 .

Then P and S are closed subspaces of H. For each n ∈ N, let Hn be a copy of H and
let hn : H → Hn be the homeomorphic
 mapping. We may assume that Hn ∩ Hm = ∅
whenever n = m. Let X = n∈N Hn . Then X is a locally compact M-space. Define
a quotient space Y of X as follows: glue h2n−1 (p) with h2n (p) for each p ∈ P and
glue h2n (s) with h2n+1 (s) for each s ∈ S. Let f : X → Y be the quotient mapping.

(16.1) f is a perfect mapping.

−1
Obviously, f is a compact mapping. Let f (y) ⊂ U −1 ∈ τ (X). If |f −1 (y)| = 1, let
V = f (U − n∈N hn (P ∪ S)), then y ∈ V ∈ τ (Y ) and f (V ) ⊂ U. If |f −1 (y)| > 1,
we may assume there exist m ∈ N and p ∈ P such that f −1 (y) = {h2m−1 (p), h2m (p)},
so there is W ∈ τ (H) such that p ∈ W , W ∩S = ∅ and h2m−1 (W )⊕h2m (W ) ⊂ U. Let
V = f (h2m−1 (W )⊕h2m (W )). Then f −1 (V ) = h2m−1 (W )⊕h2m (W ), so y ∈ V ∈ τ (Y )
and f −1 (V ) ⊂ U. Thus f is a closed mapping.
(16.2) Y is not an M-space.

Otherwise, there exist a metric space Z and a quasi-perfect mapping g : Y → Z.

Let gn = g ◦ f ◦ hn : H → Z for each n ∈ N. Then gn (H) is a compact metric
space, so gn (H) can be embedded into the Hilbert cube Iℵ0 . Since both P and S are
homeomorphic to ω1 , there is αn < ω1 such that gn takes a constant value zn on
((αn , ω1 ) × {ω1 }) ∪ ({ω1 } × (αn , ω1 )). However, g2n−1 (p) = g2n (p) when p ∈ P and
g2n (s) = g2n+1 (s) when s ∈ S, so zn = z1 for all n ∈ N, and hence g−1 (z1 ) ∩ f (Hn )
is an infinite set, it follows that there is an infinite set {yn : n ∈ N} in Y such that
yn ∈ g−1 (z1 ) ∩ f (Hn ). Since {Hn }n∈N is a locally finite family in X, {f (Hn )}n∈N is
an HCP family in Y , so {yn : n ∈ N} is a discrete set in Y , which contradicts the
countable compactness of g−1 (z1 ).
(2) Paracompact M-spaces are not preserved by closed mappings, Example 2.11.9
offers such an example.
(3) Not every closed image of a paracompact M-space can be represented as a
perfect preimage of a Lašnev space, for example the space X in Example 2.11.9. It
is easy to verify that every perfect preimage of a Lašnev space is a Σ-space.
234 3 Generalized Metric Spaces

Example 3.7.17 On relationships between M-spaces and relevant spaces.

(1) There is a space with a countable quasi-(mod k)-base which is not a wM-space
[242].
Let X = [0, ω] × [0, ω1 ] − {(ω, ω1 )}. Then X is a locally compact space. Define
Cα = {α} × [0, ω1 ] − {(ω, ω1 )} for each α  ω. Then {Cα }αω is a closed cover of
X consisting of countably compact sets. Let Un = [n, ω] × [0, ω1 ) for each n < ω.
Then Un ∈ τ (X). If Cω ⊂ V ∈ τ (X), then there is m < ω such that Um ⊂ V . Because
otherwise, suppose {xn } is a sequence such that xn ∈ Un − V . Denote xn = (an , bn ).
We may assume bn → b < ω1 . Then xn → (ω, b) ∈ Cω ⊂ V , a contradiction. That
means {Un : n < ω} ∪ {Cα : α < ω} is a countable quasi-(mod k)-base for X w.r.t.
{Cα }αω .
We prove X is not a countably paracompact space. Otherwise, since {(n, ω1 ) :
n < ω} is a discrete closed set in X, there is a locally finite family {Vn : n < ω}
of open sets in X such that (n, ω1 ) ∈ Vn . For each n < ω, take βn < ω1 such that
{(n, y) ∈ X : y > βn } ⊂ Vn . Pick β < ω1 such that β > sup{βn : n < ω}. Then {Vn }
is not locally finite at (ω, β), a contradiction.
(2) Ishiwata [205] constructed completely regular countably compact spaces X, Y
such that X × Y is not a q-space. Thus the product of M-spaces need not be an M-
space.
(3) Not every M ∗ -space is an M-space, the space Y in Example 3.7.16(1) is such
a counterexample.
(4) There is a space with a uniform base which is not a wM-space, the V -space
in Example 1.8.1 is such a space.

3.8 ℵ-Spaces

We are committed to investigating the properties of ℵ-spaces in this section. We focus

mainly on the equivalent relationships among generalized metric spaces generated
by the combinations of strict k-networks, cs-networks or k-networks and σ -discrete
families, σ -locally finite families or σ -HCP families. By the investigation of Foged
[128], Junnila, Z. Yun [221] and other researchers, the relationships among these
spaces became quite clear. The fascinating characterizations of ℵ-spaces reflect the
pursuit of goal of general topologists. This section is divided into four parts to intro-
duce the class of ℵ-spaces. First, we investigate the regular spaces with a σ -HCP
strict k-network. Second, we explore the equivalent relationships among spaces with
a σ -discrete k-network, ℵ-spaces and cs-σ -spaces. Then, we discuss the equivalent
relationships between spaces with a σ -HCP cs-network and ℵ-spaces. Finally, we
introduce some mapping theorems of ℵ-spaces.
3.8 ℵ-Spaces 235

Theorem 3.8.1 ([256]) A regular space X has a σ -HCP strict k-network if and only
if either X is an ℵ0 -space or X is an Fσ -discrete space in which every compact set is
finite.

Proof The necessity. Suppose X has a σ -HCP strict k-network P. If every compact
set in X is finite, then by Lemma 3.2.36, X is an Fσ -discrete space. Assume there is
an infinite compact set K in X. Then X is an ℵ1 -compact space. Because otherwise,
there is a discrete closed set A in X such that |A| = ℵ1 . We may assume A ∩ K = ∅.
Denote A = {xα : α < ω1 }. For each α < ω1 , let Vα = X − (A − {xα }). Then
K ∪ {xα } ⊂ Vα ∈ τ , so there is Pα ∈ P such that K ∪ {xα } ⊂ Pα ⊂ Vα . By the
structure of Vα , {Pα }α<ω1 is a family consisting
of different elements of P, and we
may assume it is an HCP family. Since K ⊂ α<ω1 Pα , any countable subset of K
is a discrete closed set in X, which contradicts the compactness of K. Thus X is an
ℵ1 -compact space. 
Now, denote P = n∈N Pn , where Pn is an HCP family. For each n ∈ N, define

En = {x ∈ X : |(Pn )x | > ℵ0 },
Fn = {P − En : P ∈ Pn } ∪ {{x} : x ∈ En }.

By Lemma 3.2.14, En is a  countable discrete closed subspace of X and Fn is a

countable family. We prove n∈N Fn is a k-network for X. If K is a compact set in
X and K ⊂ U ∈ τ , then there exist m ∈ N and P ∈ Pm such that K ⊂ P ⊂ U, so
{P − Em } ∪ {{x} : x ∈ K ∩ Em } is a finite subfamily of Fm and K ⊂ (P − Em ) ∪ (K ∩
Em ) ⊂ U. Thus X has a countable k-network, and hence X is an ℵ0 -space.
The sufficiency. We only need to prove that every Fσ -discretespace X in which
any compact set is finite has a σ -HCP strict k-network. Let X = n∈N Xn , where Xn
is a discrete closed set in X and Xn ⊂ Xn+1 . For each n ∈ N, let

Pn = {A ⊂ Xn : |A|  n}.

Then n∈N Pn is a σ -HCP strict k-network for X. 

Now, we have proved that every regular space with a σ -HCP strict k-network is
an ℵ-space, and an ℵ-space has a σ -HCP strict k-network if and only if either it is a
Lindelöf space or each compact set in it is finite.
We gave the definition of CF families in Definition 2.5.2. A family P in a space
X is said to be a CF ∗ family [347] in X if it is CF in X and moreover satisfies
that whenever K ∩ P is infinite for a compact set K ⊂ X and P ∈ P, the family
{P
∈ P : K ∩ P = K ∩ P
} is finite. Every weakly HCP family is CF ∗ . Sakai [418]
proved that every space X with a σ -CF ∗ strict k-network if and only if either X is an
ℵ0 -space or each compact subset of X is finite. This result proves a conjecture raised
by Mizokami and S. Lin [350].
In the second part of this section, we discuss the equivalent relationships among
ℵ-spaces, cs-σ -spaces and spaces with a σ -discrete k-network (or cs-network).
236 3 Generalized Metric Spaces

Lemma 3.8.2 ([128]) Suppose P is a point-countable closed k-network for a space

X which is closed under finite intersections. If {xn } is a sequence converging to x in
X and W is a sequential neighborhood of x, then there is P
∈ (P)<ω x such that
{xn } is eventually in ∪P
⊂ W .

Proof Denote

{P
∈ (P)<ω

x : {xn } is eventually in ∪ P } = {Pi }i∈N .

 the conclusion is not true, then there is a sequence {yn } in X such that yn ∈
If
in (∪Pi ) − W , and hence yn → x, which contradicts the assumption that W is a
sequential neighborhood of x. 

Lemma 3.8.3 ([378]) Suppose P is a σ -HCP family in a space X. If P is a cs∗ -

network for X, then P is a k-network for X.

Proof Let P = n∈N Pn , where each Pn is an HCP family. Suppose K is a compact
set in X and K ⊂ U ∈ τ . For each n∈ N, let Fn = ∪{P ∈ Pn : P ⊂ U}. If there is a
sequence {xn } such that xn ∈ K − in Fi , then there is a subsequence {xni } of {xn }
converging to x ∈ X. Take m ∈ N and P ∈ Pm such that {x}∪{xn i
: i ∈ N} ⊂ P ⊂ U.
Then xnm ∈ Fm , a contradiction.
 So there is k ∈ N such that K ⊂ ik Fi . By Lemma
2.5.4, there is F ∈ ( ik Pi )<ω such that K ⊂ ∪F ⊂ U. Thus P is a k-network
for X. 

Now we introduce the following beautiful Foged theorem.

Theorem 3.8.4 [128] (The Foged theorem) For every regular space X, the following
are equivalent:
(1) X has a σ -discrete cs-network.
(2) X has a σ -discrete k-network.
(3) X is a cs-σ -space.
(4) X has a σ -locally finite cs∗ -network.
(5) X is an ℵ-space.

Proof By Lemma 3.8.3, we only need to prove (5) ⇒ (1). Suppose P = m∈N Pm
is a k-network for X, where each Pm is a locally finite family of closed sets which
is closed under finite intersections in X and Pm ⊂ Pm+1 . Denote P = {Pα }α∈Λ .
For each m ∈ N, there is an open cover Um of Xsuch that every element of Um only
meets finitely many elements of Pm . Suppose n∈N {Fβ : β ∈ Γm,n } is a refinement
of Um such that each {Fβ : β ∈ Γm,n } is a discrete family of closed
 sets in X. Then
Fβ only meets finitely many elements of Pm for each β ∈ n∈N Γm,n . For each
(m, n) ∈ N × N, by Theorem 3.4.11(11.1), there is a family {Wβ : β ∈ Γm,n } of
disjoint sets in X such that Wβ is a sequential neighborhood of Fβ . Let
3.8 ℵ-Spaces 237

Am,n = {(α, β) : Pα ∈ Pm , β ∈ Γm,n and Pα ∩ Fβ = ∅}.

Then {Pα ∩ Wβ : (α, β) ∈ Am,n } is a star-finite family. Because in fact, if (Pα ∩

Wβ ) ∩ (Pγ ∩ Wδ ) = ∅ for every (α, β), (γ , δ) ∈ Am,n , then β = δ, and hence
(γ , β) ∈ Am,n , so Pγ ∩ Fβ = ∅. Thus, (Pα ∩ Wβ ) ∩ (Pγ ∩ Wδ ) = ∅ for only finitely
many (γ , δ) ∈ Am,n .
Take (m, n) ∈ N × N. For every (α, β) ∈ Am,n and i ∈ N, define

H(α, β, i) = ∪{Pα ∩ Pγ : Pγ ∈ Pi , Pγ ⊂ Wβ },
H (m, n, i) = {H(α, β, i) : (α, β) ∈ Am,n }.

Then H(α, β, i) ⊂ Pα ∩ Wβ , and hence H (m, n, i) is a star-finite family. By Lemma

2.8.5, H (m, n, i) is a σ -disjoint family. Since Pm ∧ Pi is locally finite, H (m, n, i)
is closure-preserving, so H (m, n, i) is a σ -discrete family. Define

H = ∪{H (m, n, i) : m, n, i ∈ N}.


Then H is a σ -discrete family of closed sets. Let H = k∈N Hk , where each Hk
is a discrete family of closed sets and Hi ∩ Hj = ∅ for every i, j ∈ N with i = j.
Let
R = {F ∈ H <ω : ∩F = ∅}.

For each I ∈ N<ω , define

R(I) = {F ∈ R : {k ∈ N : F ∩ Hk = ∅} = I}.

Then {∩F : F ∈ R(I)} = k∈I Hk is locally finite. If F1 , F2 ∈ R(I) with F1 =
F2 , then there exist k ∈ N and H1 , H2 ∈ Hk with H1 = H2 such that F1 ∩Hk = {H1 }
and F2 ∩ Hk = {H2 }, so H1 ∩ H2 = ∅, and hence (∩F1 ) ∩ (∩F2 ) = ∅. Thus
{∩F : F ∈ R(I)} is a discrete family of closed sets. By Theorem 3.4.11(11.1),
there is a family {V (F ) : F ∈ R(I)} of disjoint sets such that V (F ) is a sequential
neighborhood of the set ∩F . For every j ∈ N and F ∈ R(I), define

V (F , j) = ∪{H ∩ Pδ : H ∈ F , Pδ ∈ Pj and Pδ ⊂ V (F )},

V (I, j) = {V (F , j) : F ∈ R(I)}.

Then V (F , j) ⊂ V (F ). Since ( k∈I Hk ) ∧ Pj is a locally finite family, V (I, j) is
a discrete family. Let

V = ∪{V (I, j) : I ∈ N<ω , j ∈ N}.

Then V is a σ -discrete family in X. We prove V is a cs-network for X.

Suppose a sequence {xn } converges to x ∈ U ∈ τ in X. Then there exist m ∈ N
and F1 ∈ Pm<ω such that {xn } is eventually in ∪F1 ⊂ U. We may assume x ∈ ∩F1 .
238 3 Generalized Metric Spaces

Choose n ∈ N and β ∈ Γm,n such that x ∈ Fβ . Since Wβ is a sequential neighborhood

of Fβ , by Lemma 3.8.2, there exist i ∈ N and F2 ∈ Pi<ω such that {xn } is eventually
in ∪F2 ⊂ Wβ . Since x ∈ ∪F2 , x ∈ ∩{H(α, β, i) : Pα ∈ F1 }. If Pα ∈ F1 and
Pγ ∈ F2 , then Pα ∩Pγ ⊂ H(α, β, i), so (∪F1 )∩(∪F2 ) ⊂ ∪{H(α, β, i) : Pα ∈ F1 }.
Let
F = {H(α, β, i) : Pα ∈ F1 }.

Then x ∈ ∩F and {xn } is eventually in ∪F , and hence F ∈ R. By Lemma 3.8.2,

there exist j ∈ N and F3 ∈ Pj<ω such that {xn } is eventually in ∪F3 ⊂ V (F ). Since
(∪F ) ∩ (∪F3 ) ⊂ V (F , j), {xn } is eventually in V (F , j) ⊂ ∪F ⊂ ∪F1 ⊂ U. Thus
V is a cs-network for X. 

In the third part of this section, we investigate the precise relationships among
spaces with a σ -HCP k-network, spaces with a σ -HCP cs-network and ℵ-spaces.

Proposition 3.8.5 ([267, 493]) Sω1 has no σ -HCP cs-network.

Proof Otherwise, let Sω1 = {(α, 1/n) : α < ω1 , n ∈ N} ∪ {∞}, where each sequence
{(α, 1/n)} converges to ∞. Suppose P is a σ -HCP cs-network for Sω1 .
For each α < ω1 , let S (α) = {(α, 1/n) : n ∈ N}. By the transfinite inductive
method, we can prove that the subfamily {P ∈ P : |S (0) ∩ P| = ℵ0 } of P is
uncountable, which contradicts Lemma 2.5.4.
Since P is a cs-network, there is P0 ∈ P such that S (0) is eventually in P0 . Pick
p0 ∈ S (0) ∩ P0 . Suppose 0 < α < ω1 and we have chosen Pβ ∈ P, pβ ∈ S (β) ∩ Pβ
for each β < α such that |S (0) ∩ Pβ | = ℵ0 . We regard S (0) ∪ S (α) as a sequence such
that the subsequences consisting of the odd terms and the even terms of which are
S (0) and S (α) respectively. Then there is Pα ∈ P such that S (0) ∪ S (α) is eventually
in Pα ⊂ Sω1 − {pβ : β < α}. It follows that |S (0) ∩ Pα | = ℵ0 and Pβ = Pα . Thus
{P ∈ P : |S (0) ∩ P| = ℵ0 } is uncountable. 

Lemma 3.8.6 ([221]) Suppose a space X contains no closed copy of Sω1 . If F is

an HCP family in X, then for each x ∈ X, {F ∈ F : there is a sequence in F −
{x} converging to x} is countable.

Proof Suppose there is y ∈ X and an uncountable subfamily {Fα }α<ω1 of F satisfying

that there is a subset Yα = {yα,n : n ∈ N} of Fα − {y} such that yα,n → y. By
Lemma 2.5.4, we may assume F is point-finite at yα,n , so {F ∈ F : F ∩ Yα = ∅} is
countable, and hence {Yα }α<ω1 is a star-countable family. By Lemma 2.8.5, {Yα }α<ω1
is a σ -disjoint family, and we may assume  that it is a disjoint family. Since F is an
HCP family, the closed subspace {y} ∪ ( α<ω1 Yα ) of X is homeomorphic to Sω1 , a
contradiction. 
3.8 ℵ-Spaces 239

Lemma 3.8.7 ([221]) Suppose X is a regular space with a σ -HCP k-network and
X contains no closed copy of Sω1 . If D is a discrete closed subspace of X, then there
is a σ -discrete family H of closed sets in X such that for each compact set K in X
and d ∈ K ∩ D, {H ∈ H : d ∈ intK (K ∩ H)} is a net at d in X.

Proof First notethat X is a σ -space, so every compact subset of X is metrizable.

Suppose F = n∈N Fn is a k-network for X such that each Fn is an HCP family
of closed sets and Fn ⊂ Fn+1 . For each x ∈ X, let

Fx = {F ∈ F : there is a compact set K in F such that x ∈ K − {x}}.

By Lemma 3.8.6, |Fx |  ℵ0 , so we can denote

{∪F
: F
∈ Fx<ω } ∪ {{x}} = {Fk,x }k∈N .

For every n ∈ N and d ∈ D, take a closed neighborhood Gn,d of d such that Gn,d ⊂
X − ∪{F ∈ Fn : d ∈ / F}. Let Pn,d = ∪{F ∈ Fn : F ∩ D = {d}}. Then {Gn,d ∩ Pn,d :
d ∈ D} is a discrete family of closed sets in X. Define

Hk,n = {Fk,d ∩ Gn,d ∩ Pn,d : d ∈ D}, k, n ∈ N;


H = Hk,n .
k,n∈N

Then H is a σ -discrete family of closed sets in X. We prove that H satisfies our

requirement. For each compact set K in X and d ∈ K ∩ D, let U be a neighborhood of
d. Take a closed neighborhood V of d such that V ⊂ U − (D − {d}). Then there exist
n ∈ N and F
∈ Fn<ω such that K ∩ V ⊂ ∪F
⊂ U − (D − {d}). Choose k ∈ N such
that Fk,d = ∪(F
∩ Fd ) ∪ {d}. Then d ∈ Fk,d ∩ Gn,d ∩ Pn,d ⊂ U. Below we verify
that d ∈ intK (K ∩ Fk,d ∩ Gn,d ∩ Pn,d ). Since d ∈ intK (K ∩ V ) ⊂ intK (K ∩ (∪F
)),
d ∈ intK (K∩(∪(F
)d )) ⊂ intK (K∩Pn,d ). Because d ∈ int(Gn,d ), d ∈ intK (K∩Gn,d ).
For each F ∈ F
, let

W (F) = X − F ∩ K − {d},
W = ∩{W (F) : F ∈ F
and d ∈ W (F)}.

Then d ∈ W ∈ τ and {F ∈ F
: F ∩ K ∩ W = ∅} ⊂ F
∩ Fd , so K ∩ (∪F
) ∩ W ⊂
Fk,d , and hence d ∈ intK (K ∩ Fk,d ). In a summary, d ∈ intK (K ∩ Fk,d ∩ Gn,d ∩ Pn,d ).


The following theorem is the main result in the third part of this section.

Theorem 3.8.8 For every regular space X, the following are equivalent:
(1) X is an ℵ-space.
240 3 Generalized Metric Spaces

(2) X has a σ -HCP cs-network [267, 493].

(3) X has a σ -HCP k-network and contains no closed copy of Sω1 [221]. (The
Junnila–Yun theorem)

Proof By Theorem 3.8.4, Lemma  3.8.3 and Proposition 3.8.5, we only need to
prove (3) ⇒ (1). Suppose F = n∈N Fn is a k-network for X such that each
Fn is an HCP family of closed sets and Fn ⊂ Fn+1 . For each n ∈ N, let

Dn = {x ∈ X : |(Fn )x |  ℵ0 }.

Then Dn is a closed  set in X, and by (22.2) in the proof of Lemma 3.2.22, we

can denote Dn = k∈N Dn,k , where each Dn,k is a discrete closed set in X and
Dn,k ⊂ Dn,k+1 . For every n, k ∈ N, there is a discrete family Hn,k of closed sets in
X satisfying the assumption (H = Hn,k , D = Dn,k ) in Lemma 3.8.7. Define

Fn,k = ∪{F ∈ Fk : F ∩ Dn = ∅},

Fn,k = Fn|Fn,k .

 finite in X −Dn , Fn,k is also a locally finite family in X. Below we

Since Fn is locally
prove that P = n,k∈N (Hn,k ∪Fn,k ) is a k-network for X. For each compact set K in
X with K ⊂ U ∈ τ , there exist n ∈ N and F
∈ Fn<ω such that K ⊂ ∪F
⊂ U. By
Lemma 2.5.4, K ∩Dn is a finite set, so there is k ∈ N such that K ∩Dn ⊂ Dn,k . For each
d ∈ K ∩ Dn , by the property of Hn,k , there is Hd ∈ Hn,k such that d ∈ intK (K ∩ Hd )
and Hd ⊂ U. Let

L = K − ∪{intK (K ∩ Hd ) : d ∈ K ∩ Dn }.

Then there exist i ∈ N and F

∈ Fi<ω such that L ⊂ ∪F

⊂ X − Dn , so L ⊂ Fn,i .
Define
P
= {Hd : d ∈ K ∩ Dn } ∪ {F ∩ Fn,i : F ∈ F

}.

Then P
∈ P <ω and K ⊂ ∪P
⊂ U. So X is an ℵ-space. 

By (7.5) in Example 1.8.7, we have the following corollary.

Corollary 3.8.9 A regular space X is an ℵ-space if and only if X has both a σ -

HCP k-network and a point-countable cs∗ -network.

Corollary 3.8.10 ([221]) Suppose both X and Y are regular spaces with a σ -HCP k-
network. Then X × Y has a σ -HCP k-network if and only if either both X and Y are
ℵ-spaces or every compact set is finite in one of them.

Proof The necessity. If X is not an ℵ-space, then by Theorem 3.8.8, X contains a

closed copy of Sω1 . By Example 2.5.22, Y contains no copy of S1 , so every compact
set in Y is finite.
3.8 ℵ-Spaces 241


The sufficiency. Suppose n∈N Pn is a k-network for X such that each Pn is an
HCP family. Assume that every compact set in Y is finite. By Lemma 3.2.36, we can
denote Y = k∈N Yk , where each Yk is a discrete closed subspace of Y . For every
n, k ∈ N, define
Fn,k = {P × {y} : P ∈ Pn , y ∈ Yk }.

Then n,k∈N Fn,k is a σ -HCP k-network for X × Y . 


Corollary 3.8.11 Suppose the product space X = n∈N Xn is a regular space with
a σ -HCP k-network. If |Xn |  2 for each n ∈ N, then X is an ℵ-space.

Question 3.8.12 ([379]) Characterize ℵ-spaces by means of bases.

In the last part of this section, we discuss some mapping theorems on ℵ-spaces.
By Corollary 1.6.9, Propositions 2.7.2 and 2.1.13, we have the following result.

Proposition 3.8.13 Suppose f : X → Y is a mapping, X is an ℵ0 -space and Y is

a regular space. If f satisfies one of the following conditions, then Y is an ℵ0 -space.
(1) f is a sequentially quotient mapping.
(2) f is a sequence-covering mapping.
(3) f is a closed mapping.
By Theorem 3.4.12, we have the following result.

Proposition 3.8.14 Spaces with a σ -HCP closed strict k-network or with a σ -HCP
closed k-network are preserved by closed mappings.

Theorem 3.8.15 ([146, 258, 444]) Suppose f : X → Y is a closed L-mapping and

Y is a regular space. If X is an ℵ-space, then Y is an ℵ-space.

Proof By Proposition 3.8.14, Y has a σ -HCP k-network. Let P be a σ -locally

finite cs-network for X. Since f is a closed L-mapping, f (P) is a point-
countable cs∗ -network for Y . By Corollary 3.8.9, Y is an ℵ-space. 

Theorem 3.8.16 Suppose f : X → Y is a closed mapping. If X is an ℵ-space and

Y is a regular k-space, then the following are equivalent:
(1) Y is an ℵ-space.
(2) Y contains no closed copy of Sω1 .
(3) f is a peripheral L-mapping.

Proof By Theorem 3.8.15, Example 1.8.7 and Theorem 3.4.16, we get (3) ⇒ (1) ⇒
(2) ⇒ (3). 
242 3 Generalized Metric Spaces

For a space Y and a point y ∈ Y ,

χ (Y , y) = min{|By | : By is a local base of y in Y }

is called the character of y in Y and

χ (Y ) = sup{χ (Y , y) : y ∈ Y }

is called the character of Y . The following corollary is parallel to the Hanai-Morita-

Stone theorem.

Corollary 3.8.17 ([147]) Suppose f : X → Y is a closed mapping. If X is a metric

space, then the following are equivalent:
(1) Y is an ℵ-space.
(2) f is a peripheral L-mapping.
(3) (CH) χ (Y )  ℵ1 .

Proof We have proved (1) ⇔ (2) in Theorem 3.8.16. Since χ (Sω1 ) > ℵ1 , we get
(3) ⇒ (1) by Theorem 3.8.16.
(2) ⇒ (3). By Lemma 2.1.15, we may assume f is a closed L-mapping. Fix y ∈ Y .
Then f −1 (y) has a countable base. By CH, |f −1 (y)|  ℵ1 . For each x ∈ X, let Ux be
a countable local base of x. Define U = ∪{Ux : x ∈ f −1 (y)} and

B = {Y − f (X − ∪U
) : U
is a countable subfamily of U }.

Then |B|  ℵ1 and B is a local base of y. Hence χ (Y )  ℵ1 . 

By the above corollary, we obtain the following result which is similar to Theorem
2.3.10: Every closed image of a metric space X is an ℵ-space if and only if X d is a
Lindelöf subspace of X.

Theorem 3.8.18 ([417]) If f : X → Y is a countably bi-quotient closed mapping

from an ℵ-space X onto a regular space Y , then Y is an ℵ-space.

Proof Since X is an ℵ-space, we may suppose P = n∈N Pn is a σ -locally finite
closed k-network for X, such that, Pn is locally finite and Pn ⊂ Pn+1 for each n ∈
N. By Theorem 3.4.12, f is a compact-covering mapping, so f (P) = {f (P) : P ∈ P}
is a σ -HCP k-network for Y . By the Junnila–Yun theorem (see Theorem 3.8.8), we
only need to prove Y contains no closed copy of Sω1 .
Otherwise, let {∞} ∪ {yα,n : α < ω1 , n ∈ N} be a closed copy of Sω1 in Y ,
where {yα,n }n∈N is the α-th sequence converging to ∞. We construct inductively a
subset {nα }α<ω1 of N and a subfamily {Fα }α<ω1 of finite subsets of P satisfying the
following conditions:
3.8 ℵ-Spaces 243

(i) for each α < ω1 , ∪{f −1 (yα,n ) : n  nα } ⊂ ∪Fα ;

(ii) for each P ∈ Fα , P ∩ (∪{f −1 (yα,n ) : n  nα }) = ∅;
(iii) for every α < β < ω1 , Fα ∩ Fβ = ∅.
Take γ < ω1 and assume that for every α < γ , there exist nα ∈ N and a subfamily
Fα of P satisfying the above conditions. For each α < γ , by (i), we can take a
finite subset Fα of ∪{f −1 (yα,n ) : n  nα } such that Fα ∩ P = ∅ for each P ∈ Fα .
Let F = ∪{Fα : α < γ }. Then F is closed in X. For each n ∈ N, take

Qn = {P ∈ Pn : P ∩ F = ∅ and P ∩ f −1 (yγ ,k ) = ∅ for infinitely many k ∈ N}.

Obviously, Qn ⊂ Qn+1 . If there is n ∈ N such that Qn is infinite, take {Pi }i∈N ⊂ Qn

such that Pi = Pj whenever i = j. Then there is a sequence {xi } such that xi ∈ Pi and
{f (xi )} is a subsequence of {yγ ,n }n∈N . Since Qn ⊂ Pn is locally finite, {xi : i ∈ N} is a
closed set, and hence {f (xi ) : i ∈ N} is a closed set, a contradiction. As a consequence,
each Qn is finite. Below we prove there is nγ ∈ N such that

∪{f −1 (yγ ,n ) : n  nγ } ⊂ ∪Qnγ ,

and hence we can take Fγ = Qnγ to complete the induction.

Otherwise, for each n ∈ N, f −1 (yγ ,k ) ⊂ ∪Qn for infinitely many k ∈ N. Then we
can pick kn ∈ N and xn ∈ f −1 (yγ ,kn ) − ∪Qn . We may assume kn < kn+1 for each
n ∈ N. Since f is a closed mapping, any infinite subset of {xn : n ∈ N} is not closed
in X.
Claim. {xn : n ∈ N} contains a sequence converging to some point of f −1 (∞).
In fact, since X is an ℵ-space, X is of the point Gδ -property. By Proposition
2.1.13, f is a sequentially quotient mapping. Further since {f (xn )} is a subsequence
of {yγ ,n }n∈N , there is a convergent subsequence {zi } of {xn }. If {zi } converges to x0 ,
then f (x0 ) = ∞, and hence x0 ∈ f −1 (∞).
Since P is a k-network for X and the compact set {x0 } ∪ {zi : i ∈ N} ⊂ X − F,
there exist l ∈ N and P ∈ Pl such that P contains infinitely many points in {xn }
and P ⊂ X − F, so P ∈ Ql . However, for every n  l, xn ∈ X − ∪Ql , and hence
∪Ql contains at most finitely many points in {xn }, a contradiction. That means the
inductive construction is completed.
Since Fα is finite, there exist m ∈ N and an infinite set {αi : i ∈ N} ⊂ ω1 such
every i ∈ N, Fαi ⊂ Pm . Let Ei = ∪Fαi , i ∈ N. Since Pm is locally finite,
that for
{X − ji Ej : i ∈ N} is an open cover of X, so it covers f −1 (∞). Since  f is a
countably bi-quotient closed mapping, there is i0 ∈ N such that f (X − ji0 Ej )
is a neighborhood of ∞ in Y . On the other hand, by (i), f −1 (yαj ,n ) ⊂ Ej when
j  i0 , n  nαj , so
⎛ ⎞

yαj ,n ∈ Y − f (X − Ej ) ⊂ Y − f ⎝X − Ej ⎠ ,
ji0
244 3 Generalized Metric Spaces


which contradicts the assumption that f (X − ji0 Ej ) is a neighborhood of ∞. Thus,
Y contains no closed copy of Sω1 . 

Below we introduce the perfect preimage Gδ -diagonal theorem for ℵ-spaces.

Lemma 3.8.19 ([258]) Suppose a regular space X is a perfect preimage of an ℵ-

space. If U is an open cover of X, then
(1) U has a σ -discrete closed refinement F such that K (X) refines F F ;
(2) there is a sequence {Wn } of open refinements of U , such that, for each nonempty
compact set K in X, there exist m ∈ N, {Ki : i  m} ⊂ K (X) and {ni : i 
m} ⊂ N satisfying K = im Ki and |(Wni )Ki | = 1.

Proof
 Suppose f : X → Y is a perfect mapping and Y is an ℵ-space. Let P =
n∈N P n be a k-network for Y , where each Pn is a discrete family of closed sets.
By the regularity of X, we can choose an open cover V of X such that V refines U .
For each y ∈ Y , since f is a perfect mapping, there is an open neighborhood Hy of
y such that f −1 (Hy ) is contained in the union of finitely many elements of V . We
may assume P refines {Hy }y∈Y . Then for each P ∈ P, there is VP ∈ V <ω such that
f −1 (P) ⊂ ∪VP . Let −1
 F = ∪{{f (P)} ∧ VP : P ∈ P}. Then F satisfies (1).
Denote F = n∈N Fn , where each Fn = {Fα : α ∈ Λn } is a discrete family of
closed sets in X. For every n ∈ N and α ∈ Λn , take Uα ∈ U such that Fα ⊂ Uα . Let

Wα = Uα − ∪{Fβ : β ∈ Λn − {α}},
Wn = {Wα : α ∈ Λn } ∪ {U − ∪Fn : U ∈ U }.

Then {Wn } is a sequence consisting of open refinements of U . For each K ∈ K (X)−

{∅}, take a finite subfamily {Fi : i  m}
of (F )K covering K. For each i  m, take
ni ∈ N such that Fi ∈ Fni . Then K = im (K ∩ Fi ) and |(Wni )K∩Fi | = 1. So {Wn }
satisfies (2). 

Theorem 3.8.20 ([258]) ℵ-spaces satisfy the perfect preimage Gδ -diagonal theo-
rem.

Proof Suppose f : X → Y is a perfect mapping, X is a regular space with a Gδ -

diagonal and Y is an ℵ-space.
 Claim. There is a sequence {Un } of open covers of X such that K =
n∈N st(K, Un ) for each K ∈ K (X) − {∅}.
By Proposition 1.4.9, X has a G∗δ -diagonal sequence {An } such that An+1 refines
An . If x ∈ X − K for each K ∈ K (X) − {∅}, then there is k ∈ N such that
K ⊂ X − st(x, Ak ), because {X − st(x, An )}n∈N covers K. So x ∈ / st(K, Ak ), and
hence K = n∈N st(K, An ). For each j ∈ N, by Lemma 3.8.19, there is a sequence
{Bj,n } of open covers of X with the following properties:
 
(20.1) for each n ∈ N, B j,n refines ( i,l<j Bi,l ) ∧ ( kj Ak );
3.8 ℵ-Spaces 245

(20.2) {Bj,n }n∈N satisfies Lemma 3.8.19(2).

For every K ∈ K (X) − {∅} and k ∈ N, take j > k. By (20.2),  there exist a
family {Ki : i  m} ⊂ K (X) and {ni : i  m} ⊂ N such that K = im Ki and
|(Bj,ni )Ki | = 1. Then st(Ki , Bj,ni ) ⊂ st(Ki , Ak ). Take h > max{j, n1 , n2 , . . . , nm }.
Then

K⊂ st(K, Bs,t ) ⊂ st(K, Bh,1 ) ⊂ st(Ki , Bj,ni ) ⊂ st(K, Ak ).
s,t∈N im


So K = s,t∈N st(K, Bs,t ). Thus Claim is proved.
Suppose {Un } is a sequence in X satisfying Claim such that Un+1 refines Un . For
each n ∈ N, by Lemma 3.8.19, Un  has a σ -discrete closed refinement Fn such that
K (X) refines FnF . Denote Fn = m∈N Fn,m ,where each Fn,m is a locally finite
family of closed sets and Fn,m ⊂ Fn,m+1 . Let k∈N Pk be a k-network for Y such
that each Pk is a locally finite family and Pk ⊂ Pk+1 . Define

Hn,m,k = Fn,m ∧ f −1 (Pk ), n, m, k ∈ N.


Then Hn,m,k is a locally finite family in X. We verify n,m,k∈N Hn,m,k is a k-
network for X. If K is a nonempty compact set in X with K ⊂ U ∈ τ (X), then

{U} ∪ {X − st(K, Un )}n∈N

covers f −1 (f (K)), so there is n ∈ N such that f −1 (f (K)) ⊂ U ∪ (X − st(K, Un )),

and hence st(K, Un ) ∩ f −1 (f (K)) ⊂ U, it follows that (f −1 (f (K)) − U) ∩
st(K, Un ) = ∅. As a consequence, there is V ∈ τ (X) such that

f −1 (f (K)) − U ⊂ V and V ∩ st(K, Un ) = ∅.

Thus f (K) ⊂ Y −f (X−(U∪V )), and hence there exist k ∈ N and P

∈ Pk<ω such
that f (K) ⊂ ∪P
⊂ Y − f (X − (U ∪ V )). Then f −1 (f (K)) ⊂ ∪f −1 (P
) ⊂ U ∪ V .
By the definition of Fn , there exist m ∈ N and F
∈ Fn,m
<ω
such that K ⊂ ∪F
⊂

−1

st(K, Un ), so K ⊂ ∪(F ∧ f (P )) ⊂ U. Thus X is an ℵ-space. 

Corollary 3.8.21 ([318]) ℵ0 -spaces satisfy the perfect preimage Gδ -diagonal

theorem.
Example 3.8.22 On relationships between ℵ-spaces and relevant spaces.
(1) There is a Lašnev space which is not an ℵ-space. The fan space Sω1 is such a
space (see Example 1.8.7).
(2) There is an ℵ-space which has no σ -HCP strict k-network. Any uncountable
topological sum of the real number space R is such a space.
(3) There is a separable space with a σ -HCP strict k-network which is not normal.
The space X in Example 3.4.19(2) is such a space, because by Theorem 3.8.1,
X has a σ -HCP strict k-network.
246 3 Generalized Metric Spaces

Example 3.8.23 On mappings on ℵ-spaces.

(1) ℵ0 -spaces are not preserved by open compact mappings [418], because the
butterfly space (see Example 1.8.3) is an open compact image of an ℵ0 -space.
First, we show the butterfly space is an open image of an ℵ0 -space [331]. Suppose
X is the butterfly space and A = R × {0}. Let M = N ∪ {p} be the Michael space
(see Example 1.8.8). Define

Z = ((X − A) × N) ∪ (A × {p}) ⊂ X × M.

Let f = π1|Z : Z → X.
(23.1) f is an open mapping.
For each V ∈ τ (X × M), since N is dense in M,

π1 (V ∩ (B × N)) = π1 (V ∩ (B × M))

for each B ∈ τ (X). Let B = X − A. Then

π1 (V ∩ Z) = π1 (V ∩ ((X − A) × N)) ∪ π1 (V ∩ (A × {p}))

= π1 (V ∩ ((X − A) × M)) ∪ π1 (V ∩ (A × {p}))
= π1 (V ∩ ((X − A) × M)) ∪ π1 (V ∩ (X × {p})).

Since π1|X×{p} is an open mapping, f (V ∩ Z) ∈ τ (X). Thus f is an open

mapping.
(23.2) Z is an ℵ0 -space.
Obviously, Z is a regular space. Since both (X − A) × N and A × {p} are
separable metric spaces, to prove Z is an ℵ0 -space, we only need to prove
that if we set

K1 = K ∩ ((X − A) × N), K2 = K ∩ (A × {p})

for each K ∈ K (Z), then K1 , K2 ∈ K (Z). Since π2 (K) is a finite set,

π2 (K) ∩ N is a closed set in M, so K1 = K ∩ π2−1 (π2 (K) ∩ N) ∈ K (Z).
Since A × {p} is a closed set in Z, K2 ∈ K (Z).
Next, we show the space X is an open compact image of a subspace H of Z [418].
Take a sequence
 {Cn }n∈N of closed
subsets of X such that C1 = ∅, Cn ⊂ Cn+1 and
X − A = n∈N Cn . Let H = Z − n∈N (Cn × {n}). Then the space H is an ℵ0 -space
by (23.2). Let g = f|H : H → X. Obviously, g is onto X and each fiber of g is finite
(hence compact). To see g is open, let U be a basic open neighborhood of a point
z ∈ H. If z = (x, n) for some n ∈ N, then g(U) is open in X by (23.1). If z = (x, p),
then there are m ∈ N and F ∈ p such that U = (Bt(x, 1/m) × ({p} ∪ F)) ∩ H. Let
k = min F. Then g(U) = Bt(x, 1/m) − Ck is open in X. Thus g is open and compact.
3.8 ℵ-Spaces 247

(2) The property of having a σ -HCP strict k-network does not satisfy the perfect
preimage Gδ -diagonal theorem. 
Suppose Iα is a copy of I for each α < ω1 . Let X = α<ω1 Iα , Y = ω1 , and
give Y the discrete topology. Define f : X → Y by f (Iα ) = {α}. Then f is a perfect
mapping, X has a Gδ -diagonal and Y has a σ -HCP strict k-network. However, X has
no σ - HCP strict k-network.

Question 3.8.24 Is every regular space with a σ -HCP k-network a closed image of
an ℵ-space?

3.9 g-Metrizable Spaces

According to the basic problems on investigating generalized metric space, the inter-
est of researchers in g-metrizable spaces follows from its relationship with regular
spaces having a σ -discrete weak base or a σ -HCP weak base. The following ques-
tion was raised by Tanaka [457]: Is every regular space with a σ -HCP weak base a
g-metrizable space? In this section, we first give a positive answer to this question
and introduce some relevant results of it, then we discuss the mapping theorems of
g-metrizable spaces.
By the definition of weak bases, the following lemma is obvious.


Lemma 3.9.1 Let P = x∈X Px be a weak base in a space X. If a is the only
accumulation point of X, then Pa is a local base of a.

Lemma 3.9.2 Suppose P is a countably weakly HCP family in a space X. If P

is a family of sequential neighborhoods of a limit point of a nontrivial convergent
sequence in X, then P is finite.

Proof Suppose S is a nontrivial sequence in X converging to x and P is a family of

sequential neighborhoods of x. If P contains an infinite subfamily {Pn }n∈N , then S
is eventually in each Pn , so there is a subsequence {xn } of S such that xn ∈ Pn , and
hence {xn : n ∈ N} is a discrete closed set, a contradiction. Thus P is finite. 

Theorem 3.9.3 For every regular space X, the following are equivalent:
(1) X is a g-metrizable space.
(2) X has a σ -discrete weak base [127].
(3) X has a σ -HCP weak base [305].
(4) X is a g-first countable space with a σ -HCP k-network [265, 457].
248 3 Generalized Metric Spaces

Proof (2) ⇒ (1) ⇒ (3) is obvious.

(3) ⇒ (4). By Corollary 1.6.20 and Lemma 3.8.3, we only need to prove that
every
 space with a σ -HCP weak base is a g-first countable space. Suppose P =
n∈N n is a weak base for a space X such that each Pn is an HCP family. Denote
P
P = x∈X Bx , where Bx is a weak base at x. Fix x ∈ X. For each n ∈ N, let
Pn
= Pn ∩ Bx .
We may assume that x is an accumulation point of X. Then x is the limit point of
a nontrivial convergent sequence in X. Because infact, we can take a sequence {Vn }
of open sets in X such that V n+1 ⊂ Vn and {x} = n∈N Vn . For each P ∈ Pn
, x is an
accumulation point of P and there is x(P, n) ∈ (P − {x}) ∩ Vn . Let

Y = {x} ∪ {x(P, n) : P ∈ Pn
, n ∈ N}.

Then Y is a closed set in X, so Y has a σ -HCP weak base and x is the only accumulation
point in Y . By Lemma 3.9.1, Y has a σ -HCP base. By the Burke–Engelking–Lutzer
metrization theorem, Y is a metrizable subspace of X, so there is a nontrivial sequence
converging to x.
By Corollary 1.6.19 and Lemma 3.9.2, Pn
is finite. So Bx is countable. Thus X
is a g-first countable space.
(4) ⇒ (2). By Corollary 1.6.18, Sω1 is not a g-first countable space. Further by
Theorems 3.8.8, 3.8.4 and Proposition 1.6.21, X has a σ -discrete weak base. 

By Theorem 2.5.15 and Corollary 1.6.18, we can get the following metrization
theorem of g-metrizable spaces.

Corollary 3.9.4 ([425]) A space X is a metrizable space if and only if X is a g-

metrizable Fréchet–Urysohn space.
Question 3.9.5 ([126]) Is a space with a σ -locally finite closed weak base a regular
space?
Lemma 3.9.6 ([274]) If X is regular space with a point-countable cs∗ -network, then
X is a g-first countable space if and only if X is a sequential space contains no closed
copy of Sω .
Proof We only need to prove the sufficiency. Suppose P is a point-countable cs∗ -
network for X which is closed under finite intersections. For each x ∈ X, Let

Bx = {∪P
: P
∈ (P)<ω

x and ∪ P is a sequential neighborhood of x}.


Then Bx is countable. We prove B = x∈X Bx is a weak base for X.
Suppose G ⊂ X and for each x ∈ G, there is B ∈ Bx such that B ⊂ G. Then G is
a sequential neighborhood of every point in it, so G is a sequentially open set. Since
X is a sequential space, G ∈ τ .
On the other hand, suppose G ∈ τ and there is x ∈ G such that any element of
Bx is not contained in G. Define
3.9 g-Metrizable Spaces 249

P ∗ = {P ∈ (P)x : P ⊂ G} = {Pi }i∈N .


For each n ∈ N, let Fn = in Pi . Then Fn is not a sequential neighborhood of x.
Since P is a cs∗ -network, for each i ∈ N, there exist ni ∈ N and a sequence Ti
in X
converging to x such that Ti
⊂ Pni+1 − Fni and ni < ni+1 . Define

/ P and there is i ∈ N such that P ∩ Ti

= ∅} = {Qk }k∈N .
Q = {P ∈ P : x ∈

For each i ∈ N, choose a subsequence Ti of Ti
such that Ti ⊂ X − Qk . Then,
 ki
Qk ⊂ X − i>k Ti . Let
T = {x} ∪ (∪{Ti : i ∈ N}).

Then there does not exist any nontrivial sequence {xk } converging to some x
∈ X
and different xk is in different Ti . Because in fact, if x
= x, then there is i ∈ N such
that Pi contains a subsequence {xkm } of {xk }, so there exist m, j ∈ N such that j  i
and xkm ∈ Tj , and hence xkm ∈ Pi ∩ (X − Fnj ) = ∅, a contradiction. If x
= x, then
there is Q ∈ Q such that Q contains infinitely many terms of {xk }, which contradicts
the choice of Ti . Since X is a sequential space, T is a closed copy of Sω in X, a
contradiction.
In summary, X is a g-first countable space. 

Theorem 3.9.7 ([307]) A regular space X is a g-metrizable space if and only if X

is a k-space with a σ -HCP k-network and X contains no closed copy of Sω .

Proof We only need to prove the sufficiency. Since X contains no closed copy of
Sω , X contains no closed copy of Sω1 either. By Theorem 3.8.8, X is an ℵ-space. By
Lemma 3.9.6 and Theorem 3.9.3, X is a g-metrizable space. 

Question 3.9.8 ([299]) Is every regular space with a σ -compact-finite weak base
g-metrizable?

Theorem 3.9.9 ([292]) For every space X, the following are equivalent:
(1) X has a σ -compact-finite weak base.
(2) X is a k-space with a σ -weakly HCP weak base.
(3) X is a g-first countable space with a σ -weakly HCP weak base.

Proof (1) ⇒ (2). We only need to note that in a k-space, every compact-finite family
is weakly HCP.
We prove (2) ⇒ (3) ⇒ (1). Denote the set of all isolated points of X by I(X).
Suppose X is a k-space with a σ -weakly HCP weak base. By Lemma 2.5.4, X has
a σ -compact-finite network, so every compact set in X is metrizable, and hence X is
a sequential space. It follows that every accumulation point of X is a limit point of
some nontrivial convergent sequence in X. By Lemma 3.9.2, X is a g-first countable
space.
250 3 Generalized Metric Spaces


Let P = n∈N Pn be a weak base for the g-first countable space X, where each
Pn is a weakly HCP family and Pn ⊂ Pn+1 . For each x ∈ X, let

Hx = {P ∈ P : P is a sequential neighborhood of x}.

If x ∈ I(X), then {x} ∈ P, so I(X) is an Fσ -discrete subspace of X. For every n ∈ N,

P ∈ Pn , define

Dn = {x ∈ X : |(Pn )x |  ℵ0 },
Wn (P) = (P − Dn ) ∪ {x ∈ X − I(X) : P ∈ Hx }.

Then Wn (P) ⊂ P. If x ∈ X − I(X) and P ∈ Hx ∩ Pn , then Wn (P) is a sequential

neighborhood of x. Because in fact, suppose a sequence {xi } converges to x. Then
{xi } is eventually in P. By Lemma 2.5.4, ({xi : i ∈ N} ∪ {x}) ∩ Dn is finite, so {xi } is
eventually in (P−Dn )∪{x} ⊂ Wn (P), and hence Wn (P) is a sequential neighborhood
of x. Let
Wn = {Wn (P) : P ∈ Pn }.

Then Wn is point-finite. Because in fact, for each x ∈ X, since {P − Dn : P ∈ Pn } is

point-finite, we may assume x ∈ X − I(X). By Lemma 3.9.2, Hx ∩ Pn is finite, so
Wn is point-finite, and hence Wn is a point-finite and weakly HCP family. By Lemma
2.5.4, Wn is compact-finite.
For each x ∈ X, if x ∈ I(X), take Bx = {{x}}; if x ∈ X − I(X), take

Bx = {Wn (P) : n ∈ N, P ∈ Hx ∩ Pn }.

Then x∈X Bx is a weak base for X. Because first, for every x ∈ X − I(X) and
U, V ∈ Bx , there exist n, m ∈ N, P ∈ Hx ∩ Pn and Q ∈ Hx ∩ Pm such that
U = Wn (P) and V = Wm (Q), so there exist k  max{n, m} and R ∈ Hx ∩ Pk such
that R ⊂ P ∩ Q, and hence Wk (R) ⊂ Wn (P) ∩ Wm (Q). Second, for every x ∈ G ∈ τ ,
we may assume x ∈ X − I(X). Then there exist n ∈ N and P ∈ Hx ∩ Pn such
that P ⊂ G, so x ∈ Wn (P) ⊂ P ⊂ G. Finally, let G be a subset of X satisfying that
for each x ∈ G, there is some B ∈ Bx such that B ⊂ G. Then G is a sequential
neighborhood of any point in it, and hence G is a sequentially open set, so G ∈ τ .
Thus Bx is a weak
 base at x.
In summary, x∈X Bx is a σ -compact-finite weak base for X. 

Corollary 3.9.10 ([292]) Every ℵ1 -compact space with a σ -weakly HCP weak base
has a countable weak base.

Proof Suppose P = n∈N Pn is a weak base for an ℵ1 -compact space X such that
each Pn is weakly HCP. Denote the set of all isolated points in X by I(X). For each
x ∈ X − I(X), let

Hx = {P ∈ P : P is a sequential neighborhood of x}.

3.9 g-Metrizable Spaces 251

If there exist n ∈ N and an uncountable subfamily {Pα }α<ω1 of Hx ∩ Pn , then

Pα ∩ U ∩ (X − {x}) = ∅ for each α < ω1 and any open neighborhood U of x,
because X − {x} is not a closed set in X. By using the inductive method, there is a
subset {xα : α < ω1 } of X such that xα ∈ Pα ∩ (X − {xβ : β < α}) ∩ (X − {x}),
then {xα : α < ω1 } is an uncountable discrete closed set in X, a contradiction. So
Hx ∩ Pn is countable, and hence X is a g-first countable space.
By Theorem 3.9.9, X has a σ -compact-finite weak base. By Lemma 3.2.14, every
point-finite and weakly HCP family in X is countable, thus X has a countable weak
base. 

In the following, we characterize g-metrizable spaces in terms of sequences of

covers. Similar to Definition 2.10.4, we can give the definitions of locally finite weak
developments, HCP weak developments and closure-preserving weak developments.

Theorem 3.9.11 ([287]) For every regular space X, the following are equivalent:
(1) X is a g-metrizable space.
(2) X has a locally finite weak development.
(3) X has an HCP weak development.

Proof We only need to prove (3) ⇒ (1) ⇒ (2).

Suppose X has a sequence {Un } of HCP covers such that {Un } is a weak
development for X. For any sequence {xn } with xn → x ∈ V ∈ τ (X), there
is m ∈ N such that st(x, Um ) ⊂ V . By Corollary 1.6.19, there is i  m such
that {xn : n  i} ⊂ st(x, Um ). Further by Lemma 2.5.4, there is U
∈ (Um )<ωx such
that {xn : n  i} ⊂ ∪U
, so there exist U∈ Um and a subsequence {xnj } of {xn } such
that {x} ∪ {xnj : j ∈ N} ⊂ U ⊂ V . Thus n∈N Un is a cs∗ -network for X. By Lemma
3.8.3 and Theorem 3.9.3, X is a g-metrizable space.
Suppose P = n∈N Pn is a weak base for the g-metrizable  space X such that
each Pn is a discrete family of closed sets in X. Denote P = x∈X Bx , where Bx
is a weak base at x. For each n ∈ N, let

Un = {x ∈ X : Bx ∩ Pn = ∅},
Un = {Un } ∪ Pn .

Then Un is a locally finite cover of X. Note that, if Bx ∩ Pn = ∅, then

x ∈ X − ∪{P ∈ Pn : x ∈
/ P} ⊂ st(x, Un ),

we know st(x, Un ) is always a weak neighborhood of x. Thus {Un } is a weak devel-

opment for X. 
252 3 Generalized Metric Spaces

Question 3.9.12 Is every regular space with a closure-preserving weak development

g-metrizable?
By means of σ -mappings, we have the following characterizations of g-metrizable
spaces.

Theorem 3.9.13 For every regular space X, the following are equivalent:
(1) X is a g-metrizable space.
(2) X is a quotient, π and σ -image of a metric space [244].
(3) X is a quotient, compact and σ -image of a metric space.
(4) X is a compact-covering, quotient, compact and σ -image of a metric space [294].

Proof (1) ⇒ (4). Suppose X is a g-metrizable space. By the proof and notations of
(1) ⇒ (2) in Theorem 3.9.11, X has a locally finite weak development {Un }. For
every n ∈ N and K ∈ K (X), define Γn = {α : Pα ∈ Pn , Pα ∩ K = ∅}. Then Γn is
a finite set. For each α ∈ Γn , let

Kα = Pα ∩ K, and Kn = K − Kα .
α∈Γn

Then {Kα : α∈ Γn } ∪ {Kn } is a closed cover of K. For each x ∈ Kn , take a sequence
{xi } in K − α∈Γn Kα converging to x. If there is P ∈ Bx ∩ Pn , then P is a weak
neighborhood of x, so there exist m ∈ N and α ∈ Γn such that xm ∈ P ∩ Kα ,
a contradiction. Hence Bx ∩ Pn = ∅, it follows that x ∈ Un , thus Kn ⊂ Un .
Consequently, Un is a cfp-cover of X.
For each i ∈ N, let Ui = {Uα : α ∈ Λi }. Using the notations of Proposition 2.9.5,
we can define a metric space M and a compact-covering π -mapping f : (M, d) → X.
By Theorem 2.10.6, f is a quotient, compact mapping. We prove f is also a σ -
mapping. For every (αi ) ∈ M and n ∈ N, let B(α1 , . . . , αn ) = {(γi ) ∈ M : γi =
αi , i  n}. Then
f (B(α1 , . . . , αn )) = Uαi .
in

 
Because in fact, it is obvious that f (B(α1 , . . . , αn )) ⊂ in Uαi . If z ∈ in Uαi ,
then there is β
= (βi
) ∈ M such that f (β
) = z. For each k ∈ N, if k  n, let

βk = αk ; if k > n, let βk = βk . Then z ∈ k∈N Uβk . Let β = (βk ). Then  β ∈ M and

f (β) = z, so z ∈ f (B(α1 , . . . , αn )). That means f (B(α1 ,
. . . , αn )) = in Uαi . Since
{B(α1 , . . . , αn ) : (αi ) ∈ M, n ∈ N} is a base for M and in Ui is locally finite, f is
a σ -mapping.
(2) ⇒ (1). Suppose f : M → X is a quotient, π and σ -mapping, where M is a
metric space. By the mapping lemma (see Proposition 2.1.12) and Theorem 2.11.12,
X is an ℵ-space. By Theorem 2.9.9, X is a g-first countable space. By Theorem 3.9.3,
X is a g-metrizable space. 
3.9 g-Metrizable Spaces 253

Question 3.9.14 Is every g-metrizable space a quotient finite-to-one image of a

metrizable space?
By Example 3.1.19, g-metrizable spaces are not preserved by perfect mappings.
By Proposition 3.8.14 and Theorem 3.9.7, we have the following mapping theorems
of g-metrizable spaces.

Theorem 3.9.15 Suppose f : X → Y is a closed mapping, X is a g-metrizable space

and Y is a regular space. Then the following are equivalent:
(1) Y is a g-metrizable space.
(2) Y is a g-first countable space [265, 457].
(3) Y contains no closed copy of Sω [307].

Theorem 3.9.16 ([287]) g-metrizable spaces satisfy the perfect preimage Gδ -

diagonal theorem.

Proof Suppose f : X → Y is a perfect mapping, X is a regular space with a Gδ -

diagonal and Y is a g-metrizable space. By Theorem 3.8.20, X is an ℵ-space. If X
is not a g-metrizable space, then by Theorem 3.9.7, X contains a closed subspace T
homeomorphic to Sω . Let h = f|T : T → f (T ). Then h is a perfect mapping, and
hence f (T ) is a Fréchet–Urysohn closed subspace of Y . By Corollary 3.9.4, f (T )
is a metrizable space. By Theorem 2.2.12, T is a metrizable space, a contradiction.
Therefore, X is a g-metrizable space. 

Example 3.9.17 (1) There is a k-space with a point-countable cs-network which

contains no closed copy of Sω but is not a sequential space, the compactifica-
tion βN is such a space.
(2) There is an ℵ0 -space containing no closed copy of Sω which is not a k-space, the
Michael space (see Example 1.8.8) is such a space.
(3) There is a space with a σ -weakly HCP weak base which is not a k-space, the
space X in Example 2.5.16 is such a space.

3.10 Several Open Questions

In this section, to facilitate the interested readers’ research, we give a list of open
questions mentioned in the above three chapters.1 For the convenience of check,
questions still use the original serial number.

1 The following questions in the first edition have been solved: Question 1.6.20 (see Example

1.6.23), Question 2.7.7(1) (see Example 3.1.25), Question 2.7.20 (see the remark after Corollary
2.7.19), Question 3.1.16 (see Example 3.1.23), Question 3.2.29 (see Theorem 3.2.23), Question
3.4.1(5) (see Proposition 3.5.18), Question 3.8.3 (see Theorem 3.9.3). The following questions in the
second edition also have been solved: Question 2.6.5 (see Theorem 2.6.1(3)), Question 2.10.10 (see
Theorem 2.10.7), Question 3.2.4 (see Proposition 3.2.4), Question 3.5.21 (see Example 3.5.21(4)),
Question 3.8.14 (see Example 3.8.23(1)).
254 3 Generalized Metric Spaces

Question 1.4.21 ([58]) Is it true that each quasi-developable regular space is c-semi-
stratifiable?

Question 1.5.19 ([58, 319]) Is there a regular c-semi-stratifiable space that is not a
σ -space?

Question 1.7.8 ([77]) Is every wΔ-space with a point-countable base developable?

Question 1.7.9 ([125]) Is every quasi-developable β-space developable?

Question 1.7.10 ([202]) Is every wM-space with a Gδ -diagonal metrizable?

Question 1.8.16 ([334]) Is there any semi-metrizable hereditarily Lindelöf regular

space X such that X 2 is not a normal space?

Question 1.8.17 ([162]) Is there any semi-metrizable hereditarily Lindelöf regular

space Z which is not a σ -space?

Question 2.0.1 ([3]) Which spaces can be represented as images of “nice” (e.g.
metric or zero-dimensional, etc.) spaces under “nice” continuous mappings?

Question 2.0.2 ([3]) Which spaces can be mapped onto “nice” spaces by “nice”
mappings?

Question 2.0.3 ([31]) Under what circumstances can each space of a given class A
be mapped onto a space of a class B by means of a mapping belonging to a class F ?

Question 2.0.4 ([31]) If the class F (A ) of spaces are images of spaces of type A
by mappings of type F , then what internal properties can the spaces belonging to
F (A ) have?

Question 2.0.5 ([31]) Let F (A , B) denote the class of mappings whose domain
is a member of the class A and whose range is a member of the class B. Let H
be some other class of mappings. What are the properties of mappings of the class
F (A , B) ∩ H ?

Question 2.0.6 ([31]) What topological properties are preserved by these and other
mappings?

Question 2.1.17 ([341]) Characterize the class of spaces Y such that every closed
onto mapping f : X → Y is a countably bi-quotient mapping.

Question 2.1.18 ([460]) Let f : X → Y be a closed mapping. Find additional

conditions for X or Y , such that, with these conditions, ∂f −1 (y) has nice properties
for each y ∈ Y ?
3.10 Several Open Questions 255

Question 2.2.23 ([162]) Is every normal Moore space submetrizable?

Question 2.2.24 ([158]) Is the product of a strict p-space and a wΔ-space a wΔ-
space?

Question 2.3.15 Characterize the class of spaces such that every countably bi-
quotient image of these spaces is metrizable.

Question 2.4.7 ([465]) Is every first countable connected space an open image of a
connected metric space?

Question 2.4.17 ([411]) Is every quotient L-mapping from a space with a point-
countable base onto a space of pointwise countable type a countably bi-quotient
mapping?

Question 2.5.20 ([310]) Does every space with a σ -weakly HCP base have the point
Gδ -property?

Question 2.5.21 ([312]) Is every regular k-space with a σ -compact-finite k-network

a meta-Lindelöf space?

Question 2.5.25 ([448]) Find characterizations for subspaces of countable products

of Lašnev spaces.

Question 2.9.16 (1) Is every semi-metrizable space a sequence-covering π -image

of a metric space?
(2) Is every semi-metrizable space a compact-covering π -image of a metric space?

Question 2.10.11 ([200]) Is every symmetrizable space with a σ -point-finite cs-

network a quotient compact image of a metric space?

Question 2.10.23 ([31]) Is the class MOBI preserved by perfect mappings?

Question 2.10.24 ([98]) Does every space with a point-countable base belong to
MOBI2 ?

Question 3.1.5 (1) Is every β-space with a point-countable pseudo-Fq-base a semi-

stratifiable space?
(2) Is every β-space with a point-countable base a semi-stratifiable space?
(3) Is every collectionwise normal β-space with a point-countable base metrizable
[45]?

Question 3.1.21 ([82]) Are spaces with a point-countable pseudo-base preserved by

perfect mappings?
256 3 Generalized Metric Spaces

Question 3.1.22 ([167]) Are pseudo-open s-images of metric spaces preserved by

perfect mappings or pseudo-open s-mappings?

Question 3.1.24 ([416]) Can every space be represented as a closed image of a space
with a point-countable k-network?

Question 3.1.26 ([311]) Does every regular Fréchet–Urysohn space with a point-
countable k-network have a compact-countable k-network?

Question 3.2.5 ([290]) If X has a sequence {Fn } of HCP closed covers of X such
that any sequence {xn } with xn ∈ C(Fn , x) for some x ∈ X has an accumulation
point, is X a Σ ∗ -space?

Question 3.2.8 Is every space with a σ -cushioned (mod k)-network a submetacom-

pact space?

Question 3.2.12 ([450]) Does every Σ-space have a σ -discrete closed quasi-(mod k)-
network?

Question 3.2.24 ([386]) Is every perfect Σ -space a Σ-space?

Question 3.2.33 (1) Does every strong Σ ∗ -space satisfy the decomposition theorem
of type Lindelöfness [399]?
(2) Does every perfect strong Σ-space satisfy the decomposition theorem of type
compactness [97]?

Question 3.2.35 (1) If X × I is a strong Σ ∗ -space, then is X a strong Σ-space?

(2) Are Σ-spaces finitely productive [158]?

Question 3.3.8 ([354]) Does every metacompact σ -space satisfy uniform (G)?

Question 3.3.13 ([97]) Does every perfect preimage of a σ -space satisfy the decom-
position theorem of type compactness?

Question 3.3.19 (1) Does every quasi-perfect preimage of a countably paracompact

semi-stratifiable space satisfy the decomposition theorem of type countable com-
pactness [162]?
(2) Does every semi-stratifiable regular space satisfy the decomposition theorem
of type compactness?

Question 3.3.22 ([82, 110]) Is every semi-stratifiable Fréchet–Urysohn space a

closed image of a semi-metrizable space?

Question 3.3.23 ([458]) Suppose f : X → Y is a closed mapping. If X is a sym-

metrizable space and Y is g-first countable, then is Y a symmetrizable space?
3.10 Several Open Questions 257

Question 3.4.2 ([139]) Does every k-semi-stratifiable space have a σ -closure-

preserving k-network?

Question 3.4.5 Is every k-semi-stratifiable space with a point-countable (resp. σ -

locally countable) weak base a g-metrizable space?

Question 3.4.18 Does every k-semi-stratifiable space satisfy the perfect preimage Gδ -
diagonal theorem?

Question 3.4.21 Is every normal k-and σ -space a paracompact space?

Question 3.4.22 Is every regular k-space with a star-countable k-network a k-semi-

stratifiable space?

Question 3.4.23 Is every regular pseudo-open compact image of a k-semi-stratifiable

space a σ -space?

Question 3.5.1 Ceder’s questions2 :

(1) Is every M2 -space an M1 -space?
(4) Are M1 -spaces closed hereditary?
(5) Are M1 -spaces preserved by closed mappings or peripherally compact closed
mappings?
(10) Are M1 -spaces preserved by perfect mappings [88]?

Question 3.6.8 ([348]) Characterizes closed images of developable spaces in terms

of σ -HCP pair-networks.

Question 3.6.10 ([107]) Does every space satisfying open (G) have a point-countable
base?

Question 3.6.14 ([82, 156]) Are wΔ-spaces preserved by perfect mappings?

Question 3.6.17 ([97]) Does every perfect preimage of a Moore space satisfy the
decomposition theorem of type compactness?

Question 3.6.20 ([461]) Is every finite-to-one pseudo-open image of a Moore space

a developable space?

Question 3.7.8 (1) Is every M -space an M ∗ -space [365]?

(2) Is every wM-space an M-space [202]?
(3) Is every wM-space an M -space?
(4) Is every wM-space an Σ -space?

Question 3.8.12 ([379]) Characterize ℵ-spaces by means of bases.

2 Only the problems unsolved are listed and please refer to the remark after Proposition 3.5.18.
258 3 Generalized Metric Spaces

Question 3.8.24 Is every regular space with a σ -HCP k-network a closed image of
an ℵ-space?

Question 3.9.5 ([126]) Is a space with a σ -locally finite closed weak base a regular
space?

Question 3.9.8 ([299]) Is every regular space with a σ -compact-finite weak base
g-metrizable?

Question 3.9.12 Is every regular space with a closure-preserving weak development

g-metrizable?

Question 3.9.14 Is every g-metrizable space a quotient finite-to-one image of a

metrizable space?
Appendix A
Characterizations of Several
Covering Properties

This appendix is prepared for the convenience of text reading. We mainly introduce
characterizations and relevant mapping theorems of five covering properties used
in the text: paracompactness, metacompactness, subparacompactness, submetacom-
pactness and meta-Lindelöfness. For a systematic introduction of covering properties,
we recommend reading “Covering Properties” [83] or “Selected Topics in General
Topology” [212].
In this appendix, no separation axiom is assumed for spaces except it is specially
mentioned and all mappings are assumed to be continuous and onto. We first recall
some basic terminologies. Suppose X is a topological space and U = {Uα }α<γ is
a cover of X , where γ is an ordinal number. U is called a well-monotone cover of
X if Uα ⊂ Uβ for every α < β < γ . U is called a directed cover of X if for every
α, β < γ , there is δ < γ such that Uα ∪ Uβ ⊂ Uδ . For families V and W of subsets
of X , V is said to be a partial refinement of W if for each V ∈ V , there is W ∈ W
such that V ⊂ W . V said to be a refinement of W if V covers X and V is a partial
refinement of W . For other notations and terminologies, please refer to Sect. 1.1 of
the text.

A.1 Paracompact Spaces

Definition A.1.1 A space X is a called a paracompact space [112] if every open

cover of X has a locally finite open refinement. X is called a countably paracom-
pact space [116, 224] if every countable open cover of X has a locally finite open
refinement.

The following characterizations of paracompact spaces given by A.H. Stone and

Michael are well-known. Let U , V be covers of a space X . V is a star-refinement
of U if {st(V, V ) : V ∈ V } is a refinement of U .

© Atlantis Press and the author(s) 2016 259

S. Lin and Z. Yun, Generalized Metric Spaces and Mappings,
Atlantis Studies in Mathematics 6, DOI 10.2991/978-94-6239-216-8
260 Appendix A: Characterizations of Several Covering Properties

Theorem A.1.2 For every regular space X , the following are equivalent:
(1) X is a paracompact space.
(2) Every open cover of X has an open star-refinement [440].
(3) Every open cover of X has a σ -discrete open refinement [325].
(4) Every open cover of X has a σ -cushioned open refinement [327].
(5) Every open cover of X has a closure-preserving closed refinement [326].

Corollary A.1.3 ([326]) (The Michael theorem) T2 paracompactness is invariant

under closed mappings.

Proposition A.1.4 ([173]) Every perfect preimage of a paracompact space is a

paracompact space.

Proof Suppose f : X → Y is a perfect mapping and Y is a paracompact space.

For each open cover U of X and each y ∈ Y , there is U y ∈ U <ω such that
f −1 (y) ⊂ ∪U y . Take an open neighborhood Vy of y for each y ∈ Y such that
f −1 (Vy ) ⊂ ∪U y . Then the open cover {Vy : y ∈ Y } of Y has a locally finite open
refinement {W y : y ∈ Y } such that W y ⊂ Vy . So { f −1 (W y ) ∩ U : y ∈ Y, U ∈ U y }
is a locally finite open refinement of U . Thus X is a paracompact space. 

Definition A.1.5 Suppose U , V are covers of a space X . V is called a local star-

refinement of U [219] if, for each x ∈ X , there are an open neighborhood G of x
and U ∈ U such that st(G, V ) ⊂ U . V is called a pointwise W -refinement of U
[479] if, for each x ∈ X , there is U  ∈ U <ω such that (V )x is a partial refinement
of U  . V is called a local W -refinement of U [480] if, for each x ∈ X , there exist
an open neighborhood G of x and U  ∈ U <ω such that (V )G is a partial refinement
of U  .

Lemma A.1.6 ([219]) Suppose U is an open cover of a space X . Consider the

following conditions:
(1) U F
has a closure-preserving closed refinement F such that F ◦ covers X .
(2) U F
has an interior-preserving open local star-refinement.
(3) U F
has an interior-preserving open local W -refinement.
Then (3) ⇒ (2) ⇒ (1). If U is an interior-preserving open cover of X , then (1) ⇒
(3).

Proof (3) ⇒ (2) is obvious.

(2) ⇒ (1). Suppose V is an interior-preserving open local star-refinement of U F .
Let

F(U  ) = {x ∈ X : st(x, V ) ⊂ ∪U  }, U  ∈ U <ω ;

F = {F(U  ) : U  ∈ U <ω }.

Then F is a closure-preserving closed refinement of U F

and F ◦ covers X .
Appendix A: Characterizations of Several Covering Properties 261

Now suppose U is an interior-preserving open cover of X . Let F be a closure-

preserving closed refinement of U F such that F ◦ covers X . Define

Wx = (∩(U )x ) ∩ (X − ∪(F − (F )x )), x ∈ X ;

W = {Wx : x ∈ X }.

Then W is an interior-preserving open cover of X . For each x ∈ X , there is F ∈ F

such that x ∈ F ◦ , then there is U  ∈ U <ω such that F ⊂ ∪U  , and hence (W ) F ◦ is
a partial refinement of U  . Thus W is a local W -refinement of U . 
Lemma A.1.7 Suppose F is a σ -closure-preserving family of closed sets in a count-
ably paracompact space X . If F ◦ covers X , then F has a closure-preserving closed
refinement H such that H ◦ covers X .

Proof Denote F = n∈N Fn , where each Fn is closure-preserving family of closed
sets in X . For each n ∈ N, let

Pn = ∪Fn◦ and P = {Pn : n ∈ N}.

Then P has a locally finite open refinement R = {Rα : α ∈ Λ}. For each α ∈ Λ,
take n α ∈ N such that Rα ⊂ Pn α . Let

Hα = {F ∩ R α : F ∈ Fn α },
H = ∪{Hα : α ∈ Λ}.

Then H is a closure-preserving closed refinement of F such that H ◦ cover X . 

Lemma A.1.8 ([219]) Suppose {Un } is a sequence of open covers of a space X . If
each Un+1 is a pointwise (resp. local) W -refinement of Un , then U1 has a σ -point-
finite (resp. σ -locally finite) open refinement.

Proof Denote U1 = {Uα : α < γ }. For each U ∈ n∈N Un , let

α(U ) = min{α < γ : U ⊂ Uα }.

For every n > 1 and U ∈ Un , Un−1 is said to have the property Φ(U ) if α(U  ) =
α(U ) whenever U ⊂ U  ∈ Un−1 . Define

Wn = {U ∈ Un : Un−1 has the property Φ(U )}.


(8.1) n>1 Wn covers X .
For every n > 1 and x ∈ X , define αn = sup{α(U ) : x ∈ U ∈ Un }. Since
Un+1 is a pointwise W -refinement of Un , αn+1  αn < γ , so there exist β < γ and
k  3 such that αn = β when n  k − 1. Since Uk+1 is a pointwise W -refinement
of Uk , there is U ∈ (Uk )<ω
x such that (Uk+1 )x is a partial refinement of U . Pick
U ∈ U such that α(U  )  α(U ) whenever U  ∈ U . Since (Uk+1 )<ω x is a partial
262 Appendix A: Characterizations of Several Covering Properties

refinement of U , αk+1  α(U ) and α(U )  αk , so α(U ) = β. Because αk−1 = β,

for every U  ∈ (Uk−1 )x , α(U  )  β = α(U ), and if U ⊂ U  , then α(U  )  α(U ),
so α(U  ) = α(U ), it follows that Uk−1 has the property Φ(U ), and x ∈ U ∈ Wk .
Now, for each n > 1, define

Vn,α = ∪{W ∈ Wn : α(U ) = α}, α < γ ;

Vn = {Vn,α : α < γ }.

Obviously, n>1 Vn is an open refinement of U1 .
(8.2) Vn is a point-finite (resp. locally finite) family.
For each x ∈ X , take G = {x} (resp. take an open neighborhood G of x) and
<ω
U  ∈ Un−1 such that (Un )G partially refines U  . Let

Γ = {α(U  ) : U  ∈ U  },
Λ = {α < γ : G ∩ Vn,α
= ∅}.

Then Λ ⊂ Γ . Because in fact, if β ∈ Λ, then there is U ∈ Wn such that α(U ) = β

and U ∩ G
= ∅, so U ∈ (Un )G , and hence there is U  ∈ U  such that U ⊂ U  ,
it follows that α(U ) = α(U  ), consequently β ∈ Γ . Thus Vn is a point-finite (resp.
locally finite) family
 in X .
In summary, n>1 Vn is a σ -point-finite (resp. σ -locally finite) open refinement
of U1 . 

Theorem A.1.9 ([219, 317]) For every space X , the following are equivalent:
(1) X is a paracompact space.
(2) Every well-monotone open cover of X has a locally finite open refinement.
(3) Every interior-preserving directed open cover of X has an interior-preserving
open local star-refinement.
(4) Every interior-preserving directed open cover of X has a σ -closure-preserving
closed refinement F such that F ◦ covers X .
(5) Every directed open cover of X has a closure-preserving closed refinement F
such that F ◦ covers X .

Proof (1) ⇒ (5). Suppose U is a directed open cover of X . Let V be a locally finite
open refinement of U . Then V is an interior-preserving open local W -refinement of
U . By Lemma A.1.6, U F has a closure-preserving closed refinement F such that
F ◦ covers X . Since U is a directed cover, F is also a refinement of U .
(5) ⇒ (4) is obvious.
(4) ⇒ (3). By Lemmas A.1.7 and A.1.6, we only need to prove that X is a
countably paracompact space. Suppose U = {Un :  n ∈ N} is an open cover of
X . Then the interior-preserving
 directed open cover { kn Uk : n ∈ N} of X has
aclosed refinement n∈N Fn , such that, Fn is a closure-preserving family and
◦
n∈N Fn covers X . Define
Appendix A: Characterizations of Several Covering Properties 263

R0 = ∅;
⎧ ⎫
⎨   ⎬
Rn = ∪ F ∈ Fk : F ⊂ Uk , n ∈ N.
⎩ ⎭
kn kn

Then {Un − Rn−1 : n ∈ N} is a locally finite open refinement of U . Thus X is a

countably paracompact space.
(3) ⇒ (2). Suppose U is a well-monotone open cover of X . By Lemma A.1.6,
there is a sequence {Un } of interior-preserving open covers of X such that U1 =
U and Un+1  is a local W -refinement of Un . By Lemma A.1.8, U has an open
refinement n∈N Vn such that each Vn is locally finite in X . By Lemma A.1.6, X
satisfies (4), so X is a countably paracompact space, and hence {∪Vn : n ∈ N}
 a locally finite open refinement {Wn : n ∈ N} such that Wn ⊂ ∪Vn . Then
has
n∈N {Wn ∩ V : V ∈ Vn } is a locally finite open refinement of U .
(2) ⇒ (1). For each cardinal number γ , denote the following assumption by P(γ ):
every open cover of X with cardinality γ has a locally finite open refinement. Suppose
γ is an infinite cardinal number and for each λ < γ , P(λ) is true.
 If U is an open cover
of X with cardinality γ , denote U = {Uα : α < γ }, then { βα Uβ : α < γ } is a
well-monotone open cover of X , and hence it has a locally finite open refinement W .
For each W ∈ W , take α(W ) < γ such that W ⊂ ∪{Uβ : β  α(W )}. For each
α < γ , let

Pα = ∪{W ∈ W : α(W ) > α},

Pα = {Pα } ∪ {Uβ : β  α}.

By the induction hypothesis, Pα has a locally finite open refinement Hα . Define

R(W ) = {W ∩ H : H ∈ Hα(W ) and H ⊂ Uα for some α  α(W )}, W ∈ W ;

R = ∪{R(W ) : W ∈ W }.

Then R is a locally finite family of open sets in X , and for each x ∈ X , there is
W ∈ (W )x such that α(W  )  α(W ) whenever W  ∈ (W )x . Let β = α(W ) and
take H ∈ Hβ such that x ∈ H . Since x ∈ / Pβ and Hβ refines Pβ , there is α  β
such that H ⊂ Uα , so x ∈ W ∩ H ∈ R(W ) ⊂ R, and hence U has a locally finite
open refinement R. Thus P(γ ) holds. 

In the following, we investigate the relationships between paracompactness and

normality.

Definition A.1.10 A space X is said to be a collectionwise normal space [62] if, for
every discrete family {Fα }α∈Λ of closed sets in X , there is a disjoint family {Uα }α∈Λ
of open sets in X such that Fα ⊂ Uα . X is called an expandable space [225] if, for
every locally finite family {Fα }α∈Λ of closed sets in X , there is a locally finite family
{Uα }α∈Λ of open sets in X such that Fα ⊂ Uα .
264 Appendix A: Characterizations of Several Covering Properties

Collectionwise normality can be simply described as every discrete family of

closed sets has a disjoint open expansion. Also, expandability can be simply described
as every locally finite family of closed sets has a locally finite open expansion.

Theorem A.1.11 Every T2 paracompact space is a both a collectionwise normal

space and an expandable space [225].

Proof Suppose X is a T2 paracompact space. Let F = {Fα }α∈Λ be a discrete family

of closed sets in X . For each α ∈ Λ, let Vα = X −∪(F −{Fα }). Then the open cover
{Vα }α∈Λ of X has an open star-refinement V . Let Uα = st(Fα , V ). Then {Uα }α∈Λ is
a disjoint open expansion of F . Thus X is a collectionwise normal space.
Let F = {Fα }α∈Λ be a locally finite family of closed sets in X . For each Γ ∈ Λ<ω ,
define
V (Γ ) = X − ∪{Fα : α ∈ Λ − Γ }.

For each x ∈ X , let Γx = {α ∈ Λ : x ∈ Fα }. Then Γx ∈ Λ<ω and x ∈ V (Γx ). Let

V be a locally finite open refinement of the open cover {V (Γ ) : Γ ∈ Λ<ω } of X .
For each α ∈ Λ, let Uα = st(Fα , V ). Then Fα ⊂ Uα and for each x ∈ X , there exist
an open neighborhood V of x and n ∈ N such that (V )V = {Vi : i  n}. For each
i  n, take Γi ∈ Λ<ω such that Vi ⊂ V (Γi ). Let Γ = {α ∈ Λ : V ∩ Uα
= ∅}.
Then Γ ⊂ in Γi , so Γ ∈ Λ<ω . Hence {Uα }α∈Λ is a locally finite family of open
sets in X . Thus X is an expandable space. 

We did not use the axiom of T2 separation in the proof for expandability above.
Therefore, for every locally finite family {Fn }n∈N of closed sets in a countably para-
compact space X , there is a locally finite family {Un }n∈N of open sets in X such that
Fn ⊂ Un .

Proposition A.1.12 Suppose {Fα }α∈Λ is a discrete family of closed sets in a collec-
tionwise normal space X . Then there is a discrete family {G α }α∈Λ of open sets in X
such that Fα ⊂ G α .

Proof By the collectionwise normality of X , there is a family {Uα }α∈Λ of disjoint

open sets in X such that Fα ⊂ Uα . Let F = α∈Λ Fα , U = α∈Λ Uα . By the
normality of X , there is a continuous function f : X → I such that f (F) ⊂ {1} and
f (X − U ) ⊂ {0}. For each α ∈ Λ, define G α = Uα ∩ {x ∈ X : f (x) > 1/2}. Then
{G α }α∈Λ is a discrete family of open sets in X and Fα ⊂ G α . 

Proposition A.1.13 ([204]) A space X is a countably paracompact space if and

only if for every increasing open cover {G n }n∈N
 of X , there is a sequence {Hn } of
closed sets in X such that Hn ⊂ G n and X = n∈N Hn◦ .

Proof Suppose {G n }n∈N is an increasing open cover of the countably paracompact

 open cover {Vn }n∈N of X such that Vn ⊂ G n .
space X . Then there is a locally finite
For
 each n ∈ N, let H n = X − m>n Vm . Then Hn is a closed in X and Hn ⊂
V
mn m ⊂ G n . For each x ∈ X , there is an open neighborhood Ux of x which only
Appendix A: Characterizations of Several Covering Properties 265

meetsfinitely many Vn . Let k = max{n ∈ N : Ux ∩ Vn

= ∅}. Then Ux ⊂ Hk . So
X = n∈N Hn◦ .
On theother hand, suppose U = {Un }n∈N is an open cover of X . For each n ∈ N,
let G n = mn Um . Then {G n }n∈N is an increasing open cover of X , and
 hence there
is a sequence{Hn } of closed sets in X such that Hn ⊂ G n and X = n∈N Hn◦ . Let
Vn = Un − m<n Hm . Then {Vn }n∈N is an open refinement of U . For each x ∈ X ,
take n ∈ N such that x ∈ Hn◦ . Then Hn◦ ∩ Vm = ∅ whenever m > n. So {Vn }n∈N is
locally finite, and hence X is a countably paracompact space. 

A.2 Metacompact Spaces

Definition A.2.1 A space X is called a metacompact space (or weakly paracompact

space) [20] if every open cover of X has a point-finite open refinement. X is called
a countably metacompact space [155] if every countable open cover of X has a
point-finite open refinement.

Using the same method as Proposition A.1.4 and Lemma A.1.6, we can prove the
following Proposition A.2.2 and Lemma A.2.3 respectively.

Proposition A.2.2 Every perfect preimage of a metacompact space is a metacom-

pact space.

Lemma A.2.3 Suppose U is an interior-preserving open cover of a space X . Then

U F has a closure-preserving closed refinement if and only if U has an interior-
preserving open pointwise W -refinement.

Theorem A.2.4 ([219, 420]) For every space X , the following are equivalent:
(1) X is a metacompact space.
(2) Every well-monotone open cover of X has a point-finite open refinement.
(3) For every open cover U of X , U F has a closure-preserving closed refinement.

Proof By Lemma A.2.3 we get (1) ⇒ (3). By an argument same as that in the proof
of (2) ⇒ (1) in Theorem A.1.9 and replacing “locally finite” with “point-finite”, we
can prove (2) ⇒ (1). Now we prove (3) ⇒ (2).
Suppose U is a well-monotone open cover of X . By Lemma A.2.3, there is
a sequence {Un } of interior-preserving open covers of X such that U1 = U and
Un+1  is a pointwise W -refinement of Un . By Lemma A.1.8, U has an open refine-
ment n∈N Vn such that  each Vn is a point-finite family of subsets in X . Then the
directed open cover { kn (∪Vk ) : n ∈N} of X has a closure-preserving
 closed
refinement {Fn : n ∈ N} such that Fn ⊂ kn (∪Vk ). So n∈N {V − k<n Fk : V ∈
Vn } is a point-finite open refinement of U . 

By (1) ⇔ (3) of Theorem A.2.4, we have the following mapping theorem for
metacompact spaces.
266 Appendix A: Characterizations of Several Covering Properties

Corollary A.2.5 ([479] (The Worrell theorem) Metacompactness is invariant

under closed mappings.

Theorem A.2.6 ([218, 479]) For every space X , the following are equivalent:
(1) X is a metacompact space.
(2) Every open cover of X has a point-finite refinement H such that x ∈ st(x, H )◦
for each x ∈ X .
(3) Every open cover of X has an open pointwise W -refinement.

Proof (1) ⇒ (2) is obvious.

(2) ⇒ (3). Suppose each open cover U of X has a point-finite refinement H
satisfying the assumption of (2). For each H ∈ H , take U H ∈ U such that H ⊂ U H .
For each x ∈ X , define

Vx = st(x, H )◦ ∩ (∩{U H : H ∈ (H )x }).

Then {Vx : x ∈ X } is an open pointwise W -refinement of U .

(3) ⇒ (1). Suppose U is an open cover of X . Take a sequence {Un } of open covers
of X such that U1 = U and Un+1 is a pointwise W -refinement of Un . By Lemma
A.1.8, U has an open refinement n∈N Vn such that each Vn is point-finite in X . Let
W be an open pointwise W -refinement of {∪Vn : n ∈ N}. For each n ∈ N, let

Fn = {x ∈ X : st(x, W ) ⊂ ∪{∪Vi : i  n}}.

   
Then F n ⊂ in (∪Vi ) and X = n∈N F n . Thus n∈N {V − i<n F i : V ∈ Vn } is
a point-finite open refinement of U . 

Corollary A.2.7 ([219] Every pseudo-open compact image of a paracompact space

is a metacompact space.

Proof Suppose f : X → Y is a pseudo-open compact mapping and X is a para-

compact space. If U is an open cover of Y , then the open cover f −1 (U ) of X has a
locally finite open refinement V , so f (V ) is a point-finite refinement of U satisfying
y ∈ st(y, f (V ))◦ for each y ∈ Y . It follows from Theorem A.2.6 Y is a metacompact
space. 

Similar to the proof of Proposition A.1.13, we can prove the following Proposition
A.2.8.

Proposition A.2.8 ([155, 204]) A space X is a countably metacompact

space if and
only if for each decreasing sequence {Fn } of closed sets in X with
n∈N Fn = ∅,
there is a sequence {G n } of open sets in X such that Fn ⊂ G n and n∈N G n = ∅.
Appendix A: Characterizations of Several Covering Properties 267

A.3 Subparacompact Spaces

Definition A.3.1 [75] A space X is a subparacompact space if every open cover of

X has a σ -discrete closed refinement.
Proposition A.3.2 ([74]) Suppose f : X → Y is a perfect mapping and X is a T2
space. If Y is a subparacompact space, then X is a subparacompact space.
Proof Suppose U is an open cover of X . For each y ∈ Y, f −1 (y) is a compact
set in X . By Lemma 3.2.8 in the text, there is O y ∈ τ (X ) such that f −1 (y) ⊂ O y
and U|O y has a finite closed refinement in O y . Since f is a closed mapping, there is
Hy ∈ τ (Y ) such that y ∈ Hy and f −1 (Hy ) ⊂ O y . Then U| f −1 (Hy ) has a finite closed
refinement F y in f −1 (Hy ).Let H = {Hy : y ∈ Y }. Since Y is a subparacompact
space, H has a refinement i∈N Wi such that each Wi = {Wi,y : y ∈ Y } is a discrete
family of closed sets and Wi,y ⊂ Hy  . For each i ∈ N, let Pi = { f −1 (Wi,y ) ∩ F :
y ∈ Y, F ∈ F y }. We prove P = i∈N Pi is a σ -discrete closed refinement of
U . Obviously, P is a refinement of U . For each i ∈ N, since f −1 (Wi ) is discrete
and F y is finite for each y ∈ Y , Pi is σ -discrete. For every y ∈ Y , F ∈ F y and
x ∈ X − ( f −1 (Wi,y ) ∩ F), define

X − f −1 (Wi,y ), x ∈ X − f −1 (Wi,y ),
Lx =
f −1 (Hy ) − F, x ∈ f −1 (Wi,y ) − F.

Then L x is an open neighborhood of x in X and L x ∩ ( f −1 (Wi,y ) ∩ F) = ∅. Thus

f −1 (Wi,y ) ∩ F is closed, and hence X is a subparacompact space. 
Theorem A.3.3 ([75, 217]) For every space X , the following are equivalent:
(1) X is a subparacompact space.
(2) Every open cover U of X has a sequence {Un } of open refinements such that
for each x ∈ X , there exist n ∈ N and U ∈ U with st(x, Un ) ⊂ U .
(3) Every open cover of X has a σ -locally finite closed refinement.
(4) Every open cover of X has a σ -closure-preserving closed refinement.
(5) Every open cover of X has a σ -cushioned refinement.
Proof It is obvious that (1) ⇒ (3) ⇒ (4) ⇒ (5), so we only need to prove (1) ⇒
(2) ⇒ (5) ⇒ (1). 
(1) ⇒ (2). Suppose U is an open cover of X . Let n∈N Fn be a closed refinement
of U such that each Fn is discrete. For every n ∈ N and F ∈ Fn , take U F ∈ U
such that F ⊂ U F . Define

E n = ∪Fn ,
VF = U F − (E n − F),
Vn = {VF : F ∈ Fn } ∪ {U − E n : U ∈ U }.

Then the sequence {Vn } of open covers of X satisfies the requirement of (2).
268 Appendix A: Characterizations of Several Covering Properties

(2) ⇒ (5). Suppose U is an open cover of X . Let {Un } be a sequence of open

refinements of U satisfying the assumption of (2). Denote U = {Uα : α ∈ Λ}. For
every n ∈ N and α ∈ Λ, define

F(n, α) = {x ∈ X : st(x, Un ) ⊂ Uα },
Fn = {F(n, α) : α ∈ Λ}.
 
Then n∈N Fn is a refinement of U . For each Λ ⊂ Λ, suppose x ∈ X − α∈Λ Uα .
∈ N, α ∈ Λ and y ∈ F(n,
If n   α), then x ∈ / st(y, Un ), so y ∈ / st(x, Un ), and hence
x∈ / α∈Λ F(n, α). Thus, n∈N Fn is a σ -cushioned refinement of U .
(5) ⇒ (1). For the narrative brevity, we use the following notation: for every
n, k ∈ N and s = (i 1 , i 2 , . . . , i k ) ∈ Nk , denote s ⊕ n = (i 1 , i 2 , . . . , i k , n).
Let U = {Uα : α < γ } be an open cover of X . If k ∈ N and s ∈ Nk , then
we make an induction on k as follows: take an open refinement U (s) of U and a
σ -cushioned refinement F (s) of U (s). For each t ∈ N, define

Vα (t) = Wα (t) = Uα , α < γ ;

U (t) = {Vα (t) : α < γ } ∪ {Wα (t) : α < γ }.

Now assume that U (s) is an open refinement of U with the form

U (s) = {Vα (s) : α < γ } ∪ {Wα (s) : α < γ }.

Let F (s) be a σ -cushioned refinement of U (s) with the form

F (s) = {Hα (s ⊕ n) : α < γ , n ∈ N} ∪ {K α (s ⊕ n) : α < γ , n ∈ N},

where {Hα (s ⊕ n) : α < γ } and {K α (s ⊕ n) : α < γ } are cushioned in {Vα (s) : α <
γ } and {Wα (s) : α < γ } respectively. Let Uα (s) = Vα (s) ∪ Wα (s). Define

Vα (s ⊕ n) = Uα (s) − ∪{Hβ (s ⊕ n) ∪ K β (s ⊕ n) : α
= β < γ },
Wα (s ⊕ n) = Uα (s) ∩ (∪{Uβ (s) : α < β < γ })
− ∪{Hβ (s ⊕ n) ∪ K β (s ⊕ n) : β < α},
U (s ⊕ n) = {Vα (s ⊕ n) : α < γ } ∪ {Wα (s ⊕ n) : α < γ }.

(3.1) U (s ⊕ n) is acover of X .
For each x ∈ X − α<γ Vα (s ⊕ n), let δ = min{α < γ : x ∈ Uα (s)}. Then


x∈ Hβ (s ⊕ n) ∪ K β (s ⊕ n),
δ<β<γ

so there is β > δ such that x ∈ Uβ (δ), and hence x ∈ Wδ (s ⊕ n). Thus, U (s ⊕ n)

covers X .
Appendix A: Characterizations of Several Covering Properties 269

For every s ∈ Nk , n ∈ N, and α < γ , define

Tα (s ⊕ n) = Hα (s ⊕ n) − ∪{Vβ (s) : α
= β < γ }.

Since {Hα (s ⊕ n) : α < γ } is cushioned in {Vα (s) : α < γ }, {Tα (s ⊕ n) : α < γ } is

X and Tα (s ⊕ n) ⊂ Uα .
a discrete family of closed sets in 
(3.2) {Tα (s ⊕ n) : α < γ , s ∈ k∈N Nk , n ∈ N} is a cover of X  .
For each x ∈ X , let δ = min{β < γ : x ∈ H β (h) ∪ K β (h), h ∈ k∈N N }. Then
k

there exist t ∈ k∈N N and n ∈ N such that x ∈ Hδ (t ⊕ n) ∪ K δ (t ⊕ n). Take m ∈ N

k

and σ < γ such that

x ∈ Hσ (t ⊕ n ⊕ m) ∪ K σ (t ⊕ n ⊕ m) ⊂ Vσ (t ⊕ n) ∪ Wσ (t ⊕ n).

Thenx ∈/ Vσ (t ⊕ n) ∪ Wσ (t ⊕ n) whenever α > δ, so σ = δ. While it follows from

x∈/ β>δ Uβ (t ⊕ n) that x ∈ / Wδ (t ⊕ n ⊕ m), and hence x ∈ Hδ (t ⊕ n ⊕ m). Since
x ∈ Hδ (t ⊕ n) ∪ K δ (t ⊕ n), x ∈
/ Vβ (t ⊕ n) whenever β
= δ, so

x ∈ Hδ (t ⊕ n ⊕ m) − Vβ (t ⊕ n) ⊂ Tδ (t ⊕ n ⊕ m).
β
=δ

In summary, U has a σ -discrete closed refinement, thus X is a subparacompact

space. 
Corollary A.3.4 ([75]) Subparacompactness is invariant under closed mappings.

A.4 Submetacompact Spaces

Definition A.4.1 A space X is called a submetacompact space [217] (or θ -refinable

space [481]) if, every open cover of X has a sequence {Un } of open refinements
satisfying that for each x ∈ X , there is n ∈ N such that x is only in finitely many
elements of Un . The sequence {Un } is said to be an open θ -refinable sequence of
this cover.
A sequence {Un } of covers of X is said to be a pointwise W -refining sequence of
a cover U of X if for each x ∈ X , there exist U  ∈ U <ω and n ∈ N such that (Un )x
is a partial refinement of U  .
Theorem A.4.2 ([481]) Every submetacompact countably compact space is a com-
pact space.
Proof Suppose X is a submetacompact countably compact space. Let U be an open
cover of X . Then U has an open θ -refinable sequence {Un }. For each n ∈ N, denote
Un = {Uα : α ∈ Λn }. For different elements α1 , . . . , αk in Λn , let

Fn (α1 , . . . , αk ) = ∩{Uαi : i  k} − ∪{Uα : α ∈ Λn − {α1 , . . . , αk }}.

270 Appendix A: Characterizations of Several Covering Properties

Since {Fn (α) : α ∈ Λn } is a discrete family in X , it is a finite set, so there is a

finite subfamily Un,1 of Un covering {Fn (α) : α ∈ Λn }. By the same reason, there
is a finite subfamily Un,2 of Un covering ∪{Fn (α1 , α2 ) − ∪Un,1 : α1 , α2 ∈ Λn }, and
 we can obtain a sequence {Un,k } consisting of finite subfamilies
continue in this way,
of
 nU such that k∈N Un,k covers ∪{Fn (α1 , . . . , αk ) : αi ∈ Λn , i  k ∈ N}. Then
U
n,k∈N n,k is a countable subcover of n∈N Un , and hence U has a finite subcover.
Thus X is a compact space. 
In the following, we introduce the characterizations of submetacompact spaces
obtained by Junnila.
Lemma A.4.3 If every (interior-preserving) open cover of a space X has an
(interior-preserving) open pointwise W -refining sequence, then every (interior-
preserving) open cover of X has a sequence {Un } of open refinements such that
(∗) for each x ∈ X , there is a strictly increasing sequence {n i } in N such that
(Uni+1 )x is a partial refinement of Uni .
Proof For an (interior-preserving) open cover U of X , there is an (interior-
preserving) open pointwise W -refining sequence {H1,n }, and we denote it by
U → {H1,n }n1 . By the same reason, we have, for each m ∈ N,
⎛ ⎞

Hm,m ∧ ⎝ Hi,m+1 ⎠ → {Hm+1,n }nm+1 .
im

Let Un = Hn,n . Then for every x ∈ X and m ∈ N, there is n > m such that (Un )x
is a partial refinement of Um . Thus the (interior-preserving) open cover U of X has
a sequence {Un } of (interior-preserving) open refinements satisfying (∗). 
Lemma A.4.4 If a cover B of a space X has a sequence {Un } of open refinements
satisfying (∗), then B has an open θ -refinable sequence.
Proof
 Let B = {Bα : α < γ }, where γ is a cardinal number. For each U ∈
 k∈N U k
, define α(U ) = min{β < γ : U ⊂ Bβ }. For every n ∈ N and V ∈
k∈N U , we say that Un has the property Φ(V ) if {α(U ) : V ⊂ U ∈ Un } ⊂
k

{α(V )}. For every n, k ∈ N, define

Vn,k = {V ∈ Uk : Un has the property Φ(V )},

L n,k = {x ∈ X : (Uk )x is a partial refinement of Un }.

For every n > 1 and s ∈ Nn , let


Ls = L s(i),s(i+1) , where s(i) is the i-th coordinate of s(i);
i<n
  

Hs = x ∈ L s : st(x, Us(n) ) ⊂ ∪ Vs(i),s(i+1) .
i<n
Appendix A: Characterizations of Several Covering Properties 271


(4.1) {Hs : s ∈ n>1 Nn } is a cover of X .
For each x ∈ X , there exist a strictly increasing sequence {tn } in N and a sequence
{Fn }, where each Fn is a finite family of sets in X , such that, Fn ⊂ (Utn )x and
(Utn+1 )x partially refines Utn . For each n > 1, let

Pn = {F ∈ Fn : F
⊂ ∪(∪{Vti ,ti+1 : i < n})}.

Then Pn = ∅ for some n > 1. Because otherwise, for each n > 1, let αn =
max{α(P) : P ∈ Pn }. Since Pn+1 partially refines Pn , αn+1  αn , so there is
k > 3 such that αk = αk−1 = αk−2 . Take P ∈ Pk such that α(P) = αk . We prove
that Utk−1 has the property Φ(P). In fact, if P ⊂ U ∈ Utk−1 , then U ∈ (Utk−1 )x ,
so there is F ∈ Fk−2 such that U ⊂ F, and hence P ⊂ U ⊂ F, it follows that
αk = α(P)  α(U )  α(F)  αk−2 , as a consequence, α(U ) = α(P), thus Utk−1
has the property Φ(P). It follows that P ∈ Vtk−1 ,tk , which contradicts the fact that
P ∈ Pk . So there is n > 1 such that Pn = ∅. Let t = (t1 , t2 , . . . , tn+1 ). Then
x ∈ Ht and hence (4.1) is proved.
For each k ∈ N, define

Wn,k,α = ∪{V ∈ Vn,k : α(V ) = α}, α < γ ;

Wn,k = {Wn,k,α : α < γ }.

(4.2) Wn,k is a point-finite family in L n,k .

For each x ∈ L n,k , there is F ∈ Un<ω such that (Uk )x refines F . Let Λ = {α(F) :
F ∈ F }. To prove (Wn,k )x is finite, we only need to verify {α < γ : x ∈ Wn,k,α } ⊂ Λ.
Suppose α < γ and x ∈ Wn,k,α . Then there is V ∈ Vn,k such that x ∈ V and
α(V ) = α, so V ∈ (Uk )x , and hence there is F ∈ F such that V ⊂ F. Since Un
has the property Φ(V ), α(V ) = α(F), and hence α ∈ Λ. Thus, (4.2) is proved.
For every n > 1 and s ∈ Nn , define

Ws = Ws(i),s(i+1) ,
i<n
Us = {U ∈ Us(n) : U
⊂ ∪Ws },
Hs = Ws ∪ Us .

Then Hs is an open refinement of B and

(4.3) Hs is a point-finite family in H
s.
For each x ∈ Hs , st(x,Us(n) ) ⊂ ∪( i<n Vs(i),s(i+1) ) = ∪Ws , so x ∈ / ∪Us , and
hence (Hs )x = (Ws )x = i<n (Ws(i),s(i+1) )x . By (4.2),
 (H s ) x is finite.
By a combination of (4.1) and (4.3), {Hs : s ∈ n>1 Nn } is an open θ -refinable
sequence of B. 

Lemma A.4.5 If every well-monotone open cover of a space X has an open θ -

refinable sequence, then every open cover of X has an open pointwise W -refining
sequence.
272 Appendix A: Characterizations of Several Covering Properties

Proof For each cardinal number γ , the following proposition is expressed as P(γ ):
every open cover of X with cardinality γ has an open pointwise W -refining sequence.
We prove P(γ ) is true for every cardinal number γ by a method of transfinite induc-
tion.
Suppose γ is an infinite cardinal number and for each λ < γ , P(λ) is true.
If U isan open cover of X with cardinality γ , denote U = {Uα : α < γ }.
Then { βα Uβ : α < γ } is a well-monotoneopen cover of X , and hence it has
an open θ -refinable sequence {Vn }. For V ∈ n∈N Vn , take α(V ) < γ such that
V ⊂ ∪{Uβ : β  α(V )}. For every α < γ , n ∈ N, let

Pα,n = ∪{V ∈ Vn : α(V ) > α},

Pα,n = {Pα,n } ∪ {Uβ : β  α}.

By the induction hypothesis, Pα,n has an open pointwise W -refining sequence

{Wα,n,k }. For every n, i ∈ N, define

Fn,i = {x ∈ X : |(Vn )x |  i},


Gn = Fn,i .
i∈N


Then Fn,i is a closed set in X and X = n∈N G n . For each x ∈ G n , let

α(x, n) = max{α(V ) : V ∈ (Vn )x }.

Take Wn,k (x) ∈ (Wα(x,n),n,k )x , and let

⎛ ⎞

Hn,k (x) = ⎝ Wn,i (x)⎠ ∩ (∩(Vn )x ),
ik

Hn,k = {Hn,k (x) : x ∈ G n } ∪ {X − Fn,k }.

(5.1) Hn,k is a cover of X .

Because for each x ∈ X , x ∈ Hn,k (x) whenever x ∈ Fn,k .
(5.2) {Hn,k } is a pointwise W -refining sequence of U .
For each x ∈ X , take n ∈ N such that x ∈ G n . Let

Λ = {α(V ) : V ∈ (Vn )x }.

<ω
Then for each α ∈ Λ, there exist kα ∈ N and Fα ∈ Pα,n such that (Wα,n,kα )x is a
partial refinement of Fα . Define
Appendix A: Characterizations of Several Covering Properties 273

F = Fα ∩ U ,
α∈Λ
k = max{kα : α ∈ Λ} + |(Vn )x |.

Then (Hn,k )x is a partial refinement of F . In fact, let H ∈ (Hn,k )x . Then x ∈ Fn,k ,

so there is y ∈ G n such that H = Hn,k (y). Denote α = α(y, n) and take V ∈ (Vn ) y
such that α(V ) = α. Then x ∈ Hn,k (y) ⊂ V and α ∈ Λ. Let j = kα . Since j  k,
Hn,k (y) ⊂ Wn, j (y), so there is F ∈ Fα such that Wn, j (y) ⊂ F. Since y ∈ F − Pα,n ,
F ∈ U , and hence H ⊂ F ∈ F . Since F is finite, {Hn,k } is a pointwise W -refining
sequence of U .
Thus, every open cover of X has an open pointwise W -refining sequence. 
Lemma A.4.6 If every interior-preserving directed open cover of a space X has a
σ -closure-preserving closed refinement, then every interior-preserving open cover
of X has an open θ -refinable sequence.
Proof Suppose U is an interior-preserving open cover of X . Then U F is aninterior-
preserving directed open refinement of X . So U F has a closed refinement n∈N Fn
such that each Fn is closure-preserving. For every x ∈ X and n ∈ N, let

Vn (x) = ∩(U )x − ∪{F ∈ Fn : x ∈

/ F},
Vn = {Vn (x) : x ∈ X }.

Then Vn is an interior-preserving open refinement of U . We prove {Vn } is a pointwise

W -refining sequence of U . For each x ∈ X , take n ∈ N and F ∈ Fn such that
x ∈ F, then there is U  ∈ U <ω such that F ⊂ ∪U  . Let y ∈ X and x ∈ Vn (y).
Then y ∈ F. Thus there is U ∈ U  such that y ∈ U , so Vn (y) ⊂ U , and hence (Vn )x
is a partial refinement of U  . By Lemmas A.4.3 and A.4.4, U has an open θ -refinable
sequence. 
Lemma A.4.7 If an open cover U of a space X has an open θ -refinable sequence,
then there is a σ -closure-preserving closed cover F of X satisfying that for each
x ∈ X , there exist F ∈ F and U  ∈ (U )<ωx such that x ∈ F ⊂ ∪U .


Proof Suppose {Vn } is an open θ -refinable sequence of U . For every n, k ∈ N, let

Fn,k = {x ∈ X : |(Vn )x |  k}.

Then Fn,k is a closed set in X . Define

Fn (V ) = X − ∪(Vn − V ), V ∈ Vn<ω ;
Fn = {Fn (V ) : V ∈ Vn<ω };
Fn,k = Fn|Fn,k .

Since Vn|Fn,k is point-finite, it is an interior-preserving open cover for the subspace

Fn,k of X , and hence Fn,k is a closure-preserving family of closed sets in X . Let
274 Appendix A: Characterizations of Several Covering Properties

F = ∪{Fn,k : n, k ∈ N}.

For each x ∈ X , take n, k ∈ N such that x ∈ Fn,k . Further take U  ∈ (U )<ω

x such
that (Vn )x is a partial refinement of U  . Then F = Fn ((Vn )x ) ∩ Fn,k ∈ F and
x ∈ F ⊂ ∪U  . 

Theorem A.4.8 ([217]) For every space X , the following are equivalent:
(1) X is a submetacompact space.
(2) Every open cover of X has an open pointwise W -refining sequence.
(3) Every well-monotone open cover of X has an open θ -refinable sequence.
(4) Every interior-preserving directed open cover of X has a σ -closure-preserving
closed refinement.
(5) Every directed open cover of X has a σ -closure-preserving closed refinement.
(6) For every open cover U of X , there is a σ -closure-preserving family F of closed
sets in X satisfying that for each x ∈ X , there exist F ∈ F and U  ∈ (U )<ω x
such that x ∈ F ⊂ ∪U  .

Proof By Lemma A.4.7, we get (1) ⇒ (6). (6) ⇒ (5) ⇒ (4) is obvious. (4) ⇒ (3)
can be obtain by Lemma A.4.6, and (3) ⇒ (2) follows from Lemma A.4.5. By
Lemmas A.4.3 and A.4.4, (2) ⇒ (1) holds. 

Corollary A.4.9 ([217] The Junnila theorem) Submetacompactness is invariant

under closed mappings.

Theorem A.4.10 ([431]) A space X is a paracompact space if and only if X is a

submetacompact expandable space.

Proof We only need to prove the sufficiency. Suppose X is a submetacompact

expandable space. Since every expandable space is a countably paracompact space,
to prove X is a paracompact space, we only need to prove every open cover of X has
a σ -locally finite open refinement. Suppose U is an open cover of X . Let {Ui } be an
open θ -refinable sequence of U . For each i ∈ N, denote Ui = {Uα : α ∈ Λi }. We
inductively construct a sequence {Vi, j } of families of open sets in X such that
Ui ;
(1) Vi, j is a locally finite partial refinement of
(2) if |(Ui )x |  m for some x ∈ X , then x ∈ jm Vi, j , where Vi, j = ∪Vi, j ;
(3) if x ∈ Vi, j , then |(Ui )x |  j.
First, let Vi,0 = ∅. Suppose for 0  j  n, we have constructed Vi, j satisfying
conditions (1)–(3). Let A = {A ⊂ Λi : |A| = n + 1}. For each A ∈ A , define
⎛ ⎞

FA = ⎝ X − Vi, j ⎠ ∩ (X − ∪{Uα : α ∈ Λi − A}).
jn

Then F = {FA : A ∈ A } is a family of closed sets in X . For each x ∈ X , if

|(Ui )x | < n + 1, then the neighborhood jn Vi, j of x disjoints with any element of
Appendix A: Characterizations of Several Covering Properties 275

F ; if |(Ui )x |  n+1, take A ∈ A such that x ∈ ∩{Uα : α ∈ A}, then ∩{Uα : α ∈ A}

is only possible to meet the element FA of F . So F is a discrete family of closed
sets in X . By the expandability of X , there is a locally finite family {G A : A ∈ A }
of open sets in X such that FA ⊂ G A . Let

Vi,n+1 = {G A ∩ (∩{Uα : α ∈ A}) : A ∈ A }.

Then Vi,n+1 is a family satisfying conditions (1)–(3).

Since {Ui } is an open θ -refinable sequence of U , i, j∈N Vi, j is a σ -locally finite
open refinement of U , and hence X is a paracompact space. 

Proposition A.4.11 ([155]) Every submetacompact space is a countably metacom-

pact space.

Proof Suppose U = {Un }n∈N is a countable open cover of a submetacompact space

X . Let {Un } be an open θ -refinable sequence of U . Let
 
V1 = U1 , and Vn = st(X − Ui , Ui ), n > 1.
i<n i<n

Then {Vn }n∈N is a point-finite open refinement of U , and hence X is a countably

metacompact space. 

A.5 Meta-Lindelöf Spaces

Definition A.5.1 [18] A space X is called a meta-Lindelöf space if every open cover
of X has a point-countable open refinement. If every subspace of X is a meta-Lindelöf
space, then X is said to be a hereditarily meta-Lindelöf space.

A cover U of a space X is called a well-ordered cover of X if U = {Uα }α∈Λ

and Λ is a well-ordered set. For a well-ordered cover U = {Uα }α∈Λ of X , denote
α = {x ∈ X :
αx = min{α ∈ Λ : x ∈ Uα } for each x ∈ X . If α ∈ Λ, then denote U
α ⊂ Uα .
αx = α}. Obviously, U

Theorem A.5.2 ([167]) For every space X , the following are equivalent:
(1) X is a hereditarily meta-Lindelöf space.
(2) Every well-ordered open cover {Uα }α∈Λ of X has a point-countable open refine-
ment V satisfying that for each x ∈ X , there is V ∈ V such that x ∈ V ⊂ Uαx .
(3) Every well-ordered open cover {Uα }α∈Λ of X has a point-countable open refine-
α ⊂ Vα ⊂ Uα for each α ∈ Λ.
ment {Vα }α∈Λ such that U

Proof (1) ⇒ (2). Suppose X is a hereditarily meta-Lindelöf space, and there exist
an ordinal number γ and a well-ordered open cover {Uα }α<γ of X which has no
point-countable open refinement satisfying the requirement of (2). We may assume
276 Appendix A: Characterizations of Several Covering Properties

γ is the least ordinal number satisfying the above 

property. For each α < γ , the
well-ordered open cover {Uβ }β<α of the subspace β<α Uβ has a point-countable
open refinement Vα satisfying the requirement of (2). If there is an ordinal number
α such that γ = α + 1, then Vα ∪ {Uα } is a point-countable open refinement of
{Uβ }β<γ satisfying the requirement of (2), and hence γ must be a limit ordinal
number. Suppose W is a point-countable open refinement of the open cover {Uα }α<γ
of X . For each W ∈ W , pick α(W ) < γ such that W ⊂ Uα(W ) . Define

V = {W ∩ V : W ∈ W , V ∈ Vα(W )+1 }.

Then V is a point-countable family of open sets in X . For each x ∈ X , take W ∈

(W )x . Then x ∈ Uα(W ) ⊂ ∪{Uβ : β < α(W ) + 1}, so there is V ∈ Vα(W )+1 such
that x ∈ V ⊂ Uαx , it follows that W ∩ V ∈ V and x ∈ W ∩ V ⊂ Uαx . Thus, V
is a point-countable open refinement of {Uα }α<γ satisfying the requirement of (2), a
contradiction.
(2) ⇒ (3). Suppose {Uα }α∈Λ is a well-ordered open cover of X . Let V be a
point-countable open refinement of {Uα }α∈Λ satisfying the requirement of (2). For
each α ∈ Λ, let

Vα = ∪{V ∈ V : V ⊂ Uα ; if β < α, then V

⊂ Uβ }.

Then {Vα }α∈Λ is a point-countable open cover of X satisfy the requirement of (3).
(3) ⇒ (1). Suppose Y is a subspace of X and W = {Wα }α<γ is a well-ordered
open cover of Y . Take a well-ordered open cover {Uα }αγ of X such that Uγ = X
and Uα ∩ Y = Wα when α < γ . Let {Vα }αγ be a point-countable open refinement
of {Uα }αγ satisfying the requirement of (3). If y ∈ Y , then there is α < γ such that
y∈U α ⊂ Vα , so {Vα ∩ Y }α<γ is a point-countable open refinement of W . Thus Y
is a meta-Lindelöf space, and hence X is a hereditarily meta-Lindelöf space. 

Corollary A.5.3 ([167] Hereditarily meta-Lindelöf spaces are preserved by

pseudo-open s-mappings.

Proof Suppose f : X → Y is a pseudo-open s-mapping and X is a hereditarily

meta-Lindelöf space. If Z is a subspace of Y , then f | f −1 (Z ) : f −1 (Z ) → Z is also
a pseudo-open s-mapping, and hence we only need to prove Y is a meta-Lindelöf
space. Let {Uα }α∈Λ be a well-ordered open cover of Y . Then { f −1 (Uα )}α∈Λ is a well-
ordered open cover of X , and hence it has a point-countable open refinement {Vα }α∈Λ
such that f −1 (Uα )∼ ⊂ Vα ⊂ f −1 (Uα ) for each α ∈ Λ. Then f (Vα ) ⊂ Uα and
α ) ⊂ f −1 (Uα )∼ ⊂ Vα . Let Wα = f (Vα )◦ for each α ∈ Λ. Since f is a
f −1 (U
pseudo-open s-mapping, {Wα }α∈Λ is a point-countable open refinement of {Uα }α∈Λ
such that U α ⊂ Wα ⊂ Uα for each α ∈ Λ. Thus Y is a meta-Lindelöf space. 
Appendix B
The Formation of the Theory of Generalized
Metric Spaces

In the 1900 Paris International Congress of Mathematicians, Hilbert [187] pointed

out at his famous speech titled “Mathematical problems” that as long as a branch of
science can make a lot of problems, it is full of vitality, while the lack of problems
indicates that the decline or suspension of independent development. The develop-
ment of general topology in the 20th century confirmed the Hilbert quotes.
In 1944, Dieudonné [112] introduced the concept of paracompactness. This is an
important symbol of general topology into the heyday. Subsequently, the theory of
topological spaces developed rapidly. The form of its representation is a variety of
topological properties found to adapt for different purposes, and the basic direction
is to give various generalizations of paracompactness and metrizability for solving
all kinds of problems. In virtue of these researches, the theory of generalized metric
spaces became an important topic of general topology since the 1960s. The book
named “Handbook of the History of General Topology” edited by Aull and Lowen
[46–48] has made detailed description on the history of general topology. In this book,
the chapter of “A history of generalized metrizable spaces” [193] written by Hodel
focused on clarifying the main achievements of the theory of generalized metric
spaces during the period of 1950–1980. This appendix try to clarify the formation
process and some development of this theory by taking problems as the clue. Due
to space limitations, we only focus on the description of the background of some
kinds of generalized metric properties introduced during the period of 1944–1976.
It should be noticed that the division of every period here is completely in order to
illustrate the sequence of events of the problems. As for the modern development of
the theory of generalized metric spaces and related topics, in addition to this book,
readers can refer to the reference literature listed at the end of this book, especially
the book called “Encyclopedia of General Topology” [175] edited by Hart, Nagata
and Vaughan.

© Atlantis Press and the author(s) 2016 277

S. Lin and Z. Yun, Generalized Metric Spaces and Mappings,
Atlantis Studies in Mathematics 6, DOI 10.2991/978-94-6239-216-8
278 Appendix B: The Formation of the Theory of Generalized Metric Spaces

B.1 A Historical Review

The central topic of general topology in early study is problems about metrization
and compactness [4].
As pointed out by Rudin [411], “The most fundamental theorems, those which are
part of every mathematician’s background, were proved at this time”. A lot of fun-
damental work has laid a solid foundation for general topology, and made it become
an independent branch of mathematics and played a positive role in promoting the
development of other disciplines of mathematics. The most important results are as
follows.
Theorem B.1.1 (The Urysohn–Tychonoff metrization theorem, 1925) Every regu-
lar space with a countable base is a metrizable space.
Theorem B.1.2 (The Urysohn extension theorem, 1925) A space X is a normal
space if and only if every real valued continuous function from a closed subspace of
X is continuously extendable over X .
Theorem B.1.3 (The Tychonoff theorem, 1935) Every Cartesian product of com-
pact spaces is compact.
Theorem B.1.4 (The Tychonoff compact extension theorem, 1935) A space is com-
pletely regular if and only if there is a compact extension of it.
There are four classes of spaces involving in above theorems and they are metriz-
able spaces, compact spaces, normal spaces and completely regular spaces. Due to
the widespread applications, the four classes of spaces became the main object con-
cerned by topologists at that time and thus achieved fruitful results. It also produced
a series of problems to be solved. The reasons for this are mainly from two aspects:
First, there are larger gaps among metrizable spaces, compact spaces and normal
spaces. It is necessary to find the classes of spaces with nice properties and between
metrizable spaces, compact spaces and normal spaces. Second, the main topics con-
cerned by topologists in this period are certain finiteness or countability of families
of sets. In which ways some uncountable situation should be discussed? For exam-
ple, in a paper published in 1925, Urysohn raised a question on seeking metrization
theorems for general spaces such that the Urysohn–Tychonoff metrization theorem
is a natural corollary of them.
The significance of the main early achievements in general topology is to serve as
a model for future development of general topology.1 The concepts defined in that
period or more early, such as the topological sum of topological spaces (Tietze 1923),
box products (Tietze 1923), Cartesian products (Tychonoff 1930), limits of inverse
systems (Lefschetz 1931), adjunction operator (Borsuk 1937), different kinds of com-
pact extensions [compactifications (Carathéodory 1913), Alexandroff compactifica-
tion (1924), Čech-Stone compactification (1937), Wallman compactification (1938)],

1 Data mainly come from R. Engelking [119].

Appendix B: The Formation of the Theory of Generalized Metric Spaces 279

several important classes of spaces [countably compact spaces (Fréchet 1906), com-
pact spaces (Vietoris 1921), Lindelöf spaces (Alexandroff and Urysohn 1929), sepa-
rable spaces (Fréchet 1906), CCC (Suslin 1920), locally compact spaces (Alexandroff
1921), metrizable spaces (Fréchet 1906), first countable spaces (Hausdorff 1914),
second countable spaces (Hausdorff 1914), Moore spaces (Moore 1916), developable
spaces (Alexandroff and Urysohn 1923), semi-metrizable spaces (Wilson 1931), con-
nected spaces (Hausdorff 1914), various axioms of separation], mappings of differ-
ent forms [continuous mappings (Fréchet 1910), homeomorphic mappings (Fréchet
1910), closed mappings (Hurewicz 1926), open mappings (Aronszajn 1931), quo-
tient mappings (Baer and Levi 1932), monotone mappings (Whyburn 1934)], the
topology of pointwise convergence and the compact-open topology in the sets of
continuous functions (Fox 1945) etc., have become important tools for the modern
general topology research.

B.2 The Foundation Laying Period

The spaces in this appendix are at least T2 spaces, τ and τ c represent the topology
and the set of closed subsets for a space respectively, and all mappings are assumed
to be continuous and onto. The hypothesis of separation properties for individual
space is slightly different of those in the original literature.

B.2.1 The Introduction of Paracompactness

In the research of the theory of topological spaces, the key breakthrough on the
restriction of finiteness or countability for families of sets in spaces is the wild
usages of locally finite (locally countable) families and point-finite (point-countable)
families.

Definition B.2.1 Suppose P is a family of sets in a space X . P is said to be locally

finite [5] (resp. locally countable) in X if every point of X has a neighborhood which
only meets at most finitely (resp. countably) many elements of P; P is said to be
point-finite (resp. point-countable) in X if every point of X only belongs to at most
finitely (resp. countably) many elements of P.

Definition B.2.2 A space X is called a paracompact space [112] if every open cover
of X has a locally finite open refinement.

It was established by Alexandroff, in modern terminology, that every separable

metrizable space is paracompact [5]. “In the late 1940s and early 1950s, general
topology had a strong upsurge brought on by the definition of paracompactness”
[411]. Compactness ⇒ paracompactness ⇒ normality. In 1940, Tukey [469] intro-
duced fully normal spaces which are similar to paracompact spaces by the concept of
280 Appendix B: The Formation of the Theory of Generalized Metric Spaces

normal coverings and proved that every metrizable space is a fully normal space. In
1948, Stone [440] proved that the class of fully normal spaces is equivalent to the class
of paracompact spaces, and hence every metrizable space is a paracompact space.
The class of paracompact spaces is a class between the class of metrizable spaces
or compact spaces and the class of normal spaces. The properties of paracompact
spaces were concerned about by many topologists.
Every paracompact countably compact space is a compact space and this proposi-
tion implies the following general question: What kinds of countably compact spaces
are compact spaces? Arens and Dugundji [20] introduced metacompact (or weakly
paracompact) spaces by point-finite families, and proved that every metacompact
countably compact space is a compact space.

Definition B.2.3 A space X is called a metacompact space [20] if every open cover
of X has a point-finite open refinement.

The topological properties defined by open covers and their refinements, such
as paracompactness, metacompactness etc., are collectively referred to as covering
properties.

B.2.2 The Metrization Problems

In view of the role played by the metric spaces in many fields of mathematics, seeking
metrization theorems for general topological spaces is of great significance.

Definition B.2.4 Suppose P is a family of sets in a space X . P is said to be discrete

in X [62] if each point in X has a neighborhood which only meets at most one element
of P. Let Φ be a property of families. Then the family P in X is called a σ -Φ family
if P the union of countably many families in X with the Φ property.

Both σ -discrete families and σ -locally finite families are generalizations of count-
able families. By these concepts, Bing [62], Nagata [370] and Smirnov [428] obtained
the following excellent metrization theorem in general topology.

Theorem B.2.5 (The Bing-Nagata-Smirnov metrization theorem) A regular space

X is metrizable if and only if X satisfies one of the following conditions:
(1) X has a σ -discrete base.
(2) X has a σ -locally finite base.

Just like this, in 1964, S. Wang [474] for the first time gave a characterization of
ωμ -metrizable spaces as follows: a regular space X is ωμ -metrizable if and only
if X has a ωμ -base. Morita and Hanai [363] and Stone [441] proved independently
that metrizability is invariant under perfect mappings. Several years later, various
generalizations of metrizability were obtained. By excitation of the Hanai-Morita-
Stone theorem, topologists widely investigated the metrization problem for images
Appendix B: The Formation of the Theory of Generalized Metric Spaces 281

of metrizable spaces and the topological properties preserved by perfect mappings.

For example, in 1964 Y. Liu and L. Liu [315] proved that an adjunction space of two
metrizable spaces is metrizable if and only if it is first countable.
The obtaining of several metrization theorems for general spaces does not declare
that the metrization problems are completely solved. In some sense, the metriza-
tion theorems of general spaces is only a foundation to get metrization theorems
of spaces with specific topological properties, because the metrization problems for
spaces with specific topological properties are far from resolved and more important
in real life applications. Many classic metrization problems, such as the Alexandroff
problem that whether every perfectly normal topological manifold is metrizable
posed in 1923 [6], and the question whether every normal Moore space is a metriz-
able space raised by Jones [214], have been bothering and motivating the general
topologists contemporary. Just the normal Moore space conjecture dominated the
study of metrization problems and promoted the progress of related fields in gen-
eral topology for half a century. In his famous paper on metrization of topological
spaces [62], Bing defined collectionwise normality and give a partial answer to the
normal Moore space conjecture. Around this conjecture, Alexandroff [2] introduced
the concept of uniform bases.

Definition B.2.6 A base B for a space X is said to be a uniform base [2] if, for
every x ∈ U ∈ τ, {B ∈ B : x ∈ B
⊂ U } is finite.

Alexandroff made a conjecture that every normal space with a uniform base is
a metrizable space. Heath [177] gave another form of the Alexandroff conjecture
that every metacompact normal Moore is a metrizable space. As a generalization of
spaces with a uniform base, Arhangel’skiı̌ [27] introduced the concept of BCO (i.e.
base of countable order) spaces.

Definition B.2.7 A base B for a space X is called a BCO [27] if for each x ∈ X ,
{Bi } is a neighborhood base of x in X whenever {Bi } is a decreasing sequence of
sets in B containing x. A space with a BCO is called a BCO space.

Every developable space is a BCO space and every paracompact BCO space is
a metrizable space. However, a collectionwise normal BCO space may not be a
metrizable space. The another generalization of spaces with a uniform base is spaces
with a σ -point-finite base [27]. Although a paracompact space with a σ -point-finite
base may not be a metrizable space, every collectionwise normal perfect space with
a σ -point-finite base is a metrizable space. The following classic problem posed by
Heath [191, 384] is relevant to this result: Is every perfectly normal paracompact
space with a point-countable base metrizable? Todorčevic̀ [468] gave a negative
answer to this problem in 1991.
The great enthusiasm of topologists for paracompactness could not be separated
form a series of basic characterizations of paracompact spaces obtained in the 1950s.
The beautiful characterizations of paracompactness by means of discrete families,
locally finite families, closure preserving families and cushioned families given by
Michael [325–327] is a breakthrough in the study of paracompactness.
282 Appendix B: The Formation of the Theory of Generalized Metric Spaces

Definition B.2.8 Suppose P is a family of sets in a space X . P is called a closure-

preserving family [326] in X if, for any P  ⊂ P, ∪P  = ∪P  . A cover B of X is
called a cushioned refinement of P [327] if, for each B ∈ B, there is PB ∈ P such
that ∪B  ⊂ ∪{PB : B ∈ B  } for any B  ⊂ B.

Theorem B.2.9 For every space X , the following are equivalent:

(1) X is a paracompact space.
(2) Every open cover of X has a locally finite closed refinement.
(3) Every open cover of X has a closure-preserving closed refinement.
(4) Every open cover of X has a cushioned refinement.
If further assume X is a regular space, then they are equivalent to the following:
(5) Every open cover of X has a σ -discrete open refinement.
(6) Every open cover of X has a σ -locally finite open refinement.
(7) Every open cover of X has a σ -closure-preserving open refinement.
(8) Every open cover of X has a σ -cushioned open refinement.

A direct corollary of Theorem B.2.9 is that paracompactness is invariant under closed

mappings [326]. This is the beginning of investigation of the problem if some partic-
ular covering properties are invariant under closed mappings. By the enlightenment
of Michael’s characterizations of paracompactness and the Bing-Nagata-Smirnov
metrization theorem, Ceder [92] introduced Mi -spaces (i = 1, 2, 3) in a classic
paper titled “Some generalizations of metric spaces”.

Definition B.2.10 Suppose B is a family of ordered pairs B = (B1 , B2 ) of sets in a

space X . B is said to be a pair-base for X if, for each B ∈ B, B1 ∈ τ , and for every
x ∈ U ∈ τ , there is B ∈ B such that x ∈ B1 ⊂ B2 ⊂ U . B is called a cushioned
family in X if for any B  ⊂ B, ∪{B1 : B ∈ B  } ⊂ ∪{B2 : B ∈ B  }. A family P
of sets in X is called a quasi-base for X if {(P ◦ , P) : P ∈ P} is a pair-base for X .

Definition B.2.11 ([92]) Let X be a regular space. X is called an M1 -space if X has a

σ -closure-preserving base; X is called an M2 -space if X has a σ -closure-preserving
quasi-base; X is called an M3 -space if X has a σ -cushioned base.

Every M1 -space is an M2 -space, and every M2 -space is an M3 -space. However,

is every M3 -space an M2 -space? Is every M2 -space an M1 -space? These famous
questions were raised by Ceder [92]. Ceder’s paper has opened the prelude of the
study on generalized metric spaces. The study of Mi -spaces is one of the main clues
for the development of the theory of generalized metric spaces.

B.2.3 Paracompactness of Product Spaces

After proving that every product of a compact space and a paracompact space is a
paracompact space, Dieudonné [112] asked whether every product of two paracom-
pact spaces is a paracompact space. Sorgenfrey [433] constructed a paracompact
space (which was called the Sorgenfrey line later) such that the product of the space
Appendix B: The Formation of the Theory of Generalized Metric Spaces 283

multiplied by itself is not even a normal space. Compared with metric spaces or
compact spaces, the fatal defect of paracompact spaces is that not every product
of two paracompact spaces is a paracompact space. To investigate the normality of
product spaces, Dowker [116] and Katětov [224] introduced the class of countable
paracompact spaces independently.

Definition B.2.12 A space X is called a countably paracompact space [116, 224]

if every countable open cover of X has a locally finite open refinement.

Dowker proved that a normal space X is a countably paracompact space if and

only if the product of the space X and the unit closed interval I is a normal space.
After a discussion on the precise relationship between normality and countably para-
compactness, Dowker raised the question whether there is a normal space which is
not countably paracompact. Such spaces are called Dowker spaces and a Dowker
space was constructed by Rudin [409] until 1971.
Michael [325] asked whether every product of a paracompact space and a metriz-
able space is a paracompact space. Ten years later, Michael [329] constructed a
paracompact space (it was called the Michael line later), such that, the product of
this space and a separable metrizable space is not a normal space. Tamano [446]
raised the question of seeking a necessary and sufficient condition for a space X ,
such that, X × Y is a normal space for any metrizable space Y . A paper of Morita
[360] gave an appropriate solution of the Tamano problem.

Definition B.2.13 A space X is called a Morita space (or Morita’s P-space2 [360])
if, for every family of open sets {G(α1 , . . . , αi ) : α1 , . . . , αi ∈ Ω, i ∈ N} such that
G(α1 , . . . , αi ) ⊂ G(α1 , . . . , αi+1 ), there is a family {F(α1 , . . . , αi ) : α1 , . . . , αi ∈
Ω, i ∈ N} of closed sets in X with the following properties:
(1) F(α1 , . . . , αi ) ⊂ G(α1 , . . . , αi ); 
 {αi } is a sequence in Ω such that X =
(2) if i∈N G(α1 , . . . , αi ), then X =
i∈N F(α 1 , . . . , αi ).

Morita proved that a space X is a normal Morita space if and only if any product
space of X and a metrizable space is a normal space. Frolík [135] proved that any
product space of countably many Čech-complete paracompact spaces is a paracom-
pact space. Frolík’s theorem is undoubtedly the first satisfactory result about the
paracompactness of product spaces. Its shortcoming is that a metrizable space may
not be a Čech-complete space. For this reason, Morita [360] defined the following
M-spaces.

Definition B.2.14 A space X is called an M-space [360] if, there is a normal

sequence {Ui } of open covers of X , such that, for each x ∈ X and any sequence {xi }
in X , {xi } has an accumulation point in X whenever xi ∈ st(x, Ui ).

2 A space X is called a P-space if every G δ -set of X is open in X , i.e., every point of X is a P-point.
284 Appendix B: The Formation of the Theory of Generalized Metric Spaces

The concepts of M-spaces and Čech-complete spaces are independent. However,

they are coincide in the class of paracompact spaces. Countably compact spaces,
metrizable spaces and Čech-complete paracompact spaces are all M-spaces, and
every M-space is a Morita space. Any product space of countably many paracom-
pact M-spaces is a paracompact M-space.
The above discussion is based on the fact that metrizability is countably produc-
tive. For the normality of product spaces of the uncountably many metrizable spaces,
Stone [440] obtained the following equivalent conditions.
Theorem B.2.15 For every family {X α }α∈Λ of nonempty metrizable spaces, the fol-
lowing are equivalent:

(1) α∈Λ X α is a normal space.
(2) α∈Λ X α is a paracompact space.
(3) There are at most countably many α in Λ such that the X α is not a compact
space.
Thus, Nℵ1 is not a normal space. The researchers transferred interest of normality
or paracompactness of product spaces of uncountably many spaces into normality
or paracompactness of suitable subspaces of product spaces. By this idea, Corson
[108] introduced the concepts of σ -products and Σ-products, as generalizations of
finite products and countable products respectively.

Definition B.2.16 Suppose X = α∈Λ X α is a product of infinitely many spaces.
Take b = (bα ) ∈ X . For each x ∈ X , denote supp(x) = {α ∈ Λ : xα
= bα }. Let

σ (b) = {x ∈ X : |supp(x)| < ℵ0 },

Σ(b) = {x ∈ X : |supp(x)|  ℵ0 }.

Then the subspaces σ (b) and Σ(b) of X are called a σ -product and a Σ-product of
spaces {X α : α ∈ Λ} with the based point b respectively. Generally, they are denoted
as σ {X α : α ∈ Λ} and Σ{X α : α ∈ Λ} respectively.
Corson proved that every σ -product of separable metrizable spaces is a Lindelöf
space, and every Σ-product of complete metrizable spaces is a normal space. The
problem if every Σ-product of metrizable spaces is a normal space was a most
attractive famous open problem. In 1977, Gul’ko [168] solved this problem positively.
In 1983, Rudin [413] proved that every Σ-product of metrizable spaces is a shrinking
space.

B.2.4 Spaces and Classifications

In fact, early work on continuous functions, quotient mappings, open mappings and
closed mappings done by researchers in general topology has become a driving force
for the development of this field.
Appendix B: The Formation of the Theory of Generalized Metric Spaces 285

Since 1944, some of more powerful mappings are constantly emerging, such as
compact mappings (Vaı̌nšteı̌n 1947), perfect mappings (Vaı̌nšteı̌n 1947), π -mappings
(Ponomarev 1960), pseudo-open mappings (Arhangel’skiı̌ 1963), compact-covering
mappings (Michael 1966) an so on. Mappings and spaces are mutually dependent,
and hence in order to promote the vigorous development of mappings, it is necessary
to discuss the bridge bond between mappings and spaces. In 1961, one important
event which had great influence on the development of general topology was that
the international topological symposium named “General Topology and its Relations
to Modern Analysis and Algebra” held in Prague for the first time (the conference
holds every 5 years once since then) [111]. In this conference, Alexandroff [3] put
forward the idea of investigating spaces by mappings, namely, to connect various
classes of spaces by using mappings as a link, then according to the differences
between classes of spaces and mappings to arranged the study. The Alexandroff idea
was based on several known results. For example, Gale [136] characterized k-spaces
as quotient images of locally compact spaces; Ponomarev [401] characterized first
countable spaces as open images of metrizable spaces; Frolík [135] characterized
Čech-complete paracompact spaces as perfect preimages of complete metrizable
spaces. Alexandroff made a conjecture that paracompact spaces can be characterized
as perfect preimages of metrizable spaces. Although this conjecture is incorrect, the
conjecture led Arhangel’skiı̌ [25] to introduce the concept of p-spaces.

Definition B.2.17 A completely regular space X is called a p-space [25] if there

is a sequence

{Un } of families of open sets in β X covering X , such that, for each
x ∈ X , n∈N st(x, Un ) ⊂ X . The sequence {Un } is called a pluming of X .

Both Čech-complete spaces and completely regular Moore spaces are p-spaces.
Arhangel’skiı̌ characterized perfect preimages of metrizable spaces as paracompact
p-spaces. Another important role played by p-spaces is that each product space of
countably many (paracompact) p-spaces is a (paracompact) p-space. Although the
class of p-spaces and the class of M-spaces are independent, a theorem was proved
that these two classes of spaces coincide in the class of paracompact spaces.
In a word, Alexandroff’s mutual classification idea of mappings and spaces has
become an important source for further study of general topology, and has been
reached in many concrete spaces. For example,
(1) Ponomarev [401] characterized spaces with a point-countable base as open s-
images of metrizable spaces;
(2) Arhangel’skiı̌ [23] characterized spaces with a uniform base as open compact
images of metrizable spaces;
(3) Arhangel’skiı̌ [26] characterized Fréchet–Urysohn spaces as pseudo-open images
of metrizable spaces;
(4) Morita [360] characterized M-spaces as quasi-perfect preimages of metrizable
spaces;
(5) Franklin [133] characterized sequential spaces as quotient images of metrizable
spaces;
286 Appendix B: The Formation of the Theory of Generalized Metric Spaces

(6) Heath [179] characterized developable spaces as open π -images of metrizable

spaces.
As mentioned above, the development of general topology in the 20 years from
1944 to 1964 laid a solid foundation for the investigation of generalized metric prop-
erties and established a research framework of generalized metric properties. The
applications of discrete families, locally finite families and closure-preserving fami-
lies etc. in treating uncountable cases of families were the main features of this period
of research. The introduction of paracompactness generalized a lot of important theo-
rems on compact spaces or metrizable spaces to paracompact spaces (These general-
izations are often essential), numbers of nice properties of paracompact spaces have
been applied in many areas outside general topology, which accelerated the infiltra-
tion of general topology, as a basis branch, to other areas. The successful derivation
of metrization theorems for general spaces made people understand the essence of
metric spaces more profoundly, and brought a bright prospect on a more in-depth
and meticulous study of metric properties. More importantly, it allows researchers to
avoid the difficulties and complex structures of distance functions in the process of
seeking metrization theorems for specific topological spaces, and at the same time
is helpful for all kinds of generalizations and applications of metric properties. The
central topic in this period was mining properties of paracompact spaces and seek-
ing more metrization theorems. Topologists gradually shifted their early interest and
attention for compact spaces and metric spaces to paracompact spaces and general-
ized metric spaces. Specifically, the scholars focused on the following four problems
simultaneously:
(1) Among which class of spaces with weaker topological properties, countable
compactness and compactness are equivalent to each other?
(2) Among which class of more general spaces, paracompactness is closed under
countable products?
(3) Seeking more metrization theorems for spaces with some specific topology prop-
erties.
(4) Classifying spaces by mappings.

B.3 The Formation Period

In 1965 and 1966, Arhangel’skiı̌, Borges, Michael, Wicke, Worrell and others pub-
lished a number of important papers, which steered the study of general topology into
a stage of rapid development and formed the theory of generalized metric spaces.
Appendix B: The Formation of the Theory of Generalized Metric Spaces 287

B.3.1 Isocompactness and Normality

Bacon [49] called a space X isocompact if every countably compact closed subspace
of X is compact.
The role in metrization problems played by developable spaces and BCO spaces
were indicated by the work in the foundation laying period. What kind of cover-
ing properties and base properties do developable spaces have? The article titled
“Characterizations of developable topological spaces” written by Worrell and Wicke
[481] is one of the most outstanding papers on developable spaces. The concepts of
θ -refinable property and θ -base introduced by them gave a satisfactory answer to the
above question.
Definition B.3.1 A space X is said to be a θ -refinable space [481] if every open
cover of X has a sequence {Un } of open refinements, such that, for each x ∈ X , there
is i ∈ N with ord(x, Ui ) < ℵ0 , where ord(x, Ui ) = |{U ∈ Ui : x ∈ U }|.

Definition B.3.2 A family i∈N Bi of open sets in a space X is said to be a θ -base
[481] if for every x ∈ U ∈ τ , there exist i ∈ N and B ∈ Bi such that x ∈ B ⊂ U
and ord(x, Bi ) < ℵ0 .

Every developable space is a θ -refinable space with a θ -base, and the class of
developable spaces is equivalent to the class of θ -refinable BCO spaces. In 1977,
Y. Liu [314] introduced the concept of strictly quasi-paracompact spaces. Every
θ -refinable space is a strictly quasi-paracompact space and every strictly quasi-
paracompact space is an isocompact space. In 1976, Reed and Zenor [407] proved
that every normal developable manifold is metrizable, which substantively opened
a path on metrization of topological manifolds. With a slight modification of the
definition of developable spaces, Bennett [55] defined quasi-developable spaces in
1968.
Definition B.3.3 A space X is called a quasi-developable space [55] if, there is a
sequence {Bi } of families of open sets in X , such that, for each x ∈ X , {st(x, Bi ) :
i ∈ N, st(x, Bi )
= ∅} is a neighborhood base of x in X .
Different from developable spaces, a quasi-developable paracompact space may
not be a metrizable space. However, a space is a developable space if and only if it
is a perfect quasi-developable space. Bennett and Lutzer [59] proved that the quasi-
developable property is equivalent to the θ -base property. Smith [430] positively
answered the following question raised by Bennett [57]: Is every quasi-developable
space an isocompact space?
The concepts of point-countable bases and θ -bases are independent. If the assump-
tion ord(x, Ui ) < ℵ0 in the definition of θ -bases is replaced with ord(x, Ui )  ℵ0 ,
then the topological property produced is that of δθ -bases defined by Aull [44].
Obviously, every space with a point-countable base or a θ -base has a δθ -base. Many
conclusions on spaces with a point-countable base or a θ -base are also true for spaces
with a δθ -base. For example, every space with a δθ -base is an isocompact space.
288 Appendix B: The Formation of the Theory of Generalized Metric Spaces

In the process of discussion on metrization of compact spaces, Arhangel’skiı̌ and

Projzvolov [314] introduced the concept of pseudo-bases, and proved that every
compact space with a point-countable pseudo-base is metrizable.

Definition B.3.4 A family B of open sets in a space X is said to be a pseudo-

base [41] for X if, for every x, y ∈ X with x
= y, there is B ∈ B such that
x ∈ B ⊂ X − {y}.

Shiraki [422] pointed out that every space with a point-countable pseudo-base is
an isocompact space.

Definition B.3.5 A space X is said to have a G δ -diagonal [92] (resp. quasi-G δ -

is a sequence {Un } of open
diagonal [192]) if, there
covers of X (resp. families of open
sets in X ) such that n∈N st(x, Un ) = {x} (resp. n∈N {st(x, Un ) : x ∈ ∪Un } = {x})
for each x ∈ X .

Šneı̌der [432] proved every compact space with a G δ -diagonal is a metrizable

space. Anderson [16], Bennett [57] and Heath [183] asked whether every countably
compact space with a G δ -diagonal or a quasi-G δ -diagonal is a compact space. Chaber
[93] answered this question positively. Bennett, Byerly and Lutzer [58] proved the
following result.

Theorem B.3.6 Every countably compact subset of a space X with a quasi-G δ -

diagonal is a compact metrizable G δ -subset of X .

So far, the research on the isocompactness of generalized metric spaces can be

temporarily come to an end. One good result on covering properties is that Uspen-
skii [470] proved every σ -metacompact pseudo-compact space is a compact space.
Seeking characterizations of isocompact spaces is still an open problem.
For product spaces, normality and paracompactness is a pair of twin brothers, but
the former one seems more essential and difficult. In the 20th century, the excellent
work done by Dowker in the 50s and by Morita in the 60s played a valuable role for
further research. In 1969, Nagata [371] proved that a space X is a paracompact M-
space if and only if X is homeomorphic to a closed subspace of a product of a
metrizable space and a compact space. In the 1970s, a lot of excellent results emerged.
Based on the work of Rudin and Starbird [415], Przymusińsky [404] made a careful
study on problems of the normality of product spaces with a compact or metric factor,
and gave a necessary and sufficient condition for the product spaces to be normal
spaces. A series of results could be obtained as corollaries of his result. In 1977,
Burke and van Douwen [86], Kato [226] constructed an M-space X such that X is
not homeomorphic to any closed subspace of the product of a metrizable space and
a countably compact space. The another clue on normality of product spaces is the
following Morita’s conjecture [362] .

Conjecture B.3.7 The following conclusions are true:

(1) A space X is a discrete space if and only if for every normal space Y , X × Y is
a normal space.
Appendix B: The Formation of the Theory of Generalized Metric Spaces 289

(2) A space X is a metrizable space if and only if for every normal Morita space Y ,
X × Y is a normal space.
(3) A space X is a σ -locally compact metrizable space if and only if for every
normal, countably paracompact space Y , X × Y is a normal space.
Rudin [412] confirmed Conjecture B.3.7(1). Morita [362] proved that if for every
non-discrete space Y , there is a normal space X such that X ×Y is not a normal space,
then Conjecture B.3.7(3) is true (and hence by Rudin [412], Conjecture B.3.7(3) is
actually true). In 2001, Balogh [52] proved Conjecture B.3.7(2) holds.
In the following, we introduce the normality of Σ-products. A Σ-product of
compact spaces may not be a normal space (see Theorem B.3.8). So, while discussing
the normality of Σ-products, we generally assume that factor spaces are some class
of generalized metric spaces. Kombarov [234] obtained a necessary and sufficient
condition for Σ-product spaces of paracompact p-spaces to be normal spaces.

Theorem B.3.8 Suppose {X α }α∈Λ is a family of paracompact p-spaces and X =

Σ{X α : α ∈ Λ}. Then the following are equivalent:
(1) X is a normal space.
(2) X is a collectionwise normal space.
(3) X is of countable tightness.
(4) Every X α is of countable tightness.

Comparing to paracompactness of product spaces of countably many spaces,

Theorem B.3.8 enlighten people to discuss normality of Σ-product spaces of para-
compact Σ-spaces. Yajima [483] constructed an example to show that a Σ-product
of M1 -spaces may not be a normal space. In 1995, Eda, Gruenhage, Koszmider,
Tamano and Todorčevic̀ [118] proved that under the assumption CH, the Σ-product
of Lašnev spaces Sω2 is not a normal space. H. Teng [463] proved that every Σ-
product of semi-stratifiable spaces is a collectionwise subnormal space.

B.3.2 Generalizations of Bases—Networks

The Bing-Nagata-Smirnov metrization theorem evoked enthusiasm on spaces with

various base properties. However, one of the shortcomings of spaces defined by bases
is that when considering operations relevant to the “closeness”, it is inconvenient,
sometimes even very difficult. Therefore, it is necessary to make some appropriate
generalizations for the concept of bases. In order to prove the Alexandroff–Urysohn
addition theorem of arbitrary cardinal numbers, Arhangel’skiı̌ [21] made a successful
attempt.

Definition B.3.9 A family P of sets in a space X is called a network [21] for X if

every open subset of X is the union of some elements of P.

Arhangel’skiı̌ [21, 22] studied spaces with a countable network, proved every
space with a countable network is preserved by continuous mappings, and applied
290 Appendix B: The Formation of the Theory of Generalized Metric Spaces

the notion of networks to give a short and natural proof of the following theorem:
if a compact space X is the union of two subspaces of weight not greater than an
infinite cardinal number κ, then the weight of X also does not exceed κ. Every
p-space with a countable network has a countable base [25]. It is well-known that
the covering dimension dimX , the small inductive dimension indX , and the large
inductive dimension IndX of a separable metric space X coincide. Arhangel’skiı̌
[31] asked whether they agree in the class of regular continuous images of separable
metric spaces, or cosmic spaces as they are often called [331]. Cosmic spaces are also
characterized as regular spaces which have a countable network [331]. For a cosmic
space X it is known that indX = IndX , so the question is whether dimX = indX . In
2006, Charalambous [100] gave a cosmic space X such that dimX = 1 and indX = 2.
In order to investigate quotient images of separable metrizable spaces, Michael
[331] introduced the concept of strict k-networks.
Definition B.3.10 A family P of sets in a space X is said to be a strict k-network
[331] for X if, for every compact set K with K ⊂ U ∈ τ in X , there is P ∈ P
such that K ⊂ P ⊂ U . A regular space with a countable strict k-network is called
an ℵ0 -space.
Every ℵ0 -space can be characterized as a regular compact-covering image of a
separable metrizable space. However, a metrizable space may not be an ℵ0 -space.
O’Meara [387] introduced the concept of k-networks as a common generalization of
bases and strict k-networks.
Definition B.3.11 A family P of sets in a space X is said to be a k-network [387]
for X if, for every compact set K with K ⊂ U ∈ τ in X , there is a finite subfamily
P  of P such that K ⊂ ∪P  ⊂ U . A regular space with a σ -locally finite k-network
is called an ℵ-space.
Strict k-networks were called pseudo-bases in [331]. Because the term “pseudo-
base” already has been used with a different meaning by Arhangel’skiı̌ and Projzvolov
[41] (see Definition B.3.4), and the concept of pseudo-bases is stronger than that of
k-networks, we use the term “strict k-network” in this book.
The concept of ℵ-spaces generalizes that of metrizable spaces and ℵ0 -spaces.
With the help of the concept of networks, Okuyama [385] defined σ -spaces, which
generalizes ℵ-spaces and cosmic spaces.
Definition B.3.12 A regular space with a σ -locally finite network is called a σ -
space.
After investigating several basic properties of σ -spaces, Okuyama raised the fol-
lowing two questions:
(1) Is every regular space with a σ -closure-preserving network a σ -space?
(2) Is every regular closed image of a σ -space a σ -space?
Arhangel’skiı̌ [31] discussed spaces with a σ -discrete network. Siwiec and Nagata
[426] proved a beautiful characterization theorem of σ -spaces, and solved the above
two questions.
Appendix B: The Formation of the Theory of Generalized Metric Spaces 291

Theorem B.3.13 (The Nagata-Siwiec theorem) For every regular space X , the fol-
lowing are equivalent:
(1) X is a σ -space.
(2) X has a σ -discrete network.
(3) X has a σ -closure-preserving network.

Because of the characterizations of paracompact spaces given by Michael, the

questions on Mi -spaces raised by Ceder and the Nagata-Siwiec theorem, it is natural
for researchers to concern about whether the class of σ -space is equivalent to the class
of the regular spaces with a σ -cushioned network. In order to describe the problem
more clearly, we introduce the innovation work on M3 -spaces done by Borges [65].

Definition B.3.14 A space X is called a stratifiable space [65] if, there is a function
G : N × τ c → τ on X such that

(1) H ⊂ G(n, H ) and H = n∈N G(n, H );

(2) H ⊂ F ⇒ G(n, H ) ⊂ G(n, F).

The class of stratifiable spaces is equivalent to the class of M3 -spaces. Thus,

Borges solved some problems on Mi -spaces raised by Ceder [92], for example
whether M3 -spaces are preserved by closed mappings. In 1967, Michael defined
the concept of semi-stratifiable spaces [110].

Definition B.3.15 A space X is said to be a semi-stratifiable space [110] if, there is

a function G : N × τ c → τ on X such that

(1) H = n∈N G(n, H );

(2) H ⊂ F ⇒ G(n, H ) ⊂ G(n, F).

Creede [109, 110] studied semi-stratifiable spaces systematically. Stratifiable

spaces, σ -spaces and semi-metrizable spaces are all semi-stratifiable spaces. Spaces
having a σ -cushioned network were named as pseudo-stratifiable spaces by Kofner
[231]. He proved that the class of pseudo-stratifiable spaces is equivalent to the
class of semi-stratifiable spaces and at the same time constructed a regular Lindelöf
pseudo-stratifiable space which is not a σ -space.
Okuyama [385] proved that paracompact σ -spaces are countably productive in
his first paper about σ -spaces. Since the class of (paracompact) M-spaces and the
class of (paracompact) σ -spaces are independent, another question one met after the
appearing of σ -spaces is to look for the class of spaces containing both the class of
M-spaces and the class of σ -spaces, and paracompactness is countably productive
in this class. Nagami [368] successfully introduced the class of Σ-spaces satisfying
the requirements.

Definition B.3.16 Suppose P is a cover of a space X . For each x ∈ X , let

C(P, x) = ∩(P)x . A Σ-network for X is a sequence {Pi } of locally finite closed
covers of X , such that, if x is a point and {xi } is a sequence in X with xi ∈ C(Pi , x),
292 Appendix B: The Formation of the Theory of Generalized Metric Spaces

then {xi } has an accumulation point in X . A space with a Σ-network is called a

Σ-space [368]. For a Σ-network for X , let

C(x) = ∩{C(Pi , x) : i ∈ N}, x ∈ X.

Then C(x) is a countably compact closed set in X . When every C(x) is a compact set
in X , {Pi } is said to be a strong Σ-network in X and X is called a strong Σ-space
[368].

Every M-space or every σ -space is a Σ-space, and every Σ-space is a Morita

space. Paracompact Σ-spaces are countably productive. Nagami proved Σ-spaces
are preserved by perfect mappings. However, Michael [333] constructed a locally
compact paracompact space (hence a strong Σ-space) such that one closed image of
it is not a Σ-space. In view of this, Michael [333] introduced the concept of strong
Σ  -spaces.

Definition B.3.17 A family P of sets in a space X is called a (mod k)-network

[333] if, there is a cover K consisting of compact closed subsets of X , such that, for
each K ∈ K , {P ∈ P : K ⊂ P} is a network of K in X .

The class of strong Σ-spaces is equivalent to the class of spaces with a

σ -locally finite closed (mod k)-network. Michael [333] called a space with a σ -
closure-preserving closed (mod k)-network a strong Σ  -space. Obviously, strong Σ  -
spaces are preserved by closed mappings. Michael further pointed out that whether
or not the class of strong Σ  -spaces is valuable to study deeply, depends on whether
the paracompact strong Σ  -spaces are countably productive. In 1984, Patsei [390]
proved that paracompact strong Σ  -spaces are countably productive. Okuyama [386]
investigated spaces relevant to strong Σ-spaces systematically. In order to unify our
description, if we replace K in the definition of (mod k)-networks with a cover of
X consisting of some countably compact closed subsets of X , then P is said to be
a quasi-(mod k)-network for X . So every Σ-network is just a σ -locally finite closed
quasi-(mod k)-network.

Definition B.3.18 [240] A family P of sets in a space X is said to be a hereditarily

closure-preserving family in X if {H P : P ∈ P} is a closure-preserving family
whenever H P ⊂ P ∈ P.

Definition B.3.19 [386] A space with a σ -hereditarily closure-preserving closed

quasi-(mod k)-network is called a Σ ∗ -space, and a space with a σ -closure-preserving
closed quasi-(mod k)-network is called a Σ  -space.

Every Σ-space is a Σ ∗ -space and every Σ ∗ -space is a Σ  -space. But neither of

the opposite implications is true. Both Σ ∗ -spaces and Σ  -spaces are preserved by
closed mappings. Okuyama proved that a paracompact space X is a Σ-space if and
only if X × I is a Σ ∗ -space. It is worth mentioning here that paracompactness is not
countably productive in the class of semi-stratifiable spaces. Under the assumption
Appendix B: The Formation of the Theory of Generalized Metric Spaces 293

CH, Michael [334] constructed a hereditarily Lindelöf semi-metrizable space such

that the product multiplied by itself is not a normal space.
Open (mod k)-networks are called (mod k)-bases [333]. In 1970, Lutzer and
Michael [333] proved independently that a space X is a paracompact M-space if and
only if X is a regular space with a σ -locally finite (mod k)-base.
By means of various generalizations of bases, such as networks, k-networks and
(mod k)-networks, a series of generalized metric spaces were obtained, such as
cosmic spaces, σ -spaces, semi-stratifiable spaces, ℵ-spaces, Σ-spaces and so on.
Naturally, this causes the following two problems:
(1) What are the metrization theorems for these classes of spaces?
(2) What are the “factor decomposition” theorems for these classes of spaces?
Look at the metrization problems first. It is reflected in the metrization problems
of semi-stratifiable spaces and Σ-spaces. Every semi-stratifiable space has a G δ -
diagonal and Chaber [93] proved that every M-space with a G δ -diagonal is a metriz-
able space. For Σ-spaces, main discussion is the problem which class of Σ-spaces is
that of σ -spaces. This also can be regarded as a form of factor decomposition theo-
rems of σ -spaces. Siwiec and Nagata [426] defined the concepts of pseudo-networks
(it was called ct-networks at that time) and σ  -spaces.

Definition B.3.20 A family P of sets in a space X is called a pseudo-network for

X if for every x, y ∈ X with x
= y, there is P ∈ P such that x ∈ P ⊂ X − {y}. A
space with a σ -closure-preserving closed pseudo-network is called a σ  -space.

Shiraki [422] proved that every σ -space can be decomposed as a space which is
both a regular Σ  -space and a σ  -space, and every regular Σ-space with a point-
countable pseudo-base is a σ -space. Burke and Lutzer [88] constructed a Moore
space which has no point-countable pseudo-base, and answered an open question of
Reed [406] negatively.
The research on seeking this kind of theorems of generalized metric spaces is very
fruitful, such as
(1) a space is a σ -space if and only if it is a regular c-semi-stratifiable Σ  -space
[319];
(2) a space is a semi-metrizable space if and only if it is a first countable semi-
stratifiable space [110];
(3) a space is a Nagata space if and only if it is a first countable stratifiable space
[65].
Ceder [92] asked that whether every Nagata space can be decomposed as a para-
compact semi-metrizable space. After constructing a counterexample for Ceder’s
question, Heath [180] asked: What is the necessary and sufficient condition for a
paracompact semi-metrizable space being a Nagata space? Lutzer [316] gave the
definition of a k-semi-stratifiable space.
294 Appendix B: The Formation of the Theory of Generalized Metric Spaces

Definition B.3.21 A regular space X is said to be a k-semi-stratifiable space if there

is a function G : N × τ c → τ on X satisfying the requirement for being a semi-
stratifiable space, and if K ∩ H = ∅ for a compact set K and a closed set H in X ,
then there is n ∈ N such that K ∩ G(n, H ) = ∅.
Lutzer proved that a space is a Nagata space if and only if it is a semi-metrizable
k-semi-stratifiable space. Its essence is that a space is a Nagata space if and only
if it is a first countable k-semi-stratifiable space. None of the above results gives
a decomposition of stratifiable spaces. Arhangel’skiı̌ [31] asked: Which class of
cosmic spaces is the class of stratifiable spaces? Heath, Lutzer and Zenor [186]
gave the definition of monotonically normal spaces. They proved that a space is a
stratifiable space if and only if it is a monotonically normal semi-stratifiable space,
and a space is metrizable if and only if it is a monotonically normal p-space with a
G δ -diagonal.
Definition B.3.22 ([186]) A space X is called a monotonically normal space if for
each pair H, F of disjoint closed sets in X , there is an open set D(H, F) in X such
that
(1) H ⊂ D(H, F) ⊂ D(H, F) ⊂ X − F;
(2) H ⊂ H  , F ⊃ F  ⇒ D(H, F) ⊂ D(H  , F  ).
For Mi -space problems, Gruenhage [160] and Junnila [216] proved that every
M3 -space is an M2 -space independently. Since then, a lot of progresses have been
made on the problem whether every M3 -space is an M1 -space. In 2008, Tamano
[451] proved that if X is a Polish space (i.e. separable complete metric space), then
Ck (X ) is an M1 -space, and hence Ck (P) is an M1 -space. Further more, Reznichenko
[408] proved that for every separable metric space X , Ck (X ) is a stratifiable space if
and only X is a Polish space. It remains unsolved whether Ck (X ) is hereditarily M1
for any Polish space X [451].

B.3.3 The Alexandroff Idea

After Alexandroff [3] proposed his idea, results on classifying many specific spaces
by mappings emerged. However, only after Arhangel’skiı̌ [31] was published, clas-
sifying spaces by mappings formed a research direction in general topology.
In 1966, Arhangel’skiı̌ [31] published a historic literature entitled “Mappings and
spaces”, which created a new era of classifying spaces by means of mappings. This
article systematically summarized the important results achieved in the mapping the-
ory in half a century since the birth of general topology. More importantly, it gave
the concrete problems on investigating various spaces by mappings, which formed
the famous Alexandroff-Arhangel’skiı̌’s questions. Its core content is to establish
the relationships between the class of metric spaces and classes of spaces with
certain topological properties by mappings. These problems are outstanding con-
tributions to general topology, and make the idea of classifying spaces by means
Appendix B: The Formation of the Theory of Generalized Metric Spaces 295

of mapping become an important integral part of the theory of generalized metric

space [4].
In the view of describing metric spaces by distance functions, symmetrizability,
Δ-metrizability [382] and semi-metrizability [477] are the direct generalizations
of metrizability, and they are the earliest generalized metric properties. Niemytzki
[382] did pioneering work on generalized metric spaces, for example, he defined Δ-
metrics as the “metrics” that do not have to satisfy the symmetric condition and also
introduced symmetrics as the “metrics” that need not satisfy the triangle inequality.

Definition B.3.23 ([31]) A family P of sets in a space X is called a weak base for
X if, there is Px ⊂ P for each x ∈ X such that

(1) x ∈ ∩Px and P = x∈X Px ;
(2) U, V ∈ Px ⇒ there is W ∈ Px such that W ⊂ U ∩ V ;
(3) F is a closed subset of X if and only if for each x ∈ X − F, there is P ∈ Px
such that F ∩ P = ∅.
If X has a weak base P such that every Px is countable, then X is called a g-first
countable space (or weakly first countable space, g f -countable space).
Every symmetrizable space is a g-first countable space, and every first countable
space is equivalent to a g-first countable Fréchet–Urysohn space. By means of sym-
metrizability, one can obtain more general metrization theorems than by means of
semi-metrizability and also can effectively characterize certain quotient images of
metrizable spaces. For example,
(1) a space X is a quotient π -image of a metrizable space if and only if X is a
symmetrizable space satisfying the weak Cauchy condition [232];
(2) a space X is a pseudo-open π -image of a metrizable space if and only if X is a
semi-metrizable space [1, 78].

As a generalization of bases, weak bases provide a way for people to investi-

gate more general generalized metric spaces. For example, Siwiec [425] defined
g-metrizable spaces as regular spaces with a σ -locally finite weak base; in 1982,
Foged [127] proved that every g-metrizable space has a σ -discrete weak base, and
hence answered a question of Siwiec; in 2005, C. Liu [305] proved that the class of
g-metrizable spaces is equivalent to the class of regular spaces with a σ -hereditarily
closure-preserving weak base, and hence answered a question of Tanaka. It is still an
open problem that whether every regular space with a σ -compact-finite weak base
is g-metrizable [299].
One important result obtained in exploring metrizable spaces is that metrizabil-
ity implies paracompactness. Developable spaces, as a generalization of metrizable
spaces, may not be metacompact spaces. Except the θ -refinable property, does every
developable space have some covering properties parallel to those possessed by every
metrizable space?
Arhangel’skiı̌ [31] gave the definition of σ -paracompact spaces.
296 Appendix B: The Formation of the Theory of Generalized Metric Spaces

Definition B.3.24 ([31]) A space X is called a σ -paracompact space if, every open
cover U of X has a sequence {Un } of open refinements, such that, for each x ∈ X ,
there exist n ∈ N and U ∈ U with st(x, Un ) ⊂ U .

When investigating semi-metrizable spaces, McAuley [234] found that semi-

metrizability implies so called Fσ -screenable property of a space X , i.e., every open
cover of X has a σ -discrete closed refinement.
Burke [75] established the precise relationships among these properties.

Theorem B.3.25 For every space X , the following are equivalent:

(1) X is a σ -paracompact space.
(2) X is an Fσ -screenable space.
(3) Every open cover of X has a σ -locally finite closed refinement.
(4) Every open cover of X has a σ -closure-preserving closed refinement.

Burke called a space with one of the above properties a subparacompact space.
Metacompactness and subparacompactness are independent and every subparacom-
pact space is a θ -refinable space. Burke [76] and Katuta [227] raised the following
question: if every open cover of a space has a σ -cushioned refinement, then is it a
subparacompact space? In 1978, Junnila [217] gave this question a positive answer.
Although every paracompact p-space can be characterized as a perfect preimage
of a metrizable space, Arhangel’skiı̌ [31] pointed out that a p-space may not be any
perfect preimage of a developable space. Associated with this, Arhangel’skiı̌ [31]
introduced the concept of strict p-spaces.

Definition B.3.26 ([31]) A completely regular space X is called a strict p-space

if,

there is a sequence
{Un } of families of open sets in β X covering X satisfying
n∈N st(x, Un ) = n∈N st(x, Un ) ⊂ X for every x ∈ X .

Strict p-spaces are stronger than p-spaces strictly. Burke [77] proved every θ -
refinable p-space is a strict p-space. Burke [80] raised the question whether every
strict p-space is a θ -refinable space. In 1986, S. Jiang [213] answered this ques-
tion positively. On the other hand, every completely regular perfect preimage of a
developable space is a subparacompact p-space. Isiwata [206] gave an example to
illustrate the converse is not true and also gave a characterization of perfect preim-
ages of developable spaces. A characterization of perfect preimages of σ -spaces was
given by Suzuki [445].
Every closed image of a paracompact space is also a paracompact space, but a
closed image of a metrizable space may not be a metrizable space. Arhangel’skiı̌
asked in [31]: What is the intrinsic characterization of closed images of metriz-
able spaces? Lašnev [240] studied this question first, and hence closed images of
metrizable spaces are called Lašnev spaces by the researchers later. Slaughter [427]
proved that every Lašnev space is an M1 -space. The concept of hereditarily closure-
preserving families of sets (see Definition B.3.18) was introduced by Lašnev when he
investigated closed images of metrizable spaces. He characterized every closed image
of a metrizable space as a Fréchet–Urysohn space with an almost refining sequence
Appendix B: The Formation of the Theory of Generalized Metric Spaces 297

of hereditarily closure-preserving coverings comprising a network for the space.

The importance of this characterization is the introduction of concept of hereditarily
closure-preserving families. For example, Morita and Rishel [365] gave a charac-
terization of closed images of (paracompact) M-spaces by using Lašnev’s method;
the concept of Σ ∗ -spaces introduced by Okuyama [386] is based on hereditarily
closure-preserving families of sets. More interestingly, Burke, Engelking and Lutzer
[87] established a new metrization theorem.

Theorem B.3.27 (The Burke-Engelking-Lutzer metrization theorem) A space X is

a metrizable space if and only if X is a regular space with a σ -hereditarily closure-
preserving base.
Metrizable spaces and paracompact M-spaces are preserved by perfect mappings,
but an M-space may not be preserved by a perfect mapping [361].

Definition B.3.28 [201] A space X is called an M ∗ -space if, there is a sequence {Fi }
of locally finite closed covers of X , such that, for every x ∈ X and any sequence
{xi }, {xi } has an accumulation point in X if xi ∈ st(x, Fi ).

Morita and Rishel [365] and Nagata [372] proved that the class of perfect images
of M-spaces is just the class of M ∗ -spaces. If “locally finite” in the definition of M ∗ -
spaces is replaced with “closure-preserving”, then we obtain the class of M  -spaces
defined by Siwiec and Nagata [426]. The question whether every M  -space is an
M ∗ -space raised by Morita and Rishel [365] is still open.
In the above, we mainly discuss closed images of spaces. Investigation on open
images of spaces also has triggered a lot of profound work on generalized metric
spaces.
The best known for this is the class MOBI of Arhangel’skiı̌ [31] and the five
quotient mappings of Michael [336].

Definition B.3.29 The class MOBI is the smallest class of spaces such that
(1) every metric space is in this class;
(2) this class is closed under open compact mappings.

The following question has not been solved: What kind of intrinsic characteriza-
tion does the class MOBI have? The following proposition of Bennett [56] is a direct
corollary of Definition B.3.29.

Proposition B.3.30 A space X belongs to the class MOBI if and only if there is a
metrizable space M and a finite set { f 1 , f 2 , . . . , f n } of open compact mappings such
that X = f n ◦ f n−1 . . . ◦ f 1 (M).

This proposition established a fine relationship among the class MOBI, the class
of metrizable spaces and the class of open compact mappings. In a period of twenty
years from the 1970s to the 1980s, all the research work on the class MOBI took
this proposition as a basis, and the result that every space in the class MOBI has a
point-countable base was obtained [31]. After a series of work, most questions raised
298 Appendix B: The Formation of the Theory of Generalized Metric Spaces

by Arhangel’skiı̌ at that time were answered negatively. For example, Chaber [99]
showed that a space in the class MOBI can be neither a weakly θ -refinable space nor
a σ  -space. However, several difficult questions, such as whether the class MOBI is
preserved under perfect mappings, whether every completely regular metacompact
space is an open compact image of a paracompact space and so on, are still open [35].
The class of quotient mappings is a class of relatively weak mappings, and hence
specific quotient images of metrizable spaces have been widely noticed. For exam-
ple, Arhangel’skiı̌ [31] raised the question for seeking the intrinsic characterizations
of quotient s-images of metrizable spaces; Michael and Nagami [340] asked whether
every quotient s-image of a metrizable space is a compact-covering quotient s-image
of a metrizable space. Hoshina [194], Gruenhage, Michael and Tanaka [167] gave
different answers for the former question. The latter one was answered negatively by
H. Chen [101, 102]. For generalizations of open mappings, except quotient mappings
and pseudo-open mappings mentioned above, we also have bi-quotient mappings (or
limit lifting mappings) defined by Hájek [171], Filippov [123] and Michael [332],
and countably bi-quotient mappings defined by Siwiec [424]. In a survey paper [336],
Michael summarized the research work of this direction done from the 1950s to the
early 1970s, and gave intrinsic characterizations for open images, bi-quotient images,
countably bi-quotient images, pseudo-open images and quotient images of locally
compact metrizable spaces, locally compact paracompact spaces, separable metriz-
able spaces, metrizable spaces, paracompact M-spaces and M-spaces systematically.
He also characterized generalized sequentiality properties in terms of such classes
of spaces and mappings.
In the following, we take sum theorems as an example to illustrate applications
of the mapping theory. The content that discussed in the sum theorems is that under
what conditions certain topological  properties of spaces can be transferred to their
sum space. That is, suppose X = α∈Λ X α and each subspace X α of the space X has
the topological property Φ. Then under what conditions does X have the property
Φ as well? The most simple and primitive statement in this respect should be the
following question given by Alexandroff and Urysohn [8]: if a compact space X
is the union of two subspaces and each of them has a countable base, then does X
have a countable base? Smirnov [429] gave this question a positive answer for the
countable sum case which is more general. Stone [442] studied the sum theorem for
metrizable spaces. In the following, we focus on “open sum theorems” and “closed
sum theorems”.
The early research on sum theorems is about individual topological properties.
From Hodel’s paper start [188], researchers realized the intrinsic relations among
mappings, spaces and topological properties. Investigating sum theorems by means
of mappings has become a trend. It also shows the importance of mapping methods
in general topology.
Definition B.3.31 A topological property Φ is said to satisfy the point-finite open
sum theorem if, X has property Φ whenever {X α }α∈Λ is a point-finite open cover of
a space X and each X α has property Φ.
Tanaka [452] first discussed the point-finite open sum theorem of σ -spaces. Git-
tings [157] investigated the point-finite open sum theorem and proved a topological
Appendix B: The Formation of the Theory of Generalized Metric Spaces 299

property Φ satisfies the point-finite open sum theorem if Φ is invariant under finite-
to-one open mappings and preserved by topological sums.
Definition B.3.32 Suppose P is a property of families of sets. A topological prop-
erty Φ is said to satisfy the P closed sum theorem if, a space X has property Φ
whenever {X α }α∈Λ is a closed cover of X with property P and every X α has prop-
erty Φ.
When property P is countable, locally finite, hereditarily closure-preserving and
closure-preserving, the corresponding theorems are called the countable closed sum
theorem, the locally finite closed sum theorem, the hereditarily closure-preserving
closed sum theorem and the closure-preserving closed sum theorem respectively.
One of the driving forces of investigating closed sum theorems is the following
question of Tamano [447]: Does paracompactness satisfy the closure-preserving
closed sum theorem? Potoczny [403] constructed a space with a closure-preserving
cover consisting of finite sets which is not a paracompact space.
The relationship between closed sum theorems and mappings is revealed by the
following theorem.
Definition B.3.33 Suppose Φ is a topological property preserved by topological
sums.
(1) If Φ is invariant under finite-to-one closed mappings, then Φ satisfies the locally
finite closed sum theorem.
(2) If Φ is invariant under closed mappings, then Φ satisfies the hereditarily closure-
preserving closed sum theorem.
One concept more general than that of hereditarily closure-preserving families is
the concept of dominating families introduced by Morita [359].
Definition B.3.34 Suppose P is a closed cover of a space X . We say that X is
dominated by P if, a subset Z of X is closed in X if and only if there is a subfamily
P  of P such that P  covers Z and P ∩ Z is closed in X for every P ∈ P  .
Morita raised the question what kind of topological properties satisfies the domi-
nating sum theorem and proved that several topological properties, such as normality
etc., satisfy the dominating sum theorem. Singal and Arya [423] investigated general
dominating sum theorems and sum theorems for adjunction spaces.

B.3.4 g-Functions

At Wisconsin conference of set theoretic topology in 1955, McAuley [322] raised

the question of seeking pure topological characterizations of semi-metrizable space.
Brown [72] also asked the following question: in order to make a semi-metrizable
space be a developable space, what kind of topological properties should be added?
In 1962, Heath [176] introduced a class of set valued functions, which is called
g-functions later, and gave satisfactory answers to these two questions.
300 Appendix B: The Formation of the Theory of Generalized Metric Spaces

Definition B.3.35 Let X be a space. A function g : N× X → τ is called a g-function

on X if x ∈ g(n + 1, x) ⊂ g(n, x) for every x ∈ X and n ∈ N.

The importance of g-functions is reflected in the pioneering method of using this

function developed by Heath. After solving the questions of McAuley and Brown,
Heath [182] characterized stratifiable spaces and σ -spaces in terms of g-functions.
He proved every stratifiable space is a σ -space, and hence solved a question raised
by Arhangel’skiı̌ [31], which displayed the effect of g-functions.
In order to discuss the metrization problems of spaces with a G δ -diagonal, Borges
[66] introduced the class of wΔ-spaces as a generalization of the class of developable
spaces. For investigating the precise relationships among developable spaces, wΔ-
spaces and semi-stratifiable spaces, Hodel [189] gave the definition of α-spaces and
proved that a space is a developable space if and only if it is both a wΔ-space and
an α-space. Hodel [189] also defined β-spaces in order to generalize wΔ-spaces
and semi-stratifiable spaces, and proved that a space is a semi-stratifiable space if
and only if it is both a β-space and an α-space. In order to investigate metrization
problems of M-spaces, Ishii [202] introduced wM-spaces as a generalization of M  -
spaces. The following question raised by Ishii [202] is still open: is every wM-space
with a G δ -diagonal metrizable?
Hodel [190] defined wN -spaces as generalizations of wM-spaces and Nagata
spaces, and proved a space is a metrizable space if and only if it is a developable
wN -space. Hodel [190] also defined wγ -spaces, θ -spaces etc., and proved that a
space is a wM-space if and only if it is both a wγ -space and a wN -space. He also
proved that a space is a developable space if and only if it is a semi-stratifiable
θ -space. The following definition gives a list of spaces described by g-functions.

Definition B.3.36 Suppose g is a g-function on a space X . Consider the following

additional conditions:
(1) p ∈ g(n, z n ), g(n, z n ) ∩ g(n, yn )
= ∅ and yn ∈ g(n, xn ) ⇒ {xn } has an
accumulation point.
(2) g(n, p) ∩ g(n, xn )
= ∅ ⇒ {xn } has an accumulation point.
(3) p ∈ g(n, xn ) ⇒ {xn } has an accumulation point.
(4) yn ∈ g(n, p) and xn ∈ g(n, yn ) ⇒ {xn } has an accumulation point.
(5) { p, xn } ⊂ g(n, yn ) and yn ∈ g(n, p) ⇒ {xn } has an accumulation point.
(6) xn ∈ g(n, p) ⇒ {xn } has an accumulation point.
(7) { p, xn } ⊂ g(n, yn ) ⇒ {xn } has an accumulation point.

Spaces satisfying the above additional conditions are called wM-spaces, wN -

spaces, β-spaces, wγ -spaces, wθ -spaces, q-spaces and wΔ-spaces in turn. If “{xn }
has an accumulation point” in the above additional conditions is strengthen to “{xn }
converges to p”, then spaces obtained are metrizable spaces, Nagata spaces, semi-
stratifiable spaces, γ -space, θ -spaces, first countable spaces and developable spaces
in turn.
Each of the above additional conditions says that every sequence in a space satisfy-
ing certain conditions has an accumulation point. Such spaces are called generalized
Appendix B: The Formation of the Theory of Generalized Metric Spaces 301

countably compact spaces by House [195]. By this idea, several classes of general-
ized metric spaces can be generated in terms of sequences of covers of spaces.

Definition B.3.37 Suppose {Un } is a sequence of open covers of a space X . Consider

the following additional conditions:
(1) xn ∈ st( p, Un ) ⇒ {xn } has an accumulation point.
(2) Un+1 is a star-refinement of Un and xn ∈ st( p, Un ) ⇒ {xn } has an accumulation
point.
(3) xn ∈ st2 ( p, Un ) ⇒ {xn } has an accumulation point.
Spaces satisfying the above additional conditions are wΔ-spaces, M-spaces and wM-
spaces in turn. If “{xn } has an accumulation point” in the above additional conditions
is strengthen to “{xn } converges to p”, then spaces defined are developable spaces,
metrizable spaces and metrizable spaces in turn.

C W C-functions are functions relevant to g-functions. It plays a special role in

the metrization problem.

Definition B.3.38 [241] Let X be a space. A function g : N × X → P(X ) (the

power set of X ) is called a C W C-function if, x ∈ g(n + 1, x) ⊂ g(n, x) for every
x ∈ X and n ∈ N, and U ∈ τ if and only if for each x ∈ U , there is n ∈ N such that
g(n, x) ⊂ U . A C W C-function is also called a weak base g-function [149].

The period from 1965 to 1975 is 10 years for rapid development of the theory of
generalized metric spaces. Especially, the survey paper entitled “Recent advances in
the theory of generalized metric spaces” written by Burke and Lutzer [88] in 1976
declared the formation of the theory of generalized metric spaces, and established
the status of this theory in general topology. Journals “General Topology and its
Applications” founded in Holland in 1971 and “Topology Proceedings” founded in
USA in 1976 both fully reflected this trend. A significant symbol is that the research
work for solving four problems formed in the foundation period greatly enriched the
theory of generalized metric spaces. At the same time, many new problems should
be solved. By the stimulating of the Bing-Nagata-Smirnov metrization theorem and
the organic combining of various concepts produced by generalizations of bases,
the research on generalized metric spaces merged into a torrent and was pushed to
a climax. Theorems of Siwiec and Nagata on characterizations of σ -spaces [426],
theorems of Nagami on countable products of paracompact Σ-spaces [368] and the-
orems of Foged on characterizations of ℵ-spaces [128] later, formed a wonderful
chapter in these important progresses. Through the systematic investigations on the
Alexandroff idea in Prague, mappings played an important role in the big stage of
general topology. Combined with many classic methods, mappings have become
essential means in studying the theory of spaces [4]. The deep problems and inci-
sive arguments of Arhangel’skiı̌ in “Mappings and spaces” [31], a comprehensive
description of Michael in “A quintuple quotient quest” [336], became one of the
important sources for general topology to move forward.
References

1. Alexander, C.C., Semi-developable spaces and quotient images of metric spaces, Pacific J.
Math., 37(1971), 277–293. MR47#2543. Zbl0216.19303.
2. Aleksandrov, P.S., On the metrisation of topological spaces (in Russian), Bull. Acad. Pol. Sci.,
Sér. Sci. Math. Astron. Phys., 8(1960), 135–140. MR22#5024. Zbl0124.37903.
3. Aleksandrov, P.S., On some results concerning topological spaces and their continuous map-
pings, in: Proc. 1st Topological Symp. (Prague, 1961), General Topology and its Relations
to Modern Analysis and Algebra I, Academic Press, New York, 1962, 41–54. MR26#3003.
Zbl0113.16506.
4. Aleksandrov, P.S., Fedorchuk, V.V., Zaı̌cev, V.I., The main aspects in the development of set-
theoretic topology (in Russian), Uspechi Mat. Nauk, 33(1978)(3), 3–48 (English translation:
Russian Math. Surveys, 33(1978)(3), 1-53). MR58#12879. Zbl0409.54001.
5. Alexandroff, P.S., Sur les ensembles de la premiére classe et les ensembles abstraits, C. R.
Acad. Paris., 178(1924), 185–187.
6. Alexandroff, P.S., On local properties of closed sets, Ann. Math., 36(1935)(2), 1–35.
MR1503204. Zbl0011.03901.
7. Alexandroff, P.S., Niemytzki, V.V., The condition of metrizability of topological spaces and
the axiom of symmetry (in Russian), Math. Sb., 3(1938), 663–672. Zbl0019.23504.
8. Alexandroff, P.S., Urysohn, P.S., Sur les espaces topologiques compacts, Bull. Intern. Acad.
Pol. Sci. Ser. A, 1923, 5–8.
9. Alexandroff, P.S., Urysohn, P.S., Mémoire sur les espaces topologiques compacts, Verh.
Koninkl. Akad. Wetensch., 14(1929), 1–96. JFM55.0960.02.
10. Alleche, B., Arhangel’skiı̌, A.V., Calbrix, J., Weak developments and metrization, Topology
Appl., 100(2000), 23–38. MR2001g:54029. Zbl0935.54027.
11. Alster, K., Metric spaces all of whose decompositions are metric, Bull. Acad. Pol. Sci., Sér.
Sci. Math. Astron. Phys., 20(1972), 395–400. MR46#6308. Zbl0242.54032.
12. Alster, K., Burke, D.K., Davis, S., The wΔ-space problem, Topology Appl., 30(1988), 175–
181. MR90b:54017. Zbl0659.54020.
13. Amono, M., Mizokami, T., On the position of Ceder space and McAuley space, Topology
Proc., 32(2008), 89–100. MR2009i: 54043. Zbl1158.54010.
14. An, T.V., Tuyen, L.Q., On an affirmative answer to S. Lin’s problem, Topology Appl.,
158(2011), 1567–1570. MR2812465. Zbl1222.54018.
15. An, T.V., Tuyen, L.Q., On π -images of separable metric spaces and a problem of Shou Lin,
Mat. Vesnik, 64(2012)(4), 297–302. MR2965962. Zbl1289.54058.
16. Anderson, B.A., Metric topologies, Ph. D. Thesis, University of Iowa, 1966.
17. Anderson, B.A., Topologies comparable to metric topologies, in: Proc. Arizona State Univ.
Topological Conf. (Tempe, 1967), 1969, 15–21. Zbl0211.25602.

© Atlantis Press and the author(s) 2016 303

S. Lin and Z. Yun, Generalized Metric Spaces and Mappings,
Atlantis Studies in Mathematics 6, DOI 10.2991/978-94-6239-216-8
304 References

18. Aquaro, G., Point countable open covering in countably compact spaces, in: Proc. 2nd Prague
Topological Symp. (Prague, 1966), General Topology and its Relations to Modern Analysis
and Algebra II, 1967, 39–41. Zbl0157.29403.
19. Arens, R., Note on convergence in topology, Math. Mag., 23(1950), 229–234. MR12,271h.
Zbl0041.31502.
20. Arens, R., Dugundji, J., Remark on the concept of compactness, Portug. Math., 9(1950),
141–143. MR12,434b. Zbl0039.18602.
21. Arhangel’skiı̌, A.V., An addition theorem for the weight of sets lying in bicompacts (in
Russian), Dokl. Akad. Nauk SSSR, 126(1959)(2), 239–241. MR21#5176. Zbl0087.37602.
22. Arhangel’skiı̌, A.V., Concerning the weight of topological spaces, in: Proc. 1st Topological
Symp. (Prague, 1961), General Topology and its Relations to Modern Analysis and Algebra
I, Academic Press, New York, 1962, 72–74. MR30#4237.
23. Arhangel’skiı̌, A.V., On mappings of metric spaces (in Russian), Dokl. Akad. Nauk
SSSR, 145(1962), 245–247 (English translation: Soviet Math. Dokl., 3(1962), 953-956).
MR25#2579. Zbl0124.15802.
24. Arhangel’skiı̌, A.V., On open and almost open mappings of topological spaces (in Russian),
Dokl. Akad. Nauk SSSR, 147(1962), 999–1002. MR26#739.
25. Arhangel’skiı̌, A.V., On a class of spaces containing all metric and all locally bicompact spaces
(in Russian), Dokl. Akad. Nauk SSSR, 151(1963), 751–754 (English translation: Soviet Math.
Dokl., 4(1963), 1051–1055). MR27#2959. Zbl0124.15801.
26. Arhangel’skiı̌, A.V., Some types of factor mappings and the relations between classes of topo-
logical spaces (in Russian), Dokl. Akad. Nauk SSSR, 153(1963), 743–746 (English translation:
Soviet Math. Dokl., 4(1963), 1726-1729). MR28#1587. Zbl0129.38103.
27. Arhangel’skiı̌, A.V., Some metrization theorems (in Russian), Uspechi Mat. Nauk,
18(1963)(5), 139–145. MR27#6242. Zbl0128.16701.
28. Arhangel’skiı̌, A.V., Bicompact sets and the topology of spaces (in Russian), Trudy Moskov
Mat. Obsch., 13(1965), 3–55 (English translation: Trans. Moscow Math. Soc., 13(1965), 1-
62). MR33#3251. Zbl0162.26602.
29. Arhangel’skiı̌, A.V., Behavior of metrizability in factor mappings (in Russian), Dokl. Akad.
Nauk SSSR, 164(1965)(2), 247–250. MR33#697. Zbl0136.43602.
30. Arhangel’skiı̌, A.V., On a class of spaces containing all metric and all locally bicompact
spaces (in Russian), Mat. Sb. (N.S.), 67(1965)(109), 55-88 (English translation: Amer. Math.
Soc. Transl., 92(1970), 1-39). MR27#2959.
31. Arhangel’skiı̌, A.V., Mappings and spaces (in Russian), Uspechi Mat. Nauk, 21(1966)(4),
133–184 (English translation: Russian Math. Surveys, 21(1966), 115-162). MR37#3534.
Zbl0171.43603.
32. Arhangel’skiı̌, A.V., Perfect maps and condensations (in Russian), Dokl. Akad. Nauk
SSSR, 176(1967), 983–986 (English translation: Soviet Math. Dokl., 8(1967), 1217-1220).
MR38#6552. Zbl0172.24403.
33. Arhangel’skiı̌, A.V., Intersection of topologies, and pseudo-open bicompact mappings (in
Russian), Dokl. Akad. Nauk SSSR, 226(1976)(4), 745–748 (English translation: Soviet Math.
Dokl., 17(1976)(1), 160-163). MR53#6487. Zbl0344.54010.
34. Arhangel’skiı̌, A.V., General Topology III, Encyclopaedia of Mathematical Sciences, V. 51,
Springer-Verlag, Berlin, 1995. MR97f:54001. Zbl0826.00014.
35. Arhangel’skiı̌, A.V., Notes on the history of general topology in Russia, Topology Proc.,
25(2000)(Spring), 353–395. MR2003c:01029. Zbl1006.54001.
36. Arhangel’skiı̌, A.V., Buzyakova, R.Z., Addition theorems and D-spaces, Comment. Math.
Univ. Carolinae, 43(2002), 653–663. MR2005a:54033. Zbl1090.54017.
37. Arhangel’skiı̌, A.V., Franklin, S.P., Ordinal invariants for topological spaces, Michigan Math.
J., 15(1968), 313–320. MR39#2112. Zbl0167.51102.
38. Arhangel’skiı̌, A.V., Just, W., Wicke, H., Not all pseudo-open maps are compositions of closed
maps and open maps, Topology Proc., 19(1994), 3–14. MR96k:54011. Zbl0843.54021.
39. Arhangel’skiı̌, A.V., Ponomarev, V.I., Fundamentals of General Topology: Problems and Exer-
cises (in Russian), Hayka, Moscow, 1974 (English edition translated from the Russian by Jain
References 305

V.K., Mathematics and its Applications, V. 13, Kluwer Academic Publishers, Dordrecht,
1984). MR87i:54001. Zbl0568.54001.
40. Arhangel’skiı̌, A.V., Pontryagin, L.S., General Topology I, Encyclopaedia of Mathematical
Sciences, V. 17, Springer-Verlag, Berlin, 1990. MR91g:54001. Zbl0778.00007.
41. Arhangel’skiı̌, A.V., Projzvolov, V.V., On the connection between pointwise cardinality of
systems of subsets of a compact space and its weight (in Russian), Vestnik Moskov Univ. Ser.
I Mat. Meh., 21(1966)(3), 75–77. MR33#3252. Zbl0163.17203.
42. Aronszajn, N., Über ein urbildproblem, Fund. Math., 17(1931), 92–121. Zbl0003.02703.
43. Atkins, J.M., Slaughter, jr.F.G., On the metrizability of preimages of metric spaces under
closed continuous functions, in: Proc. Univ. Oklahoma Topological Conf., Norman, 1972,
13–22. MR54#13864. Zbl0246.54009.
44. Aull, C.E., Quasi-developments and δθ-bases, J. London Math. Soc., 9(1974), 197–204.
MR52#9171. Zbl0295.54023.
45. Aull, C.E., A survey paper on some base axioms, Topology Proc., 3(1978), 1–36.
MR80m:54044. Zbl0409.54038.
46. Aull, C.E., Lowen, R., Handbook of the History of General Topology, 1, Kluwer Academic
Publishers, Dordrecht, 1997. MR99a:01001. Zbl0888.54001.
47. Aull, C.E., Lowen, R., Handbook of the History of General Topology, 2, Kluwer Academic
Publishers, Dordrecht, 1998. MR2001e:54001. Zbl0902.54001.
48. Aull, C.E., Lowen, R., Handbook of the History of General Topology, 3, Kluwer Academic
Publishers, Dordrecht, 2001. MR2002k:54001. Zbl0980.00038.
49. Bacon, P., The compactness of countably compact spaces, Pacific J. Math., 32(1970), 587–
592. MR41#2624. Zbl0175.49503.
50. Baer, R.W., Levi, F., Stetige funktionen in topologischen räumen, Math. Zeitschr., 34(1932),
110–130. MR1545244. Zbl0002.16002.
51. Balogh, Z.T., Topological spaces with point-networks, Proc. Amer. Math. Soc., 94(1985),
497–501. MR87a:54040. Zbl0584.54024.
52. Balogh, Z.T., Nonshrinking open covers and K. Morita’s duality conjectures, Topology Appl.,
115(2001), 333–341. MR2002j:54030. Zbl0983.54023.
53. Balogh, Z.T., Gruenhage, G., When the collection of -balls is locally finite, Topology Appl.,
124(2002)(3), 445–450. MR2003h:54038. Zbl1015.54011.
54. Banakh, T., Bogachev, V., Kolesnikov, A., k ∗ -metrizable spaces and their applications, J.
Math. Sci., 155(2008)(4), 475–522. MR2731965. Zbl06406133.
55. Bennett, H.R., Quasi-developable spaces, in: Proc. Arizona State Univ. Topological
Conf. (Tempe, 1967), 1969, 314–317. Zbl0236.54021.
56. Bennett, H.R., On Arhangel’skiı̌’s class MOBI, Proc. Amer. Math. Soc., 26(1970), 178–180.
MR42#2425. Zbl0197.48502.
57. Bennett, H.R., On quasi-developable spaces, General Topology Appl., 1(1971), 253–262.
MR44#5921. Zbl0222.54037.
58. Bennett, H.R., Byerly, R., Lutzer, D.J., Compact G δ sets, Topology Appl., 153(2006), 2169–
2181. MR2008a:54038. Zbl1101.54034.
59. Bennett, H.R., Lutzer, D.J., A note on weak θ-refinability, General Topology Appl., 2(1972),
49–54. MR46#853. Zbl0229.54022.
60. Bennett, H.R., Lutzer, D.J., The β-space property in monotonically normal spaces and GO-
spaces, Topology Appl., 153(2006), 2218–2228. MR2007e:54028. Zbl1101.54028.
61. Berney, E.S., A regular Lindelöf semi-metric space which has no countable network, Proc.
Amer. Math. Soc., 26(1970), 361–364. MR42#5225. Zbl0198.55602.
62. Bing, R.H., Metrization of topological spaces, Canad. J. Math., 3(1951), 175–186.
MR13,264f. Zbl0042.41301.
63. Boone, J.R., Some characterizations of paracompactness in k-spaces, Fund. Math., 72(1971),
145–153. MR45#4359. Zbl0223.54013.
64. Boone, J.R., Siwiec, F., Sequentially quotient mappings, Czech. Math. J., 26(1976), 174–182.
MR53#6505. Zbl0334.54003.
306 References

65. Borges, C.R., On stratifiable spaces, Pacific J. Math., 17(1966), 1–16. MR32#6409.
Zbl0175.19802.
66. Borges, C.R., On metrizability of topological spaces, Canad. J. Math., 20(1968), 795–804.
MR37#6910. Zbl0167.21201.
67. Borges, C.R., Metrizability of adjunction spaces, Proc. Amer. Math. Soc., 24(1970), 446–451.
MR41#7623. Zbl0189.53704.
68. Borges, C.R., Lutzer, D.J., Characterizations and mappings of Mi -spaces, in: Topology Con-
ference (Virginia Polytech. Inst. and State Univ., Blacksburg, 1973), Lecture Notes in Math.,
V. 375, Springer-Verlag, Berlin, 1974, 34–40. MR50#14681. Zbl0286.54014.
69. Borsuk, K., Quelques rétracts singuliers, Fund. Math., 24(1935), 249–258. Zbl0011.04004.
70. Bourbaki, N., Topologie Générale, Ch. I et II, Hermann, Paris, 1940. MR3,55e.
Zbl0026.43101.
71. Bourbaki, N., General Topology, Part 1, Addison-Wesley Educational Publishers Inc., U.S.,
1966. Zbl0301.54001.
72. Brown, M., Semi-metric spaces, in: Summer Institute on Set Theoretic Topology (Wisconsin,
1955), 1958, 62–64.
73. Buck, R.E., Some weaker monotone separation and basis properties, Topology Appl.,
69(1996), 1–12. MR97g:54038. Zbl0848.54015.
74. Buhagiar, D., Lin, S., A note on subparacompact spaces, Mat. Vesnik, 52(2000)(3–4), 119–
123. MR2002j:54022. Zbl1054.54508.
75. Burke, D.K., On subparacompact spaces, Proc. Amer. Math. Soc., 23(1969), 655–663.
MR40#3508. Zbl0187.19902.
76. Burke, D.K., Subparacompact spaces, in: Proc. Washington State Univ. Topological Conf.,
1970, 39–49. MR42#1066. Zbl0198.27501.
77. Burke, D.K., On p-spaces and wΔ-spaces, Pacific J. Math., 35(1970), 285–296. MR43#3986.
Zbl0189.53403.
78. Burke, D.K., Cauchy sequences in semimetric spaces, Proc. Amer. Math. Soc., 33(1972),
161–164. MR44#7512. Zbl0233.54015.
79. Burke, D.K., A nondevelopable locally compact Hausdorff space with a G δ -diagonal, General
Topology Appl., 2(1972), 287–291. MR47#7702. Zbl0243.54017.
80. Burke, D.K., Spaces with G δ -diagonal, in: TOPO 72–general topology and its applications,
Proc. 2nd Pittsburgh Internat. Conf. (Pittsburgh, 1972), Lecture Notes in Math., V. 378,
Springer-Verlag, Berlin, 1974, 95–100. MR51#1755. Zbl0286.54018.
81. Burke, D.K., Preservation of certain base axioms under a perfect mapping, Topology Proc.,
1(1976), 269–279. MR56#6621. Zbl0392.54007.
82. Burke, D.K., Closed mappings, in: Reed, G.M. ed., Surveys in General Topology, Academic
Press, New York, 1980, 1–32. MR81c:54014. Zbl0441.54005.
83. Burke, D.K., Covering properties, in: Kunen, K., Vaughan, J.E. eds., Handbook of
Set-theoretic Topology, Elsevier Science Publishers B. V., Amsterdam, 1984, 347–422.
MR86e:54030. Zbl0569.54022.
84. Burke, D.K., Weak-bases and D-spaces, Comment. Math. Univ. Carolinae, 48(2007)(2), 281–
289. MR2008g:54009. Zbl1199.54065.
85. Burke, D.K., Davis, S. W., Spaces with a σ -weakly hereditarily closure perserving base,
Topology Proc., 35(2010), 9–18. MR2010c:54034. Zbl1181.54028.
86. Burke, D.K., van Douwen, E.K., On countably compact extensions of normal locally compact
M-spaces, in: Reed, G.M. ed., Set-theoretic Topology (Papers, Inst. Medicine and Math.,
Ohio Univ., Athens, 1975–1976), Academic Press, New York, 1977, 81–89. MR55#13381.
Zbl0436.54023.
87. Burke, D.K., Engelking, R., Lutzer, D.J., Hereditarily closure-preserving collections and
metrization, Proc. Amer. Math. Soc., 51(1975), 483–488. MR51#6746. Zbl0307.54030.
88. Burke, D.K., Lutzer, D.J., Recent advances in the theory of generalized metric spaces, in:
Topology: Proc. 9th Ann. Spring Topological Conf. (Memphis State Univ., 1975), Lecture
Notes in Pure and Applied Mathematics, V. 24, Marcel Dekker Inc., New York, 1976, 1–70.
MR55#1318. Zbl0353.54020.
References 307

89. Burke, D.K., Michael, E.A., On a theorem of V.V. Filippov, Israel J. Math., 11(1972), 394–
397. MR47#1022. Zbl0236.54015.
90. Burke, D.K., Michael, E.A., On certain point-countable covers, Pacific J. Math., 64(1976),
79–92. MR57#7542. Zbl0341.54022.
91. Burke, D.K., Stoltenberg, R.A., A note on p-spaces and Moore spaces, Pacific J. Math.,
30(1969), 601–608. MR40#3507. Zbl0183.27502.
92. Ceder, J.G., Some generalizations of metric spaces, Pacific J. Math., 11(1961), 105–125.
MR24#A1707. Zbl0103.39101.
93. Chaber, J., Conditions which imply compactness in countably compact spaces, Bull. Acad.
Pol. Sci., Sér. Sci. Math. Astron. Phys., 24(1976), 993–998. MR58#24189. Zbl0347.54013.
94. Chaber, J., Primitive generalizations of σ -spaces, in: Császár, Á. ed., Topology, Colloq. Math.
Soc. János Bolyai (Budapest, 1978), V. 23, North-Holland, Amsterdam, 1980, 259–268.
MR81m:54054. Zbl0442.54020.
95. Chaber, J., Perfect images of p-spaces, Proc. Amer. Math. Soc., 85(1982), 609–614.
MR83h:54037. Zbl0491.54007.
96. Chaber, J., Perfect preimages of Moore spaces, Bull. Pol. Acad. Sci., Math., 31(1983), 31–34.
MR85e:54030. Zbl0542.54026.
97. Chaber, J., Generalizations of Lašnev’s theorem, Fund. Math., 119(1983), 85–91.
MR85f:54022. Zbl0547.54009.
98. Chaber, J., On the class MOBI, in: Frolík, Z. ed., Proc. 6th Prague Topological Symp. (Prague,
1986), General Topology and its Relations to Modern Analysis and Algebra VI, Heldermann
Verlag, Berlin, 1988, 77–82. MR90b:54009. Zbl0636.54016.
99. Chaber, J., More nondevelopable spaces in MOBI, Proc. Amer. Math. Soc., 103(1988)(1):
307–313. MR89k:54030. Zbl0648.54017.
100. Charalambous, M.G., Resolving a question of Arkhangel’skiı̌’s, Fund. Math., 192(2006)(1),
67–76. MR2007m:54030. Zbl1107.54027.
101. Chen, H., Weak neighborhoods and Michael-Nagami’s question, Houston J. Math., 25(1999),
297–309. MR2000d:54015. Zbl0965.54031.
102. Chen, H., Compact-covering maps and k-networks, Proc. Amer. Math. Soc., 131(2003), 2623–
2632. MR2004f:54013. Zbl1025.54022.
103. Choban, M.M., Perfect mappings and spaces of countable type (in Russian), Vestnik Moskov
Univ. Ser. I Mat. Meh., 22(1967)(6), 87–96. MR36#3313.
104. Choban, M.M., The open mappings and spaces. Ren. Circolo. Math. Palermo, 29(1992)(1),
51–104. MR93m:54026.
105. Chow, S., Le probl’eme intégral de la localisation des ensembles ponctuels plans bornés à
paratingent incomplet, Fund. Math., 29(1937), 12–21. Zbl0017.03802.
106. Collins, P.J., Roscoe, A.W., Criteria for metrisability, Proc. Amer. Math. Soc., 90(1984), 631–
640. MR85c:54041. Zbl0541.54034.
107. Collins, P.J., Reed, G.M., Roscoe, A.W., Rudin, M.E., A lattice of conditions on topological
spaces, Proc. Amer. Math. Soc., 94(1985), 487–496. MR87b:54018. Zbl0562.54043.
108. Corson, H.H., Normality in subsets of product spaces, Ann. J. Math., 81(1959), 785–796.
MR21#5947. Zbl0095.37302.
109. Creede, G.D., Semi-stratifiable spaces, in: Proc. Arizona State Univ. Topological
Conf. (Tempe, 1967), 1969, 318–323. Zbl0211.25702.
110. Creede, G.D., Concerning semi-stratifiable spaces, Pacific J. Math., 32(1970), 47–54.
MR40#8006. Zbl0189.23304.
111. Czechoslovak Academy of Sciences, Inter. Math. Union., General Topology and its Relations
to Modern Analysis and Algebra I, in: Proc. 1st Topological Symp. (Pragu, 1961), Academic
Press, New York, 1962, pp. 5.
112. Dieudonné, J., Une généralisation des espaces compacts, J. Math. Pures Appl., 23(1944),
65–76. MR7,134f. Zbl0060.39508.
113. van Douwen, E.K., Simultaneous extension of continuous functions, Thesis, Free University,
Amsterdam, 1975. Zbl0309.54013.
308 References

114. van Douwen, E.K., Pfeffer, W.F., Some properties of the Sorgenfrey line and related spaces,
Pacific J. Math., 81(1979), 371–377. MR80h:54027. Zbl0409.54011.
115. Dow, A., Martínez, R.R., Tkachuk, V.V., A glance at spaces with closure-preserving local
bases, Topology Appl., 157(2010)(3), 548–558. MR2010m:54004. Zbl1187.54021.
116. Dowker, C.H., On countably paracompact spaces, Canad. J. Math., 3(1951), 219–224.
MR13,264c. Zbl0042.41007.
117. Dugundji, J., Topology, Allyn and Bacon Inc., Boston, 1966. MR33#1824. Zbl0144.21501.
118. Eda, K., Gruenhage, G., Koszmider, P., Tamano, K., Todorčevic̀, S., Sequential fans in topol-
ogy, Topology Appl., 67(1995), 189–220. MR96k:54007. Zbl0868.54001.
119. Engelking, R., General Topology (revised and completed edition), Heldermann Verlag, Berlin,
1989. MR91c:54001. Zbl0684.54001.
120. Fedeli, A., Le Donne, A., On good connected preimages, Topology Appl., 125(2002), 489–496.
MR2003g:54062. Zbl1019.54008.
121. Filippov, V.V., The perfect image of a paracompact feathered space (in Russian), Dokl. Akad.
Nauk SSSR, 176(1967)(3), 533–535 (English translation: Soviet Math. Dokl., 8(1967), 1151–
1153). MR36#5903. Zbl0167.51202.
122. Filippov, V.V., Preservation of the order of a base under a perfect mapping (in Russian), Dokl.
Akad. Nauk SSSR, 181(1968), 1077-1079 (English translation: Soviet Math. Dokl., 9(1968),
1005-1007). MR38#683. Zbl0186.56004.
123. Filippov, V.V., Quotient spaces and multiplicity of a base (in Russian), Mat. Sb.,
80(1969)(4), 521–532 (English translation: Math. USSR Sb., 9(1969), 487–496). MR41#6179.
Zbl0202.53801.
124. Fletcher, P., Lindgren, W.F., Orthocompactness and strong Čech completeness in Moore
spaces, Duke Math. J., 39(1972), 753–766. MR47#1029. Zbl0251.54013.
125. Fletcher, P., Lindgren, W.F., On wΔ-spaces, wΔ-spaces and Σ  -spaces, Pacific J. Math.,
71(1977), 419-428. MR58#12919. Zbl0361.54012.
126. Foged, L., Weak bases for topological spaces, Ph. D. Thesis, Washington University, Missouri,
1979.
127. Foged, L., On g-metrizability, Pacific J. Math., 98(1982), 327–332. MR84c:54054.
Zbl0478.54025.
128. Foged, L., Characterizations of ℵ-spaces, Pacific J. Math., 110(1984), 59–63. MR85d:54037.
Zbl0542.54030.
129. Foged, L., Sequential coreflections of stratifiable spaces, Proc. Amer. Math. Soc., 92(1984),
470–472. MR86a:54034. Zbl0549.54021.
130. Foged, L., A characterization of closed images of metric spaces, Proc. Amer. Math. Soc.,
95(1985), 487–490. MR87c:54020. Zbl0592.54027.
131. Foged, L., Normality in k-and ℵ-spaces, Topology Appl., 22(1986), 223–240. MR87i:54045.
Zbl0602.54030.
132. Foged, L., Point-countable bases and k-networks, Topology Appl., 69(1996), 101–114.
MR97b:54036. Zbl0857.54024.
133. Franklin, S.P., Spaces in which sequences suffice, Fund. Math., 57(1965), 107–115.
MR31#5184. Zbl0132.17802.
134. Frolík, Z., Generalizations of compact and the Lindelöf spaces (in Russian), Czech. Math. J.,
9(1959), 172–217. MR21#3821. Zbl0098.14201.
135. Frolík, Z., On the topological product of paracompact spaces, Bull. Acad. Pol. Sci., Sér. Sci.
Math. Astron. Phys., 8(1960), 747–750. MR23#A2859. Zbl0099.38601.
136. Gale, D., Compact sets of functions and function rings, Proc. Amer. Math. Soc., 1(1950),
303–308. MR12,119d. Zbl0037.35501.
137. Gao, G., On hereditarity of paracompactness (in Chinese), J. Jiangsu Normal Univ. (Natural
Science), (1979)(1), 10–13 (This paper is completed in 1964).
138. Gao, G., On the closure-perserving sum theorems (in Chinese), Acta Math. Sin., 29(1986),
58–62. MR87m:54065. Zbl0607.54006.
139. Gao, G., On k-networks and bases (in Chinese), J. Suzhou Univ. (Natural Science), 2(1986),
107–111.
References 309

140. Gao, G., Theory of Topological Spaces (in Chinese), Science Press, Beijing, 2000.
Zbl0969.54001.
141. Gao, G., Theory of Topological Spaces, Second Edition (in Chinese), Science Press, Beijing,
2008.
142. Gao, G., et al., Paracompactness and Generalized Metric Spaces, Jiangsu Science and Tech-
nology Press, Nanjing, 1988.
143. Gao, Y., A note concerning the Collins, Reed, Roscoe, Rudin metrization theorem, Topology
Appl., 74(1996), 73–82. MR98b:54037a. Zbl0883.54028.
144. Gao, Z., Some results on K -semi-stratifiable space (in Chinese), J. Northwest Univ., (1985)(3),
12–16. MR87g:54064.
145. Gao, Z., On g-function separation, Questions Answers in General Topology, 4(1986), 47–57.
MR87j:54051. Zbl0597.54027.
146. Gao, Z., ℵ-space is invariant under perfect mappings, Questions Answers in General Topology,
5(1987), 271–279. MR89b:54033a. Zbl0636.54026.
147. Gao, Z., The closed images of metric spaces and Fréchet ℵ-spaces, Questions Answers in
General Topology, 5(1987), 281–291. MR89b:54033b. Zbl0643.54035.
148. Gao, Z., On J. Nagata’s question, Math. Japonica, 51(2000)(1), 49–52. MR2001b:54034.
Zbl0948.54027.
149. Gao, Z., Metrizability of spaces and weak base g-functions, Topology Appl., 146/147(2005),
279–288. MR2006b:54021. Zbl1059.54023.
150. Gao, Z., Hattori, Y., A characterization of closed s-images of metric spaces, Questions Answers
in General Topology, 4(1986/87), 147–151. MR0917897. Zbl0617.54024.
151. Gartside, P.M., Reznichenko, E.A., Near metric propertics of function spaces, Fund. Math.,
164(2000)(2), 97–114. MR2002b:54026. Zbl0971.46012.
152. Ge, Y., On pseudo-sequence coverings, π -images of metric spaces, Mat. Vesnik, 57(2005)(3–
4), 113–120. MR2194600. Zbl1100.41013.
153. Ge, Y., On three equivalences concerning Ponomarev-systems, Arch. Math., 42(2006)(3),
239–246. MR2007j:54023. Zbl1164.54363.
154. Gillman, L., Jerison, M., Rings of continuous functions, Van Nostrand, Princeton, 1960 (Grad-
uate Texts in Math., V. 43, Springer-Verlag, Berlin, 1976). MR53#11352. Zbl0327.46040.
155. Gittings, R.F., Some results on weak covering conditions, Canad. J. Math., 26(1974), 1152–
1156. MR50#5738. Zbl0268.54018.
156. Gittings, R.F., Concerning quasi-complete spaces, General Topology Appl., 6(1976), 73–89.
MR52#11855. Zbl0323.54025.
157. Gittings, R.F., Open mapping theory, in: Reed, G.M. ed., Set-theoretic Topology (Papers,
Inst. Medicine and Math., Ohio Univ., Athens, 1975–1976), Academic Press, New York,
1977, 141–191. MR58#2687. Zbl0363.00011.
158. Gittings, R.F., Products of generalized metric spaces, Rocky Mountain J. Math., 9(1979)(3),
479–497. MR80d:54004. Zbl0417.54007.
159. Good, C., Knight, R.W., Mohamad, A.M., On the metrizability of spaces with a sharp base,
Topology Appl., 125(2002)(3), 543–552; erratum ibid, 143(2004), 291–292. MR2003h:54039.
Zbl1018.54021.
160. Gruenhage, G., Stratifiable spaces are M2 , Topology Proc., l(1976), 221–226. MR56#6614.
Zbl0389.54019.
161. Gruenhage, G., On the M3 ⇒ M1 question, Topology Proc., 5(1980), 77–104. MR82h:54043.
Zbl0463.54023
162. Gruenhage, G., Generalized metric spaces, in: Kunen, K., Vaughan, J.E. eds., Handbook
of Set-theoretic Topology, Elsevier Science Publishers B. V., Amsterdam, 1984, 423–501.
MR86h:54038. Zbl0555.54015.
163. Gruenhage, G., Generalized metric spaces and metrization, in: Hušek, M., van Mill, J.
eds., Recent Progress in General Topology (Papers from the Prague Topological Symp.,
Prague, 1991), Elsevier Science Publishers B. V., Amsterdam, 1992, 239–274. MR1229128.
Zbl0794.54034.
310 References

164. Gruenhage, G., Metrizable spaces and generalizations, in: Hušek, M., van Mill, J. eds., Recent
Progress in General Topology II (Papers from the Prague Topological Symp., Prague, 2001),
Elsevier Science Publishers B. V., Amsterdam, 2002, 201–225. MR1969999. Zbl1029.54036.
165. Gruenhage, G., A survey of D-spaces, Contemp. Math., 533(2011), 13–28. MR2012e:54027.
166. Gruenhage, G., Generalized metrizable spaces, in: Hart, K.P., van Mill, J., Simon, P. eds.,
Recent Progress in General Topology III (Papers from the Prague Topological Symp., Prague,
2011), Atlantis Press, Pairs, 2014, 459–492. MR3205490. Zbl1314.54001.
167. Gruenhage, G., Michael, E.A., Tanaka, Y., Spaces determined by point-countable covers,
Pacific J. Math., 113(1984), 303–332. MR85m:54018. Zbl0561.54016.
168. Gul’ko, S.P., On properties of subsets of Σ-products (in Russian), Dokl. Akad. Nauk SSSR,
237(1977)(3), 505–508 (English translation: Soviet Math. Dokl., 18(1977), 1438–1442).
MR57#1395. Zbl0397.54012.
169. Guthrie, J.A., A characterization of ℵ0 -spaces, General Topology Appl., 1(1971), 105–110.
MR44#5922. Zbl0216.19103.
170. Guthrie, J.A., Mapping spaces and cs-networks, Pacific J. Math., 47(1973), 465–471.
MR49#3821. Zbl0253.54025.
171. Hájek, O., Note on quotient maps, Comment. Math. Univ. Carolinae, 7(1966), 319–323.
MR34#1992. Zbl0149.19304.
172. Halfar, E., Compact mappings, Proc. Amer. Math. Soc., 8(1957), 828–830. MR89388.
Zbl0078.36004.
173. Hanai, S., On closed mappings II, Proc. Japan Acad., 32(1956), 388–391. MR18,225b.
174. Hart, K.P., van Mill, J., Simon, P., Recent Progress in General Topology III (Papers from
the Prague Topological Symp., Prague, 2011), Atlantis Press, Pairs, 2014. MR3204729.
Zbl1282.54001.
175. Hart, K.P., Nagata, J., Vaughan, J.E., Encyclopedia of General Topology, Elsevier Science
Publishers B. V., Amsterdam, 2004. MR2005d:54001. Zbl1059.54001.
176. Heath, R.W., Arc-wise connectedness in semi-metric spaces, Pacific J. Math., 12(1962), 1301–
1319. MR29#4032. Zbl0113.16501.
177. Heath, R.W., Screenability, pointwise paracompactness and metrization of Moore spaces,
Canad. J. Math., 16(1964), 763–770. MR29#4033. Zbl0122.17401.
178. Heath, R.W., On spaces with point-countable bases, Bull. Acad. Pol. Sci., Sér. Sci. Math.
Astron. Phys., 13(1965), 393–395. MR32#4656. Zbl0132.18402.
179. Heath, R.W., On open mappings and certain spaces satisfying the first countability axiom,
Fund. Math., 57(1965), 91–96. MR31#4006. Zbl0134.41802.
180. Heath, R.W., A paracompact semi-metric space which is not an M3 -space, Proc. Amer. Math.
Soc., 17(1966), 868–870. MR33#3256. Zbl0151.30201.
181. Heath, R.W., On certain first countable spaces, Ann. Math. Stud., 60(1966), 103-113.
Zbl0147.41603.
182. Heath, R.W., Stratifiable spaces are σ -spaces, Notices Amer. Math. Soc., 17(1969), 761–761.
183. Heath, R.W., Problem I, General Topology Appl., 1(1971), iv.
184. Heath, R.W., Hodel, R.E., Characterizations of σ -spaces, Fund. Math., 77(1973), 27l–275.
MR47#9549. Zbl0258.54027.
185. Heath, R.W., Junnila, H.J.K., Stratifiable spaces as subspaces and continuous images of M1 -
spaces, Proc. Amer. Math. Soc., 83(1981), 146–148. MR82f:54045. Zbl0476.54024.
186. Heath, R.W., Lutzer, D.J., Zenor, P.L., Monotonically normal spaces, Trans. Amer. Math. Soc.,
178(1973), 481–493. MR51#9030. Zbl0269.54009.
187. Hilbert, D., Mathematical problems (in German), Göttinger Nachrichten, (1900), 253–297
(English translation: Bull. Amer. Math. Soc., 8(1902), 437–479). MR1557926.
188. Hodel, R.E., Sum theorems for topological spaces, Pacific J. Math., 30(1969), 59–65.
MR40#6502. Zbl0181.50502.
189. Hodel, R.E., Moore spaces and wΔ-spaces, Pacific J. Math., 38(1971), 641–652. MR46#6290.
Zbl0219.54024.
190. Hodel, R.E., Spaces defined by sequences of open covers which guarantee that cer-
tain sequences has cluster points, Duke Math. J., 39(1972), 253–263. MR45#2657.
Zbl0242.54027.
References 311

191. Hodel, R.E., Some results in metrization theory, 1950 – 1972, in: Topology Conference
(Virginia Polytech. Inst. and State Univ., Blacksburg, 1973). Lecture Notes in Math., V. 375,
Springer-Verlag, Berlin, 1974, 120–136. MR50#8459. Zbl0287.54029.
192. Hodel, R.E., Metrizability of topological spaces, Pacific J. Math., 55(1974), 441–459.
MR51#6747. Zbl0286.54017.
193. Hodel, R.E., A history of generalized metrizable spaces, in: Aull, C.E., Lowen, R. eds.,
Handbook of the History of General Topology, V. 2, Kluwer Academic Publishers, Dordrecht,
1998, 541–576. MR2002a:54001. Zbl0940.54003.
194. Hoshina, T., On the quotient s-images of metric spaces, Sci. Rep. Tokyo Kyoiku Daigaku Sect.
A, 10(1970), 265–268. MR43#1115. Zbl0214.49503.
195. House, V.D., Countable products of generalized countably compact spaces, Pacific J. Math.,
57(1975), 183–197. MR51#11427. Zbl0301.54025.
196. Hurewicz, W., Ueber stetige bilder von punktmengen, Proc. Akad. Amsterdam, 29(1926),
1014–1017. JFM 52.0595.03.
197. Hušek, M., van Mill, J., Recent Progress in General Topology (Papers from the Prague
Topological Symp., Prague, 1991), Elsevier Science Publishers B. V., Amsterdam, 1992.
MR95g:54004. Zbl0782.00072.
198. Hušek, M., van Mill, J., Recent Progress in General Topology II (Papers from the Prague
Topological Symp., Prague, 2001), Elsevier Science Publishers B. V., Amsterdam, 2002.
MR2004e:54001. Zbl1005.00037.
199. Hyman, D.M., A note on closed maps and metrizability, Proc. Amer. Math. Soc., 21(1969),
109–112. MR38#6534. Zbl0174.25803.
200. Ikeda, Y., Liu, C., Tanaka, Y., Quotient compact images of metric spaces, and related matters,
Topology Appl., 122(2002)(1-2), 237–252. MR2003f:54063. Zbl0994.54015.
201. Ishii, T., On closed mappings and M-spaces I, II, Proc. Japan Acad., 43(1967), 752–761.
MR36#5904. Zbl0153.52402; Zbl0155.31305.
202. Ishii, T., On wM-spaces I, II, Proc. Japan Acad., 46(1970), 5–15. MR41#6147.
Zbl0198.27202; Zbl0198.27203.
203. Ishii, T., wM-spaces and closed maps, Proc. Japan Acad., 46(1970), 16–21. MR41#7648.
Zbl0198.27204.
204. Ishikawa, F., On countably paracompact spaces, Proc. Japan Acad., 31(1955), 686–687.
MR17,650a. Zbl0066.41001.
205. Isiwata, T., The product of M-spaces need not be an M-space, Proc. Japan Acad., 45(1969),
154–156. MR39#6242.
206. Isiwata, T., Inverse images of developable spaces, Bull. Tokyo Gakugei Univ. IV Math.,
23(1971), 11–21. MR46#6291. Zbl0346.54011.
207. Itō, M., The closed image of a hereditary M1 -space is M1 , Pacific J. Math., 113(1984), 85–91.
MR85j:54039. Zbl0503.54035.
208. Itō, M., M3 -spaces whose every point has a closure preserving outer base are M1 , Topology
Appl., 19(1985), 65–69. MR86g:54039. Zbl0567.54012.
209. Yakovlev, N.N., On g-metrizable spaces (in Russian), Dokl. Akad. Nauk SSSR, 226(1976)(3),
530–532 (English translation: Soviet Math. Dokl., 17(1976)(1), 156–159). MR54#3662.
Zbl0402.54030.
210. Jayanthan, A.J., Kannan, V., Spaces every quotient of which is metrizable, Proc. Amer. Math.
Soc., 103(1988), 294–298. MR89f:54017. Zbl0646.54015.
211. Jiang, J., On paracompactness and metrizability of topological spaces (in Chinese), Acta Math.
Sin., 29(1986), 679–701. MR0876351. Zbl0612.54035.
212. Jiang, J., Selected Topics in General Topoloy (in Chinese), Sichuan Education Press, Chengdu,
1991.
213. Jiang, S., Every strict p-space is θ-refinable, Topology Proc., 11(1986), 309–316.
MR89j:54030. Zbl0637.54024.
214. Jones, F.B., Concerning normal and completely normal spaces, Bull. Amer. Math. Soc.,
43(1937), 671–677. MR1563615. Zbl0017.42902.
312 References

215. Jones, F.B., R.L. Moore’s axiom 1’ and metrization, Proc. Amer. Math. Soc., 9(1958), 487–
487. MR20#278. Zbl0091.36101.
216. Junnila, H.J.K., Neighbornets, Pacific J. Math., 76(1978), 83–108. MR58#2734.
Zbl0353.54016.
217. Junnila, H.J.K., On submetacompactness, Topology Proc., 3(1978), 375–405. MR80j:54015.
Zbl0413.54027.
218. Junnila, H.J.K., Paracompactness, metacompactness, and semi-open covers, Proc. Amer.
Math. Soc., 73(1979), 244–248. MR81e:54020. Zbl0404.54016.
219. Junnila, H.J.K., Metacompactness, paracompactness, and interior-preserving open covers,
Trans. Amer. Math. Soc., 249(1979), 373–385. MR80b:54024. Zbl0404.54017.
220. Junnila, H.J.K., Three covering properties, in: Reed, G.M. ed., Surveys in General Topology,
Academic Press, New York, 1980, 195–245. MR81e:54019. Zbl0449.54018.
221. Junnila, H.J.K., Yun, Z., ℵ-spaces and spaces with a σ -hereditarily closure-preserving k-
network, Topology Appl., 44(1992), 209–215. MR93m:54045. Zbl0772.54024.
222. Kanatani, Y., Sasaki, N., Nagata, J., New characterizations of some generalized metric spaces,
Math. Japonica, 30(1985), 805–820. MR86m:54036. Zbl0586.54033.
223. Kao, K. (Gao, G.), A note on M1 -spaces, Pacific J. Math., 108(1983), 121–128. MR85b:54047.
Zbl0487.54029.
224. Katětov, M., Measures in fully normal spaces, Fund. Math., 38(1951), 73–84. MR14,27c.
Zbl0045.17101.
225. Katětov, M., Extension of locally finite covering (in Russian), Colloq. Math., 6(1958), 145–
151. MR21#2219. Zbl0085.16901.
226. Kato, A., Solution of Morita’s problems concerning countably-compact-ifications, General
Topology Appl., 7(1977)(1): 77–87. MR55#11211. Zbl0344.54029.
227. Katuta, Y., Expandability and its generalizations, Fund. Math., 87(1975), 231–250.
MR51#13986. Zbl0312.54026.
228. Kelley, J.L., General Topology, van Nostrand, New York, 1955 (Graduate Texts Math., V. 27,
Springer-Verlag, Berlin, 1975). MR16,1136c. Zbl0066.16604.
229. Kemoto, N., Yajima, Y., Submetacompactness of β-spaces, Topology Proc., 22(1997)(Spring),
265–279. MR2000b:54035. Zbl0917.54025.
230. Kodama, Y., Nagami, K., Theory of topological spaces (in Japanese), Iwanami, Tokyo, 1974.
231. Kofner, Ja., A new class of spaces and some problems from symmetrizability theory (in
Russian), Dokl. Akad. Nauk SSSR, 187(1969), 270–273 (English translation: Soviet Math.
Dokl., 10(1969), 845–848). MR40#1964. Zbl0202.53702.
232. Kofner, Ja., On two problems from the theory of symmetrizability (in Russian), Bull. Acad.
Pol. Sci., Sér. Sci. Math. Astron. Phys., 18(1970)(2), 81–87. MR41#7620. Zbl0199.57503.
233. Kofner, Ja., Closed mappings and quasi-metrics, Proc. Amer. Math. Soc., 80(1980), 333–336.
MR81m:54049. Zbl0463.54010.
234. Kombarov, A.P., On the tightness and normality of Σ-products (in Russian), Dokl. Akad.
Nauk SSSR, 239(1978), 775–778 (English translation: Soviet Math. Dokl., 19(1978), 403–
407). MR58#12889. Zbl0397.54013.
235. Kullman, D.E., Developable spaces and p-spaces, Proc. Amer. Math. Soc., 27(1971), 154–160.
MR42#6780. Zbl0209.53903.
236. Kunen, K., Weak P-points in N∗ , in: Császár, Á. ed., Topology, Colloq. Math. Soc. János
Bolyai, V. 23 (Budapest, 1978), North-Holland, Amsterdam, 1980, 741–749. MR82a:54046.
Zbl0435.54021.
237. Kunen, K., Set Theory: An Introduction to Independence Proofs, North-Holland, Amsterdam,
1980. MR82f:03001. Zbl0443.03021.
238. Kunen, K., Vaughan, J.E., Handbook of Set-theoretic Topology, Elsevier Science Publishers
B. V., Amsterdam, 1984. MR85k:54001. Zbl0546.00022.
239. Lašnev, N., Continuous decompositions and closed mappings of metric spaces (in Russian),
Dokl. Akad. Nauk SSSR, 165(1965), 756–758 (English translation: Soviet Math. Dokl.,
6(1965), 1504–1506). MR33#703. Zbl0145.19603.
References 313

240. Lašnev, N., Closed images of metric spaces (in Russian), Dokl. Akad. Nauk SSSR, 170(1966),
505–507. MR34#3547. Zbl0153.24203.
241. Lee, K.B., On certain g-first countable spaces, Pacific J. Math., 65(1976)(1), 113–118.
MR54#11286. Zbl0359.54022.
242. Li, J., Some counter examples on generalized metric spaces (in Chinese), J. Math. Study,
28(1995)(4): 105–107. MR1412277. Zbl0913.54026.
243. Li, J., Images on locally separable metric spaces and related results, Ph. D. Thesis, Shandong
University, Jinan, 2000.
244. Li, J., ℵ-spaces and σ -mappings, Italian J. Pure Applied Math., 17(2005), 55–62. MR2203461.
Zbl1098.54029.
245. Li, J., Jiang, S., Tanaka, Y., Point-countable k-networks and maps, Questions Answers in
General Topology, 17(1999)(1), 101–108. MR2000a:54029. Zbl0939.54017.
246. Li, P., Mou, L., Wang, S., Remarks on sym-wg-spaces, Questions Answers in General Topol-
ogy, 30(2012), 67–72. MR2954283. Zbl1263.54037.
247. Lin, F., Lin, S., Uniform covers at non-isolated points, Topology Proc., 32(2008)(1), 259-275.
MR2010d:54023. Zbl1158.54009.
248. Lin, F., Lin, S., Open uniform (G) at non-isolated points and maps, Questions Answers Gen
Topology, 28(2010)(2), 147-156. MR2011h:54020. Zbl1223.54018.
249. Lin, F., Lin, S., Maps on submetrizable spaces, Questions Answers in General Topology,
28(2010)(2), 203–215. MR2011h:54035. Zbl1223.54041.
250. Lin, F., Lin, S., Uniform bases at non-isolated points and maps, Houston J. Math., 37(2011)(2),
677–688. MR2012g:54026. Zbl1225.54004.
251. Lin, F., Lin, S., Some weak versions of the M1 -spaces, Topology Appl., 158(2011)(8), 1019–
1024. MR2012h:54031. Zbl1214.54026.
252. Lin, F., Lin, S., Junnila, H.J.K., Regular bases at non-isolated points and metrization theorems,
Stud. Sci. Math. Hungarica, 49(2012)(1), 91–105. MR3059790. Zbl1299.54056.
253. Lin, S., On normal separable ℵ-spaces, Questions Answers in General Topology, 5(1987),
249–254. MR0917881. Zbl0633.54003.
254. Lin, S., A note on K -semistratifiable spaces (in Chinese), J. Suzhou Univ., 4(1988), 357–363.
255. Lin, S., On a generalization of Michael’s theorem, Northeast Math. J., 4(1988), 162–168.
MR90d:54023. Zbl0667.54011.
256. Lin, S., A study of pseudo-bases, Questions Answers in General Topology, 6(1988), 81–97.
MR89j:54031. Zbl0655.54020.
257. Lin, S., On a problem of K. Tamano, Questions Answers in General Topology, 6(1988),
99–102. MR89e:54025. Zbl0648.54026.
258. Lin, S., Mapping theorems on ℵ-spaces, Topology Appl., 30(1988), 159–164. MR89k:54063.
Zbl0663.54017.
259. Lin, S., Spaces with a locally countable k-network, Northeast Math. J., 6(1990), 39–44.
MR91e:54078. Zbl0704.54017.
260. Lin, S., The countable products of Lašnev spaces (in Chinese), Adv. Math. (China), 20(1991),
192–194. MR92d:54041. Zbl0736.54025.
261. Lin, S., A note on M-spaces (in Chinese), J. Suzhou Univ. (Natural Science), 7(1991), 267–
269.
262. Lin, S., On R-quotient, ss-mappings (in Chinese), Acta Math. Sin., 34(1991), 7–11.
MR92i:54019. Zbl0760.54009.
263. Lin, S., On the quotient compact images of metric spaces (in Chinese), Adv. Math. (China),
21(1992), 93–96. MR93c:54024. Zbl0786.54011.
264. Lin, S., A note on metrization theorem, J. Math. Res. Exposition, 12(1992), 153–155.
MR1161139. Zbl0769.54030.
265. Lin, S., On g-metrizable spaces (in Chinese), Chinese Ann. Math. Ser. A, 13(1992), 403–409.
MR94b:54087. Zbl0770.54030.
266. Lin, S., σ -mappings and Alexandroff’s problem, in: Proc. 1st Academic Annual Meeting of
Youth of Fujian Association for Science and Technology, Fujian Science and Technology
Press, Fuzhou, 1992, 5–8. MR1252903. Zbl0847.54031.
314 References

267. Lin, S., Spaces having σ -hereditarily closure-preserving k-networks, Math. Japonica,
37(1992), 17–21. MR92m:54050. Zbl0744.54011.
268. Lin, S., On Arhangel’skiı̌’s “Mappings and spaces” (in Chinese), J. Suzhou Univ. (Natural
Science), 8(1992), 393–400; 9(1993), 11–19.
269. Lin, S., The sequence-covering s-images of metric spaces, Northeast Math. J., 9(1993), 81–85.
MR94f:54066. Zbl0841.54028.
270. Lin, S., On (mod K )-metrizable spaces (in Chinese), J. Math. (PRC), 13(1993), 456–460.
MR1286670. Zbl0849.54020.
271. Lin, S., Generalized metric space and mappings (in Chinese), Science Press, Beijing, 1995.
MR96k:54002. Zbl0940.54002.
272. Lin, S., On the quotient images of normal metric spaces, Math. Japonica, 43(1996), 483–485.
MR97d:54052. Zbl0864.54010.
273. Lin, S., On sequence-covering s-mappings (in Chinese), Adv. Math. (China), 25(1996), 548–
551. MR1453163. Zbl0864.54026.
274. Lin, S., A note on the Arens’ space and sequential fan, Topology Appl., 81(1997), 185–196.
MR98m:54008. Zbl0885.54019.
275. Lin, S., Mapping theorems on k-semistratifiable spaces, Tsukuba J. Math., 21(1997), 809–815.
MR99b:54023. Zbl1025.54501.
276. Lin, S., Point-countable Covers and Sequence-covering Mappings (in Chinese), Science Press,
Beijing, 2002. MR2003k:54001. Zbl1004.54001.
277. Lin, S., A note on sequence-covering mappings, Acta Math. Hungar., 107(2005), 193–197.
MR2005m:54053. Zbl1081.54025.
278. Lin, S., The images of connected metric spaces (in Chinese), Chinese Ann. Math. Ser. A,
26(2005), 345–350. MR2006b:54015. Zbl1081.54016.
279. Lin, S., Covering properties of k-semistratifiable spaces, Topology Proc., 29(2005), 199–206.
MR2006m:54041. Zbl1086.54017.
280. Lin, S., A note on D-spaces, Comment. Math. Univ. Carolinae, 47(2006)(2), 313–316.
MR2007b:54049. Zbl1150.54340.
281. Lin, S., On open compact maps and an Arhangel’skiı̌’s problem (in Chinese), Chinese Ann.
Math. Ser. A, 27(2006)(5): 719-722. MR2266076.
282. Lin, S., Generalized Metric Space and Mappings, Second Edition (in Chinese), Science Press,
Beijing, 2007. MR2323470. Zbl1159.54002.
283. Lin, S., Some problems on generalized metrizable spaces, in: Pearl, E. ed., Open Problems in
Topology II, Elsevier Science Publishers B. V., Amsterdam, 2007, 731–736. MR2008j:54001.
Zbl1158.54300.
284. Lin, S., Boundary-s-mappings, boundary-compact-mappings and k-semistratifiable spaces
(in Chinese), Chinese Ann. Math. Ser. A, 32(2011)(2), 229–236. MR2012f:54026.
Zbl1240.54058.
285. Lin, S., Point-countable Covers and Sequence-covering Mappings, Second Edition (in Chi-
nese), Science Press, Beijing, 2015.
286. Lin, S., Cai Z., Liu, C., The closed mappings on k-semistratifiable spaces, Houston J. Math.,
35(2009)(1), 139-147. MR2010h:54035. Zbl1173.54005.
287. Lin, S., Chen, H. (Chen, Huanran), The perfect preimages of g-metrizable spaces (in Chinese),
Adv. Math. (China), 24(1995), 338–341. MR1358894. Zbl0867.54031.
288. Lin, S., Li, Z., Li, J., Liu, C., On ss-mappings, Northeast Math. J., 9(1993), 521–524.
MR94m:54037. Zbl0817.54024.
289. Lin, S., Liu, Z., A new characterization of developable spaces, Ann. Math. Res., 26(1993)(2),
55–57.
290. Lin, S., Shen, R., A note on Σ  -spaces, Topology Proc., 32(2008)(1), 253-257.
MR2009m:54003. Zbl1157.54011.
291. Lin, S., Tanaka, Y., Point-countable k-networks, closed maps, and related results, Topology
Appl., 59(1994), 79–86. MR95e:54019. Zbl0817.54025.
292. Lin, S., Yan, L., A note on spaces with a σ -compact-finite weak base, Tsukuba J. Math.,
28(2004), 85–91. MR2082222. Zbl1062.54027.
References 315

293. Lin, S., Yan, P., On sequence-covering compact mappings (in Chinese), Acta Math. Sin.,
44(2001), 175–182. MR2001m:54030. Zbl1005.54031.
294. Lin, S., Yan, P., Notes on c f p-covers, Comment. Math. Univ. Carolinae, 44(2003), 295–306.
MR2004i:54039. Zbl1100.54021.
295. Lin, S., Zhang, J., Weak bases and the compact-covering images of metric spaces (in Chinese),
Acta Math. Sci. Ser. A, 33(2013)(3), 483–493. MR3113716. Zbl1299.54030.
296. Lin, S., Zheng, C., The almost-open images of a class of connected paracompact spaces,
Chinese Ann. Math. Ser. A, 30(2009)(1), 107–114. MR2512215. Zbl1199.54129.
297. Lin, S., Zhou, Y., Yan, P., On sequence-covering π -mappings (in Chinese), Acta Math. Sin.,
45(2002), 1157–1164. MR2004a:54036. Zbl1024.54020.
298. Lin, S., Zhu, Z., Workshop lecture on the images of some metric spaces (in Chinese), Adv.
Math. (China), 42(2013)(2), 129-137. MR3112898. Zbl1299.54001.
299. Liu, C,. Spaces with a σ -compact-finite k-network, Questions Answers in General Topology,
l0(1992), 81–87. MR92m:54057. Zbl0748.54007.
300. Liu, C., A note on compact covering mappings (in Chinese), J. Guangxi Univ., 18(1993)(4),
49–51.
301. Liu, C., Spaces with a σ -hereditarily closure-preserving k-network, Topology Proc., l8(1993),
179–188. MR96c:54049. Zbl0821.54018.
302. Liu, C., k-mapping images of metric spaces (in Chinese), J. Math. (PRC), 14(1994)(2), 233–
236. MR1313434. Zbl0900.54007.
303. Liu, C., On point-countable covers (in Chinese), J. Math. Res. Exposition, 16(1996)(1), 121–
124. MR1391382. Zbl0866.54026.
304. Liu, C., On weakly bisequential spaces, Comment. Math. Univ. Carolinae, 41(2000)(3), 611–
617. MR2002c:54026. Zbl1038.54004.
305. Liu, C., On weak bases, Topology Appl., 150(2005), 91–99. MR2005k:54048. Zbl1081.54026.
306. Liu, C., Notes on closed maps, Houston J. Math., 33(2007)(1), 249–259. MR2287858.
Zbl1133.54018.
307. Liu, C., Dai, M., g-metrizability and Sω , Topology Appl., 60(1994), 185–189. MR95j:54018.
Zbl0806.54025.
308. Liu, C., Dai, M., The compact-covering s-images of metrizable spaces (in Chinese), Acta
Math. Sin., 39(1996), 41–44. MR1412902. Zbl0865.54014.
309. Liu, C., Lin, S., On countable-to-one maps, Topology Appl., 154(2007), 449–454.
MR2007i:54031. Zbl1129.54010.
310. Liu, C., Ludwig, L.D., Nagata-Smirnov revisited: Spaces with σ -wHCP bases, Topology
Proc., 29(2005)(2), 559–565. MR2007e:54037. Zbl1123.54009.
311. Liu, C., Tanaka, Y., Spaces with certain compact-countable k-networks, and questions, Ques-
tions Answers in General Topology, 14(1996), 15–37. MR1384050. Zbl0847.54032.
312. Liu, C., Tanaka, Y., Spaces having σ -compact-finite k-networks, and related matters, Topology
Proc., 21(1996), 173–200. MR99c:54036. Zbl0896.54018.
313. Liu, C., Tanaka, Y., Spaces and mappings: special networks, in: Pearl, E. ed., Open Problems
in Topology II, Elsevier Science Publishers B. V., Amsterdam, 2007, 23–34. MR2008j:54001.
Zbl1158.54300.
314. Liu, Y., On a class of topological spaces containing all metacompact and subparacompact
spaces (in Chinese), Acta Math. Sin., 20(1977), 212–214. MR80a:54035. Zbl0373.54020.
315. Liu, Y., Liu, L., The metrization of adjuction spaces (in Chinese), Acta Math. Sin., 22(1979),
241–243. MR80i:54008. Zbl0423.54003.
316. Lutzer, D.J., Semimetrizable and stratifiable spaces, General Topology Appl., 1(1971), 43–48.
MR45#5952. Zbl0211.25704.
317. Mack, J., Directed covers and paracompact spaces, Canad. J. Math., 19(1967), 649–654.
MR35#2263. Zbl0147.22805.
318. Mancuso, V.J., Inverse images and first countability, General Topology Appl., 2(1972), 29–44.
MR45#9298. Zbl0234.54003.
319. Martin, H.W., Metrizability of M-spaces, Canad. J. Math., 25(1973), 840–841. MR48#7217.
Zbl0247.54031.
316 References

320. Martin, H.W., Contractibility of topological spaces onto metric spaces, Pacific J. Math.,
61(1975), 209–217. MR53#14432. Zbl0304.54026.
321. Martin, H.W., Weak bases and metrization, Trans. Amer. Math. Soc., 222(1976), 337–344.
MR54#11290. Zbl0341.54039.
322. McAuley, L.F., On semi-metric spaces, in: Summer Institute on Set Theoretic Topology (Wis-
consin, 1955), 1958, 58–62.
323. McAuley, L.F., A relation between perfect separability, completeness, and normality in semi-
metric spaces, Pacific J. Math., 6(1956), 315–326. MR18,325c. Zbl0072.17802.
324. McAuley, L.F., A note on complete collectionwise normality and paracompactness, Proc.
Amer. Math. Soc., 9(1958), 796–799. MR20#6086. Zbl0109.15301.
325. Michael, E.A., A note on paracompact spaces, Proc. Amer. Math. Soc., 4(1953), 831–838.
MR15,144b. Zbl0052.18701.
326. Michael, E.A., Another note on paracompact spaces, Proc. Amer. Math. Soc., 8(1957), 822–
828. MR19,299c. Zbl0078.14805.
327. Michael, E.A., Yet another note on paracompact spaces, Proc. Amer. Math. Soc., 10(1959),
309–314. MR21#4406. Zbl0092.15403.
328. Michael, E.A., A theorem on semi-continuous set-valued functions, Duke Math. J., 26(1959),
647–651. MR22#229. Zbl0151.30805.
329. Michael, E.A., The product of a normal space and a metric space need not be normal, Bull.
Amer. Math. Soc., 69(1963), 375–376. MR27#2956. Zbl0114.38904.
330. Michael, E.A., A note on closed maps and compact sets, Israel J. Math., 2(1964), 173–176.
MR31#1659. Zbl0136.19303.
331. Michael, E.A., ℵ0 -spaces, J. Math. Mech., 15(1966), 983–1002. MR34#6723. Zbl0148.16701.
332. Michael, E.A., Bi-quotient maps and Cartesian products of quotient maps, Ann. Inst. Fourier
Grenoble, 18(1968), 287–302. MR39#6277. Zbl0175.19704.
333. Michael, E.A., On Nagami’s Σ-spaces and some related matters, in: Proc. Washington State
Univ. Topological Conf., 1970, 13–19. MR42#1067. Zbl0195.24503.
334. Michael, E.A., Paracompactness and the Lindelöf property in finite and countable Cartesian
products, Compositio Math., 23(1971), 199–214. MR44#4706. Zbl0216.44304.
335. Michael, E.A., On representing spaces as images of metrizable and related spaces, General
Topology Appl., 1(1971), 329–343. MR45#2681. Zbl0227.54009.
336. Michael, E.A., A quintuple quotient quest, General Topology Appl., 2(1972), 91–138.
MR46#8156. Zbl0238.54009.
337. Michael, E.A., ℵ0 -spaces and a function space theorem of R. Pol, Indiana Univ. Math. J.,
26(1977), 299–306. MR55#6352. Zbl0327.54013.
338. Michael, E.A., σ -locally finite maps, Proc. Amer. Math. Soc., 65(1977), 159–164.
MR56#1253. Zbl0356.54034.
339. Michael, E.A., A problem, in: Baayen, P.C., van Mill, J. eds., Topological Structures II, Proc.
of Symp. on Topology and Geometry (Amsterdam, 1978), Mathematical Centre Tracts, V. 115,
Amsterdam, 1979, 165–166. MR81a:54001a. Zbl0413.00020.
340. Michael, E.A., Nagami, K., Compact-covering images of metric spaces, Proc. Amer. Math.
Soc., 37(1973), 260–266. MR46#6269. Zbl0228.54008.
341. Micheal, E.A., Olson, R.C., Siwiec, F., A-spaces and countably bi-quotient maps, Disserta-
tiones Math., 33(1976), 1–43. MR54#6067. Zbl0338.54006.
342. van Mill, J., Reed, G.M., Open Problems in Topology, Elsevier Science Publishers B. V.,
Amsterdam, 1990. MR92c:54001. Zbl0718.54001.
343. van Mill, J., Tkachuk V.V., Wilson, R.G., Classes defined by stars and neighbourhood assign-
ments, Topology Appl., 154(2007), 2127–2134. MR2008g:54032. Zbl1131.54022.
344. Miščenko, A.S., Spaces with a pointwise denumerable basis (in Russian), Dokl. Akad. Nauk
SSSR, 144(1962), 985–988 (English translation: Soviet Math. Dokl., 3(1962), 855–858).
MR25#1537. Zbl0122.17304.
345. Mizokami, T., On a certain class of M1 -spaces, Proc. Amer. Math. Soc., 87(1983), 357–362.
MR0681849.
346. Mizokami, T., On M-structures, Topology Appl., 17(1984), 63–89. MR0730024.
References 317

347. Mizokami, T., On C F families and hyperspaces of compact subsets, Topology Appl., 35(1990),
75–92. MR91a:54013. Zbl0715.54018.
348. Mizokami, T., On the closed images of a developable space, Houston J. Math., 19(1993)(3),
455–467. MR94i:54059. Zbl0784.54030.
349. Mizokami, T., On closed subsets of M1 -spaces, Topology Appl., 141(2004), 197–206.
MR2005g:54052. Zbl1058.54013.
350. Mizokami, T., Lin, S., On spaces with a σ -C F ∗ pseudo-base, Math. Japonica, 46(1997),
377–384. MR98h:54009. Zbl0992.46008.
351. Mizokami, T., Shimane, N., On the M3 versus M1 problem, Topology Appl., 105(2000), 1–13.
MR2001f:54032. Zbl0954.54011.
352. Mizokami, T., Shimane, N., The M3 versus M1 Problem in Generalized Metric Spaces, Yoko-
hama Publishers, Yokohama, 2008. MR2009k:54001. Zbl1146.54001.
353. Mizokami, T., Shimane, N., Kitamura, Y., A characterization of a certain subclass of M1 -
spaces, JP J. Geom. Topology, 1(2001), 37–51. MR1876154. Zbl0996.54041.
354. Moody, P.J., Concerning the Collins, Reed, Roscoe, Rudin metrisation theorem, Bull. London
Math. Soc., 25(1993), 476–480. MR94h:54034. Zbl0789.54039.
355. Moody, P.J., Reed, G.M., Roscoe, A.W., Collins, P.J., A lattice of conditions on topological
spaces II, Fund. Math., 138(1991), 69–81. MR92e:54024. Zbl0745.54008.
356. Moore, J.T., A solution to the L space problem, J. Amer. Math. Soc., 19(2006)(3): 717–736.
MR2008c:54022. Zbl1107.03056.
357. Moore, R.L., On the foundations of plane analysis situs, Trans. Amer. Math. Soc., 17(1916)(2),
13l–164. MR1501033. JFM46.0828.02.
358. Morita, K., On the simple extension of a space with respect to a uniformity, IV, Proc. Japan
Acad., 27(1951), 632–636. MR14,571e. Zbl0045.11702.
359. Morita, K., On spaces having the weak topology with respect to closed coverings, Proc. Japan
Acad., 29(1953), 537–543. MR15,977b. Zbl0053.12405.
360. Morita, K., Products of normal spaces with metric spaces, Math. Ann., 154(1964), 365–382.
MR29#2773. Zbl0117.39803.
361. Morita, K., Some properties of M-spaces, Proc. Japan Acad., 43(1967), 869–872.
MR37#3517. Zbl0153.52403.
362. Morita, K., Some problems on normality of products of spaces, in: Proc. 4th Prague Topology
Symp. (Prague, 1976), General Topology and its Relations to Modern Analysis and Algebra
IV, Part B, 1977, 296–297. MR58#2715. Zbl0374.54004.
363. Morita, K., Hanai, S., Closed mappings and matric spaces, Proc. Japan Acad., 32(1956),
10–14. MR19,299a. Zbl0073.17803.
364. Morita, K., Nagata, J., Topics in General Topology, Elsevier Science Publishers B. V., Ams-
terdam, 1989. MR91a:54001. Zbl0684.00017.
365. Morita, K., Rishel, T., Results related to closed images of M-spaces I, II, Proc. Japan Acad.,
47(1971), 1004–1011. MR46#9936. Zbl0254.54038. Zbl0254.54039.
366. Mrówka, S.G., On completely regular spaces, Fund. Math., 41(1954), 105–106. MR16,157b.
Zbl0055.41304.
367. Mrówka, S.G., On normal metrics, Amer. Math. Monthly, 72(1965), 998–1001. MR32#1677.
Zbl0132.18302.
368. Nagami, K., Σ-spaces, Fund. Math., 65(1969), 169–192. MR41#2612. Zbl0181.50701.
369. Nagami, K., Ranges which enable open maps to be compact-covering, General Topology
Appl., 3(1973), 355–367. MR49#9794. Zbl0278.54011.
370. Nagata, J., On a necessary and sufficient condition of metrizability, J. Inst. Polyt. Osaka City
Univ., 1(1950), 93–100. MR13,264e. Zbl0041.09801.
371. Nagata, J., Mappings and M-spaces, Proc. Japan Acad., 45(1969), 140–144. MR39#4813.
Zbl0188.28101.
372. Nagata, J., Characterizations of some generalized metric spaces, Notices Amer. Math. Soc.,
18(1971), 71T–G151.
373. Nagata, J., A survey of the theory of generalized metric spaces, in: Proc. 3th Prague Topo-
logical Symp. (Prague, 1971), General Topology and its Relations to Modern Analysis and
Algebra III, 1972, 321–331. MR50#8446. Zbl0307.54025.
318 References

374. Nagata, J., On closed mappings of generalized metric spaces, Proc. Japan Acad., 47(1971),
181–184. MR47#1033. Zbl0225.54010.
375. Nagata, J., Modern Dimension Theory, Heldermann Verlag, Berlin, 1983. MR84h:54033.
Zbl0518.54002.
376. Nagata, J., Modern General Topology, 2nd rev. ed., North-Holland Math Library, V. 33,
Elsevier Science Publishers B. V., Amsterdam, 1985. MR87g:54003. Zbl0598.54001.
377. Nagata, J., Metrizability, generalized metric spaces and g-functions, Comment. Math. Univ.
Carolinae, 29(1988)(4): 715–722. MR90d:54060. Zbl0686.54019.
378. Nagata, J., Generalized metric spaces I, in: Morita, K., Nagata, J. eds., Topics in General
Topology, Elsevier Science Publishers B. V., Amsterdam, 1989, 315–366. MR91e:54071.
Zbl0684.00017.
379. Nagata, J., Remarks on metrizability and generalized metric spaces, Topology Appl., 91(1999),
71–77. MR2000c:54022. Zbl0926.54019.
380. Nedev, S.J., Symmetrizable spaces and final compactness (in Russian), Dokl. Akad. Nauk
SSSR, 175(1967): 532–534 (English translation: Soviet Math. Dokl., 8(1967), 890–892).
MR35#7293. Zbl0153.52701.
381. Nedev, S.J., o-metrizable spaces (in Russian), Trudy Moskov Mat. Obsch, 24(1971), 201-
236 (English translation: Trans. Moscow Math. Soc., 24(1974), 213–247). MR51#4177.
Zbl0295.54039.
382. Niemytzki, V.W., On the “third axiom of metric space”, Trans. Amer. Math. Soc., 29(1927),
507–513. MR1501402.
383. Nyikos, P.J., Problem section, classic problem IV, Topology Proc., 1(1976), 360–365.
384. Nyikos, P.J., Classic problems, in: Pearl, E. ed., Problems from Topology Proceedings, Topol-
ogy Atlas, Toronta, 2003, 69–89.
385. Okuyama, A., Some generalizations of metric spaces, their metrization theorems and product
spaces, Sci. Rep. Tokyo Kyioku Daigaku. A, 9(1968), 236–254. MR37#5846. Zbl0153.52404.
386. Okuyama, A., On a generalization of Σ-spaces, Pacific J. Math., 42(1972), 485–495.
MR47#2547. Zbl0219.54018.
387. O’Meara, P., A new class of topological spaces, University of Alberta Dissertation, 1966.
388. O’Meara, P., On paracompactness in function spaces with the compact open topology, Proc.
Amer. Math. Soc., 29(1971), 183–189. MR43#2659. Zbl0214.21105.
389. Orihuela, J., Smith, R.J., Troyanski, S., Strictly convex norms and topology, Proc. London
Math. Soc., 104(2012), 197–222. MR2876968. Zbl1241.46005.
390. Patsei, I.P., The σ -product of strong Σ  -spaces (in Russian), Vestnik Moskov Univ. Mat.,
39(1986)(2), 87–89. MR87k:54054. Zbl0612.54022.
391. Pearl, E., Problems from Topology Proceedings, Topology Atlas, Toronta, 2003.
392. Pearl, E., Open Problems in Topology II, Elsevier Science Publishers B. V., Amsterdam, 2007.
Zbl1158.54300.
393. Peng, L., The decomposition theorem for Σ ∗ -spaces with G δ -points (in Chinese), Adv. Math.
(China), 33(2004), 110–114. MR2058449.
394. Peng, L., On Σ ∗ -spaces and strong Σ ∗ -spaces of countable pseudocharacter, Topology Appl.,
148(2005), 233–238. MR2005k:54040. Zbl1062.54026.
395. Peng, L., The D-property of some Lindelöf spaces and related conclusions, Topology Appl.,
154(2007), 469–475. MR2007k:54031. Zbl1110.54014.
396. Peng, L., A special point-countable family that makes a space to be a D-space, Adv. Math.
(China), 37(2008)(6), 724–728. MR2010i:54043.
397. Peng, L., On finite unions of certain D-spaces, Topology Appl., 155(2008)(6), 522–526.
MR2009g:54051. Zbl1143.54013.
398. Peng, L., On weakly monotonically monolithic spaces, Comment. Math. Univ. Carolinae,
51(2010)(1), 133–142. MR2011d:54061. Zbl1224.54078.
399. Peng, L., Lin, S., A note for the Σ ∗ -spaces (in Chinese), Adv. Math. (China), 29(2000),
354–356. MR1853859. Zbl0997.54040.
400. Peng, L., Wang, L., On CSS-spaces and related conclusions (in Chinese), Acta Math. Sci. Ser.
A, 30(2010)(2), 358–363. MR2664833. Zbl1224.54065.
References 319

401. Ponomarev, V.I., Axioms of countability and continuous mappings (in Russian), Bull. Acad.
Pol. Sci., Sér. Sci. Math. Astron. Phys., 8(1960), 127–134. MR22#7109. Zbl0095.16301.
402. Popov, V., A perfect map need not preserved a G δ -diagonal, General Topology Appl., 7(1977),
31–33. MR55#4095. Zbl0343.54035.
403. Potoczny, H.B., A non-paracompact space which admits a closure-preserving cover of com-
pact sets, Proc. Amer. Math. Soc., 32(1972), 309–311. MR44#5923. Zbl0208.50802.
404. Przymusiński, T.C., Products of normal spaces, in: Kunen, K., Vaughan, J.E. eds., Handbook
of Set-theoretic Topology, Elsevier Science Publishers B. V., Amsterdam, 1984, 781–826.
MR86c:54007. Zbl0559.54009.
405. Qu, Z., Gao, Z., Spaces with compact-countable k-networks, Math. Japonica, 49(1999)(2),
199–205. MR2001g:54033. Zbl0927.54031.
406. Reed, G.M., On subspaces of separable first countable T2 -spaces, Fund. Math., 91(1976),
189–202. MR54#13863. Zbl0341.54016.
407. Reed, G.M., Zenor, P.L., Metrization of Moore spaces and generalized manifolds, Fund.
Math., 91(1976), 203–210. MR54#13868. Zbl0339.54030.
408. Reznichenko, E.A., Stratifiability of Ck (X ) for a class of separable metrizable X , Topology
Appl., 155(2008)(17-18), 2060–2062. MR2009j:54022. Zbl1180.54039.
409. Rudin, M.E., A normal space X for which X × I is not normal, Fund. Math., 73(1971/72),
179–186. MR45#2660. Zbl0224.54019.
410. Rudin, M.E., The normality of products with one compact factor, General Topology Appl.,
5(1975), 45–59. MR50#14656. Zbl0296.54004.
411. Rudin, M.E., Lecture on Set Theoretic Topology, Conference Board of the Mathemati-
cal Sciences Regional Conference Series in Math., V. 23, Providence, 1975. MR51#4128.
Zbl0318.54001.
412. Rudin, M.E., κ-Dowker spaces, Czech. Math. J., 28(1978), 324–326. MR57#17588.
Zbl0383.54012.
413. Rudin, M.E., The shrinking property, Canad. Math. Bull., 26(1983), 385–388. MR85h:54040.
Zbl0536.54013.
414. Rudin, M.E., Starbird, M., Products with a metric factor, General Topology Appl., 5(1975),
235–248. MR52#1606. Zbl0305.54010.
415. Rudin, W., Homogeneity problems in the theory of Čech compactifications, Duke Math. J.,
23(1956), 409–419. MR0080902.
416. Sakai, M., Remarks on spaces with special type of k-networks, Tsukuba J. Math., 21(1997),
443–448. MR98g:54045. Zbl0886.54023.
417. Sakai, M., Mapping theorems on ℵ-spaces, Comment. Math. Univ. Carolinae, 49(2008)(1):
163–167. MR2009j:54022. Zbl1212.54049.
418. Sakai, M., Mizokami and Lin’s conjecture on σ -C F ∗ pseudo-bases, Topology Appl.,
157(2010), 152–156. MR2010k:54018. Zbl1186.54020.
419. Sakai, M., On k-networks and weak bases for spaces, Topology Appl., 157(2010), 2383–2388.
MR2011f:54019. Zbl1203.54010.
420. Sconyers, W.B., Metacompact spaces and well-ordered open coverings, Notices Amer. Math.
Soc., 17(1970), 230–230.
421. Shibakov, A., On spaces with point-countable k-networks and their mappings, Serdica Math.
J., 20(1994)(1), 48–55. MR95h:54020. Zbl0827.54015.
422. Shiraki, T., M-spaces, their generalization and metrization theorems, Sci. Rep. Tokyo Kyoiku
Daigaku A, 11(1971), 57–67. MR46#4495. Zbl0233.54016.
423. Singal, M.K., Arya, S.P., Weak topology sum theorems, in: Császár, Á. ed., Topology, Colloq.
Math. Soc. János Bolyai (Budapest, 1978), V. 23, North-Holland, Amsterdam, 1980, 1095–
1109. MR82e:54033. Zbl0449.54007.
424. Siwiec, F., Sequence-covering and countably bi-quotient mappings, General Topology Appl.,
1(1971), 143–154. MR44#5933. Zbl0218.54016.
425. Siwiec, F., On defining a space by a weak base, Pacific J. Math., 52(1974), 233–245.
MR50#3198. Zbl0285.54022.
320 References

426. Siwiec, F., Nagata, J., A note on nets and metrization, Proc. Japan Acad., 44(1968), 623–727.
MR39#3450. Zbl0181.25902.
427. Slaughter, jr.F.G., The closed image of a metrizable space is M1 , Proc. Amer. Math. Soc.,
37(1973), 309–314. MR46#9930. Zbl0255.54024.
428. Smirnov, Yu.M., On metrization of topological spaces (in Russian), Uspechi Mat. Nauk,
6(1951)(6), 100–111 (English translation: Amer. Math. Soc. Transl. Ser. 1, 8(1962), 63–77).
MR14,70a. Zbl0045.11704.
429. Smirnov, Yu.M., On the metrizability of bicompacts decomposable into a sum of sets with
countable bases (in Russian), Fund. Math., 43(1956), 387–393. MR18,813d. Zbl0071.38401.
430. Smith, J.C., Properties of weak θ-refinable spaces, Proc. Amer. Math. Soc., 53(1975), 511–
517. MR52#1628. Zbl0338.54013.
431. Smith, J.C., Krajewski, L.L., Expandability and collectionwise normality, Trans. Amer. Math.
Soc., 160(1971), 437–451. MR44#2190. Zbl0224.54020.
432. Šneı̌der, V.E., Continuous images of Suslin and Borel sets, metrization theorems (in Russian),
Dokl. Akad. Nauk SSSR, 50(1945), 77–79. MR14,782d. Zbl0061.39705.
433. Sorgenfrey, R.H., On the topological product of paracompact spaces, Bull. Amer. Math. Soc.,
53(1947), 631–632. MR8,594f. Zbl0031.28302.
434. Soukup, D., Xu, Y., The Collins-Roscoe mechanism and D-spaces, Acta Math. Hungar.,
131(2011)(3), 275–284. MR2012b:54016. Zbl1249.54055.
435. Soukup, D., Szeptycki, P., A counterexample in the theory of D-spaces, Topology Appl.,
159(2012)(10-11), 2669–2678. MR2923437. Zbl1246.54018.
436. Soukup, D., Szeptycki, P., The union of two D-spaces need not be D, Fund. Math.,
220(2013)(2), 129–137. MR3024041. Zbl1264.54042.
437. Steen, L.A., Seebach, jr.J.A., Counterexamples in Topology, Second Edition, Springer-Verlag,
New York, 1978 (Dover Publications Inc., New York, 1995). MR80a:54001. Zbl0386.54001.
438. Stephenson, R.M.jun, Near compactness and separability of symmetrizable spaces, Proc.
Amer. Math. Soc., 68(1978), 108–110. MR56#16575. Zbl0381.54019.
439. Stoltenberg, R.A., A note on stratifiable spaces, Proc. Amer. Math. Soc., 23(1969), 294–297.
MR39#6245. Zbl0183.27204.
440. Stone, A.H., Paracompactness and product spaces, Bull. Amer. Math. Soc., 54(1948), 977–982.
MR10,204c. Zbl0032.31403.
441. Stone, A.H., Metrizability of decomposition spaces, Proc. Amer. Math. Soc., 7(1956), 690–
700. MR19,299b. Zbl0071.16001.
442. Stone, A.H., Metrisability of union of spaces, Proc. Amer. Math. Soc., 10(1959), 361–366.
MR21#4410. Zbl0090.38705.
443. Svetlichny, S.A., Open mappings of submetrizable spaces (in Russian), Vestnik Moskov Univ.
Mat., 43(1988)(6), 18–20. MR90h:54021. Zbl0689.54005.
444. Sun, S., The class of ℵ-spaces is invariant of closed mappings with Lindelöf fibers, Comment.
Math. Univ. Carolinae, 29(1988), 351–354. MR90d:54026. Zbl0656.54021.
445. Suzuki, J., On pre-σ -spaces, Bull. Tokyo Gakugei Univ. IV Ser., 28(1976), 22–32. MR55#6349.
Zbl0345.54011.
446. Tamano, H., On compactifications, J. Math. Kyoto Univ., l(1962), 161–193. MR25#5489.
Zbl0106.15601.
447. Tamano, H., A characterization of paracompactness, Fund. Math., 72(1971), 189–201.
MR45#5956. Zbl0224.54022.
448. Tamano, K., Closed images of metric spaces and metrization, Topology Proc., 10(1985),
177–186. MR87j:54011. Zbl0616.54026.
449. Tamano, K., Generalized metric spaces II, in: Morita, K., Nagata, J. eds., Topics in General
Topology, Elsevier Science Publishers B. V., Amsterdam, 1989, 367–409. MR91e:54074.
Zbl0698.54024.
450. Tamano, K., Definitions of Σ-spaces, Topology Proc., 22(1997)(Summer), 529–532.
MR2000g:54051. Zbl0945.54021.
451. Tamano, K., A base, a quasi-base, and a monotone normality operator for Ck (P), Topology
Proc., 32(2008), 277–290. MR2009j:54042. Zbl1161.54008.
References 321

452. Tanaka, Y., On open finite-to-one maps, Bull. Tokyo Gakugei Univ. IV Ser., 25(1973), 1–13.
MR49#11455. Zbl0355.54008.
453. Tanaka, Y., Closed maps on metric spaces, Topology Appl., 11(1980), 87–92. MR82k:54017.
Zbl0436.54010.
454. Tanaka, Y., Metrizability of certain quotient spaces, Fund. Math., 119(1983), 157–168.
MR86c:54028. Zbl0542.54022.
455. Tanaka, Y., Point-countable covers and k-networks, Topology Proc., 12(1987), 327–349.
MR90e:54060. Zbl0676.54035.
456. Tanaka, Y., Symmetric spaces, g-developable spaces and g-metrizable spaces, Math. Japonica,
36(1991), 71–84. MR92d:54018. Zbl0732.54023.
457. Tanaka, Y., σ -hereditarily closure-preserving k-networks and g-metrizability, Proc. Amer.
Math. Soc., 112(1991), 283–290. MR91h:54046. Zbl0770.54031.
458. Tanaka, Y., Closed maps and symmetric spaces, Questions Answers in General Topology,
11(1993)(2), 215–233. MR94h:54014. Zbl0788.54033.
459. Tanaka, Y., Li, Z., Certain covering-maps and k-networks, and ralated matters, Topology Proc.,
27(2003), 317–334. MR2005a:54021. Zbl1075.54010.
460. Tanaka, Y., Liu, C., Fiber properties of closed maps, and weak topology, Topology Proc.,
24(1999)(Spring), 323–344. MR1802696. Zbl0965.54019.
461. Tanaka, Y., Murota, T., Generalizations of wΔ-spaces, and developable spaces, Topology
Appl., 82(1998)(1-3), 439–452. MR99c:54011. Zbl0891.54011.
462. Tanaka, Y., Yajima, Y., Decompositions for closed maps, Topology Proc., 10(1985), 399–411.
MR88a:54032. Zbl0613.54003.
463. Teng, H., On a problem of Y. Yajima, Topology Appl., 38(1991), 39–43. MR92d:54009.
Zbl0714.54006.
464. Teng, H., Xia, S., Lin, S., Closed images of some kinds of generalized countably compact
spaces (in Chinese), Chinese Ann. Math. Ser. A, 10(1989), 554–558 (English translation:
Chinese J. Contemp. Math., 10(1989)(3), 239–245). MR91e:54058. Zbl0705.54015.
465. Tkachuk, V.V., When do connected spaces have nice connected preimages, Proc. Amer. Math.
Soc., 126(1998), 3437–3446. MR99a:54015. Zbl0903.54004.
466. Tkachuk, V.V., Monolithic spaces and D-spaces revisited, Topology Appl., 156(2009), 840–
846. MR2009m:54017. Zbl1165.54009.
467. Tkachuk, V.V., The Collins-Roscoe property and its applications in the theory of function
spaces, Topology Appl., 159(2012), 1529–1535. MR2891418. Zbl1245.54022.
468. Todorčevic̀, S., A topology on sequences of countable ordinals, Bull. Pol. Acad. Sci., Math.,
39(1991)(1-2), 137–140. MR93k:54005. Zbl0780.54035.
469. Tukey, J.M., Convergence and Uniformity in Topology, Ann. Math. Studies, V. 2, Princeton,
1940. MR2,67a. Zbl0025.09102.
470. Uspenskii, V.V., Pseudocompact spaces with a σ -point-finite base are metrizable, Comment.
Math. Univ. Carolinae, 25(1984), 261–264. MR87f:54035. Zbl0574.54021.
471. Vaı̌nšteı̌n, I.A., On closed mappings of metric spaces (in Russian), Dokl. Akad. Nauk SSSR,
57(1947), 319–321. MR9,153b.
472. Veličko, N.V., Symmetrizable spaces (in Russian), Mat. Zametki, 12(1972)(5), 577–582 (Eng-
lish translation: Math. Note, 12(1972), 784–786). MR48#5021. Zbl0254.54031.
473. Veličko, N.V., Quotient spaces of metrizable spaces (in Russian), Sibirskii Mat. Zhur-
nal, 28(1987)(4), 73–81 (English translation: Siberian Math. J., (1988)(4), 575–581).
MR88g:54016. Zbl0683.54032.
474. Wang, S., Remarks on ωμ -additive spaces, Fund. Math., 55(1964), 101–112. MR29#4022.
Zbl0135.40901.
475. Wicke, H.H., On the Hausdorff open continuous images of Hausdorff paracompact p-spaces,
Proc. Amer. Math. Soc., 22(1969), 136–140. MR39#4801. Zbl0176.51801.
476. Willard, S., Metric spaces all of whose decompositions are metric, Proc. Amer. Math. Soc.,
21(1969), 126–128. MR39#919. Zbl0174.25802.
477. Wilson, W.A., On semi-metric spaces, Amer. J. Math., 53(1931), 361–373. MR1506824.
Zbl0001.22804.
322 References

478. Worrell, jr.J.M., Upper semicontinuous decompositions of developable spaces, Proc. Amer.
Math. Soc., 16(1965), 485–490. MR31#6207. Zbl0132.18305.
479. Worrell, jr.J.M., The closed continuous images of metacompact topological spaces, Portug.
Math., 25(1966), 175–179. MR38#676. Zbl0171.21303.
480. Worrell, jr.J.M., Paracompactness as a relaxation of full normality, Notices Amer. Math. Soc.,
15(1968), 661–661.
481. Worrell, jr.J.M., Wicke, H.H., Characterizations of developable topological spaces, Canad. J.
Math., 17(1965), 820–830. MR32#427. Zbl0132.18401.
482. Wu, L., On K -semistratifiable spaces, J. Suzhou Univ. (Natural Science), (1983)(1), 1–4.
483. Yajima Y. On Σ-products of Σ-spaces, Fund. Math., 123(1984), 29–37. MR86d:54035.
Zbl0556.54008.
484. Yan, P., The compact images of metric space (in Chinese), J. Math. Study, 30(1997)(2), 185–
187, 198. MR1468151. Zbl0918.54029.
485. Yan, P., Jiang, S., On the compact-covering π -maps (in Chinese), J. Math. (PRC), 24(2004),
429–432. MR2005g:54025. Zbl1053.54534.
486. Yan, P., Lin, S., The compact-covering s-images on metrizable spaces (in Chinese), Acta
Math. Sin., 42(1999), 241–244. MR1701751. Zbl1011.54029.
487. Yan, P., Lü, C., Compact images of spaces with a weaker metric topology, Czech. Math. J.,
58(2008), 921–926. MR2009j:54049.
488. Yang, C., On paracompactness spaces, Proc. Amer. Math. Soc., 5(1954), 185–189.
MR15,976d. Zbl0055.41401.
489. Yang, E., Shi, W., Weak base g-functions and metrizability of topological spaces, Topology
Appl., 158(2011), 238–243. MR2011k:54036. Zbl1213.54039.
490. Yu, Z., Yun, Z., D-spaces, α D-spaces and finite unions. Topology Appl., 156(2009)(8), 1459–
1462. MR2010d:54034. Zbl1162.54012.
491. Yun, Z., On a problem of J. Nagata, Comment. Math. Univ. Carolinae, 30(1989)(4), 811–815.
MR91b:54059. Zbl0692.54017.
492. Yun, Z., On point-countable closed k-network, Questions Answers in General Topology,
7(1989), 139–140. MR90m:54032. Zbl0714.54007.
493. Yun, Z., A new characterization of ℵ-spaces, Topology Proc., 16(1991), 253–256.
MR94c:54046. Zbl0784.54029.
494. Yun, Z., On closed mappings, Houston J. Math., 31(2005), 193–197. MR2005k:54045.
Zbl1069.54020.
495. Yun, Z., Junnila, H.J.K., On a special metric, Houston J. Math., 26(2000)(4), 877-882.
MR2001k:54048. Zbl0973.54030.
496. Zhang, J., On compact-covering and sequence-covering images of metric spaces, Mat. Vesnik,
64(2012)(2), 97–107. MR2890958. Zbl1289.54064.
Index

Symbols V -space, 39, 146

(P ) A , 3 X ∪ f Y , 63
B-refinement, 141 I, 2
Bt (x, 1/n), 39 N, 2
C(P , x), 160 NN , 43
C F family, 87 P, 2
C F ∗ family, 235 Q, 2
C W C-function, 301 R, 2
D-space, 10 R+ , 2
D-space question, 10 S1 , 2
Fσ -discrete set, 168 Δ-metric, 295
Fσ -discrete space, 94 Δ-system, 96
Fσ -screenable space, 296 Δ-system lemma, 96
G δ -diagonal, 19 Δα∈Γ f α , 4
G δ -diagonal sequence, 19 Φ point-star network, 101
G ∗δ -diagonal, 19 Σ-network, 291
G ∗δ -diagonal sequence, 19 Σ-product, 284
H C P, 87 Σ-space, 160
H C P at a point, 92 Σ{X α : α ∈ Λ}, 284
H C P weak development, 251 Σ ∗ -space, 160
I (X ), 135 Σ  -function, 161
L-mapping, 55 Σ  -space, 160
L-space problem, 51 ℵ-space, 29
M-sequence, 36 ℵ1 , 2
M-space, 36 ℵ1 -compact space, 7
M ∗ -space, 228 ℵ0 , 2
M  -space, 228 ℵ0 -space, 28
M1 -space, 14 β-function, 35
M2 -space, 14 β-space, 35
M3 -space, 15 β N, 45
Mi -space, 15 β
R, 77
P-point, 157 , 3
P-space, 283 α∈Γ Pα , 3
Sω , 44, 156 χ(Y ), 242
S2 , 43, 156 χ(Y, y), 242
Sω1 , 45 δθ-base, 287
T1 -separating cover, 27 γ -space, 300
© Atlantis Press and the author(s) 2016 323
S. Lin and Z. Yun, Generalized Metric Spaces and Mappings,
Atlantis Studies in Mathematics 6, DOI 10.2991/978-94-6239-216-8
324 Index

c, 2 k X , 101
K (X ), 3 k ∗ -metrizable space, 101
P <ω , 3 m 1 -space, 212
PF, 3 m 2 -space, 212
S (X ), 3 m 3 -space, 212
I (X ), 135 p-sequence, 67
st n (A, P ), 3 p-space, 66
ω, 2 q-function, 35
ωμ -metrizable space, 280 q-space, 35
ω1 , 2 s-connected space, 75
π -mapping, 123 s-mapping, 55
πβ , 4

ss-mapping, 116
α∈Γ f α , 4 wM-sequence, 36
ψ(D), 41 wM-space, 36
ψ(N), 42 wN -space, 300
σ -Φ family, 6 wΔ-condition, 36
σ -discrete family, 6 wΔ-function, 35
σ -function, 180 wΔ-sequence, 35
σ -locally finite mapping, 141 wΔ-space, 35
σ -mapping, 144 wγ -space, 300
σ -paracompact space, 296 wθ-space, 300
σ -product, 284 (G), 183
σ -space, 24 (mod k)-base, 226
σ {X α : α ∈ Λ}, 284 (mod k)-metrizable space, 226
σ  -function, 27 (mod k)-network, 26
σ  -space, 27 Čech-complete space, 69
θ-base, 287
θ-refinable sequence, 269
θ-refinable space, 269 A
θ-space, 300 Additive, 4
c-semi-stratifiable space, 23 Adjunction mapping, 63
c f p-cover, 123 Adjunction space, 63
cs-σ -space, 30 Alexandroff-Arhangel’skiı̌’s question, 54,
cs-network, 30 294
cs ∗ -cover, 123 Alexandroff double-arrow space, 46
cs ∗ -network, 30 Alexandroff idea, 285, 294
ct-network, 27 Alexandroff problem, 281
fB, 4 Alexandroff’s question, 53
f cs-cover, 123 Almost disjoint family, 41
g-first countable space, 31 Almost (mod k)-network, 27
g-function, 8 Almost open mapping, 99
g-metrizable space, 31 Almost quasi-(mod k)-network, 27
g-second countable space, 31 Arens space, 43
g f -countable space, 31 Arhangel’skiı̌ theorem, 69
k-closed set, 101 Arhangel’skiı̌’s question, 53
k-coreflection, 101
k-mapping, 101
k-metrizable space, 101 B
k-network, 29 Baire category theorem, 5
k-semi-stratifiable function, 23 BCO space, 281
k-semi-stratifiable space, 19 Bernstein set, 42
k-semi-stratification, 19 Bing metrization criterion, 14
k-space, 57 Bing metrization theorem, 13
Index 325

Bing-Nagata-Smirnov metrization theorem, D

12 Decomposition theorem, 176
Bi-quotient mapping, 113 Decomposition theorem of type compact-
Bi-sequential space, 115 ness, 176
Bowtie set, 39 Decomposition theorem of type countable
Bowtie space, 39 compactness, 176
Bowtie topology, 40 Decomposition theorem of type Lindelöf-
Burke-Engelking-Lutzer metrization theo- ness, 176
rem, 92 Decreasing g-function, 8
Butterfly space, 40, 85, 246 Determined, 71
Butterfly topology, 40 Developable function, 11
Developable space, 11
Developable space at non-isolated points,
223
C Development, 10
Cauchy sequence, 126 Development at non-isolated points, 135
Ceder’s questions, 203 Diagonal theorem, 5
CH, 5 Directed cover, 259
Character of a space, 242 Discrete at a point, 6
Class P , 205 Discrete family, 6
Class MOBI, 140 Distance, 5
Closed hereditary, 4 Dominating family, 299
Closed mapping, 54 Dowker space, 283
Closed sum theorem, 299
Closure-preserving at a point, 14
Closure-preserving family, 14 E
Closure-preserving weak development, 251 Expandable space, 26, 263
Collectionwise normal space, 263 Expansion, 118
Compact mapping, 55
Compact-countable family, 159
Compact-covering mapping, 55 F
Compact-finite family, 13 Fan space, 45
Continuum hypothesis, 5 Filippov theorem, 113
Convention, 218 Finitely productive, 4
Foged theorem, 236
Convergence lemma, 37
Fortissimo space, 96, 178
Cosmic space, 25
Fréchet-Urysohn space, 8
Countable tightness, 111
Frolík space, 71
Countable-to-one mapping, 112
Countable type space, 85
Countably H C P family, 86 G
Countably bi-quotient mapping, 54 Generalized countably compact space, 34
Countably compact mapping, 55 Generalized sequentiality property, 85
Countably hereditarily closure-preserving GMS-system, v
family, 86
Countably metacompact space, 265
Countably paracompact space, 259 H
Countably productive, 4 Half-disc topology, 70, 121
Countably weakly H C P family, 86 Hanai-Morita-Stone theorem, 62
Countably weakly hereditarily closure- Heath-Junnila theorem, 209
preserving family, 86 Hereditarily closure-preserving family, 86
Covering property, 280 Hereditarily meta-Lindelöf space, 275
Cushioned family, 15 Hereditarily quotient mapping, 56
Cushioned refinement, 282 Hereditary, 4
326 Index

I Morita theorem, 64
Idea of mutual classifications of spaces and Morita’s P-space, 283
mappings, 54 Morita’s conjecture, 288
Interior-preserving family, 38 Morita–Hanai–Stone theorem, 62
Irreducible mapping, 210 Morita–Rishel theorem, 230
Isocompact space, 68 Mrówka space, 41, 100, 224

J N
Jiang theorem, 213 Nagata condition, 194
Junnila–Katuta question, 163 Nagata space, 205
Junnila theorem, 274 Nagata–Siwiec theorem, 182
Junnila–Yun theorem, 240 Nagata-Smirnov metrization theorem, 13
Natural mapping, 63
Neighborhood assignment, 10
K Neighborhood base of A in X , 13
König lemma, 78 Net, 26
Network, 24
Network hyperspace, 102
L Network of A in X , 26
Lašnev space, 86 Nontrivial sequence, 3
Least upper bound property, 47 Normal metric, 73
Left half-open interval topology, 47 Normal Moore space conjecture, 281
Local W -refinement, 260 Nw(X ), 102
Local star-refinement, 260 Nw(x), 102
Locally countable family, 115
Locally finite at a point, 12
Locally finite family, 12 O
Locally finite weak development, 251 Obvious mapping, 63
Open (G), 220
Open hereditary, 4
M Open mapping, 54
Mapping lemma, 58 Open sum theorem, 298
Metacompact space, 265 Open uniform (G), 220
Meta-Lindelöf space, 275 Open uniform (G) at non-isolated points, 222
Metric space, 6 Orthocompact space, 195
Metric topology, 6 Outer base, 78
Metrizable space, 6
Michael line, 42, 121, 139, 191
Michael space, 45 P
Michael theorem, 260 Pair-base, 14
Michael-Lutzer theorem, 226 Pair-cs ∗ -network, 31
Michael-Nagami problem, 113 Pair-cs-network, 31
Michael-Nagami theorem, 112 Pair-k-network, 31
Minimal cover, 108 Pair-(mod k)-network, 160
Minimal interior cover, 108 Pair-network, 25
Miščenko lemma, 109 Pair-quasi-(mod k)-network, 160
MOBI, 140 Pair-strict k-network, 31
MOBIi , 141 Paracompact space, 259
Monotone normality operator, 22 Partial refinement, 259
Monotonically monolithic space, 153 Perfect mapping, 55
Monotonically normal space, 22 Perfect preimage G δ -diagonal theorem, 66
Moore space, 11 Perfect preimage σ  -theorem, 66
Morita space, 283 Perfect space, 11
Index 327

Peripheral L-mapping, 55 R
Peripherally compact mapping, 55 Refinement, 259
Peripherally countably compact mapping, Regular closed set, 16
55 Regular open set, 16
Pluming, 66 Right half-open interval topology, 46
Point-countable family, 11
Point-finite development, 131
Point-finite family, 13 S
Point-finite semi-development, 131 Second category, 5
Point-finite weak development, 131 Semi-developable space, 8
Point G δ -property, 29 Semi-development, 8
Point-network, 183 Semi-metric, 8
Point-regular base, 134 Semi-metrizable function, 9
Point-regular base at non-isolated points, Semi-metrizable space, 8
134 Semi-stratifiable function, 19
Point-star network, 101 Semi-stratifiable space, 19
Point-star sequential neighborhood network, Semi-stratification , 19
101 Sequence-covering mapping, 55
Point-star weak base, 101 Sequential fan, 44
Pointed irrational extension topology, 111, Sequentially closed set, 33
134 Sequentially connected space, 75
Pointwise W -refinement, 260 Sequentially open set, 33
Pointwise W -refining sequence, 269 Sequentially quotient mapping, 55
Pointwise countable type, 84 Sequential neighborhood, 33
Polish space, 294 Sequential neighborhood network, 33
Ponomarev theorem, 80 Sequential space, 32
Precisely refine, 106 Sharp base, 224
Productive, 4 Sorgenfrey line, 46
Proper mapping, 101 Space of countable type, 85
Property P at non-isolated points, 92 Space of pointwise countable type, 84
Pseudo-base, 28, 148 Star-countable family, 116
Pseudo-distance, 5 Star-finite family, 17
Pseudo-Fq-base, 148 Star-refinement, 259
Pseudo-F-quasi-base, 148 Stone theorem, 6
Pseudo-k-network, 148 Stratifiable function, 18
Pseudo-metric space, 6 Stratifiable space, 18
Pseudo-network, 27 Stratification, 18
Pseudo-open mapping, 54 Strict k-network, 28
Pseudo-stratifiable space, 291 Strictly quasi-paracompact space, 287
Strict pluming, 66
Strict p-sequence, 67
Q Strict p-space, 66
Quasi-base, 14 Strong Σ-space, 26
Quasi-developable space, 11 Strongly Fréchet-Urysohn space, 57
Quasi-development, 11 Strong Σ-network, 292
Quasi-G δ -diagonal, 19 Strong Σ ∗ -space, 160
Quasi-G δ -diagonal sequence, 19 Strong Σ  -space, 160
Quasi-(mod k)-base, 226 Submetacompact space, 269
Quasi-(mod k)-metrizable space, 226 Submetrizable space, 65
Quasi-(mod k)-network, 160 Subparacompact space, 267
Quasi-open mapping, 210 Sum theorem, 298
Quasi-perfect mapping, 55 Symmetric, 7
Quotient mapping, 54 Symmetrizable space, 7
328 Index

T Weakly bi-sequential space, 115

Topological sum metric, 6 Weakly developable space, 218
Tychonoff compact extension theorem, 5 Weakly first countable space, 31
Tychonoff theorem, 5 Weakly H C P family, 86
Weakly hereditarily closure-preserving fam-
ily, 86
U Weakly monotonically monolithic space,
Uniform base, 134 153
Uniform base at non-isolated points, 134 Weakly paracompact space, 265
Uniform (G), 183 Weak neighborhood, 31
Urysohn extension theorem, 5 Weak P-point, 157
Urysohn lemma, 5 Weak topology, 71
Urysohn-Tychonoff metrization theorem, 5 Well-monotone cover, 213, 259
Well-ordered cover, 275
Worrell theorem, 266
W
Weak base, 31
Weak Cauchy condition, 126 Z
Weak development, 32 Zermelo theorem on well-ordering, 5