Professional Documents
Culture Documents
VOLUME3
History of Topology
Volume 3
The titZes published in this series are listed at the end 0/ this voZume.
Handbook of the History
of General Topology
Volume 3
Edited by
C. E. Aull
Department 0/ Mathematics,
Virginia Polytechnic Institute and State University,
Blacksburg, Virginia, U.SA
and
R. Lowen
Department 0/ Mathematics and Computer Science,
University 0/ Antwerp, RUCA,
Antwerp, Belgium
Introduetion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Combinatorial Topology Versus Point-set Topology ................. 809
I.M. farnes
Elements of the History of Loeale Theory . . . . . . . . . . . . . . . . . . . . . . . . .. 835
Peter lohnstone
Nonsymmetrie Distanees and their Associated Topologies: About the
Origins of Basic Ideas in the Area of Asymmetrie Topology . . . . . . . . .. 853
Hans-Peter A. Knzi
Supereategories of Top and the Inevitable Emergenee of
Topologieal Construets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 969
E. Lowen-Colebunders and R. Lowen
Topological Features of Topological Groups . . . . . . . . . . . . . . . . . . . . . . .. 1027
Michael G. Tkachenko
History of Shape Theory and its Applieation to General Topology ..... 1145
Sibe Marddic and lack Segal
A History of the Normal Moore Spaee Problem. . . . . . . . . . . . . . . . . . . .. 1179
Peter l. Nyikos
Index ......................................................... 1213
Introduction
This account of the History of General Topology has grown out of the
special session on this topic at the American Mathematical Soeiety meeting in
San Antonio, Texas, 1993. It was there that the idea grew to publish a book
on the historical development of General Topology. Moreover it was feit that
it was important to undertake this project while topologists who knew some of
the early researchers were still active.
Since the first paper by Frechet, "Generalisation d'un theoreme de Weier-
strass", C.R.Acad. Sei. 139, 1904, 848-849, and Hausdorff's c1assic book,
"Grundzge der Mengenlehre", Leipzig, 1914, there have been numerous de-
velopments in a multitude of directions and there have been many interactions
with a great number of other mathematical fields. We have tried to cover as many
of these as possible. Most contributions concern either individual topologists,
speeific schools, speeific periods, speeific topics or a combination of these.
The first and second volumes, which were published in 1997 and 1998,
contain the following artic1es:
Felix Hausdorff(1868-1942) (G. Preu)
Frederic Riesz' Contributions to the Foundations ofGeneral Topology (WJ. Thron)
The Contributions of L. Vietoris and H. Tietze to the Foundations of General
Topology (H. Reitberger)
Some Aspects ofthe Work and 1nfiuence of R.L. Moore (B. Fitzpatrick Jr.)
The Works of Bronislaw Knaster (1893-1980) in Continuum Theory (J.J. Chara-
tonik)
Witold Hurewicz - Life and Work (K. Borsuk, transl. by K. Kuperberg, A. Kuper-
berg)
The Early Work of F.B. Jones (M.E. Rudin)
The Beginning ofTopology in the United States and the Moore School (F.B. Jones)
Some Topologists ofthe 1940s (A.H. Stone)
Miroslav Katetov (1918-1995) (Petr Simon)
Origins of Dimension Theory (M. Katetov, P. Simon)
General Topology, in Particular Dimension Theory, in The Netherlands: the De-
cisive 1nfiuence of Brouwer's 1ntuitionism (T. Koetsier, J. van Mill)
The Flowering ofGeneral Topology in Japan (1. Nagata)
V111
schools, periods and subareas of the field that we are seeking authors to write
about.
Most of the authors for this work either were contacted personally by one
of the editors or were recommended by experts in the field. The first drafts of
papers were sent to readers and their suggestions were forwarded to the authors.
We expect that there will be some dis agreement among some authors, but we
also consider this to be healthy. We hope that this work will encourage, not only
further study in the his tory of the subject, but also further mathematical research
in the field.
We would like to thank all colleagues who willingly contributed to what we
hope will become a standard reference work on the History of General Topology.
In view of the fact that most contributors would consider themselves primarily
mathematicians rather than historians of mathematics, we are especially grateful
for their efforts.
Finally, we would like to thank Kluwer Academic Publishers for their
professional support in the publication of this book.
I.M. JAMES
Mathematical Institute
24-29 St Giles, Oxford OXI 3LB
Oxford
Introduction
Point-set topology seems to have become separated from the rest of topology
around the middle of the twentieth century. For most of the period we shall be
considering in this article the term combinatorial topology meant l "practically
everything which could not better be described as point-set topology". Most of
the pioneers were equally at horne in both disciplines, but more recently that
has become rather unusual. One of the purposes of this article is to try and
throw some light on this situation. It is easier to describe what happened than
to explain why, but some insight may be gained if we try to understand what it
must have been like to be a topologist in the first half of the century, especially
in the second quarter. At the end of this article I will attempt an explanation.
Mathematical terminology tends to change in the course of time. The Latin
term 'Analysis Situs' is due to Leibniz and the German term 'Topologie' to
Listing, a member of Gauss's circle. In the English language 'topology' was
used in its mathematical sense as early as 1883 and occasionally later in the
nineteenth century but its widespread adoption was delayed because it had other
scientific meanings, for example in botany where it is recorded in 1659. However
it gradually superseded the older term, partly because it was easy to form deriva-
tives such as 'topologist' and 'topological'. Poincare chose 'Analysis Situs' as
the title for his famous series ofmemoirs, published from 1895 onwards, as did
Veblen for his infiuential Colloquium volume [52] of 1922. Alexander was using
'topological' in the titles of his research papers in the twenties but, as we shall
see, preferred 'analysis situs' when he addressed the International Congress of
t The research on which this article is based was carried out while I held a Leverhulme
Emeritus Fellowship.
1 I am grateful to Dr Shaun Wylie for this information, and for sharing with me his interesting
reminiscences of Princeton in the mid-thirties, referred to below.
809
C. E. Aull and R. Lowen (eds.), Handbook ofthe History ofGeneral Topology, Volurne 3, 809-834.
2001 Kluwer Acadernic Publishers.
810 I.M.JAMES
2 These terms seem to have been interchangeable in the period under consideration, although
more recently a distinction has been drawn.
COMBINATORIAL TOPOLOGY VERSUS POINT-SET TOPOLOGY 811
theory, whereas what Leibnitz had in mind was a new and independent type of
mathematics, especially designed to avoid the complications of function theory
and to deal directly with the purely quantitative aspects of geometrical problems.
No doubt combinatorial analysis situs is more nearly a development ofLeibnitz'
original idea' .
Alexander then gave some examples from the theory ofknots, but for present
purposes we might choose instead the weIl-known problem ofthe seven bridges
of Knigsberg. To consider this problem one has no need of a proper map
of the city, only the cmdest of diagrams is sufficient. One does not need to
know precisely where the pedestrian went, only which bridges were crossed
and in what order. That is the essence of the combinatorial approach, in which
topological spaces are viewed as made up of ceIls, usually siraplexes, whose
topology is weIl understood, and the important thing is how these are fitted
together. In what Alexander calls the point-theoretical approach, it is the points
of the space which matter, and different methods are used. He went on to say
why he preferred the former approach:
'The vogue for point-theoretical analysis situs seems to be due, in large part,
to the predominating inftuence of analysis on mathematics in general. Nowadays
we tend, almost automatically, to identify physical space with the space of three
variables and to interpret physical continuity in the classical function theoretical
manner. But the space of three real variables is not the only possible model of
physical space, nor is it a satisfactory model for dealing with certain types of
problems. Whenever we attack a topological problem by analytic methods it
almost invariably happens that to the intrinsic difficulties ofthe problem, which
we can hardly hope to avoid, there are added certain extraneous difficulties in
no way connected with the problem itself, but apparently associated with the
particular type of machinery used in dealing with it' .
Menger, speaking later in the same Congress, put the case for the point-
theoretic approach. One might be inclined to maintain that this also came
into existence early in the present century, specifically with the publication of
Frechet's work on abstract metric spaces in 1906. But again there is a lengthy
prehistory, as described by Manheim [36], which is quite different from the
prehistory of combinatorial topology. During the nineteenth century there was
a perception that more rigour was needed in analysis, especially in dealing with
sets of functions. The first revolution of rigour is particularly associated with
Cauchy, the second with Weierstrass. Cantor and Dedekind, working indepen-
dently, placed the definition of the real number system on a secure foundation.
Later Cantor developed the theory of abstract sets and this led in due course
to the general notion of topological space, as formulated by Hausdorff in his
classic monograph [19] of 1914.
Hausdorff's work opened up a whole new world to the point-set topologists.
For combinatorialists, however, it was quite sufficient to consider triangulable
812 I.M. JAMES
spaces and at first it seemed that combinatorial methods could not be extended to
more general elasses of spaces. However an idea of Alexandroff's, dating from
1928, enabled combinatorial methods to be applied in great generality. As he
put it in [2]: 'The transfer of homological objects to more general geometrical
objects than polyhedra ... based on my ... concept of the nerve of a covering of a
given space makes it possible to transfer to practically all topological spaces the
fundamental concepts of combinatorial topology'. Remarkably it was this idea
which led to Hausdorff developing a late interest in combinatorial topology3. Not
long afterwards singular homology theory was invented and provided another
way of extending homology theory to spaces in general, and several more ways
of doing so were introduced later [15].
Topology in Europe
Within Europe there was a tendency for the combinatorial tradition to be empha-
sized in the west, and the set-theoretic in the east; to some extent this remains
true even today. To suggest that combinatorial topology fiourished in Western
Europe, that set-theoretic topology fiourished in Eastern Europe, and that both
kinds fiourished in Central Europe, while not entirely false, is to oversimplify a
complex situation. In the Netherlands Brouwer revolutionized both sides of the
subject. In France, despite the impetus given by Poincare's work, research tended
to be more concerned with differential geometry than topology, while in Italy
a strong school of algebraic geometry developed but hardly any combinatorial
topology. In Great Britain, apart from some early interest in knots and graphs,
the study of topology began with Newman, who studied in Vienna.
In the quarter century from 1910 to 1935, or thereabouts, Central Europe
was a hotbed of research in topology. The University of Vienna played a leading
role during the first part of this period. The golden age for topology in Vienna
seems to have begun with the appointment of Wirtinger (see [27]), a elose
friend of Klein. Topology, particularly knot theory, was one of Wirtinger's
research interests, although not the main one. Before long, Vienna became the
leading centre for topology in an its aspects, with faculty members of the calibre
of Menger, Reidemeister and Tietze, and some remarkable students, such as
Hurewicz and Vietoris.
In the early years of the twentieth century the number of mathematicians who
were trying to understand Poincare's work and to develop it further was not large.
Tietze was one of the leaders, and as he put it 'In the domain of analysis situs
Poincare has recently brought us an abundance of new results, but at the same
3 I am grateful to Professor Klaus Volkert for this information, and to the Editors of the
Hausdorff Nachlass for sending me copies of the correspondence between Alexandroff and
Hausdorff.
COMBINATORIAL TOPOLOGY VERSUS POINT-SET TOPOLOGY 813
time raised an abundance of new questions which still await settlement'. Most
of these pioneers were equally at horne in both branches of topology. That was
certainly true of Tietze and Vietoris, the authors of the artic1e [48] on topology
which appeared in the 1931 edition ofthe Enzyklopdie. While the subject had
developed considerably in the intervening 25 years this was not just an up-date
of the earlier artic1e by Dehn and Heegard. For example point-set topology had
come of age, during that period, and so Tietze and Vietoris naturally inc1uded it,
although they preferred to call it general topology4. In fact the title of their artic1e
translates as 'Relations between the different branches of topology'. Unlike the
Dehn-Heegard artic1e, which introduced new ideas, it has the character of a
survey. The first chapter, entitled 'Point-sets in n-dimensional number-space',
is concerned with euc1idean spaces and their subspaces. This is intended as an
introduction to the abstract concept of topological space in the second chap-
ter, simply entitled 'General topology', which inc1udes a discussion of metric
spaces. Covering spaces are also considered. The third chapter, entitled 'n-
dimensional topology' is mainly about manifolds, the fourth is 'Combinatorial
topology'. The artic1e conc1udes with a discussion of dimension theory.
Although specialists in point-set topology were not always interested in
combinatorial topology it would be difficult, at this period, to find a combinato-
rialist who was not well-versed in point-set topology. Possibly as a result there
was a tendency, especially in the early days of homotopy theory, to impose
unnecessary restrictions on the nature of the spaces considered. For example
the statement of a theorem might inc1ude a dimensional condition, which would
mean a restriction to separable metric spaces, but it was so on recognised that
this could be avoided if homological dimension was used instead of the ordinary
dimension of point-set topology. The Borsuk school made a special study of
the c1ass of spaces known as absolute neighbourhood retracts, which in some
respects provided an alternative to the c1ass of triangulable spaces. The concept
of fibre bundle, which developed in the thirties, is an interesting hybrid. It
takes various forms (see [46]) but usually the base space is treated combinato-
rially while the fibre is not. For example the base may be a complex of some
description and the fibre an absolute neighbourhood retract.
By the time the Tietze-Vietoris survey appeared combinatorial topology was
becoming increasingly algebraic, so much so that it evolved into what became
known as algebraic topology. A crucial step in the process of algebraization
was the recognition that homology, seen by Poincare and others as a set of
numerical invariants, was far better thought of as a family of abelian groups. In
his autobiography [3] Alexandroff describes when this happened:
'In the middle ofDecember (1925) Emmy Noether came to spend amonth in
Blaricum. This was a brilliant addition to the group of mathematicians around
as it stood at the outbreak of the second world war, but by the time an English
translation appeared in 1956 the subject had developed so much further that the
work was feIt to be too out-of-date to serve as a textbook. I shall be discussing
textbooks more generally later.
Topology in America
Outside Europe the situation varied from country to country. From about 1925
America was beginning to riYal Europe for mathematical research. While Amer-
ican mathematicians continued to study in Europe it was becoming quite normal
for European mathematicians to spend some time in America. During the twen-
ties the United States became increasingly important for research in topology.
However a strong polarization developed at an early stage, with Lefschetz at
Princeton leading the combinatorialists while R.L. Moore at Austin led the set-
theoreticians. In their different ways both were remarkable teachers as well as
major researchers. The antipathy between these two formidable personalities,
and their adherents, has been exaggerated, but it is part of the his tory of topology
in the United States. It is interesting to try and trace the origins of the two
schools. To do so we need to retrace our steps somewhat since the story begins
in the nineteenth century.
Around the end of the century E.H. Moore was regarded as the leading
mathematician in America. He was greatly inuenced by a post-doctoral year he
spent in Germany. From 1892 he was in charge of the mathematics department at
the University of Chicago. The following year he brought Klein over to lecture,
and this served to strengthen the German inuence. When Hilbert's Grundlagen
der Geometrie appeared in 1899, it stimulated a great deal of interest in the
axiomatic approach to foundational questions, particularly in geometry. Among
E.H. Moore's graduate students both Oswald Veblen and R.L. Moore wrote their
theses in that area. Veblen was a little senior to R.L. Moore (who we will simply
refer to as Moore from now on) and to some extent acted as his advisor. Veblen
was recruited by Princeton in 1905 as one ofPresident Woodrow Wilson's new
'preceptors', as was Moore shortly afterwards, but Moore moved on to settle
down in his native Texas while Veblen made his career in Princeton.
Veblen
his address at the 1932 Congress. In 1924 Alexander in turn persuaded the
University to appoint Lefschetz, and together they made Princeton a leading
centre for research in algebraic topology. Veblen's interests had returned more
to geometry.
Veblen's abilities as a negotiator, fund-raiser and organiser were exceptional.
Not the least of his achievements was the design of Fine Hall, for many years
the home of the Princeton Department of Mathematics. This building was con-
structed of red brick and limestone in the "college style" that Veblen had so
admired when he spent a year at Oxford in 1928-9. He worked c10sely with a
high-quality decorating firm from New York on the furnishings and insisted on
extensive sound-testing of the c1assrooms. There was a first-c1ass departmental
library, a common room and other facilities. Faculty members had "studies" ,not
"offices"; some of these were large rooms lavishly appointed with fireplaces,
carved oak panelling, leather sofas, oriental rugs, concealed blackboards, and
leaded widows.
On the foundation of the Institute of Advanced Study in 1930 mathematics
was chosen as the initial mission. Veblen was given the first faculty appointment
and assumed responsibility for selecting the other original members of the
Institute's School ofMathematics. Alexander, von Neumann, Einstein and Weyl
were appointed in 1933, and Marston Morse two years later. The Institute had
many visitors each year, known as fellows, and a few research associates, known
as permanent members, but the six constituted the regular faculty of the school
of mathematics for the first decade.
Thus the University had lost two of its leading mathematicians to the In-
stitute. Fortunately, however, both institutions recognized the advantages of
cooperation. Institute seminars were open to university personnel and vice versa,
and various facilities were shared. Both gained from having a larger community
ofpermanent and visiting mathematicians. At the University, the able leadership
of Eisenhart and Lefschetz carried the department through this period. While
Eisenhart administered the department and its relations with the rest of the
university, Lefschetz built up the research and graduate programs. The graduate
students were welllooked after and the process of selection was rigorous.
At the Institute it was Veblen who was largely responsible for the policy of
concentrating on postdoctoral work, his views on the subject having been formed
by his experience at the University. As well as being an academic member he was
a trustee of the Institute from its early years and played a large part in arranging
the purchase of the land on which its buildings stand. Outside Princeton Veblen
was also active in the affairs of the American Mathematical Society at a critical
time, and in raising money for mathematical research [16]. During the Nazi
period he was instrumental in helping many European mathematicians to settle
in the United States, no easy matter in the years of the Depression when many
Americans were also looking for academic posts. A full account of all the
COMBINATORIAL TOPOLOGY VERSUS POINT-SET TOPOLOGY 817
Lefschetz
Moore
Let us now leave Princeton for a time and turn our attention to the University
of Texas at Austin, where Moore had come to specialise in point-set theory
(he eschewed the terms analysis situs and topology). Hausdorff's infiuential
COMBINATORIAL TOPOLOGY VERSUS POINT-SET TOPOLOGY 819
7 I am grateful to the Center for American History of the University of Texas at Austin for
permission to quote from the Wilder (Raymond Louis) papers and the Princeton Mathematics in
the 1930's collection in the Archives of American Mathematics.
820 I.M. JAMES
Since this Handbook already contains articles about the Austin school, where so
many American point-set topologists received their training, I should like to try
and balance this by writing more about the Princeton school of combinatorial
topology, which was even more infiuential. Of course other American institu-
tions were important but if we focus our attention on Princeton as it was during
what many see as a golden age, we can begin to understand how, in Whitney's
words, 'topology (was) moving towards America'. Under Alexander, Lefschetz
and others, Princeton became a stronghold of the combinatorialists, who tended
to look down on the other topologists. Even before the Institute for Advanced
Study came into existence visiting mathematicians were particularly attracted to
Princeton. Its high reputation for research in combinatorial topology originated
in the twenties, and was later reinforced by the decline of the universities of the
German-speaking world. Alexandroff, Cech, Hopf, Kuratowski and Whitehead,
amongst others, came to Princeton to work with Alexander, Lefschetz and
Veblen. Of course this was greatly to the benefi.t of the students, especially
young Americans, inc1uding some, such as Fox and Steenrod, who were to join
the Princeton faculty later.
I was fortunate enough to meet some of the leading figures personally, when
I first went to America in 1955/6, and perhaps I might be permitted to say a few
words about this here. I was working initially with Norman Steenrod who I had
got to know when he had visited Oxford and lectured at the topology seminar
on the cohomology operations he had just discovered, soon to be known as the
Steenrod squares. At first I was living in the Graduate College of the University
and working in Fine Hall. However in the New Year Steenrod arranged for his
research student Emery Thomas and myself to accompany him to Berkeley,
where he was spending the rest of the academic year on sabbatical at the
University of Califomia. The following year I retumed to Princeton as a member
of the Institute for Advanced Study, but retained my links with topologists at
the University, especially Fox and Steenrod. In the summer I drove down to
Mexico with the Fox family to attend the important 1956 symposium in algebraic
topology organized by Lefschetz. These and other experiences remain vivid in
my memory.
However this is not the place to describe the Princeton School of Topology
as I found it some years after the second world war. At Oxford Henry Whitehead
had told me much about his experiences at Princeton in the period 1928-31 when
he was studying geometry under Veblen, but growing increasingly interested in
topology. He had told me stories of Alexander, Lefschetz, Veblen and others.
While I was at Princeton I naturally wanted to meet these topologists of an
older generation. Lefschetz was not in Fine Hall much, although I soon had
some experience of his brusque manner. I had tea with the Veblens and dinner
COMBINATORIAL TOPOLOGY VERSUS POINT-SET TOPOLOGY 821
with the Morses. Alexander, I was told, would never appear, so Emery Thomas
and I called on him one day and were received most courteously. No doubt
the Whitehead connection helped to open doors. Gradually I began to form a
picture of what life had been like at Fine Hall before the second world war.
Fortunately, there is a fair amount of relevant material, both published and
unpublished, to show what it was like in those days. For example, interesting
reminiscences of Princeton topologists by Saunders MacLane [35] and Gian-
Carlo Rota [43] can be found in the second of the Centennial volumes of the
American Mathematical Society. The Archives of American Mathematics at
Austin, Texas, contain the transcript of an interview of Tucker by Albert Lewis
and I would like to quote further from this.
Tucker tells what happened when Lefschetz and Moore c1ashed. In 1931-2
Moore was touring American universities as a Visiting Lecturer on behalf of
the American Mathematical Society:
'I remember the time when Moore came to give his visiting lectures. I
think it just for sheer meanness, Moore scheduled the first one of his lectures
for Saturday evening, and then they went on Monday afternoon and Tuesday
afternoon. Well, Lefschetz and Alexander went round to all the graduate students
and said "This is one time when your attendance is required. You have no option.
You must be there". So the room was full. Lefschetz accorded Moore the same
show of interest he would any other speaker, namely by asking questions. Also
Lefschetz did this partly to sort ofhelp the graduate students. Moore started out
by writing out his axioms on the board. Something like five minutes went by,
and there was no sound except this chalk on the blackboard.
'WeH, Lefschetz was very, very restless, and he finally broke in with a
question. Moore turned around slowly and looked at Lefschetz and said, "If you
would read what I have written on the board, you wouldn't need to ask your
question". You could tell when Lefschetz was getting angry because the back of
his neck would get pink. Lefschetz repeatedly asked questions, which we knew
was just Lefschetz's style, but every one of these Moore regarded somehow
as being a planted barb, so he would come back very hard in some scornful
way. Alexander actually tried to pitch in and help Lefschetz a bit, but on points
Moore won easily.
'At the next talk Lefschetz did not say a word. Again he sat where we could
watch the back of his neck, and it would get red and then the colour would
recede, but he didn't say a word. As far as I couldjudge, everybody went out of
his way to show interest and concern and respect towards Moore. But afterwards
at other places he went, I understand he made very caustic remarks about the
822 I.M.JAMES
way he'd been treated, especially by Professor Lefschetz, at Princeton. But this
heckling was just a standard thing with Lefschetz'.
'Though I've been in Princeton only two months it seems like an age. It must
be that I'm enjoying it all, but I've really not had the opportunity to sit down
and come to a conc1usion on the matter - or maybe I don't want to .... This is a
leisurely joint - but I don't imagine I'm telling you anything. It took them two
weeks to get started. It was a month before they put me to work. My job breaks
into two parts. 1) sitting in the library two nights a week, 2) checking over
manuscripts for the Annals (officially as 'assistant editor'). The library work is
no burden. The only things the place lacks are spittoons - a fellow can't have
his chew. The editorial work is a bit different. Just before the proofs are sent out
(i.e. before the printer gets to work) they hand me fifteen or more manuscripts
and I have to go over them with an eye for c1arity - so's the printer Can read
COMBINATORIAL TOPOLOGY VERSUS POINT-SET TOPOLOGY 823
them. Greek underlined in red, German in green, etc ... The main thing 1 don't
like about it is that there isn't enough time allowed for me to read the papers on
topology.
'I was here two weeks before meeting Zippin and Lefschetz. 1 met Tucker
shortly after arrival. 1 outlined for hirn my plans for studying topology. He made
no comment except to recommend seeing Lefschetz, who, he was sure, would
have something to say about it. A few days later Tucker stopped me and said
Zippin had asked ab out me. So 1 set about looking Zippin up. The first step
was to see Miss Blake. 1 trailed her all over the building looking for him. When
this failed she caIled him on the phone, it was about 11.30 in the morning -
so it came about that 1 invited myself to luncheon at Zippin's place. He's a
pretty sweIl sort of person. We talked for over two hours. Among other things,
he strongly recommended my going after the recent efforts of Pontrjagin. He
gave (loaned, 1 mean) me a set of galleys of Pontrjagin's paper in the April
issue of the Annals. This paper, by the way, I've just about finished. It's nice,
easy reading with some red-hot theorems. When the October Annals appeared
1 skimmed through the Pontrjagin paper on the duality theorem to see how he
applied his results. It doesn't seem to me as though there is any other theorem
in mathematics quite as beautiful as the one he's got.
'Lefschetz held a conference with all the new graduate students. He first gave
us a short lecture and then began on us one by one and recommended courses. He
invariably insisted on projective geometry; so when it came to my turn 1 swore up
and down that 1 was weIl versed in the subject. So he finally agreed that perhaps 1
was weIl enough prepared to take his topology course. My other two courses are
Differentiallnvariants - Thomas, and Functional Operators - von Neumann.
About a week later someone, perhaps Tucker, reminded Lefschetz that 1 was
Prof. Wilder' s pet, for he stopped me after one of his lectures and introduced me
to Mr. Wallman and inveigled me into collaborating with Mr.Wallman in writing
up his (Lefschetz's) lecture notes for the topology course. Heaven knows why
this is necessary.
'You mentioned in one of your letters that Lefschetz, in his topology course,
might develop the subject from the point of view of Pontrjagin's work. As yet
it doesn't appear that he is doing this. He did attempt to introduce the notion of
chains with coefficients which are rational numbers modulo 1. But he got stuck
when someone pointed out that they didn't form a ring (i.e. no distributive law).
The next day he insisted that it was all right since all he needed was that they
should form a group. Everyone is still suspicious of the matter.
'Also Lefschetz is having me read some work ofHopf's on the mappings of
complexes. Lefschetz feels that the American topologists have been concern-
ing themselves too much with general problems and have been neglecting the
smaller ones. Hence 1 am to report on these papers at the topology seminar
sometime in January, and thus institute a new order of things.
824 I.M. JAMES
'This place is 'group' crazy. One comes here fresh and innocent. But it isn't
long before you go around babbling about groups. Weyl is giving a continuation
of his course of last year on continuous groups. Eisenhart also on continuous
groups from a slightly different point of view, Lefschetz (although, in his own
words, he is no 'groupologist') can't seem to avoid them, Alexander is lecturing
(in the topology seminar) on Abelian groups - discrete and continuous, and
Zippin abets him occasionally by giving an analysis of the torsion groups.
Besides, as mentioned above, I'm reading Pontrjagin. And in order not to grow
weak on group theory, Nathan, Murray and myself are conducting our own
personal seminar on group theory. Our efforts are centred on trying to discover
(from notes) what Weyl was talking about last year. Nathan and Murray are Nat.
Research fellows; Nathan was at Harvard last year. Both of them are working
under von Neumann trying to solve the problem of unbounded operators in
Hilbert space.
'There is nothing in the way of research I have been able to do .. .I sought
the aid of Lefschetz in order to discover what linked what and how in higher
dimensions. It seems that nothing is known about the latter problem, so at least
I have something to work on. When I questioned Lefschetz about the linking
business he expressed curiosity as to how I came to worry about it. So I told him
about Alexandroff's problem and Antoine's example. Well, Lefschetz launched
into a lecture. He discussed the matter of young mathematicians acquiring the
habit of publishing numerous papers on trivial problems. It appears that the
true Princetonian method is to work only on general problems and to publish
only when some step in theory has been accomplished. What had the whole
Polish school accomplished? There wasn't one thing that he could remember
off-hand. And the only person to come out of the R.L. Moore school that was
worth a damn was Wilder, and probably in spite of it al1. And unless I happened
to be a 'Wilder', I couldn't possibly hope to become a great mathematician by
publishing a lot of papers on trivial problems.
'It was a good lecture, and I agreed that his main thesis was true. I didn't
begin to worry about the matter though until the following day when Zippin
comered me. He said that he had heard that Lefschetz had given me a curtain
talk. He advised me not to worry about it. He said that Lefschetz had ahabit of
getting under people's skin; but he did it entirely without malice. I expressed
surprise that anyone should be upset by Lefschetz, - for surely everyone liked
him. Zippin agreed that everyone liked Lefschetz, but insisted that Lefschetz
had the faculty of getting a person riled.
'It is c1ear that Zippin is an extremely sensitive person, and, thinking that
I was such, had endeavoured to sooth my wounded spirits. However he did
make c1ear that one must not entirely disregard Lefschetz's advice. I notice that
Princeton takes good care of its graduate students. So it seems proper that I
should make an effort to conciliate Lefschetz and work only on such problems
COMBINATORIAL TOPOLOGY VERSUS POINT-SET TOPOLOGY 825
as he recommends and publish only in the Annals. Unfortunately I'm not quite
up to it. I can only work on such problems as I can get my fingers into. If they
don't happen to be important, as is sure to happen, 1'11 be out of luck. So just
now, I'm entirely without inspiration in the matter of research. I feel somewhat
Bat'.
'I'm having a pretty swe11 time here lately. Really getting soused in topology. It
is a whale of ajob writing up the notes for Lefschetz's course. As I mentioned in
my last letter Wallman and I are co11aborating in the matter. The results I think
will be very peculiar. Lefschetz, as you know, is a very sketchy lecturer; so we
have to organize the material, and fi11 in the details. At the beginning Wallman
and I tried working together. But we couldn't get anything done because we
spent the whole time arguing. So we agreed that Wallman would be responsible
for the first half of the first semester, and I would take the latter half. As a result
there will be a certain lack of coherence.
'My duties began at the point in the course where Lefschetz tried to introduce
coefficient groups other than the integers for defining chains. He considered
integers modulo m, rational numbers, and rational numbers modulo I (the last in
deference to Pontrjagin's methods). Remembering that you thought there might
be something screwy about bis methods, I listened very carefu11y. I won't say
he did anything wrong, but he didn't do it right either. I was quite disgusted and
in a fog. So I sat down and wrote a thirty page treatise on the matter. I was very
proud of the result: considered an arbitrary Abelian coefficient group; defined
the homology group to correspond; and showed how, for a complex, to compute
this group in terms of the properties of the coefficient group and then the ordinary
homology group related to the integers; and in case the coefficient group was
topological, I topologized the homology group; using the real numbers modulo
1 (the circ1e); it came out that the homology group was the direct sum of a toral
group (whose dimension was the Betti number) and a finite group (isomorphie
to the ordinary torsion group of one lower dimension). Since this overlapped
somewhat with what Wallman was supposed to do, I gave him the artic1e and
suggested that he write up his part to correspond. We11, he did. He chopped it
up (Le. the artic1e) and inserted parts here and there, substituting his own proofs
when he feIt like it. This disgusted me; it took all the kick out of it - imagine!
Butchering up my brainchild. However I was very pleased to note that one ofhis
substituted proofs was all wet - he made an error of a group-theoretieal nature.
I can't make out whieh of the two is more stubborn, Wallman or myself. As a
case in point we were arguing about the notion of an oriented simplex. lasserted
that it was the product of a feeble inte11ect, or perhaps a strong mind at a weak
moment, that it had caused more topological headaches than anything else, that
826 I.M. JAMES
the word orientation ought to be thrown into discard, and that the notion of chain
was fundamental. Wallman insisted that it had geometrical significance. He said
that he could visualize a line segment with an arrowhead attached, a triangle
containing a circle with an arrow, and a tetrahedron containing a corkscrew. As
a compromise we agreed that each would write the section pertaining to the
definition of chain and its boundary - he, using the notion of chain as basic -
I, avoiding it. We did, and compared results. I proved to him that I used exact1y
two less logical steps than he did. But he stuck to his conviction. As he was
responsible for that part, it will appear with his interpretation of orientation.
'It was the week preceding Xmas that the notes covering the first two weeks
of the course appeared. Lefschetz had spent that time in a somewhat general
discussion of the nature of topology and its applications to other branches of
math. Wallman did a swell job of writing it up. He put in everything Lefschetz
said and a lot more. He had a very beautiful and lengthy discussion on Abelian
integrals on 2-dimensional manifolds. When Lefschetz read it over, he had a fit.
He went tearing around Fine Hall shouting for Steenrod and Wallman. Wallman
had left on his vacation. So Lefschetz collared me, and gave me a lecture.
This was his course in topology. He wanted the notes written accordingly. In
particular he wasn't giving a course in Abelian integrals.
'As a result Wallman and I are trying to be a bit more circumspect. We
decided to throw out the part I had written about topologizing the homology
group in case the coefficient group was topological, and the results for the
modulo 1 case. This breaks my heart. And it seems stupid since it is clear,
from Pontrjagin's results, that a topological homology group is the thing to play
around with.
'I don't think Pontrjagin's results have soaked into Lefschetz's head yet. He
insists on sticking to the rational numbers and the rational numbers modulo 1.
He likes to divide by integers; and rational numbers seem to be all he needs for
this. The modulo 1 is so that he can have a torsion group. I don't see why he
doesn 't use the rest of the continuum in both cases. There is no operation that you
perform on the homology groups of aspace that are not essentially continuous.
If you use only the rationals, the continuity involved does not appear on the
surface ....
'I've been getting a lot ofhelp from Zippin. Perhaps you notice his influence.
I think I mentioned that he made me read Pontrjagin's paper of last April. Once
in a while I ron into a difficulty concerning groups. After a long struggle I figure
out a fact and a proof. Then I take it to Zippin and state the fact. He always says
'Oh sure'. Then he explains how it can be seen very easily by using theorems
on character groups ....
'This semester Lefschetz is tackling the homology groups of an abstract
space. He's using Cech's methods - following the Fundamenta paper closely.
COMBINATORIAL TOPOLOGY VERSUS POINT-SET TOPOLOGY 827
He says he's only read the first part of this paper. I bet he follows Ceeh, and
uses rational ehains.
'I've leamed how to get help from Lefsehetz. Tbe thing to do is to refuse
to pretend to understand him when he isn't c1ear. He eertainly has a buneh of
good ideas. Still, though, he does misfire oeeasionally.
'Tbis Wallman is a nice ehap. He has energy and ideas. I get a kick out of
working with him' .
I started this letter two months ago - I'm just a punk letter writer. In the first
edition of this I started to tell you what Ceeh is like. But you will be seeing
him shortly so it isn't neeessary. However my impression is that he is a pretty
swell guy. He enjoys having people disagree with him, and doesn't have to be
handled with gloves.
'Life has been very enjoyable for me this year. Last Oetober I passed the
exams that admit me to eandidaey for the doetor's degree. Sinee then I've been
struggling with the thesis. Lefsehetz wouldn't look at the thing until it was
typewritten. I gave it to him two weeks ago. Tbe ehanges he is making are large
and numerous. I feel slightly put upon. Tbe typing was a terrifie job - the thing
is a hundred pages long (a good bit ofit is putting things together). Tbe ehanges
shouldn't take me more than a month so relief is on the horizon. I notieed when
writing up the thesis that I have a very sensitive eonscienee. This I ean traee to
the training you gave me. I remember distinetly the time you refused to aeeept as
a proof a very niee intuitive argument I had. Tbe same sort of thing eame up in a
very extensive way in my thesis. In terms of bases of a buneh of groups I made
a eonstruetion. I was certain that the result of the construction was dependent
only on the invariants of the groups. But it took me a long time to find out what
I meant by this, and to work out the details of the proof.
'During the past semester Alexander leetured on eombinatorial topology.
He's the best leeturer around here. He eoneentrated on the infinite eomplex
(very infinite-non-eountable and loeally finite). He had all sorts of eyc1es-finite,
infinite, infinite in the large but finite in the small, ordinary and dual eyc1es for
all these types. He was trying to develop an interseetion theory so that he would
have a homology ring, but he ran into diffieulties. So he stopped leeturing and
said he would begin again onee the trouble had been ironed out. Sinee then both
Ceeh and Whitney have proposed definitions of interseetions. It is believed that
they are equivalent. Anyhow they agree that only dual eyc1es and a dual eyc1e
with an ordinary eyc1e ean be multiplied. Ceeh has generalized the definition to
eyc1es of an arbitrary spaee. When the definition has been decided 1'11 have to
eonsider the problem of universal eoefficient domains for the homology ring.
828 I.M. JAMES
'Beginning in January last, I've been running a seminar in topology; and doing
it in the spirit of the R.L. Moore school. So far I've been following the notes
to the course you gave in Michigan ... The thing I've noticed about Princeton
is that the students who come here without having done research work have
considerable difficulty in knowing how to do it. So I talked up the virtues of the
Moore system to Lefschetz. He finally agreed that it was worth trying. However
he complained bitterly about the material I proposed using. So far he has been
satisfied. He attended several of the meetings and was very pleased with the
way things were going. He was so pleased in fact that he is starting a seminar in
algebraic geometry to be conducted in as nearly the same fashion as possible.
'Hurewicz gave us aseries of lectures on his work. It is beautiful stuff and
plenty of it' .
Recognition of Topology
Due recognition of the place of topology in mathematics was not won easily.
Analysis became increasingly rigorous, in the course of the nineteenth century,
and once some early misconceptions had been cleared up point-set topology
achieved a high standard of rigour as well. Combinatorial topology, however,
was considered to be less rigorous, not without justification. The intuitive ar-
guments often used in combinatorial topology were not always reliable, even
in the hands of a master like Poincare. The history of the Poincare conjecture,
which is still undecided, contains many examples of unsuccessful proofs. The
attempts to prove Dehn's lemma provide other examples. It is hardly surprising
that it took some time before the status of combinatorial topology was generally
accepted.
In the early International Mathematical Congresses topology hardly appears
on the programme at all, although often mentioned in the major addresses.
For example, in the very first Congress Hurwitz drew attention to the need for
the classification of three-dimensional manifolds, something still not achieved
over a century later. But it was not until the Zurich Congress of 1932, already
COMBINATORIAL TOPOLOGY VERSUS POINT-SET TOPOLOGY 829
mentioned, that the subject first seems to have won appropriate recognition.
One recaHs that Hadamard, in a famous lecture given in New York in 1911,
described it as 'the revenge of geometry on analysis'; presumably when he said
this it was combinatorial topology which he meant, since point-set topology was
was then in its infancy and in any case was more in the nature of an outgrowth
of analysis.
In 1935 the first specialist conference on topology was held in Moscow.
This brought together most of the active researchers at that time, about forty
in number. Whitney has given his personal impressions of what took place in
[54]. Another such meeting took place in Geneva shortly afterwards. Plans for
a major topology conference to be held in Warsaw in 1939 were abandoned due
to the gravity of the international situation. After the second world war was over
international conferences on topology began to be held quite frequently, but they
were normallyon either combinatorial topology or on point-set topology, not
on both together as was the practice before the war.
Textbooks on Topology
There seem to have been two major attempts to combine combinatorial and
point-set topology in one treatise. One was by von Kerekjmo, whose first vol-
ume [29], covering set-theoretic topology and the topology of surfaces, appeared
in 1923. The second volume, which would have dealt with higher-dimensional
spaces, was drafted by Kneser8 , after von Kerekjmo himself abandoned the
project, but was never completed. The other project was by Alexandroff and
Hopf, whose first volume [4] appeared in 1935. The titles ofthe four parts ofthis
volume are: 'Grundbegriffe der mengentheoretischen Topologie', 'Topologie
der Komplexe', 'Topologische invarianz satze und ansschliessende Begriffsbil-
dungen ' and 'Verschlingungen im Euklidischen Raum. Stetige Abbildungen von
Polyedern' .
At the end there is an appendix on abelian groups and another on convex
subspaces of euclidean space, also an annotated list of 25 relevant books on
topology. Apparently the projected second volume was to have been on point-
set topology, which is not taken very far in the first volume, and the third on
the topology of manifolds. However although the first volume at once became
a standard work and remained so until weH into the fifties, nothing further ever
appeared, unless the work [1] on combinatorial topology written by Alexandroff
alone and described above may bear some relation to the original project.
At a more introductory level there were a number ofbooks which dealt with
both point-set and combinatorial topology, such as those ofHocking and Young
8 I arn grateful to members of the Kneser family for a copy of this draft, and to Dr Moritz
Epple for his comments thereon.
830 I.M.JAMES
[24] and Schubert [44], which appeared in 1961 and 1964, respectively. These
books combined basic point-set topology with aselection of combinatorial
topics, such as simplicial homology. However since the combinatorial topology
made little use of the point-set topology, there did not seem much point in
putting them together, and this kind of textbook is no longer published.
Of course there were plenty of books to choose from which dealt with either
point-set topology or combinatorial topology but did not try to combine them.
The first publication which could be regarded as a textbook of combinatorial
topology was Veblen's Analysis Situs [52] of 1922, based on his colloquium
lectures of 1916. This appeared too early to have treated homology from the
group-theoretical viewpoint, and Lefschetz' Topology [31] of 1930 was an at-
tempt to bring it up-to-date. Reidemeister published several textbooks, of which
his Kntentheorie [40] of 1932 was particularly useful. The 1934 Lehrbuch
der Topologie [45] of Seifert and Threlfall, which inc1uded a particularly good
introduction to the topology of manifolds, remained a standard work for many
years. In the case of point-set topology, Hausdorff's c1assic Grundzge der
Mengenlehre [19], which appeared in 1914, laid the foundations. His later Men-
genlehre [20] of 1927 is a much shorter book, largely because topology is given
less space. By that time several accounts of elementary point-set topology had
appeared, and gradually the standard exposition we know today was perfected,
as for example in Bourbaki [8]. Most ofthe later textbooks are written with the
needs of analysts in mind; sometimes the fundamental group is discussed but
otherwise there are no links with other kinds of topology.
Historical Literature
The history of algebraic topology up to (hut not inc1uding) the time of Poincare
is the subject ofthe thesis ofBollinger [6]. The monograph [40] of J.-C. Pont, is
largely based on this thesis. The only similar history of general topology seems
to be that of Manheim [36], although a number of artic1es have been published
on particular topics. The little-known survey [22] by Hirsch provides a well-
informed overview of the development of topology generally. Dieudonne wrote
several relevant studies, of which [14] is perhaps the most relevant. A number of
artic1es ofhistorical interest are listed in the bibliography. These inc1ude surveys
such as [21], [49] and [53], also the biographies and autobiographies of certain
individuals.
A useful list of early publications may be found at the end of the van der
Waerden artic1e [51] on combinatorial topology, and an annotated list of early
textbooks at the end of Alexandroff and Hopf [4]. It must be aCknowledged,
however, that the history of topology, especially twentieth-century topology,
has not yet received the attention it deserves from the professional historians.
Much valuable unpublished material has not yet been studied. All too few of the
COMBINATORIAL TOPOLOGY VERSUS POINT-SET TOPOLOGY 831
Conclusion
With all this background let us now return to the original question: why did the
separation between the two disciplines occur? Perhaps the subject had simply
become too large to hold together. Perhaps cultural factors, such as the compara-
tive lack of contact between topologists in Western Europe and those in Eastern
Europe, were also important. Perhaps the appearance of abstract topological
spaces in full generality distracted the attention ofpoint-set topologists from the
more geometrically-motivated questions which interested the combinatorialists.
Or perhaps it was feIt that, once the basic canon of point-set topology had been
established, the motivation for further research came rather from analysis than
from elsewhere in topology, as Whitney suggested when introducing the 1950
Conference on Topology held at Cambridge (Mass.):
'The subject of algebraic topology and applications was chosen ... because
of its great growth in recent years, and the increasingly large contact with other
fields of mathematics, in geometry, algebra and analysis. The subject of general
topology has moved considerably into the domain of analysis. It was with great
regret that the field of point set topology had to be omitted altogether' .
Until then it had been usual for both branches of the subject to be represented
at conferences. But while Whitney was surely right to say that point-set topology
was moving away from combinatorial topology it would also be true to say
that combinatorial topology was moving away from point-set topology. The
introduction of algebraic methods may have had some effect but the advent of
homotopy theory seems to have been more important. Although I have already
touched on aspects of the history of homotopy theory incidentally, a brief sketch
of its development may help to explain the reasons for making this assertion.
Of course the notion of continuous deformation, especially of paths, has a
long history. The term homotopy was introduced by Dehn and Heegard in their
Enzyklopdie artic1e of 1907, although the meaning they gave to it was not the
one which is now standard. It was more like simplicial homotopy, but not the
same. Although Veblen adopted their definition in his 1922 colloquium volume
the much simpler standard definition soon replaced it. While Hopf is genera11y
regarded as the founder of algebraic topology it was Hurewicz who was the
founder, or at least co-founder, of homotopy theory. Many of the fundamental
ideas of homotopy theory were introduced by Hurewicz in 1935/6 and as we
have seen he was lecturing on them in America soon afterwards.
832 I.M. JAMES
The most significant of these ideas, perhaps, was the concept of homotopy
type, the c1assification of spaces by homotopy equivalence. This circumvented
many of the difficulties of the older c1assification, by homeomorphism. Homol-
ogy, of course, is a homotopy invariant but many other topological invariants,
such as compactness, are not. In fact the methods of point-set topology are
largely irrelevant to homotopy theory, where algebraic methods played an ever-
increasing role. Later of course it turned out that homotopy theory could also be
used to help solve many of the old problems of c1assical topology. Nothing was
happening in point-set topology which could rival such spectacular successes.
However, although fifty years aga point-set topology seemed to have exhausted
itself, before long it took on a new lease of life, as described elsewhere in this
collection of artic1es.
References
[16] Feffer, L.B., Oswald Veblen and the capitalization of American Mathematics:
raising money for research 1923-1928, Isis 89 (1998), 474-497.
[17] Feigl, G., Geschichtliche Entwicklung der Topologie, lahresber. Deutsche Math.
Vereinig., 37 (1928), 273-280.
[18] Freudenthai, H., Topologie in den N ederlanden: das erste Halbjahrhundert, Nieuw
Arch. Wiskunde III Ser., 26 (1978), 22-40.
[19] Hausdorff, E, Grundzge der Mengenlehre, Von Leit, Leipzig, 1914.
[20] _ _ , Mengenlehre, De Gruyter, Berlin, 1927.
[21] Henn, H.-W. and Puppe, D., Algebraische Topologie, Ein Jahrhundert Mathe-
matik 1890-1900, Deutsche Math. Vereinig., 1992,673-716.
[22] Hirsch, G., Topologie, Abrege d'histoire des mathematiques 1700-1900,
Dieudonne (ed.), Hermann, Paris, 1976, 211-266.
[23] Hirzebruch, EH., Emmy Noether and topology, Proc. Israel Math. ConJ., 12
(1999), 57-65.
[24] Hocking, I.G. and Young, G.S., Topology, Addison-Wesley, Reading, Mass.,
1961.
[25] Hodge, W.V.D., Solomon Lefschetz 1884-1972, Biogr. Mem. Roy. Soc., 19
(1973),433-453.
[26] Hopf, H., Ein Abschnitt aus der Entwicklung der Topologie, lahresber. Deutsche
Math. Vereinig., 34 (1925), 1-14.
[27] Hornich, H., Wilhelm Wirtinger, Monatshefte fr Mathematik (1), 52 (1948),
1-12.
[28] James, I.M. (ed.), History o/Topology, North Holland, Amsterdam, 1999.
[29] von Kerekjarto, B., Vorlesungen ber Topologie, Julius Springer, Berlin, 1923.
[30] Kneser, H., Die Topologie der Mannigfaltigkeiten, lahresber. Deutsche Math.
Vereinig., 34 (1925), 1-14.
[31] Lefschetz, S., Topology, Amer. Math. Soc., Providence, RI, 1930.
[32] _ _ , A page of mathematical biography, Bull. Amer. Math. Soc., 74 (1968),
854-879.
[33] _ _ , Reminiscences of a mathematical migrant in the U.S., Amer. Math.
Monthly, 77 (1970), 344-350.
[34] _ _ , James WaddelI Alexander (1888-1971), Biog. Mem. Amer. Phil. Soc.
(1973), Philadelphia, 1974, 110-114.
[35] MacLane, S., Topology and logic at Princeton, ABrief History, A Century 0/
Mathematics in America, Duren (ed.), Amer. Math. Soc., Providence, RI, part II,
1988, 127-154.
[36] Manheim, J.H., The Genesis 0/ Point Set Topology, Pergamon Press, Oxford,
1964.
[37] Marcus, L., Solomon Lefschetz: an appreciation in memoriam, Bull. Amer. Math.
Soc., 79 (1973), 663-680.
[38] Montgomery, D., Oswald Veblen, Bull Amer. Math. Soc., 69 (1963), 26-36.
[39] Parshall, K.H. and Rowe, D.E., American Mathematics comes of age: 1875-
1900, A Century 0/ Mathematics in America, Duren (ed.), Amer. Math. Soc.,
Providence, RI, part III, 1989,3-28.
[40] Pont, J.-c., La Topologie Algebrique des Origines cl Poincare, Presses Univ. de
France, Paris, 1974.
834 I.M. JAMES
PETER JOHNSTONE
Department of Pure Mathematics
University of Cambridge, England
Contents
835
C. E. Aull and R. Lowen (eds.), Handbook ofthe History ofGeneral Topology, Volume 3,835-851.
2001 Kluwer Academic Publishers.
ELEMENTS OF THE HIS TORY OF LOCALE THEORY 837
1. Prehistory: 1914-54
In the early his tory of topology, Felix Hausdorff [27] is generally (and rightly)
credited with being the first person to emphasize the importance of the notion
of open set in formulating the definitions of topological concepts. (Some credit
for popularizing the idea should perhaps also be given to the Polish school of
topologists led by Sierpmski and Kuratowski, in the years after the First World
War.) The idea has therefore been around for almost as long as topology itself
that a topological space is something which possesses (indeed, is defined by) a
lattice of open subsets, as well as a set of points.
However, the idea that a topological space is a lattice of open sets, and that
the points themselves are a secondary construct, took a much longer time to
emerge. In part, this was because such an idea was dependent on the devel-
opment of lattice theory as an autonomous branch of abstract algebra, which
did not take place until the late 1930s with the work of Birkhoff and Stone,
culminating in the first edition of Birkhoff's well-known textbook [14]. Thus,
although Vietoris's astonishing paper [95] of 1922 contains what is in effect
an entirely lattice-theoretic description of how to construct the hyperspace (the
space of all nonempty c10sed subsets) of a compact Hausdorff space, it could
not be recognized in those terms because the concepts in abstract lattice theory,
which were needed to formulate it as such, did not then exist. It was not until
sixty years later [38], [41] that a purely lattice-theoretic translation ofVietoris's
construction could be published.
Whilst Birkhoff's work was enormously influential, in that it defined and
standardized the language that mathematicians needed in order to talk about the
concepts of abstract lattice theory, it was really the work ofMarshall Stone [88],
[89] on the topological representation of Boolean algebras that laid the true
foundation-stone (forgive the pun!) ofthe lattice-theoretic approach to topology
- just as it was the starting-point for so many other important developments of
mathematics in the middle third of the twentieth century. I have analysed some
of these in the Introduction to my book Stone Spaces [38], and any reader who
is familiar with that Introduction will have to forgive me for inducing here what
will undoubtedly be a sensation of deja vu: it is in my view almost impossible
to overstate the influence of Stone's representation theorem on the subsequent
development of mathematics.
The key idea that made Stone's work a genuinely new starting-point was the
fact that topological spaces could be defined from purely algebraic data such as
a Boolean algebra. Prior to this, although topologists had gained considerable
expertise in the construction of 'exotic' or 'pathological' spaces in order to
demonstrate the independence of different topological properties, the motivation
for studying topological spaces at all had come entirely from geometrical roots
- the spaces one wished to study were those constructed by geometrical means
838 PETER JOHNSTONE
Strietly speaking, loeale theory was not born in 1957: the word 'loeale' did
not aequire a mathematical meaning until Isbell's paper [29] fifteen years later.
Nevertheless, it is fair to regard 1957 as the true origin of the subjeet which we
now eall by the name 'Ioeale theory'. (In this survey, I am anaehronistieally using
the word 'loeale' whenever I wish to refer to the geometrie objeet popularly
known as a 'spaee without points', and 'frame' when I refer to the algebraie
objeet which represents it set-theoretieally. The different writers in the field
ELEMENTS OF THE HISTORY OF LOCALE THEORY 839
gave the subject its real impetus. The very first expose of the new seminar [67]
was one in which the two Paperts reported on the work they were doing in
Cambridge; it was followed by one in which Benabou [13] determined the free
objects in the category of frames, and observed that they are a11 spatial (in fact
they are the topologies ofpowers ofthe Sierpinski space). From this simple (and
essentially categorical) observation follow two important facts: first, that every
frame is a quotient of a spatial one (equivalently, every locale is a sublocale of a
space - hence the title of Isbell's paper 'Atomless parts of spaces' [29], which
(as we shall see in the next section) brilliantly exploited the idea that locales are
not just a generalization but an improvement of the category of spaces, which is
necessary in order to make the concept of subspace behave better), and secondly,
that the algebraic theory of frames is exactly the infinitary algebraic theory (in
the sense of Unton [54]) of the contravariant 'forgetful functor' from spaces to
sets which sends aspace to its open-set lattice. (The latter point has been further
developed in recent years by Barr and Pedicchio [11], [12].)
The Ph.D. theses of the two Paperts [66], [68] have already been mentioned.
Whilst Seymour Papert subsequently left the subject of locales for the greener
pastures of automata theory, Dona Papert (latterly as Dona Strauss) came back
to it from 1966 onwards in a long sequence of joint papers [17], [18], [19],
[20], [21] with Hugh Dowker, culminating in [22] which was published after
Dowker's death. Whilst this sequence stretches weH into the periods covered by
later sections of this artic1e, it has its origins firmly in the period of initial growth;
and Dowker himself, though he published nothing on locales before 1966, had
certainly been thinking about pointless topology as a useful generalization of
c1assical topology from the late 1950s, as is made c1ear in [90]. It is interesting
to note that Dowker's motivation for this development, like that of Ehresmann,
came from sheaf theory (on which he had written a well-known set of lecture
notes in 1956), and the way in which sheaves mediate the passage between local
and global structures on spaces; consideration of this led him, like Ehresmann,
to observe that the requirement that spaces should 'consist of points' actually
gets in the way when one works in this area, and one can give a smoother
development of the theory in the pointless context. (The same observation, in the
context of the sheaves used in algebraic geometry, led Alexander Grothendieck
to the still more general notion of topos; I shall have more to say about this
development in the next section.) Dowker's other undisputed contribution to
the subject was the introduction of the word 'frame' to describe the type of
lattice which was needed to carry the structure of a 'pointless space'; although
this usage did not immediately carry all before it, it has come to be universally
accepted as the ideal choice of terminology, being short, memorable and not
readily confused with any other usage in lattice theory or elsewhere.
Progress in the study of frames continued through the 1960s on several
other fronts. In particular, Bernhard Banaschewski, who had been interested in
ELEMENTS OF THE HISTORY OF LOCALE THEORY 841
the lattice-theoretic aspects of topology from an early stage [2], published his
first paper to contain an explicit mention of frames in 1969 [3]. Banaschewski
was to become one of the most prolific authors on pointless topology, and in
particular on its constructive and choice-free aspects, in the next three decades.
However, a further (and more literal) revolution was needed before pointless
topology was able to occupy its true place in the spectrum of tools available to
the modem topologist.
Up to about 1970, if anyone had asked what was the reason for studying topology
on a lattice-theoretic foundation, the only answer that any of its practitioners
could have given was the mountaineer's 'Because it's there' . But at the beginning
of the 1970s there came two new developments, one very specific and the other
much broader and more diffuse, which radically changed the way in which
'pointless topologists' looked at their subject. The specific development was
the publication of John Isbell's celebrated paper [29], and the more diffuse one
was the development of elementary topos theory and the discovery of its links
with constructive topology.
On a purely technicallevel, Isbell did two things which decisively changed
the way in which (most) locale-theorists viewed their subject. One was the
introduction of the word 'locale' itself (an inspired choice, which conveyed all
the right overtones about the spatial nature of these objects without causing
their algebraic underpinnings to obtrude, and which at the same time was
easily capable of all the necessary inflections), and the other - linked to it
- was the decision that the notion of locale morphism must be seen as going in
the 'geometric' direction, opposite to that of the frame homomorphism which
represents it algebraically. The latter change, though not without its opponents
- even today there are distinguished locale-theorists who still write their papers
in terms of frames rather than locales, and who insist that it is not necessary to
dualize the category in order to 'see' it in topological terms - is surprisingly
important psychologically (cf. [45]): ifyou have to turn the arrows round every
time you pass from (traditional) spaces to frames, you cannot help thinking
of the latter as a representation (and generalization) of the former, whereas
making the functor covariant allows you to see locales as 'just another category
in which one can do topology', on the same footing as c1assical topological
spaces, Choquet pseudotopological spaces, neamess spaces, or whatever else
may take your fancy. (Of course, it's not quite on the same footing: for example
it is not a topological category in the technical sense of that term. But it has
just as much right to be viewed as a valid description of the domains on which
'continuously varying quantities' are defined.)
842 PETER JOIINSTONE
But both these things are mere teehniealities beside the more substantial
ehanges that Isbell brought about in how we view loeales. The key one was
the observation that, in eertain respeets at least, the eategory of loeales is
aetually an improvement on the eategory of topologie al spaees: it eonforms
more c10sely to our intuition about how the objeets whieh are the domains of
eontinuously varying quantities ought to behave. The partieular instanees of this
good behaviour addueed by Isbell were eoneerned with the preservation of topo-
logieal properties under produets - a notoriously diffieult area of traditional
topology, many ofwhose diffieulties ean now be seen as arlsing direetly from the
(unreasonable) requirement that produet spaees ought to 'live on their points'.
But one eould equally well point to the good behaviour of dense subloeales:
the fact that any interseetion of dense subloeales is dense, again eontrasting
with what happens if we impose the requirement that, in order to be dense,
the interseetion must contain points. (The two are of course inter-related; both
are instanees of the good behaviour of (eategorieal) limits in the eategory of
loeales.) Thus it beeame possible to see the absence of points as a 'liberating'
feature ofpointless topology, rather than a (literally) pointless generalization, in
mueh the same way that Hausdorff's advoeaey of open sets as the right language
in whieh to define topologieal spaees had a liberating effeet on topology sixty
years earlier.
Seeondly, Isbell's paper lays great stress on the lattiee of subloeales of a
loeale, and on the way in whieh it (and not the 'enve1oping Boolean algebras'
favoured by Nbeling and earlier writers) provides the eorreet substitute for the
lattiee of all subspaces of aspace. Of course, it is not a Boolean algebra, in
general; but it is at least a eoframe (that is, it is isomorphie to the lattiee of
c10sed subloeales of another loeale) and is generated as such by eomplemented
elements. (In passing, I believe that [29] contains the first published proof of this
fact, although I suspeet that it may have been known to one or two others slightly
earlier than 1972. However, the eorresponding result for toposes was not known
to the Grothendieek sehool, sinee they posed as an open problem in [1] the
question whether the lattiee of subtoposes of a topos is always distributive.) As
Isbell wrote, 'Lattices of subloeales eannot be said to be a teehnieal improvement
on Boolean algebras of subspaees, but they are very good lattiees (and contain
more information)'. In that spirit, he made extensive use of these lattiees in his
investigation of loeales, opening the way for further studies later in the deeade
by Harold Simmons [85], [86], [87], and more reeently for the development
of 'deseriptive loeale theory' by Isbell himself and his student Till Plewe [31],
[32], [33], [70], [72].
Turning to the seeond infiuenee that revolutionized the study of locales
in the early 1970s, namely the rise of elementary topos theory, one is faeed
immediately with a question: why was not the eonnection between loeales and
toposes made explicit by the Grothendieek sehool [1] in the early 1960s? They
ELEMENTS OF THE mSTORY OF LOCALE THEORY 843
were perfeetly weIl aware that the geometrie notion of topos whieh they were
developing was a generalization of the notion of topologieal spaee (indeed,
the name 'topos' was ehosen preeisely for that reason), although they were
perhaps over-optimistie as to just how general the new notion was - for an
analysis of this point, see [40]. They were also aware that their generalization
inc1uded the ease of a 'spaee without points', sinee the example of an atomless
eomplete Boolean algebra appears in [1]. But they seem to have totally ignored
the eonneetion between their notion and the work of Ehresmann and Benabou:
this appears to have been another instanee of Ehresmann's legendary isolation
within the Freneh mathematieal eommunity.
Whatever the reason for this omission, the developers of elementary topos
theory were weIl aware that, in its geometrie aspeets, a topos is more eorrectly
seen as a generalized loeale than as a generalized spaee. This is already apparent
in [51], and made mueh more explieit in Lawvere's later artic1e [52]. The
input from eonstruetive logic which elementary topos theory brought, and made
available to help the understanding of problems on the geometrie side of the
subjeet, was most expertly applied by Andre Joyal in his early work (most of it
unpublished, but see [24], [25] for some of its fruits). The key to understanding
the link between eonstruetive logic and loeale theory is the observation that
'singling out' a point of a spaee is an essentially c1assieal operation, in that
it separates the spaee into the point under eonsideration and whatever is not
the point; in a eonstruetive eontext, sueh deeompositions do not exist exeept
in fortunate special eases, so we are obliged to eonsider spaee as eonsisting
essentially of its open parts. (The same observation also arises naturally in
theoretieal eomputer scienee; see [93].)
Related to this observation is the faet that loeale theory is (for the most
part - but see [83] for an exeeption) inherently choice-free: many theorems
and definitions, whieh require the axiom of ehoice if we eonsider spaee as
eonsisting of points, are valid in loeale theory without assuming ehoice. Perhaps
the first explieit instanee of this faet was the present author's observation [37]
that Tyehonoff's theorem for loeales has a ehoice-free proof; however, es-
sentially the same observation was made independently by Banasehewski and
Mulvey [7], [8], and my original proof (which, although ehoiee-free, was not
eonstruetive) has been suecessively simplified and further eonstruetivized by
several authors [50], [5], [47], [91], [15], [62]. Related to this development
was a good deal of further work on produet loeales and related eonstruets sueh
as funetion-Ioeales [28] and hyperloeales [41], all of whieh tended to eonfirm
Isbell's insight that produets are better-behaved in the eategory of loeales than
they are in spaees.
The eonstructivization of topology is not in itself of any direet benefit to
c1assical topology. However, the impetus to construetivize had adefinite pur-
pose; it was motivated by the observation that any topos (in particular, the topos
844 PETER JOHNSTONE
This final seetion is of neeessity mueh more sketehy than its three predeeessors.
To summarize all the lines of development in loeale theory sinee 1984, and to
relate them to one another, would require a whole book rather than part of a
ehapter. All I ean do within the eompass of the present article is to indieate a
few of the direetions that the development has taken, and to provide referenees
for those who wish to follow them in greater detail.
Some of these direetions were foreshadowed in my survey article [39],
written shortly after the eompletion of [38]. For example, I foresaw the need for a
theory of uniform loeales, and this has been developed by a number of authors,
most notably by Ales Pultr [73], [74], [75], [76], [77]. (However, somewhat
against my expeetations, there seems to be a genuine divergenee between the
eonstruetive and classical theories at the uniform level, which is not present at
the 'eontinuous' level.) Related to this is the theory of loealic groups, which
had begun slightly earlier with Wraith's observation [97] that many familiar
examples oftopological groups (such as Galois groups of infinite extensions) are
more naturally viewed as loealie groups. In unpublished work in 1981, Franees
Kirwan had shown that, beeause of the differenee between loeale produets and
ELEMENTS OF THE HIS TORY OF LOCALE THEORY 845
spatial produets, the spaee of rationals did not admit a loealie group strueture; a
few years later, this gave rise to the surprising observation [34], [42], [43], [98]
that every loealie subgroup of a loealie group is c10sed as a subloeale. Onee
again, this ean be seen as an instanee where working in the eategory of loeales
simply eliminates 'pathologies' which oeeur in the eategory of spaees.
John Isbell [30] took up my ehallenge in [39] to develop a dimension
theory for loeales, although less has followed from this than I had expeeted.
As mentioned earlier, Isbell and his student Till Plewe have sinee developed the
loealie analogue of deseriptive set theory [31], [32], [33], [70], [72], whieh now
seems to be a flourishing growth area.
Many of the themes mentioned in the previous seetion have eontinued to
develop. On the subjeet of good behaviour of loeale produets, a genuine surprise
was the observation by Madden and Vermeer [59] that realeompaetness (whieh
had earlier been defined for loeales by Reynolds [78]) was equivalent to the
Lindelf property for regular loeales. (See also [69] and [83] for further develop-
ments in this area.) Less of a surprise was the proofby Moerdijk and Wraith [60]
that loeal and global eonneetedness together imply path-eonneetedness for 10-
eales - yet again, the eategory of loeales exc1udes pathologies to whieh one
is aeeustomed in the eategory of spaees. Jim Madden [58] has also done im-
portant work on the generalization of the notion of frame where one assumes
the existenee of joins only for subsets of restrieted eardinality - as also has
Bemhard Banasehewski [6]. The connections between loeale theory and fibre-
wise topology were given a new impetus by the present author's diseovery [43]
that the c1assieal notion of c10sure bifureates eonstruetively into 'fibrewise' and
'absolute' versions; see also [35], [44], [49], [91] for applieations of this idea.
And further investigations of openness have been earried out by Banasehewski
and Pultr [9], [10]; the extension of the Joyal-Tiemey des cent theorem to new
eontexts by Vermeulen [92] and Plewe [71] has already been mentioned.
In fact loeale theory itself seems to have bifureated into c1assical and eon-
struetive versions: inereasingly, the problems which ean be taekled using c1as-
sicallogic have moved away from the areas in which eonstruetive teehniques
ean be applied. In the former camp, an interesting development has been the
emergenee of a flourishing Chinese sehool of loeale-theorists, largely fostered by
Professors Wang Guo-Jun in Xi'an and Liu Ying-Ming in Chengdu: important
developments which ean be eredited to this sehool inc1ude the diseovery by Sun
Shu-Hao [46] of a notion of 'weak produet' for loeales, which sheds interesting
light on the differenee between loeale produets and spaee produets, and Li Yong-
Ming's [53] introduction of a usable notion of 'quotient map of loeales', whieh
has sinee been developed further by Till Plewe.
On the eonstruetive side of things, mention should be made of Giovanni
Sambin's introduetion [80], [81] of the notion of 'formal spaee' (see also [16],
[62], [65] for some applieations of this notion). Striet1y speaking, this is not
846 PETER JOIINSTONE
a development of locale theory, since its motivations are different: its aim
is to develop the notions of topology within the predicative environment of
Martin-Lf type theory, rather than the (constructive, but impredicative) logic
encountered within a topos. (Much of the motivation for working in such a
context comes from the demands of theoretical computer science.) The require-
ment of predicativity imposes the need for great care in formulating the basic
definitions, but once this is done the resulting development has striking paralleis
with locale theory - it seems likely that the two subjects will reinforce and
enrich each other as time goes on.
Theoretical computer science has also had a more direct input into the
development of locale theory (and even into topos theory) , again as a result
of the inherent constructivity of its logic. The idea that datatypes are not simply
sets but 'domains' which carry intrinsic order-theoretic and topological structure
has been around since the work of Dana Scott [84] in the early 1970s, but
increasingly the need to view them constructively has led computer scientists to
represent them as locales rather than spaces. In particular, the need to consider
spaces of subsets (known to computer scientists as powerdomains rather than
hyperspaces) has led several people (for example, Schalk [82] and Vickers [94])
to revisit and extend the ideas of [41].
Finally, mention must be made of the spectacular growth (not foreseen
in [39]) of 'non-commutative locale theory': the theory of quantales, which
owes its origin to Chris Mulvey's enigmatically-titled paper [61], and which
has had notable success in providing a c1ear and unified language in which to
describe the representation theory of non-commutative rings and C* -algebras.
Already, the theory of quantales has developed into a subject which would
require another artic1e as long as this one to do justice to its history; all I shall
do is to refer the interested reader to the book [79] by Rosenthal.
References
[1] Artin, M., Grothendieck, A. and Verdier, J.L., Theorie des Topos et Cohomologie
Etale des Schemas, Seminaire de Geometrie Algebrique du Bois-Marie, annee
1963-64; second edition published as Lect. Notes Math. vols. 269, 270 and 305
(Springer-Verlag, 1972).
[2] Banaschewski, B., Untersuchen ber Filterrume, Doctoral dissertation, Univer-
sitt Hamburg (1953).
[3] _ _ , Frames and compactifications, in Extension Theory 0/ Topological
Structures and its Applications (Deutscher Verlag Wissen., 1969),29-33.
[4] _ _ , Coherent frames, in Continuous Lattices, Lect. Notes Math. vol. 871
(Springer-Verlag, 1981),1-11.
[5] _ _ , Another look at the localic Tychonoff theorem, Comment. Math. Univ.
Carolin. 29 (1988), 647-656.
[6] _ _ , The frame envelope of a a-frame, Quaest. Math. 16 (1993), 51-60.
ELEMENTS OF THE IllSTORY OF LOCALE THEORY 847
[31] _ _ , First steps in descriptive theory of locales, Trans. Amer. Math. Soc. 327
(1991), 353-371; corrigenda ibid. 341 (1994),467-468.
[32] _ _ , Some problems in descriptive locale theory, in Category Theory 1991,
C.M.S. Conf. Proc. no. 13 (Amer. Math. Soc., 1992), 243-265.
[33] _ _ , Some structure of Borel locales, Proc. Amer. Math. Soc. 126 (1998),
2477-2479.
[34] Isbell, J.R., KTIz, 1., Pultr, A. and Rosicky, J., Remarks on localic groups, in
Categorical Algebra and its Applications, Lect. Notes Math. vol. 1348 (Springer-
Verlag, 1988), 154-172.
[35] Jibladze, M. and Johnstone, P.T., The frame of fibrewise closed nuclei, Cahiers
Top. Geom. DijJ. Cat. 32 (1991), 99-112.
[36] Johnstone, P.T., Open maps of toposes, Manuscripta Math. 31 (1980),217-247.
[37] _ _ , Tychonoff's theorem without the axiom of choice, Fund. Math. 113
(1981),21-35.
[38] _ _ , Stone Spaces, Cambridge Studies in Advanced Math. no. 3 (Cambridge
Univ. Press, 1982).
[39] _ _ , The point of pointless topology, Bull. Amer. Math. Soc. (N.S.) 8 (1983),
41-53.
[40] _ _ , How general is a generalized space?, in Aspects ojTopology in Memory
oj Hugh Dowker, L.M.S. Lect. Notes Sero vol. 93 (Cambridge Univ. Press, 1985),
77-111.
[41] _ _ , Vietoris locales and localic semilattices, in Continuous Lattices and their
Applications, Lect. Notes Pure Appl. Math. vol. 101 (Marcel Dekker, 1985),
155-180.
[42] _ _ , A simple proof that localic subgroups are closed, Cahiers Top. Geom.
DijJ. Cat. 29 (1988), 157-161.
[43] _ _ , A constructive "c1osed subgroup theorem" for localic groups and
groupoids, Cahiers Top. Geom. DijJ. Cat. 30 (1989),3-23.
[44] _ _ , Fibrewise separation axioms for locales, Math. Proc. Camb. Phi/os. Soc.
108 (1990), 247-256.
[45] _ _ , The art ofpointless thinking: a student's guide to the category oflocales,
in Category Theory at Work, Research Expos. Math. vol. 18 (Heldermann Verlag,
1991), 85-107.
[46] Johnstone, P.T. and Sun, S.H., Weak products and Hausdorff locales, in Categor-
ical Algebra and its Applications, Lect. Notes Math. vol. 1348 (Springer-Verlag,
1988),173-193.
[47] Johnstone, P.T. and Vickers, S.J., Preframe presentations present, in Category
Theory, Lect. Notes Math. vol. 1488 (Springer-Verlag, 1991), 193-212.
[48] Joyal, A. and Tiemey, M., An Extension of the Galois Theory of Grothendieck,
Mem. Amer. Math. Soc. no. 309 (1984).
[49] Kock, A., A Godement theorem for locales, Math. Proc. Camb. Phi/os. Soc. 105
(1989),463-471.
[50] KTIZ,I., A constructive proof ofTychonoff's theorem for locales, Comment. Math.
Univ. Carolin. 26 (1985), 619-630.
[51] Lawvere, F.W., Quantifiers and sheaves, in Actes du Congres International des
Mathematiciens, Nice 1970, Tome I (Gauthier-Villars, 1971),329-334.
ELEMENTS OF THE mSTORY OF LOCALE THEORY 849
[97] Wraith, G.C., Localic groups, Cahiers Top. Geom. Diff. 22 (1981), 61-66.
[98] _ _ , Unsurprising results on localic groups, 1. Pure Appl. Alg. 67 (1990),
95-100.
NONSYMMETRIC DISTANCES AND TBEIR ASSOCIATED
TOPOLOGIES: ABOUT TBE ORIGINS OF BASIC
IDEAS IN TBE AREA OF ASYMMETRIC TOPOLOGY
Contents
1. Introduction
We begin with some remarlcs explaining the strueture of this article. After
some introduetory statements in the following paragraphs, we summarize the
historie development of what is now often ealled "Nonsymmetrie or Asymmetrie
Topology" in Seetion 2. In the following, more speeme seetions we diseuss the
historie development of some of the main ideas of the area in greater detail. The
list of seetions and keywords given above should help the specialist to find his
way through the various seetions.
For further information about the area we refer the reader to the two mono-
graphs written by Murdeshwar and Naimpally [415] (Quasi-Uniform Topolog-
ical Spaees) and Fleteher and Lindgren [197] (Quasi-Uniform Spaces), respee-
tively. The first book was published in 1966; the seeond one appeared in 1982.
A short look at the lists of eontents of the two books reveals that there was mueh
progress in the area of quasi-uniform spaees between 1966 and 1982:
The book of Murdeshwar and Naimpally diseusses the basie properties of
quasi-uniformities and quasi-pseudometries and their indueed topologies. In par-
tieular, the authors study how quasi-uniformities generate topologies satisfying
eertain separation axioms; for instanee it is shown that a quasi-uniform spaee
induees an Ro-topology if and only if the interseetion taken over its family of
entourages is symmetrie [414]. They also deseribe how quasi-uniform struetures
induee bitopological spaees in the sense of Kelly [277]. A erucial idea of the
book is to use Pervin-Sieber completeness to extend the classical eoneept of
eompleteness known from the theory of uniform spaees:
In order that eonvergent filters be Cauehy, Pervin and Sieber [448] had
ealled a filter ~ on a quasi-uniform spaee (X, OlL) a Cauchy filter provided that
for eaeh entourage U E OlL there exists x E X sueh that U (x) E ~. Sinee in
general sueh filters do not eonverge to their cluster points, Murdeshwar and
Naimpally defined a quasi-uniform spaee (X, OlL) to be complete provided that
eaeh Cauehy filter has a cluster point in (X, OlL). (As usual now, in the following
we shall use the term "eonvergenee eompleteness" for the stronger property
that eaeh sueh filter eonverges. It is known that the two properties eoincide in
quasi-uniform spaees that are loeally symmetrie [197, Corollary 3.9].)
Note that in uniform spaees the given definition yields the usual eoneepts
of Cauehy filter and eonvergenee. One dis advantage of this definition is that
eomplete subspaees of (Hausdorft) quasi-uniform spaees need not be closed
(see e.g. [81] or [197, Example 3.10]).
The book ended with some elementary, but important applieations of quasi-
uniformities to the theory of funetion spaces.
A bitopological variant of Pervin-Sieber completeness was later discussed
in a paper ofRichardson [485]. Recently Romaguera and Schellekens [511] also
studied the condition that in a quasi-uniform space (X, OlL) each Pervin-Sieber
856 HANS-PETER A. KNZI
Cauchy filter clusters with respect to the topology 'l' (CU v CU -I); their investiga-
tions were motivated by the observation that the quasi-metric complexity space
introduced by Schellekens possesses the latter property, which implies Smyth
completeness and (in TI-spaces) small-set symmetry.
The monograph of Fletcher and Lindgren included many additional ideas:
The basic connections between quasi-proximities and quasi-uniformities are
discussed. The theory of the bicompletion of a quasi-uniform space is dealt
with. Nachbin's theory ofuniform ordered spaces is explained and the theory of
the bicompletion is applied to the concept of ordered completions and compact-
ifications. Deep connections between notions from the theory of quasi-uniform
spaces and topological concepts like neighbomets and open coverings are care-
fully discussed. The authors also pay special attention to quasi-uniformities
possessing a base consisting of transitive entourages or having weak symmetry
properties. Basic results on quasi-metric spaces and quasi-metrizability of topo-
logical spaces are treated. In a final section the authors collect some notes that
comment on the historical development of their subject.
Since the book ofFletcher and Lindgren was published, eighteen years have
passed. Much work was done in the meantime for which that book has provided
a basis so that our understanding of quasi-uniform structures has improved
considerably durlng the last eighteen years. Let us mention some of the central
new ideas a future book on quasi-uniform structures will certainly have to deal
with:
Numerous researchers developed the theory of extensions further. Com-
pietion theories for balanced quasi-metrics and quiet (resp. stable) quasi-
uniformities were established.
Quasi-uniform structures were applied to problems in theoretical computer
science by numerous mathematicians and computer scientists: To this end the
concepts of strongly sober compactification, weighted quasi-metric, continuity
space, topological quasi-uniform space and Yoneda-completion were introduced
and investigated.
Methods and tools from category theory, nonstandard analysis and de-
scriptive set-theory were applied successfully to the study of quasi-uniform
structures.
Quasi-uniformities in function spaces as well as hyperspaces attracted the
attention of various mathematicians. Although our knowledge is still insufficient
in these areas, parts of a structured theory begin to emerge.
Besides the classical theory of quasi-uniform spaces the theories of fuzzy
quasi-uniformities, probabilistic quasi-metrics, quasi-uniform frames and ap-
proach quasi-uniformities were developed. The study of quasi-uniform struc-
tures in topological algebra was intensified. Further applications of methods
from the theory of quasi-uniform spaces to problems in approximation the-
NONSYMMETRIC DISTANCES AND THEIR ASSOCIATED TOPOLOGIES 857
ory and functional analysis were discovered. Many classical problems about
quasi-uniformities were at least partially solved.
Since no book on these topics is available at present, the reader may wish to
consult some of the survey articles that were written after 1982 and that contain
much information about several of these topics:
In [124] Deak: discussed extensions of quasi-uniformities and quasi-pseudo-
metrics from subspaces to larger spaces, in particular completions. A survey on
extensions of quasi-uniformities is also due to Csaszar [108].
In [320] Knzi dealt with recent results on quasi-uniform spaces that are
related to problems and facts discussed in the afore-mentioned book ofFletcher
and Lindgren. In [322] he collected many results about monotonic, stable and
quiet quasi-uniformities. Furthermore he reviewed the theories of weighted
quasi-metrics and topological quasi-uniform spaces. Finally, in [324] he dis-
cussed quasi-uniformities on function spaces, mainly homeomorphism groups.
Kopperman's survey article [297] is devoted to the study of asymmetry and
duality with respect to topological spaces. Classical examples of mathematical
structures that come in symmetric and asymmetric versions include commuta-
tive and noncommutative algebraic structures as well as symmetric preorders
(equivalence relations) and asymmetric preorders (partial orders). His basic idea
is that in such cases there is always a duality available whose use simplifies their
study.
Finally, in [61] Brmmer surveyed several categorical aspects ofthe bicom-
pietion of functorial quasi-uniformities.
Let us also observe that there exists a still very useful, although now slightly
obsolete survey article about quasi-metrizable spaces, which was published by
Kofner [285] in 1980. More information about quasi-metrizable spaces can be
found in Gruenhage's survey artic1e on generalized metric spaces [226] in the
Handbook of Set-Theoretic Topology. The history of these spaces is also treated
in Hodel's article [250].
A bibliography on papers dealing with quasi-metric spaces was compiled
by Reilly [474] in 1992. Though the list of references given at the end of
the present paper (for simplicity in strict1y alphabetic order!) has been made
sufficiently large so that it should be useful to many readers having different
goals, we did not attempt to obtain a complete list of articles published in the
area of asymmetric topology.
We shall concentrate in this survey on the c1assical concepts of quasi-
uniformities and quasi-metrics. Thus with the exception of Lowen's approach
spaces [378],[382], which as an especially important concept of asymmetric
distance structure will be discussed in the last section of this article, related struc-
tures like Csaszar's syntopogenous structures [98], [99], Brown's confluence
para-quasi-uniformities [51], Szaz's relator spaces [565], Hodel's neighborhood
assignments [249], pointwise quasi-uniformities on completely distributive lat-
858 HANS-PETER A. KNZI
today is due to Pervin and yields the finest totally bounded quasi-uniformity of a
topological space (e.g. [355]), which is now called the Pervin quasi-uniformity.
Adescription of the same quasi-uniformity was given by Nielsen and Sloyer
[431] with the help of bounded semicontinuous functions, as was noted for
instance by Hunsaker and Lindgren [258]. The set of filter quasi-uniformities
on a (nonempty) set is readily seen to be a complete lattice under set-theoretic
inc1usion (see e.g. [415]).
Alternative characterizations of quasi-uniform spaces in terms of families
of quasi-pseudometrics (resp. in terms of families of pair-covers) were given by
Csaszar, Ganter, Steinlage, Reilly and Salbany [98], [216], [469], [470], [516].
The term quasi-proximity first appeared in the articles of Pervin and Steiner
[447], [553]. Pervin's definition was not quite correct and Steiner had to fill a
gap in Pervin's original definition. At about the same time quasi-proximities
and related concepts were studied in terms of strong inc1usions by various other
mathematicians, e.g. by Dowker [144] and in a more general form by Csaszar
[98].
According to a statement made in the book ofFletcher and Lindgren [197],
after the publication of the monograph by Murdeshwar and N aimpally the theory
of quasi-uniform spaces developed in the following five main directions:
(1) The work of Brmmer [54], Salbany [518], Carlson and Hicks [80],
Csaszar [100], [101], Hunsaker and Lindgren [258] and Stoltenberg [554], [555]
investigated relationships between quasi-uniformities and quasi-proximities;
these authors established the existence of completions and compactifications
and constructed a theory of quasi-uniformities which is similar to the established
theory of uniformities. Some of these researchers made use of categorical tools
to formulate and obtain their results. In their investigations the bitopological
aspect of the theory was not neglected. The filter of inverse relations of a
quasi-uniformity is also a quasi-uniformity. Similarly, each quasi-pseudometric
has an obvious conjugate by interchanging the order of points. Hence quasi-
uniformities and quasi-metrics naturally generate bitopological spaces; i.e. sets
equipped with two topologies. In 1963 Kelly ([277], see [157] for a correction)
had initiated the study of such spaces. He defined pairwise regular and pairwise
normal bitopological spaces and obtained interesting bitopological analogues
of c1assical topological theorems. He noted that if one studies the bitopological
space induced by a quasi-pseudometric, then one regains some of the symmetry
of the c1assical metric situation and in consequence one can obtain systematic
generalizations of many standard results. Lane continued the study ofbitopolog-
ical spaces. In [358] he defined pairwise completely regular spaces and proved
that these are precisely the quasi-uniformizable bitopological spaces. This was
also done, independently, by Fleteher [169]. (A bitopological space (X, r;},~)
is called pairwise completely regular if for each x E X and disjoint r;}-c1osed
set A there is a r;}-lower semicontinuous and Q-upper semicontinuous function
860 HANS-PETER A. KNZI
f from X into [0, 1] such that f(x) = 1 and f(A) = 0; and for each 22.-
c10sed set B not containing x, there is a \!JI-Iower semicontinuous and 22.-upper
semicontinuous function g from X into [0, 1] such that g(x) = 0 and g(B) = 1.
A subbase characterization ofthis property related to quasi-proximities was later
given by Aarts and Mrsevic [1].)
The notion of the bicompletion of a quasi-uniform space, first investi-
gated by Csaszar [98] in a more general context, was further analysed by
Salbany [518]. He established that there is a one-to-one correspondence be-
tween (bi)compactifications of pairwise Tychonoff spaces and compatible quasi-
proximities. In particular, he studied the bitopological Cech-Stone compactifica-
tion. The latter is characterized by the unique extension property ofbicontinuous
maps into the bitopological unit interval equipped with the quasi-pseudometric
u(x, y) = (y - x) V o. Pairwise Tychonoff (sup)compact bitopological spaces
were shown to admit a unique compatible quasi-uniformity. Bach bicontin-
uous map from such aspace into a quasi-uniform space is quasi-uniformly
continuous.
In [194] Fletcher and Lindgren described the now c1assical construction of
the bicompletion as the paircompletion.
(2) Another c1assical part of the theory consisted of the work mainly due
to Junnila [266] and Scott [534] on covering properties associated with quasi-
uniformities. The best known of these properties are undoubtedly "orthocom-
pactness" and its generalizations like preorthocompactness.
(3) The study of Fox [206], Junnila [265] and Kofner [284], [288] about
neighbornets of topological spaces showed among other things that the c1asses
of y-spaces, quasi-metrizable spaces and non-archimedeanly quasi-metrizable
spaces are all distinct and that the fine quasi-uniformity of metrizable and
suborderable (= generalized ordered) spaces has a base consisting of transitive
entourages.
(4) Important investigations due to Fox and Kofner (published later in [209]),
Heath [235], [236], Junnila [266], Nedev [430], and Stoltenberg [556] dealt
with quasi-metrizable spaces and their relations to developable spaces and other
generalized metric spaces.
(5) The work ofBlatter and Seever [45], LaI and Singal [354] and Redfield
[466] tried to exploit the use of quasi-uniformities in the study of Nachbin's
completely regular ordered spaces.
In the foHowing paragraphs we similarly intend to c1assify roughly the main
directions of research after the monograph "Quasi-Uniform Spaces" had been
published by Fletcher and Lindgren.
Csaszar [108], Deak [124], Doitchinov [136], [137], [139] and Render [479]
continued the work on extensions and completions of quasi-uniform spaces.
Related to their investigations Fletcher and Hunsaker [178], [180], [181] as weH
as Romaguera [496] studied different kinds of symmetry conditions and filters in
NONSYMMETRIC DISTANCES AND THEIR ASSOCIATED TOPOLOGIES 861
In the light of all the interesting developments that have taken place since
publication of the monograph "Quasi-Uniform Spaces" by Fletcher and Lind-
gren, we can certainly be confident that also in the years to come progress in the
area of asymmetric topology will not slow down and many new and surprising
results will still be detected and published.
3. Basic Constructions
For the following discussion recall that a quasi-uniformity Oll, is called totally
bounded if the coarsest uniformity Oll,* finer than it (the so-called associated
supremum uniformity Oll, v Oll, -1) is precompact, and it is said to be precompact
if each (quasi-)uniform cover has a finite subcover. It is well known that the
concepts of total boundedness and precompactness are equivalent for uniform
spaces. It was observed very early (see e.g. in [415] and the paper by Lambrinos
[357]) that in the setting of quasi-uniformities precompactness is strictly weaker
than total boundedness and need not be preserved under finite suprema.
In [415], [448] it was proved that a quasi-uniform space is precompact
(resp. totally bounded) if and only if each ultrafilter is Cauchy (resp. Cauchy
with respect to the supremum uniformity). In partieular a quasi-uniform space
is compact if and only if it is precompact and Pervin-Sieber complete. Pervin
and Sieber [448] showed that a topologie al space is compact if and only if each
compatible quasi-uniformity is convergence complete.
The connection between quasi-uniformities and quasi-proximities is analo-
gous to the symmetric case. Each quasi-uniformity naturally induces a quasi-
proximity and there is a one-to-one correspondence between quasi-proximities
and totally bounded quasi-uniformities; in fact by defining the appropriate maps,
namely quasi-proximally continuous maps (resp. quasi-uniformly continuous
maps), this correspondence yields two isomorphie categories.
These results were essentially established by Hunsaker and Lindgren [258]
who - improving on results of Fletcher [170] - showed that given a quasi-
proximity space (X, 8) the collection of all sets of the form T(A, B) = [(X \
A) x X] U [X x (X \ B)], where A8B, is a subbase for a totally bounded quasi-
uniformity Oll,8, whieh induces 8, Le. A8B if and only if there is an entourage
V satisfying V (A) n B = 0. Moreover, Oll,8 is the coarsest quasi-uniformity
that induces the quasi-proximity 8. Furthermore it is the unique totally bounded
quasi-uniformity inducing 8.
By definition, the quasi-proximity dass of a quasi-uniformity Oll, consists of
those quasi-uniformities that induce the quasi-proximity induced by Oll,. Recently
Losonczi [372] observed that for an arbitrary quasi-uniformity Oll, the coarsest
(and totally bounded) member Oll, w belonging to the quasi-proximity dass of Oll,
is equal to the infimum of Oll, and the Pervin quasi-uniformity corresponding to
the topology induced by Oll,.
NONSYMMETRIC DISTANCES AND THEIR ASSOCIATED TOPOLOGIES 863
tota11y bounded quasi-uniformities of some topological space and that for a topo-
logical space admitting a coarsest quasi-uniformity the lattice of its compatible
tota11y bounded quasi-uniformities need not be modular.
A sufficient condition for the existence of a nontransitive tota11y bounded
compatible quasi-uniformity on a topological space was obtained in [219].
The condition was used to prove that each infinite completely regular T2-
space admits a nontransitive totally bounded quasi-uniformity. It is still un-
known whether each infinite T2-space admits a nontransitive totally bounded
quasi-uniformity.
Fletcher and Lindgren had proved in their book that the coarsest compatible
quasi-uniformity of a locally compact T2-space is a uniformity if and only
if the space is compact [197, Proposition 1.47]. In [308] Knzi character-
ized those topological spaces that admit a coarsest quasi-unijormity, equiv-
alently a coarsest quasi-proximity. This c1ass of spaces generalizes the c1ass
of core-compact spaces. In particular each locally compact space (i.e. each
point has a neighborhood base consisting of compact sets) admits a coarsest
quasi-uniformity.
To this end he introduced the following auxiliary concept: Let (X, ~) be a
topological space and let GI, G2 E ~. Write GI< G2 if for each ultrafilter '9
on X containing GI there exists a finite collection .M. of open sets of X such that
each element of.M. contains a limit point of '9, and n.M. ~ G2. (As above, the
convention is used that, n0 = X.) If GI < G2, then GI is said to be handy in G2
(with respect to X). Knzi showed that a topological space X admits a coarsest
quasi-uniformity if and only if its handy-relation is approximating (i.e. for each
open set G of X we have G = U{G ' : G' < G and G' is open in X}). He also
established that a topological space which admits a coarsest quasi-uniformity
and in which each convergent ultrafilter has an irreducible convergence set is
core-compact and that a topological space admits a unique quasi-proximity if
and only if each of its open sets is handy in itself.
Furthermore, he observed that some well-known results on locally compact
spaces generalize to the c1ass of topological spaces admitting a coarsest quasi-
uniformity; for instance a nonempty product of topological spaces admits a
coarsest quasi-uniformity if and only if each factor space admits a coarsest
quasi-uniformity and all but finitely many factor spaces are compact. Finally
he constructed a sober TI-space that is not core-compact, but has an open base
consisting of sets that are handy in themselves (and thus admits a coarsest
quasi-uniformity); he also noted that if a quasi-uniformity is minimal among
the compatible quasi-uniformities on a topological space X, then it is neces-
sarily the coarsest compatible quasi-uniformity on X. It is also known that if
a topological space admits a coarsest transitive quasi-uniformity then it is the
coarsest compatible quasi-uniformity (compare [370]).
868 HANS-PETER A. KNZI
4. Functorial Quasi-uniformities
Let T denote the (obvious) forgetful functor from the category Quu of quasi-
uniform spaces and quasi-uniformly continuous maps to the category Top of
topological spaces and continuous maps. A Junctorial admissible quasi-
uniformity on the topological spaces is a functor F : Top ~ Quu such that
T F = 1, i.e. F is a right inverse or section of T, briefly a T -seetion. Functorial
admissible quasi-uniformities on subcategories of Top are defined similarly.
Brmmer [54] was first to consider explicitly the dass of a1l functorial
quasi-uniformities, although some basic work on canonical covering quasi-
uniformities was done at about the same time by Fletcher and Lindgren [188],
[189]. He proved that the dass of all T -sections has a complete lattice structure
with respect to the partial ordering "coarser than". In particular he described the
coarsest functorial quasi-uniformity on Top. Indeed, he showed that the coarsest
Junctorial quasi-uniformity on Top is the Pervin quasi-uniformity by observing
that this quasi-uniformity is the initial quasi-uniformity for a1l continuous maps
into the SierpiIiski space with its unique quasi-uniformity. Of course, in gen-
eral a topological space admits many nonfunctorial quasi-uniformities (some
of which may be strictly coarser than the Pervin quasi-uniformity). The fine
quasi-uniformity of a topological space determines the finest T -section. It will
be discussed thoroughly in Section 5.
V sing the idea of initiality, Brmmer also popularized a method, now known
as the spanning construction, for building sections of the forgetful functor T
[54]. Much of Brmmer's work (see e.g. [55]) was motivated by the search
for a natural extension of dassical uniform space results to quasi-uniform and
bitopological spaces. His slogan was that in general such an extension is not
unique [57], [58]. For instance he showed that the realcompactness reflection
has several extensions to bitopological spaces.
In general functorial quasi-uniformities are not preserved under the usual
topological operations. While the restriction of the Pervin quasi-uniformity to
an arbitrary subspace is the Pervin quasi-uniformity, the corresponding result for
the fine quasi-uniformity does not hold; but it is true for subspaces that are the
intersection of an open and a dosed set, as is observed in the book by Fletcher
and Lindgren [197, Corollary 2.19]. In general functorial quasi-uniformities
also are not preserved under products; for instance Fletcher and Lindgren [197,
Corollary 2.18] showed that if the product of the Pervin quasi-uniformity with
itself of a Hausdorff space X is the Pervin quasi-uniformity on X 2 , then X is
finite.
870 HANS-PETER A. KNZI
Salbany [518] studied the (pairwise) completely regular spaces as the initial
hull ofthe bitopological unit interval Ib in Bitop and showed that (sup)compact
pairwise Tychonoff bitopological spaces yield the epireflective hull of Ib. He
also proved that the first coordinate functor K 1 : Pcreg ~ Top from the
category of (pairwise) completely regular bitopological spaces has a unique
right inverse Ql : Top ~ Pcreg. For X in Top, the first topology of QIX
is that of X, and the second topology has the c10sed sets of X as a base of
open sets. Indeed, it is the bitopological space generated by the Pervin quasi-
uniformity CfP of X. The associated supremum topology r(CfP*) is usually called
the Skula or b-topology of X [546]. Only recently compact Skula topologies
were characterized by Dow and Watson [143].
Salbany also discovered that a map / : X ~ Y between completely regular
To bitopological spaces is epi if and only if / is dense with respect to the
supremum topology of Y. Brmmer subsequently showed that any map between
To quasi-uniform spaces is epi if and only if it is dense with respect to the
associated supremum uniformity.
As in the last section, a quasi-uniformity is called transitive if it has a base
consisting of transitive entourages; for instance any Pervin quasi-uniformity has
this property. Much of the initial work on transitive spaces is due to Fletcher and
Lindgren [188], [189]. Birsan [42] and Reilly [471] proved that a bitopological
space is pairwise zero-dimensional if and only if it admits a transitive quasi-
uniformity. In [110] it was shown how the concept of topogenous order can be
used in answering questions conceming transitive quasi-uniformities.
The following construction of transitive quasi-uniformities is now known as
the Fletcher construction (see [172]).
Let (X, 5") be a topological space and let .stl be a collection of interior-
preserving open covers C(6 such that U.stl is a subbase for 5". For any C(6 set
U'il, = UXEX({X} x n{D : x E D E C(6}). Then the collection {U'il, : C(6 E .stl} is a
subbase for a compatible transitive quasi-uniformity on X.
On the other hand, if OU is any transitive quasi-uniformity compatible with
5", then c1early there is a collection .stl of interior-preserving open covers of X
such that U.stl is a subbase for 5" and OU = OUd.
Examples for the Fletcher construction abound (see e.g. [197]): If .stl is the
collection of all finite open covers of a topological space X, then OUd is the
Pervin quasi-uniformity of X; if .stl is the collection of all interior-preserving
open covers, then OUd is the fine transitive quasi-uniformity of X, that is the
finest compatible transitive quasi-uniformity on X; if .stl is the collection of all
weH-monotone open covers (i.e. open covers well-ordered by set-theoretic in-
c1usion), then we obtain the weIl-monotone (open covering) quasi-uniformity 0/
X introduced by Junnila [266]. It is preserved under restriction to subspaces and
coincides with the Pervin quasi-uniformity if and only if the underlying space
NONSYMMETRIC DISTANCES AND THEIR ASSOCIATED TOPOLOGIES 871
In [310] Knzi also showed that a weakly Lindelf almost ~o-fuIly nor-
mal (in the sense of Mansfield [400]) Tl -space that admits a complete quasi-
uniformity is realcompact.
Recently, Knzi and Watson [352] gave an example of a quasi-metrizable
space whose finest compatible quasi-uniformity is not complete. As they noted,
such aspace cannot be countably metacompact, because the finest compatible
quasi-uniformity on a countably metacompact quasi-pseudometrizable space is
complete. Since their space does not satisfy the Hausdorff separation axiom, the
problem to construct such a regular example remains open (see Problem 9 of
[320]). It seems also unknown whether the fine quasi-uniformity of any Moore
space is complete (see [336]).
By a result of Ferrario and Knzi the finest compatible quasi-uniformity of
each quasi-pseudometrizable space is bicomplete [150]. Subsequently
Romaguera and Salbany [503] showed that indeed the finest compatible quasi-
uniformity of a quasi-pseudometrizable bitopological space is bicomplete (see
also [445]). In [499] Romaguera also proved that the fine quasi-uniformity
of any sub-To-quasi-pseudometrizable (pairwise) completely regular bispace is
bicomplete.
Moreover, Ferrario and Knzi established that a topological space admits a
bicomplete quasi-uniformity if and only if its fine quasi-uniformity is bicom-
plete. The fine transitive quasi-uniformity of any countable or any first-countable
TI-spaces is bicomplete; on the other hand the fine (= unique) quasi-uniformity
of the cofinite topology on an uncountable set is not bicomplete. Further-
more the finest quasi-uniformity of each quasi-sober space is bicomplete. In
fact, as mentioned above, they noted that the sobrification of a topological
space can be obtained by constructing the bicompletion of its weIl-monotone
quasi-uniformity. Recently some far reaching extension of the latter result was
obtained by Flagg, Kopperman and Snderhauf [165] (see Section 11).
As was noted by Perez-Peiialver and Romaguera [444], the result of [336]
that the fine (indeed the weIl-monotone quasi-uniformity) of each topological
space is left K -complete, can be deduced from the fact that any filter that is
stable with respect to the conjugate of the weIl-monotone quasi-uniformity of
a topological space X contains its set of adherence points (with respect to the
topology of X).
The fine quasi-uniformity of a topological space need neither be right K-
complete nor D-complete, according to observations due to Knzi and Ryser
[345] resp. Knzi and Romaguera [336].
In [443] Perez-Peiialver and Romaguera pointed out that no quasi-metrizable
(Hausdorff) space seems to be known whose finest quasi-uniformity is not right
K -complete, although simple quasi-pseudometrizable spaces of this kind can
readily be constructed. It can be shown that the fine quasi-uniformity of each
submetacompact space is right K -complete.
878 HANS-PETER A. KNZI
Hicks and Sharma [246] considered topological spaces that have the property
that every neighbornet belongs to the fine quasi-uniformity. By an observation
due to Fox [205] (resp. Junnila, see [197, Proposition 6.25]) the Michaelline
(resp. any topological space that possesses a c1osure-preserving c10sed cover by
finite sets) is of this kind. Junnila [266] noted that also each preorthocompact
space that is the countable union of c10sed discrete subspaces has this property.
Among other things Hicks and Sharma showed that spaces satisfying their
condition are scattered provided that they are locally compact Hausdorff spaces
having a G &-diagonal.
A quasi-uniform space (X, OU) is called bicomplete provided that the coarsest
uniformity OU * finer than OU is complete.
It is now well known that each quasi-uniform To-space (X, OU) has an (up to
quasi-uniform isomorphism) unique bicompletion (X, OlL) in the sense that the
space (X, OlL) is a bicomplete extension of (X, OU) in which (X, OU) is r (0lL*)-
dense. The uniformities (0lL)* and oU* coincide. Furthermore if D is a r(OU*)-
dense subspace of a quasi-uniform space (X, OU) and f : (D, OUID) -+ (Y, "V) is
a quasi-uniformly continuous map where (Y, "V) is a bicomplete quasi-uniform
To-space, then there exists a (unique) quasi-uniformly continuous extension
1: X -+ Yof f.
It seems that any reasonable extension of the completion theory of uniform
spaces to arbitrary quasi-uniform spaces first naturally leads to Csaszar's theory
of bicompleteness [98] outlined above and originally developed in the realm of
syntopogenous spaces.
In the context of quasi-proximities and quasi-uniformities the construction
was popularized and further developed mainly by Salbany [518] and Fletcher
and Lindgren [194]. In particular Fletcher and Lindgren showed that a quasi-
uniformity is bicomplete if and only if each linked Cauchy filter pair (pairwise)
converges.
In [63] Brmmer and Knzi noted that for a totally bounded quasi-uniform
(To-)space (X, OU) the topological space (X, r(OlL is a locally compact space
in which the convergence set of any ultrafilter is the c10sure of some (unique)
singleton. (Such spaces are called locally compact and strongl~sober. They will
be discussed more thoroughly in Section 11.) Furthermore OU is the coarsest
quasi-uniformity that the topological space (X, g-(0lL admits, and the uniform
topology r (OlL *) is compact.
It follows that the b-c1osure S of a totally bounded quasi-uniform space
(X, OU) in (i, O) is homeomorphic to the sobrification ofthe space (X, r(OU.
They also proved that S is equal to i if and only if (X, r(OU is a core-compact
NONSYMMETRIC DISTANCES AND THEIR ASSOCIATED TOPOLOGIES 879
compact, while there are precompact D-complete quasi-uniform spaces that are
not compact. The concept of a D-Cauchy filter was also considered by Balaguer
[27] in bis study of the notion of a compressed filter for quasi-proximity struc-
tures. He observed that a quasi-uniform space (X, OU) is Cauchy bounded if and
only if the set of D-Cauchy filters coincides with the set of filters compressed
with respect to the induced quasi-proximity 8ou.
A quasi-uniform space (X, OU) is called strongly-D-complete provided that
each co-D-Cauchy filter has a -r(OU)-c1uster point (see [295], [180]). As the
name suggests, each strongly D-complete quasi-uniform space is D-complete.
From the definition given above for filters it should be c1ear what is meant
by a Cauchy pair 0/ nets in a quasi-uniform space. In fact Doitchinov's whole
theory can easily be based on Cauchy pairs of nets, as he did in [136]. A quasi-
pseudometric space will be called d-complete provided that each D-Cauchy
sequence converges.
Doitchinov observed that in order to obtain a reasonable completion theory it
is necessary to put restrictions on the c1ass of quasi-uniform spaces considered.
First he developed a conjugate invariant completion theory for his so-called bal-
anced quasi-metric spaces. A quasi-metric space (X, d) is called balanced [134]
provided that for each Cauchy pair (Xk, Yn) of sequences and all x, y E X we
have that d (x, y) ~ sUPnEw d (x, Yn) + SUPkEW d (Xk, y). The usual quasi-metric
on the Sorgenfrey line is balanced and d-complete. A d-complete subspace of
a balanced d-complete quasi-metric space need not be c1osed.
He showed that each balanced quasi-metric space (X, d) has a standard
(necessarily unique up to quasi-isometries) balanced d-completion (X+, d+)
(in the sense that X is quasi-isometrically embedded as a dense subspace in
X+ and that each quasi-uniformly continuous map / : (X, d) ~ (Y, c) into
a d-complete balanced quasi-metric space (Y, c) can be uniquely extended to
a quasi-uniformly continuous map on X+). Furthermore the completion of the
conjugate of a balanced quasi-metric space can be identified with the conjugate
of its completion. For metric spaces the construction yields the usual metric
completion.
Then Doitchinov went on to extend his theory to quasi-uniform spaces by
introducing the concept of a quiet quasi-uniformity. Essentially the same concept
was independently found by Deak [117] using the idea of a weakly concentrated
Cauchy filter pair. Deak called a Cauchy filter pair (~, cg) on a quasi-uniform
space weakly concentrated if for any entourage U there exists an entourage V
such that V-1(y) E ~ and V(x) E cg imply that (x, y) E U. He showed that
each Cauchy filter pair that it is weakly concentrated contains a coarsest one
among the Cauchy filter pairs coarser than it.
He also said that a family of Cauchy filter pairs is uniformly weakly concen-
trated if the above condition holds for each filter pair, with V depending only on
U, but not on the filter pair. Using this terminology, a quasi-uniformity is called
NONSYMMETRIC DISTANCES AND THEIR ASSOCIATED TOPOLOGIES 881
quiet if the collection of its Cauchy filter pairs is uniformly weakly concen-
trated [135]. Quietness is a productive and hereditary property of quasi-uniform
spaces.
The importance of quiet quasi-uniformities stems from the fact that they
possess a satisfactory completion theory. Doitchinov [135], [136] showed that
each quiet quasi-uniform To-space (X, OU) has a standard D-completion, which
is now called its Doitchinov completion, i.e. there is an (up to isomorphism)
uniquely determined quiet D-complete quasi-uniform To-space (X+, OU+) into
which (X, OU) is embedded as a dense subspace and onto which each quasi-
uniformly continuous map f : (X, OU) ~ (Y, "V) into an arbitrary quiet D-
complete quasi-uniform To-space (Y, "V) can be uniquely extended such that the
extension is quasi-uniformly continuous.
The Doitchinov completion of the conjugate of a quiet quasi-uniform To-
space can be identified with the conjugate of its Doitchinov completion. Hence
the conjugate of a D-complete quiet quasi-uniform space is quiet and D-
complete. For uniform To-spaces the construction of the Doitchinov completion
coincides with the usual completion.
The induced quasi-uniformity OUd of any balanced quasi-metric space (X, d)
is quiet. If (X, d) is a quasi-metric space such that OUd is quiet, then (X, d) is
d-complete if and only if OUd is D-complete.
In [126] Dek showed that for a quiet quasi-uniform To-space X, up to
quasi-uniform isomorphism, X+ is the only D-complete quiet quasi-uniform
To-space in which X is doubly dense.
He extended parts of the theory of quiet quasi-uniform spaces to two larger
c1asses of quasi-uniform spaces that he had called weakly quiet and subquiet
[126], although at cost of losing some good properties of Doitchinov's original
construction. For instance he gave an example of a tota1ly bounded subquiet
quasi-uniformity that is not a uniformity, while each quiet totally bounded quasi-
uniform space is known to be uniform [182], [314]. Interestingly, the original
proof of the latter result due to Fletcher and Hunsaker is based on the Doitchinov
completion.
A discussion of some aspects of Doitchinov's completion theory from a
nonstandard point of view was given by Render [479]. Among other things he
observed that a totally bounded quasi-uniform space has a uniformly regular
D-completion if and only if it is uniform.
More generally, in [127] Dek looked for quasi-uniform completeness prop-
erties and corresponding completions that are symmetric in the following sense:
(i) OU is complete if and only if OU -1 is so; and () the completion of OU- 1
is isomorphic to the conjugate of the completion of OU. Several bitopological
notions of quasi-uniform completeness were analysed by him. For example
he showed that each quasi-uniformity has an St-completion where a quasi-
uniformity is called St-complete provided that each stable Cauchy filter pair
882 HANS-PETER A. KNZI
converges. (In order to avoid a dash with the concept of Smyth completeness
we have slightly modified his terminology.)
Although the dass of quiet quasi-uniform spaces has numerous nice prop-
erties, many quasi-uniformities encountered in practical applications lead to
nonregular spaces and, since each quiet quasi-uniformity induces aregular topol-
ogy [136], cannot be quiet. For this reason, other dasses of quasi-uniform spaces
that have canonical D-complete extensions in which a copy of the original space
is densely embedded have been investigated.
In his studies on D-completions [103], [104] Csaszar showed among other
things that every quasi-uniform space has a D-complete uniformly loose exten-
sion.
With the help of the envelopes of certain filters, Doitchinov [139] built a
D-complete extension for an arbitrary stable quasi-uniform To-space, which he
called its (standard) E-completion (envelope completion).
Let us recall that for a filter cg;; on a quasi-uniform space (X, CU) the filter
generated by {U(F) : F E cg;;, U E CU} is said to be the envelope or roundifica-
tion of cg;;. A filter that coincides with its envelope is called round (see e.g. [525],
[101]). A filter cg;; on a quasi-uniform space (X, CU) is called stable provided that
nFe'!fU (F) E cg;; whenever U E CU. In the theory of quasi-uniform spaces the
latter concept was introduced by Csaszar [101].
In [139] Doitchinov said that a quasi-uniformity is stable provided that each
D-Cauchy filter is stable. Obviously, all uniformities are stable.
Note that neighborhood filters of points are stable D-Cauchy filters. It is
readily verified that the (nonempty) intersection of a family of stable filters is
stable; moreover a filter is stable if and only if its envelope is stable [117].
There exist quiet quasi-uniform spaces that are not stable, as well as sta-
ble quasi-uniform spaces that are not quiet. The standard E-completion of a
stable quasi-uniform To-space need not be stable. Furthermore the standard E-
completion of a quiet stable quasi-uniform To-space can be different from its
Doitchinov completion, although both extensions yield the usual completion for
uniform spaces ([127], [139]).
Subsequently, Deak [127] extended Doitchinov's construction of the E-
completion of a stable quasi-uniform space to substable quasi-uniform To-
spaces. (A quasi-uniform space is called substable if every D-Cauchy filter
is finer than some stable D-Cauchy filter.)
Some basic properties of stable quasi-uniformities were exhibited by Jun-
nila and Knzi in [268]: There exist stable quasi-uniformities which are not
the supremum of stable quasi-pseudometric quasi-uniformities. Stability is pre-
served by initial constructions such as subspaces and products. The conjugate of
a quasi-uniformity is hereditarily precompact if and only if each (ultra)filter is
stable. (That result was independently obtained by Deak [127].) Consequently a
quasi-uniform space whose conjugate is hereditarily precompact is stable. Fur-
NONSYMMETRIC DISTANCES AND THEIR ASSOCIATED TOPOLOGIES 883
OlL has a compatible extension to the whole space if and only if the trace filters
are stable. In this case such an extension is unique.
In [367] Losonczi conducted a thorough study of finite (dense) extensions
of quasi-uniformities for prescribed topologies. In particular he investigated the
lattice (or semilattice) of compatible extensions. In [101] Csaszar had shown
for strict extensions that if Y \ X is finite then it is necessary and sufficient for
the existence of an extension of OlL to ~ that the trace filters are round and tarne.
(A filter rg; on a quasi-uniform space (X, OlL) is called tarne if for any U E OlL
there is S E rg; such that S ~ U (x) whenever rg; converges to x in the topology
induced by OlL.) In [367, Corollary 4.16] Losonczi obtained a similar result for
arbitrary finite extensions.
The problem on extensions has further variants, some of which we consider
next. Given a quasi-uniformity OlL and a (doubly dense) extension ofthe induced
bitopology, Dek [117] looked for conditions guaranteeing that there is an
extension of OlL compatible with the bitopological extension. Most of his results
were formulated in terms of trace filter pairs. So he showed that there exists an
extension of a quasi-uniformity compatible with a given system of trace filter
pairs if and only if each trace filter pair is round and Cauchy.
In [121] he established that if a quasi-uniformity can be extended to an
extension of the induced topology (bitopology), then it can be extended to any
finer topological (completely regular bitopological) extension belonging to the
same system of trace filters (filter pairs). The problem of extensions with small
quasi-uniform weight was also considered.
He also noted [117] (compare [105]) that a supdense extension compatible
with given trace filter pairs exists if and only if they are linked, round and
Cauchy. Such an extension is necessarily unique.
Furthermore he proved in [129] that there exists a (so-called) doubly uni-
formly strict extension of a quasi-uniformity for prescribed trace filter pairs
if and only if they are uniformly weakly concentrated and minimal Cauchy.
Similarly, he showed that there exists a doubly uniformly regular extension for
prescribed trace filter pairs if and only if they are minimal Cauchy and the
ultrafilter pairs finer than trace filter pairs are uniformly weakly concentrated.
In [106] Csaszar investigated quasi-uniform extensions for prescribed traces
of entourages. In particular, uniformly strict extensions were considered.
The general problem of simultaneous quasi-uniform extensions in a topolog-
ical space is the following: In a topological space (X, 5") there are given subsets
Xi ~ X (i Elf. 0) and on each Xi, a (totally bounded) quasi-uniformity OlLi.
The problem is to find necessary and sufficient conditions under which there
exists a (totally bounded) quasi-uniformity OlL on X such that r(OlL) = 5" and
the restriction OlLlXi of OlL to Xi coincides with OlLi (i E I) (see e.g. [112]). The
special case of finite partitions was dealt with by Csaszar in [107]. His work
was continued by Losonczi [368].
886 HANS-PETER A. KNZI
to help the reader to understand better some of the results mentioned in other
seetions.
The following two c1asses of quasi-uniformities [415, p. 37 and p. 41] are
dealt with in the book by Murdeshwar and Naimpally in eonnection with their
diseussion ofthe Ro resp. regularity separation axiom (see also [424]): A quasi-
uniformity OlL is ealled point-symmetrie if r(OlL) ~ r(OlL- i ). It is said to be
loeally symmetrie if for eaeh U E OlL and x E X there is V E OlL sueh that
V-i V(x) ~ U(x). Murdeshwar and Naimpally proved that a topological spaee
is an Ro- (resp. regular) spaee if and only if it admits a point-symmetrie (resp.
loeally symmetrie) quasi-uniformity. In faet for Ro- (resp. regular) spaees the
Pervin quasi-uniformity has the eorresponding property.
The eoneept of point-symmetry was named by Salbany [518]. Point-
symmetrie quasi-uniformities were for instanee also studied by Carlson [79]
(under the name "Ioeally right symmetrie"). He showed that this property
eharaeterizes those quasi-uniform spaees for whieh the eolleetion of quasi-
uniform eoverings forms a nearness strueture with the same c10sure operator.
In [556] Stoltenberg ealled a quasi-metrie strong provided that it induees a
point-symmetrie quasi-uniformity.
The eonjugate of a point-symmetrie quasi-uniformity is now often ealled
small-set symmetrie. This eoneept was originally defined by Fleteher and Hun-
saker [179]. It is equivalent to the eondition that for eaeh open set G and
eaeh entourage V, V(G) eontains the c10sure of G (see [181]). The given
eharaeterization was obtained subsequently in [333]. The eoneept seems to be
very useful in the theory of funetion spaees (see Seetion 9).
A quasi-uniformity OlL is ealled equinormal [195] if for eaeh pair of c10sed
disjoint sets A and B there is U E OlL sueh that U (A) n B = 0. The study of
these quasi-uniform spaees began in a paper by Fleteher and Lindgren [195]
were the following results were established: Every topologie al spaee admits
an equinormal quasi-uniformity, namely the Pervin quasi-uniformity. A regular
spaee admits an equinormal quasi-metrie if and only if it admits an equinormal
metrie. A quasi-uniformity with the Lebesgue property is equinormal and any
equinormal quasi-uniformity eompatible with a regular (resp. Ro-) topology is
loeally symmetrie (resp. point-symmetrie).
A quasi-uniformity OlL on set X is ealled Lebesgue provided that for eaeh
open eover 6 of X there is an entourage V sueh that {V (x) : x E X} refines
6. The eoneept of a Lebesgue quasi-uniformity was diseussed by Fleteher and
Lindgren in [188], [197]. They showed for instanee that eaeh eompaet quasi-
uniform spaee satisfies it, that eaeh Lebesgue quasi-uniformity is eonvergenee
eomplete and that a topologieal spaee is orthoeompaet if and only if each open
eover is Lebesgue with respeet to the fine transitive quasi-uniformity. Mrsevic,
Reilly and Vamanamurthy [413] devoted a paper to the Lebesgue property,
888 HANS-PETER A. KNZI
(resp. right K -Cauchy) provided that for any entourage U E Oll, there is d E D
such that d2, dl E D and d2 :::: dl :::: d imply that (Xd" Xd2) E U (resp.
(Xd2' Xd,) E U). In papers dealing with problems in theoretical computer science
the terms forward resp. backward Cauchy nets are usually used for such nets.
Some categorical motivation for the concept of a left K -Cauchy sequence
was recently provided by Bentley and Hunsaker [39]. Unfortunately there are
convergent sequences in quasi-pseudometric spaces that are not left K -Cauchy.
In fact, a regular quasi-metric space in which each convergent sequence has a
left K -Cauchy subsequence is metrizable [333].
The corresponding concept for filters was introduced and mainly studied by
Romaguera [492], [496]: A filter c:; on a quasi-uniform space (X, OU) is said to
be left (resp. right) K -Cauchy provided that for each U E OU there is an FE'?;
such that U(x) E '?; (resp. U-l(x) E '?;) whenever x E F. A quasi-uniformity
is called left (right) K-complete provided that each left (right) K-Cauchy filter
converges.
Some basic results on these concepts are the following: Each left (resp.
right) K -Cauchy filter converges to its cluster points. Any right K -Cauchy filter
is stable. An ultrafilter is right K -Cauchy if and only if it is stable. A filter is
stable and left K -Cauchy if and only if it is Cauchy with respect to the supremum
uniformity. Using a technique due to Snderhauf, Knzi [322] showed that left
K -completeness defined by filters is equivalent to left K -completeness defined
by nets. In [338] he and Romaguera proved that the corresponding result also
holds for right K -completeness. Obviously, completeness in the sense of Pervin
and Sieber implies left K -completeness.
The behaviour of conditions from the area of K -completeness under various
kinds of maps was investigated by Knzi in [328]. For instance he showed that
the property that each co-stable filter clusters is preserved under uniformly
open continuous surjections between quasi-uniform spaces. (Here a filter was
called co-stable provided that it is stable in the conjugate space.) Note that the
latter property implies left K -completeness, which need not be preserved under
uniformly open continuous surjections.
Results of the Niemytzki-Tychonoff type have been established for many
completeness properties (compare also [343]). As noted above, Pervin and
Sieber [448] had shown that a topological space is compact if and only if each
compatible quasi-uniformity is complete (in their sense). Typical are also the
following results established by Alemany, Romaguera and Salbany: A quasi-
pseudometrizable space is countably compact if and only if every compatible
quasi-pseudometric is left K -complete (resp. right K -sequentially complete)
[502], [9]. A topological space is compact if and only if each compatible
quasi-uniformity is left (resp. right K -complete) [496]. On the other hand,
they observed that a quasi-metrizable space which admits only bicomplete
quasi-metrics is finite [503].
890 HANS-PETER A. KNZI
precompact if and only if each sequence has a left K -Cauchy subsequence (see
[333], [321]).
In [322] Knzi characterized precompactness of a quasi-uniform space by
the condition that each maximal co-round filter is left K -Cauchy. With the help
of that result he showed that a quasi-uniform space is compact if and only if it is
precompact and left K -complete. The analogous result for quasi-pseudometrics
can readily be obtained by an application of Knig's Lemma [333], [321].
Romaguera [496] showed that the conjugate of each uniformly regular left
K -complete quasi-uniform space is right K -complete and that a uniformly reg-
ular quasi-uniform space is Smyth complete if and only if it is left K -complete.
The property of right K -completeness has tumed out to be useful in hyper
and function spaces (see Section 9).
A filter cg; on a quasi-uniform space (X, OlL) is called S-Cauchy provided
that for any U E OlL and S E cg; there is x E S with U (x) E cg;. The name was
chosen in honour of Smyth. These filters were called hereditarily Cauchy by
Deale [128]. A quasi-uniformity is said to be Smyth complete provided that each
round S-Cauchy filter is a neighborhood filter of a point. (Originally Smyth
required that the neighborhood filter should be unique in order to obtain To-
spaces.) This concept will be discussed more carefu1ly in Section 11. It can be
verified that a topological space is quasi-sober if and only if it admits a Smyth
complete quasi-uniformity.
In [322] Knzi explained the c10se connections that hold between left K-
Cauchy filters and S-Cauchy filters: A filter cg; on a quasi-uniform space is an
S-Cauchy filter if and only if there exists a left K -Cauchy filter 'je finer than
cg; such that the envelope of 'je is coarser than cg;. It follows that each left K-
Cauchy filter on a quasi-uniform space converges if and only if each S-Cauchy
filter converges.
A quasi-uniformity OlL on a set X is called unijormly regular provided that
for any U E OlL there is V E OlL such that V{x) S; U{x) whenever x E X.
Originally the concept of uniform regularity was introduced by Csaszar when
studying extensions [100].
Since quiet quasi-uniform spaces are (doubly) uniformly regular and quiet
spaces attracted a lot of attention recently, the property of uniform regularity
was studied thoroughly during the last few years.
Fletcher and Hunsaker [178] proved that every point-symmetric uniformly
regular D-complete quasi-uniform space is quiet. Every continuous quasi-metric
induces a uniformly regular quasi-uniformity and every Lebesgue quasi-
uniformity that is uniformly regular is quiet. (In this artic1e we call a quasi-
metric d continuous on a set X provided that for each p E X the function
dp : X -+ R defined by dp{x) = d{p, x) is r{d)-continuous [556], [490]; note
however that other authors have used a different terminology.)
892 HANS-PETER A. KNZI
into details. For instance he called a quasi-uniformity fully Cauehy if for any
Cauchy filter pair there is a coarsest one among the Cauchy filter pairs coarser
than it; according to him a quasi-uniformity is said to be pointwise Cauehy if
any Cauchy filter pair whose first member is fixed has a coarsest one among the
Cauchy filter pairs coarser than it.
Among the results that he obtained let us mention the following ones [129]:
Each locally quiet quasi-uniformity is fully Cauchy. Each weakly co-regular
quasi-uniformity is pointwise Cauchy. Furthermore, each fully Cauchy quasi-
uniformity is Cauchy, where he called a quasi-uniformity Cauehy provided that
(~l n 'l, ~2 n '(2) is Cauchy whenever (~l, ~2) and ('l, '(2) are Cauchy filter
pairs that are linked (i.e. each member of ~i hits each member Of'i (i E {l, 2}).
He called a quasi-uniformity filter-symmetrie provided that (~l, ~2) is
a Cauchy filter pair whenever (~2, ~t> is a Cauchy filter pair. Any proxi-
mally symmetric quasi-uniformity is filter-symmetric. He verified that a quasi-
uniformity is filter-symmetric if and only if it is quiet and doubly co-stable;
any Cauchy bounded filter-symmetric quasi-uniformity was shown to be a uni-
formity. The latter result generalized the corresponding result of [333] about
Smyth symmetry.
Improving on results of Fletcher and Hunsaker [182] and Knzi [314], he
also established that any totally bounded Cauchy quasi-uniformity is a unifor-
mity; therefore in particular each quiet totally bounded quasi-uniformity is a
uniformity.
In [498] Romaguera considered a kind of (symmetrized) weakening of the
condition of quietness in quasi-uniform spaces, which he calledfitting and which
is preserved under conjugation and bicompletion. He showed that each fitting
totally bounded quasi-uniformity is a uniformity and that each fitting quasi-
uniform space satisfies a weakened (symmetrized) version ofuniformregularity
which he named unijormly Ro.
A quasi-uniformity Oll, on a set X is called monotonie [95] provided that
there exists an operator M : Oll, --+ Oll, such that M(U) ~ M(V) whenever
U, V E Oll, and U ~ V, and such that M(U)2 ~ U whenever U E Oll,. Obviously,
each quasi-uniformity with a base that is well-ordered by reverse (set-theoretic)
inc1usion is monotonic.
The concept of monotonicity for uniform structures is due to Gartside and
Moody [218], [95].1t came up in their study of monotonic topological properties.
They showed that a topological Tl-space admits a monotonic uniformity if and
only if it is monotonically normal and has an orthobase, i.e. it is protometrizable.
For such spaces the finest compatible uniformity is monotonic.
In [95] it was shown that the unique compatible quasi-uniformity of the
cofinite topology on an uncountable set is monotonic.
The concept of a monotonie quasi-unijormity was investigated to some
extent by Junnila and Knzi in their study of monotonically orthocompact
NONSYMMETRIC DISTANCES AND THEIR ASSOCIATED TOPOLOGIES 895
spaees [267]. Among other things they showed that the fine quasi-uniformity
of eaeh eountable first-eountable TI-spaee is monotonie and that the fine quasi-
uniformity of the reals, endowed with the usual topology, is not monotonie.
Furthermore they established that eaeh Tl -spaee with an orthobase admits a
monotonie quasi-uniformity and that there is a eountable spaee that does not
admit any monotonie quasi-uniformity. It still seems to be unknown whether
eaeh monotonieally orthoeompaet spaee admits a monotonie quasi-uniformity.
Lindgren [196] that the completely regular ordered spaces with total order are
exact1y the suborderable spaces.
In [44] Blatter extended results on algebras of real-valued continuous func-
tions due to Hewitt, which are related to the c1assical Stone-Weierstrass theorem,
to Nachbin's completely regular ordered spaces. He used his results to charac-
terize those completely regular ordered spaces that have a unique order com-
pactification. Subsequently Salbany [520] obtained a related characterization of
bitopological spaces that admit a unique quasi-uniformity.
Fletcher and Lindgren [196] used their theory of the bicompletion to advance
the theory of order compactifications and completions of completely regular
ordered spaces further. Their approach to constructing order completions of
completely regular ordered spaces with the help of the bicompletion of deter-
mining quasi-uniformities c1arified many of the results concerning Nachbin's
uniform ordered spaces obtained earlier (see also the papers by Redfield [466],
[467]). Their results were inc1uded in their book [197] where also various open
problems on topological ordered spaces were listed.
Some of them were studied in three papers by Knzi. In [309] he observed
that in a suborderable space the uniformity C(6(X) is convex if and only if each
c10sed discrete set is countable. (Recall that a quasi-uniformity Oll is called
convex provided that for each U E Oll there is V E Oll such that V ~ U and
V(x) is orderconvex whenever x EX. For instance, for any To-quasi-uniformity
Oll, the uniformity Oll * is convex with respect to nou. Also the fine uniformity of
any suborderable space is convex [196].) Moreover he proved that a completely
regular ordered space for which every compatible convex uniformity is totally
bounded, need not be pseudocompact.
In [318] he studied those topological ordered spaces X for which the set of
all neighborhoods of the partial order in the product topology forms a quasi-
uniformity that determines X. It was shown for instance that a topological
ordered space with a countably compact sequential topology satisfies the stated
condition if and only if the set of all neighborhoods of its diagonal is a uni-
formity. While each suborderable space has the property under consideration,
the real plane with its usual order and topology does not satisfy it. His in-
vestigations were motivated by the following problem posed by Fletcher and
Lindgren [197, p. 9]: For which quasi-uniform spaces (X, Oll) is the family of all
r(Oll* x Oll*)-neighborhoods of nou a quasi-uniformity compatible with r(Oll)?
In [319] he also obtained some characterization ofthose quasi-uniformities
Oll on a set X for which there does not exist a strictly coarser quasi-uniformity
on X that determines the same uniform ordered space as Oll. His characterization
sheds some light on the theory of minimal order compactifications of completely
regular ordered spaces.
In the spirit of Taimanov's c1assical theorem Hunsaker [257] obtained an ex-
tension theorem for continuous increasing mappings. He deduced that the quasi-
898 HANS-PETER A. KNZI
very slowly. Recently, however the interest in quasi-uniform function and hy-
perspaces increased considerably. Partially this fact can be explained by their
applications in theoretical computer science (see e.g. Snderhauf's thesis [558]
or Section 11).
The research dealing with quasi-uniformities on function spaces was initi-
ated by a paper of Naimpally [423] and investigations due to Fletcher and his
collaborators Liu and Seyedin (see e.g. [171], [174], [199], [535], [536], [537]),
who were mainly interested in applications of quasi-uniformities to the study of
topological homeomorphism groups.
Naimpally [423] showed that if Xis a topological space, (Y, 'V) is a quasi-
uniform space and ~(X, Y) is the family of continuous maps from X to (Y,
r ('V equipped with the quasi-uniformity 9Lv of quasi-uniform convergence,
then the evaluation map e : ~(X, Y) x X -+ Y defined by e(g, x) = g(x) is con-
tinuous, i.e. (~(X, y), 9Lv) is jointly continuous. Furthermore with Murdeshwar
he established that the topology induced on ~(X, Y) by the quasi-uniformity of
compact convergence is finer than the compact-open topology [414, Theorem
7.5].
In [171] Fletcher considered the full homeomorphism group H(X) of a so-
called representable (= strongly locally homogeneous [28]) topological space
X equipped with an appropriate topology of quasi-uniform convergence and in
[174] he observed that if (X, dU) is a quasi-uniform space and Ga subgroup of
H (X) such that each member of G is dU-quasi-uniformly continuous, then G is
a paratopological group (see Section 12) provided that it carries the topology of
quasi-uniform convergence with respect to dU. In [199, Example 3.2] Fletcher
and Liu (compare [324]) gave an example showing that in general G is not a
topological group.
Seyedin [535] proved that for an Ro-space X equipped with a compatible
point-symmetric quasi-uniformity dU and for a subgroup G of H (X) such that
G is quasi-equicontinuous with respect to dU, G is a topological group under
the topology of pointwise convergence; for an arbitrary quasi-uniformity dU the
method yields a paratopological group.
Several results obtained by some of these authors were over-optimistic and
had to be corrected later (e.g. by Cao and Render [70], [481]). Their investiga-
tions were continued by Kouphos and Papadopoulos in [298]. These two authors
studied exponentiallaws for function spaces equipped with the quasi-uniformity
of quasi-uniform convergence.
Concepts of boundedness in connection with function spaces were investi-
gated by Papadopoulos [437], [439]. A subset A ofaquasi-uniform space (X, dU)
is called bounded if given any entourage V there are n and finite F ~ X such
that A ~ V n (F). In the theory of quasi-uniform spaces the concept ofbounded-
ness was introduced by Murdeshwar and Theckedath [416]. Some bitopological
NONSYMMETRIC DISTANCES AND THEIR ASSOCIATED TOPOLOGIES 901
uniform spaee X sueh that the set of all (eontinuous) funetions from X to
X is not (Pervin-Sieber) eomplete for the quasi-uniformity of quasi-uniform
eonvergenee.
Cao [71] (see also [338]) proved that if Xis a topological spaee and (Y, OU) is
a quiet (or small-set symmetrie) quasi-uniform space, then <f6(X, Y) need not be
c10sed in the set of all functions from X to Y equipped with the quasi-uniformity
of quasi-uniform convergence; however, it is known that <f6(X, Y) is c10sed in
that space y X provided that OU is locally symmetric [423], [438].
Recently, Porter continued the research on quasi-uniformities on homeo-
morphism groups with her investigations on variants of the open-open topology
[453], [454], [455], [456]. Among other things she showed that the open-open
topology on the group H (X) of all self-homeomorphisms of a topological space
X coincides with the topology of quasi-uniform convergence transmitted by the
Pervin quasi-uniformity of X. Fletcher [174] had characterized those topological
(Tl- )spaces for which the paratopological group H (X) is not discrete.
Porter also observed that for a semiregular space X, H(X) is a topological
group provided that it is equipped with the topology of quasi-uniform conver-
gence transmitted by the Pervin-type quasi-uniformity generated by the regular
open sets of X.
In his survey artic1e [324] Knzi continued her work by applying results
about function spaces to quasi-uniform isomorphism groups. Among other
things he observed that the group of all quasi-uniform isomorphisms of a
bicomplete quasi-uniform To-space equipped with the quasi-uniformity of quasi-
uniform eonvergenee yields a paratopologieal group whose two-sided quasi-
uniformity is bicomplete. He also showed that if X is a topological space
in which each c10sed set is the intersection of finitely many regular c10sed
sets (for instance, X is metrizable), c;P is the Pervin quasi-uniformity of X
and :iQl> is the quasi-uniformity of quasi-uniform convergence on H(X), then
(H (X), 0, 'l' (:iQT) is a topological group. He also argued that the same construc-
tion applied to the fine quasi-uniformity of a completely metrizable space yields
a topological group, while this is not the case for the fine quasi-uniformity of
the Sorgenfrey line.
Moreover [500] Romaguera and Ruiz-G6mez introduced and studied bitopo-
logical versions of theorems of pointwise, compaet and (quasi-)uniform compaet
convergence on spaces of continuous functions, extending several c1assical
theorems on function spaces to the bitopological case.
Similarly, the investigations on quasi-uniform hyperspaces started only
slowly. In the light of the celebrated investigations due to Michael [409] it
had already been observed by Levine and Stager [362] as weH as Berthiaume
[40] that for a quasi-uniform spaee X we can define on the set of (nonempty)
sub sets of X a Hausdorff (= Bourbaki) quasi-uniformity in a way analogous
to the construetion for uniform spaces. In those two papers the construction
NONSYMMETRIC DISTANCES AND THEIR ASSOCIATED TOPOLOGIES 903
was mainly studied for the case in which the underlying quasi-uniformity is
the Pervin quasi-uniformity of a topological space. In particular, Levine and
Stager proved that the Hausdorff quasi-uniformity of a Pervin quasi-uniformity
always induces the Vietoris topology. Berthiaume analysed the connections
between the topology induced by the upper (resp. lower) quasi-uniformity and
the upper (resp. lower) Vietoris topology further. He employed his results about
quasi-uniformities on hyperspaces to show that an infinite product of upper
semicontinuous multivalued mappings, with compact sets as values, is again
upper semicontinuous.
His investigations were continued by Francaviglia, Lechicki and Levi [210]
who used quasi-uniformities to define and study convergence of nets of semi-
continuous multifunctions. Among other things they observed that the bitopo-
logical space determined by the upper and lower Vietoris topology is quasi-
uniformizable if and only if the original space is normal. They also noted that
for a locally compact space X the bitopological hyperspace consisting of the
lower Vietoris topology and the so-called cocompact topology is (uniquely)
quasi-uniformizable. It is weIl known that the join of these two topologies,
which is usually called the Fell topology of X, is compact and Hausdorff for a
locally compact space X.
The study of the Hausdorff quasi-uniformity induced by arbitrary quasi-
uniformities was continued by Cao [69] and Knzi and Ryser [345]. In par-
ticular, Cao showed that the locally finite topology defined on the collection
of nonempty closed sets coincides with the topology of the Hausdorff quasi-
uniformity of the locally finite covering quasi-uniformity. Knzi and Ryser
observed that the Hausdorff quasi-uniformity of a compact (resp. hereditarily
precompact) quasi-uniform space need not be compact (resp. hereditarily pre-
compact), while it is precompact (resp. totally bounded, supcompact) if and
only if the underlying space is precompact (resp. totally bounded, supcompact).
Bicompleteness on the other hand (even for quasi-metric spaces) behaves rather
badly under the Hausdorff hyperspace construction. Furthermore extending the
Burdick-Isbell Theorem from uniform spaces to the quasi-uniform setting the
two authors established that the Hausdorff quasi-uniformity of a quasi-uniform
space (X, OU) is right K -complete if and only if each stable filter on (X, OU) has
a cluster point. They also observed that for a bounded quasi-pseudometric space
(X, d), its Hausdorff quasi-pseudometric d* is right K -sequentially complete if
and only if d is right K -sequentially complete. They verified that for instance
the point-finite open covering quasi-uniformity of a metacompact space has the
property that each stable filter has a cluster point.
Continuing this line of research, in [340] Knzi and Romaguera showed that
the Hausdorff quasi-uniformity of a quasi-uniform space (X, OU) is compact if
and only if (X, OU) is compact and (X m , OU-1IXm ) is hereditarily precompact,
where X m denotes the set of the minimal elements of X with respect to the
904 HANS-PETER A. KNZI
(specialization) preorder nou. They also noted that the problem to eharaeter-
ize hereditary preeompaetness of the Hausdorff quasi-uniformity is related to
classieal results on better quasi-ordering. Furthermore in [72], together with
Cao and Reilly, they investigated the relationship between the Bourbaki quasi-
uniformity and the Vietoris topology on the family 'j{o(X) of all nonempty
eompaet subsets of a quasi-uniform spaee (see also [509]). Extending clas-
sical results about uniform spaees due to Morita, they established that for
a so-ealled eompaet1y symmetrie quasi-uniform spaee (X, Oll) the Hausdorff
quasi-uniformity on 'j{o(X) is (pervin-Sieber) eomplete if and only if Oll is
eomplete, and that the Bourbaki quasi-uniformity induees the Vietoris topology.
Their eoneept of eompaet symmetry strengthens the well-known notion of loeal
symmetry, which is not sufficient to prove these results. They also noted that
a eompaet1y symmetrie quasi-uniform spaee indueing a k-topology is small-set
symmetrie.
Some partial results on left K -eompleteness of the Hausdorff quasi-
uniformity were obtained in [339] by Knzi and Romaguera. In partieular it
was proved that the Hausdorff quasi-uniformity of the well-monotone quasi-
uniformity of any topological spaee is left K -eomplete. It was also observed
that the eondition that eaeh eo-stable filter clusters is only neeessary, but not
sufficient for left K -eompleteness of the Hausdorff quasi-uniformity.
In [341] Knzi and Romaguera showed that for any small-set symmetrie
quasi-uniform spaee, uniform loeal eompaetness is preserved by the Hausdorff
quasi-uniformity restrieted to eompaet sets.
Hohl and Levi [253] diseussed hit-and-miss hyperspaee topologies in terms
of their upper and lower parts, foeusing among other things on their quasi-
uniformization by appropriate Urysohn families. The splitting up ofuniformities
on hyperspaees into upper and lower quasi-uniformities was also treated by
Pasquale [442].
The results of Knzi and Ryser cited above were used by Cao, Reilly and
Romaguera [75] to explore various kinds of eompleteness of the multifunction
space (equipped with its natural Hausdorff quasi-uniformity) in terms of suitable
properties of the range spaee. Also a classical result due to Hunsaker and
Naimpally about loeal eompaetness of the multifunetion spaee was extended
from the uniform to the quasi-uniform setting.
In [73] Cao and Reilly introdueed and investigated a eoneept of almost
quasi-uniform eonvergenee for multifunetions; in [76], in eollaboration with Va-
manamurthy, they eompared various kinds of eonvergenees for multifunetions.
Furthermore in [74] Cao and Reilly diseussed almost eontinuity for multifune-
tions between bitopological spaees. They also established bitopological ana-
logues of closed graph and open mapping theorems for multihomomorphisms
between paratopological groups.
NONSYMMETRIC DISTANCES AND THEIR ASSOCIATED TOPOLOGIES 905
Using similar methods, in [281] Khanh had presented general open mapping
theorems for families of multifunctions in his so-called directly complete quasi-
metric spaces. (In our terminology "directly complete" means that the conjugate
is right K -sequentially complete.) His theorems generalize many known open
mapping theorems, c10sed graph theorems, theorems of the Lustemik type,
subtraction theorems, and theorems on approximation and semicontinuity (see
also [133, p. 20]).
Chou and Penot discussed criteria for uniform openness of multifunctions
betweeen quasi-uniform spaces [90]. To this end they introduced a concept of
convergence for infinite products of relations.
Related to these investigations, in [328] Knzi considered asymmetric ver-
sions of the fact that each almost uniformly open (multivalued) map with c10sed
graph from a supercomplete uniform space into an arbitrary uniform space is
uniformly open.
Rodriguez-L6pez and Romaguera [486] compared some graph topologies
with the topology of quasi-uniform convergence on the set of lower semi-
continuous real-valued functions of a quasi-pseudometric space. In this way
they obtained several analogues of known results about the set of continuous
real-valued functions on a metric space.
To show that Y is paracompact, take any open cover. Then there is a locally
finite refinement that covers Y x (R \ {rl , r2, ... }) where rl, r2, ... E AU B.
We can add a discrete family of open sets that will cover the rest of Y x A
together with a dopen subset of Y x B. This leaves a dopen sub set of Y x B,
which we can then cover using the paracompactness of Y.
908 HANS-PETER A. KNZI
n=k
In [95] it was shown that for each infinite cardinal there exists a y-space
(resp. quasi-metrizable space) X such that q(X) > (resp. tq(X) > )
where q(X) (resp. tq(X is the minimal cardinality of a base of a compatible
quasi-uniformity (resp. transitive quasi-uniformity) on X.
Let us mention that numerous other quasi-metrization theorems have been
obtained, e.g. by Romaguera and his collaborators. In the following we mention
just a few:
In [489] Romaguera gave a characterization of strongly quasi-metrizable
spaces in terms of a collection of pairs of real-valued functions in the style of a
characterization of developable spaces due to Brandenburg.
In [490] he showed that a topological space (X, ;?T) is (continuously) quasi-
metrizable if and only if there exists a family ?:F = U~o?:Fn of functions from
X into the unit interval [0, 1] such that
(B) for each n E w and ?:F' ~ ?:Fn, inf?:F' is lower semicontinuous (resp.
continuous) and
(e) {f-I(E, 1] : E > 0, f E ?:F} is a base for;?T.
In joint work with Antonino [16] he obtained a related characterization of
quasi-metrizability by equinormal quasi-metrics.
A different kind of characterization of quasi-pseudometrizability was given
by Vitolo [573]. He showed that a To-space is quasi-pseudometrizable if and only
if it is topologically embeddable into the lower Hausdorff quasi-pseudometric
hyperspace of a metric space restricted to the nonempty c10sed subsets.
It should finally be mentioned, however, that many people working in the
area of quasi-metrizable spaces think that the problem of finding a striking
topological characterization of this c1ass of spaces is still open.
Results leading toward a characterization of quasi-metrizability for bitopo-
logical spaces began in the early sixties when the first papers by Kelly [277]
and Lane [358] were published. Kelly showed that a bitopological space is
quasi-pseudometrizable if it is pairwise regular and both topologies are second
countable. Improving on a result of Lane [358] and confirming a conjecture of
Patty [441], Salbany [517] subsequently showed that a pairwise regular bispace
(X, qp, 21) is quasi-pseudometrizable provided that qp has a a - 21-Iocally finite
base and 21 has a a - qp-Iocally finite base. A related result was also obtained
by Romaguera [488]. These conditions however are all not necessary.
The first satisfactory general solution to the bitopological quasi-pseudo-
metrization problem was given by Fox [207], but unfortunately never published.
He showed that a bitopological space (X, qp, 21) is quasi-pseudometrizable if and
only if qp has both a a-qp-cocushioned pairbase and a a-21-cushioned pairbase
and 21 has both a a-21-cocushioned pairbase and a a-qp-cushioned pairbase. An
essentially equivalent formulation of his result is the following: A pairwise TI-
space (X, ;?T, :1) is quasi-metrizable if and only if it is pairwise stratifiable and
910 HANS-PETER A. KNZI
both (X,5") and (X, '[f) are y-spaces. An elegant variant of Fox's ideas was
later described by Kopperman [296] using the concept of a dual.
Again, many more quasi-metrization theorems for bitopological spaces can
be found in the literature, e.g. the ones by Pareek [440] and Raghavan and
Reilly [459], [465], as weIl as some others by Romaguera. For instance, in
the spirit of a metrization theorem due to Guthrle and Henry, in [487] Roma-
guera proved that a bitopological space is quasi-pseudometrizable if and only
if it is the initial bispace induced by a a-pairwise relatively complete family
of real-valued functions, where pairwise relatively complete means pairwise
semi-equicontinuous and punctually bounded. Some related characterizations
of non-archimedeanly quasi-metrizable bitopological spaces were obtained by
Gutierrez and Romaguera in [228]. Fox had noted [207] that a bitopological
space (X, Cl},~) is non-archimedeanly quasi-pseudometrizable if and only if ~
has a base consisting of Cl}-c1osed sets that is the countable union of ~-interior
preserving and Cl}-c1osure preserving collections and Cl} has a base consisting of
~-c1osed sets that is the countable union of Cl}-interior preserving and ~-c1osure
preserving collections.
In [493] Romaguera presented some necessary and sufficient condition for
quasi-pseudometrization of bitopological spaces in terms of g-functions by
generalizing a corresponding metrization theorem due to Nagata.
In the spirit of Katetov's c1assical theorem Romaguera and Salbany [501]
showed that a (pairwise) Hausdorff (pairwise) compact bispace (X, Cl},~) is
quasi-metrizable if and only if the bispace (X x X x X, Cl} x ~ x Cl}, ~ x Cl} x
52) is hereditarily (pairwise) normal. Similarly, they generalized the c1assical
theorem of Sneider that every Hausdorff compact space with a Go-diagonal
is metrizable to bitopological spaces. In [403] Marin and Romaguera obtained
some quasi-metrization theorems for pairwise monotonically normal bispaces.
In [523] Salbany called a 5"1 X 5"2-open cover C(& of the diagonal of a
bispace (X, 5"1, 5"2) even ifthere is a 5"1 x 5"2-neighborhood V ofthe diagonal
such that {V- I (x) x V(x) : x E X} refines C(&. In his paper he gave a necessary
and sufficient condition for quasi-pseudometrizability of those bispaces for
which every 5" 1 x 5"2 -open cover of is even. Recently [499] Romaguera
discussed the connections between Salbany's notion of evenness and some
concept of a pairwise Lebesgue quasi-uniformity. Furthermore Brown [53]
obtained a bitopological version of Arhangel'skiI's theorem that a Hausdorff
space is metrizable if and only if it is fully normal and has a base of countable
order.
Nevertheless it is important to realize that - although the search has gone on
for decades - no reasonable property of bitopological quasi-pseudometrizable
spaces has been found that could substitute for paracompactness, which has
tumed out to be so useful in the area of metric spaces.
NONSYMMETRIC DISTANCES AND THEIR ASSOCIATED TOPOLOGIES 911
For some early attempts to find such a concept due to Fletcher, Hoyle III
and Patty as well as Datta, we refer the reader to [177], [115], [292]. Cooke and
Reilly [97] presented a survey of some related concepts of pairwise compactness
discussed in the literature.
Extending a characterization of paracompactness due to Junnila to the bi-
topological setting, Marin and Romaguera [402] gave a definition of pairwise
paracompactness that allowed them to prove that a Tl-bitopological space is
quasi-metrizable if and only if it is pairwise developable and pairwise para-
compact. Generalizing that concept of pairwise paracompactness Romaguera
[499] introduced the concept of a (pairwise) almost 2-fully normal bispace: A
bispace (X, CZP,~) is called almost 2-fully normal provided that the set of all
~ x CZP-neighborhoods of the diagonal in X x X forms a quasi-uniformity on
X compatible with (CZP, ~). Among other things, he showed that abispace is
quasi-pseudometrizable if and only if it is pairwise developable and (pairwise)
almost 2-fully normal.
In [52], [215] Brown resp. Ganster and Reilly compared the concept of
pairwise paracompactness due to Marin and Romaguera with related notions,
e.g. the 8-pairwise paracompact spaces due to Raghavan and Reilly [464] or the
concepts of (semi-)sequential normality introduced by Brown [52]. (A bitopo-
logical space (X, 5"1, 5"2) is called 8-pairwise paracompact if every 5"i open
cover admits a 5"1 v 5"2-open refinement which is 5"1 v 5"2 locally finite for
i = 1,2.)
Certainly the following unpublished example [203] due to Fox will allow
the reader to understand better the difficulties involved in that matter.
Part of a letter jrom Fox to Salbany, dated June 19, 1979.
It seems impossible to prove the latter theorem (namely, a stratifiable devel-
opable space is metrizable) above without intimate use ofparacompactness,
and it would appear that to find a bitopological quasi-metrization analogue
one would need a satisfactory concept of paracompactness for bitopological
spaces. We are in fact going to present a counterexample to the obvious
analogue, namely a pairwise stratifiable, pairwise developable bitopological
space which is not quasi-metrizable.
Let (X, CZP,~) be a bitopological space. We say that CZP is developable with
respect to ~ if there exists a countable family {Dn : n E N} of ~ x CZP-
neighborhoods of the diagonal tJ. in X x X such that for each point x E X
the family {Dn(x) : n E N} is a CZP-neighborhood base at x. Such a family
{Dn : n E N} we call a development of CZP with respect to ~. If CZP is
developable with respect to ~, and ~ is developable with respect to CZP, we
say that (X, CZP,~) is pairwise developable.
With this definition, ... quasi-metrizable bitopological spaces are in fact
pairwise developable.
912 HANS-PETER A. KNZI
computer science have established some kind of fixed point theorems (see e.g.
the papers by Matthews resp. Flagg and Kopperman [407], [162]). Several
papers are entirely devoted to fixed point theorems, e.g. the ones by Harder
and Hicks [233], [243] as weIl as the ones by Reilly and Subrahmanyam [472],
[475]. A fixed point theorem for functions in a quasi-uniform space which
generalizes the Banach contraction principle was for instance given by Morales
[412].
Some contributions to the work of Hicks, in particular related to his version
of Caristi's theorem were subsequently made by Ciric [93], [94], Checa and
Romaguera [85], Jachymski [262] and Rhoades [483].
Moreover Kada, Suzuki and Takahashi [272] introduced and studied some
special kind of asymmetric distance function in metric spaces that they called
w-distance to improve various classical fixed point theorems. In [25] Baisnab
and Jas extended the concept of positive definiteness, defined by Dugundji for
metric spaces, to quasi-proximity spaces in order to prove some fixed point
theorem.
Semi-Lipschitz functions in quasi-metric spaces were defined and studied by
Romaguera and Sanchis [507] and then applied to problems ofbest approxima-
tion. (A function f : X ~ R from a quasi-metric space (X, d) into the reals is
said to be semi-Lipschitz ifthere exists k ~ 0 such that f(x) - f(y) ::: kd(x, y)
whenever x, y EX.)
We conclude this section with a discussion of applications of quasi-metrics
in other areas of mathematics (see also Sections 11 and 12). In [263] Jawhari,
Misane and Pouzet investigated in depth graphs and ordered sets as a kind
of quasi-metric space where - instead of real numbers - the values of the
distance function belong to an ordered semigroup equipped with an involution.
It was shown that many known results on retractions and fixed point property
for classical metric spaces extend to these spaces. The authors concluded that
their approach supports the idea that certain concepts of infinistic nature, like
those which inspired metric spaces, can perfectly apply to the study of discrete,
or even finite structures.
Cunningham [114] discussed the concept of nonymmetric distance in math-
ematical psychology with the help of bidirectional trees. Numerical methods
were presented that can be used to derive a representation for any given set of
dissimilarities.
An application ofnonsymmetric distance functions to questions ofbiomath-
ematics was outlined by Beyer, Smith and Waterman [41]. These authors used
quasi-pseudometrics to measure the distance between biological sequences,
such as amino acid sequences or nucleotide sequences and developed various
algorithms to compute those distances.
920 HANS-PETER A. KNZI
of uniform approximation ensures that such spaces are compact ordered in the
sense ofNachbin. The canonical quasi-uniformity ofthese (so-called uniformly
approximated) spaces X is generated by functions in the sense that there exists
a base consisting ofhypergraphs of continuous functions, where the hypergraph
of a function f : X --+ X is defined by {(x, y) : f (x) E {y}}.
The category of totally bounded bicomplete quasi-uniform spaces also yields
the starting point of Snderhauf's construction of a function space constructor
yielding a symmetric monoidal c10sed category [560]. Since, for the sake of
completeness, the author had to consider multivalued functions, also a suit-
able powerspace functor was defined. His powerspace construction (related to
Hausdorff's construction) inc1udes some well-known constructions like the
Smyth, Hoare and Plotkin powerdomains.
As an application, he gave a convenient model for the real numbers in
that category. In [564] he generalized the construction to the wider c1ass of
uniformly locally totally bounded quasi-uniform spaces, which is contained in
the c1ass of all S-completable spaces (see the following paragraph). In this
artic1e a refined function space construction was introduced. His basic idea is
that quasi-uniformities, being a common refinement of uniformities and partial
orders, may serve as a useful tool to introduce the notion of quantity into domain
theory.
Another important idea due to Smyth [550] was the suggestion to endow a
quasi-uniform space (X, OlL) with a topology r:J that is not necessarily its standard
topology r(OlL), but is linked to its quasi-uniform structure by some additional
axioms. Any topological quasi-uniform space (X, OlL, r:J) satisfies r:J ~ r(OlL);
furthermore nOlL coincides with the specialization preorder of r:J.
Moreover Smyth defined an appropriate concept of a Cauchy filter (now
often called S-Cauchy filter) on a topological quasi-uniform space and called
a topological quasi-uniform space complete if every round Cauchy filter is the
r:J -neighborhood filter of a (unique) point. Let us note that for topological quasi-
uniform spaces carrying the standard topology the concepts of S-Cauchy filter
and Smyth completeness (= S-completeness) reduce to the concepts considered
in Section 7. Smyth formulated much of the material of his paper in terms of
the syntopological spaces due to Csaszar.
Correcting an error in a preliminary version of Smyth' s artic1e, Snderhauf
[557] verified that by choosing an appropriate set ofaxioms one can indeed
obtain a categorically nice completion theory in Smyth's sense, which extends
the c1assical completion theory for uniform spaces.
Furthermore he called a quasi-uniform space equipped with its standard
topology S-completable (i.e. Smyth-completable) if its Smyth completion also
carries the standard topology and proved that a quasi-uniformity is S-
completable if and only if each (round) S-Cauchy filter is stable. For Smyth-
completable spaces the construction of the Smyth completion coincides with
924 HANS-PETER A. KNZI
topological convergence, the usual filter-net translation does not work in this
case and it is not a trivial task to derive such characterizations.
If we equip an arbitrary topological space with the weIl-monotone quasi-
uniformity (and its standard topology), then we obtain the sobrification equipped
with its weIl-monotone quasi-uniformity as its S-completion. Developing this
result, essentially due to Ferrario and Knzi [150], [322], in ajoint paper Flagg,
Kopperman and Snderhauf [165] introduced a functor w from the category
of topological quasi-uniform To-spaces and continuous, quasi-uniformly con-
tinuous maps to the category of quasi-uniform To-spaces and quasi-uniformly
continuous maps that equips the underlying set X of a topological quasi-uniform
space (X, OU, '!J) with some kind of weIl-monotone quasi-uniformity OU~ derived
from OU and '!J and acts as identity on morphisms. They showed that t' (OU~) = '!J
and t' OU~) -1) = t' (OU -1). Furthermore, they proved that for any topological
quasi-uniform space (X, OU, '!J) with Smyth completion (X, au, ~) the roundifi-
cation Rd of filters yields a quasi-uniform isomorphism from the bicompletion
(X, aU;) of (X, OU~) to (X, au'). Indeed, they verified that Rd is a natural
isomorphism b 0 W --+ W 0 s where b (resp. s) denotes the bicompletion (resp.
Smyth completion) functor. In particular, it follows from their results that the
Smyth completion (X, au, ~) of (X, au., '!J) can be obtained via the bicompletion
(X, aU;) and that a topological quasi-uniform space is Smyth complete exactly
when its image under the functor w is bicomplete.
In [166] Flagg and Snderhauf introduced and studied a concept of a topo-
logical 'V -continuity space. They showed that the Smyth completion of a quali-
tative domain with its Alexandroff topology yields the ideal completion together
with its Scott topology. For the case of posets this theorem reduces to a result
due to R.-E. Hoffmann: The sobrification of aposet in its Alexandroff topology
equals the ideal completion ip its Scott topology.
A further important attempt to reconcile orders and metric spaces is due to
Matthews [407]. It arose in the context ofhis studies of denotational semantics
of dataflow networks. Based on his idea that points do not necessarily have zero
distance from themselves he introduced the concept of a partial metric. He also
showed that an equivalent theory can be obtained by his so-called weighted
quasi-pseudometrics. A quasi-pseudometric space (X, q) is weightable [407] if
there exists a (so-called weight) function I . I : X --+ R+ (here R+ denotes the
set of the nonnegative reals) satisfying
that the new theory comprises the classical theory of quasi-uniform spaces as
weH as the theory of frame uniformities and is, in the setting of biframes,
equivalent to Frith's theory of covering quasi-uniformities. Many results about
compactifications and completions have been established by these three authors
for frame quasi-uniformities which extend the c1assical theory of quasi-uniform
spaces. For instance in their article [184] they considered totally bounded quasi-
uniformities and quasi-proximities for frames and proved that for a given quasi-
proximity < on a frame L there is a totally bounded quasi-uniformity on L that is
the coarsest quasi-uniformity and the only totally bounded quasi-uniformity, that
determines <. Furthermore they constructed the compactification of a totally
bounded frame quasi-uniformity.
In [213], [214] Frith, Hunsaker and Walters-Wayland presented the
(bi)completion of a quasi-uniform frame. Applied to a uniform frame their
procedure yields its (unique) completion. The notion of Samuel compactification
was extended to quasi-uniform frames. As expected, the Samuel compactifica-
tion of a quasi-uniform frame is defined to be the completion of its totally
bounded coreection. So-called Cauchy completions of quasi-uniform frames
were constructed by Kim [282]. It is known that each complete quasi-uniform
frame is Cauchy complete.
Doitchinov's paper [138] contains a discussion of such theories. While he
considered the entourage approach an important step in the development of
the theory of frames, he thought that the work done so far relies too heavily
on the idea of a conjugate "companion" frame. In his opinion the idea of a
conjugate space which appears quite naturally when considering a quasi-uniform
space consisting of points, is in fact something foreign in the pointless case. In
particular he suggested a different definition of u-smallness for frame elements
- a concept which is central in the theory of Fletcher, Hunsaker and Lindgren.
In any case, as in the spatial setting, entourage-like theories seem to be more
manageable than covering ones in the study of quasi-uniformities. An interest-
ing alternative (but equivalent) approach to entourage (quasi-)uniformities on
frames was recently presented by Picado [449], [450]. It clarified some points
raised in Doitchinov's discussion [138]. Picado's arguments rely on the presenta-
tion of the coproduct of the underlying frame by itself as a frame of C -ideals and
his entourages are special elements of that coproduct. He stated that the concept
of a quasi-uniform frame should be obtained from the corresponding symmetric
concept just by dropping the symmetry condition. In [260] Hunsaker and Picado
investigated the frame (quasi-uniformities) with a transitive base and established
results analogous to the corresponding spatial ones for zero-dimensional frames
and zero-dimensional biframes.
According to aresult ofNeville every frame is isomorphic to the generalized
Gleason algebra of an essentially unique bi-Stonian space (X, ff, ~). In [175]
Fletcher, Frith, Hunsaker and Schauerte showed that the unique quasi-uniformity
934 HANS-PETER A. KNZI
that detennines rzJ and the specialization order of :J' (in Nachbin's sense) is
compatible with :J' and coarser than the Pervin quasi-unifonnity of:J'.
The study ofjuzzy quasi-uniformities was initiated by Hutton [261]. Since
then fuzzy concepts in asymmetrie topology have been investigated by several
authors. In general, they tried more or less successfully to extend major c1assical
results about quasi-unifonnities to some fuzzy setting. The basic idea ofHutton's
fuzzification of quasi-unifonnities is the observation that if (X, Oll) is an ordinary
quasi-uniform space, then each entourage U can be identified with a function
U : 2 x -* 2 x by defining A -* U (A) as usual; obviously A 5; U (A) for every
A E 2 x and U preserves unions.
Hutton hirnself showed that every fuzzy topological space is fuzzy quasi-
unifonnizable. In [274] Katsaras proved that there exists a one-to-one correspon-
dence between the family of all fuzzy quasi-proximities, on a set X, and the
family of all totally bounded fuzzy quasi-proximities on X. In [275] he and Peta-
las considered a concept ofthe fuzzy fine (transitive) quasi-unifonnity. Wu and
Wu [585] proved that a fuzzy bitopological space is fuzzy quasi-unifonnizable
if and only if it is pairwise completely regular. Moreover it was shown that if
the (bi)space is also (pairwise) compact, then the fuzzy bitopology is induced
by a unique fuzzy quasi-unifonnity. Fuzzy Hausdorff-Bourbaki structures and
their relations with the fuzzy Vietoris structures were investigated by Maes in
[399].
In [145] Dzhajanbajev and Sostak introduced a concept of a fuzzy (quasi-)
uniformity which is consistent with fuzzy topology as it was defined by the
second author. According to Sostak a fuzzy topology on a set X is a fuzzy
subset of the fuzzy powerset I x (i.e. a mapping rzJ : I x -* l) satisfying certain
axioms.
However the approach to quasi-unifonnities outlined by these two authors,
which contains Hutton's theory as an important, but very special crisp case, led
to relations between fuzzy quasi-unifonnities and fuzzy quasi-proximities that
are essentially different from c1assical ones. Therefore, in his investigations on
fuzzy syntopogenous structures Sostak [552] suggested an alternative definition
of a fuzzy quasi-unifonnity which seems to be more satisfactory. In a related
artic1e [227], various quasi-uniform counterparts of even fuzzy topologies were
considered.
In the spirit of Nachbin, the three authors Allam, EI-Saady and Mashhour
[11] investigated the relationship between order and fuzzy unifonnities with
the help of fuzzy quasi-unifonnities. Fuzzy syntopogenous preordered spaces
were studied in [10]. Moreover the artic1e [26] contains a discussion of a notion
of a fuzzy proximity ordered space. Some concept of a stratified fuzzy quasi-
uniformity was considered in [35].
NONSYMMETRIC DISTANCES AND THEIR ASSOCIATED TOPOLOGIES 935
In [223], [224] Gregori and Romaguera proved some fixed point theorems
for fuzzy mappings in Smyth complete and left K -complete quasi-pseudometric
spaces.
In this context we should also mention the theories of pointwise topology
and pointwise quasi-uniformities on completely distributive lattices, developed
mainly in the Chinese mathematicalliterature [575], [588], [539]. One of its
purposes, according to Wang [575], is the intention to bring the theories ofpoint-
set topology and fuzzy topology together in a unified theory. In [540] Shi proved
for instance that each topological molecular lattice is quasi-uniformizable.
approach distance 8 on X we obtain the gauge ' associated with 8 as the set of
all extended quasi-pseudometrics d on X which satisfy 8d ::; 8.
The gauge ' of an approach space (X, 8) induces (= generates) a quasi-
uniformity ~(') on X, which is compatible with the topology 5'8. By means
of this construction we obtain for topological approach spaces the fine quasi-
uniformity of the induced topology, and for quasi-pseudometric approach spaces
their standard quasi-pseudometric quasi-uniformity.
Symmetric gauges (i.e. gauges having bases consisting of extended pseudo-
metrics) describe the so-called uniform approach spaces, which in the obvious
way are related to uniform spaces.
In the following years Lowen and his students studied the concept of an ap-
proach space more carefully. As in topological and metric spaces a convergence
theory can be developed in AP.
Approach spaces also seem to be the right setting for a unified treatment of
certain topological and metric properties, as well as for the development of a
theory of approximation (see e.g. [379]).
Some of the advantages of the notion of an approach space (see the dis-
cussion contained in [384]) can readily be seen by taking arbitrary products of
metric spaces in AP: The notion of a distance is not lost, but the point-to-point
distances no longer suffice to describe the structure. All point-to-set distances
are required, and in general approach spaces these cannot be derived from their
point-to-point distances.
In the book "Approach Spaces: The Missing Link in the Topology - Uni-
formity - Metrie Triad" [382], which appeared in 1997, Lowen explained
the theory of approach spaces developed over the past ten years in a systematic
form: Several completely different, but equivalent characterizations of approach
spaces were discussed, e.g. approach systems (related to neighborhood systems
in topology), regular function frames (paralleling c10sed sub sets in topology),
hulls (similar to topological c1osures), towers (special families of pre-topological
c10sure operators) and approach limits (corresponding to the description of a
topology via convergence of filters). The author also demonstrated that cer-
tain well-known spaces, like hyperspaces, function spaces, probabilistic metric
spaces, spaces of measures and spaces of random variables are equipped with a
natural approach structure (see also [381], [390]).
Indeed, many interesting applications of approach distances in the theory
of hyperspaces of metric spaces were found. In [386], for a metric space X
and a fairly arbitrary collection L of subsets of X, the hyperspace CL(X) of
all nonempty c10sed subsets of X (to be identified with their distance func-
tionals) was endowed with a canonical approach distance function having the
topology of uniform convergence on members of L as topological corefiection
and the Hausdorff metric as metric eorefiection. It was shown that for particular
choices of L, canonical approach distance functions overlying the Wijsman and
NONSYMMETRIC DISTANCES AND THEIR ASSOCIATED TOPOLOGIES 937
structures might wish to dispense with that restriction and imitate the model of
continuity spaces.
Using the language of approach spaces for problems related to theoretical
computer science, in [582] WindeIs introduced a Scott (approach) distance
which is a canonical numerification of the Scott topology and allows quantita-
tive considerations. He showed that the Scott distance shares many important
features of the Scott topology.
In [429] Nauwelaerts introduced a family of Cartesian c10sed topologi-
cal subconstructs in the category of semi-approach uniform limit spaces and
uniform contractions, which is a topological universe containing AUDif and
the category of semi-uniform limit spaces in the sense of Preuss. He argued
that the objects that he obtained in this way satisfy some kind of regularity
condition and observed that analogous results and examples can be established
in a non-symmetric setting.
References
[1] Aarts, I.M. and Mrsevic, M., Pairwise complete regularity as a separation axiom,
J. Austral. Math. Soc. (Ser. A) 48 (1990), 235-245.
[2] _ _ , A bitopological view on cocompact extensions, Topology Appl. 42
(1991), 1-16.
[3] Albert, G.E., A note on quasi-metric spaces, Bull. Amer. Math. Soc. 47 (1941),
479-482.
[4] Alegre, C., Ferrer, I. and Gregori, v., Quasi-uniform structures in linear lattices,
Rocky Mountain J. Math. 23 (1993), 877-884.
[5] _ _ , Quasi-uniformities on real vector spaces, Indian 1. Pure Appl. Math. 28
(1997),929-937.
[6] _ _ , On a dass of real normed lattices, Czechoslovak Math. J. 48(123) (1998),
785-792.
[7] _ _ , On the Hahn-Banach theorem in certain linear quasi-uniform structures,
Acta Math. Hung. 82 (1999), 325-330.
[8] Alemany, E. and Romaguera, S., On half-completion and bicompletion of quasi-
metric spaces, Comment. Math. Univ. Carol. 37 (1996), 749-756.
[9] _ _ , On right K -sequentially complete quasi-metric spaces, Acta Math.
Hung. 75 (1997), 267-278.
[10] Allam, A.A., EI-Saady, K. and Hussein, S.A., Fuzzy syntopogenous structures
and order, Fuzzy Sets Syst. 63 (1994), 91-98.
[11] Allam, A.A., EI-Saady, K. and Mashhour, A.S., Fuzzy uniform structures and
order, 1. Fuzzy Math. 2 (1994), 57-67.
[12] A16, R.A. and Deeba, E.Y., A note on uniformities of a BCK-algebra, Math.
Japonica 30 (1985), 237-240.
[13] Andrikopoulos, A. and Stabakis, J., On Nachbin's problem conceming uni-
formizable ordered spaces, Acta Math. In/orm. Univ. Ostrav. 5 (1997), 65-70.
942 HANS-PETER A. KNZI
[37] Bentley, H.L., Herrlich, H. and Hunsaker, W.N., A Baire category theorem for
quasi-metric spaces, Indian J Math. 37 (1995), 27-30.
[38] Bentley, H.L., Herrlich, H. and Husek, M., The historical development of uni-
form, proximal, and nearness concepts in topology, in Handbook of the History
ofGeneral Topology, C.E. Aull and R. Lowen (eds.), vol. 2, Dordrecht, Kluwer,
1998, 577--629.
[39] Bentley, H.L. and Hunsaker, W.N., Cauchy sequences in quasi-uniform spaces:
categorical aspects, Lecture Notes Computer Science, vol. 393, Springer, Berlin
(1989), 278-285.
[40] Berthiaume, G., On quasi-uniformities in hyperspaces, Proc. Amer. Math. Soc.
66 (1977), 335-343.
[41] Beyer, W.A., Smith, T.F. and Waterman, M.S., Some biological sequence
metrics, Adv. Math. 20 (1976), 367-387.
[42] Birsan, T., Transitive quasi-uniformities and zero dimensional bitopological
spaces, An. Sti. Univ. "Al. I. Cuza" IaJi Sect. Ia Mat. 20 (1974), 317-322.
[43] Blatter, 1., Order compactifications of totally ordered topological spaces, J.
Approximation Theory 13 (1975), 56--65.
[44] _ _ , Hewitt's Stone-Weierstrass theorems for ordered topological spaces,
in Functional Analysis, de Figueiredo (ed.), Proc. of the Brazilian Math. Soc.
Symposium 1974, Lecture Notes in Pure and Applied Math., vol. 18, Dekker,
New York (1976), 9-25.
[45] Blatter, J. and Seever, G.L., Interposition and lattice cones of functions, Trans.
Amer. Math. Soc. 222 (1976), 65-96.
[46] Bonsangue, M.M., Rutten, J.J.M.M. and van Breugel, F., Alexandroff and Scott
topologies for generalized metric spaces, in Papers on General Topology and
Applications, Eleventh Summer Conference, Southem Maine, 1995, Annals
New York Acad. Sei., vol. 806, 1996, 49--68.
[47] _ _ , Generalized metric spaces: completion, topology, and powerdomains
via the Yoneda embedding, Theoretical Computer Science 193 (1998), 1-51.
[48] Brack, P. and Kent, D.C., Approach spaces, limit tower spaces, and probabilistic
convergence spaces, Appl. Categ. Struct. 5 (1997), 99-110.
[49] _ _ , On convergence approach spaces, Appl. Categ. Struct. 6 (1998), 117-
125.
[50] Brown, L.M., On topological spaces with a unique compatible quasi-uniformity,
Glasgow Math. J. 18 (1977), 11-12.
[51] _ _ , Confluence para-quasi-unijormities, Topology theory and applications,
5th Colloq., EgerlHung. 1983, Colloq. Math. Soc. Janos Bolyai, vol. 41, 1985,
125-152.
[52] _ _ , Semi-sequentially normal bitopological spaces, Topology Appl. 44
(1992),57--62.
[53] _ _ , P-Q-metrizability ofbitopological spaces with a dual base of countable
order, Acta Math. Hung. 83 (1999), 315-325.
[54] Brmmer, G.C.L., Initial quasi-uniformities, Nederl. Akad. Wetensch. Proc. Sero
A 72 =Indag. Math. 31 (1969),403-409.
[55] _ _ , On certain factorizations of functors into the category of quasi-uniform
spaces, Quaestiones Math. 2 (1977), 59-84.
944 HANS-PETER A. KNZI
[75] Cao, J., Reilly, LL. and Romaguera, S., Some properties of quasiuniform
multifunetion spaees, 1. Austral. Math. Soc. (Ser. A) 64 (1998), 169-177.
[76] Cao, J., Reilly, LL. and Vamanamurthy, M.K., Comparison of convergenees for
multifunetions, Demon. Math. 30 (1997), 171-182.
[77] Carlson, J.W., On the eategory of quasi-uniform spaees, Kyungpook Math. J. 16
(1976),63-69.
[78] _ _ , Quotient struetures for quasi-uniform spaees, Colloq. Math. 36 (1976),
63-68.
[79] _ _ , Nearness and quasi-uniform struetures, Topology Appl. 15 (1983), 19-
28.
[80] Carlson, J.W. and Hieks, T.L., On eompleteness in quasi-uniform spaces, J.
Math. Anal. Appl. 34 (1971), 618-627.
[81] _ _ , Some properties of quasi-uniform struetures, Arch. Math. (Basel) 24
(1973),81-86.
[82] Carlson, J.W., Hieks, T.L. and Huffman, S.M., Complete quasi-uniform spaces,
Canad. Math. Bull. 23 (1980), 497-498.
[83] Carter, K.S. and Hieks, T.L., Some results on quasi-uniform spaees, Canad.
Math. Bull. 19 (1976), 39-51.
[84] Chattopadhyay, K.e. and Hazra, RN., Basic quasi-proximities, grills and
eompaetifieations, Rocky Mountain J. Math. 15 (1985), 835-851.
[85] Cheea, E. and Romaguera, S., Continuity of eontraetive mappings on eomplete
quasi-metrie spaees, Math. Japonica 35 (1990), 137-139.
[86] Chicourrat, M., RE-proximites et ultrafiltres, CR. Acad. Sei. Paris 318 (1994),
143-148.
[87] _ _ , RE-proximities as fixed points of an operator on pseudo proximities,
Topology Appl. 104 (2000), 39-51.
[88] Choban, M.M. and Nedev, S.L, On the theory of O-metrizable spaees, I, 11, m
(Russian), Vestnik Moskov. Univ. Sero l. Math. Meh. 27 1, 8-15; 2, 10-17; 3,
10-15 (1972).
[89] Choe, T.H., PartiallY ordered topologie al spaees, An. Acad. Bras. Cienc. 51
(1979), 53-63.
[90] Chou, C.C. and Penot, J.-P., Infinite produets ofrelations, set-valued series and
uniform openness ofmultifunetions, Set-Valued Anal. 3 (1995), 11-21.
[91] Ciesielski, K., Flagg, Re. and Kopperman, R, Charaeterizing topologies
with bounded eomplete eomputational models, Electronic Notes Theoreti-
cal Computer Science 20 (1999), URL: http://www.elsevier.nl/loeate/entes/
volume20.html
[92] _ _ , Polish spaees, eomputable approximations, and bitpological spaees,
Topology Appl., to appear.
[93] Ciric, L., Periodic and fixed point theorems in a quasi-metrie spaee, 1. Austral.
Math. Soc. (Ser. A) 54 (1993),80-85.
[94] _ _ , Semi-eontinuous mappings and fixed point theorems in quasi metric
spaees, Publ. Math. Debrecen 54 (1999), 251-261.
[95] Collins, P.J., Gartside, P.M., Kopperman, RD., Knzi, H.P.A. and Moody, P.J.,
On topologies generated by filters of relations, in Papers on General Topology
946 HANS-PETER A. KNZI
[183] Fletcher, P., Hunsaker, W. and Lindgren, W., Characterizations of frame quasi-
uniformities, Quaestiones Math. 16 (1993), 371-383.
[184] _ _ , Totally bounded frame quasi-uniformities, Comment. Math. Univ. Carol.
34 (1993), 529-537.
[185] _ _ , Frame quasi-uniformities, Monatsh. Math. 117 (1994), 223-236.
[186] Fletcher, P. and Knzi, H.P.A, A topological space without a complete quasi-
uniformity, Proe. Amer. Math. Soe. 90 (1984), 611-615.
[187] _ _ , Extension properties induced by complete quasi-uniformities, Paeifie 1.
Math. 120 (1985), 357-384.
[188] Fletcher, P. and Lindgren, w.F., Transitive quasi-uniformities, 1. Math. Anal.
Appl. 39 (1972), 397-405.
[189] _ _ , Quasi-uniformities with a transitive base, Paeifie 1. Math. 43 (1972),
619-631.
[190] _ _ , Orthocompactness and strong Cech completeness in Moore spaces, Duke
Math. 1. 39 (1972), 753-766 (Correction, ibid. 40 (1973), 959).
[191] _ _ , Locally quasi-uniform spaces with countable bases, Duke Math. J. 41
(1974),231-240.
[192] _ _ , Some unsolved problems conceming countably compact spaces, Roeky
Mountain 1. Math. 5 (1975), 95-106.
[193] _ _ , C-complete quasi-uniform spaces, Areh. Math. (Basel) 30 (1978),175-
180.
[194] _ _ , A construction of the pair completion of a quasi-uniform space, Canad.
Math. Bull. 21 (1978), 53-59.
[195] _ _ , Equinormal quasi-uniformities and quasi-metrics, Glas. Mat. Ser. III 13
(33) (1978), 111-125.
[196] _ _ , A theory of uniformities for generalized ordered spaces, Canad. 1. Math.
31 (1979), 35-44.
[197] _ _ , Quasi-Uniform Spaces, Leeture Notes Pure Appl. Math. 77, Dekker,
New York 1982.
[198] _ _ , Compactifications of totally bounded quasi-uniform spaces, Glasgow
Math. 1. 28 (1986), 31-36.
[199] Fletcher, P. and Liu, P., Topologies compatible with homeomorphism groups,
Paeifie 1. Math. 56 (1975), 77-86.
[200] Fleteher, P., McCoy, RA and Slover, R, On boundedly metacompact and
boundedly paracompact spaces, Proe. Amer. Math. Soe. 25 (1970),335-342.
[201] Fleteher, P. and Naimpally, S.A, On almost complete and almost precompact
quasi-uniform spaces, Czeehoslovak Math. 1. 21(96) (1971), 383-390.
[202] Fora, AA, Strongly zero-dimensional bitopological spaces, 1. Univ. Kuwait
(Sei.) 11 (1984), 181-190.
[203] Fox, R, Letter to S. Salbany, 1979.
[204] _ _ , A short proof of the Junnila quasi-metrization theorem, Proe. Amer.
Math. Soe. 83 (1981), 663-664.
[205] _ _ , Pretransitivity and products of suborderable spaces, Topologyand Order
Struetures, MC Traet Part I, Amsterdam, 142 (1981), 115-118.
[206] _ _ , Solution of the y-space problem, Proe. Amer. Math. Soe. 85 (1982),
606-608.
NONSYMMETRIC DISTANCES AND THEIR ASSOCIATED TOPOLOGIES 951
[230] Haddad, L., Sur quelques points de topologie generale. Theorie des nasses et
des tramails, Ann. Fac. Sei. Clermont-Ferrand 44, fascicule 7, 1970,3-80.
[231] Halpin, M.N., Transitive quasi-uniform spaces, M.Sc. thesis, Univ. Cape Town,
1974.
[232] HanselI, RW., Borel measurable mappings for nonseparable metric spaces,
Trans. Amer. Math. Soc. 161 (1971), 145-169.
[233] Harder, AM. and Hicks, T.L., Fixed point theory and iteration procedures,
Indian 1. Pure Appl. Math. 19 (1988), 17-26.
[234] Hausdorff, F., Grundzge der Mengenlehre, Leipzig 1914.
[235] Heath, RW., A postscript to a note on quasi-metric spaces, Notices Amer. Math.
Soc. 19 (1972), A-338.
[236] _ _ , A construction of a quasi-metric Souslin space with a point-countable
base, Set-theoretic topology (papers, Inst. Medicine and Math., Ohio Univ.,
Athens,Ohio, 1975-1976),219-224.
[237] _ _ , Some nonmetric, first countable, cancellative topological semigroups
that are generalized metric spaces, Topology Appl. 44 (1992), 167-173.
[238] Heckmann, R, Approximation of metric spaces by partial metric spaces, Appl.
Categ. Struct. 7 (1999), 71-83.
[239] Hejcman, J. and Vilimovsky, J., On a property of pseudometrics and uniformities
near to convexity, Czechoslovak Math. 1. 38(113) (1988), 366--380.
[240] Henriksen, M., Kopperman, R., Mack, J. and Somerset, D.W.B., Joincompact
spaces, continuous lattices, and C* -algebras, Algebra Univ. 38 (1997), 289-323.
[241] Henriksen, M., Kopperman, R, Raybum, M. and Todd, AR., Oxtoby's
pseudocompleteness revisited, Topology Appl. 100 (2000), 119-132.
[242] Herden, G., On a lifting theorem of Nachbin, Mathematical Soeial Seiences 19
(1990), 37--44.
[243] Hicks, TL., Fixed point theorems for quasi-metric spaces, Math. Japonica 33
(1988),231-236.
[244] Hicks, T.L. and Huffman, S.M., A note on locally quasi-uniform spaces, Canad.
Math. Bull. 19 (1976), 501-504.
[245] Hicks, T.L. and Satterwhite, RE., Quasi-pseudometrics over Tikhonov semi-
fields, Math. Japonica 22 (1977), 315-321.
[246] Hicks, T.L. and Sharma, P., Properties of z-spaces, Topology Proc. 4 (1979),
109-113.
[247] Hodel, RE., Spaces defined by sequences of open covers which guarantee that
certain sequences have cluster points, Duke Math. 1. 39 (1977), 253-263.
[248] _ _ , Metrizability of spaces satisfying Nagata's condition, Math. Japonica
47 (1998), 287-293.
[249] _ _ , Neigbhorhood assignments and cardinal functions: a unified approach
to metrization and uniformity, Topology Appl. 90 (1998),31-56.
[250] _ _ , A history of generalized metrizable spaces, in Handbook ofthe History
ofGeneral Topology, C.E. Aull and R Lowen (eds.), vol. 2, Dordrecht, Kluwer
1998,541-576.
[251] Hoffmann, R-E., On the sobrification remainder S X \ X, Pacific 1. Math. 83
(1979), 145-156.
[252] Hogan, D.T., Quotient quasi-uniformities, Math. Chronicle 4 (1976), 108-111.
NONSYMMETRIC DISTANCES AND THEIR ASSOCIATED TOPOLOGIES 953
[253] Hohi, L. and Levi, S., Decomposition properties of hyperspace topologies, Set-
Valued Anal. 5 (1997), 309-321.
[254] Howes, N.R., Modem Analysis and Topology, Springer, Berlin 1995.
[255] Hung, H.H., Shrinkings of open neighbourhoods and a topologieal description
of metrizability, Topology Proc. 22 (Summer 1997), 139-153.
[256] _ _ , Quasi-metrizability, Topology Appl. 83 (1998), 39-43.
[257] Hunsaker, W.N., Extensions of continuous increasing functions, Topology Proc.
5 (1980), 105-110.
[258] Hunsaker, W.N. and Lindgren, w.F., Construction of quasi-uniformities, Math.
Ann. 188 (1970), 39-42.
[259] _ _ , Upper semi-continuous functions and quasi-proximity classes, Bul. Inst.
Politehn. Iai (N.S.) 19(23) (1973) sect. I, 23-30.
[260] Hunsaker, W.N. and Pieado, J., Frames with transitive structures, Appl. Categ.
Struct., to appear.
[261] Hutton, B., Uniformities on fuzzy topologieal spaces, 1. Math. Anal. Appl. 58
(1977), 559-571.
[262] Jachymski, J., A contribution to fixed point theory in quasi-metric spaces, Publ.
Math. Debrecen 43 (1993), 283-288.
[263] Jawhari, E.M., Misane, D. and Pouzet, M., Retracts: Graphs and ordered sets
from the metric point ofview, Contemp. Math. 57 (1986), 175-226.
[264] Jung, A. and Snderhauf, P., Uniform approximation of topological spaces,
Topology Appl. 83 (1998), 23-37.
[265] Junnila, H.J.K., Neighbornets, Pacific J. Math. 76 (1978),83-108.
[266] _ _ , Covering properties and quasi-uniformities oftopological spaces, Ph.D.
Thesis, Virginia Polytechnic Institute and State University, Blacksburg, Va.
1978.
[267] Junnila, H.J.K. and Knzi, H.P.A., Ortho-bases and monotonie properties, Proc.
Amer. Math. Soc. 119 (1993), 1335-1345.
[268] _ _ , Stability in quasi-uniform spaces and the inverse problem, Topology
Appl. 49 (1993), 175-189.
[269] _ _ , Characterizations of absolute Fu8-sets, Czechoslovak Math. J. 48(123)
(1998),55-64.
[270] Junnila, HJ.K., Knzi, H.P.A. and Watson, S., On a class of hereditarily para-
compact spaces, Fundam. Prikl. Mat. 4 (1998), 141-154 (Russian); preprint in
Topology Atlas: URL: http://at.yorku.ca/p/a/a/g/09.htm (English).
[271] Junnila, H.J.K., Smith, J.C. and Telegarski, R., Closure preserving covers by
small sets, Topology Appl. 23 (1986), 237-262.
[272] Kada, 0., Suzuki, T. and Takahashi, W., Nonconvex minimization theorems
and fixed point theorems in complete metric spaces, Math. Japonica 44 (1996),
381-391.
[273] Karthikeyan, c., Bicompletion of quasi-bitopologieal near-rings and quasi-
bitopological N -groups, Indian J. Pure Appl. Math. 30 (1999), 211-220.
[274] Katsaras, A.K., Fuzzy quasi-proximities and fuzzy quasi-uniformities, Fuzzy
Sets Syst. 27 (1988), 335-343.
[275] Katsaras, A.K. and Petalas, C.G., Fuzzy topologies and fuzzy quasi-
uniformities, Radovi Mat. 4 (1988), 15-29.
954 HANS-PETER A KNZI
[276] Keimei, K. and Roth, w., Ordered Cones and Approximation, Lecture Notes
Math., vol. 1517, Springer, Berlin 1992.
[277] Kelly, J.C., Bitopological spaces, Proc. London Math. Soc. 13 (1963), 71-89.
[278] Kent, D.C. and Liu, D.M., Ordered compactifications and families of maps,
Internat. J. Math. Math. Sei. 20 (1997), 105-110.
[279] Kent, D.C. and Richardson, G.D., Ordered probabilistic metric spaces, J.
Austral. Math. Soc. (Ser. A) 46 (1989), 88-99.
[280] Kent, D.C. and Richmond, T.A, Ordered compactification of totally ordered
spaces, Internat. J. Math. Math. Sei. 11 (1988), 683-694.
[281] Khanh, P.Q., On general open mapping theorems, J. Math. Anal. Appl. 144
(1989),305-312.
[282] Kim, Y.K., Cauchy completions of quasi-uniform frarnes, Kyungpook Math. J.
38 (1998), 429-437.
[283] Kimmie, Z., Functorial transitive quasi-uniformities and their bicompletion,
Ph.D. Thesis, University of Cape Town, August 1995.
[284] Kofner, Ja. A, Ll-metrizable spaces (Russian), Mat. Zametki 13 (1973), 277-
287; (Translation =Math. Notes 13 (1973), 168-174).
[285] Kofner, J., On quasi-metrizability, Topology Proc. 5 (1980), 111-138.
[286] _ _ , Quasi-metrizable spaces, Pacific J. Math. 88 (1980), 81-89.
[287] _ _ , Closed mappings and quasi-metrics, Proc. Amer. Math. Soc. 80 (1980),
333-336.
[288] _ _ , Transitivity and the y-space conjecture in ordered spaces, Proc. Amer.
Math. Soc. 81 (1981), 629-635.
[289] _ _ , Transitivity and ortho-bases, Canad. J. Math. 33 (1981), 1439-1447.
[290] _ _ , Open compact mappings, Moore spaces and orthocompactness, Rocky
Mountain J. Math. 12 (1982), 107-112.
[291] Kong, T.Y., Kopperman, R. and Meyer, P.R., Which spaces have metric analogs,
General Topology and Applications, Proc. 5th Northeast Conf., New YorkINY
(USA) 1989, Lect. Notes Pure Appl. Math., vol. 134, 1991, 209-215.
[292] Konstandilaki-Savopoulou, Ch. and Reilly, I.L., 'On Datta's bitopological
paracompactness', Indian J. Pure Appl. Math. 12 (1981), 799-803.
[293] Kopperman, R.D., Lengths on semigroups and groups, Semigroup Forum 25
(1982),345-360.
[294] _ _ , All topologies come from generalized metrics, Amer. Math. Monthly 95
(1988), 89-97.
[295] _ _ , Total boundedness and compactness for filter pairs, Ann. Univ. Sei.
Budapest. 33 (1990), 25-30.
[296] _ _ , Which topologies are quasimetrizable? Topology Appl. 52 (1993), 99-
107.
[297] _ _ , Asymmetry and duality in topology, Topology Appl. 66 (1995), 1-39.
[298] Kouphos, Th.K. and Papadopoulos, B.K., Exponentiallaws for function spaces
equipped with the quasi-uniformity ofthe quasi-uniform convergence, Ann. Soc.
Sei. Bruxelles Sero I, 101 (1987), 89-97.
[299] Krishnan, V.S., A note on semiuniform spaces, J. Madras Univ. Sect. B 25
(1955), 123-124.
NONSYMMETRIC DISTANCES AND THEIR ASSOCIATED TOPOLOGIES 955
[342] Knzi, H.P.A, Romaguera, S. and Salbany, S., Topological spaces that admit
bicomplete quasi-pseudometrics, Ann. Univ. Sei. Budapest. 37 (1994),185-195.
[343] _ _ , Bispaces admitting only bicomplete or only totally bounded quasi-
metrics, Rend. Circ. Mat. Palermo 1I, to appear; preprint partially in: H.P.A
Knzi and S. Romaguera, On hereditarily compact quasi-pseudometrizable
spaces, in Seminarberichte FernUniversitt Gesamthochschule in Hagen, Fach-
bereich Mathematik, 63(3) (1998), 439-446.
[344] Knzi, H.P.A, Romaguera, S. and Sipacheva, O.v., The Doitchinov completion
of a regular paratopological group, Serdica Math. J 24 (1998), 73-88.
[345] Knzi, H.P.A and Ryser, c., The Bourbaki quasi-uniformity, Topology Proc.
20 (1995), 161-183.
[346] Knzi, H.P.A and Schellekens, M.P., On the Yoneda-completion of a quasimet-
ric space, Theoretical Computer Seience, to appear.
[347] Knzi, H.P.A and Vajner, V., Weighted quasi-metrics, in Papers on General
Topology andApplications, Eighth Summer Conf. Queens College 1992, Annals
New York Acad. Sci., vol. 728,1994,64-77.
[348] Knzi, H.P.A and Wajch, E., Borel classification via quasi-metrics, Topology
Appl. 77 (1997), 183-192.
[349] _ _ , On a-discrete Borel mappings via quasi-metrics, Czechoslovak Math.
J 48(123) (1998), 439-455.
[350] Knzi, H.P.A and Watson, S., A metrizable completely regular ordered space,
Comment. Math. Univ. Carol. 35 (1994), 773-778.
[351] _ _ , A nontrivial Tl-space admitting a unique quasi-proximity, Glasgow
Math. J. 38 (1996), 207-213.
[352] _ _ , A quasi-metric space without complete quasi-uniformity, Topology Appl.
70 (1996),175-178.
[353] _ _ , A nontransitive space based on combinatorics, Boll. U.M.1. (8) 2-B
(1999),315-317.
[354] LaI, S. and Singal, M.K., Proximity ordered spaces, J. London Math. Soc. 14
(1976),393-404.
[355] Lambrinos, P.Th., Quasi proximal continuity, Bull. Austral. Math. Soc. 9 (1973),
89-98.
[356] _ _ , Quasi-uniform characterizations of (weak) boundedness and (weak)
compactness, Annal. Soc. Sei. Bruxelles 90 (1976), 307-316.
[357] _ _ , On precompact (quasi-)uniform structures, Proc. Amer. Math. Soc. 62
(1977), 365-366.
[358] Lane, E.P., Bitopological spaces and quasi-uniform spaces, Proc. London Math.
Soc. 17 (1967), 241-256.
[359] Lawson, J.D., Order and strongly sober compactifications, in Topology and
Category Theory in Computer Seience, G.M. Reed, AW. Roscoe and R.E
Wachter (eds.), Clarendon Press, Oxford 1991, 179-205.
[360] Lawvere, EW., Metric spaces, generalized logic, and closed categories, Rend.
Sem. Mat. Fis. Milano 43 (1973), 135-166.
[361] Lee, YJ. and Lowen, R., Approach theory in merotopic, Cauchy and conver-
gence spaces, I and TI, Acta Math. Hung. 83 (1999), 189-207; 209-229.
958 HANS-PETER A KNZI
[362] Levine, N. and Stager, W.J., Jr., On the hyper-space of a quasi-uniform space,
Math. 1. Okayama Univ. 15 (1971-72), 101-106.
[363] Lindgren, W.F., Topological spaces with unique quasi-uniform structure, Arch.
Math. (Basel) 22 (1971), 417-419.
[364] _ _ , Topological spaces with a unique compatible quasi-uniformity, Canad.
Math. Bull. 14 (1971), 369-372.
[365] Lindgren, W.F. and Nyikos, P.J., Spaces with bases satisfying certain order and
intersection properties, Pacific 1. Math. 66 (1976), 455-476.
[366] Losonczi, A, The cardinality of the coarsest quasi-proximity class of locally
compact T2-spaces, Topology Proc. 23 (Spring 1998),245-262.
[367] _ _ , Finite quasi-uniform extensions I, 11, Acta Math. Hung. 79 (1998), 85-
116; 82 (1999), 35-55.
[368] _ _ , Special simultaneous quasi-uniform extensions, Acta Math. Hung. 84
(1999), 9-18.
[369] _ _ , On the cardinality of compatible quasi-uniformities, Topology Appl. 103
(2000), 43-54.
[370] _ _ , Topological spaces with a coarsest compatible quasi-uniformity, Quaes-
tiones Math. 23 (2000), 67-75.
[371] _ _ , On the cardinality of 1r(8), Comment. Math. Univ. Carol., to appear.
[372] _ _ , Notes on the coarsest element of1r(8), preprint.
[373] Lowen, E. and Lowen, R., A quasitopos containing CONV and MET as full
subcategories, Internat. 1. Math. Math. Sei. 11 (1988),417-438.
[374] _ _ , Topological quasitopos hulls of categories containing topological and
metric objects, Cah. Topologie Geom. DijJ. Cat. 30 (1989), 213-228.
[375] Lowen-Colebunders, E., Lowen, R and Nauwelaerts, M., The Cartesian closed
hull of the category of approach spaces, Cah. Topologie Geom. DijJ. Cat., to
appear.
[376] Lowen, E., Lowen, R and Verbeeck, C. Exponential objects in the construct
PRAP, Cah. Topologie Geom. DijJ. Cat. 38 (1997), 259-276.
[377] Lowen, R, Kuratowski's measure of non-compactness revisited, Q. J. Math.
Oxf. II, 39 (1988), 235-254.
[378] _ _ , Approach spaces: a common supercategory of TOP and MET, Math.
Nachr. 141 (1989), 183-226.
[379] _ _ , A topological category suited for approximation theory? 1. Approxima-
tion Theory 56 (1989), 108-117.
[380] _ _ , Cantor-connectedness revisited, Comment. Math. Univ. Carol. 33
(1992),525-532.
[381] _ _ , Approximation of weak convergence, Math. Nachr. 160 (1993), 299-
312.
[382] _ _ , Approach Spaces: The Missing Link in the Topology - Uniformity -
Metric Triad, Oxford Mathematical Monographs, Oxford University Press 1997.
[383] Lowen, R. and Robeys, K., Completions of products of metric spaces, Q.1.
Math. Oxf. II 43 (1992),319-338.
[384] _ _ , Products ofmetric spaces, in Recent Developments ofGeneral Topology
and its Applications, Intern. Conf. in Memory of F. Hausdorff, Math. Research,
vol. 67, Akademie Verlag, Berlin 1992,203-211.
NONSYMMETRIC DISTANCES AND THEIR ASSOCIATED TOPOLOOIES 959
[408] McCartan, S.D., Separation axioms fortopological ordered spaces, Proc. Camb.
Phil. Soc. 64 (1968),965-973.
[409] Michael, E., Topologies on spaces ofsubsets, Trans. Amer. Math. Soc. 71 (1951),
152-183.
[410] Mooney, D.D. and Richmond, T.A., Cardinality and structure of semilattices
of ordered compactifications, in Papers on General Topology and Applications,
Ninth Summer Conference at Slippery Rock University 1993, Annals New York
Acad. ScL, vol. 767, 1995, 188-193.
[411] Mooney, D.D. and Richmond, T.A., The lattice of ordered compactifications of
a direct sum of totally ordered spaces, Order 15 (1998), 1-19.
[412] Morales, P., Topological contraction principle, Fundamenta Math. 110 (1980),
135-144.
[413] Mrsevic, M., Reilly, IL. and Vamanamurthy, M.K, Lebesgue sets in quasi-
pseudo-metric spaces, Math. Japonica 37 (1992), 189-194.
[414] Mordeshwar, M.G. and Naimpally, S.A., Tennungsaxioms in quasi-uniform
spaces, Nieuw Archiefvoor Wiskunde (3) 14 (1964), 97-101.
[415] _ _ , Quasi-Uniform Topological Spaces, Noordhoff, Groningen 1966.
[416] Mordeshwar, M.G. and Theckedath, KK, Boundedness in a quasi-uniform
space, Canad. Bull. 13 (1970), 367-370.
[417] Nachbin, L., Sor les espaces topologiques ordonnes, C.R. Acad. Sei. Paris 226
(1948), 381-382.
[418] _ _ , Sor les espaces uniformisables ordonnes, C.R. Acad. Sei. Paris 226
(1948), 547.
[419] _ _ , Sor les espaces uniformes ordonnes, C.R. Acad. Sci. Paris 226 (1948),
774-775.
[420] _ _ , Topology and Order, D. van Nostrand, Princeton 1965.
[421] Nagel, R., (ed.), in W. Arendt et al., One-parameter Semigroups of Positive
Operators, Springer Lecture Notes Math., vol. 1184 Berlin, 1986.
[422] Nailana, KR., Strict complete regularity in the categories of bitopological
spaces and ordered topological spaces, Publ. Math. Debrecen, to appear.
[423] Naimpally, S.A., Function spaces of quasi-uniform spaces, Nederl. Akad. Wet.
Proc. Sero A 68 =Indag. Math. 27 (1966), 768-771.
[424] _ _ , Separation axioms in quasi-uniform spaces, Amer. Math. Month. 74
(1967),283-284.
[425] Nauwelaerts, M., The Cartesian closed topological huH of the category of
(quasi-)uniform spaces (revisited), Rend. Ist. Mat. Univ. Trieste, to appear.
[426] _ _ , Cartesian closed hull for (quasi-)metric spaces (revisited), Comment.
Math. Univ. Carolin. 41 (2000), 559-573.
[427] _ _ , The hulls ofthe category ofuniform approach spaces, preprint.
[428] _ _ , Some Cartesian closed topological constructs of convergence-approach
spaces, preprint.
[429] _ _ , Some Cartesian closed topological constructs in the category of semi-
approach uniform limit spaces, Acta Math. Hung. 88 (2000),59-71.
[430] Nedev, S.l, Generalized-metrizable spaces (Russian), C.R. Acad. Bulgare Sei.
20 (1967), 513-516.
NONSYMMETRIC DISTANCES AND THEIR ASSOCIATED TOPOLOGIES 961
[431] Nielsen, R. and Sloyer, C., Quasi-uniformizability, Math. Ann. 182 (1969), 273-
274.
[432] Niemytzki, v., ber die Axiome des metrischen Raumes, Math. Ann. 104
(1931),666-671.
[433] Norman, L.J., A suffieient condition for quasi-metrizability of a topological
space, Portugal. Math. 26 (1967),207-211.
[434] O'Neill, S.1., Partial metrics, valuations and domain theory, in Papers on Gen-
eral Topology and Applications, Eleventh Summer Conference, Southem Maine,
1995, Annals New York Acad. Sei., vol. 806, 1996,304-315.
[435] Papadopoulos, B.K., Ascoli's theorem in a quiet quasi-uniform space, Panamer-
ican Math. J. (4) 3 (1993), 19-22.
[436] _ _ , Quasi-uniform convergence on function spaces, Questions Answers Gen.
Topology 12 (1994), 121-131.
[437] _ _ , (Quasi)-uniformities on the set of bounded maps, Internat. J. Math.
Math. Sei. 17 (1994), 693-696.
[438] _ _ , A note on the paper "Quasi-uniform convergence on function spaces",
Questions Answers Gen. Topology 13 (1995), 55-56.
[439] _ _ , Boundedness and (quasi-)uniform convergence, Math. Japonica 41
(1995),565-571.
[440] Pareek, C.M., Bitopological spaces and quasi-metric spaces, J Univ. Kuwait
Sei. 6 (1979), 1-7.
[441] Patty, c.w., Bitopological spaces, Duke Math. J. 34 (1967), 387-391.
[442] Pasquale, A., Hypertopologies induced by sc ales offunctional kind, Boll. U.M.1.
(7) 7-B (1993), 431-449.
[443] Perez-Pefialver, M.1. and Romaguera, S., On right K -completeness of the fine
quasi-uniformity, Questions Answers Gen. Topology 14 (1996), 245-249.
[444] _ _ , Weakly Cauchy filters and quasi-uniform completeness, Acta Math.
Hung. 82 (1999), 217-228.
[445] _ _ , Cofinal bicompleteness and quasi-metrizability, Rend. Inst. Mat. Univ.
Trieste 30 Suppl. (1999), 165-172.
[446] Pervin, W.J., Quasi-uniformization of topological spaces, Math. Ann. 147
(1962),316-317.
[447] _ _ , Quasi-proxirnities for topological spaces, Math. Ann. 150 (1963), 325-
326.
[448] Pervin, w.J. and Sieber, J.L., Completeness in quasi-uniform spaces, Math. Ann.
158 (1965), 79-81.
[449] Picado, J., Weil uniformities for frames, Comment. Math. Univ. Carol. 36
(1995),357-370.
[450] _ _ , Frame quasi-uniformities, Proc. Symposium on Categorical Topology,
University of Cape Town 1994, B. Banaschewski, c.R.A. Gilmour and H.
Herrlich (eds.), Dept. Math., University ofCape Town 1999,161-175.
[451] Pol, R., A perfectly normallocally metrizable non-paracompact space, Fund.
Math. 97 (1977), 37-42.
[452] Popa, E., Completion of quasi-uniform spaces, Math. Ann. 186 (1970), 297-298.
[453] Porter, K.F. Serni-reasonable topologies for homeomorphism groups, Houston
J Math. 17 (1991), 367-373.
962 HANS-PETER A. KNZI
[454] _ _ , The open-open topology for funetion spaees, Internat. J. Math. Math.
Sei. 16 (1993),111-116.
[455] _ _ , The regular open-open topology for funetion spaees, Internat. J. Math.
Math. Sei. 19 (1996), 299-302.
[456] _ _ , Some equivalent topologies on homeomorphism groups, Topology Proc.
22 (Summer 1997), 373-383.
[457] Priess-Crampe, S. and Ribenboim, P., Generalized ultrametrie spaees I, 11, Abh.
Math. Sem. Hamburg 66 (1996), 55-73; 67 (1997), 19-31.
[458] Priestley, H.A., Ordered topological spaees and the representation of distributive
lattices, Proc. London Math. Soc. 24 (1972), 507-530.
[459] Raghavan, T.G., On quasi-metrizability, Indian J Pure Appl. Math. 15 (1984),
1084-1089.
[460] Raghavan, T.G. and Reilly,l.L., Metrizability of quasi-metrie spaees, J London
Math. Soc. 15 (1977),169-172.
[461] _ _ , On non-symmetrie topological struetures, Colloquia Math. Soc. Janos
Bolyai, vol. 23, Topology, Budapest (Hungary) 1978, pp. 1005-1014.
[462] _ _ , On the eontinuity of group operations, Indian J Pure Appl. Math. 9
(1978), 747-752.
[463] _ _ , Uniformization of quasi-uniform spaees, Bull. Austral. Math. Soc. 23
(1981),413-422.
[464] _ _ , A new bitopologieal paraeompaetness, J Austral. Math. Soc. (Ser. A)
41 (1986), 268-274.
[465] _ _ , Charaeterization of quasi-metrizable bitopologieal spaees, J. Austral.
Math. Soc. (Ser. A) 44 (1988),271-274.
[466] Redfield, R.H., Ordering uniform eompletions ofpartially ordered sets, Canad.
J Math. 26 (1974), 644-664.
[467] _ _ , Uniformly eonvex totally ordered sets, Proc. Amer. Math. Soc. 51 (1975),
289-294.
[468] Reichei, H.-C., Basis properties oftopologies eompatible with (not necessarily
symmetrie) distanee funetions, Topology Appl. 8 (1978),283-289.
[469] Reilly,l.L., On generating quasi-uniformities, Math. Ann. 189 (1970), 317-318.
[470] _ _ , Quasi-gauge spaees, J. London Math. Soc. 6 (1973), 481-487.
[471] _ _ , Zero dimensional bitopological spaees, Nederl. Akad. Wetensch. Proc.
Sero A 76 =Indag. Math. 35 (1973), 127-133.
[472] _ _ , A generalized eontraetion principle, Bull. Austral. Math. Soc. 10 (1974),
359-363.
[473] _ _ , A note on quasi-metrie spaees, Proc. Japan Acad. 52 (1976), 428-430.
[474] _ _ , A quasi-metric bibliography, The University of Auekland, Dept. Math.
Stat., Report Series No. 269, June 1992, New Zealand.
[475] Reilly, I.L. and Subrahmanyam, P.v., Some fixed point theorems, J Austral.
Math. Soc. (Ser. A) 53 (1992),304-312.
[476] Reilly,l.L., Subrahmanyam, P.V. and Vamanamurthy, M.K., Cauehy sequenees
in quasi-pseudo-metrie spaees, Monatsh. Math. 93 (1982), 127-140.
[477] Reilly,l.L. and Vamanamurthy, M.K., On oriented metrie spaees, Math. Slovaca
34 (1984),299-305.
NONSYMMETRIC DISTANCES AND THEIR ASSOCIATED TOPOLOGIES 963
[521] _ _ , Injective objects and morphisms, in Categorical Topology and its Rela-
tion to Analysis, Algebra and Combinatorics (Prague, 1988), I. Adamek and S.
MacLane (eds.), World Sci. Publishing, Teaneck, NI 1989,394-409.
[522] _ _ , Linearly ordered topologies and quasi-metrization, in Papers on Gen-
eral Topology and Applications, Ninth Summer Conference at Slippery Rock
University, 1993, Annals New York Acad. Sci., vol. 767, 1995,217-227.
[523] _ _ , On even quasimetrization, Proc. Symposium on Categorical Topology,
University of Cape Town 1994, B. Banaschewski, C.R.A. Gilmour and H.
Herrlich (eds.), Dept. Math., University of Cape Town (1999), 225-236.
[524] Salbany, S. and Todorov, T., Nonstandard and standard compactifications of
ordered topological spaces, Topology Appl. 47 (1992), 35-52.
[525] Samuel, P., Ultrafilters and compactifications of uniform spaces, Trans. Amer.
Math. Soc. 64 (1948), 100-132.
[526] Schauerte, A., Functorial quasi-uniformities over partially ordered spaces,
M.Sc. Thesis, Univ. Cape Town, 1988.
[527] Schellekens, M., The Smyth Completion: a Common Foundation for Denota-
tional Semantics and Complexity Analysis: in Proc. MFPS 11, Electronic Notes
in Theoretical Computer Science, vol. 1, 1995. URL: http://www.elsevier.nVlo-
cate/entcs/volume l.html
[528] _ _ , Complexity Spaces Revisited, Proceedings of the 8th Prague topological
symposium, Prague, Simon, Petr. (ed.) Czech Republic, 1996, North Bay, ON:
Topology Atlas (1997) URL: http://at.yorku.ca/p/p/a/a/OO.htm
[529] _ _ , On upper weightable spaces, in Papers on General Topology and Ap-
plications, Eleventh Summer Conference, Southem Maine 1995, Annals New
York Acad. Sei., vol. 806, 1996,348-363.
[530] _ _ , Complexity spaces: Lifting and directedness, Topology Proc. 22
(Summer 1997), 403-425.
[531] _ _ , The correspondence between partial metries and semivaluations; M-
spaces: an abstract study of the lifting process; A characterization of partials
matrizability, preprints.
[532] Schwarz, F. and Weck-Schwarz, S., Is every partially ordered space with a com-
pletely regular topology already a completely regular partially ordered space?
Math. Nachr. 161 (1993), 199-201.
[533] Scott, B.M., Toward a product theory for orthocompactness, Studies in topology,
Proc. Cont Univ. North Carolina, Charlotte, N.C. 1974, Academic Press, New
York 1975, pp. 517-537.
[534] _ _ , More about orthocompactness, Topology Proc. 5 (1980), 155-184.
[535] Seyedin, M., Quasi-uniform spaces and topological homeomorphism groups,
Canad. Math. Bull. 17 (1974), 97-98.
[536] _ _ , R3 -quasi-uniform spaces and topological homeomorphism groups, Proc.
Amer. Math. Soc. 52 (1975), 465-468.
[537] _ _ , On quasiuniform convergence, in General Topology and its Relations to
Modem Analysis and Algebra, IV, Proc. Fourth Prague Topological Symposium,
1976, Part B, Prague 1977, 425-429.
[538] Shchepin, E.Y., On topological products, groups, and a new class of spaces more
general than metric spaces, Sov. Math. Dokl. 17 (1976), 152-155.
966 HANS-PETER A. KNZI
[579] Williams, J., Locally uniform spaces, Trans. Amer. Math. Soc. 168 (1972), 435-
469.
[580] Wilson, W.A., On quasi-metric spaces, Amer. 1. Math. S3 (1931), 675-684.
[581] WindeIs, B., Unifonn Approach Theory, Doctoral Thesis, Universiteit Antwer-
pen, 1997.
[582] _ _ , The Scott approach structure: an extension of the Scott topology for
quantitative domain theory, Acta Math. Hung. 88 (2000), 35-44.
[583] _ _ , On the metrization lemma for uniform spaces, Quaestiones Math., to
appear.
[584] Woods, R.G., Topological extension properties, Trans. Amer. Math. Soc. 210
(1975), 365-385.
[585] Wu, C. and Wu, J., Fuzzy quasi-uniformities and fuzzy bitopological spaces,
Fuzzy Sets Syst. 46 (1992), 133-137.
[586] Wyler, 0., Compact ordered spaces and prime Wallman compactifications, in
Categorical Topology, Heldermann, Berlin 1984, pp. 618--635.
[587] Yajima, Y., A characterization of submetacompactness in terms of products,
Proc. Amer. Math. Soc. 112 (1991), 291-296.
[588] Yang, L., Pointwise quasi-uniformities on completely distributive lattices
(Chinese), 1. Math. Res. Expo. 8 (1988), 187-194.
[589] Zaustinsky, E.M., Spaces with non-symmetric distance, Memoirs Amer. Math.
Soc. 34 (1959).
SUPERCATEGORIES OF TOP AND
THE lNEVITABLE EMERGENCE OF
TOPOLOGICAL CONSTRUCTS
E. LOWEN-COLEBUNDERS
Vrije Universiteit Brussel
Departement Wiskunde
Pleinlaan 2
1050 Brussels, Belgium
AND
R.LOWEN
Universiteit Antwerpen
Departement Wiskunde en Informatica
Groenenborgerlaan 171
2020 Antwerp, Belgium
Contents
1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 971
1.1 Scope of the work ...................................... , 971
1.2 Notation and terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 972
2 The Abundance of Examples from the Beginning of the 20th Century . . . .. 972
2.1 Early approaches to "topology" . . . . . . . . . . . . . . . . . . . . . . . . . . .. 972
2.2 Examples of nontopological structures . . . . . . . . . . . . . . . . . . . . .. 973
3 Generalized Topological Structures Motivated by Examples ............. 974
3.1 Pseudotopological spaces ................................. 974
3.2 Pretopological spaces .................................... 975
3.3 Limit spaces ............................................ 976
3.4 Convergence spaces. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 977
3.5 Moving on ............................................. 977
3.6 Drifting towards uniformity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 979
4 The Introduction of Categorical Ideas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 981
4.1 The early days ......................................... , 981
4.2 Unification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 982
5 Topological Constructs ............................................ 984
5.1 The past. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 984
969
C. E. Aull and R. Lowen (eds.), Handbook ofthe History ofGeneral Topology, Volurne 3,969-1026.
2001 Kluwer Acadernic Publishers.
970 EVA LOWEN-COLEBUNDERS AND BOB LOWEN
1. Introduction
Before embarking on our journey let us point out what we will not be doing,
thereby stifling unjustified "great expectations" in the budo
To start with, we will of course avoid any duplication of topics treated
in the comprehensive artic1e of Herrlich and Strecker, "Categorical Topology
- its Origins as Exemplified by the Unfolding of the Tbeory of Topological
Refiections and Corefiections Before 1971" which appeared in the first volume
of the "Handbook of the History of General Topology". This is not a great
problem since that paper mainly dealt with subcategories of Top whereas our
work will deal with supercategories. We will also not deal with refiective and
corefiective subcategories and generalizations of these concepts in general, but
only in specific instances for specific categories.
We will also avoid duplication of items which were treated in the paper
by Bentley, Herrlich and Husek, ''The historical Development of Uniform,
Proximal and Neamess Concepts in Topology" which appeared in the second
volume of the "Handbook of the History of General Topology". However we
will sometimes be dealing with the categories which were also treated in that
paper, but from another angle.
We will not deal with special properties of morphisms, sinks, sources or
factorization structures, except in those instances where they are required to
define or explain certain concepts which fit in the topics which we do consider.
We will not treat the category Loc of locales, although of course this is also
an important "almost" supercategory of Top, mainly because there is aseparate
paper by Johnstone in this volume which treats this topic in detail.
Some categories will appear in the text in spite of the fact that they are not
supercategories of Top, such as for instance Chy, Met and Unif, simply because
they too are subcategories of supercategories of Top which are being discussed.
Such categories will not be treated in their own right but only in function of the
other categories.
So, what are we going to do? We will try to give a historical overview
of the interplay between, on the one hand, the numerous supercategories of
Top which have been introduced from the beginning of the 20-th century,
and on the other hand, the theory of topological constructs, which emerged
from the need of unification and systematization. We will pay attention to both
directions of the interplay and in our discussion we will follow the historical
path: we start with the first direction, the one that goes from the examples of
nontopological structures and supercategories of Top to the creation of topolog-
ical constructs. Once topological constructs were introduced, there originated
an important opposite direction in the interplay. We will try to show how the
972 EVA LOWEN-COLEBUNDERS AND BOB LOWEN
Notation and terminology has changed over the years; sometimes greatly and
sometimes frequently and sometimes both. We believe that in most cases these
changes were for the better, and hence we will adhere as much as possible to
modem and present-day notation and terminology, certainly for all our defini-
tions and statements of results. We will only refer to, or use, older terminology
whenever this might be required or is interesting for the story.
In all diagrams which involve categories, an r next to an arrow means that the
sma1ler category is bireflectively embedded in the larger one and analogously ac
next to an arrow means that the smaller category is bicoreflectively embedded in
the larger one (see 5.8). An arrow .sIl L - - C(6 stands for an inc1usion functor.
Sometimes an inc1usion functor may also be denoted simply as .sIl - - C(6 We
have not been consistent in the use of arrows and also not in our choices as to
whether smaller or larger categories are depicted on a lower or higher level in
diagrams. Usually such choices were dictated more by, our personal, aesthetical
feeling. It should however always be perfectly clear what is embedded in what.
(CU) cl(0) = 0.
(CI2) VA E 2 x : A C cl(A).
Also, from the early days of Topology, it was clear that several important con-
crete situations simply could not be captured in the setting of topological spaces.
All these early examples, in one way or another, had to do with convergence. Of
course the very early examples were expressed in terms of convergent sequences.
Only later, after Moore and Smith in 1922 [275] had introduced their so-called
Moore-Smith convergence of nets, and after Cartan in 1937 [85] provided the
notion of filter, net and filter versions of these examples became available.
The use of the measure theoretic concept of convergence almost everywhere
dates already from the very beginning of probability and measure theory. This
convergence is nontopological. In 1954 Kowalsky gave a filter description of
this convergence. In the American Mathematical Monthly, in 1966, a proof, of
this by then well known fact, was presented by Ordman [295].
On the set of continuous maps '(X, Y) between topological spaces X and
Y, Hahn in 1921 [157] and Caratheodory in 1929 [83] considered the very
natural notion of continuous convergence of sequences of maps. Again for the
filter version of this convergence, one had to wait until after Cartan's work
of 1937 [85]. A filter version appeared in the work of Bastiani in 1964 [33]
under the name "convergence locale". In 1965 Cook and Fisher [99] proved
974 EVA LOWEN-COLEBUNDERS AND BOB LOWEN
(Ps) If (Oll, x) E q for every ultrafilter Oll finer than a given filter ~ then
(~,x) E q.
In the same paper, Choquet also introduced the notion of a pretopology under
the French name "Pre-topologie", as a pseudotopological structure satisfying
a supplementary axiom, namely that the interseetion of all filters converging
to a given point should also converge to that point. This means that there is
a smallest filter converging to each point. This smallest filter is what is called
the neighborhoodfilter. His equivalent description of a pretopology in terms of
neighborhoodfilters is presented here.
(Pr) VV E "V(x) : x E V.
Choquet also observed that these structures are linked in a direct way to what
he called "Pre-adherences". These coincide with the Cech-c1osures mentioned
earlier, but he does not make explicit reference to the existing notions. Moreover
the correspondence is only established at the level ofthe objects. Choquet further
describes a way of associating a pretopological and a topological modification
with a pseudotopological space. Later in this overview we come back to these
modifications. In his proofs he uses the notion of a grill which was introduced
one year earlier, in 1947, by himself [88]. The interaction between convergence
and topology is the main subject of the paper of Sonner, 1953 [345], which
appeared shortly afterwards.
In 1964, Kent, in order to capture also order convergence within the theory,
proposed yet another weakening of the axioms, thus defining what are known
as convergence structures. Further weakening of the convergence axioms was
considered by Kent in 1964 [230], and by Ghler in his books of 1977-78,
"Grundstrukturen der Analysis", volume I [148] and volume 11 [149]. In the
book of 1974 by Poppe, "Compactness in General Function Spaces" [299] and
also in the papers by Dolecki and Greco of 1986 [115], [116] convergence
structures without axioms are studied.
3.5. MOVING ON
From that time onwards, convergence theory evolved rapidly in several direc-
tions. A first important direction is compactification theory with contributions
from Wyler in 1968 [372], Richardson in 1970 [316], Ramaley and Wyler in
1970 [311], Richardson and Kent in 1972 [317], Cook in 1973 [98], Cochran
and Trail in 1973 [96], Gazik in 1974 [150], Rao in 1974 and 1975 [313], [314],
Vinod Kumar in 1977 [241] and Herrmann in 1979 [195]. In 1979 Kent and
Richardson wrote a survey paper, "Compactifications of convergence spaces"
[233]. Some of the last open problems were finally solved by Butzmann and
Kneis in their paper of 1986 [79].
A second important direction which convergence theory tumed to is that of
analysis. The application of continuous convergence on c1asses of continuous
functions opened many new research areas. In the context of infinite dimen-
sionallocally convex spaces, Bastiani in 1964 [33] considered the "convergence
locale", i.e. continuous convergence, on spaces of continuous maps, linear
continuous maps and differentiable maps. Bastiani thus succeeds in proving
results parallel to those from the classical finite dimensional locally compact
case with function spaces endowed with the compact-open topology. An Ascoli
theorem is proved for locally convex spaces. A great stimulus for the study of
978 EVA LOWEN-COLEBUNDERS AND BOB LOWEN
continuous convergence on function spaces came from the Binz school. Early
material developed in the late sixties by this school appeared in papers of Binz
[45], [46], Keller [226], and Binz and Keller [49]. The main results obtained
by this team are found in the monograph of Binz of 1975 [48]. Application
of continuous convergence also led to other generalized Ascoli theorems as in
the work of Cook and Fischer of 1965 [99] and Poppe of 1965 [298]), Stone
Weierstrass theorems as in the work of Binz of 1970 [47], Feldman of 1971
[131], Butzmann of 1974 [78] and Schroder of 1971 [328] and theory on duality
and reexivity as in the work of Schroder of 1971 [328], Butzmann of 1971 and
1972 [76], [77] and Mller of 1975 [279].
From the very beginning of convergence theory compatibility of conver-
gence structures with group- or vectorspace structures received a lot of attention.
These considerations proved to be very useful for instance for the study of
distributions as in the work of Wloka of 1963 [370].
Around the same period also new examples of concrete nontopological
convergences appeared in relation to analysis. The theory of distributions also
benefitted from the work of Marinescu who introduced a special type of limit
vector spaces, later called Marinescu spaces [266]. Wagner 1966 [361] studied
convergence and Mikusinsky operators.
Quasi-bounded convergence was defined by Frlicher and Bucher in 1966
[143] in relation with differentiation theory. For more details on convergence in
the setting of differentiation we refer to the books by Keller, 1974, [228] and
by Ghler, 1978, [149]. More recent applications in non-linear analysis were
developed by Khrennikov in 1983, [234]. A differentiation theory which was
strongly inuenced by the role of continuous convergence was developed by
Frlicher and Kriegl in the eighties, the main results of which appeared in their
monograph of 1988, ''Linear Spaces and Differentiation Theory" [144] and by
Nel, in the late eighties and early nineties, [290], [292], [293], [294].
In all these aformentioned fields of applications of convergence theory,
compactification theory, analysis, with in particular limit vector spaces with
differentiation theory, the fact that the setting of convergence structures is an
extremely convenient one was the basic motivation for the development of the
general theory. The exact mathematical meaning of these convenience properties
became c1ear later, these will be discussed in Section 7.
Also in more applied fields new examples were studied that had impor-
tant applications. Epi- and hypoconvergence or the more general gamma-
convergence of De Giorgi [151] was introduced in 1977 and has proved to
be useful in many branches of optimization theory. Epi- and hypoconvergence
coincide with closed convergence of the epigraphs and hypographs of the re-
lated functions and their usefulness comes from the fact that together with the
convergence of the functions also the maxima and minima converge.
TOPOLOGICAL CONSTRUCTS 979
Finally we come back to the sequential structures that were introduced at the
time when Topology was born. The introduction of L- and L *-spaces by Frechet
and Urysohn was the starting point for what became a very large research area
devoted to sequential convergence developed mainly by the Cech school. The
area continued to grow until today and has many applications in functional
analysis. The structures of L-spaces and of L *-spaces that were the subject of
these investigations, are not generalized topological. Therefore we do not go
into further details here, but we refer to recent survey papers on this topic and
to the paper by Fric in the first volume ofthis handbook [27, pp. 343-356].
More details on convergence theory can for instance be found in the intro-
ductory chapter ofLowen-Colebunders, 1989, [259]. Recently a historical paper
was written by Ghler 1997 [147] and a monograph on convergence structures
in functional analysis is in preparation by Beattie and Butzmann. A very detailed
bibliographicallist with papers on convergence spaces up to 1980 can be found
in the proceedings of the conferences on convergence spaces in Reno 1976 and
Cameron 1980.
While outside Topology the vast field of generalized topological structures was
growing rapidly, on a parallel track, attention also focused on uniformlike struc-
tures. Uniform spaces were created by Weil in his work of 1937 [367], proximity
spaces were introduced by Efremovic in 1951 [121], [122], contiguities were
introduced in 1959 by Ivanova and Ivanov [217]. We recall the definition of
proximity structures since this will figure in our diagrams and comparisons later
on. Although proximity structures as we know them today were introduced by
Efremovic, similar structures but with weaker axioms were already studied by
Krishna Murti in 1940 [282], Wallace in 1942 [362] and Szymanski in 1941
[350].
The pair (X, ) is called a proximity space and a function f : X ---+ Y between
proximity spaces (X, X) and (Y, y) is called a proximity mapping iffor any
subsets A and B of X : Ax B => f (A) y f (B). The category of proximity
spaces and proximity mappings is denoted Prox.
At some stage in the development of generalized uniform spaces the paths of
Uniformity and Topology crossed. Motivated by the applications of convergence
theory in analysis and in particular by the development of the theory of limit
vectorspaces, an appropriate notion of completeness was needed. Kowalsky in
1954, starting with a limit space, gave an axiomatic description for Cauchy filters
compatible with a given "Limitierung". In 1967 Cook and Fischer, [tOO] intro-
duced uniform convergence structures as a generalization of uniform spaces,
thus providing a setting for the study of uniform continuity. Instead of consid-
ering one filter of entourages as Weil did, they gave axioms for a collection of
filters on X x X.
(UC1) A E;E.
(UC2) If!!:f E ;E and !!:f C '9 then '9 E ;E.
(UC5) If!!:f E ;E and '9 E ;E then !!:f 0 '9 E ;E where !!:f 0 '9 is the filter generated
by all sets FoG for FE !!:f and G E '9.
Soon after that, in 1968 in his paper [227], Keller formulated "Cauchyness"
as a primitive concept, independent of any structure given in advance, thus
introducing what are known as Cauchy spaces. He proved that his set ofaxioms
is necessary and sufficient for a collection of filters to be the collection of Cauchy
filters in some uniform convergence space.
(Chy3) If fIJi E'(6 and cg E'(6 and fIJi and cg have a supremum, then fIJi n cg E '(6.
The pair (X, '(6) is called a Cauchy space. A function f : X --+ Y between
Cauchy spaces (X, '(6x) and (Y, '(6y) is called Cauchy continuous if for each
filter fIJi E '(6 X : f (fIJi) E '(6 y. The category of Cauchy spaces and Cauchy
continuous maps is denoted Chy.
Keller formulates an axiom for limit spaces called axiom P which is a sort
of symmetry condition that allows him to describe every such limit space as a
Cauchy space.
(P) For two points x and y the collections qx and qy of all filters converging
to x and y respectively, coincide whenever they intersect, in other words,
if there exists a filter converging to both x and y then the same filters
converge to x and to y.
structures was presented in a systematic way, first the objects, then the mor-
phisms and then the fundamental constructions. In "Chapitre I" which deals
with topology alone this takes almost 30 pages, which is unusual for a topology
book of that time and even for modem books on topology. In "Chapitre 11" on
uniform spaces about 7 pages are spent on fundamental constructions. In both
cases initial and final constructions are comprehensively treated and all special
cases are given separately: subspaces, products and quotients. The parallel
presentation of the two chapters, makes the similarity between the structures
and the fundamental constructions on them apparent. Almost all the ingredients
are there to deduce that both topological spaces and uniform spaces form special
types of categories wherein it is possible to describe arbitrary initial and final
structures.
To fundamental constructions, in particular initial and final structures, a lot
of attention is also paid by the authors of Cech, ''Topological Spaces", [86],
1966. Stability under fundamental constructions is investigated in a general
setting.
4.2. UNIFICATION
Recognizing the similarities also far beyond just topological spaces and uniform
spaces, several attempts have been made to provide a setting to treat topolog-
ical and uniformlike notions at the same time. New systems ofaxioms were
intended to provide simultaneous generalizations of both topologieal, proximal
and uniform ideas thus providing a unifying view.
One such attempt for presenting a unifying theory for topological, uniform
and proximal structures was presented by Czaszar in 1960 in the book "Fonde-
ments de la Topologie Generale" [102]. The common primitive notion which
Czaszar described is based on transitive relations on subsets, which he called
"structures syntopogenes". In the topological case the relation A < B means
that A is contained in the interior of B, in the case of a proximity, A < B means
that A is not near the complement of B, and in the uniform case A < B means
Bis a uniform neighborhood of A. At the same time he captured examples of
generalized uniform structures. After introducing the appropriate morphisms the
author pays a lot of attention to the fundamental constructions in the setting of
syntopogeneous structures. Remark that syntopogeneous structures on the same
underlying set can be equivalent (in the sense that the identity is an isomorphism)
without being equal. This means that this example does not fulfil all the required
axiom for being a "topological construct" as we will see later in chapter 5.
Another attempt at unification was made by Doicinov in 1964 [114]. He
constructed a type of structure, appropriately called a "supertopology", which
is a common generalization of topologies and proximities. In the paper by
TOPOLOGICAL CONSTRUCTS 983
(M3) If .stl corefines ~ (i.e., for each A E .stl there exists a B E ~ such that
B C A) and.stl E .M then also ~ E .M.
(M4) If .stl and ~ are collections of subsets of X such that .stl U ~ E .M then
either .stl E .M or ~ E .M.
The pair (X,.M) is called a merotopic space and the members of.M are usually
referred to as micromeric collections. A function f : X ---+ Y between
merotopic spaces (X, .Mx) and (Y, .My) is called uniformly continuous if for
each micromeric collection .stl E .Mx : f(.stl) E .My. The category consisting of
merotopic spaces and continuous maps is denoted Mer.
As it turned out the setting of merotopic spaces is of such generality that
every symmetric convergence [28, pp. 577-630] can be described as a mero-
topic space. Katetov has shown how the embedding is constructed. He used
convergent nets for that purpose. In 1975 Robertson translated Katetov's results
into the setting of filter convergence. Katetov paid a lot of attention to a subc1ass
of objects defined as filtermerotopic spaces, these are merotopic spaces having
a base consisting of filters. He proved that in the setting of filterspaces nice
function spaces exist.
One of the pioneers of using categorical methods in General Topology is
Herrlich. Under the influence of his 1968 lecture notes on ''Topological reflec-
tions and coreflections" Categorical Topology started with the investigation of
special objects and subcategories of the categories Top and Haus. In 1974 at
the same time he published his papers introducing the nearness concept, there
were his contributions on Topological Functors and cartesian c10sed topolog-
ical categories [172], [173], [171]. So his work on nearness spaces in these
984 EVA LOWEN-COLEBUNDERS AND BOB LOWEN
s. Topological Constructs
5.1. THEPAST...
In the sixties and beginning of the seventies several authors started formal-
izing these ideas.
1. If cP and 1/1 are Cf6-morphims such that Dcp = D1/I, Rep = R1/I and Tcp = T1/I
then cP = 1/1.
4. For each 2t'-object A the c1ass T- 1 (A) is a completely ordered set with
respect to the order: X :s Y if there exists cP E hom~ (X, Y) for which
Tcp = idA.
ep E hom~ (sup Depi, sup RCPi) and 1/1 E hom~ (~nf Depi, ~nf RCPi)
iEI iEI lEI lEI
In that same paper Rusek also shows that conditions 4 and 5 can be replaced
by the following condition:
4'. If F is a presheaf in Cf6 and if the inductive and projective limits, ind(T F)
and proj(T F) exist in 2t' then the inductive and projective limits, ind(F)
and proj(F) exist in Cf6 and moreover Tind(F) = ind(T F) and Tproj(F) =
proj(T F).
986 EVA LOWEN-COLEBUNDERS AND BOB LOWEN
Antoine then coneludes that this last property is equivalent to the existence
of arbitrary set-indexed final structures. However it was later observed that
this equivalence can only be proved if arbitrary elass-indexed initial (or final)
structures are assumed to exist. Apart from this however, Antoine's concept
comes very elose to what today we call topological constructs.
1. Given categories <:g and ge and a functor T : <:g ---+ ge two new categories
are defined. First, the category geiT is defined with objects those of<:g and
with morphisms between X and Y the ge-morphisms between T X and TY.
The morphisms of geiT are called T-premorphisms. A T-premorphism a
is called a T -morphism if there exists a <:g-morphism q; such that T q; = a.
Second, the set of T-morphisms determines a subcategory ofgelT which is
denoted <:g IT .
TOPOLOGICAL CONSTRUCTS 987
Roberts further says: "It should be clear that the indexing is merely a
matter of notational convenience and that in the above definition it is only
the cardinality ofthe index set which is at issue. In deciding whether T is initial
it is enough to consider index sets belonging to a sufficiently large universe" .
2. Given a top structure s on C{&, a top category C{&S is defined with objects all
pairs (E, ~) where E is a C{&-object and where ~ E sE, and with morphisms
between (E, ~E) and CF, ~F) those C{&-morphisms f : E ~ F such that
~E C {y E SE I (Sf)(y) E ~F}.
In [374] Wyler gives an even more general definition which we will not recall
here.
Hoffmann starts the introduction of his thesis by saying: "Diese Arbeit ging
aus von der Definition der Initial- und Finaltopologie, wie sie im Werk von
Bourbaki gegeben sind".
TOPOLOGICAL CONSTRUCTS 989
R = inf{K E ~ I T(K) = W}
and
S = inf{K E ~ I T(K) = V}
3. Smallness conditions.
Things finally came to rest with now generally accepted definitions and termi-
nology as they can be found in the book of Adamek, Herrlich and Strecker,
where the three principles above are translated in a simple and precise1y fitting
way.
First we have to say when a category is called a construct and what we mean
by initial lifts.
1. U is topologieal.
3. (<(6, U) is amnestie.
2. For every set X the U -}ibre of X is a complete lattice. The smallest element
is the so-called indiscrete structure and the largest element is the so-called
discrete structure.
3. U has a left-adjoint right inverse which takes any set to the discrete object
"on that set" and it also has a right-adjoint right inverse which takes any
set to the indiscrete object "on that set".
7. For morphisms in (Cf6, U), being an isomorphism, being initial and bijective
or being final and bijective are equivalent properties.
11. The projective objects in (Cf6, U) are precisely the discrete objects.
12. The injective objects in (Cf6, U) are precisely the indiscrete objects with
nonempty underlying sets.
TOPOLOGICAL CONSTRUCTS 993
4. lf.'il contains an object with a nonempty underlying set then .'il is corejiective
in ~ if and only if it is bicorejiective in ~ if and only if it is closed under
the formation of coproducts and quotients.
994 EVA LOWEN-COLEBUNDERS AND BOB LOWEN
From the moment that topological constructs were firmly established, a system-
atic investigation of topological constructs and their relationship to each other
became possible.
The fact that all arrows in the diagram are reflective embeddings was some-
how known from the beginning, although not the categorical aspects and impli-
cations. The reflections were usually called "modifications", and it was known
that if .stl is the reflective subconstruct of ~ then the .stl-modification (or re-
flection) of a ~-structure on a given set was obtained by taking the finest
.stl-structure coarser on that same set: a fact which is implied by reflectivity,
but not conversely!
Also compactification theory, largely in Conv, benefitted from categorical
considerations.
In a paper of 1987 Tozzi and Wyler took a second look at the definition
of supertopologies given by Doicinov in [114]. They noticed that in order to
embed the theory of supertopologies in the framework of topological constructs
the original definition had to be slightly modified. In the paper they consequently
wrote [354] they say: "-Yler had the possibility to discuss this problem with
Doicinov when he was visiting him in the University ojPittsburgh and he agreed
with the proposal". We give here the adapted definition. JF(X)* stands for the
set of all filters on X, inc1uding the degenerate filter ~(X).
(ST5) If A E .M and U E () (A) then there exists aVE () (A) such that
V E ()(B) for each B E .M for which B C V.
~I/
STop
Wyler continued the line of thought of Cook and Fisher and slightly modified
the axioms of uniform convergence structures in 1971 [374]. The purpose for
this change in the axioms is to make sure that nice function spaces exist, and
so it is motivated by categorical observations. We'll come back to that aspect in
the section on convenience properties, and here we include Wyler's definition.
ULim
So, modulo the R-property, Chy, ULim and Fil are supereategories of Top
via the embeddings.
We now eome baek to the topological eonstruet of neamess spaees. Its relation
to the eategories of symmetrie topological, uniform and merotopic spaees is
c1arified in the diagram below. We reeall that a topological spaee is ealled
symmetrie if for any pair of points x, y E X we have that x E y if and only if
y EX. The subeategory of symmetrie topologieal spaees is denoted Tops
Unif
Near
Mer
A new topologieal eonstruet was also ereated to solve the bad behaviour of
a forgetful funetor: namely the forgetful funetor from pMet OO to Top which
TOPOLOGICAL CONSTRUCTS 997
does not preserve initial structures. pMetOO stands for the topological construct
consisting of extended pseudometric spaces and nonexpansive maps. At the
same time this topological construct turned out to be a remarkable supercategory.
It was introduced by Lowen in 1988-89 [252], [253].
The pair (X, 8) is called an approach space and the function 8 is called a distance.
A function f : X -+ Y between approach spaces (X, 8x) and (Y, 8y) is called
a contraction if for each x E X and for each A C X : 8y(f(x), f(A ~
8x(x, A). The category of approach spaces and contractions is a topological
construct, denoted Ap.
Probably the most surprising feature of Ap is that it contains both Top
and pMet OO as full and isomorphism-c1osed subconstructs. Moreover there are
no less than 8 equivalent, but conceptually and technically totally different,
structures which define an approach space, as many as for topological spaces
[255]. pMetOO is bicorefiectively embedded in Ap and Top is even birefiectively
and bicorefiectively embedded as indicated in the diagram below.
Top pMetOO
~
Ap
The methods of eategorieal topology that were developed first in the eontext
of Top and its subeonstruets, beeame a useful tool in the study of general-
ized struetures. Isomorphie eategories were deteeted sueh as Mer [224] and
S - Near [173] on the one hand and Fil [224] and Grill [322] on the other
hand.
Embedding topologieal eategories in a eommon supercategory led to a bet-
ter understanding of the interrelation between them. Comparison of canonical
eonstructions became possible.
998 EVA LOWEN-COLEBUNDERS AND BOB LOWEN
So while the abundance of examples was the original trigger for the creation
of the theory of topological constructs, the application of the theory that evolved
largely contributed to a better understanding of the existing examples. At the
same time these ideas inspired the creation of new examples, where some arose
as initially c10sed extensions of existing ones [41], [42], [70], [110], [164],
[165], [176], [180], [184], [203], [211], [259], [267], [268], [306], [214].
7. Convenience Properties
Once the similarities between the examples were recognized and formulated in
a categorical setting attention went to the differences and their expression in
categorical terms. Clearly, some of the examples behave better than others. For
instance, in Conv and PsTop nice function spaces exist, namely the hom-sets
of continuous functions endowed with continuous convergence. Similarly FiI
too has ''nice'' function spaces [224]. For well-fibred topological constructs the
existence of "nice" function spaces is equivalent to the Cartesian c10sedness of
the category. In some of the c1assical constructs such as Top, PrTop and Unif
nice function spaces do not exist [25], [186].
The lack of natural function spaces in a topological construct which is not carte-
sian c10sed such as e.g. Top, has long been recognized as an awkward situation
for various applications. In 1967 Steenrod published a paper [348], suggesting
to replace Top by the subcategory of all compact1y generated Hausdorff spaces
for use in homotopy theory and topological algebra. In their paper from 1971
Dubuc and Porta show the importance of cartesian c10sedness in the setting of
topological algebra, in particular Gelfand duality theory. In infinite dimensional
differential calculus the advantage of working in a cartesian c10sed setting has
convincingly been demonstrated by several authors. The first papers to go in
that direction were by Bastiani [33], Frlicher and Bucher [143] and Seip [337].
We refer also to the more recent work of Frlicher and Kriegl, see e.g. [142],
[239], [238], wherein they present overwhelming evidence ofthe importance to
have canonical function spaces in their setting. In even more recent work, also
on infinite dimensional calculus, Nel comes to the same findings in a serles of
papers between 1990 and 1998, [291], [294], [274] (with Monadi), [273] (with
Min), [246] (with Lee). Also in the first parts of an as yet unfinished manuscript
"Introduction to Categorical Methods parts I-Irr" [292], [293] Nel pays much
attention to cartesian c10sedness once he embarks on the development of bis
nonnormable linear analysis.
TOPOLOGICAL CONSTRUCTS 999
A x [A,B] ~ B
idAXf*i /t
AxC
None of the topological constructs Top, PrTop, Mer, Near, Ap, pMetOO
or Unif are cartesian c1osed.
The constructs PsTop, Lim, Conv and Chy are cartesian c1osed. For most of
these constructs this positive resuit was not known when they were introduced,
it was only discovered several years later. That PsTop is cartesian c10sed is due
to Machado in his paper from 1973 [263] which is 25 years after its introduction
by Choquet in 1948. That Lim is cartesian closed was observed by Binz in [49],
12 years after the introduction of limit spaces by Kowalsky in 1954. That Conv
is cartesian c10sed was shown by Nel in [287], more than 10 years after Kent
had introduced Conv. The cartesian c10sedness of Chy was shown by Bentley,
Herrlich and Lowen-Colebunders in 1987 in their paper [40], 19 years after the
introduction ofCauchy spaces by Keller in 1968. That ULim is cartesian c10sed
was shown by Lee in her paper [244] of 1976,9 years after the introduction of
uniform convergence spaces by Cook and Fischer in [100] and 5 years after the
modification ofWyler published in his paper [374] of 1971.
In 1974 Edgar introduced a cartesian c10sed supercategory of Top, which he
called the category of L *-spaces [119]. The definition of an L *-space is obtained
by replacing the sequences in the definitions of Frechet and Urysohn by nets. It
is however immediately seen that this category is isomorphic to PsTop where
the axioms are formulated in terms of nets rather than in terms of filters. So this
result can be derived from the one for PsTop.
1000 EVA LOWEN-COLEBUNDERS AND BOB LOWEN
(Ca3) For any pair of filters '!J and'9 : (A.('!J n '9) = maxA.('!J), 1..('9.
The pair (X, 1..) is called a convergence approach space and the function I..
is called a limit operator. A function 1 : X ~ Y between convergence
approach spaces (X,A.x) and (Y,A.y) is called a contraction if for each filter
'!J : A.y(f('!J 0 1 ~ A.x('!J). The category of convergence approach spaces and
contractions is a topological construct, denoted CAp.
Ap is embedded as a full subconstruct of CAp. The relationship between a
distance and a limit operator is given by the formulas: A.('!J)(x) = sUPAesec(~)
/S(x, A) and /S(x, A) = inf{A.('!J)(x) IA E '!J}.
The second one is the construct of so-called pseudo approach spaces which
is a superconstruct of both PsTop and pMetoo . We will come back to this
construct in the following section.
Cartesian c10sed topological constructs have many supplementary nice prop-
erties. We only mention a few ofthem [170].
For certain purposes there is another convenience property which is more in-
teresting than cartesian closedness. It was shown by Herrlich, Salicrup and
Vazquez in 1979 [191] that the investigation of connectedness in a topologi-
cal construct benefits from the following property which is independent from
cartesian closedness.
None of the constructs Top, Chy, Near, Ap or Unif are extensional [177],
[180], [181].
The constructs PrTop, Lim, Conv and CAp are extensional [180], [181],
[186].
Extensional topological constructs too have nice properties.
The last two properties are in fact equivalent to being extensional as was
shown by Herrlich in 1988 [181].
1002 EVA LOWEN-COLEBUNDERS AND BOB LOWEN
The topological universes, the definition of wbich follows below, basically first
appeared under the name concrete quasi-topoi. This is a tenn wbich is still often
used today. However in the work ofDubuc of 1979 [117], where this tenn was
introduced, no fibre-smallness was required. Later these constructs were called
strongly topological by Herrlich in his 1983 work [177]. The tenn topological
universe is due to Nel, and it first appeared in bis paper [290] from 1984.
One of the reasons that topological universes are important and interesting
is given in the following equivalent characterization from Herrlich [180].
Xi
- hi Yi
gi!
f
!/i
X ~
Y
None of the constructs Top, Chy, Near, Ap or Unif are topological uni-
verses.
The constructs Lim [296], Conv [180], CAp [249], Grill [42], PsTop [375]
are topological universes. Motivated by the fact that Chy is not a topological
universe, Preu in [308] introduced a superconstruct SChy which is a topologial
universe.
course it are precisely those extensions which are as small as possible which
are of special interest. Such extensions are called hul1s.
A fundamental paper on cartesian c10sed topological hulls is the paper by
Herrlich and Nel from 1977, [187] where they showed that any construct with
finite products which can be fully embedded in some cartesian c10sed topological
construct can also be embedded in a smallest such construct. This of course
made it possible to envisage the construction of many more cartesian c10sed
topological hulls since quite a few topological constructs were known to be
subconstructs of other, also well known, cartesian c10sed topological constructs.
One of the first constructed such hulls was the cartesian c10sed topological
hull of Top. This was achieved in aseries of papers by Antoine, Machado and
Bourdaud in the period 1966-1976, [24], [22], [263], [62] and [63]. Antoine
gave the start with his description of the objects of the cartesian c10sed topo-
logical hull as those which are initial for a particular source, but he did not
give an internal description of these objects. Machado made a first step towards
this internal description, especially in the case of Hausdorff spaces. Bourdaud
finally rounded the internal description off with the elegant characterization
given below. The ideas that came out of these papers can not be overestimated,
they were the source for many generalizations and for developing techniques to
find cartesian c10sed hulls of other constructs.
The cartesian c10sed topological hull of Top is the construct EpiTop of so-
called epi-topological spaces (also called Antoine spaces). That means it is the
full subconstruct of PsTop having as objects those spaces (X, q) which satisfy
the two supplementary conditions (EpiTl) and (EpiT2) below. We denote by
c1q the c10sure operator of the Top-bireflection of q. Further, for a filter I?; on
X we denote by Liml?; the set of all points x E X such that (I?;, x) E q and by
I?;* the filter generated by all the sets {x E X I c1q (x) n F i= 0}, F E I?;.
N ext it was shown by Bourdaud in 1975-76 that the cartesian c10sed topolog-
ical hull of the construct PrTop is the construct PsTop. Also in [63] Bourdaud
proves that the cartesian c10sed topological hull of the construct of all Tychonoff
spaces is exactly the construct of c-embedded convergence spaces [48].
The cartesian c10sed topological hull of Unif was constructed by Adamek
and Reiterman in 1981 in [11]. It is the construct of so-called bomological
uniform spaces, denoted BUnif. First recall that a bomology on a set is a
collection ~ of subsets of X such that all finite subsets are in ~ and such that ~ is
c10sed under the formation of subsets and of finite unions. The elements of ~ are
called bounded sets. Uniformities are characterized by their generating farnilies
ofpseudometrics. Now a bomological uniform space is a tripie (X, 9ll, ~) where
9ll is a uniformity on X and ~ is a bomology on X, satisfying the following
properties:
(BU2) ~ contains every set M with the property that for each pseudometric
d E 9ll and for each B > 0, there exists a bounded set B such that
d(x, B) < B for all x E M.
In the original paper by Adamek and Reiterman a third condition was men-
tioned, but recently, while generalizing the result of Adamek and Reiterman it
was shown by Nauwelaerts that this condition is redundant [286].
In 1992 Bentley and Lowen-Colebunders [43] give adescription of the carte-
sian c10sed topological hull of the construct of completely regular filterspaces.
TOPOLOGICAL CONSTRUCTS 1005
where lF(X) stands for the set of all filters on X and where 5"r stands for the
topology on [0, 00] generated by all half-open intervals extending to 00.
It was shown by Machado in 1973 that Top [263] is finally dense in PrTop
and by Bourdaud in 1975 [62] that the one-point extension of the Sierpinski
space is initially dense in PrTop. In 1988 Herrlich deduced from this that
PrTop (introduced by Choquet in 1948) is the extensional topological hull of
Top [181].
In 1989 Lowen-Colebunders and Lowen described the extensional topolog-
ical hull of Ap. It is the construct PrAp of pre-approach spaces, Le. the full
subconstruct of CAp with those objects for which the limit operator fulfils the
following strengthening of (Ca3).
1006 EVA LOWEN-COLEBUNDERS AND BOB LOWEN
In 1993 the extensional topological hull of the construct Chy was described
by Alderton and Schwarz in [18].
TUH(~) = CCTH(ETH(~
This result is due to Schwarz [334]. It was observed by Schwarz in 1989 that
the order of taking hulls on the right-hand side can not be interchanged [334].
In general ETH(CCTH(~ is strict1y smaller than CCTH(ETH(~ and need
not be cartesian c1osed.
The relation among the various hulls of a topological construct ~ (provided
they exist) is given in the following diagram, see the paper by Schwarz [334].
TUH(~) = CCTH(ETH(~
----------- ETH(CCTH(~
ETH(~)
~I CCTH(~)
~/ ~
TOPOLOGICAL CONSTRUCTS 1007
r r
Top
(PsA) For any filter CJF on X : )"(CJF) = sup{),,(CU) I CUultra, CJF c CU}.
This subconstruct is called PsAp, and its objects are called pseudo-approach
spaces.
The relationship among the various hulls of Ap is depicted in the following
diagram.
1008 EVA LOWEN-COLEBUNDERS AND BOB LOWEN
r r
r r
Ap
(1) If (Y, "1/') E A and (Y', "1/") < (Y, "1/') then (Y', "1/") E A.
(2) For any two spaces (Y, "1/'), (Y', "1/") in A there exists aspace (y II , "1/'")
in A such that both (Y, "1/') < (Y", "1/'") and (Y', "1/") < (Y", "1/'").
(3) For each x E X there exists aspace (Y, "1/') E A such that x E Y.
Now if we consider two sets X and X' both equipped with a filter of semi-
uniform spaces A and A' respectively, then a map ! : X ~ X' is a morphism
if for each (Y, "1/') E A there exists a (Y', "1/") E A' such that ! (Y) C Y'
and such that !IY : (Y, "1/') ~ (Y', "1/") is uniformly continuous. These objects
together with these morphisms defines a topological construct which we will not
designate since we now immediately consider a subconstruct. A semi-uniformity
is called simple if it is generated by a single entourage of the diagonal. A filter of
semi-uniformities is called simple if it has a basis of simple semi-uniformities.
This means, (X, A) is simple if for each (Y, "1/') E A there exists a simple
semi-uniform space (Y', "1/") E A such that (Y, "1/') < (Y', "1/"). Finally a filter
of semi-uniformities is called saturated if it is an intersection of simple filters
of semi-uniformities. The full subconstruct of the aforementioned topological
construct with objects the saturated filters of semi-uniformities is the topological
universe hull SF of Unif.
In the following table we give the present day situation concerning the
various hulls of the "basic" topological constructs which we have introduced in
the foregoing sections. An equality sign in a column means that the construct of
that row equals its respective hull, i.e. is already respectively cartesian c1osed,
extensional or a topological universe. If only a reference is given this means
that the hull has been described but that as yet no name has been given to
it. The first block contains genuine supercategories and the second block "al-
most" supercategories of Top. The third block contains subconstructs of various
supercategories of Top.
1010 EVA LOWEN-COLEBUNDERS AND BOB LOWEN
Near ? ? ?
Mer ? = [180] ?
ULim = [244] ? ?
Prox (*) ? ?
References
[30] Banaschewski, B., Aremark on extensions of Hausdorff spaces, Gen. Top. Appl.
4 (1974), 283-284.
[31] _ _ , Essential extensions of To spaces, Gen. Top. Appl. 7 (1977),233-246.
[32] Banaschewski, B. and Bruns, G., Categorical characterization of the MacNeille
completion, Ann. 0/ Math. 18 (1967), 369-377.
[33] Bastiani, A, Applications differentiables et varietes differentiable de dimension
infinie,1. Analyse Math. 13 (1964), 1-114.
[34] Beattie, R and Butzmann, H.-P., Convergence Structures in Functional Analysis,
manuscript for a book, 2000.
[35] Bentley, H.L., T -categories and some representation theorems, Portug. Math.
32 (1973), 201-222.
[36] Bentley, H.L. and Herrlich, H., Extensions of topological spaces, in Proc.
Memphis Top. ConJ, 1976, 129-184.
[37] _ _ , Completion as reflection, Comment. Math. Univ. Carolinae 19 (1978),
541-568.
[38] Bentley, H.L., Herrlich, H. and Husek, M., The Historical Development 0/
Uniform, Proximal and Neamess Concepts in Topology, Kluwer Academic
Publishers 1998,577-630.
[39] Bentley, H.L., Herrlich, H. and Lowen, R, Improving constructions in topology,
in Category Theory at Work, Heldermann Verlag 1991, 3-20.
[40] Bentley, H.L., Herrlich, H. and Lowen-Colebunders, E., The category of Cauchy
spaces is cartesian c1osed, Topology Appl. 27 (1987), 105-112.
[41] _ _ , Convergence,1. Pure Appl. Algebra 68 (1990),27-45.
[42] Bentley, H.L., Herrlich, H. and Robertson, W.A, Convenient categories for
topologists, Comment. Math. Univ. Carolinae 17 (1976) 207-227.
[43] Bentley, H.L. and Lowen-Colebunders, E., On the cartesian c10sed hull of the
category of completely regular filter spaces, Cahiers Top. Geom. Diff. Categ. 33
(1992),345-360.
[44] Bentley, H.L. and Naimpally, S.A, Wallmann Tl-compactifications as epireflec-
tions, Gen. Top. Appl. 4 (1974), 29-41.
[45] Binz, E., Ein Differentierbarkeitsbegriff in limitierten Vektorrumen, Comment.
Math. Helvetica 41 (1966), 137-156.
[46] _ _ , Bemerkungen zulimitierten Funktionenalgebren, Math. Ann. 175 (1968),
169-184.
[47] _ _ , Notes on a characterization offunction algebras, Math. Ann. 186 (1970),
314-326.
[48] _ _ , Continuous convergence on C(X), vol. 469, Lecture Notes Math.
Springer Verlag 1975.
[49] Binz, E. and Keller, H.H., Kunktionenrume in der Kategorie der Limesrume,
no. 383 in Sero AI. Ann. Acad. Sci. Fennicae, 1966.
[50] Birkhoff, G., Moore-Smith convergence in general topology, Ann. 0/ Math. 38
(1937),39-56.
[51] _ _ , Lattice Theory, vol. XXV, Americam mathematical Society Colloquium
Publications. AMS 1967.
[52] Booth, P. and Tillotson, J., Monoidal c1osed, cartesian c10sed and convenient
categories of topological spaces, Pacific 1. Math. 88 (1980), 35-53.
TOPOLOGICAL CONSTRUCTS 1013
[53] Brger, R., Connectivity spaces and component categories, Heldermann Verlag
1984, 71-89.
[54] Bourbaki, N., Topologie Generale, chapter 9, Utilisation des nombres reels en
topologie generale, Hermann, Paris 1958.
[55] _ _ , Topologie Generale, chapter 3, Groupes topologiques, 4. Nombres reels.
Hermann, Paris 1960.
[56] _ _ , Topologie Generale, chapter 5, Groupes a un parametre, 6, Espaces
numeriques et espaces projectifs, 7, Les groupes additifs ~n, 8, N ombres
complexes. Hermann, Paris 1963.
[57] _ _ , Topologie Generale, chapter 1, Structures topologiques, 2, Structures
uniformes, Hermann, Paris 1965.
[58] _ _ , Topologie Generale, chapter 10, Espaces fonctionnels. Hermann, Paris
1967.
[59] _ _ , Theorie des Ensembles, Hermann, Paris 1970.
[60] Bourdaud, G., Structures d' Antoine associees aux semi-topologies et aux
topologies, Compt. Rendus Aead. Sei. Paris 279 (1974), 591-594.
[61] _ _ , Deux characterizations des c-espaces, Compt. Rendus Aead. Sei. Paris
281 (1975), 313-316.
[62] _ _ , Espaces d' Antoine et semi-espaces d' Antoine, Cahiers Top. Geom. Diff.
Categ. 16 (1975), 107-134.
[63] _ _ , Some cartesian c10sed topological categories of convergence spaces,
Leeture Notes Math. 540 (1976), 93-108.
[64] Brace, J.W., The topology of almost uniform convergence, Pae. J Math. 9
(1959),643-652.
[65] Brandenburg, H. and Husek, M., Aremark on cartesian c1osedness, in Category
Theory, Proe. Gummersbaeh Con! 1981, Springer-Verlag 1982,33-38.
[66] Brock, P. and Kent, D.C., Approach spaces, limit tower spaces and probabilistic
convergence spaces, Applied Categ. Struet. 5 (1997), 99-110.
[67] _ _ , On convergence approach spaces, Applied Categ. Struct. 6 (1998), 117-
125.
[68] Brmmer, G.c.L., A eategorieal study of initiality in uniform topology, Ph.D.
thesis, University of Cape Town, 1971.
[69] _ _ , Topological functors and structure functors, Lecture Notes Math. 540
(1976), 109-135.
[70] _ _ , Topological categories, Topology Appl. 18 (1984), 27-41.
[71] Brmmer, G.c.L. and Giuli, E., A categorical concept of completion of objects,
Comment. Math. Univ. Carolinae 33 (1992),131-147.
[72] Brmmer, G.c.L. and Hoffmann, R.-E., An external characterization of
topological functors, Leeture Notes Math. 540 (1976), 136-151.
[73] Bruns, G., Die struktur der univerzweigten punktalen raumtypen, Math.
Japoniea 4 (1957), 123-132.
[74] Bruns, G. and Schrnidt, J., Die puntalen typen topologischer rume, Math.
Japoniea 4 (1957),133-177.
[75] Burroni, A., T-categories, Cahiers Top. Geom. Diff. Categ. 12 (1971), 215-321.
[76] Butzmann, H.-P, Dualitten in 't6 c (X), Ph.D. thesis, University of Mannheim,
1971.
1014 EVA LOWEN-COLEBUNDERS AND BOB LOWEN
[202] Hong, Y.H., On initially structured functors, J. Korean Math. Soc. 14 (1978),
159-165.
[203] Hunsaker, W.H. and Sharma, P.L., Proximity spaces and topological functors,
Proc. Amer. Math. Soc. 45 (1974), 419-425.
[204] Husek, M., Generalized proximity and uniform spaces I, Comment. Math. Univ.
Carolinae 5 (1964), 247-266.
[205] _ _ , Generalized proximity and uniform spaces 11, Comment. Math. Univ.
Carolinae 6 (1964), 119-139.
[206] _ _ , S-categories, Comment. Math. Univ. Carolinae 5 (1964), 37-46.
[207] _ _ , Categoricalmethods in topology, inProc. Symp. Gen. Top. Appl. Prague,
1966, 190-194.
[208] _ _ , Remarks on reflections, Comment. Math. Univ. Carolinae 7 (1966),
249-259.
[209] _ _ , Categorial connections between generalized proximity spaces and com-
pactifications, in Proceedings Symposium Berlin 1967 on Extension Theory of
Topological Structures, 1967, 127-132.
[210] _ _ , Construction of special functors and its applications, Comment. Math.
Univ. CaroUnae 8 (1967), 555-566.
[211] _ _ , Lattices of reflections and coreflections in continuous structures, Lect.
Notes Math., Springer 540 (1976), 404-424.
[212] _ _ , History and development of Hausdorff's work in extension in metric
spaces, Recent Development ofGen. Top. and its Appl., Intern. Con! in memory
ofF. Hausdorff(J868-1942}, BerUn 1992, Akademie Verlag 1992, 160-169.
[213] Husek, M. and Pumpln, D., Disconnectedness, Quaest. Math. 13 (1990), 449-
459.
[214] Husek, M. and Tozzi, A., Generalized reflective cum coreflective classes in Top
and Unif, Applied Categ. Struct. 4 (1996), 57-68.
[215] Isbell, J.R., Some remarks conceming categories and subspaces, Canad. 1. Math.
9 (1957), 563-577.
[216] _ _ , Uniform Spaces, AMS Providence, RI, 1964.
[217] Ivanova, V.M. and Ivanov, A.A., Contiguity spaces and bicompact extensions,
Isvestia Akad. Nauk. SSSR 23 (1959), 613-634.
[218] Kannan, V., Reflexive cum coreflexive subcategories in topology, Math. Ann.
195 (1972), 1972.
[219] Katetov, M., Measures in fully normal spaces, Fund. Math. 38 (1951), 73-84.
[220] _ _ , On real-valued functions in topological spaces, Fund. Math. 38 (1951),
85-91.
[221] _ _ , On two results from general topology (Czech), Cas. pest. mat. 82 (1957),
367.
[222] _ _ , ber die Berhrungsrume, Wiss. Zeitschrift Humboldt Univ. Berlin,
Math. Reihe 9 (1959), 685-691.
[223] _ _ , Allgemeine Stetigkeits strukturen, in Proc. Int. Congr. Math. Stockholm,
1963,473-479.
[224] _ _ , On continuity structures and spaces ofmappings, Comment. Math. Univ.
Carolinae 6 (1965), 257-278.
1020 EVA LOWEN-COLEBUNDERS AND BOB LOWEN
[249] Lowen, E. and Lowen, R, A quasitopos containing CONV and MET as full
subcategories, Intern. J Math. Math. Sei. 11 (1988),417-438.
[250] _ _ , Topological quasitopos hulls of categories containing topological and
metric objects, Cahiers Top. Geom. Dijf. Categ. 30 (1989), 213-228.
[251] Lowen, E., Lowen, R. and Verbeeck, c., Exponential objects in the construct
PRAP, Cahiers Top. Geom. Dijf. Categ. 38 (1997), 259-276.
[252] Lowen, R, Kuratowski's measure of noncompactness revisited, Quarterly J.
Math. Ox/ord 39 (1988), 235-254.
[253] Lowen, R, Approach Spaces: a common supercategory of TOP and MET,
Math. Nachrichten 141 (1989), 183-226.
[254] _ _ , Cantor-connectedness revisited, Comment. Math. Univ. Carolinae 33
(1992),525-532.
[255] _ _ , Approach Spaces: the Missing Link in the Topology-UniJormity-Metric
Triad, Oxford Mathematical Monographs, Oxford University Press 1997.
[256] Lowen, R and Robeys, K., Compactifications of products of metric spaces and
their relation to Smirnov and Cech-Stone compactifications, Topology Appl. SS
(1994), 163-183.
[257] Lowen,R and Sioen, M., A smaIl note on separation in AP. to appear.
[258] Lowen-Colebunders, E., Uniformizable Cauchy spaces, Intern. J. Math. Math.
Sei. 5 (1982), 513-527.
[259] _ _ , Function Classes 0/ Cauchy Continuous Maps, Marcel Dekker, New
York 1989.
[260] Lowen-Colebunders, E., Lowen, R and Nauwelaerts, M., The cartesian c10sed
hull of the category of approach spaces, Cahiers. Top. G60m. Diff. Categ., to
appear.
[261] Lowen-Colebunders, E. and Sonck, G., Exponential objects and cartesian
c10sedness in the construct Prtop, Applied Categ. Struct. 1 (1993), 345-360.
[262] _ _ , On the largest coreflective cartesian c10sed subconstruct of Prtop,
Applied Categ. Struct. 4 (1996), 69-79.
[263] Machado, A., Espaces d' Antoine et pseudo-topologies, Cahiers Top. Geom.
Dijf. Categ. 14 (1973), 309-327.
[264] Malan, S.W., Relational structures, Ph.D. thesis, Rand Afrikaans Univ. Johan-
nesburg 1976.
[265] Marinescu, G., Topologische und pseudotopologische Vektorrume, Ed. Acad.
Rep. Pop. Romine 1959.
[266] _ _ , Espaces Vectoriels Pseudotopologiques et Theorie des Distributions,
Deutscher Verlag der Wissenschaften 1963.
[267] Marny, T., On epireflective subcategories of topological categories, Topology
Appl. 10 (1970),175-181.
[268] _ _ , Rechts-Bikategoriestrukturen in toplogischen Kategorien, Ph.D. thesis,
Free University Berlin, 1973.
[269] _ _ , Limit-metrizability of limit spaces and uniform limit spaces, Lecture
Notes Math. 719 (1979), 234-242.
[270] Mathews, J.C. and Anderson, RE, A comparison of two modes of order
convergence, Proc. Amer. Math. Soc. 18 (1967), 100-104.
1022 EVA LOWEN-COLEBUNDERS AND BOB LOWEN
[271] Menu, J. and Pultr, A., On categories determined by Poset- and Set-valued
functors, Comment. Math. Univ. Carolinae 15 (1974), 665-678.
[272] _ _ , Simply (co)reflective subcategories of the categories determined by
Poset-valued functors, Comment. Math. Univ. Carolinae 16 (1975),161-172.
[273] Min, K.C. and Nel, L.D., Newton's method and Frobenius-Dieudonne theorem
in nennormable analysis, Applied Categ. Struet. 5 (1997),205-216.
[274] Monadi, A. and Nel, L.D., Holomorphy in convergence spaces, Applied Categ.
Struct. 1 (1993), 233-245.
[275] Moore, E.H. and Smith, H.L., A general theory of limits, Amer. J. Math. 44
(1922), 102-121.
[276] Mr6wka, S., On the convergence of nets of sets, Fund. Math. 45 (1958), 237-
246.
[277] Mrowka, S. and Pervin, W. On uniform connectedness, Proc. Amer. Math. Soc.
15 (1964), 446-449.
[278] Mller, B., Uber den c-Dual eines Limesvektorrumes, Ph.D. thesis, University
of Mannheim 1972.
[279] _ _ , Dualittstheorie fr vektorunderverbnde von ~c (x), Math. Nachrichten
69 (1975), 273-284.
[280] Mller, G., Allgemeine Konvergenzbegriffe im topologischen Vereinen und
Verbnden, Math. Ann. 139 (1959), 76-86.
[281] Mulvey, C., A categorical characterization of compactness, J. London Math.
Soc. 17 (1978), 429-436.
[282] Krishna Murti, S.B., A set ofaxioms for topological algebra, J. Indian Math.
Soc. 4 (1940), 116-119.
[283] Naimpally, S.A. and Warrack, B.D., Proximity spaces, Cambridge University
Press 1970.
[284] Nakagawa, R., Natural relations, separations and connections, Questions
Answers Gen. Topoi. 3 (1985), 28-46.
[285] _ _ , Categorical topology. In K. Morita and J. Nagata: (eds.), Topics in
General Topology, North Holland 1989, 563-623.
[286] Nauwelaerts, M., Cartesian closed hull of (quasi-)metric spaces revisited, to
appear.
[287] Nel, L.D., Initially structured categories and cartesian closedness, Canadian J.
Math. 27 (1975), 1361-1377.
[288] _ _ , Cartesian closed topological categories, Lecture Notes Math. 540 (1976),
439-451.
[289] _ _ , Cartesian closed coreflective hulls, Quaest. Math. 2 (1977), 269-283.
[290] _ _ , Topological universes and smooth Gelfand-Naimark duality, Contem-
porary Math. 30 (1984), 244-276.
[291] _ _ , Infinite dimensional calculus allowing nonconvex domains with empty
interior, Monatsh. Math. 110 (1990), 145-166.
[292] _ _ , Introduction to categorical methods I and II, Carleton-Ottawa Mathe-
matical Lecture Notes Series 11, Carleton University 1991.
[293] _ _ , Introduction to categorical methods III, Carleton-Ottawa Mathematical
Lecture Notes Series 12, Carleton University 1991.
TOPOLOGICAL CONSTRUCTS 1023
[367] _ _ , Sur les espaces a structure unifonne et sur la topologie generale, Actual.
Sei. Ind., 1937, p. 551.
[368] Wischnewsky, M.B., Initialkategorien, Ph.D. thesis, University of Munchen,
1972.
[369] _ _ , Aspects of universal algebra in initialstructure categories, Cahiers Top.
Geom. Diff. Categ. 14 (1974), 1-27.
[370] Wloka, J., Limesrurne und distributionen, Math. Ann. 152 (1963), 351~09.
[371] Wolff, H., On the external characterization oftopological functors, Manuscripta
Math. 22 (1977), 63-76.
[372] Wyler, 0., The Stone-Cech cornpactification for limit spaces, Notices Amer.
Math. Soc. 15 (1968), 169.
[373] _ _ , On the categories of general topology and topological algebra, Arch.
Math. 22 (1971), 7-17.
[374] _ _ , Top categories and categorical topology, General Topology Appl. 1
(1971), 17-28.
[375] _ _ , Are there topoi in topology? in Categorical Topology, E. Binz and
H. Herrlich (eds.), Springer Verlag 1976,699-719.
[376] _ _ , Function spaces in topological categories, Lecture Notes Math. 719
(1979), 411~20.
[377] _ _ , On convergence of filters and ultrafilters to subsets, in Categorical meth-
ods in computer seience, H. Ehrig, H. Herrlich, H.J. Kreowski, and G. Preu
(eds.), Springer Verlag 1989, 340-350.
[378] _ _ , Lecture Notes on Topoi and Quasitopoi, World Scientific 1991.
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS
MICHAEL G. TKACHENKO
Universidad Aut6noma Metropolitana
Mexico
Contents
Notation. The symbols Z, Q, lR and 'lI' denote, respectively, the integers, ratio-
nals, reals, and the cirele with their usual algebraic operations and topology. As
usual, N is used for non-negative integers. The minimal subgroup of a group G
eontaining a subset A of Gis denoted by (A). The subgroup of G generated by
an element x E Gis (x).
We identify cardinals with the corresponding ordinals, but keep (J) for the first
infinite ordinal and ~o for the first infinite cardinal. The power of the continuum
is denoted by c, so that c = 2~o. The topological terminology we use is standard
and almost coincides with that of Engelking [Eng]. The only difference is that
we denote the tightness and the Lindelf number of aspace X by t(X) and
L(X), respectively.
Here we discuss the tools needed for working with topological groups, that is, the
main operations that enable us, on the one hand, to construct more complicated
groups starting with relatively simple ones, and, on the other, to reduce the
study of general groups to simpler ones, decomposing them somehow into a
combination of well-known groups.
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1031
It is natural to expect that the compatibility of the group operations and the
topology of a topological group should result in better topological properties of
the underlying space. This is probably the main idea of all specialists working
in the area and having a topologie al bias. The first very simple result justifying
this expectation is that all group topologies are regular [We3], [Da2], [pon4].
Indeed, every topological group G is a homogeneous space due to the fact
that all translations ta : G ~ G defined by ta(x) = a . x for a, x E G, are
homeomorphisms of G. So we have only to verify the regularity of G at the
identity eGo Since eG . eG = eG and the group operations are continuous, for
every open neighborhood U of eG there exists an open neighborhood V of eG
such that V-I = V (V is symmetric) and V . V ~ U. It is c1ear that the open
sets V and V . F are disjoint, where F = G \ U. Thus, the c10sure of V is
contained in U, and hence G is regular at the identity.
Note that we have just used the abbreviation G for a topological group
instead of the rigorous (G, ., 5"). We will always do this in the sequel if it is
c1ear what the group operation and the group topology on Gare. If the group
operation is commutative, we also use the plus sign.
The regularity of topological groups is probably not a very surprising fact.
But it is really surprising that all topological groups are completely regular;
1032 MICHAEL G. TKACHENKO
1. N(eG) = 0;
2. N(x- 1 ) = N(x);
{x E G : N(x) < I} ~ U.
It is not easy to prove that ZWj contains two c10sed disjoint subsets Fl and
F2 which cannot be separated by disjoint open neighborhoods; the proof given
by Stone requires the fact that every continuous real-valued function on ZWj
depends on, at most, countably many coordinates. Furthermore, the group ZW\
is quite large, its cardinality is 2Wj . So, are most topological groups normal or
not? Are there "smali" non-normal topological groups?
The reader will probably be satisfied with the following somewhat inexact
answer: there are as many non-normal topological groups as non-normal com-
pletely regular spaces (Pestov [PeI4] embeds the category ofTikhonov spaces as
a subcategory to the category of Hausdorff topological groups, and this explains
the foregoing phrase). The reason for that is certainly not obvious: it turns out
that every completely regular space X can be embedded as a closed subspace
into a corresponding topological group, say F(X). It was A. A. Markov [MarI]
who, in 1941, gave the first construction of such an embedding. The whole con-
struction is quite complicated, the original complete exposition takes about fifty
pages (see [Mar2]). The topological group F(X) corresponding to aspace X
was called by Markov thefree topological group over X. The wordfree appears
in the name of the group F (X) due to the fact that Markov's construction makes
X to be the set of free algebraic generators of the group F(X), so that F(X) is
algebraically a free group over X. In spite of the algebraic "simplicity" of F (X),
the topological nature of this group is fairly complicated and its investigation is
still far from complete. A detailed discussion of free topological groups will be
given in Section 5.
In fact, the above theorem was proved in 1936 by Kakutani [KakI] and
Birkhoff [Bir] , fifteen years before the Bing-N agata-Smirnov criterion of metriz-
ability of regular topological spaces was obtained. The book [Pon4] by Pontrya-
gin also contains an argument that implies Theorem 1.3. This theorem is based
on Theorem 1.1 and is in fact its easy corollary. Indeed, let {Un : n E N} be
a base of a topological group G at the identity eGo According to Theorem 1.1,
one can define a continuous pseudonorm Non G such that {x E G : N(x) <
2- n } ~ Un for each n E N. It is dear that N(x) > 0 for each x i= eG, and
hence the function d defined by d(x, y) = N(x-1y) for all x, y E Gis a left
invariant metric generating the topology of G.
Another interesting feature of topological groups is that convergence prop-
erties in this dass are significantly better than in general topological spaces.
Every first-countable space X is Fn!chet (that is, if a point x EX belongs to the
dosure of a subset A of X, then x is a limit of a convergent sequence lying in A).
One cannot generalize Theorem 1.3 by showing that a Fn!chet topological group
is metrizable: the b-product K of uncountably many copies of the cirde group
1[' (with the topology induced from the Cartesian product) is a non-metrizable
Fn!chet topological group. The Frechet property of the group K follows from
the general result proved by N. Noble [Nob] in 1970: any b-product of first-
countable spaces is Frechet. It is worth mentioning that the group K is countably
compact but not compact, an observation made by Pontryagin [Pon4] in 1939.
We still have not seen any difference in the convergence properties of topo-
logical spaces on the one side and topological groups on the other. Nyikos
[Nyi] was the first to note that difference. Let us say that X is an cx4-space (or
(4 - FU)-space in the terminology of [Arh5]) if, for every point x E X and
any countable family {Sn: n E N} of sequences converging to x, one can find
a sequence S converging to x which meets infinitely many sequences Sn. Note
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1035
that the sequentialfan Sw (obtained as the quotient space ofthe disjoint sum of
a countably infinite family of convergent sequences via identifying limit points
of all these sequences) is normal Frechet, but not an a4-space. Surprisingly,
this is impossible in topological groups: from a result of Nyikos, every Frechet
topological group is an a4 -space.
To formulate another result of Nyikos [Nyi] , let us recall that aspace X
is called sequential if every subset A of X containing the limits of all con-
vergent sequences lying in A is dosed in X. The dass of sequential spaces
is considerably wider than the dass of Frechet spaces, but Nyikos proves that
every sequential topological group with the a4-property is Frechet. Again, an
analogous result is not valid for topological spaces: there exist normal sequential
a4-spaces which are not Frechet, for example the Arens space S2 (see [Are] or
[NST, p. 205]).
Let us turn back to the metrizability of topological groups. Theorem 1.3 is,
without any doubt, a perfect solution to the metrization problem. Surprisingly,
the condition ofbeing first countable in that theorem can be weakened! Aspace
X is called weakly first countable if it is possible to assign to each point x E X
a sequence {B(n, x) : n E N} of subsets of X containing x in such a way that
B(n + 1, x) S; B(n, x) for each n E N, and so that a subset U of X is open
if and only if, for each x E U, there exists n such that B(n, x) S; U. We just
recall here that all first countable spaces, as well as the symmetrizable spaces,
are weakly first countable. (Some additional information on the properties of
weakly first countable spaces can be found in [Siw].) The following unexpected
result also belongs to Nyikos [Nyi].
bisequential space is Frechet and has the Cl4 -property, but bisequentiality by no
means implies metrizability (even for countable regular spaces). On the other
hand, bisequentiality is stronger than both the Frechet-Urysohn property and
the Cl4-property as the following theorem of [Arh12] shows:
The reader can find further generalizations of Theorem 1.5 in the article
[NST] by Nogura, Shakhmatov, and Tanaka, where much information about the
relations between various kinds of convergence properties in topological spaces
and groups is given. Two difficult consistency results ab out topological groups
satisfying Cl2 and Cl3 convergence conditions (see [Arh5] for the definition) have
been proved in [Sh8].
One of the most important things in any branch of mathematics is to know how
"new" and complicated objects can be constructed from a few well-known and
simple objects. Here we discuss two (of four or five) main ways of constructing
new topological groups: the operations of taking subgroups and quotient groups.
Given a sub set H of a topological group G, we call H a topological subgroup
of G if H is a subgroup of the group G and carries the topology inherited from
G. Thus, a topological subgroup H of G is not necessarily a closed sub set of G,
but H inherits both the algebraic structure and topology from G. The rationals
Q form a subgroup of the additive group IR of the reals, and if IR has the interval
topology, then the induced topology of Q is also interval. The set Z of integers
is a subgroup of the topological groups IR and Q, and Z inherits the discrete
topology from these groups. Why do the topologies on Z induced by IR and Q
coincide? In fact, the operation of taking a topological subgroup is transitive:
if H is a topological subgroup of a group G and K is a topological subgroup
of H, then the topology of K induced by H coincides with the topology of K
induced direct1y by G. This simple fact is very useful and enables us to use
the word "subgroup" instead of "topological subgroup" if there is no danger
of ambiguity. The fact, to some extent surprising, is that every open subgroup
of a topological group G is closed in G. This result is specific for topological
groups, but one can prove it in a few lines as folIows.
Let H be an open subgroup of G and let x E G be a cluster point of H.
Since H is open, x . H is a neighborhood of x and, hence, the set x . H meets
H. Thus, x . hl = h2 for some hl, h2 E H, whence x = h2 h i l EH.
It is also useful to know how open subgroups of a given group G can arise.
A general way to construct them is to take a symmetric open neighborhood U
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1037
The answer to Problem 1.6 is almost always the same as in the general
case of topological spaces: if (dosed) subspaces of completely regular spaces
inherit a topological property c;P, then the same is true for (dosed) subgroups
of topological groups, and vice-versa (the direct implication is trivial, but the
inverse one is less obvious). There are quite a few non-trivial exceptions of this
empirical rule, and some additional conditions on subgroups are needed in those
exceptional cases.
Problem 1.7 is considerably more difficult than Problem 1.6. For example,
there is no satisfactory characterization of subgroups of Lindelf topological
groups (see [Arh8, Problem 14]). On the other hand, subgroups of compact
topological groups were characterized in 1937 by A. Weil [We3] as totally
bounded (or preeompaet) groups, that is, the topological groups which can be
covered by finitely many translates of any neighborhood of the identity. We thus
have the following result.
Lindelf topological groups, one of them is lRW \. This was noted by I. Guran,
cited in [Arh8].
It is wortb mentioning that the subgroups of a-compact groups are exactly
the a-preeompaet groups, that is, topological groups representable as a union of
countably many preeompaet subsets. We call a subset A of a topological group
G preeompaet if, for every neighborhood U of the identity in G, there exists a
finite set K ~ G such that A ~ (K . U) n (U . K). It is also important to note that
the latter c1ass is narrower than the c1ass of subgroups of Lindelf topological
groups (see Section 3.1). Indeed, the a-product H ofuncountably many copies
of the discrete group Z2 endowed with the ~o-box topology considered by
Comfort [ComI] is a Lindelf topological group in which every Gs-subset is
open. Therefore, H is closed in every topological group which contains H as
a topological subgroup. Since H is not a-compact, it cannot be embedded into
any a -compact group. Sections 2 and 3 contain more information about Problem
1.7.
Many interesting problems in the theory of topological groups concern
quotient groups. Given a topological group G and a c10sed normal subgroup H
of G, we consider the (algebraic) quotient group GI H and the corresponding
quotient homomorphism rr: G ~ GI H. The quotient topology in GI H is
defined by dec1aring the homomorphism rr open and continuous. The latter
means that open sets in GI H are exactly the images of open sets in G. One can
easily verify that GI H with the quotient topology is a topological group. The
definition of quotient topological groups goes back to the early thirties (see van
Dantzig [DaI], and also R. Baer and F. Levi [BaLeD. Open homomorphisms
play an important rle in that definition. The following theorem, stated explicitly
by Freudenthal [FreI], c1arifies this rle.
Let us first discuss results conceming Problem 1.10. Since the quotient
homomorphism 1r: G --+ G / H is open, it is c1ear that the answer to this prob-
lem is affirmative for every property CZJl which is preserved by open continuous
mappings. So, we can immediately conc1ude that the properties ofbeing first (or
second) countable, locally compact, locally connected, sequential, the Frechet
property, and many others are preserved when taking quotient groups. Note that
first countability is equivalent to metrizability for topo10gical groups (Theorem
1.3), so a quotient group of a metrizable topological group is also metrizable. It
is worth mentioning that an open continuous image of a metrizable space need
not be metrizable. This makes the above result particularly interesting.
Problem 1.11 seems more difficult, though some results are immediate.
For example, if anormal subgroup H of G and the quotient group G / H are
separable, then G is also separable. A similar result for connectedness was
proved by Chevalley [Che]. The first non-trivial theorem in this direction is
due to Freudenthal [Fre 1]: If both groups H and G / H are compact, then G is
also compact. A similar assertion for local compactness appeared in the book
of Montgomery and Zippin [MoZ2] twenty years later, though the proof of the
fact was not complicated.
Some efforts are required to show that the answer to Problem 1.11 is
affirmative for metrizability. This was done by Vilenkin [Vii] in 1948 (for
the proof, see Section 10 of [Gr2] or 5.38(e) of [HRD. Furthermore, if H and
G / H are second countable, then G is also second countable. The best way to
deduce this fact is to combine the above preservation results for metrizability
and separability.
One of the latest positive results conceming Problem 1.11 involves pseudo-
compact groups. Aspace X is pseudocompact if every continuous real-valued
function on X is bounded. It is c1ear that a11 compact and countably compact
spaces are pseudocompact, but the c1ass of pseudocompact spaces is consider-
ably wider than both c1asses just mentioned. We shall discuss the properties of
pseudocompact topological groups in Section 1.5, but this is a good place to
mention the following result proved in 1988 by Comfort and Robertson [CRob].
Theorem 1.12. If the groups Hand G / H are hoth pseudocompact, then the
group G is also pseudocompact.
The reader will find many results about quotient groups in Seetions 2-6.
The above theorem implies that a 10ca1ly compact NSS group does not
contain dopen non-void sets and, hence, is connected. Let G be an arbitrary
locally compact connected group and suppose that U is a symmetrie open
neighborhood of the identity in G. Since H = U~l U n is a dopen subgroup of
G (see Seetion 1.3) and G is connected, we must have H = G. This equality
means that every open neighborhood of the identity generates the group G.
Note that the condusion does not require local compactness of G. On the other
hand, if every open (symmetrie) neighborhood ofthe identity in G generates the
group G, then G is connected. This follows immediately from Theorem 1.14.
The condition of local compactness cannot be omitted: the group Q of rationals
1042 MICHAEL G. TKACHENKO
with the interval topology is generated by every open neighborhood of zero, but
Q is not connected.
It is also interesting to note the difference in the conc1usion ofTheorem 1.14
when the group G is compact: an open subgroup H of G contained in a c10pen
neighborhood of the identity can even be chosen to be normal (in the algebraic
sense). Can one prove the same assertion for locally compact (in groups? Making
use of the semidirect product of topological groups, Montgomery and Zippin
[MoZ2] constructed a locally compact zero-dimensional group of countable
weight which has no compact open subgroups. This gives a strongly negative
answer to our question. The term zero-dimensional space will always mean that
the space has a base of c10pen sets.
Let us say that aspace X is hereditarily disconnected if all non-void con-
nected subsets of X are singletons. Every zero-dimensional space is hereditarily
disconnected, but not vice-versa, as an example of Erds shows (see Exam-
pIe 6.2.19 of [Eng]). These notions, however, coincide for Hausdorff locally
compact spaces: every hereditarily disconnected locally compact space is zero-
dimensional (this follows from Corollary 6.2.10 of [Eng]). Summarizing, we
can reformulate Theorem 1.14 as follows:
Again, local compactness is essential in Theorem 1.15: the group in Erds 's
example is second countable and hereditarily disconnected, but not zero-dimen-
sional. Several wide c1asses of topological groups in which hereditary discon-
nectedness implies zero-dimensionality were recently found by Dikranjan [Dil],
[Di2]. For example, countably compact groups form such a c1ass.
Let us also mention some facts about the component of a topological group
G, Le., the maximal connected subset of G that contains its identity. First, the
component C of an arbitrary topological group G is a c10sed normal subgroup
of G, the result due to Schreier [Schr]. If the group G is locally compact, then
the component C is the intersection of all open subgroups of G. The quotient
group G / C is hereditarily disconnected for every topological group G, and if
G is locally compact, then G / C is zero-dimensional. The assumption of local
connectedness of Ginthe last result cannot be dropped as Erds's example
shows.
Let us now consider quotients of zero-dimensional topological groups. First,
quotient groups of zero-dimensional groups need not be zero-dimensional. A
relatively simple example of this kind can be obtained by means of the following
slight modification of the construction given by Cartan and Dieudonne [CaDi].
We do not pretend to originality, but the arrangement of details in the example
below seems to be new to some extent.
TOPOLOGICAL FEATIJRES OF TOPOLOGICAL GROUPS 1043
Example 1.16. The group of reals lR with the usual topology is the quotient
group of a second countable zero-dimensional Abelian topological group.
Some work has been done to find other dasses ~ of topological groups with
the property that quotients of zero-dimensional groups from ~ are also zero-
dimensional. It tums out that pseudocompact groups have this property; more
generally, the dass of locally pseudocompact groups is as required [Tkl1]. This
result generalizes Theorem 1.17. The reader will find more details on the subject
in Section 6.
In fact, Exarnple 1.16 is notexceptional: Arhangel'skil [Arh2, p. 139] proved
that every second countable topological group can be represented as a quotient
group of a second-countable zero-dimensional topologieal group. The following
result also proved by Arhangel'skil (see [Arh9] or Section 7 of [Arh8]) shows
even more:
Thus, the absence of local compactness can produce some fairly strange
effects. The proof ofTheorem 1.18 given in [Arh9] as weIl as its weaker form
for second countable groups are based on the use of free topological groups.
1044 MICHAEL G. TKACHENKO
The c1ass of locally compact groups is interesting in many other respects. Let
us mention two more results each of which leads to unexpected generalizations.
It is known that topological groups need not be normal spaces. However, all
locally compact groups are paracompact and, hence, normal. (W. Comfort on
page 1161 of [Com2] attributes this result to E. Michael.) The reason for this is
quite simple: every locally compact group contains an open (J' -compact subgroup
generated by a compact neighborhood of identity. This result was extended to
locally Cech-complete topological groups by Brown [Bro].
The second fact is even simpler: if a subgroup H of a topological group G
is locally compact, then H is c10sed in G. Sometimes the topological groups
c10sed in every "enveloping" topological group are called absolutely closed (see
[Ale]). From the fifties, the groups satisfying this condition have borne the name
eomplete. We shall deal with complete topological groups and completions in
Section 1.6.
The facts presented in this section form a base from which a more profound
study oflocally compact groups starts. Pontryagin-van Kampen's duality theory,
for example, requires some "elementary" results on the structure of finite-
dimensionallocally compact Abelian groups (see Section 2 for details) and
certain (still "elementary") reduction to compact groups. To give the first idea
of the technique elaborated by Pontryagin and van Kampen in the thirties, we
mention one more fact here. Let us say that a topological group G is monothetie
if G contains a dense cyclic subgroup. As far as we know, the following result
should be attributed to Weil [We4].
and H = naeA Ha, then the quotient group GIH is topologically isomorphic
to the direct product naeA Gal Ha. All these results appeared in [We4].
The question about normality of topological groups goes back to the mid-
thirties. Again, the direct product operation was used in [Sto] to construct a
non-normal topological group. Let us take, after Stone, all factors equal to the
simplest non-compact group: the discrete group of integers Z. It turns out that
the product group ZWI is not normal (see also Section 1.1). Note that the group
ZWI is separable by the Hewitt-Marczewski-Pondiczery theorem, so a separable
topological group need not be normal.
Another question concems normal separable topological groups. There are
many separable normal spaces which are not Lindelf. For aperiod of time,
however, there was a suspicion that anormal separable topological group had to
be Lindelf. In 1976, Hajnal and Jubasz [HaJu] used the Continuum Hypothesis
(CH) to construct a hereditarily separable, hereditarily normal topological group
which was not Lindelf. Their group was a dense subgroup of the product group
Z(2)Wl, where Z(2) = {O, I} is the discrete two-element group. This result
answered the question negatively, but under CH. Nine years later, in 1985, Hart
and van Mill [HvM2] gave a counterexample in ZFC, without any extra axioms.
Their construction involves free topological groups instead of products.
One of the most interesting and difficult problems conceming direct products
is the (finite) productivity of topological properties in the dass of topological
groups. The main idea is to find topological properties which may fall to be
productive in general, but which, in the presence of the topological group
structure on factors, become productive. We have to confess that the quest for
such properties has not been very fruitful. Up to now, the only known productive
property specific for topological groups is pseudocompactness, a truly amazing
result. The theorem about the productivity of pseudocompactness was proved
by Comfort and Ross in [CRos] almost forty years after the definition of a
topological group had been given. Its formulation (but not its proof) is easy:
To deduce the Comfort-Ross theorem, one can simply take in Theorem 1.23
every X a to be equal to Ga. Furthermore, a topological property oftopological
groups responsible for the productivity of functional boundedness is given in
[Tk9]. It turns out that every topological group has an w-directed lattice of
continuous open mappings onto Dieudonne-complete spaces, and the existence
of such lattices for factors makes the functional boundedness stable under the
product operation. An alternative proof of Theorem 1.23 was given in 1992 by
Husek [Hush] via a factorization of continuous functions defined on subsets of
Cartesian products.
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1049
he founded the theory of uniform spaces, showed that topological groups have
natural uniform structures (the left and right group uniformities) and defined
the completion of uniform spaces and, hence, of topological groups. We assume
that the reader is familiar with the basic notions of the theory of uniform spaces
and simply give a brief sketch ofWeil's approach to completions oftopological
groups.
Let G be a topological group. Denote by .N"G (e) the system of all open
neighborhoods of the identity e in G. For every V E .N"G(e), consider the
following entourages of the diagonal in G 2 :
is again a left Cauchy filter. Consider the set G of equivalence c1asses in <fol with
the corresponding uniform structure *V, a base of which consists of the sets
where V E .N"G (e). The uniform structure just defined induces a topology on G
which agrees with the original topology of the subset G ~ G. It turns out that
the uniform space (G, *V) is complete, that is, every Cauchy filter converges
in this space.
Weil's basic idea was to extend the group operations from G to continuous
operations on G. A multiplication in Gis defined by means of multiplication in
<fol: if <!f, 'tJe E <fol, then <!f 'tJe is the left Cauchy filter in G with the base
{F . H : F E <!f, H E 'tJe}.
1052 NUCHAELG.TKACHENKO
It turns out that the inverse of a left Cauchy filter is a right Cauchy filter and
vice-versa. There are, however, many topological groups G in which the family
ofleft Cauchy filters C(6~ does not coincide with the family of right Cauchy filters
C(6o' and this is the main difficulty in the theory developed by Weil. But Weil
notes that if the inverse function x ~ x-I is uniformly continuous on some
V E .NG (e), then C(61 = C(6r. Thus, the set G with the multiplication and inverse
operations defined above becomes a group in this case. It is easy to verify that
both operations are eontinuous in the topology induced by the uniformity *V
and, hence, G is a topological group with the dense subgroup G. The topological
group G is called the Weil completion of G.
The following particularly useful fact is also due to Weil [We3]. Let us call
a topologie al group G locally bounded if G contains a bounded neighborhood
V of identity. [The latter means that V can be covered by finitely many trans-
lations of any set W E .NG(e).] So, the crucial fact is that the inverse operation
x ~ x-I is uniformly continuous on any bounded neighborhood of identity
in G. Summarizing, every locally bounded group G is a dense subgroup of the
topological group G.
Recall that the uniform spaee (G, *V) is complete, that is, every left Cauchy
filter in the group G converges to some point of G. Topological groups having
this property are called Weil-complete.
Let V be a bounded neighborhood of the identity in G. Then the closure V-
of V in Gis also bounded in G. It is known, however, that a closed bounded
sub set of a complete uniform spaee is compact and, hence, V- is a eompact
neighborhood of identity in G. The latter means that the group G is locally
compact. On the other hand, every subgroup of a locally compact topological
group is obviously locally bounded, thus giving us the following result [We3]:
One can reformulate Theorem 1.25 by saying that the class of subgroups of
locally compact groups coincides with the class of locally bounded topological
groups. This result is a natural generalization of a similar theorem characterizing
subgroups of compact topological groups (see Theorem 1.8).
We mentioned above that the inverse of a left Cauchy filter is a right one,
and vice-versa. In general, the inverse of a left Cauchy filter may fail to be
a left one, thus making it impossible to extend the inverse operation from G
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1053
V E N G(e), define
-
V = {~E
-
G: F ~ V for some F E ~}.
An easy verification shows that the family {V : V ENG (e)} forms a base of
a group topology r at the identity of G and the restrietion of r to G coincides
with the original topolo~ r of the group G, thus making G a dense topological
subgroup of the group G. The main fact established by Ralkov was the com-
pleteness of Gin the sense that every minimal Cauchy filter on Gconverges. It
was shown in the same artic1e [Rai] that the completeness of a topological group
is equivalent to the property of being an absolutely closed group, as defined in
1942 by Alexandrov [Ale]. In other words, a complete group H is c10sed in
any topological group which contains H as a topological subgroup. These two
results show that Ralkov's construction presents exact1y what is expected of
the completion of topological groups. The construction just given is called the
Ralkov completion or simply the completion of topological groups.
A non-compact completely regular space can have many compact exten-
sions. This fact gives rise to the problem of whether a non-complete topological
group G has different group completions, that is, if G embeds as a dense
subgroup in different complete topological groups. If so, is there a maximal
group completion of G? It turns out that the Ralkov completion G is maximal
among all group extensions of G and it is unique up to topologie al isomorphisms
fixing points of G.
It is easy to see that every Weil-complete topological group is Ralkov-
complete, but not vice-versa as the example of the group of autohomeomor-
phisms of the unit interval shows [Die]. Ralkov's approach permitted the char-
acterization of Weil-complete groups [Gr2]: a topological group G is Weil-
complete if and only if it is Ralkov-complete and every left Cauchy filter on G
is a right Cauchy one.
Our description of the Ralkov completion is adopted from Graev's work
[Gr2] , where some technical improvements to the original construction were
introduced. In fact, the Ralkov completion of a topological group G is the
completion of the uniform space (G, *"V*). The symbol *"V* stands for the
bilateral group uniformity on Gwhich is the upper bound of the left and right
group uniformities on G. It is useful to note that the three uniformities *"V, "V*
and *"V* coincide for an arbitrary Abelian topological group G, so the Weil
completion and the Ralkov completion coincide for Abelian groups. A similar
conc1usion holds for the wider c1ass of topological groups with an invariant
base, that is, for groups with a base at the identity consisting of sets invariant
under all inner automorphisms [HR].
Graev [Gr2] mentions that a topological group locally isomorphie to a
(Weil-) complete group is (Weil-) complete as weIl. In particular, if N is a
discrete normal subgroup of a (Weil-) complete topological group G, then the
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1055
quotient group GIN is also (Weil-) complete. On the other hand, Vilenkin
[ViI] shows that the c1ass ofWeil-complete topological groups is c10sed under
extensions: if N is a c10sed normal subgroup of a topological group G and both
N and GI N are Weil-complete, then G is Weil-complete. Following Vilenkin's
reasoning, Graev [Gr2] proves a similar result for complete topological groups.
These results give rise to the problem of whether a quotient of a (Weil-)
complete topological group is (Weil-) complete. Vilenkin [ViI] constructed an
example which shows that a quotient group of a Weil-complete topological
group need not be Weil-complete. Since the groups in Vilenkin's example are
Abelian, this implies that quotients of complete groups can lose this property.
On the other hand, Freudenthal [FreI] proved that if a Weil-complete group G
is metrizable, then all quotients of G are Weil-complete and metrizable (see also
[Gr2]). The proof given by Freudenthal could not be (automatically) translated
into a proof of a similar result for complete groups, and the completeness
preservation theorem for quotients of metrizable groups appeared many years
later.
There is an interesting relation between completeness of topological groups
and Cech-completeness (a completely regular space X is called Cech-complete
if Xis a Ga-set in the Cech-Stone compactification X). Choban [Chol] and,
independently, Brown [Bro] ~roved that a first countable topological group is
complete if and only if it is Cech-complete. In fact, the result they proved is
more general. Following Pasynkov [pas3], we call a topological group G almost
metrizable if G contains a non-empty compact set which has countable character
in G .It is immediate that all metrizable and alliocally compact groups are almost
metrizable, but the c1ass of almost metrizable groups is wider than the union of
the two c1asses just mentioned. An almost metrizable group G contains a c10sed
compact subgroup K such that the quotient group GI K is first countable, and
this property characterizes almost metrizable groups [Pas3]. The result proved
in [Chol] and [Bro] can now be formulated as folIows: an almost metrizable
topological group is complete if and only if it is Cech-complete. Note that this
fact and Hausdorff's theorem [Hau] on open mappings (if f: X --+ Y is an
open continuous mapping of a complete metrizable space X onto a metrizable
space Y, then Y is complete) together imply completeness of the quotients of
first countable complete groups.
The articles [Sul], [Gral], [Gra2] , [LeiI] and [Lei2] contain many results
about completeness of quotient groups. Usually, additional conditions on a
complete group or its c10sed normal subgroup are to be imposed. A useful
source of further information about complete topological groups is the book
[RoDi] by Roe1cke and Dierolf.
It frequently happens that a continuous epimorphism between two complete
topological groups is open. As far as we know, the first example of this phe-
nomenon was found in 1931 by Banach [Ba]: if f: G --+ H is a continuous
1056 NUCHAELG.TKACHENKO
epimorphism of first countable topological groups and both G and H are com-
plete as metric spaces, then f is open. A corollary to Banach's theorem is the fact
that a continuous epimorphism between Weil-complete second countable groups
is always open [FreI]. It is also well known that a continuous homomorphism
of a locally compact a-compact group onto a locally compact group is open
[Pon4], [Bou].
The coincidence of the c1ass of all topological groups with the c1ass of
quotients of complete groups would follow if we knew that the free topological
group F(X) over a paracompact space X is complete. The latter result was
recently announced by O. Sipacheva.
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1057
G' n H ' - {e L'}, G' 0 H = L' and G' is normal in L' (no matter what a
homomorphism<p: H ~ Aut(G) is). Note that an "internal" semidirect product
of G and H need not even be homeomorphic to the topological product G x H.
There are few situations either in abstract groups or in topological groups
when a semidirect splitting is direct. First, this obviously happens if H is normal
in L. This is also true for a divisible subgroup G of an Abelian group L or for
p-primary components of Abelian torsion groups - they are always direct
summands. The case of topological groups is more complicated. Let us say that
a topological group G is a vector group if it is topologically isomorphic to the
additive group of a finite-dimensional vector space jRn. A semidirect splitting
becomes direct in each of the following cases:
(1) L is a locally compact Abelian group compact modulo its component and
G is a vector subgroup of L or the maximal compact subgroup [Kam2] ,
[We4];
The theory of locally compact groups is the richest part of the general the-
ory of topological groups in results as well as in the techniques involved. At
the beginning the study of locally compact groups was very much stimulated
by Hibert's fifth problem as to whether every locally Euclidean topological
group is a Lie group. The efforts made by many contributors gave rise to the
Pontryagin-van Kampen duality theory for loeally compact Abelian groups,
representation theory of compact topological groups, structure theory of locally
compact (Abelian) groups, etc. We briefly discuss here the main coneepts and
results concerning locally compact topological groups.
[(f) = L f(x) dm
for each f E Co(G). The integral [ is left invariant in the sense that
i f(gx)dm = i f(x) dm
for every f E Co (G). The function b. is called the modular function of a locally
compact group G. If b,. (g) = 1 for all g E G, the group G is called unimodular.
Strictly speaking, we have to write b,.(i g ) = 1 in the definition ofunimodularity,
where ig is the inner automorphism of G defined by ig (x) = gx g -1 for each x E
G. It turns out that the mapping b.: G --+ R+ is a continuous homomorphism
of G to the multiplicative group of positive reals R+. If the left and right group
uniformities of a locally compact group G coincide, then G is unimodular. In
particular, every compact group is unimodular.
If a locally compact group has a finite Haar measure mG(G) < 00, then G
is compact. If G is a compact group, the Haar measure mG on G can always
be chosen to satisfy mG(G) = 1, and this condition defines the measure mG
uniquely.
By the Peter-Weyl theorem [PeWe], for every element gof a compact group
G with g i= eG, there exists a continuous finite-dimensional representation
Q: G --+ V(n, C) such that g fj. ker(Q). This result immediately implies that
every compact group is projectively Lie, and this is the first step towards the
solution of Hilbert's fifth problem.
On the other hand, there exist locally compact (even discrete) topological
groups which do not admit non-trivial finite-dimensional representations. The
situation completely changes if infinite-dimensional representations of a locally
compact group G are admitted: for each element g of G different from the
identity there exists a continuous, unitary, irreducible representation of G in
an infinite-dimensional Hilbert space lHI whose kernel does not contain g (see
[GR1], [GR2]).
Fortunately, the study of compact groups does not need the infinite-dimen-
sional representation theory founded by Gel'fand and Ralkov. It turns out that
every continuous, irreducible, unitary representation of a compact group is
finite-dimensional [Koo] , [Nach]. In fact, every continuous unitary represen-
tation of a compact group is the direct sum of irreducible finite-dimensional
representations.
Let G be a locally compact group with the component of identity Go, and
suppose that G / Go is compact. Then the intersection of the kemels of all finite-
dimensional irreducible unitary representations of G coincides with {eG} iff G
is topologieally isomorphie to Rn X K, where n ~ 0 and K is a compact group
[We4].
the natural group operations in the set G/\ of all characters of a group G as
follows: if X, Xl, X2 E G/\ and g E G, then
(Xl + X2)(g) = XI(g) + X2(g) and (-X)(g) = -X(g)
Later, Pontryagin [Pon2], [Pon4] introduced a group topology into the dual
group G/\ known as the compact-open topology or the topology of uniform
convergence on compact sets. A base of open neighborhoods of the neutral
element in G/\ is defined as the family of all sets of the form
W(K, U) = {X E G/\ : X(K) ~ U},
where K is an arbitrary compact subset of G and U runs through open neigh-
borhoods of zero in 1I'. van Kampen [Kam2] gave a formally different definition
of a group topology for the dual group G/\, but in fact it coincides with the
one considered by Pontryagin for alilocally compact Abelian groups (LCA for
short). This was shown by Yosida and Iwamura [YI].
The first basic fact of the duality theory for LCA groups is that the dual
group G/\ with the compact-open topology is LCA for every LCA group G
[Pon4], [Kam2] , [We4]. This makes it possible to consider the second dual
group G/\/\ = (G/\)/\ for an LCA group G.
Let us note that every continuous, irreducible, unitary representation of
an LCA group is one-dimensional, so the Peter-Weyl theorem implies that
characters of an LCA group G separate the points of G. This is the second basic
fact of the duality theory. Another proof of this fact which does not depend on
the use ofthe representation theory ofLCA groups was given in 1954 by F!i>lner
[Foll], [FoI2].
Let G and H be LCA groups and cp: G -+ H be a continuous homo-
morphism. The dual homomorphism cp/\: H /\ -+ G/\ defined by cp/\ (X) (g) =
X(cp (g)) for all X E H /\ and g E G is continuous. In addition, if cp is open, then
so is cp/\.
Since the group G/\ is LCA, the second dual group G/\/\ is weIl defined and
we can consider the canonical homomorphism PG: G -+ G/\/\ given by
PG(g)(X) = X(g)
for X E G/\ and g E G. The following duality theorem is due to Pontryagin and
van Kampen.
Theorem 2.1. For every LCA group G, PG is a topological isomorphism ofG
onto G/\/\. If cp: G -+ H is a continuous homomorphism between LCA groups
G and H, then PHCP = cp/\/\ PG
This theorem enables us to identify the groups G and G/\/\ as weIl as the
homomorphisms cp and cp/\/\. One of the main steps towards the proof of Theo-
rem 2.1 is the duality of compactness and discreteness: if an Abelian topological
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1065
In both cases, the corresponding isomorphisms are induced by the dual homo-
morphism ep", where ep is the canonical homomorphism of G onto G / H in the
first case and the embedding of H to G in the second one.
Theorem 2.1 implies that the properties of an LCA group G are completely
determined by the properties of its dual group G". Passing from G to G/\
and back makes it possible to study in detail the structure of LCA groups. We
have just mentioned that an LCA group G is compact if and only if the dual
group G/\ is discrete. An LCA group G is connected iff every non-zero element
of G" generates a subgroup topologieally isomorphie to the discrete group Z
(such elements are called pure). An LCA group G is totally disconnected iff all
elements of G/\ are compact, that is, the dosure ofthe cydie subgroup (g) in G
is compact for each g E G. A compact Abelian group G has finite dimension iff
the freerank of G/\, say r(G/\), is finite. Furthermore, dim G = r(G/\). An LCA
group G is metrizable iff the dual group G/\ is a -compact. The weight of an
LCA group G coincides with the weight of G/\. All these results illustrate how
one can "translate" certain properties of an LCA group into the duallanguage
of the group of characters.
Let us describe some results on the structure of LCA groups. First, every
LCA group G is topologieally isomorphie to the direct product jRn x H, where
n E N and H is an LCA group that contains an open compact subgroup. An
LCA group is called elementary if it is topologieally isomorphic to a group of
the form jRn x ']I' x Zl x F, where 0 ~ 1, m, n < 00 and F is a finite group.
The dass of elementary LCA groups is exactly the dass of compactly generated
Abelian Lie groups. Every LCA group is an inductive limit of open compactly
generated subgroups.
The dass of compactly generated LCA groups admits a dear description: an
LCA group H is compactly generated iff H ~ jRn X Zl x K, where 0 :s 1, n < 00
1066 MITCHAELG.TKACHENKO
The results of this section concern locally compact topologieal groups elose in
some sense to Abelian groups. A measure "of being elose to Abelian" can have
an algebraic or topologie al nature as we will see below.
One of the main difficulties arising in the study of non-Abelian locally
compact groups is that the Pontryagin-van Kampen duality theory cannot be
applied directly (or is useless). The strategy of investigation in the non-Abelian
case consists, therefore, in the study of some special elasses of "generalized"
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1067
commutative groups which then will play the rle of Abelian ones in subsequent
steps.
A topological group G is called soluble if there exists a finite sequence
{e} = Go ~ GI ~ ... ~ Gn =G
of c10sed normal subgroups of G such that Gi +1/ Gi is Abelian for each i =
0, 1, ... ,n -1. A soluble group G is called nilpotent if Gi+I/ Gi is in the center
ofG/Gi,O:'.S.i <no
A locally compact group G is said to be pronilpotent (prosoluble) if every
neighborhood of the identity in G contains a c10sed normal subgroup G such that
G j N is a nilpotent (soluble) group. In other words, pronilpotent (prosoluble)
groups are exactly projective limits of nilpotent (soluble) groups.
A locally compact group G is called induetively pronilpotent (induetively
prosoluble) if every finite subset of G topologically generates a pronilpotent
(prosoluble) subgroup. This means that the c10sure of (K) in G is a pronilpotent
(prosoluble) for each finite subset K ~ G. The theory of inductively pronilpo-
tent groups was founded by Glushkov [Glul] and Platonov [Pla]. Let us give a
brief exposition of the main results concerning inductively pronilpotent groups.
First of all, a connected inductive1y pronilpotent group is nilpotent. A com-
pact1y generated inductively pronilpotent group is a projective limit of nilpotent
Lie groups. The theorem below gives a detailed description of the structure of
inductively pronilpotent groups.
(3) the eomponent (tG)o ofthe identity in tG is a central subgroup ofGo and
(tG)o = tG n Go;
(4) the quotient group GjtG is an induetively nilpotent Lie group with no
eompaet elements;
(i) Ql is linearly ordered by inc1usion and contains the trivial groups {e} and G;
(ii) for every subfamily X S; Ql, the subgroups c1G (UX) and nX belong to Ql;
prosoluble group belongs to each of the c1asses RN, RI. Secondly, a locally
compact inductively pronilpotent group belongs to each of the c1asses Z, N. In
particular, a locally compact inductively prosoluble group G different from a
cyclic group of prime order is not topologically simple, that is, G contains a
non-trivial c10sed normal subgroup.
Classes of locally compact groups satisfying various compactness and dis-
creteness type conditions were described in the survey of Mukhin [Muk].
We have to mention that the c1ass of profinite (that is, compact tota1ly
disconnected) groups occupies a special place among locally compact groups.
The fact that stimulated the study of profinite groups was that the Oalois groups
of infinite extensions of fields endowed with the Krull topology are profinite. It
is not surprising that the first period of the development of the theory of profinite
groups was very much infiuenced by the ideas and techniques from field c1asses
and the Oalois theory of c1assic fields. The monographs by Serre [Ser] and Ribes
[Rib] are good sources for the results obtained in this way. Chapter 11 of [OLl
presents the state of the art in profinite groups by the end of the eighties.
where U is open in G and K runs through compact sub sets of G. The topological
space ::E(G) is compact for every locally compact group G. The Chabauty
topology on ::E(G) was studied in [Cha] , [Mac], [MaSw], [Wanl], [Wan2] ,
[Mam], [porn], [Scho], [PoHu], [FZ], [Stil, [Pr2], [PrTsI] and [PrTs2].
The space ::E(G) with the Vietoris topology was an object of a systematic
investigation in [PrCh], [Pr2], [Pr3], [Pr4], [Pr7]-[Pr9], [PrIO], [KoPr], [PrSa],
[Sar] , [Kom], [Pil] , [Pi2] , and [MiPo]. Let us only mention that the Vietoris
1070 NUCHAELG.TKACHENKO
topology on ;;e(G) is locally compact if and only if the locally compact group
G belongs to one of the following dasses:
(1) compact groups;
(2) Cpoo x ... x Cp;;o x K, where PI, ... , Pn are distinct prime numbers,
C p'oo, ... , C p;;o are quasieydie groups, K is a finite group and PI, ... , Pn
do hot divide the order of K;
(3) Qp x K, where Qp is the additive group of the field of p-adie numbers, K
is a finite group and P does not divide the order of K.
It is worth mentioning that the Vietoris topology on 5:E(G) (and its mod-
ifications [Pr5]) is better suited for the investigation of subtle algebraic and
topological properties of a locally compact group G than the Chabauty topology.
Topologies on ;;e(G) distinct from Chabauty and Vietoris ones were considered
in [PrCha] and [PrStu].
The set 5:E(G) has also a natural structure of a lattiee whieh was studied by
Mukhin [Muk].
In the dass of topologie al groups, we can take direct products, subgroups and
quotient groups. These are the three main operations which produce a large stock
of elaborated examples of topologieal groups starting from relatively simple
ones. There can be, however, small c1asses of topological groups which generate
all topologieal groups by means of these operations. What are these dasses?
For example, do metrizable groups generate all topologie al groups? A similar
problem arises if we take a dass At of topological groups and study the -variety
of groups generated by At, that is, the minimal c1ass "V(At) that contains At and
is c10sed under direct products, taking subgroups and quotient groups. We only
mention here that the circ1e group 1I' generates the dass of all totally bounded
Abelian groups (this follows from results presented in Section 3.4).
One more line of investigation is to establish whether the groups of a given
dass "V are topologieally isomorphie to subgroups of the groups from another
given dass W. The latter problem leads directly to the question of the exis-
tence of universal topologie al groups with respect to certain properties. As an
example, we can mention Arhangel'skil's problem (see Problem 23 of [Arh8])
whether there exists a universal topological group of countable weight, that is,
a group G with weG) .:5 ~o such that every second countable topological group
H embeds into G as a topological subgroup.
These problems have been very stimulating for the development of the theory
of topologieal group during the last thirty years and many deep results were
established on the way of solving them. Let us start with metrizable groups.
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1071
It is dear that subgroups and quotient groups of metrizable groups are metriz-
able (see Theorem 1.3). Therefore, the direct product operation is the main
source of "new" groups. It was not dear in the forties how the subgroups of
direct products of metrizable groups look like. The problem was the fact that
metrizable topological groups need not have invariant metrics (general linear
groups GL(n, IR) and GL(n, q with n > 1 do not admit such metrics).
It is not surprising, therefore, that the first step towards solving the subgroup
characterization problem was done for the "invariant" case. Let us say that a
topological group G has an invariant basis if every neighborhood of the identity
in G contains a smaller neighborhood V of the identity such that x . V . X -1 = V
for all x E G. In other words, G has a base at the identity which consists of sets
invariant under inner automorphisms of G. It is dear that every metrizable group
with an invariant metric has an invariant basis, and vice versa. The following
theorem of Graev [Gr2] gives a complete characterization of topological groups
with invariant bases.
Since both Abelian and compact topological groups have invariant bases,
Graev [Gr2] condudes that such groups can be embedded into direct products
of first countable topological groups as topological subgroups. In fact, every
compact group can be embedded into a direct product of second countable
topological groups because first countable compact groups are second countable.
A complete description of subgroups of direct products of metrizable groups
was given in 1953 by Katz. Following [Katz], we say that a topological group
G has a quasi-invariant basis if for every neighborhood U of the identity in G
there exists a countable family y of neighborhoods of the identity satisfying the
following property: for every x E Gone can find V E Y with x . V . X -1 ~ U. In
[Arh8], [Usp9], the groups with a quasi-invariant basis are called ~o-balaneed.
It is immediate that metrizable groups have quasi-invariant bases. Katz's result
[Katz] can now be formulated as folIows.
It is easy to see that the dasses of topological groups with invariant and
quasi-invariant bases are dosed with respect to direct products, taking sub-
groups and quotient groups. A scheme defining an increasing chain of dasses
1072 MICHAEL G. TKACHENKO
of topological groups with this property was considered in 1981 by Guran [Gu 1].
For an infinite cardinal -r, a topological group G is cal1ed -r -bounded if G can
be covered by -r translations of any neighborhood of the identity. Sporadical1y,
~o-bounded groups had appeared before in [Bro] and [Gra2] with the name
weakly separable groups. Guran [Gul] was the first to discover the remarkable
categorical properties of -r-bounded groups. First, it is easy to see that direct
products, subgroups and continuous homomorphic images of -r-bounded groups
are -r-bounded. In other words, -r-bounded topological groups form a J -variety
[M02]. Further, a topological group containing a dense -r-bounded subgroup is
-r-bounded as well. However, the main result about -r-bounded groups is the
following theorem that appeared in [Gul] almost without proof. Its proof can
be found in [Usp1o], [TkI8] or in Chapter IV of [THVR].
Let us briefly discuss a scheme of the proof given by Pestov. Every topolog-
ical space X can be represented as an open continuous image of a completely
regular submetrizable space Y (in other words, Y admits a continuous one-
to-one mapping onto a metrizable spaee) - the eorresponding construction
is given on p. 331 of [Eng]. For a topological group G, we ean, therefore,
find a Tikhonov submetrizable spaee Y and an open continuous onto mapping
1: Y ---+ G. Then the free topological group F (Y) is of eountable pseudochar-
T
acter. Let
r
F (Y) ---+ G be a continuous homomorphism extending 1. Sinee
the mapping 1 is open, the homomorphism is open as weIl. This means that
r.
G is topologically isomorphie to the quotient group F(Y)j K where K is the
kernel of
Pestov's construction also implies that every topological group is a quo-
tient of an NSS-group (see Seetion 1.4). Indeed, sinee the above space Y is
submetrizable, the free topological group F(Y) is an NSS-group by a theorem
of [SiUs] (see also [MoTl] and [Tho]). Another proof of the fact that the free
topologie al group on a submetrizable space is an NSS-group has been given by
Pestov [Pe16] with the use of the concept of afree Banach-Lie algebra.
In 1979, Arhangel'skii introdueed the ~o-representable topologie al groups
as subgroups of direct produets of groups of countable pseudoeharaeter (see
1074 MICHAEL G. TKACHENKO
where f..l, is the Lebesgue measure on J. One easily verifies that the sets 0 (V, s)
form a base of a topologieal group topology at the identity of G#, thus making
G# a topological group.
It remains to embed G into G#. To this end, we assign to each x E G
the element x of G# deftned by x(r) = x for all r E J. Then the funetion
i: G -+ G#, i (x) = x, is a topologieal monomorphism of G to G#, and i (G)
is a dosed subgroup of G#. A simple argument shows that the group G# is
path-eonneeted and loeally path-eonneeted.
It can be useful to note that the eorrespondenee G 1-+ G# is of functorial
nature: every eontinuous homomorphism 1/1: G -+ Hextends uniquely to a
continuous homomorphism 1/1#: G# -+ H# [BrMo]. This defines a eovariant
funetor # in the eategory of topologieal groups and continuous homomorphisms.
It turns out that the groups G and G# share many properties. For instanee,
if G is Abelian, divisible, torsion or torsion-free, then so is G#. A similar
assertion is valid for the metrizability [BrMo], weight, charaeter, separability
(see [Bie]) , ~o-boundedness (I. Guran, personal eommunieation), and some
others. However, the group G# is never totally bounded (except for the ease
IGI = 1). There are, nevertheless, many ways to embed a totally bounded group
into a conneeted totally bounded group. For instance, every totally bounded
group is topologieally isomorphie to a dosed subgroup of a pseudocompact
eonnected group [Ur]. On the other hand, the author is not aware of the existence
of a funetorial construetion of this kind.
Path-connectedness of the group G# makes it very useful in many respeets.
In [Arh2], Arhangel'skii used Hartman and Mycielski's construetion to simplify
Kakutani's proof [Kak4] of the existenee of free topological groups. Pestov
[Pe13] applied it to give a short version of the original eonstruetion of exten-
sions of continuous pseudometries to universal topologie al algebras presented
by Choban in [Ch05].
group, where U(n) is the group of unitary complex matrices of size n. This
follows from the fact that every compact topological group has a complete
system of finite-dimensional irreducible representations [PeWe], [Pon3].
The problem of the existence of a universal group of countable weight
(in the sense of embeddings) was explicitly posed in 1981 by Arhangel'skii
[Arh8]. Uspenskij responded affirmatively and presented two "different" uni-
versal topological groups of countable weight [Usp8], [UspI2]. The first one is
the group Aut(lW) of homeomorphisms of the Hilbert cube /w onto itself in the
compact-open topology. Universality of the group Aut /w is proved by Uspenskij
[Usp8] with the help of Keller's theorem from the infinite-dimensional topology.
The second example of Uspenskij's is "elementary" in the sense that it does
not require any knowledge of the infinite-dimensional topology. Let U be a
complete, separable metric space which contains an isometric copy of every
complete, separable metric space. Such aspace U was constructed by Urysohn
[Uryl] , [Ury2]. In addition, the space U has the following property: every
isometry h: F ~ U with F ~ U and IF I < ~o can be extended to an isometry
h: U ~ U. Making use of these properties, Uspenskij [UspI2] shows that the
group Is(U) of isometries of U in the topology of pointwise convergence is
universal in the dass of second countable groups. It is not known, however, if
the groups Aut /w and Is(U) are topologically isomorphic (see [UspI2]).
The problem whether there exists a universal group in the dass of topological
groups of weight ~ t' for a uncountable cardinal t' is still open. It is not dear
whether Aut r: is such a group. A similar problem posed by Arhangel'skii in
[ArhlO] for Abelian groups has recently been solved (in the affirmative) by
Shkarin for the special case t' = ~o.
The problem of the existence of projectively universal topological groups in
the sense of open homomorphic images was considered by Pelant, Shakhmatov,
and Watson in [PSW]. The main result of [PSW] is the following theorem.
Theorem 3.5. For every infinite cardinal t', there exists a complete metric
Abelian group Gt' ofweight t' such that every complete metric Abelian group
ofweight ~ t' is a quotient group of Gt'.
In particular, there exists a Polish Abelian group G such that every Polish
Abelian group is a quotient group of G. In fact, the group G in question is
the completion of the free Abelian group A(B(w on the Baire metric space
B(w) = ZW with the topology generated by Graev's extension d ofthe standard
metric d on B (w). It is interesting to note that the complete groups appear in
Theorem 3.5 not by accident. As is shown in [PSW], for every topological group
G of weight t' ~ ~o, there exists a metric Abelian group H of weight t' which
is not a continuous homomorphic image of G.
Much less is known in the non-Abelian case. By a theorem of [PSW], for
each infinite cardinal t' , there is an invariant complete metric group Ht' of weight
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1077
't' such that every invariant complete metric group of weight ~ 't' is a quotient
group of H r . It is again an open problem whether the result remains valid if one
omits "invariant". For't' = ~o, this problem was posed by Kechris [Kec].
Following the idea of the previous chapters, we are going to present many results
illustrating the fact that cardinal functions behave much better in topological
groups than in (Tikhonov) spaces. Sometimes the difference is enormous. For
example, a first countable topological group G is metrizable by the Birkhoff-
Kakutani theorem (see Theorem 1.3) and, hence, w(G) = d(G) = c(G).
Another particular phenomenon is the coincidence of several cardinal functions
in the dass of topological groups while they are different, even for com-
pact spaces. A good example of this kind is the equality of the character and
Jr-character of every topological group [Arh7].
The strongest results about relations between cardinal functions have been
obtained for (locally) compact topological groups. The information on the
subject known to date seems to be almost exhaustive.
As usual, w(G), nw(G), d(G), X(G), 1/!(G), L(G) and c(G) are the
weight, network weight, density, character, pseudocharacter, Lindelf number
and cellularity of G. We denote by Jrw(G) and JrX(G) the Jr-weight and Jr-
character of G. The tightness of Gis t(G), and o(G) is the number of open sets
in G. If ~ (G) is the family of all compact subsets of G, we define
k(G) = min{IYI : y S; ~(G), Uy = G}.
is a base for G. Therefore, w(G) ~ d(G) . X(G). This fact implicitly appears
in [Pon4] and it admits several generalizations. Let us note that if Du S; G and
G = Du . U for each U E 00, then {xV: U E 00, x E Du} is also a base for
G. This observation readily implies the following.
1078 MICHAEL G. TKACHENKO
The last assertion of Theorem 4.1 implies all the others and apparently is due
to Guran [GuI]. In fact, the inequalities in (1), (3) and (4) of the theorem are
equalities because ib(G) ~ w(G) and X(G) ~ w(G) for every topological
group G. Note that the previous theorem is not valid, even for compact spaces:
the two arrows space Z (see Exercise 3.1 O.C of [Eng]) is compact, first countable
and separable, but w(Z) = c. Therefore, neither (1), (2) nor (3) ofTheorem 4.1
can be extended to compact spaces.
Let y be a n: -base at the identity of a topological group G. One easily
verifies that the family {U . U- I : U E y} is a base at the identity. This implies
that the character and n:-character of G coincide. Since d(G) ~ n:w(G) and
n: X(G) ~ n:w(G), from Theorem 4.1 (1) it follows that
"gap" between the pseudocharacter and diagonal number can be arbitrarily large
because for every cardinal 't', the Tikhonov cube rc contains a dense subspace
Y of countable pseudocharacter such that b.(Y) = 't' [Ami]. The existence of a
topological group structure on aspace changes the situation: b.(G) = 1/I(G) for
every topological group G [Arh?]. Indeed, if y is a pseudobase at the identity
of G, put Ou = {(x, y) E G x G : x-1y E U} for every U E y. Then the
intersection of the family {Ou : U E y} of open subsets of X 2 coincides with
the diagonal b.x of X 2 This simple observation is applied in [Arh?] to prove
the following result.
Theorem 4.3. lf G is a topological group, then pw(G) :::: iw(G) :::: wL(G) .
1/1 (G).
Since wL(G) :::: min{L(G), c(G)}, we conclude that every topological
group G satisfies iw(G) :::: L(G) . 1/I(G) and iw(G) :::: c(G) . 1/I(G) [Arh?].
Note that if Xis a TI-space, then lXI:::: 2iw (X). Therefore, Theorem 4.3 implies
the following.
Theorem 4.5. Every 0' -compact topological group has countable cellularity.
1080 MICHAEL G. TKACHENKO
Theorem 4.6. Let G be a topological group. Then c(G) ::: k(G) . ~o.
Corollary 4.8. lf His a subgroup 0/a topological group G, then c(H) ::: 2c (G).
It is immediate from the definition that Nag(X) ~ k(X) for every Tikhonov
space X. In particular, every a-compact space X satisfies Nag(X) ~ ~o. The
spaces X with Nag(X) ~ ~o are called Lindel/ I:,-spaces. One can also verify
that L(X) ~ Nag(X) ~ nw(X) for every Tikhonov space X.
The advantage of working with the class of Lindelf I:, -spaces is that many
results valid for a -compact spaces remain true in the former class. In addition,
the class of Lindelf I:,-spaces is closed under countable products, continuous
images and taking Fa-subsets. Combining these facts, one easily deduces that
a topological group generated by a Lindelf I:,-subspace is also a Lindelf
I:,-space [Uspl].
In 1982, Uspenskij extended Theorem 4.5 to Lindelf I:,-groups. A more
general form of this result is the inequality c(G) ~ Nag(G), which holds for
every topological group G (see Theorem 1 of [Uspl]).
Another refinement of Theorem 4.5 is related to the notion of a L-cellular
space introduced by Arhangel'skil in [ArhlO]. The idea of the corresponding
definition goes back to Efimov [Efl], who proved that every family y of G,,-
sets in {O, l}K contains a countable subfamily p, such that cl(Up,) = cl(Uy). In
general, if every family y of G.".-sets in aspace X contains a subfamily p, such
that cl(Up,) = cl(Uy) and 1p,1 ~ L, the space X is called L-cellular. Thus, for
any cardinal /C, {O, l}K is ~o-cellular by Efimov's theorem. Note that c(X) ~ L
for every L-cellular space X. The following result proved by Uspenskij in [Usp5]
is a wide generalization of Efimov's theorem which also implies Theorem 4.5.
In 1989, Uspenskij mentioned in [Uspll], without proof, that both ass er-
tions of the above theorem can be naturally extended to topological groups H
satisfying Nag(H) ~ L (in this respect, see Theorem 1.5 of [Tk12]). Tkacenko
[Tk12] proved that Theorem 4.10 is also valid for arbitrary products of Lindelf
I:,-groups.
Theorem 4.7 and Corollary 4.8 made it interesting to find out whether
2~o is the least upper bound for the cellularity of subgroups of ~o-bounded
topological groups. Uspenskij [Usp14] constructed a subgroup H of K C with
c(H) = c, where K is the discrete free group with countably many generators.
Thus, subgroups of separable topological groups can have cellularity equal to
c, and the upper bound in Theorem 4.7 is exact.
1082 MICHAEL G. TKACHENKO
To finish the section, we have to mention the following resuit of [Juh]: the
cardinality o(G) ofthe family of all open sets in an infinite topological group G
satisfies o(G) = o(GytJ. In fact, if G contains a disjoint family of open sets of
cardinality., then O(G)T = o(G) (see Theorem 3.16 of [Com2]). It is known
that the existence of a Hausdorff space X with o(X) < o(X)W is consistent with
ZFC (the result is due to SheIah).
The first and second assertions of the above theorem were proved in 1943 by
Kakutani in [Kak3] for compact Abelian groups with the heip of the Pontryagin-
van Kampen duality theory. Fifteen years latter, Hartman and Hulanicki
[HaHu] gave a direct proof of the same facts. Both artic1es make use of the
generalized continuum hypothesis (GCH for short). The first argument proving
(1) and (2) ofTheorem 4.12 without extra set-theoretic assumptions appeared in
the artic1e [Itz] by Itzkowitz and this was based on the Ivanovskll-Kuz'minov
theorem.
The third assertion ofTheorem 4.12 follows from the Ivanovskll-Kuz'minov
theorem and the following result of [Efl]: If X is a dyadic compact space and
x EX, X(x, X) = 2: ~o, then X contains a discrete subspace D of cardinality
such that D U {x} is homeomorphic to the one-point compactification of D.
Indeed, let G be a compact topological group with w(G) = 2: ~o. >From
(2) of Theorem 4.1, it follows that w(G) = X(G) and, hence, G contains
1084 MICHAEL G. TKACHENKO
Theorem 4.13. Let G be an infinite compact topological group with w(G) = t'.
Then there exists a continuous onto mapping I: G -+ [0, 1]t'.
The first assertion of Theorem 4.14 was proved by Arhangel'skll [Arh7] for
compact groups; a minor modification of the corresponding reasoning yields
the results in the present form. The part ib(G) = k(G) . ~o of (5) follows
immediately from local compactness of G and the definition of the index of
boundedness. The inequality c(G) ~ k(G) . ~o is valid for every topological
group G by Theorem 4.6. To deduce (5) it remains to note that always ib(G) ~
c(G) [Gu1].
The equality w(G) = k(G) . X(G) is due to W. Comfort (see Theorem
3.5 (iii) of [Com2]), which is apart of Theorem 4.14 (2). Combining the latter
equality with (5) of the theorem, we obtain (2). Assertions (3), (4) and (6) of
the theorem follow from (5) and Theorem 3.12 (ii)-(iv) of [Com2].
Theorem 4.14 has several interesting corollaries, two of which we present
below.
Formally, the above result does not generalize Theorem 4.18 because there
exist almost metrizable topological groups which are not ~o-balanced [pe4].
However, every compact G J-set in a quotient space of an almost metrizable
topological group is homeomorphic to a compact GJ-set in a quotient space of
an ~o-balanced topological group [Ch03], so the conc1usion of Theorem 4.18
is covered by Uspenskij's result.
The proofs of Theorems 4.19 and 4.20 given in [Usp9] are essentially shorter
than the proofs of Theorems 4.17 and 4.18 presented in [Ch03]. In fact, the
brevity of Uspenskij's exposition is due to the use of Haydon's theorem on a
spectral decomposition of Dugundji compact spaces [Hay].
Since ~o-bounded groups form a proper subc1ass of ~o-balanced groups, it
is natural to ask whether Theorem 4.19 can be extended to ~o-balanced groups.
However, Uspenskij noted in [Usp9] that discrete groups are ~o-balanced and a
compact space that admits a continuous transitive action of a discrete group is
plainly a homogeneous compact space and, hence, need not be dyadic (the two-
arrows compact space is a counterexample). It is also mentioned in [Usp9] that
the two arrows compact space is homeomorphic to a quotient space of some
topological group of countable pseudocharacter, which means that Theorem
4.20 cannot be extended to the groups of countable pseudocharacter.
By Scepin's theorem (see Theorem 9 of [Scei]), an infinite, zero-
dimensional, homogeneous Dugundji compact space X is homeomorphic to the
Cantor cube {O, IV, where t' = w(X). Combining this result and Theorem 4.19,
Uspenskij [Usp9] deduces that a compact zero-dimensional space of weight t'
which admits a continuous transitive action of an ~o-bounded topological group,
is homeomorphic to the Cantor cube {O, ur. A similar assertion is valid for
compact zero-dimensional quotient spaces of ~o-balanced topological groups.
Another surprising fact established by Uspenskij in [Uspll] is that compact
retracts of topological groups are Dugundji. Recall that a continuous mapping
f: X ~ X is called retraction if f Cf (x = f (x) for each x EX, and that the
image feX) under the retraction f is called a retract of X.
(i) e(U) nY = U;
Note that every a-compact group G satisfies Nag(G) ::::: ~o, so Theorem 4.23
implies that w(X) = t(X) for a compact continuous image X of a a-compact
group G. In fact, Theorem 4.23 is a special case of a more general result proved
in [TkI2]. Recall that if Xis a dyadic compact space of an uncountable regular
weight 1", then there exists a continuous onto mapping I: X -+ [0, IY [Ef2],
[Gerl]. Let us define the index id(X) of aspace X as folIows:
It is known that every compact space X satisfies id(X) ::::: t (X) and, in addition, if
Xis dyadic, then w(X) = id(X) [Sapl]. The former assertion was generalized
in [TkI2]: id(X) ::::: t(X) . Nag(X) for every Tikhonov space X. The latter
result admits the following generalization, which also implies Theorem 4.23
(see [TkI2]):
(3) X is ~o-monolithic;
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1091
It is a well-known fact (see [Rob] , [KaMe]) that for every non-empty set X,
there exists a group G that contains X as a set of generators and satisfies the
following condition:
(FG) every function cp: X -+ Hof X to an arbitrary group H can be extended
to a homomorphism $: G -+ H.
The set X is called the Jree basis of G. It turns out that the group G in
question is unique up to an isomorphism fixing the points of X; this group is
usually called the Jree group on X and denoted by F(X). If one changes the
word "group" to "Abelian group" in the above lines, we get the definition of the
free Abelian group on X, which will be referred to as A(X). The word "free"
appears here due to the fact that there are no non-trivial algebraic relations
in F(X) which means that any "word" g = Xfl ..... x~n in F(X) (with
Xl, ... , X n E X and EI, ... , En = 1) is different from the identity of F(X)
whenever gis irreducible. A similar assertion (with the obvious changes for the
commutative case) is valid for A(X).
In 1945, A. Markov published the work [Mar2] in which he showed that there
exists a profound analogy between algebraic and topological groups (the first
announcement goes back to 1941, see [MarI]). Let X be a completely regular
topological space and suppose that G is a topological group which contains X as
a subspace. Following Markov, we call G the free topological group on X if X
algebraically generates G and the pair (G, X) satisfies the following condition
resembling (FG):
(FTG) every continuous function cp: X -+ H of X to a topological group H
can be extended to a continuous homomorphism $: G -+ H.
The space X is called the free topological basis of G. A usual notation
for the group G is F(X). Again, if all groups in question are Abelian, G is
called thefree Abelian topological group on X and we designate it A(X). The
definition of a free (Abelian) topological group does not say anything about its
existence, and the first existence proof in [Mar2] based on the use of multinorms
was really difficult. Different approaches to the existence proof were proposed
by Nakayama [Nak], Kakutani [Kak4] and Graev [Grl]. Let us discuss briefly
Kakutani's construction.
Given a completely regular space X, consider the family <g; of all continuous
functions J: X -+ Hf of X to topological groups Hf satisfying IHf I :::: IX I ~o.
[The fair objection that <g; is not a set but a proper c1ass can be eliminated by
choosing one representative for every c1ass of equivalent functions J, g E <g; in
the sense that J '" g if there exists a topological isomorphism 1/1 : Hf -+ Hg
such that g = 1/1 0 J. Standard cardinal estimates show that <g; / '" is a set
of cardinality not greater than exp2i, the second exponent of i = lXI . ~o.]
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1093
Graev also applies this technique to show that for every c10sed subset Y of X,
the subgroup Fy = (Y) and the normal subgroup F Ny of F(X) generated by
Y, are both c10sed in F(X).
The relation between F(X) and A(X) was found out by Markov [Mar2]:
the free Abelian topological group A(X) is topologically isomorphie to the
quotient group F(X)jK, where K is the derived subgroup of F(X) generated
by all commutators. In particular, K is c10sed in the free topological group
F(X). Again, Seetion 4 of [Grl] contains a shorter proof of this result.
It is natural to try to obtain information on the properties of F(X) and
A(X) relating them to the properties of the "generating" space X. For example,
neither F(X) nor A(X) is anormal space if X is not (because X is c10sed in
these groups). Therefore, if Xis a complete1y regular non-normal space, then
the free Abelian topological group A(X) fails to be anormal topological space.
This was exactly Markov's solution to the normality problem in the c1ass of
Hausdorff topological groups [MarI], [Mar2].
In fact, the topological complexity of the groups F(X) and A(X) is higher
than the complexity of the space X as the following result shows [Grl].
Theorem 5.2. lf aspace X is not discrete, then the groups F(X) and A(X)
are not first countable.
In other words, the groups A(X) and F(X) are metrizable if and only if
X is discrete. Since the free (Abelian) topological group on a discrete space
is discrete, we conclude that metrizability is equivalent to discreteness for free
(Abelian) topological groups. It also follows from Graev's results [Grl] that the
groups A(X) and F(X) are never locally compact, except for a discrete space
X. A generalization of this result to topological (Abelian) groups with a free
algebraic basis was obtained by Dudley [Dud].
By definition, X is a subspace of F(X) and A(X). It turns out that these
groups contain all finite powers of X as closed subspaces by Arhangel'skii's
result (see [Arhl]). This fact was rediscovered eight years later in [Joi], [HMT]
and is known as Joiner's lemma. This gives us even more examples of non-
normal topological groups: if S is the Sorgenfrey line, then the groups A(S) and
F(S) are not normal because S2 fails to be normal.
For every positive integer n, let Fn(X) (An (X be the subset of F(X)
(A(X which consists of a11 elements of length at most n with respect to the
basis X. It is easy to see that both Fn(X) and An (X) are continuous images of
the space X n , where X is the topological sum of X, its copy X-I and the identity
e of F(X) (A(X: X = X EEl {e} EEl X-I. Therefore, if Xis compact, then the
sub sets Fn(X) and An(X) are c10sed in F(X) and A(X), respectively. (Making
use of the Stone-Cech compactification X of X, Arhangel'skiI [Arhl] shows
that the same conc1usion holds for any completely regular space X; see also
[Joi] and [HMT].)
In fact, a continuous mapping of X n onto F n (X) can be simply defined
by jn(XI, ... , x n ) = Xl ..... Xn. The continuity of jn follows from the conti-
nuity of the group multiplication in F(X). For A(X), it suffices to replace the
multiplication by the sum operation.
If X is compact, then the mapping jn is c10sed and we can uniquely re-
construct the topology of Fn (X) (An (X for every n E N+. The following
theorem, proved in 1948 by Graev, gives adescription of the topology of the
groups F(X) and A(X) (see [Gr2]).
We can reformulate the above theorem, saying that the free topological group
F(X) on a compact space Xis the inductive limit ofits c10sed subspaces Fn(X).
The proof of this fact given by Graev can be applied in a more general situation
when X is a so-called kw-space, that is, there exists an increasing sequence
{Kn : n E w} of compact subsets of X such that X = U{Kn : n E w} and
X is the inductive limit of Kn 'so Every kw-space is obviously O"-compact, but
not vice-versa (the rationals are a counterexample). Theorem 5.3 can now be
reformulated once again: the free topological group F(X) is a kw-space with
the decomposition F(X) = U{Fn(X) : n E w}. By a result of Mack, Morris
and Ordman [MMO], the free topological group F(X) on a kw-space X with
a decomposition X = UnEwKn is the inductive limit of its compact subspaces
{Fn(Kn ) : n E w}. This gives a more symmetric form to Graev's theorem.
Another generalization of Theorem 5.3 was given by Tkacenko in [Tk7] for
spaces c10se to being compact.
Theorem 5.4. Jf all finite powers oJ aspace X are normal and countably
compact, then F (X) is the inductive limit oJ its closed subspaces Fn (X), n E
N+.
A similar assertion also holds for free Abelian topological groups. The
c1ass of spaces X satisfying the conditions of Theorem 5.4 contains c10sed
1096 MICHAEL G. TKACHENKO
(1) The free topological group F(X) has the inductive limit property iJf X is
locally compact separable or discrete.
(2) Thefree Abelian topological group A(X) has the inductive limit property iJf
X is locally compact and the set 0/ all non-isolated points in X is separable.
Finally, Sipacheva [Si2] has shown that, for every countable (not necessarily
metrizable) space X with one non-isolated point a, F(X) has the inductive limit
property if and only if, for any collection fUn : n E w} of open neighborhoods
of a in X, there exists a neighborhood V of a such that V n (Un \ Un +l) is finite
for all n E w (the original proof in [Si2] requires, however, certain correction).
The problem of completeness of free topological groups was first investi-
gated by Graev in [Grl]. Making use of Theorem 5.3, he proved the following
basic result.
Theorem 5.6. For every compact space X, the free topological group F(X)
and the free Abelian topological group A(X) are complete. Furthermore, these
groups are Weil-complete.
This had been the only known fact about completeness offree groups unti11973,
when Hunt and Morris [HuMo] extended it to free (Abelian) topological groups
on kw-spaces. These results, however, leave open the problem of characterization
of the spaces X for which the groups A(X) and F(X) are complete. This
problem was completely solved in 1983 for free Abelian topological groups
by Tkacenko [Tkl] (however, Pestov [PeI7] notes that the solution to the
completeness problem was earlier obtained by Flood in [Flo]).
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1097
Theorem 5.7. Thefree Abelian topological group A(X) is complete if and only
if the space X is Dieudonne-complete.
In particular, the group A(X) is complete for every metrizable or, more gener-
ally, every paracompact space X. The above theorem also permits us to describe
completions of free Abelian topological groups. Let /l-X be the Dieudonne
completion of aspace X, which is the maximal subspace of the Stone-Cech
compactification X containing X and having the property that every continuous
function f: X ~ M of X to an arbitrary metrizable space M can be extended
to a continuous function f: /l-X ~ M. It is not difficult to see that X is
--
Dieudonne-complete iff /l-X = X. Thus, Theorem 5.7 applied to A(/l-X) implies
the equality A(X) = A(/l-X) for every completely regular space X, where A(X)
is the completion of A(X) (see [TkID.
--
All the results of this section are equally valid for A(X) and F(X). The
completeness of free groups is not an exception to this particular rule, but the
his tory of the problem and the methods applied for its solution in the Abelian and
non-Abelian cases are different. Roughly speaking, the reason for this distinction
is based in the complexity of a topological description of the topology of F (X)
[Tk3], while a corresponding topological description for A(X) is relatively
simple (see [Nil], [MoNi]) and, in fact, goes back directly to Graev's artic1e
[Grl].
The first advance in the non-Abelian case is due to Uspenskij [Usp6] who
showed that the free topological group F(X) is Weil-complete for a metrizable
space X (a complete proof ofthat is given in [Usp13D. In fact, Uspenskij proved
that this result is valid for all X which are products of metrizable spaces, so
the remaining problem was to extend the result to c10sed subsets of products of
metrizable spaces. It was also shown in [Usp6] that F(X) is Weil-complete for
every Lindelf space X.
The final solution of the completeness problem was announced in 1988 by
O. Sipacheva [Sii], but as far as the author knows, nobody was able to verify
the proof given there. Apart from the exposition style of [Sii], the Western
reader faces another problem: the artic1e has never been translated into English.
Recently, the artic1e [Si3] by Sipacheva gave results which imply a positive
solution to the completeness problem.
1098 MICHAEL G. TKACHENKO
Both mappings in and jn are continuous because of the continuity of the multi-
plication and inverse operations in F(X). It is c1ear that the image in (X n ) is the
sub set Fn (X) of F (X) consisting of all words of length at most n with respect
to the basis X.
If Xis compact, then all mappings in are c10sed (and hence quotient), thus
defining the topology of each of the spaces Fn (X). Since F (X) is topologically
the inductive limit of the c10sed subspaces Fn (X), n E N+, this gives us a
topological description of the space F(X) [Grl]. In fact, the same approach
works in the case of a kw-space X [MMO]: again al1 mappings in are quotient
and F(X) is the inductive limit of the c10sed subspaces Fn(X). Thus, if X
is additionally sequential, then so is X n for each n E N+, and since in is
quotient, the image Fn(X) = in(X n ) is also sequential. Using the inductive
limit property of F(X), we conc1ude that F(X) is sequential as weH [OrTh]. A
similar cargument shows that F(X) has countable tightness if Xis a countably
tight kw-space (the result implicitly foHows from [Arh7]).
Suppose that the free topological group F(X) is sequential. What will the
sequential order of F(X) be then? This was exactly the question answered by
Ordman and Smith-Thomas [OrTh]: if X is not discrete and F(X) is sequen-
tial, then the sequential order of F(X) is equal to (VI. In particular, F(X) is
Frechet-Urysohn iff X is discrete. These results were extended to subgroups of
free topological groups by Morris and Thompson [MoT2]: every non-discrete
sequential subgroup of a free topological group has sequential order (VI. In
fact, Morris and Thompson show that if a subgroup G of a free (Abelian)
topological group F(X) (A(X has a non-trivial sequence YI, Y2, ... converg-
ing to the identity e, then G contains the free (Abelian) topological group on
Y = {e} U {yn : n E N+} and, hence, G also contains the Franklin space Sw
which is of sequential order (VI .
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1099
The results just mentioned depend on the assumption that the free topo-
logical group F(X) or its subgroup is sequential. Is it really the case for good
spaces X, say, for metrizable ones? Surprisingly, in 1979 Fay, Ordman and
Thomas [FOT] proved that the free topological group over the rationals is neither
sequential nor a k-space. In fact, they showed that even the c10sed subspace
F6(Q) of F(Q) is not a k-space. This pathology made it interesting to study the
metrizable spaces X forwhich the group F(X) (a subspace Fn(X is sequential
or a k-space. Since the rationals Q are very far from being locally compact or
complete, there was hope that locally compact or complete metrizable spaces
generate free topological groups with the k-property. However, Borges [Bor]
noted that there exists a locally compact zero-dimensional metrizable space
X such that F(X) and F6(X) are not k-spaces. The space X in question is
simply the free sum of c copies of the convergent sequence (with its limit). It
is also shown in [Bor] that the mapping i6: x!' ~ F6(X) is not quotient, thus
answering (negatively) a question posed in [FOT].
The next step to realizing the topological structure of free groups on metriz-
able spaces was done by Arhangel'skil, Okunev, and Pestov in [AOP]. They
gave a complete characterization of metrizable spaces X for which F(X) or
A(X) has the k-property. We summarize the results of [AOP] in the following
two theorems.
Theorem 5.8. For a metrizable space X, the /ollowing conditions are equiva-
lent:
(a) F(X) is a k-space;
Theorem 5.9. lf X is metrizable and X' is the set 0/ all non-isolated points 0/
X, then the /ollowing conditions are equivalent:
The rationals Q and the free sum S of c copies of the convergent sequence
both produce the free topological groups F(Q) and F(S) which fall to be k-
spaces. The reason for this phenomenon is that these groups contain small
1100 MICHAEL G. TKACHENKO
Theorem 5.10. For an arbitrary metrizable space X, the Jollowing are equiv-
alent:
It is also shown in [Yaml] that A2(X) is a k-space for any metrizable space X,
and there exist metrizable spaces X and Y such that A3(X) and A4(Y) are not
k-spaces, but A3(Y) iso
A characterization of spaces X for which all mappings in: X n ~ Fn (X)
are quotient is still open. The Abelian version of the problem is also far from
being solved. For n = 2, however, the problem was completely solved in
1985 by Pestov in [Pe7]: the mapping i2 is quotient iff every neighborhood
of the diagonal in X 2 belongs to the finest uniform structure on X compatible
with the topology on X. In particular, the mapping i2 is quotient for every
paracompact space X. An analogous result is valid for free Abelian topological
groups. Therefore, there are locally compact spaces X for which the mapping
i2: X ~ A2(X) is not quotient: every locally compact non-normal space X
suits. On the other hand, if M is the free sum of Wl copies of the convergent
sequence, then the mapping in: M" ~ An(M) is not quotient for each n ~ 4
[Yaml]. This result refines one of the facts established by Borges in [Bor].
Several facts conceming the mappings in were obtained by Tkacenko in
[Tk7], [Tk16].1t is proved in [Tk7] that if X n is normal and countably compact
for some n E N+, then in is a closed mapping. Further, if X n is pseudocompact,
then the conclusion is weaker: in is z-closed, that is, in takes zero-sets in X n to
closed sets in Fn (X) [Tk16]. These results, however, leave open the problem of
characterization of the spaces for which the mappings in are closed or z-closed.
We finish this section with a very isolated theme: the reexivity of free
Abelian topological groups. In other words, the problem is to characterize
the spaces X for which the free Abelian topological group A(X) satisfies the
Pontryagin-van Kampen duality. As far as the author knows, only tbree publi-
cations [PelD], [pe18] and [GH] exist conceming this area (the second article
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1101
being a complete version ofthe first one). In [pe18] it is proved that if A(X) is
reflexive, then X is path-disconnected and if, additionally, X is pseudocompact,
then the first homotopy group of X is trivial. In addition, if X is a strongly zero-
dimensional space which is either metrizable or compact, then A(X) is reflexive.
The latter result left open the problem of characterization of reflexivity of A (X)
in terms of X. Galindo and Hernandez [GH] made significant progress in this
direction. They proved that, for a JL-space X (the c10sure of every functionally
bounded subset of Xis compact), the group A(X) is reflexive iff every compact
subset of X is totally disconnected and all compact sub sets of Ck(X, '1[') are
equicontinuous. Here Ck(X, '1[') is the space of all continuous functions from X
to '1[' endowed with the topology of uniform convergence on compact subsets.
Theorem 5.12. For a completely regular space X, the following are equiva-
lent:
In 1945, Markov posed the following problem (see [Mar2]). Let X and Y be
completely regular spaces such that the free topological groups F(X) and
F(Y) are topologically isomorphie. Are then X and Y homeomorphic? In
1104 MICHAEL G. TKACHENKO
Theorem 5.13. The equality dim X = dim Y is valid for an arbitrary pair X,
Y of l-equivalent completely regular spaces.
Burov [Burl] noted that Theorem 5.13 remains valid for the cohomologieal
dimension dimG, where the group of coefficients G is finitely generated and
Abelian.
It is interesting to note that even the strong result of [Pe3] admits a certain
generalization to a weaker equivalence relation: Gul'ko proved in [Gul] an
analog ofTheorem 5.13 forcompletelyregular spaces X and Y such that Cp(X)
and Cp(Y) are uniformly homeomorphie. Pestov's theorem is in contrast with
the following result of [Bur2]: there exist M-equivalent compact spaces X and
Y such that ind X =1= ind Y and Ind X =1= Ind Y. If, however, X and Y are
metrizable and locally compact (or separable), then ind X = ind Y (see [pavi]).
By another interesting result of [Pavi], two non-zero-dimensional polyhedra
are l-equivalent iff their dimensions coincide. The situation changes for the
M-equivalence. L. S. Pontryagin noted (cited in [Grl]) that if X and Y are M-
equivalent compact spaces, then for any topological group G, the groups [X, G]
and [Y, G] of the homotopy equivalence c1asses of continuous functions from
X and Y to G are isomorphie. In particular, the c10sed unit interval and the
circumference are not M-equivalent. Pontryagin's idea was rigorously realized
in [PaVa]: the cohomologieal groups Hn(X, n) and Hn(y, n) are isomorphie
for any countable Abelian group n and n = 1,2. It was mentioned (without
proof) in [PaVa] that the result remains valid for an arbitrary Abelian group n
and any nE N.
Let us turn to negative results related to the M -equivalence relation. The
first construction whieh allows us to produce various examples of M -equivalent
spaces with different topologieal properties was discovered by Tkachuk [TkaI]
in 1983. His idea was to use the Alexandroff duplicate X* of aspace X. It is
shown in [TkaI] that if Xis compact and lXI = ~ ~o, then X* is M -equivalent
to the free sum X E9 C(r), where C(.) is the one-point compactification of a
discrete space of cardinality . In particular, the spaces X* and X* E9 C (.) are
M -equivalent for any compact space X of an infinite cardinality . Another pair
of M-equivalent spaces given in [TkaI] is WI with the interval topology and
WI E9 C(wt}. Thus, Tkachuk conc1udes that there exist M-equivalent spaces X
and Y satisfying one of the following:
(c) X and Y are compact, X does not contain non-trivial convergent sequences,
but Y does contain convergent sequences;
(d) all compact subsets of X are countable, but Y contains a compact subspace
of uncountable weight.
A general method for construction of M -equivalent spaces was developed by
Okunev in [Okt]. In fact, his method is a far-going generalization of Graev's idea
(see [GrID to show that the countable fan and the countable topological sum of
convergent sequences are M -equivalent. We can describe Okunev's construction
as follows. Let X be a completely regular space and K be a retract of X. If X+
is the space obtained by adding an isolated point to X and Y is the topological
sum of K and the R-quotient space XI K obtained from X by gluing K to a
point, then X+ and Y are M -equivalent. Another variation of this construction
makes use of a bunch of topological spaces. Let (X, xo) and (Y, YO) be spaces
with fixed points xo and YO. Then the quotient space (X E9 Y)/{xo, YO} is called
the bunch of (X, xo) and (Y, YO). This space is denoted by (X, xo) v (Y, YO).
Again, let K be a retract of a completely regular space X and xo e K. It is
shown in [Okl] that the spaces X and (XI K, *) v (K, xo) are M-equivalent,
where * is the image of K under the canonical projection of X onto XI K .
Applying this method to various spaces (sometimes fairly complicated ones),
Okunev presents in [Okl] a long list of properties which are not M-invariant.
This inc1udes the Baire property, local connectedness and path connectedness
(even within the c1ass of metric continua), countable 1l' -weight (even for compact
spaces), countable tightness, Frechet-Urysohn property, sequentiality, bisequen-
tiality, the property ofbeing a k-space, homotopy groups and, hence, homotopic
equivalence (even for finite-dimensional polyhedra), normality, collectionwise
normality (for countably compact spaces), hereditary normality, hereditary para-
compactness and hereditary Dieudonne-completeness (for compact scattered
spaces), and some others. It is also known (unpublished observation by Okunev
and Tkacenko) that dyadicity is not M-invariant within the c1ass of compact
spaces. However, under certain additional restrictions, F - and A-equivalence
relations become more stable. For example, if the spaces X and Y are metriz-
able, then the dispersion weights of X and Y coincide, where the dispersion
weight of aspace M is the minimum weight of non-empty open subsets of M
(see [GuOkD. It is also proved in [GuOk] that l-equivalence preserves local
compactness in metrizable spaces.
Cellularity is not an M -invariant cardinal function either. This fact was
announced without proofby Uspenskij in [Usp7]. Earlier, an analogous negative
result was established by Tkachuk in [Tka4] for A-equivalence. We recall that
M-equivalence implies A-equivalence, but no examples of A-equivalent spaces
that were not M -equivalent are known up to now. It is worth mentioning that
the cellularity becomes M -invariant in the c1ass of compact spaces (this follows
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1107
implicitly from the results of [Arh7]). Tkachuk [Tka2], [Tka3] extended this
result to l-equivalent spaces with the Baire property (in fact, this was done for
the so-called t-equivalent spaces).
On the other hand, some topological properties of aspace X are "easily" M-
invariant just because they admit a characterization in terms of the properties of
the free topological group F(X). For example, X is separable iff F(X) is separa-
ble, thus separability is an M -invariant property. Countable network weight and
a-compactness are other examples of such properties. Let us say that aspace X
with a a-discrete network is a a-space. We also call aspace X a-metrizable if
X is a union of countably many c10sed metrizable subspaces. By a subtle result
proved by Arhangel'skii in [Arh8], Xis a paracompact a-space (paracompact
a-metrizable space) iff so is the free topological group F(X). Therefore, the
properties ofbeing a paracompact a-space or a paracompact a-metrizable space
are "easy" M-invariant properties. By Theorem 5.7, DieudonmS-completeness
is also an "easy" M -invariant property.
A general result about the structure of M -equivalent spaces was obtained
by Pestov [Pe6] in 1984: if X and Y are M-equivalent spaces, then Y is a
countable union of subspaces each of which is homeomorphic to a subspace
of X. Therefore, Pestov conc1udes that hereditary cellularity (that is, spread),
hereditary separability and hereditary Lindelf number are M -equivalent cardi-
nal functions. In particular, if X and Y are M -equivalent Lindelf spaces and
one of them is perfectly normal, so is the other.
It is interesting to compare briefly various equivalence relations introduced
previously. By Graev's result in [Grl], M- and A-equivalence preserve com-
pactness. In 1982, Uspenskij [Usp2] generalized this result extending it to
l-equivalence. In fact he proved a stronger assertion: if the spaces Cp(X)
and Cp(Y) are uniformly homeomorphic and X is compact, then so is Y.
However, there exist Tikhonov spaces X and Y such that Cp(X) and Cp(Y)
are homeomorphic, but only one of the spaces X, Y is compact. In fact, one
can take X = [0. 1] and Y = lR [GuKh]. Tkachuk [Tka5] showed that if X
is a non-empty Tikhonov space and Y = X E9 N, then the free topological
groups F(X) and F(Y) are uniformly homeomorphic. This implies that neither
compactness, countable compactness nor pseudocompactness are preserved by
uniform homeomorphisms of free topological groups. Finally, we mention the
following interesting result of [Baa]: F - and l-equivalence relations coincide in
the c1ass of locally compact zero-dimensional separable metric spaces.
We finish our discussion of the 1- and M -equivalence relations with several
results conceming products of spaces. In 1982, Pavlovskii [Pav2] proved that if
spaces X, Y and Z are Dieudonne-complete, X and Y are l-equivalent and the
products X x Z, Y x Z are k-spaces, then X x Z and Y x Z are l-equivalent.
Eight years later, Arhangel'skll [Arhll], [ArhI3] showed that if compact spaces
X and Y are l-equivalent, then for any space Z. the product spaces X x Z and
1108 MICHAEL G. TKACHENKO
The proof of the existence of free topological groups given by Kakutani [Kak4]
makes use of two operations: the direct product of topological groups and taking
subgroups. Therefore, if we want to generalize the notion of a free topological
group to certain c1asses of topological groups, it is natural to restrict ourselves
to classes closed under these operations. Further, it is not indispensable to work
only with c1asses c10sed with respect to subgroups. Sometimes it suffices to
know that closed subgroups of the groups in a given c1ass belong to this c1ass.
For example, the compact topological groups form a c1ass like this.
Let 'V be a c1ass of topological groups. Suppose that X is a Tikhonov space
and a group G contains X as a subspace. We say that G = G(X, 'V) E 'V is a
free 'V-group on X ifthe pair (X, G) satisfies the following condition:
(FGC) An arbitrary continuous mapping f: X ~ H of X to a topolog-
ical group H E 'V can be uniquely extended to a continuous homomorphism
1: G~ H.
Kakutani's argument (see [Kak4]) shows that the free 'V-group G(X, 'V)
exists for every completely regular space X and for each of the following c1asses
'V: totally bounded (Abelian) groups, ~o-bounded (Abelian) groups (see Section
1.3), groups with a (quasi) invariant basis, etc. For each of these c1asses, the
subset X ~ G(X, 'V) is c10sed in G(X, 'V) [M02]. The existence of the free
6-group G(X, ~) for the class ~ of compact groups (or simply free compact
group) requires some changes in Kakutani's construction because subgroups of
compact groups need not be compact. However, every continuous homomor-
phism ({J: G ~ H of topological groups is uniformly continuous with respect
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1109
to the bilateral uniformities and, hence, ({J can be continuously extended to the
completions of the groups. This proves, in fact, that the completion of the free
totally bounded group G(X, "\1') on X (with "\I' being the class oftotally bounded
groups) is the free compact group on X. It is clear that X is closed in G(X, '(6) iff
Xis compact. Further, X generates a dense subgroup of G(X, '(6) (topologieally
isomorphie to G(X, "\1' and the latter subgroup never coincides with G(X, '(6).
Free (Abelian) compact groups are well-studied objects from topological and
algebraic points of view. The study of properties of free compact (Abelian)
groups is related mainly to the names of Hofmann and Morris [Hof], [HM1]-
[HM3], [HM5], [HM6]. A good description ofthe material is given in the survey
[CHR]. Since the algebraic aspects dominate there, we omit the details.
Morris [Mo 1] found necessary and sufficient conditions for the existence of
the free topologie al group G(X, "\1') relative to the variety "\1'.
An even more general approach to free topological groups was suggested in
the article [CM1] by Comfort and van Mill. Given two classes OlL and"\l' oftopo-
logieal groups and a Tikhonov space X, a topological group G = G(X, OlL, "\1')
containing X as a subspace is called the free (OlL, "\I')-group on X if G E OlL
and every continuous mapping f: X ~ H of X to an arbitrary topologi-
cal group H E "\I' can be uniquely extended to a continuous homomorphism
1: G ~ H. Let CZP A and '(6A be the classes of pseudocompact Abelian and
countably compact Abelian groups, respectively. Comfort and van Mill [CM1]
show that there is a free (CZP A, CZP A)-group over X iff X = 0, and for every
completely regular space X, there exists a free (CZP A, '(6A)-group in whieh X is
closed. Tkacenko [Tk8], [Tk14] and Fokkink [Fok] generalized the first of the
above results by proving that the free (CZP A, CZP A)-group on a non-empty space
X does not exist, even if we do not require that a homomorphism rextending
a mapping f: X ~ H should be unique.
For anormal topologieal space, there exist three different approaches to define
its dimension, due to Lebesgue and Cech (covering dimension dim), Urysohn
and Menger (smali inductive dimension ind), and Poincare, Brouwer, and Cech
(large inductive dimension Ind). All the three dimensions coincide for separable
metric spaces (Tumarkin-Urysohn-Hurewitz's theorem proved in the late 20s),
the equality dim X = Ind X holds for any metric space X (Katetov-Morita's
theorem, 1954), while in 1962, Roy [Roy] constructed his famous example of
a complete metric space Y such that ind Y = 0, but dim Y = Ind Y = 1.
In general, the functions ind, Ind and dim behave completely independently
of each other, subject to the trivial restrietion ind X ~ Ind X, which always
holds. Well-known examples show that the dimension functions dim, ind and
Ind need not coincide, even in the class of compact spaces [Nag] , [Pea]. In
1110 MICHAEL G. TKACHENKO
addition, continuous open mappings of compact metrizable spaces can raise all
these dimensions simuItaneousIy [KeI] and neither dim nor Ind are monotone
functions (see Chapter VII of [Eng]).
We will see in this Section how the topological group structure improves
the behaviour of dimensional characteristics of spaces (but still leaves some
possibilities for singularities). Let us start with Iocally compact topological
groups which Ieave no space for pathologies.
In particular, the equality ind G = Ind G holds for every (J' -compact topo-
logical group G. It is still unknown, however, if dim G and ind G coincide for
every (J'-compact group G (see Question 1.5 of [Sh7]).
To present further results on dimensions of topological groups, we need the
notion of an lR-factorizable group introduced by Tkacenko [TkI2], [Tk13].
Theorem 6.8. The equalities ind G = 0 and dim G = 0 are equivalent for
every lR-factorizable topological group G.
This problem has been mentioned by many specialists, but we are unable to
attribute it to anybody. There are no examples raising dimensions of subgroups,
and the problem remains unsolved to date. In some special cases, however,
the monotonicity problem has a positive solution. In 1990, Shakhmatov [Sh6]
proved that if H is a totally bounded subgroup of an arbitrary topological group
G, then dim H ::::: dim G. It is interesting to note that this result does not depend
on any additional normality condition and the group H is not assumed to be
c1osed.
In the same artic1e [Sh6] , Shakhmatov found two c1asses of topological
groups in which the covering dimension dirn is monotonous with respect to all
subgroups.
1114 MICHAEL G. TKACHENKO
The second assertion of Theorem 6.12 can be extended to retracts and G8-
subsets ofLindelf E-groups [Sh3]. It is not known, however, ifTheorem 6.12
is valid for ~o-bounded groups (see Question 3.4 of [Sh7]). There are no results
on the logarithmic inequality ind G x H ::::: ind G +ind H for topological groups
G and H (except for locally compact and u-compact groups, when one can
apply, respectively, Theorem 6.2 or Theorems 6.6 and 6.12 (2. The problem is
open, even for totally bounded groups [Sh7].
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1115
In 1978, Bel'nov [Bel] showed that every completely regular space X can be
embedded as a c10sed subspace into a completely regular homogeneous space Y
with dim Y = dim X. Since topological groups are homogeneous, this motivated
Bel'nov to ask whether every completely regular space X can be embedded into
a topological group G such that dim G = dim X. The following result of [Sh6]
answers this question in the negative.
Theorem 6.13. Ifn =1= 0, 1,3,7, then the n-dimensional sphere sn cannot be
embedded into an n-dimensional topological group (no matter which dimension
ind, Ind or dim is considered).
The condition n =1= 0, 1,3,7 appears in Theorem 6.13 because its proof
makes use of the Adams theorem [Ada] that sn is not an H -space unless n :/=
0,1,3,7. The importance of the Adams theorem was first noted by Uspenskij
who proved in 1987 that if n =1= 0,1,3,7, then Sn cannot be a retract of a
topological group. Both SO and SI are c1early topological groups themselves, S3
can easily be embedded as a c10sed subgroup into the group of quartemions and,
hence, is a topological grOUp. The compact space S7 also admits a topological
group structure compatible with its topology [Ada]. Actually, Kato [Kat] showed
that Theorem 6.13 is a special case of a general result about manifolds: A
compact n-dimensional manifold Mn without boundary can be embedded into
an n-dimensional topological group if and only if Mn can be equipped with a
group structure compatible with the topology of Mn (in other words, if Mn is a
topological group itselt).
The "gap" n =1= 0, 1,3,7 in Theorem 6.13 and the use of Adams' theorem
in its proof made it interesting to find out if the numbers 0, 1,3,7 are really
exceptional, and give "purely" topological counterexamples to the problem
posed by Bel'nov. The first step in this direction was done by Kimura [Kim].
He proved that the figure eight cannot be embedded into a topological group
of dimension one and asked if the "letter" T (a subspace of the figure eight)
admits an embedding into a one-dimensional topological group. Kulesza [Kul]
answered the latter question in the negative. In the same artic1e, Kulesza proved
that for every integer n ::: 1, there exists a compact metric space X n of dimension
n which cannot be embedded into an n-dimensional topological group, and the
metric hedgehog with (01 spines embeds in no finite-dimensional topological
group.
Unlike the general negative results, the zero-dimensional case is surprisingly
good. The following theorem is due to Shakhmatov [Sh6].
Theorem 6.14. Every completely regular space X with ind X = can be
embedded into a pseudocompact Abelian topological group G with dim G 0.
=
1116 MICHAEL G. TKACHENKO
In particular, Theorem 6.15 implies that the equality dim F(X) = 0 holds for
every zero-dimensional metrizable space X.
The investigation was then continued by Shakhmatov [Sh3], [Sh4], [Sh6],
Sipacheva [Sill and Tlcacenko [Tk4], [Tk6]. It is natural to try to extend the
above theorem to a wider c1ass of spaces X. In 1983, Tkacenko [Tk4] proved
that ind A(X) = 0 for every completely regular space X satisfying dim X = O.
Six years later, Sipacheva [Si1] showed that the conc1usion remains valid for
F(X). Combining these results, we obtain the following theorem.
Theorem 6.17. There is a Tikhonov space X with ind X = 0 such that the
groups F(X) and A(X) satisfy ind F(X) t=
0 and ind A(X) t=
O. In addition,
the space X can be chosen normal or pseudocompact.
The foregoing results leave open Arhangel' skil' s problem of whether ind X =
oimplies ind F(X) = 0 or ind A(X) = 0 for a metrizable space X (see [Arh8]).
Another problem posed by Arhangel'skiI was to find out if one can omit the
condition on X to be a paracompact cr-space in Theorem 6.15. In other words,
does dimX = 0 imply dimF(X) = 0 or dimA(X) = 0 for every completely
regular space X?
While these problems remain open, an analog of the second one was solved
in the affirmative by Shakhmatov [Sh6] for free topological groups relative to
special varieties of topological groups (see Section 6.2). Recall that a c1ass 'V
of topological groups is called a variety if 'V is c10sed with respect to direct
products and taking arbitrary subgroups of the groups in 'V.
Let us use TB (AT B) to denote the variety of totally bounded (Abelian)
topological groups and B(w) (AB(w to denote the variety of ~o-bounded
(Abelian) topologie al groups. It is easy to see that relative free topological
groups for these varieties exist and, for every Tikhonov space X, there are
continuous isomorphisms F(X) -+ G(X, B(w -+ G(X, TB) and A(X) -+
G(X, AB(w -+ G(X, AT B) [Sh6]. The following theorem of [Sh6] answers
Arhangel'skil's question for free topological groups relative to TB and AT B.
Note that to prove Theorem 6.18 it suffices to show that ind G(X, TB) =
0= ind G(X, AT B) and then apply Theorem 6.8 (and the fact that every totally
bounded group is lR-factorizable). It is also shown in [Sh6] that one cannot
weaken dim X = 0 to ind X = 0 in Theorem 6.18 (1). In fact, Shakhmatov
presents two different spaces X (one normal and the other pseudocompact)
such that ind X = 0 but ind G(X, 'V) t= 0 for each 'V E {TB, AT B, B(w),
AB(w)}.
1118 NUCHAELG.TKACHENKO
7. Miscellaneous
The limitation of the volume does not allow us to present all the interesting and
important areas in topological groups adjacent to General Topology. Here we
mention briey only some of them and give references to the main sources of
information.
(2) G / N with the quotient topology is a topological group for every N E .N.
Then (G,5') is called a protopological group. From a result of [Cov], every
protopological group with continuous multiplication is a topological group.
The following highly non-trivial theorem was proved by Reznichenko in
[Rez]: every pseudocompact paratopological group is a topological group. In
fact, Reznichenko's theorem finishes the work started by Korovin [Kor].
A different problem is to find conditions on a topological semigroup Gwhich
imply that G is a topological group. By a theorem of Numakura [Numa], every
compact topological semigroup satisfying the left and right cancellation laws
is a topological group. In 1955, A. D. Wallace [Wal] asked whether this result
could be extended to countably compact topological semigroups. Mukherjea and
1120 MICHAEL G. TKACHENKO
Tserpes [MuTs] answered Wallace's question in the affirmative for the special
case of first countable countably compact semigroups. Pfister [Pfi] showed that
every completely regular countably compact topological semigroup which is
algebraically a group is, in fact, a topological group. Grant [Gra3] gave a positive
answer to Wallace's question in the case of a completely regular sequentially
compact cancellative topological semigroup. Yur'eva [Yur] generalized both
Grant's and Mukherjea-Tserpes's results as follows: every countably com-
pact sequential topological semigroup is a topological group. Finally, Robbie
and Svetlichny [RoSv] constructed a countably compact subsemigroup of TC
in 1996 which fails to be a group, thus answering the Wallace question in
the negative. Rowever, their construction makes use of CR. Recently, Tomita
[Tom2] presented a construction of a similar subsemigroup of TC that depends
on MAcountable. It is not known whether such a subsemigroup can be constructed
in ZFC.
ties ofR-factorizable groups, as well as the list of open problems, in the survey
artic1e [Tk18].
paet [DTIl]. The problem about the existenee (in ZFC) of a pseudoeompaet
topological group without suitable sets has been solved in the affirmative in
[OTII]. The artic1e [OTI2] contains many results about suitable sets in quotient
groups, minimal groups, in produets, direet sums and in the groups equipped
with the Bohr topology. The existenee of suitable sets in free topologieal groups
on eompaet spaees is eonsidered in [OkTk] where it is shown that the free
topologie al group F(X) on a dyadie eompaet spaee X contains a suitable set.
Suitable sets in finite produets of topological groups and in Abelian groups with
the Bohr topology are studied by Tomita and Trigos-Arrieta in [TI]. The fol-
lowing result has been proved independently in [DTI2] and [TI]: every loeally
eompaet Abelian group endowed with the Bohr topology (i.e., with the finest
tota1ly bounded group topology eoarser than the original one) has a suitable set.
In addition, a diserete Abelian group with the Bohr topology contains a c10sed
diserete subset which algebraieally generates the group [DTI2].
References
[Ada] Adams, J.F., On the non-existence of elements of Hopf invariant one, Ann.
o! Math. 72 (1960),20-104.
[Ale] Alexandrov, A.D., On extension of a Hausdorff space to an H -closed space,
Dokl. AN SSSR 37 (1942), 138-141.
[Ami] Amirdzhanov, G.R., On dense subspaces of countable pseudocharacter and
other generalizations of separability, Dokl. AN SSSR 235 (1977), 993-996.
[Are] Arens, R., Note on convergence in topology, Math. Mag. 23 (1950), 229-
234.
[Arhl] Arhangel'skiI, A.v" Mappings related to topological groups, Soviet Math.
Dokl. 9 (1968), 1011-1015.
[Arh2] _ _ , Topological spaces and mappings, Remarks on Topological Groups,
Moscow Univ. Pub!., Moscow 1969 (in Russian).
[Arh3] _ _ , Structure and classification of topological spaces and cardinal in-
variants, Russian Math. Surveys 33 (1978), 33-96. Russian original in:
Uspekhy Mat. Nauk 33 (1978), 29-84.
[Arh4] _ _ , On a-expanded spaces, Dokl. Akad. Nauk SSSR 239 (1978), 505-
508.
[Arh5] _ _ , The frequency spectrum of a topological space and the product
operation, Trans. Moscow Math. Soc. 40 (1981), 163-200. Russian original
in: Trudy Moskov. Mat. Obshch. 40 (1979),171-206.
[Arh6] _ _ , Cardinal invariants oftopological groups. Embeddings and conden-
sations, Soviet Math. Dokl. 20 (1979), 783-787. Russian original in: Dokl.
AN SSSR 247 (1979), 779-782.
[Arh7] _ _ , Relations among the invariants of topological groups and their
subspaces, Russian Math. Surveys 35 (1980), 1-23. Russian original in:
Uspekhy Mat. Nauk 35 (1980), 3-22.
1124 MICHAEL G. TKACHENKO
[EIl] EIlis, R., Locally compact transformation groups, Duke Math. 1. 24 (1957),
119-125.
[Eng] Engelking, R., General Topology, Heldermann Verlag, Berlin 1989.
[EV] Esenin-Vol'pin, AS., On the relation between the local and integral weight
in dyadic bicompacta, Dokl. Akad. Nauk SSSR 68 (1949), 441-444.
[FOT] Fay, T.H., Ordman, E.T. and Thomas, B.Y.S., The free topological group
over rationals, Topology Appl. 10 (1979), 33-47.
[FZ] Flachsmayer, J. and Ziechang, H., Uber die schwache konvergenz der
Haarschen Mae von untergruppen, Math. Ann. 156 (1964), 1-8.
[Flo] Flood, J., Free Topological Vector Spaces, Ph.D. thesis (Australian National
Univ., Canberra, 1975).
[Fok] Fokkink, R., A note on pseudocompact groups, Proc. Amer. Math. Soc. 107
(1989),569-571.
[FolI] F~lner, E., Generalization of a theorem of Bogoliouboff to topological
Abelian groups, Math. Scand. 2 (1954), 5-18.
[FoI2] _ _ , Note on a generalization of a theorem of Bogoliouboff, Math.
Scand. 2 (1954),224-226.
[FreI] Freudenthai, H., Einige Stze ber topologische Gruppen, Ann. of Math.
37 (1936), 46-56.
[Fre2] _ _ , Topologische Gruppen mit gengend vielen fastperiodischen Funk-
tionen, Ann. of Math. 13 (1951), 57-77.
[Fre3] _ _ , La structure des groupes a deux bouts et des groupes triplement
transitifs, Nederl. Akad. Wetensch. Indag. Math. 13 (1951), 288-294.
[Fro] Frobenius, G., ber Gruppencharaktere, Sitzungsber. Preuss. Akad. Wiss.,
phys.-math. Kl. 1989,985-1021.
[GH] Galindo, J. and Hemandez, S., Pontryagin-van Kampen reflexivity for free
Abelian topological groups, Forum Math. 11 (1999), 399-415.
[GRS] Gartside, P.M., Reznichenko, E.A and Sipacheva, O.Y., Maltsev and retral
spaces, Topology Appl. 80 (1997), 115-129.
[GR1] Gel'fand, I.M. and Ra'ikov, D., Irreducible unitary representations oflocally
bicompact groups, Mat. Sbornik, N.S. 13 (55)(1943), 301-316 (in Russian).
[GR2] _ _ , Irreducible unitary representations of locally bicompact groups,
Dokl. AN SSSR 42 (1944), 199-201 (in Russian).
[Gerl] Gerlitz, J., On subspaces of dyadic compacta, Studia Scientarum Math.
Hungarica 11 (1976), 115-120.
[Ger2] _ _ , Continuous functions on products of topological spaces, Fundam.
Math. 106 (1980), 67-75.
[GiRS] Ginsburg, J., Rajagopalan, M. and Saks, Y., On the density character of
closed subgroups, Proc. Amer. Math. Soc. 57 (1976), 148-150.
[GL] General Algebra, vol. I, Mathematical Library. Nauka, Moscow 1990.
[Gle2] Gleason, AM., Groups without small subgroups, Ann. Math. 56 (1952),
193-212.
[Glul] Glushkov, VM., Locally nilpotent locally bicompact groups, Trudy Moskov.
Mat. Obshch. 4 (1954), 291-332 (in Russian).
[Glu2] _ _ , Structure oflocally compact groups and the fifth problem ofHilbert,
Uspekhy Mat. Nauk 12 (1957),3-41 (in Russian).
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1129
[Gd] Graev, M.I., Free topological groups, in Topology and Topological Algebra,
Translations Series 1, vol. 8, 1962, pp. 305-364, American Mathematical
Society, Russian original in: Izvestiya Akad. Nauk SSSR Sero Mat. 12 (1948),
279-323.
[Gr2] _ _ , Theory oftopological groups I, Uspekhy Mat. Nauk 5 (1950),3-56
(in Russian).
[Gr3] _ _ , On free products of topological groups, Izv. AN SSSR Sero Mat. 14
(1950),343-354 (in Russian).
[Gral] Grant, D.L., Topological groups which satisfy an open mapping theorem,
Pacific J. Math. 68 (1977), 411-423.
[Gra2] _ _ , Products of topological groups which satisfy an open mapping
theorem, Topology Proc. 17 (1992), 131-135.
[Gra3] _ _ , Sequentially compact cancellative topological semigroups: some
progress on the Wallace problem, in Papers on General Topology and
Applicationa, VlIth Summer Conference at the University of Wisconsin
(Madison, 1991), Annals of the New York Acad. Sci., vol. 704, 1993,
150--154.
[Gul] Gul'ko, S.P., On uniform homeomorphisms of spaces of continuous func-
tions, Proc. ofthe Steklov Institute ofMath., No. 3 (1993), 87-93. Russian
original in: Trudy Steklov. Inst. 193 (1992), 82-88.
[GuKh] Gul'ko, S.P. and Khmyleva, T.B., Compactness is not preserved by the
relation of t-equivalence, Mat. Zametky 39 (1986), 895-903 (in Russian).
[GuOk] Gul'ko, S.P. and Okunev, O.G., Local compactness and M-equivalence,
in Problems in Geometry and Topology, Petrozavodsk 1982, 14-23 (in
Russian).
[Gu1] Guran, 1., On topological groups close to being Lindelf, Soviet Math.
Dokl. 23 (1981),173-175.
[Gu2] _ _ , Topology of an infinite symmetric group and condensations,
Comment. Math. Univ. Carolin. 22 (1981), 311-316 (in Russian).
[Gu3] _ _ , An example of a non-metrizable minimal topological group of
countable pseudocharacter, Dokl. AN USSR no. 1 (1986),6-9 (in Russian).
[Haa] Haar, A., Der Mabegriff in der Theorie der kontinuierlichen Gruppen,
Ann. of Math. 34 (1933), 147-169.
[Hag] Hagler, 0., On the structure of Sand C(S) for S dyadic, Trans. Amer. Math.
Soc. 214 (1975), 415-428.
[HaJu] Hajnal, A. and Jubasz, 1., A separable normal topological group need not
be Lindelf, General Topol. Appl. 6 (1976), 199-205.
[HaI] Halmos, P.R., Measure Theory, Van Nostrand, New York 1950.
[HMT] Hardy, J.P., Morris, S.A. and Thompson, H.B., Application of the Stone-
Cech compactification to free topological groups, Proc. Amer. Math. Soc.
55 (1976), 160--164.
[HvM1] Hart, K. and van Mill, J., A countably compact group H such that H x H
is not countably compact, Trans. Amer. Math. Soc. 323 (1991), 811-821.
[HvM2] _ _ , A separable normal topological group which is not Lindelf,
Topology Appl. 20 (1985), 279-287.
1130 MICHAEL G. TKACHENKO
[HaHu] Hartman, S. and Hulanicky, A., Sur les ensembles de puissance minimum
dans les groupes topologiques, Colloq. Math. 6 (1958), 187-191.
[HaMi] Hartman, S. and Mycielski, J., On embeddings of topological groups into
connected topological groups, Colloq. Math. 5 (1958), 167-169.
[Hau] Hausdorff, F., ber innere Abbildungen, Fund. Math. 25 (1934), 279-291.
[Hay] Haydon, R., On a problem of PelczyDski: Milutin spaces, Dugundji spaces
and AE(O-dim), Studia Math. 52 (1974), 23-31.
[HeSa] Hemandez, S. and Sanchis, M., Ga-open functionally bounded subsets in
topological groups, Topology Appl. 53 (1993), 289-299.
[HST] Hemandez, S., Sanchis, M. and Tkacenko, M., Bounded sets in spaces and
topological groups, Topology Appl. 101 (2000), 21-43.
[HeTk] Hemandez, C. and Tkacenko, M., Subgroups of R-factorizable groups,
Comment. Math. Univ. Carolin. 39 (1998), 371-378.
[Hew] Hewitt, E., A problem of set-theoretic topology, Duke Math. 1. 10 (1943),
309-333.
[HR] Hewitt, E. and Ross, K, Abstract Harmonic Analysis, vol. I, Springer-
Verlag, Berlin-Gttingen-Heidelberg 1979.
[Hof] Hofmann, KH., An essay on free compact groups, Lecture Notes in
Mathematics, vol. 915, Springer-Verlag, Berlin 1982, 171-197.
[HM1] Hofmann, K.H. and Morris, S.A., Free compact groups I: Free compact
Abelian groups, Topology Appl. 23 (1986), 41-64; Errata 28 (1988), 101-
102.
[HM2] _ _ , Free compact groups TI: the center, Topology Appl. 28 (1988), 215-
231.
[HM3] _ _ , Free compact groups ill: Free semisimple compact groups, in Cat-
egorical Topology, J. Adamek and S. MacLane (eds.), World Scientific,
Singapore, pp. 208-219.
[HM4] _ _ , Weight and c, J. Pure Appl. Algebra 68 (1990), 181-194.
[HM5] _ _ , Free compact groups IV: Splitting the component and the structure
of the commutator group, 1. Pure Appl. Algebra 70 (1991), 89-96.
[HM6] _ _ , Free compact groups V: Remarks on projectivity, in Category The-
ory at Work, H. Herrlich and H.-E. Porst (eds.), Heldermann Verlag, Berlin
1991, 177-198.
[HM7] _ _ , The Structure of Compact Groups: A Primer for Students - A
Handbookfor Experts, Walter de Gruyter Pubi., Berlin 1998.
[HoMo] Hofmann, KH. and Mostert, P., Splitting in topological groups, Memoirs
of Amer. Math. Soc. 43, Providence, Rhode Island 1963.
[HuMo] Hunt, D.C. and Morris, S.A., Free subgroups of free topological groups,
in Proc. Second Internat. Conf. on the Theory of Groups 1973, M. F.
Newman (ed.), Lecture Notes in Mathematics, vol. 372, Springer-Verlag,
Berlin 1974, pp. 377-387.
[Hur] Hurwitz, A., ber die Erzeugung der Invarianten durch Integration, Nachr.
k. Ges. Gttingen, math.-phis. Kl. 1897,71-90.
[Hus] Husain, T., Introduction to Topological Groups, W.B. Saunders Co.,
Philadelphia and London 1966.
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1131
[Mall] Malykhin, v.1., Extremally disconnected and similar groups, Soviet Math.
Dokl. 16 (1975),21-25.
[MaI2] _ _ , Topology and forcing, Uspekhy Mat. Nauk 38 (1983), 69-118.
[Mal3] _ _ , Nonpreservation of properties of topological groups at raising them
to square, Siberian Math. J. 28 (1987), 639-645. Russian original in: Sib.
Mat. Zh. 28 (1987), 154-161.
[MaPr] Malykhin, v.1. and Protasov, LV., Maximal resolvability ofbounded groups,
Topology Appl. 73 (1996), 227-232.
[Mam] Mamford, D., Aremark on Mahler's compactness theorem, Proc. Amer.
Math. Soc. 28 (1971), 289-294.
[MarI] Markov, A.A., On free topological groups, Dokl. AN SSSR 31 (1941),299-
301 (in Russian).
[Mar2] _ _ , On free topological groups, in Topology and Topological Algebra,
Translation Series 1, vol. 8, 1962, 195-272. American Math. Soc. Russian
original in: Izvestiya Akad. Nauk SSSR 9 (1945), 3-64.
[Mau] Maurice, M.A., Compact Ordered Spaces, Math. Centre Tracts, vol. 6,
Math. Centrum, Amsterdam 1964.
[Megi] Megrelishvili, M.G., Group representations and construetion of minimal
topologie al groups, Topol. Appl. 62 (1995), 1-19.
[Meg2] Megrelishvili Levi, M., G-minimal topologieal groups, in Abelian Groups,
Module Theory and Topology, D. Dikranjan and L. Sake (eds.), Leeture
Notes in Pure and Appl. Math., vol. 201, Mareel Dekker Ine., New York,
1998,289-299.
[Mie] Michael, E., A quintuple quotient quest, General Topology Appl. 2 (1972),
91-138.
[Mil] Mill, J. van, A homogeneous Eberlein eompact spaee which is not
metrizable, Pacific J. Math. 101 (1982), 141-146.
[MiPo] Mironets, A.A. and Poletskih, V.M., The spaee of normal subgroups of a
nilpotent group, Dokl. AN Ukr. SSR (5) 1986, 19-22.
[MoZ1] Montgomery, D. and Zippin, L., Small subgroups of finite-dimensional
groups, Ann. Math. 56 (1952), 213-241.
[MoZ2] _ _ , Topological Transformation Groups, Interscienee Tracts in Pure
and Applied Mathematics, vol. 1, Interscienee Publishers, Ine., New York-
London 1955.
[Mol] Morris, S.A., Varieties of topological groups, Bull. Austral. Math. Soc. 1
(1969), 145-160.
[M02] _ _ , Varieties oftopologieal groups: a survey, Colloq. Math. 46 (1982),
147-165.
[Mo3] _ _ , Free Abelian topological groups, in Proc. Conf Toledo, Ohio 1983,
Helderman Verlag, Berlin 1984, 375-391.
[MoNi] Morris, S.A. and Niekolas, P., Loeally invariant topologies on free groups,
Pacific J. Math. 103 (1982), 523-537.
[MNPS] Morris, S.A., Nickolas, P., Pestov, v.G. and Svetliehny, S., Quotients of
free topological groups on metrizable spaces, preprint.
[MoPe] Morris, S.A. and Pestov, V.G., Open subgroups offree Abelian topological
groups, Math. Proc. Cambr. Philos. Soc. 114 (1993), 439-442.
1134 MICHAEL G. TKACHENKO
[MoTl] Morris, S.A. and Thompson, H.B., Free topological groups with no small
subgroups, Proc. Amer. Math. Soc. 46 (1974), 431-437.
[MoT2] _ _ , Metrizability of subgroups of free topological groups, Bull. Austral.
Math. Soc. 33 (1986),103-112.
[Mos] Mostert, P.S., Sections in principal fibre spaces, Duke Math. 1. 23 (1956),
57-72.
[MuTs] Mukherjea, A. and Tserpes, N.A., A note on countably compact semi-
groups, 1. Austral. Math. Soc. 13 (1972), 180-184.
[Muk] Mukhin, Yu.N., Topological groups, J. Soviet Math. 28 (1985),825-870.
[Nach] Nachbin, L., On the finite dimensionality of every irreducible representation
of a compact group, Proc. Amer. Math. Soc. 12 (1961), 11-12.
[Nag] Nagami, K., Dimension Theory, New York 1970.
[Nak] Nakayama, T., Note on free topological groups, Proc. Imp. Acad. Sei. 19
(1943),471-475.
[Neu 1] Neumann, J. von, Zum Raarschen Ma in topologischen Gruppen, Com-
positio Math. 1 (1934), 106-114.
[Neu2] _ _ , The uniqueness of Haar's measure, Mat. Sbomik, N. S. 43 (1936),
721-734.
[Nil] Nickolas, P., Free Topological Groups and Free Products 0/ Topological
Groups, Ph.D. dissertation, Univ. ofNew South Wales, Australia 1976.
[Ni2] _ _ , Subgroups of the free topological group on [0, 1],1. London Math.
Soc. 12 (1976),199-205.
[Ni3] _ _ , A Kurosh subgroup theorem for topological groups, Proc. London
Math. Soc. 42 (1981), 461-477.
[Nob] Noble, N., The continuity offunctions on Cartesian products, Trans. Amer.
Math. Soc. 149 (1970), 187-198.
[NST] Nogura, T., Shakhmatov, D. and Tanaka, Y., Metrizability of topological
groups having weak topologies with respect to good covers, Topology Appl.
S4 (1993), 203-212.
[Nov] Novak, 1., On the Cartesian product of two compact spaces, Fund. Math.
40 (1953), 106-112.
[Numa] Numakura, K., On bicompact semigroups, Math. 1. Okayama Univ. 1
(1952),99-108.
[Numm] Nummela, E.C., Uniform free topological groups and Samuel compactifi-
cations, Topology Appl. 13 (1982), 77-83.
[Nyi] Nyikos, P., Metrizability and the Frechet-Urysohn property in topological
groups, Proc. Amer. Math. Soc. 83 (1981), 793-801.
[Ok1] Okunev, O.G., A method for constructing examples of M -equivalent spaces,
Topology Appl. 36 (1990), 157-171, Correction, Topology Appl. 49 (1993),
191-192.
[0k2] _ _ , M-equivalence of products, Trans. Moscow Math. Soc. S6 (1995),
149-158. Russian original in: Trudy Moskov. Mat. Obsch. S6 (1995).
[Ok3] _ _ , Aremark on the tightness of products, Comment. Math. Univ.
Carolin. 37 (1996), 397-399.
TOPOLOGICAL FEATURES OF TOPOLOGICAL aROUPS 1135
[OkTk] Okunev, o.a. and Tkacenko, M.a., On thin generated sets in topological
groups, Abelian Groups, Module Theory, and Topology, D. Dikranjan and
L. Salce (eds.), Marcel Dekker, Inc., New York 1998. pp. 327-342.
[OrTh] Ordman, E.T. and Smith-Thomas, B.V., Sequential conditions and free
topological groups, Proc. Amer. Math. Soc. 79 (1980), 319-326.
[Pas 1] Pasynkov, B.A., On coincidence of different definitions of dimension
for locally compact groups, Dokl. AN SSSR 132 (1960), 1035-1037 (in
Russian).
[Pas2] _ _ , On coincidence of different definitions of dimension for quotient
spaces oflocally compact groups, Uspekhy Mat. Nauk 17 (1962),129-135
(in Russian).
[Pas3] _ _ , Almost metrizable topological groups, Dokl. AN SSSR 161 (1965),
281-284 (in Russian).
[Pas4] _ _ , On spaces with a compact group of transformations, Soviet Math.
Dokl. 17 (1976), 1522-1526.
[Pas5] _ _ , On the relative cellularity of Lindelf subspaces of topological
groups, Topology Appl. 57 (1994), 249-258.
[PaVa] Pasynkov, B.A. and Valov, V.M., On free groups of topological spaces,
Dokl. Bulgar. Acad. Sei. 34 (1981), 1049-1052.
[Pavi] PavlovskiI, D.S., On spaces of continuous functions, Dokl. AN SSSR 253
(1980), 38-41.
[Pav2] _ _ , On spaces having linearly homeomorphic spaces of continuous
functions in the topology of pointwise convergence, Russian Math. Surveys
37 (1982). Russian original in: Uspekhy Mat. Nauk 37 (1982), 185-186.
[Pea] Pears, A.R., Dimension Theory ofGeneral Spaces, Cambridge Univ. Press,
Cambridge 1975.
[PSW] Pelant, J., Shakhmatov, D.B. and Watson, S., A universal complete metric
Abelian group of a given weight, in Topology with Applications, Szekszard
(Hungary), Bolyai Soeiety Mathematical Studies, 4 (1993), 431-439.
[Pel] Pelczynski, A., Linear Extensions, Linear Averagings and their Applica-
tions, Diss. Math. 58. Warsaw, PWN 1968.
[PeWe] Peter, F. and Weyl, H., Die Vollstndigkeit der primitiven Darstellungen
eigen geschlossenen kontinuierlichen Gruppe, Math. Ann. 97 (1927), 737-
755.
[pe 1] Pestov, v.a., Some properties of free topological groups, Moscow Univ.
Math. Bull. 37 (1982), 46-49.
[Pe2] _ _ , On embeddings and condensations of topological groups, Math.
Notes 31 (1982), 228-230.
[Pe3] _ _ , The coincidence of the dimensions dim of l-equivalent topological
spaces, Soviet Math. Dokl. 26 (1982), 380-382.
[pe4] _ _ , Relations between classes of almost metrizable, projectively metriz-
able, and ~o-representable topological groups, in: Topoi. Spaces and Their
Mappings, Latvian Univ. Press, Riga 1983, pp. 80-86 (in Russian).
[Pe5] _ _ , Topological groups and algebraic envelopes oftopological spaces,
Ph.D. thesis, Moscow Univ., Moscow 1983 (in Russian).
1136 MICHAEL G. TKACHENKO
[PrI5] _ _ , Resolvability of free groups, Dokl. NAN Ukr. (8) (1997), 38-40 (in
Russian).
[pr16] _ _ , Resolvability of groups, Mat. Stud. 9 (1998), 130--148 (in Russian).
[Pr17] _ _ , Nonresolvable topologies on groups, Ukr. Mat. Zh. 50 (1998),
1646-1655 (in Russian).
[PrCh] Protasov, LV. and Charin, V.S., On the projection of topological groups,
Mat. Zametky 24 (1978), 383-389 (in Russian).
[PrCha] Protasov, LV. and Charyev, A., Bourbaki spaces oftopological groups, Ukr.
Mat. Zh. 42 (1990), 542-549 (in Russian).
[PrSa] Protasov, LV. and Saryev, A., Topological Abelian groups with a locally
compact lattice of closed subgroups, Dokl. AN Ukr. SSR Sero A (3) (1980),
29-32 (in Russian).
[PrStu] Protasov, LV. and Stukotilov, V.S., Ochan's topologies on the space of
closed subgroups, Ukr. Mat. Zh. 41 (1989), 1337-1342 (in Russian).
[PrTsl] Protasov, LV. and Tsybenko, Yu.V., Connectedness in the space of sub-
groups of a topological group, Ukr. Mat. Zh. 35 (1983), 382-385 (in
Russian).
[PrTs2] _ _ , Chabauty topology in the lattice of closed subgroups, Ukr. Mat. Zh.
36 (1984), 207-213 (in Russian).
[Przl] Przymusitiski, T., Normality and paracompactness in finite and Cartesian
products, Fund. Math. 105 (1980), 87-104.
[Prz2] _ _ , On the dimension of product spaces and an example of M. Wage,
Proc. Amer. Math. Soc. 76 (1979), 315-321.
[Rai] Ralkov, D.A., On completion of topological groups, Izv. AN SSSR 10
(1946),513-528 (in Russian).
[Rem] Remus, D., A short solution of Markov's problem on connected group
topologies, Proc. Amer. Math. Soc. 110 (1990), 1109-1110.
[ReSt] Remus, D. and Stoyanov, L., Complete minimal and totally minimal groups,
Topoi. Appl. 42 (1991), 57-69.
[Rez] Reznichenko, E.A., Extensions of functions defined on products of pseu-
docompact spaces and continuity of the inverse in pseudocompact groups,
Topology Appl. 59 (1994), 233-244.
[Rib] Ribes, L., Introduction to Profinite Groups and Galois Cohomology,
Queen's Papers Pure Appl. Math. 1970, No. 2.
[RoSv] Robbie, D. and Svetlichny, S., An answer to A.D. Wallace's question about
countably compact cancellative semigroups, Proc. Amer. Math. Soc. 124
(1996), 325-330.
[Rob] Robinson, D.J.F., A Course in the Theory 0/ Groups, Springer-Verlag,
Berlin 1982.
[RoDi] Roelcke, W. and Dierolf, S., Uniform Structures on Topological Groups and
their Quotients, McGraw-Hill International Book Company, New York-
Toronto 1981.
[Roy] Roy, P., Failure of equivalence of dimension concepts for metric spaces,
Bull. Amer. Math. Soc. 68 (1962), 609-613.
[Rud] Rudin, M.E., Lectures on Set-Theoretic Topology, Conf. Sero in Math.,
No. 23, Amer. Math. Soc., Providence 1975.
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1139
[She] Shelah, S., On P-points, w and other results in general topology, Notices
Amer. Math. Soc. 25 (1978), A-365,87T-G49.
[Shil] Shirokov, L., On bicompacta which are continuous images of dense sub-
spaces of topological products, manuscript deposited by VINITI (Moscow,
1981),2946-81 Dep. (in Russian).
[Shi2] _ _ , An external characterization of Dugundji spaces and /C-metrizable
compact Hausdorff spaces, Soviet Math. Dokl. 25 (1982), 507-510. Russian
original in: Doklady Akad. Nauk SSSR 263 (1982), 1073-1077.
[Sil] Sipacheva, O.V., Zerodimensionality and completeness in free topological
groups I, n Serdica 15 (1989), 119-140; 141-154 (in Russian).
[Si2] _ _ , On free topological groups with the inductive limit topologies,
Annals 0/ New York Acad. 0/ Sei. 788 (1996), 188-196.
[Si3] _ _ , Free topological groups of spaces and their subspaces, Topology
Appl. 101 (2000), 181-212.
[SiTk] Sipacheva, O.V. and Tkacenko, M.G., Thin and bounded subsets of free
topological groups, Topology Appl. 36 (1990), 143-156.
[Sius] Sipacheva, O.v. and UspenskiI, V.V., Free topological groups with no small
subgroups and Graev metrics, Moscow Univ. Math. Bull. 42 (1987), 24-29.
[Siw] Siwiec, E, Generalizations of the first axiom of countability, Rocky
Mountain J. Math. 5 (1975), 1-60.
[SkI] SkIyarenko, B.G., On topological structure of locally compact groups and
their quotient groups, Matem. Sb. 60 (1963),63-88 (in Russian).
[SS] SkIyarenko, B.G. and Smirnov, Ju.M., Some problems of dimension the-
ory, Trans. IVth All-Union Math. Congress I (plenary papers) (Leningrad,
1961), 219-226 (in Russian).
eSte] Stephenson, R.M., Minimal topological groups, Math. Ann. 192 (1971),
193-195.
[Stil Stieglitz, A., Chabauty und Weil-Raume topologischer Gruppen, Math.
Nachr. 85 (1978), 215-233.
[Sto] Stone, M.H., On the compactification of topological spaces, Ann. Soc.
Polon. Math. 21 (1948), 153-160.
[Sul] Sully, L., A note on B- and Br-complete topological groups, Proc. Camb.
Phil. Soc. 66 (1969), 275-279.
[Tho] Thompson, H.B., Aremark on free topological groups with no small
subgroups, J. Austral. Math. Soc. 18 (1974), 482-484.
[THVR] Tkacenko, M.G., Hernandez, C., Villegas, L.M. and Rend6n, 0., Topo-
logical Groups. Universidad Aut6noma Metropolitana, Mexico 1998 (in
Spanish).
[Tkl] Tkacenko, M.G., On the completeness of free Abelian topological groups,
Soviet Math. Dokl. 27 (1983), 341-345. Russian original in: Dokl. AN SSSR
269 (1983), 299-303.
[Tk2] _ _ , The Souslin property in free topological groups over compact
spaces, Mat. Notes 34 (1983), 790-793.
[Tk3] _ _ , Free topological groups and related topics, Colloq. Math. 41 (1983),
609-623.
TOPOLOGICAL FEATURES OF TOPOLOGICAL GROUPS 1141
[UsplO] _ _ , Why compact groups are dyadic, in Proc. ofthe Vlth Prague Topo-
logical Symposium 1986, Frolfk Z. (ed.), Heldermann Verlag, Berlin 1988,
601-610.
[Usp11] _ _ , Topological groups and Dugundji compacta, Math. USSR Sbomik
67 (1990), 555-580. Russian original in: Matem. Sbomik 180 (1989), 1092-
1118.
[UspI2] _ _ , On the group of isometries of the Urysohn universal metric space,
Comment. Math. Univ. Carolin. 31 (1990), 181-182.
[Usp13] _ _ , Free topological groups ofmetrizable spaces, Math. USSR Izvestiya
37 (1991), 657-680. Russian original in: Izvestiya Akad. Nauk SSSR Sero
Mat. 54 (1990), 1295-1319.
[UspI4] _ _ , On the Souslin number of subgroups of products of countable
groups, Acta Univ. Carolin. Math. Phys. 36 (1995), 85-87.
[UspI5] _ _ , On sequentially continuous homomorphisms oftopological groups,
preprint.
[UspI6] _ _ , On subgroups ofminimal topological groups, preprint.
[ViI] Vilenkin, N.Y., On the theory of weakly separable topological groups, Mat.
Sbomik 22 (1948),135-177 (in Russian).
[Vi2] _ _ , On the dyadicity of the group space of bicompact commutative
groups, Uspekhy Mat. Nauk 13 (1958), 79-80 (in Russian).
[Vill] Villegas-Silva, L.M., On resolvable spaces and groups, Comment. Math.
Univ. Carolin. 36 (1995), 579-584.
[Wal] Wallace, A.D., The structure of topological semigroups, Bull. Amer. Math.
Soc. 61 (1955), 95-112.
[Wanl] Wang, S.P., On the limits of subgroups in a group, Amer. 1. Math. 92 (1970),
708-724.
[Wan2] _ _ , Limits of lattices in a Lie group, Trans. Amer. Math. Soc. 133
(1968),519-526.
[Wel] Weil, A., La mesure invariante dans les espaces de groupes et les espaces
homogenes, Enseingment Math. 35 (1936), 24l.
[We2] _ _ , Sur les groupes topologiques et les groupes mesures, C. R. Acad.
Sei. Paris 202 (1936), 1147-1149.
[We3] _ _ , Sur les espaces a structure uniforme et sur la topologie generale,
Hermann & Cie, Paris. Publ. Math. Univ. Strasbourg 1937.
[We4] _ _ , L'integration dans les groupes topologiques et ses applications,
Hermann & Cie, Paris. Actualites Scientifiques et Industrielles, Publ. Math.
Institut Strasbourg 1941 and 1951.
[Will Willis, G., The structure of totally disconnected locally compact groups,
Math. Ann. 300 (1994), 341-361.
[Yal] Yamabe, H., On conjecture of Iwasawa and Gleason, Ann. of Math. 58
(1953), 48-54.
[Ya2] _ _ , A generalization of a theorem of Gleason, Ann. of Math. 58 (1953),
351-365.
[Yaml] Yamada, K., Characterizations of a metrizable space X such that every
An (X) is a k-space, Topology Appl. 49 (1993), 75-94.
1144 MICHAEL G. TKACHENKO
[Yam2] _ _ , Free Abelian topologieal groups and k-spaees, Glas. Mat. Sero 3 1
(1996), no. 1,83-91.
[Yama] Yamanoshita, T., On the dimension of homogeneous spaees, J. Math. Soc.
Japan 6 (1954),151-159.
[YI] Yosida, K. and lwamura, T., Equivalenee of two topologies of Abelian
groups, Proc. Imp. Acad. Tokyo 20 (1944), 451-453.
[Yur] Yur'eva, A.A., Countably eompaet sequential topologieal semigroup is a
topologieal group, Math. Stud. 2 (1993),23-24 (in Russian).
[Zam] Zambahidze, L.G., On relationships between dimensions of free bases of
free topologie al groups, Soobshch. AN Gruz. SSR 97 (1980), 569-572 (in
Russian).
[Zar] Zarelua, A.v., Equality of dimensions and eompaet extensions, Dokl. AN
SSSR 144 (1962), 713-716 (in Russian).
[Zell Zelazko, w., A theorem on Bo division algebras, Bult. Acad. Polon. Sei. 8
(1960), 373-375.
[Zele] Zelenyuk, E.G., Resolvability of topological groups, Ukr. Mat. Zh. 51
(1999), No. 1 (in Russian).
[ZePr] Zelenyuk, E.G. and Protasov, LV., Nonresolvable and extremally diseon-
neeted topologies on groups, Dokl. NAN Ukr. 1997,7-11 (in Russian).
HISTORY OF SHAPE THEORY AND ITS
APPLICATION TO GENERAL TOPOLOGY
Abstract
This paper consists of a short his tory of shape theory and of a survey of
the various applications of shape theory. In particular, the paper describes
applications in several areas of general topology (topology of continua, hy-
perspaces, compactifications, fixed points) and in dynarnical systems (Conley
index, attractors).
funetor S : H(Top) -+ Sh(Top). This functor keeps objects fixed, i.e., SeX) =
X, and on HPol, it is an isomorphism. Consequently, for spaces having the
homotopy type of polyhedra, shape theory reduces to standard homotopy theory.
It is generally considered that shape theory is the correct substitute for homotopy
theory when one works with spaces beyond the dass HPol.
The above mentioned approximation process involves the replacing oftopo-
10gical spaces X by suitable inverse systems of spaces having the homotopy type
ofpolyhedra X = (X).., Pw, A). One views the terms X).. as better and better
approximations of X as A progresses within the directed index set A. If one
adopts this point of view, one can trace the origins of shape theory to the early
papers on inverse systems by P.S. Aleksandrov [4], [5], [6], A. Kurosh [171],
S. Lefschetz [177] and H. Freudenthal [99]. Ceeh homology and cohomology
groups are shape invariants and undoubtedly belong to shape theory. They were
developed by Aleksandrov [5], L. Vietoris [274], E. Cech [48], A. Kolmogoroff
[156], 1.W. Alexander [2], C.H. Dowker [77], [78], K. Morita [220] and others.
The shape-theoretic analogue of the fundamental group was defined already
in 1937 by A. Komatu [157]. The corresponding higher dimensional groups,
called shape groups, were introduced by D.E. Christie in 1944 in bis paper
[63], which contains the beginnings of a theory of ordinary and strong shape.
In 1959 E.L. Lima defined the stahle shape eategory [180]. Cell-like spaees
and cell-like mappings (as weIl as UVn-spaces and UVn-mappings) are further
examples of early shape-theoretic notions (see [172]). They were preceded by
the notion of cellular set, introduced in 1960 by M. Brown in [46].
Modem shape theory begins in 1968, when K. Borsuk published his paper:
Coneeming homotopy properties of eompaeta [29]. In that paper he defined the
shape category Sh(CM), for the dass CM of metric compacta, and the shape
functor S : H(CM) -+ Sh(CM). He used the name shape for the first time in
[30]. Borsuk's technique of fundamental sequenees was based on embedding
metric compacta in the Hilbert cube Q and approximating them by open neigh-
borhoods, which are ANR's. This technique was improved and generalized to
arbitrary metric spaces by R.H. Fox in 1972 [95].
In 1970 S. Mardesic and 1. Segal noticed that inverse systems represent
the right tool for approximating spaces in shape theory. Their approach was
applicable to compact Hausdorff spaces [200], [201]. In 1973, several authors
used natural transformations to extend the shape category to arbitrary topo-
10gical spaces [188], [179], [278]. It was K. Morita who in 1975 defined the
same shape category using the inverse system technique. He associated with
every space X suitable inverse systems X = (X).., [p)..)../], A) in the homotopy
category H(HPol) and suitable morphisms of systems [p] = ([p)..]) : X -+ X
referred to as homotopy expansions [219]. The following Morita conditions
must be satisfied. (MI) Every homotopy dass [f] : X -+ P to a polyhedron
factors through X. (M2) This factorization is unique in the following sense: If
mSTORY OF SHAPE THEORY 1147
[fAJ, [fD : XI.. ~ P are homotopy c1asses such that [JA][PA] = [fD[PA]' then
[fA][PAA'] = [J{][PAA']' for some )..' 2: )... The Cech system gives an example
of a polyhedral homotopy expansion [220]. Shape morphisms F : X ~ Y are
given bytriples ([p], [q], [f]), where [p] : X ~ X and [q] : Y ~ Y are HPol-
homotopy expansions (the terms of X and Y belong to HPol) and [I] : X ~ Y
is a morphism of the pro-category pro-H(Top). For an arbitrary category 'fb, the
category pro-'fb was defined by A. Grothendieck in 1959 in his work on algebraic
geometry [106].lts objects are inverse systems in 'fb. If X = (XI.., Pu', A) and
Y = (YJ.L' PJ.LJ.L' , M) are two objects ofpro-'fb, one considers increasing functions
J : M ~ A and collections of morphisms JJ.L : Xf(J.L) ~ YJ.L such that, for
/Lo ::: /LI, the following diagram commutes.
P f(J.LO)f(J.L 1)
I I
X f(J.LO) X f(J.L
JJ.LO JJ.Ll
(1)
is still open [87]. Using 3-manifolds, A. Kadlof has determined the structure of
pointed I-movable continua embedded in R 3 . In particular, he proved that every
FANR in R 3 has the shape of a compact polyhedron [127].
For movable spaces various shape-theoretic results assume a simpler form.
E.g., if J : (X, *) ~ (Y, *) is a pointed shape morphism between pointed
movable metric continua, which induces isomorphisms of shape groups h :
7rk(X, *) ~ 7rk(Y, *), for all k, and if the spaces X, Y are finite-dimensional,
then J is a pointed shape equivalence. This is a consequence of the shape-
theoretic Whitehead theorem and the fact that such an J induces isomorphisms
ofhomotopy pro-groups !I.k(X, *) ~ !I.k(y, *) [140], [83].
Borsuk also introduced the notion of n-movability. A metric compactum
X ~ Q is n-movable provided every neighborhood U of X in Q admits
a neighborhood U' of X so small that every mapping J : K ~ U' of a
metric compactum K of dimension dim K ~ n can be deformed within U
arbitrarily elose to X. Every LC n - 1 compactum is n-movable [34]. The notion
of movability and n-movability was quickly generalized to compact Hausdorff
spaces [200], [160] and to arbitrary spaces [161]. In particular, G. Kozlowski
and 1. Segal proved that every paracompact LCn - 1 space is n-movable [161].
In aseries ofpapers J.E. Keesling studied the structure ofthe Cech cohomology
groups iIk(X) (integer coefficients) of movable and n-movable compacta and
spaces. In particular, in [139] he proved that for a movable compact Hausdorff
space X the group iIk(X)/Tor (iIk(X)) is an ~l-free abelian group. Keesling
used his algebraic criteria for n-movability to show the existence of metric
LCn - 1 continua X, which do not have the shape of any n-movable para-
compact space [141]. The notion of n-movability was the beginning of the
n-shape theory, which was developed in aseries of papers by A. Chigogidze
(see [60]). The n-shape theory proved to be especially useful in the theory
of n-dimensional Menger manifolds, a theory initiated by M. Bestvina [20]
(see [62]).
In 1972 T.A. Chapman applied infinite-dimensional topology to the shape
ofmetric compacta [57], [59]. In particular, he proved that two compacta X, Y,
embedded in the Hilbert cube Q as Z-sets, have the same shape if and only
if their complements Q\X, Q\Y are homeomorphic (for a different proof see
[257]). Chapman also exhibited an isomorphism of categories T. The domain of
T is the weak proper homotopy category W9P, whose object are complements
M = Q\X of Z-sets X of Q. The morphisms ofW9P are equivalence elasses
of proper mappings J : M ~ N = Q\Y, where two such mappings J, g :
M ~ N are considered equivalent provided every compact set B ~ N admits
a compact set A ~ M and a homotopy H : M x I ~ N such that H connects
J to g and HM\A) x /)) ~ N\B. The codomain of T is the restriction of
the shape category Sh(CM) to Z-sets X of Q. On objects M = Q\X ofW9P,
T(M) = X.
HISTORY OF SHAPE THEORY 1151
Junctor E : SSh(Top) -+ Sh(Top) such that S = ES, Le., the shape functor S
factors through the strong shape category.
The strong shape category was first introduced by J.B. Quigley in 1973
[241] and in a somewhat more general form in 1979 by Y. Kodama and 1. Ono
[154]. The objects of his category are c10sed subsets of the Hilbert cube Q.
The morphisms X -+ Y are homotopy c1asses of approaching mappings, where
the latter are mappings </J : Q x [0, (0) -+ Q such that, for sufficiently large
r ::: 0 and sufficiently small neighborhood U of X in Q, </J(U x [r, (0 is
arbitrarily c10se to Y. In 1976 D.E. Edwards and H.M. Hastings independently
discovered the category SSh(CM) [91] (also see [92]). They first endowed the
category pro-Top with the structure of a c10sed model category in the sense of
D.G. Quillen [242]. Localizing pro-Top at weak homotopy equivalences, they
obtained a homotopy category, which they denoted by Ho(pro-Top). Using this
category instead ofpro-H(Top), they obtained the category SSh(CM). Edwards
and Hastings showed that the analogue of Chapman's category isomorphism
theorem in the case of strong shape assumes a much simpler form, because the
role of weak proper homotopy category is taken up by the proper homotopy
category of the complements of Z-sets of Q. It was later proved by T. Porter
that Ho(pro-Top) can also be obtained by localizing pro-Top at level homotopy
equivalences [238].
F. W. Bauer was the first to define a strong shape category for arbitrary
topological spaces [14]. In his approach he used 2-categories. Bauer's theory
was further developed by his student B. Gnther, who used the technique of
simplicial sets [107]. Z. Miminoshvili [211], [212] and Yu.T. Lisitsa [184],
[185] made the first steps towards defining the strong shape category using
the inverse system approach. What was missing in these papers was the right
method of approximating spaces by polyhedra. An adequate tool was developed
independently by K. Morita, for topologically complete spaces [219], and by
S. Mardesic, for arbitrary topological spaces [190]. This tool is the resolution
of spaces (also see [202]). Resolutions p : X -+ X can be viewed as special
inverse limits of inverse systems in Top, satisfying conditions (Rl), (R2), which
strengthen Morita's conditions (MI), (M2). In particular, (Rl) requires that every
mapping ! : X -+ P to a polyhedron can be factored through some member
XI.. of X up to a given covering 'V of P, i.e., the mappings !APA and! are 'V-
near mappings. Condition (R2) refers to uniqueness of this factorization. It was
proved in [191] that every space admits an ANR-resolution. In the definition of
the shape category F.W. Cathey and 1. Segal have replaced Morita's expansions
by ANR-resolutions and have replaced the category pro-H(Top) by the Edwards-
Hastings category Ho(pro-Top). In this way they have obtained the strong shape
category SSh(Top) [47].
In aseries of papers published from 1983 (see [181], [182]), Lisitsa and
Mardesic have defined a coherent homotopy category CH(pro-Top), whose
HISTORY OF SHAPE THEORY 1153
objects are cofinite inverse systems (every element of the index set has a fi-
nite number of predecessors). Morphisms are homotopy c1asses of coherent
mappings f : X ~ Y. The latter consist of an increasing function I and
of mappings I J.LO : X! (J.LO) ~ Y J.Lo such that, for J1,0 ~ J1, t, the diagram (1)
is commutative up to homotopy. Homotopies f J.LOJ.Ll' which realize this com-
mutativity must be chosen and make up part of the structure of f. Similarly,
for increasing sequences J1,0 ~ .. ~ J1,n, one has homotopies of higher order
IJ.Lo.J.Ln : X!(J.Ln) X .n ~ YJ.LO' which also belong to the structure of f. The
strong shape category was defined using ANR-resolutions and CH(pro-Top).
It was proved in 1991 by Gnther [108] that the categories Ho(pro-Top) and
CH(pro-Top) are isomorphie, because each ofthem is isomorphie to a category
he defined using simplicial sets [109]. A direct proof that the two categories are
isomorphie is given in [195]. A consequence of these results is that the strong
shape categories of [47] and of [181] are also isomorphie. More general versions
of coherent homotopy categories are present in the literature since the work of
J.M. Boardman and R.M. Vogt [24], [275] and J.M. Cordier and T. Porter [69],
[12], [13].
Even though resolutions can be used to develop strong shape theory, it soon
became c1ear that this is too fine a tool for the task and that the right notion
to develop strong shape is that of a strong expansion. A mapping of systems
p : X ~ X is a strong expansion if it has Morita's property (MI) and the
following property (S2), which strengthens property (M2). For every polyhedron
P, mappings 10, ft : XA ~ P and homotopy F : X x I ~ P, whieh connects
10PA and ItPA' there exist a 'A' ~ 'A and a homotopy H : X A, x I ~ P, which
connects 10PAA' and ItPw, and thehomotopies H(PA' xl) and F are connected
by a homotopy (X x 1) x I ~ P, which is fixed on X x aI. Strong expansions
were first introduced in Gnther's 1989 Ph.D. thesis [107]. Independently, J.
Dydak and S. Nowak introduced the more general notion of a strong shape
equivalence p : X ~ Y between inverse systems [84]. It was proved in [193]
that resolutions are strong expansions. Moreover, strong expansions of spaces
from the c1ass HPol coincide with coherent expansions, abasie notion in the
Lisitsa-Mardesic strong shape theory [192], [194], [196].
Appropriate variations of the basic ideas of shape led to new types of shape
theories. In particular, there is fibered shape [133], [284], equivariant shape
[7], [54], stable shape [233], [19], proper shape [11], [9], [53], uniform shape
[255], [215], extension shape [61], [125].
in the work of L.G. Brown, R.G. Douglas and P.A. Fillmore [45]. Then D.S.
Kahn, J. Kaminker and C. Schochet constructed exact homologies, which on
finite CW-complexes coincide with arbitrary extraordinary homologies [130],
[129], [131]. In aseries of papers F.W. Bauer considered extraordinary strong
homology for compact metric and arbitrary spaces and proved their strong shape
invariance [14], [15], [16], [17], [18]. To the same topic are devoted the papers
[12] by A.V Batanin and [240] by M.A. Prasolov. The relation between shape
theory and proper homotopy is described in [239].
Shape theory also has applications to pattern recognition. Contacts between
these areas were established in the books by M. Pavel [235] and by Cordier and
Porter [71], as weIl as in some artic1es, e.g., [234], [70] and [104].
There are four main areas of general topology, where shape theory has been
successfully applied. These are continua, compactifications, hyperspaces and
fixed points. We will now describe some of the results obtained in these areas.
Continua. Most applications of shape theory to continua theory refer to the
notion of movability and its variants. In [163] J. Krasinkiewicz proved that a
circ1e-like continuum X is movable if and only if its Cech cohomology group
iI1(X) equals 0 or Z, or equivalently, iI1(X) is finitely divisible (no element
different from 0 is divisible by arbitrarily large natural numbers). Hence, by a
theorem ofM.C. McCord [206], a circ1e-like continuum X embeds in the plane if
and only if it is movable. In the same paper Krasinkiewicz also proved that every
nonmovable circ1e-like continuum is indecomposable. Finally, he answered a
question ofG.W. Henderson [118] by characterizing continua X, which cannot
be mapped onto any non-planar circ1e-like continuum, as continua whose Cech
cohomology group iI 1(X) is finitely divisible.
In [169] J. Krasinkiewicz and M. Smith showed that a hereditarily in-
decomposable continuum X, which is the limit of an inverse sequence of
locally connected unicoherent continua, must be tree-like and thus, of dimension
dim X ~ 1. This implies that hereditarily indecomposable continua of trivial
shape must be tree-like.
In [208] D.R. McMillan, Jr. obtained the result that pointed movable con-
tinua coincide with continua which are movable and pointed I-movable. He
proved that a continuous image of a pointed l-movable continuum is again such
a continuum and every subcontinuum of a 2-manifold is pointed movable. These
results were obtained independently by Krasinkiewicz in his paper [165], which
contains a detailed study of (pointed) movability and (pointed) l-movability
of continua. In particular, Krasinkiewicz characterizes pointed I-movable con-
tinua as continua whose fundamental pro-group has the Mittag-Leffler property.
Since for I-dimensional continua movability and pointed l-movability coincide
[270], it follows that al-dimensional continuous image of al-dimensional
movable continuum is itself a movable continuum. Krasinkiewicz also showed
that a continuum which is not pointed I-movable contains an indecomposable
1156 SIBE MARDESIC AND JACK SEGAL
It appears that [244] was the first paper, which explicitly linked differential
equations to shape theory. In that paper J.T. Rogers, Jr. used various facts from
shape theory to analyze which continua can appear as cross-sections of the
solution funnel of an ordinary differential equation. E.g., he showed that a
necessary condition for a continuum to be such a cross-section is that it be
pointed l-movable.
By a dynamical system we mean a continuous system (flow), i.e., an action
x 1---+ Xl, X E X, t E ~,ofthe group ~ ofreals or adiscrete system (cascade), i.e.,
an action x 1---+ x n , x EX, n E Z, of the group Z of integers. The phase space X
is assumed to be locally compact, usually a manifold. A semi-dynamical system,
is an action ofthe semigroups ~+ = {t E ~ : t 2: O} and Z+ = {t E Z : t 2: O},
respectively. In the discrete case this amounts to iterates fn of a homeomorphism
(of a self-mapping) f : X --+ X, for n E Z (for n E Z+). Here we discuss
more c10sely two areas of application: the Conley index and the attractors of a
dynamical system.
Originally, the Conley index, introduced by e.e. Conley under the name
Morse index, was defined for isolated invariant sets S of a fiow [66]. It provides
useful information conceming the dynamical structure of S. To define the Conley
index one considers index pairs for S, Le., compact pairs (N, L) in X such that
N is an isolating neighborhood of S and L is the exit set, Le., every orbit which
exits N, goes through L first. Moreover, L is positively invariant relatively
to N. Index pairs exist and, for all of them, the quotient space N / L has the
same homotopy type, which is that of a compact polyhedron. By definition, that
homotopy type is the Conley index h(S).
In [243] J.W. Robbin and D. Salamon generalized the Conley index by
introducing the shape index s(S) of an isolated invariant set S. This notion
is weH defined for both, differentiable fiows and diffeomorphisms on smooth
manifolds. Robbin and Salamon first defined an index pair for S as a compact
pair (N, L) in X such that N\L is an isolating neighborhood for S and the
induced semi-dynamical system on N / L is continuous. The action on N / L
gives rise to an inverse system, whose limit W#(S) is a pointed compactum,
which does not depend on the choice of the pair (N, L) and is a one-point
compactification of the unstable manifold W U (S) of S (the set of all points
of X, whose orbit tends to S in backwards time), endowed with a particular
topology, called intrinsic. The shape index s(S) is the shape type of W#(S). In
the case of fiows, the shape index coincides with the Conley index. If S is an
attractor, s(S) is just the shape of S [113]. M. Mrozek [227] has generaHzed the
work of Robbin and Salamon by introducing a shape index for locally defined
mappings of locally compact Hausdorff spaces. His theory unifies several earlier
index theories, inc1uding his own cohomological Conley index [225]. Mrozek's
theory essentially uses basic shape theory of compact spaces. In his paper [227]
Mrozek raised the question whether the shape index can be constructed for
mSTORY OF SHAPE THEORY 1163
mappings on spaces which are not locally compact, but satisfy the Rybakowski
conditions. The difficulty lies in the fact that in noncompact situations the inverse
limit functor and the shape functor do not necessarily commute. In [247] ER.
Ruiz deI Portal and J.M. Salazar have overcome this difficulty by successfully
using resolutions (in the sense of Mardesic). Arecent survey of the Conley
index theory is contained in [214].
In [252] lM.R. Sanjurjo applied the shape-theoretic Lusternik-Schnirel-
mann category TJ (in the sense of [40]) to the study of the Conley index and
obtained simpleInequalities which involve the categories of isolated invariant
sets, of isolating neighborhoods, of the exit sets and of the unstable manifolds
of invariant sets and Morse sets. Por example, if {MI, ... , Mn} is a Morse
decomposition of a regular isolated invariant set K and W U is the unstable
manifold of K, then TJ(W U ) ~ TJ(Mt> + ... + ,,(Mn). The obtained inequalities
yield useful information conceming the dynaiiiics of K. In particular, one can
obtain lower bounds for the number of Morse sets in a given decomposition,
one can detect the presence of connecting orbits in an attractor-repeller pair and
prove the existence of critical points. One can also obtain criteria for K to be a
saddle set or criteria for the non-triviality of shape of an attractor (repeller) in
an attractor-repeller decomposition.
Attractors of a dynamical system are defined as compact subsets A ~ X,
which are invariant and positively asymptotically stahle. The first papers on
shape and attractors are H.M. Hastings' papers [114], [115], inspired by the
celebrated Poincare-Bendixson theorem. In the classical version of this theorem,
one considers an annulus M embedded in the plane R 2 and a Cl fiow, which has
the property that orbits through the boundary of M enter Int(M) as t increases
and the fiow has no rest points. Then there exists a limit cycle K in Int(M).
Hastings' theorem refers to a compact n-dimensional submanifold M ~ Rn and
a continuous semi-dynamical system such that orbits through the boundary of
M enter Int(M) as t increases. The theorem asserts that there exists an attractor
K ~ Int(M) such that the inclusion K ~ M is a shape equivalence.
Using different methods, several authors proved that every attractor of a fiow
has the shape of a compact polyhedron (see [26], for differentiable fiows, and
[113] and [250], for continuous fiows). Using more sophisticated shape theory,
in particular, the finite-dimensional Chapman complement theorem [58], B.
Gnther and J. Segal have also proved the converse implication. More precisely,
they showed that every finite-dimensional compactum A, which has the shape
of a compact polyhedron, can be embedded in Rn in such a way that it is the
attractor of some fiow on Rn [113]. Subsequently, Gnther proved that the same
result also holds for differentiable fiows of class er, 1 ~ r < 00 [111].
The above characterization of compacta which appear as attractors does not
hold for discrete systems. B.g., the dyadic solenoid is the attractor of a discrete
system in spite of the fact that it does not have the shape of a compact polyhedron
1164 SIBE MARDESIC AND JACK SEGAL
(it falls to be movable). On the other hand, Gnther exhibited a class of solenoids,
which cannot be embedded as attractors of a discrete system and cannot even be
attractors of a self-mapping [110]. Recently, H. Kato gave a characterization of
compacta, which are attractors of a discrete system on a manifold [138]. These
are compacta K, which admit a homeomorphism I : K ~ K and admit a
mapping g : P ~ P of a compact polyhedron P such that I : K ~ K is
topologically conjugate to the shift map on the compactum
Acknowledgements
References
[1] Adams, J.F., On the groups J(X) - IV, Topology 5 (1966), 21-71.
[2] Alexander, I.W., A theory of connectivity in terms of gratings, Ann. Math. 39
(1938),883-912.
[3] Alexandroff, P.S., ber die Struktur der bikompakten topologischen Rumen,
Math. Ann. 92 (1924), 267-274.
[4] _ _ , Simpliziale Approximationen in der allgemeinen Topologie, Math. Ann.
96 (1926), 489-511.
[5] _ _ , Une definition des nombres de Betti pour un ensemble ferme quel-
conque, C. R. Acad. Sei. Paris 184 (1927),317-320.
[6] _ _ , Untersuchungen ber Gestalt und Lage abgeschlossener Mengen
beliebiger Dimension, Ann. Math. 30 (1929), 101-187.
[7] Antonian, S.A. and Mardesic, S., Equivariant shape, Fund. Math. 127 (1987),
213-224.
[8] Artin, M. and Mazur, B., Etale Homotopy, Lecture Notes in Math. vol. 100,
Springer, Berlin 1969.
[9] Ball, B.I., Alternative approaches to proper shape theory, in Studies in Topology
(Charlotte conference, 1974), Academic Press, New York 1975, pp. 1-27.
[10] _ _ , Quasicompactifications and shape theory, Pacific J Math. 84 (1979),
251-259.
[11] Ball, B.I. and Sher, R.B., A theory of proper shape for locally compact metric
spaces, Fund. Math. 86 (1974), 163-192.
[12] Batanin, M.A., On the coherent prohomotopy category 0/ Lisica-Mardesic and
generalized steenrod homology theories (in Russian), USSR Academy of Sei.,
Siberian Section, Novosibirsk 1986, preprint, pp. 1-23.
[13] _ _ , Coherent categories with respect to monads and coherent prohomotopy
theory, Cahiers Topoi. Geom. Diff. Categor. (4) 34 (1993), 279-304.
[14] Bauer, F.w., A shape theory with singular homology, Pacific J. Math. 64 (1976),
25-65.
[15] _ _ , Duality in manifolds, Ann. Mat. Pura Appl. 136 (1984), 241-302.
[16] _ _ , Extensions of generalized homology theories, Pacific J Math. 128
(1987), 25-61.
1166 SillE MARDESIC AND JACK SEGAL
[17] _ _ , Generalized homology theories and chain complexes, Ann. Mat. Pura
Appl. 155 (1989), 143-191.
[18] _ _ , The existence of strong homology theory, Comm. Algebra (9) 22, 3419-
3432.
[19] _ _ , A strong shape theory admitting an S-dual, Topology Appl. 62 (1995),
207-232.
[20] Bestvina, M., Characterizing k-dimensional universal Menger compacta, Mem.
Amer. Math. Soc. (380) 71 (1988), pp. 1-110.
[21] _ _ , Local homology properties of boundaries of groups, Michigan Math. 1.
(1) 43 (1996), 123-139.
[22] Bestvina, M. and Mess, G., The boundary of negatively curved groups, 1. Amer.
Math. Soc. 4 (1991), 469~81.
[23] Blackadar, B., Shape theory for C*-algebras, Math. Scand. 56 (1985), 249-275.
[24] Boardman, J.M. and Vogt, R.M., Homotopy Invariant Aigebraic Structures on
Topological Spaces, Lecture Notes in Math. vol. 347, Springer, Berlin 1973.
[25] Bogatyl, S.A., Approximate and fundamental retracts (in Russian), Mat. Sbomik
(135) 93 (1974), 90-102 (Math. USSR Sbomik 22 (1974), 91-103).
[26] Bogatyi, S.A. and Gutsu, V.I., On the Structure of attracting compacta (in
Russian), Differentsial'nye Uravneniya 25 (1989), 907-909.
[27] Borsuk, K., Sur un continu acyclique qui se laisse transformer topologiquement
en lui meme sans points invariants, Fund. Math. 24 (1935),51-58.
[28] _ _ , On some metrization of the hyperspace of compact sets, Fund. Math.
41, (1954), 168-202.
[29] _ _ , Concerning homotopy properties of compacta, Fund. Math. 62 (1968),
223-254.
[30] _ _ , Conceming the Notion 0/ the Shape 0/ Compacta, in Proc. Internat.
Symp. Topology and its App. (Herceg-Novi 1968), D. Kurepa, (ed.), Savez
Drustava Mat. Fiz. Astronom., Beograd 1969, pp. 98-104.
[31] _ _ , Fundamental retracts and extensions of fundamental sequences, Fund.
Math. 64 (1969), 55-85.
[32] _ _ , On movable compacta, Fund. Math. 66 (1969),137-146.
[33] _ _ , A note on the theory of shape of compacta, Fund. Math. 67 (1970),
265-278.
[34] _ _ , On the n-movability, Bull. Acad. Polon. Sei. Ser. Sei. Math. Astronom.
Phys. 20 (1972),859-864.
[35] _ _ , Theory 0/ Shape, Polish Scientific Publishers Warszawa 1975.
[36] _ _ , On nearly extendable maps, Bull. Acad. Polon. Sei. Sero Sei. Math.
Astronom. Phys. 23 (1975), 753-760.
[37] _ _ , On the Lefschetz-Hopf fixed point theorem for nearly extendable maps,
Bull. Acad. Polon. Sei. Ser. Sei. Math. Astronom. Phys. 23 (1975), 1273-1279.
[38] _ _ , On a metrization of the hyperspace of ametrie space, Fund. Math. 94,
(1977), 191-207.
[39] _ _ , Aremark on shape of continua, Bull. Acad. Polon. Sei. Ser. Sei. Math.
Astronom. Phys. 25 (1977), 1141-1147.
[40] _ _ , On the Lusternik-Schnirelmann category in the theory of shape, Fund.
Math. 99 (1978), 35~2.
mSTORY OF SHAPE THEORY 1167
[65] Conley, C. and Easton, R, Isolated invariant sets and isolating blocks, Trans.
Amer. Math. Soe. 158 (1971), 35-61.
[66] Conley, C.C., Isolated invariant sets and the Morse index, CBMS Regional
Conference Series in Math. 38, Amer. Math. Soc., Providence, R.I. 1978.
[67] Coram, D.S. and Duvall, P.F., Approximate fibrations, Roeky Mountain 1. Math.
7 (1977), 275-288.
[68] _ _ , Approximate fibrations and a movability condition for maps, Paeifie 1.
Math. 72 (1977), 41-56.
[69] Cordier, J.M. and Porter, T., Vogt's theorem on categories ofhomotopy coherent
diagrams, Math. Proe. Cambridge Phil. Soe. 100 (1986),65-90.
[70] _ _ , Pattern recognition and categorical shape theory, Pattern Reeognition
Letters 7 (1988), 73-76.
[71] _ _ , Shape Theory - Categorieal Methods of Approximation, EIlis Horwood,
Chichester 1989.
[72] Curtis, D.W. and Schori, RM., 2x and C(X) are homeomorphic to the Hilbert
cube, Bull. Amer. Math. Soe. 80 (1974),927-931.
[73] Cuchillo-Ibafiez, E., Mor6n, M.A., Ruiz deI Portal, F.R and Sanjurjo, lM.R,
A topology for the sets of shape morphisms, Topology and its Appl. 94 (1999),
51-60.
[74] Dadarlat, M., Shape theory and asymptotic morphisms for C*-algebras, Duke
Math.1. 73 (1993), 687-711.
[75] Dlidrlat, M. and Nemethi, A., Shape theory and connective K -theory, 1. Oper.
Theory 23 (1990), 207-291.
[76] Deleanu, A. and Hilton, P., On the categorical shape of a functor, Fund. Math.
97 (1977), 157-176.
[77] Dowker, C.H., Cech cohomology theory and the axioms, Ann. Math. 51 (1950),
278-292.
[78] _ _ , Homology groups of relations, Ann. Math. 56 (1952), 84-95.
[79] Dranishnikov, A., Keesling, J.E. and Uspenskij, V.V., On the Higson corona of
uniformly contractible spaces, Topology 37 (1998), 791-803.
[80] Draper, J. and Keesling, J.E., An example concerning the Whitehead theorem
in shape theory, Fund. Math. 92 (1976), 255-259.
[81] Dydak, J., Some remarks on the shape of decomposition spaces, Bull. Aead.
Polon. Sei. Sero Sei. Math. Astronom. Phys. 23 (1975), 561-563.
[82] _ _ , A simple proof that pointed FANR-spaces are regular fundamental re-
tracts of ANR's, Bull. Aead. Polon. Sei. Sir. Sei. Math. Astronom. Phys. 25
(1977), 55-62.
[83] _ _ , The Whitehead and the Smale theorems in shape theory, Dissertationes
Math. 156 (1979), 1-55.
[84] Dydak, J. and Nowak, S., Strong shape for topological spaces, Trans. Amer.
Math. Soe. 323 (1991), 765-796.
[85] Dydak, J., Nowak, S. and Strok, S., On the union of two FANR-sets, Bull. Aead.
Polon. Sci. Sb: Sei. Math. Astronom. Phys. 24 (1976), 485--489.
[86] Dydak, J. and Segal, J., Shape Theory. An Introduetion, Lecture Notes in Math.
vol. 688, Springer-Verlag, Berlin 1978.
HISTORY OF SHAPE THEORY 1169
[107] Gnther, B., Starker Shapefiir beliebige topologisehe Rume, Dissertation, J.w.
Goethe-Universitt, Frankfurt a.M. 1989.
[108] _ _ , Comparison of the coherent pro-homotopy theories of Edwards-Ha-
stings, Lisica-Mardesic and Gnther, Glasnik Mat. 26 (1991), 141-176.
[109] _ _ , The use of semisimplicial complexes in strong shape theory, Glasnik
Mat. (47) 27 (1992), 101-144.
[110] _ _ , A compactum that cannot be an attractor of a selfmap on a manifold,
Proe. Amer. Math. Soe. 120 (1994), 653--655.
[111] _ _ , Construction of differentiable fiows with prescribed attractor, Topology
Appl. 62 (1995), 87-91.
[112] _ _ , On the homotopic structure of dynamical systems containing global
attractors, in Continua with the Houston Problem Book, H. Cook et al. (ed.), M.
Dekker, Lecture Notes in Pure and Appl. Math. vol. 170, (1995), 227-248.
[113] Gnther, B. and Segal, J., Every attractor of a fiow on a manifold has the shape
of a finite polyhedron, Proe. Amer. Math. Soe. vol. 119, (1993), 321-329.
[114] Hastings, H.M., Shape Theory and Dynamieal Systems. The Strueture ofAttrae-
tors in Dynamieal Systems (Proc. Conf., North Dakota State Univ., Fargo, N.D.,
1977), pp. 150-159, Lecture Notes in Math., vol. 668, Springer, Berlin, 1978.
[115] _ _ , A higher dimensional Poincare-Bendixson theorem, Glasnik Mat. 14
(1979),263-268.
[116] Hastings, H.M. and Heller, A., Homotopy idempotents on finite-dimensional
complexes split, Proe. Amer. Math. Soe. 85 (1982), 619--622.
[117] Hausdorff, F., Grundzuge der Mengenlehre, von Weit, Leipzig, 1914.
[118] Henderson, G. W., Continua which cannot be mapped onto any non-planar circle-
like continuum, Colloq. Math. 23 (1971), 241-243.
[119] Hemandez-Paricio, L.J., Fundamental pro-groupoids and covering projections,
Fund. Math. 156 (1998), 1-31.
[120] Holsztytiski, w., Une generalization du theoreme de Brouwer sur les points
invariants, Bull. Aead. Polon. Sei. Ser. Sei. Math. Astronom Phys. 12 (1964),
603--606.
[121] _ _ , Universal mappings and fixed point theorems, BUll' Aead. Polon. Sei.
ser. Sei. Math. Astronom. Phys. 15 (1967), 433-438.
[122] _ _ , Aremark on the universal mappings of I-dimensional continua, Bull.
Aead. Polon. Sei. Ser. Sei. Math. Astronom. Phys. 15 (1967), 547-549.
[123] Hopf, H., New proof of the Lefschetz formula on invariant points, Proe. Nat.
Aead. Sei. 14 (1928), 149-153.
[124] Husch, L.S. and Ivansic, 1., Embedding compacta up to shape, in Shape Theory
and Geometrie Topology, Lecture Notes in Math. vol. 870, Springer, Berlin
1981, pp. 119-134.
[125] Ivansic, 1., Rubin, L.R. and Schapiro, P.J., Extension shape theory, Kyungpook
Math.1. 40 (2000),157-172.
[126] Ivansic, 1., Sher, R.B. and Venema, G.A., Complement theorems beyond the
trivial range, Illinois 1. Math. 25 (1981), 209-220.
[127] Kadlof, A., On the shape of pointed compact connected subsets of E3, Fund.
Math. 115 (1983), 163-193.
HISTORY OF SHAPE THEORY 1171
[128] Kahn, D.S., An example in Cech cohomology, Proc. Amer. Math. Soc. 16 (1965),
584.
[129] Kahn, D.S., Kaminker, J. and Schochet, c., Generalized homology theories on
compact metric spaces, Michigan Math. J. 24 (1977), 203-224.
[130] Kaminker, J. and Schochet, C., Steenrod homology and operator algebras, Bull.
Amer. Math. Soc. 81 (1975), 431-434.
[131] _ _ , K -theory and Steenrod homology: Applications to the rown-Douglas-
Fillmore theory of operator algebras, Trans. Amer. Math. Soc. 227 (1977), 63-
107.
[132] Kapitanski, L. and Rodnianski, I., Shape and Morse theory of attractors, Comm.
Pure Appl. Math. 53 (2000), 218-242.
[133] Kato, H., Fiber shape categories, Tsukuba J. Math. 5 (1981), 247-265.
[134] _ _ , Shape properties ofWhitney maps for hyperspaces, Trans. Amer. Math.
Soc. 297 (1986), 529-546.
[135] _ _ , On some sequential strong Whitney-reversible properties, Bull. Polish
Acad. Sei. Math. 37 (1989), 517-524.
[136] _ _ , Movability and strong Whitney-reversible properties, Topology Appl. 31
(1989), 125-132.
[137] _ _ , A note on fundamental dimensions of Whitney continua of graphs, J.
Math. Soc. Japan 41 (1989), 243-250.
[138] _ _ , Attractors in euclidean spaces and shift maps on polyhedra, Houston J.
0/ Math. 24 (1998), 671-679.
[139] Keesling, J.E., An algebraic property of the Cech cohomology groups which
prevents local connectivity and movability, Trans. Amer. Math. Soc. 190 (1974),
151-162.
[140] _ _ , On the Whitehead theorem in shape theory, Fund. Math. 92 (1976),
247-253.
[141] _ _ , The Cech cohomology ofmovable and n-movable spaces, Trans. Amer.
Math. Soc. 219 (1976), 149-167.
[142] _ _ , The Stone-Cech compactification and shape dimension, Topology
Proceedings 2 (1977), 483-508.
[143] _ _ , Decomposition of the Stone-Cech compactification which are shape
equivalences, Pacific J. Math. 75 (1978), 455-466.
[144] _ _ , The dimension of closed sets in the Stone-Cech compactification, Trans.
Amer. Math. Soc. 299 (1987),413-428.
[145] _ _ , The one-dimensional Cech cohomology ofthe Higson compactification
and its corona, Topology Proceedings 19 (1994), 129-148.
[146] _ _ , Subcontinua ofthe Higson corona, Topology Appl. 80 (1997),155-160.
[147] Keesling, J.E. and Sher, R.., Shape properties of the Stone-Cech compactifi-
cation, General Top. Appl. 9 (1978), 1-8.
[148] Keesling, J.E. and Wilson, D.C., Embedding Tn-like continua in Euclidean
space. Topology Appl. 21 (1985),241-249.
[149] Kelley, J.L., Hyperspaces of a continuum, Trans. Amer. Math. Soc. 52 (1942),
22-36.
[150] Klee, V.L., Stability of the fixed point property, Colloq. Math. 8 (1961), 43-46.
1172 SffiE MARDESIC AND JACK SEGAL
[151] Knaster, B., Kuratowski, K and Mazurkiewicz, S., Ein Beweis des Fixpunkt-
satzes fr n-dimensionale Simplexe, Fund. Math. 14 (1929), 132-137.
[152] Ko~ak, ~., On the fundamental theorem of overlays, Note di Matematica, Lecce
10 (1990), 355-362.
[153] Kodama, Y. and Koyama, A, Hurewicz isomorphism theorem for Steenrod
homology, Proc. Amer. Math. Soc. 74 (1979),363-367.
[154] Kodama, Y. and Ono, J., On fine shape theory, Fund. Math. 105 (1979), 29-39.
[155] Kodama, Y., Spiez, S. and Watanabe, T., On shape o/hyperspaces, Fund. Math.
100 (1978), 59-67.
[156] Kolmogoroff, A, Les groupes de Betti des espaces localement bicompacts, C.R.
Acad. Sei. Paris 202 (1936), 1144-1147.
[157] Komatu, A, Bemerkungen ber die Fundamentalgruppe eines Kompakturns,
Proc. Imp. Acad. Tokyo 13 (1937), 56-58.
[158] _ _ , A note on some strong Whitney-reversible properties, Tsukuba 1. Math.
4 (1980), 313-316.
[159] Kozlowski, G., Images 0/ ANR's, Mimeographed notes, Seattle 1974.
[160] Kozlowski, G. and Segal, J., n-movable compacta and ANR-systems, Fund.
Math. 85 (1974), 235-243.
[161] _ _ , Locally well-behaved paracompacta in shape theory, Fund. Math. 95
(1977),55-71.
[162] Krasinkiewicz, J., Certain properties ofhyperspaces, Bull. Acad. Polon. Sei. Sero
Sei. Math. Astronom. Phys. 21 (1973),705-710
[163] _ _ , Mappings onto eircle-like continua, Fund. Math. 91 (1976), 39-49.
[164] _ _ , Local connectedness and pointed I-movability, Bull. Acad. Polon. Sei.
Sero Sei. Math. Astronom. Phys. 25 (1977), 1265-1269.
[165] _ _ , Continuous images of continua and 1-movability, Fund. Matk. 98
(1978), 141-164.
[166] _ _ , Hereditarily Indecomposable Representatives 0/ Shapes, Proc. Internat.
Conf. on Geometrie Topology, Polish Sei. Publ., Warszawa 1980, pp. 245-252.
[167] Krasinkiewicz, J. and Nadler, S.B., Jr., Whitney properties, Fund. Math. 98
(1978), 165-180.
[168] Krasinkiewicz,1. and Mine, P., Generalized paths and pointed movability, Fund.
Math. 104 (1979), 141-153.
[169] Krasinkiewicz, J. and Smith, M., Hereditarily indecomposable continua with
trivial shape, Fund. Math. 119 (1983), 133-134.
[170] Kuperberg, K, An isomorphism theorem of Hurewicz type in Borsuk's theory
of shape, Fund. Math. 77 (1972), 21-32.
[171] Kurosch, AG., Kombinatorischer Aufbau der bikompakten topologischen
Rumen, Compositio Math. 2 (1935), 471-476.
[172] Lacher, R.C., Cell-like mappings and their generalizations, Bull. Amer. Math.
Soc. 83 (1977), 495-552.
[173] Laguna, V.P., Mor6n, M.A, Nguyen To Nhu and Sanjurjo, J.M.R., Movability
and limits of polyhedra, Fund. Math. 143 (1993), 191-201.
[174] Lefschetz, S., Interseetions and transformations of complexes and manifolds,
Trans. Amer. Math. Soc. 28 (1926), 1-49.
mSTORY OF SHAPE THEORY 1173
[175] _ _ , Manifolds with a boundary and their transformations, Trans. Amer. Math.
Soe. 29 (1927), 429-462.
[176] _ _ , Topology, Coll. Publ. vol. 12, Amer. Math. Soc., New York 1930.
[177] _ _ , On compact spaces, Ann. Math. 32 (1931), 521-538.
[178] _ _ , On the fixed point formula, Annals 0/ Math. 38 (1937),819-822.
[179] LeVan, J.H., Shape Theory, Thesis, University of Kentucky, Lexington, Ken-
tucky 1973.
[180] Lima, E.L., The Spanier-Whitehead duality in new homotopy categories, Summa
Brasiliensis Math. (3) 4 (1959), 91-148.
[181] Lisica, Ju.T. and Mardesic, S., Steenrod-Sitnikov homology for arbitrary spaces,
Bull. Amer. Math. Soe. 9 (1983), 207-210.
[182] _ _ , Coherent prohomotopy and strong shape theory, Glasnik Mat. (39) 19
(1984),335-399.
[183] _ _ , Strong homology of inverse systems of spaces, 11, Topology Appl. 19
(1985), 45-64.
[184] Lisitsa, Yu.T., Duality Theorems and Dual Shape and Coshape Categories
(Russian), Doklady Akad. Nauk SSSR no. 3, vol. 263, 532-536 (Soviet Math.
Dokl. no.2, vol. 25 (1982), 373-378).
[185] _ _ , Strong shape theory and the Steenrod-Sitnikov homology (Russian),
Sibirski Mat. Z. 24 (1983),81-99.
[186] _ _ , The theorems ofHurewicz and Whitehead in strong shape theory, Dokl.
Akad. Nauk SSSR vol. 283, no. I, (1985), 31-35.
[187] MacDonald, J.L., Categorical shape theory and the back and forth property, J.
Pure Appl. Algebra 12 (1978), 79-92.
[188] Mardesic, S., Shapes of topological spaces, General Topology Appl. 3 (1973),
265-282.
[189] _ _ , On the Whitehead theorem in shape theory I, Fund. Math. 91 (1976),
51-64.
[190] _ _ , Approximate polyhedra, resolutions ofmaps and shape librations, Fund.
Math. 114 (1981), 53-78.
[191] _ _ , Inverse limits and resolutions, in Shape Theory and Geometrie Topology,
Proc. Conference, Dubrovnik 1981, Lecture Notes in Math. vol. 870, Springer
Verlag, Berlin, pp. 240-253.
[192] _ _ , Strong expansions and strong shape theory, Topology Appl. 38 (1991),
275-291.
[193] _ _ , Resolutions of spaces are strong expansions, Publ. Inst. Math. Beograd
(63) 49 (1991),179-188.
[194] _ _ , Coherent and strong expansions of spaces coincide, Fund. Math. 158
(1998),69-80.
[195] _ _ , Coherent homotopy and localization, Topology Appl. 94 (1999), 253-
274.
[196] _ _ , Strong shape and homology, Springer monographs in mathematics,
Berlin-Heidelberg-New York 2000, XII.
[197] Mardesic, S. and Matijevic, V., Classifying overlay structures of topological
spaces, Topology Appl. (to appear).
1174 SIBE MARDESIC AND JACK SEGAL
[198] Mardesic, S. and Rubin, L.R., Approximate inverse systems of compacta and
covering dimension, Pacijic J. Math. 138 (1989), 129-144.
[199] Mardesic, S. and Rushing, T.B., Shape fibrations I, Topology Appl. 9 (1978),
193-215.
[200] Mardesic, S. and Segal, J., Movable compacta and ANR-systems, Bull. Acad.
Polon. Sei. Sero Sei. Math. Astronom. Phys. 18 (1970), 649-654.
[201] _ _ , Shapes of compacta and ANR-systems, Fund. Math. 72 (1971), 41-59.
[202] _ _ , Shape Theory. The Inverse System Approach, North-Holland, Amster-
dam 1982.
[203] Mardesic, S. and Watanabe, T., Approximate resolutions of spaces and
mappings, Glasnik Mat. 24 (1989), 587-637.
[204] Mazurkiewicz, S., Sur l'hyperespace d'un continu, Fund. Math. 18 (1932), 171-
177.
[205] _ _ , Sur l'existence des continus indecomposables, Fund. Math. 25 (1935),
327-328.
[206] McCord, M.C., Embedding (l}I-like compacta in manifolds, Canad. 1. Math. 19
(1967), 321-332.
[207] McMillan, D.R., Jr., A criterion for cellularity in a manifold, Ann. Math. 79
(1964), 327-337.
[208] _ _ , One-dimensional shape properties and three-manifolds, in Studies in
Topology, N.M. Stavrakas and KR. Allen (eds.), Academic Press, New York
1975.
[209] Mihalik, M.L., Ends of fundamental groups in shape and proper homotopy,
Pacijic 1. Math. 90 (1980), 431~58.
[210] _ _ , On the concept of attractor, Commun. Math. Phys. 99 (1985),177-195.
[211] Miminoshvili, Z., On the strong homotopy in the category oftopological spaces
and its applications to the theory of shape (in Russian), Soobsc. Akad. Nauk
Gruzin. SSR 98 (1980), 301-304.
[212] _ _ , On a strong spectral shape theory (in Russian), Trudy Tbilissk. Mat. Inst.
Akad Nauk Gruzin. SSR 68 (1982), 79-102.
[213] Mioduszewski, J. and Rochowski, M., Remarks on fixed point theorem for
inverse limit spaces, Colloq. Math. 9 (1962), 67-71.
[214] Mischaikow, K, Conley Index Theory. Dynamical Systems (Montecatini Terme,
1994), Lecture Notes in Math., 1609, Springer, Berlin, 1995, 119-207,
[215] Miyata, T., Uniform shape theory, Glasnik Mat. (49) 29 (1994), 123-168.
[216] Miyata, T. and Segal, J., Shape and uniform properties of hyperspaces of
noncompact spaces, Glasnik Mat. 32 (1997), 99-124.
[217] Montgomery, J. T., Cohomology of isolated invariant sets under perturbation, 1.
Differential Equations 13 (1973), 257-299.
[218] Morita, K, The Hurewicz and the Whitehead theorems in shape theory, Sei.
Reports Tokyo Kyoiku Daigaku, Sec. A 12 (1974), 246-258.
[219] _ _ , On shapes oftopological spaces, Fund. Math. 86 (1975),251-259.
[220] _ _ , Cech cohomology and covering dimension for topological spaces, Fund.
Math. 87 (1975),31-52.
[221] Mor6n, M.A. and Ruiz deI Portal, ER., Ultrametrics and infinite dimensional
Whitehead theorems in shape theory, Manuscripta Math. 89 (1996), 325-333.
HISTORY OF SHAPE THEORY 1175
[247] Ruiz deI Portal, F.R. and Salazar, J.M., Shape index for discrete dynamical
systems on metric spaces (to appear).
[248] Sanjurjo, J.M.R., On fundamental approximative absolute neighborhood re-
tracts, Canadian Math. Bull. 27 (1984), 134-142.
[249] _ _ , Stability of the fixed point property and universal maps, Proc. Amer.
Math. Soc. 105 (1989), 221-230.
[250] _ _ , Multihomotopy, Cech spaces of loops and shape groups, Proc. London.
Math. Soc. 69 (1994), 330-344.
[251] _ _ , On the structure of uniform attractors, J. Math. Analysis and Appl. 192
(1995),519-528.
[252] _ _ , Lustemik-Schnirelmann category and Morse decompositions Mathe-
matika (to appear).
[253] Schori, R.M. and West, J.E., The hyperspace of the closed unit interval is a
Hilbert cube, Trans. Amer. Math. Soc. 213 (1975), 217-235.
[254] Segal, J. Hyperspaces of the inverse limit space, Proc. Amer. Math. Soc. 10
(1959), 706-709.
[255] Segal, J., Spiez, S. and Gnther, B., Strong shape of uniform spaces, Topology
and Appl. 49 (1993), 237-249.
[256] Segal, J. and Watanabe, T., Cosmic approximate limits and fixed points, Trans.
Amer. Math. Soc. 333 (1992), 1-61.
[257] Siebenmann, L.C., Chapman's classification of shapes - a proofusing collapsing,
Manuscripta Math. 16 (1975), 373-384.
[258] Sklyarenko, E.G., Homology and Cohomology of General Spaces (Rus-
sian), Itogi nauki i tehniki, Series - Contemporary problems of Mathematics,
Fundamental directions, vol. 50, (1989), Acad. Nauk SSSR, Moscow, pp.
125-228.
[259] _ _ , Hyper (co) homology left exact covariant functors and homology theory
of topological spaces (Russian), Uspehi Mat. Nauk (3) 50 (1995), 109-146.
[260] Smale, S., A Vietoris mapping theorem for homotopy, Proc. Amer. Math. Soc.
8 (1957), 604-610.
[261] Spiez, S., An example of a continuum X with Fd(X x SI) = Fd(X) = 2, Bull.
Acad. Polon. Sei. Sero Sei. Math. 27 (1979), 923-927.
[262] _ _ , On the fundamental dimension of the cartesian product of compacta
with fundamental dimension 2, Fund. Math. 116 (1983), 17-32.
[263] Steenrod, N.E., Regular cycles of compact metric spaces, Ann. Math. Soc. 41
(1940), 833-851.
[264] Stone, M.H., Applications of the general theory of Boolean rings to general
topology, Trans. Amer. Math. Soc. 41 (1937), 375-481.
[265] Taylor, J.L., A counterexample in shape theory, Bull. Amer. Math. Soc. 81
(1975),629-632.
[266] Tezer, C., Shift equivalence in homotopy, Math. Z 210 (1992), 197-201.
[267] Toda, H., On unstable homotopy of spheres and classical groups, Proc. Nat.
Acad. Sei. 46 (1960), 1102-1105.
[268] Torwiczyk, H., On CE-images of the Hilbert cube and characterization of Q-
manifolds, Fund. Math. 106 (1980), 31-40.
mSTORY OF SHAPE THEORY 1177
[269] Tychonoff, A., ber die topologische Erweiterung von Rumen, Math. Ann. 102
(1930),544-561.
[270] Trybulec, A., On Some Properties of Movable Continua (in Polish), Thesis,
Warszawa 1974.
[271] Venema, G.A., Embeddings of compacta with shape dimension in the trivial
range, Proc. Amer. Math. Soc. 55 (1976), 443-448.
[272] Vietoris, L., Bereiche zweiter Ordnung, Monatsh. fr Math. und Phys. 32 (1922),
258-280.
[273] _ _ , Kontinua zweiter Ordnung, Monatsh. fr Math. und Phys. 33 (1923),
49-62.
[274] _ _ , ber den hheren Zusammenhang kompakter Rume und eine Klasse
von zusammenhang streuen Abbildungen, Math. Ann. 97 (1927), 454-472.
[275] Vogt, R.M., Homotopy limits and colimits, Math. Z. 134 (1973), 11-52.
[276] Watanabe, T., Approximative shape II-Generalized ANR's, Tsukuba J. Math.
11 (1987), 303-339.
[277] Waiewski, T., Sur un continu singulier, Fund. Math. 4 (1923),214-235.
[278] Weber, C., La forme d'un espace topologique est une completion, CR. Acad.
Sei. Paris 277 (1973), A7-A9.
[279] Whitney, H., Regular families of curves, I, Ann. Math. 34 (1933), 244-270.
[280] Williams, R.F., Classification of one-dimensional attractors, Proc. Symp. Pure
Math., Amer. Math. Soc. 14 (1970), 341-361.
[281] Winslow, A., There are 2c nonhomeomorphic continua in rin - rin , Pacific J.
Math. 84 (1979),233-239.
[282] Wojdyslawski, M., Sur la contractilite des hyperespaces des continus localement
connexes, Fund. Math. 30 (1938), 247-252.
[283] _ _ , Retractes absolus et hyperespaces des continus, Fund. Math. 32 (1939),
184-192.
[284] Yagasaki, T., Fiber shape theory, Tsukuba J. Math. 9 (1985), 261-277.
[285] _ _ , Movability of maps and shape fibrations, Glasnik Mat. 21 (1986), 153-
177.
A mSTORY OF THE NORMAL MOORE SPACE PROBLEM
PETERJ.NYIKOS
Department 01 Mathematics
University 01 South Carolina
Columbia sc 29208
Contents
1179
C. E. Aull and R. Lowen (eds.), Handbook ofthe History ofGeneral Topology, Volume 3, 1179-1212.
2001 Kluwer Academic Publishers.
A mSTORY OF THE NORMAL MOORE SPACE PROBLEM 1181
This is a thematic history of the normal Moore space problem, which was
for many years the general topology problem whose solution was most eagerly
sought after. This was partly due to the dominance of the RL Moore school
of point-set topology, and partly due to the fact that it became enmeshed in set
theory from the beginning, and gave rise to a whole area of topology that has
been, in the words ofTall [1984] "at the cutting edge of set-theoretic topology,
very frequently being the first topological consumer of a new set-theoretic
technique" .
It is the second factor which predominates in the account that follows. I
leave it to those more experienced in historical writing than myself to fiesh
out the varlous personal infiuences that helped make the normal Moore space
problem the unique thing it was in the years between its statement and solution.
[Paradoxically, Moore himself does not seem to have infiuenced people in this
direction: see Tall [1981] where it is said, "Burton [Jones] later toId me Moore
was not fond of the normal Moore space problem".] While the chronological
order of events is generally followed below, the primary focus is on the pro-
gression of mathematical ideas that shaped research on the problem; and so,
the narrative will sometimes spring forward and sometimes fall back a good
number of years.
In Section 11 there is a list of problems which are referred to by number
from time to time. There are many problems and results that could have been
mentioned but were not inc1uded due to lack of space. A good source for others
is Tall [1984].
Throughout this account, "space" will mean "Hausdorff topological space"
with the exception of one comment on Frechet spaces-L in Section 2.
In a metric space we can let OUn be the set of all open disks of radius 1/ n, and
it is an easy exercise to show that this gives a development. Moore spaces were
named after R.L. Moore, the well-known founder of what came to be known as
the Texas school of topology. They are the spaces given by the first three parts
ofhis lengthy Axiom 1 in Moore [1932]. The whole axiom gave the definition
1182 PETER J. NYIKOS
of a complete Moore space, and was one of a set ofaxioms in Moore [1935]
characterizing the Euc1idean plane in an intrinsic way. The history ofAxiom 1
has been recounted in Volume 1 by Iones [1997].
As the supply of Moore spaces increased, it became natural to ask:
Or, to use the words of Iones [1966], what is the "right reason" that non-
metrizable Moore spaces fail to be metrizable? What was wanted was a property
that is strong enough to apply to all Moore spaces, not just certain c1asses such
as separable Moore spaces or locally compact Moore spaces. At the same time,
"overkill" was to be avoided: it should be as simple and general a property as
possible, just strong enough to make all Moore spaces satisfying it metrizable
without also making, say, all regular spaces metrizable. And so, as the store-
house of nonmetrizable Moore spaces grew, it was only a matter of time before
someone focused on the elementary separation axiom of normality:
Definition 1.2. Aspace is normal if for every pair Fl, F2 of disjoint closed
sets, there are disjoint open sets Ui such that Fi C Ui for i = 1, 2.
All metrizable spaces are normal. This is a fact whose easy proof appears
in most textbooks of general topology. On the other hand, as if by accident (hut
as we now know, it was no accident), all the nonmetrizable Moore spaces found
up to 1933 (and a good way beyond) failed to be normal.
It was on October 28, 1933 that F. Burton Iones, then a student of R.L. Moore,
public1y announced, at an AMS meeting, both the following problem and a
tantalizing partial solution:
The Normal Moore Space Conjecture (NMSC) was that the answer to this
problem is Yes. This conjecture sounded at first like many other questions
circulating in R.L. Moore's c1asses, and no one suspected for a long time that
it was inextricably bound up with set-theoretic independence results, let alone
A mSTORY OF THE NORMAL MOORE SPACE PROBLEM 1183
The special case ID I = ~o is what J ones explicitly proved, but he pointed out
immediately that the argument, with slight changes, also shows that the axiom
2~o < 2~1 implies that no separable normal space can have an uncountable
c10sed discrete subspace, and the argument can also be used to show Lemma 2.1
with similarly slight changes. Jones also showed, in the IDI = ~o case, that a
separable normal space cannot have a c10sed discrete subspace of cardinality
c = 2~o; this is immediate from Lemma 2.1 and the fact that 2K > K for an
cardinal numbers K. Applications of this will be given in the next seetion.
When Jones first obtained bis result, he had no inkling that 2~o < 2~1 might
not follow from the usual (ZFC) axioms of set theory. In fact, as an obedient
student of R.L. Moore, he had done almost no reading of the papers of set
theorists, and so the following exchange took place between them one day,
as recounted by J ones many years later. Jones said to Moore, "There must be
something wrong with me. Ijust can't seem to prove that 2/oto is less than 2/otl".
Moore responded, "Ha, ha, ha. Neither can Sierpinski". Jones also wrote in his
1937 paper, "The author has tried for some time without success to prove that
2~o < 2~1". This was not, however, the real explanation for the long delay
between his 1933 presentation and the publication. The real reason was that
Jones was trying to settle the normal Moore space problem outright; he would
have been happy to try any reasonable-looking axiom if it would achieve that
goal, one way or the other. But about forty years were to elapse before that was
to happen.
In addition to his normality theorems, Jones [1937] gave, in effect, the first
example of a nonmetrizable pseudonormal Moore space.
1184 PETER J. NYIKOS
There are some c10sely related spaces which can be used in place of r in
this theorem. One is the Cantor tree, with the points on its top level identified
with the Cantor set.
Definition 3.4. A tree is a partially ordered set (T, ~) in which the set; =
{x E T : x ~ t} is well-ordered for each t E T. Given an ordinal number a, the
Jull binary tree 01 height a is the tree of all transfinite sequences 1 : ~ K,
for some ordinal < a, and the order on the tree is end extension: 1 ~ g iff
dom 1 c dom g and g r dom 1 = I. The Cantor tree is the full binary tree of
height w + 1.
Some have used the term "Cantor tree" for any uncountable sub set of the
full binary tree of height w + 1 that inc1udes the full binary tree of height w.
The points on the top level of what I here call the Cantor tree are identified
with the Cantor set by identifying each infinite sequence of O's and I 's with
the real number between 0 and I whose digits, in temary notation, are formed
by replacing all I 's in the sequence with 2's. There is a way of extending this
identification, matching the whole Cantor tree with a subset of the plane, given
in detail in Nyikos [1989]. The sequences of finite length are put directly above
the center points in the deleted middle thirds of the points of the Cantor set
in a systematic fashion. The empty sequence is matched up with (1/2,1/2).
Sequences with n terms are identified with points on the horizontalline 1/2n + 1
above the x-axis, and the two one-term extensions of each sequence (J' of n - 1
terms are matched up with points on this line that meet lines of slope 1 and -1
passing thru the point matched with (J'
The topology on the Cantor tree can be visualized by using this match-
up, along with the relative topology obtained by rotating the topology of the
Sorgenfrey plane 45 degrees counterc1ockwise. That is, we erect above each
point p on the x -axis a sequence of isosceles right triangles with horizontal
bases and with their apexes at the point p, of height 1/ n for each positive integer
n, and use those triangles as a base for the neighborhoods of p. All points except
those on the top level of the Cantor tree are isolated in this topology. Just as with
the tangent disk space, a subspace which exc1udes all points on the x-axis except
for an uncountable '}..-set gives a nonmetrizable pseudomormal Moore space. In
addition, this example is locally compact. In Nyikos [1989] it is explained how
this topology coincides with the interval topology on the Cantor tree.
1186 PETER J. NYIKOS
Definition 3.5. The interval topology on a tree T is the one whose base is all
sets of the form (s, t] = {x E T : s < x ~ t}, together with all singletons {tl
such that t is a minimal member of T.
Also in the 1930's, Iones defined a completely different kind of tree in the
hope that it would be a nonmetrizable normal Moore space. He called it a "tin
can space" because of a visualization of it as aseries of (VI horizontal shelves
of tin cans, with countably many cans on each shelf. A lively exposition of this
example T, which Iones apparently never published (but which is a subspace
of aspace in Iones [1966]) can be found in Bing [1981]. Here is a bare-bones
description:
Cantor tree. We will encounter these spaces again in Sections 5 and 6, where
similar things are done in the quest for normal subspaces.
Jones's hypothesis 2~o < 2~1 is a weakening of the venerable old Continuum
Hypothesis (CH), the truth or falsity of which was the first in a list of problems
singled out for special treatment by Hilbert in a famous lecture a hundred years
ago. This axiom, which can be stated as "2~o = ~1", had seen a lot of use by
the time Jones used its consequence 2~o < 2~1, but Jones still deserves credit
for first using 2~o < 2~1 itself in a proof which (as we now know) cannot be
done using just the usual (ZFC) axioms.
Another already-old problem, posed by Souslin [1920], was destined to
have a fascinating indirect effect on the normal Moore space problem. Souslin
took a characterization of ~ as the only separable, connected linearly ordered
space with no greatest or least point, and asked whether countable cellularity-
the condition that every disjoint collection of open sets is countable-could
be substituted for separability. An affirmative answer to this question came to
be known as Souslin's Hypothesis (SH). It turned out to be equivalent to the
nonexistence of a Souslin tree; see Kurepa [1935], [1942] where, in asense,
the whole modem theory of trees was born. Both Souslin trees and Jones's tin
can space T are examples of what is known as Aronszajn trees, another concept
introduced in Kurepa's articles.
Notation 4.1. If T is a tree, then T (0) is its set of minimal members. Given
an ordinal ol, if T() has been defined for a1l < ol, then T ra
= U{T() :
< ol}, while T(ot) is the set of minimal members of T \ T rot. The set T{ot)
is called the Ol-th level of T.
For example, in the case of Jones's T, the Ol-th level is simply the set of all
functions f E T with domain ol.
theory. Of course, at the time Souslin posed his question, Gdel's incomplete-
ness theorems were yet to be discovered or even conjectured. Hilbert was still
optimistic about being able to put all of set theory, hence all of mathematics,
on a firm footing. But already in 1930, Gdel delivered the first of his blows
against the program with his incompleteness theorems. The first showed that any
recursively axiomatizable system sufficient to do "arithmetic" (more precisely:
elementary number theory) would have undecidable statements. Even more
devastating to the Hilbert program was the second, which stated that any such
system could not even be used to show its own consistency. Then Gdel [1939]
gave a partial answer to Hilbert's First Problem, by defining the universe L
of constructible sets, and showing that it satisfies eH along with all the usual
(ZFC) axioms. Taken together, these results told mathematicians that they could
apply eH with no more (or less!) fear of arriving at a contradiction than ifthey
did ordinary everyday mathematics. To some, this settled the status of eH, but
Gdel himself was not satisfied, and in a highly readable artic1e Gdel [1947]
gave reasons why he thought eH was an 'unreasonable' axiom.
Gdel's reasons would make no sense to a strict formalist, but then, such
people are rare if nonexistent. On the other hand, many mathematicians are
Platonists without even having thought about it, and some of the leading set
theorists are also Platonists. A rather common naIve view about eH and other
special axioms was expounded on by RH Bing in a colorful artic1e on the normal
Moore space problem Bing [1981], where he commented:
Mathematicians speak of the set of real numbers (or of the reals) rather
than some set of reals. This suggests the underlying assumption that there
is only one set of reals. If one c1ass in some school is studying the reals
and another c1ass somewhere else is studying them also, the two c1asses are
studying the same thing. However, we are told in axiomatic set theory that
there are many sets of reals.
For example, in a sufficiently complicated universe, the set of constructible
reals, which are the real numbers in the model L, is a countable set. Yet they are
the reals in Gdel's constructible universe L. [As we shall see in Seetion 9, any
universe where all normal Moore spaces are metrizable is such a "sufficiently
complicated" universe.] Bing was less pleasant in some of his oral public state-
ments, but one can hardly distike such homespun comments as the following
one, which occurs further down in the same page of Bing [1981] as the one just
quoted:
The terms uncountable, countable, infinite, finite have different meanings
to different people. These concepts are used to indicate how many elements
a set has. Would one call a drove of homets infinite or uncountable just
because it is impracticable (or impossible) to look inside their nest? I hope
not. An uncountable set has more elements than there are integers. To many
A mSTORY OF THE NORMAL MOORE SPACE PROBLEM 1189
Definition 4.3. A Q-set is an uncountable subset Y of R such that every sub set
of Y is a G li in the relative topology of Y.
[Some leave out the word "uncountable" in the definition, thus extending "Q-
set" to include every countable subset of R] The axiom Rothberger formulated
and applied is nowadays denoted l' > ~ 1, where l' is one of the standard
uncountable cardinal numbers systematically treated by van Douwen [1984].
1190 PETER J. NYIKOS
It involves the idea of an infinite set being "almost contained" in every member
of a family of sets:
Some authors omit the word "infinite" in the preceding definition, thereby
allowing any finite sub setof X to be a pseudo-intersection of any filter on X.
Definition 4.5. Let qJ be the dass of all filters whose members are all infinite
subsets of a countable set. Then
p = min {I ~ I : ~ is a base for a member of qJ that has no pseudo-intersection}.
Three years after Rothberger's paper appeared, Bing [1951] featured a similarly
bold theorem that applied Q-sets to the normal Moore space problem:
Theorem 5.1. Any subspace 0/ the Moore plane that meets the x-axis in a
Q-set is normal.
The Moore plane is separable, but no subset of it that meets the x-axis in
an uncountable subset is second countable; so the following is immediate from
Theorem 5.1 and the Urysohn metrization theorem:
This theorem obviously runs in the opposite direction from Jones's theorem,
and yet its use of a Q-set is just an extension of Jones's use of a -set to produce
his Frechet space-L. It is therefore interesting to speculate on how 10ng Bing
knew of Theorem 5.1 before he actually published it; perhaps he was waiting
to come up with related results that did not go out on a set-theoretic limb. And
come up with them he did! The second notable accomplishment ofBing [1951]
was the definitive metrization theorem for Moore spaces, using a concept that
is very similar to normality.
A mSTORY OF THE NORMAL MOORE SPACE PROBLEM 1191
Theorem 5.5. Aspace is metrizable if, and only if, it is a collectionwise normal
Moore space.
Theorem 5.6. Aspace is metrizable if, and only if, it is a regular space with a
a -discrete base.
Lemma 5.7. Every Moore space is subparacompact; that is, every open cover
has a a -discrete closed refinement.
The word "subparacompact" was introduced later, and often it is defined us-
ing "a-Iocally finite" in place of"a-discrete". The two definitions are equivalent,
cf. Theorem 3.1 Burke [1984a], which inc1udes another equivalent condition
from which Lemma 5.7 follows immediately.
Another major achievement in Bing [1951] was the first ZFC construction
of anormal space that is not collectionwise normal, widely known as "Bing's
Example G". This space is obtained by isolating all but ~1 carefully chosen
points in the product of2~1 two-point discrete spaces; these ~1 points are chosen
so that they will be a a (c1osed) discrete subspace. Since each neighborhood of
a nonisolated point contains a neighborhood that is open in the usual product
topology, we cannot expand these uncountably many points to disjoint open
1192 PETER J. NYlKOS
Definition 5.8. The product measure on a product of copies of {O, I} is the one
that assigns to each basic dopen set B the measure 2-n , where n is the number
of coordinates in which B is restricted.
For example, if B is the set B(4In) of all elements of the product {O, l}lR
that take on the value 0 at 4 and the value 1 at n, then the product measure of
B(4In) is 1/4. The entire product has measure 1, and every nonempty open set
has positive measure, so that there cannot be uncountably many disjoint open
sets.
The year before, Arens and Dugundji [1950] had introduced a concept that
would playa signmcant indirect role in the normal Moore space problem:
Alexandroff [1960] posed the question whether every normal space with a
uniform base is metrizable. In effect, he was asking:
A HISTORY OF THE NORMAL MOORE SPACE PROBLEM 1193
The undedying set of the tangent V space is the c10sed upper half plane. Points
above the x-axis are isolated, while a point p on the x-axis has a neighborhood
base whose members are the union of two line segments with slopes of 1 and
-1, ending in the point p.
It is easy to see that Heath's tangent V space has a uniform base: just take
the basic neighborhoods of height 1/ n as n ranges over the positive integers.
Thus, by Heath's theorem, it is a nonmetrizable metacompact Moore space.
Also, Heath [1964] showed that the subspace formed by removing a11 but a
Q-set of points on the x -axis is normal, and also gave the definitive solution
to the "separable normal Moore space problem" by showing the converse of
Corollary 5.2, thus giving:
Not long after, Bing [1964] gave a somewhat complicated set theoretic
characterization of when there is a normal Moore space that is not collectionwise
Hausdorff. Despite the title of Bing [1964] he gave no translation of the normal
Moore space problem itself into set theory, and we do not have a convenient
translation of it even now. But Bing's characterization was significant for the
reason that he also showed the set-theoretic property in question to be equivalent
to the existence of a first countable normal space that is not collectionwise
Hausdorff. [A space is first countable if every point has a countable base for its
neighborhoods.] In other words, Bing showed:
Since the c1ass of first countable spaces is far more extensive than the c1ass
of Moore spaces, this result had the effect of diverting a lot of attention from
the normal Moore space problem to:
1194 PETER J. NYIKOS
Problem 5.14. Is there a first countable normal space that is not collectionwise
normal?
We lack a set-theoretic translation for this problem too. Tall [1972] has
come as elose as any, giving a set-theoretic translation for a related problem
that applies to all first countable spaces. It involves the following concepts.
Definition 5.15. A family :t of subsets of aspace X is normalized if, whenever
:t' c :t then there are disjoint open sets U and V such that U:t' c U and
U :t \:t' c V. :t is completely separated if it expands to a disjoint collection
of open sets.
What Tall [1972] succeeded in doing was to find a set-theoretic character-
ization of the statement, ''There is a first countable space in which there is a
normalized collection of elosed sets that is not completely separated" . As with
Bing's result, one can substitute "Moore" for "first countable". Naturally, this
led to some more speculation that Problem 5.14 is equivalent to the normal
Moore space problem, but this is still not settled (Seetion 11, Problem 7). These
ideas were further developed in Tall [1974] where the construction in Bing
[1965] was modified to produce a nonmetrizable metacompact normal Moore
space in any model where there is a normal first countable (or: Moore) space
that is not collectionwise Hausdorff.
Now we return to the early 1960's to recall the revolution in set theory begun
by Cohen' s revolutionary discovery that the negation of CH is consistent if ZFC
is consistent. Credit for the Promethean task of generalizing Cohen's method
(known as "forcing") and making it understandable to a large mathematical
audience goes to Scott and Solovay, who introduced the general use of Boolean-
valued models, and to Shoenfeld, who showed how to do forcing with posets
in general. The new technique was to revolutionize the field of set-theoretic
topology, with the normal Moore space problem and related problems like 5.14
in the vanguard. The most common way of looking at it is to view it as taking a
countable transitive model M of the ZFaxioms, known as ''the ground model"
and carefully adding a new set G, which is a generic subset of aposet, known as
"the forcing poset". The set G is taken from the larger universe V which "knows
M is countable". A new countable transitive model, M[G], is then defined to
be the smallest transitive model of ZF within V to contain G as a member and
M as a subelass. Known as "the forcing extension", M[G] satisfies the axiom
of choice if M does, but many other axioms, like CH, do not always carry over.
Cohen himself did what is nowadays referred to as "adding ~2 Cohen reals".
If the ground model satisfies 2~1 = ~2, as it did in Cohen's original treatment,
A HISTORY OF THE NORMAL MOORE SPACE PROBLEM 1195
then 2~o < 2~1 is satisfied, too. This, however, is not enough to give Q-sets,
and as it later turned out, the process of adding ~2 Cohen reals destroys all
Q-sets. The first forcing argument to produce Q-sets arose through work on the
Souslin problem. The results were curiously complementary. On the one hand,
Tennenbaum [1968] and Solovay and Tenenbaum [1971] showed that adding
~l Cohen reals [later reduced to just one Cohen real by Shelah] produces a
Souslin tree in the forcing extension. On the other hand, Martin and Solovay
[1970] showed the consistency ofthere being no Souslin trees by using an axiom
whose consistency they proved there. This axiom, known as "Martin's axiom
and the negation of the continuum hypothesis" (MA + ...,CH) was later used
to show the consistency of there being Q-sets and hence of there being normal
nonmetrizable Moore spaces. Not long thereafter, the axiom came to be applied
all over general topology, measure theory, and the theory of Boolean algebras,
and in a number of other branches of mathematics.
The Q-set result has a curious history. In 1968, while a student at the
University of Wisconsin, Franklin Tall told Jack Silver, who was there at the
time, about Bing's Q-set result, letting Silver know that he was working to show
Q-sets consistent. Silver gave an iterated forcing proof of what is called Solo-
vay's lemma in Rudin [1975], crediting it to Solovay, and using it to show the
consistency of there being Q-sets. Silver never published his proof, and it turned
out to be redundant, in the following way. Not long after Silver's result, a fellow
student ofTall at Wisconsin, David Booth, proved that Martin's Axiom implies
p = c, a result which is called "Booth's lemma" in Rudin [1975] and is also
an immediate corollary of Solovay's lemma. [Later, van Douwen showed the
conc1usions ofthe two lemmas to be equivalent.] Then Booth discovered, while
browsing through old issues of Fundamenta Mathematicae, that Rothberger had
already shown Theorem 4.6, thereby giving Silver's contribution the status of a
rediscovery.
The upshot of all this is that suddenly, the existence of nonmetrizable normal
Moore spaces was known to be consistent, inc1uding examples that were (l)
separable and locally compact (subspaces of the Cantor tree) and (2) separable
and locally connected (subspaces of the Cantor road space or of the Moore
road space or of the Moore plane). But this was far from being the end of the
applications of Martin's Axiom to the normal Moore space problem. It was
soon also applied successfully to Jones's tin can space and his road space. First,
Baumgartner, Malitz, and Reinhardt [1970] improved on Martin and Solovay's
result, showing that MA + ...,CH implies all Aronszajn trees are special. Then
Fleissner [1975] showed that special Aronszajn trees, as weIl as Jones's road
space, were normal under the same axiom. Earlier, Jones had proved, in effect,
that an Aronszajn tree is a Moore space if, and only if, it is special (for a proof
see Todorcevic [1984]). Putting these results together and extending one proof
slightly, we obtain:
1196 PETER J. NYIKOS
Theorem 6.1. lf MA + -.CH, then every Aronszajn tree, and every road space
canonically obtained from one, is a nonmetrizable normal Moore space.
This was not the first consistency result, nor is it the last, on the subject of
normality in special Aronszajn trees; in fact, the last word on it is yet to be said
[cf. Problems 13 and 14]. We will have more to say on this at the beginning and
the end of the next section.
The first result on when special Aronszajn trees are normal is a corollary of a
theorem in the dissertation of Tall [1977]:
G and wl \ G then there are open sets in the ground model that mimic them
weIl enough and meet in sufficiently many points that Uo and UI will meet in
some of the same points. First countability comes into play in guaranteeing the
existence of these mimics.
Theorem 7.1 is actually the K = WI case of a general theorem of Tall [1969]
which works for all regular cardinals K, but requires that the spaces be A-cwH
for all A < K (which is automatic if K = wt) in addition to being normal. One
can work one's way up inductively for a while: after the first step we get all
normal first countable spaces to be wl-cwH, and are ready to apply the forcing
for K = W2., and so forth. This runs into complications at singular cardinals of
cofinality WI, and it took another five years before Fleissner [1974] showed the
way.
Fleissner did not use forcing for his theorem; instead he utilized some
exceptionally deep analyses by Jensen of Gdel's Constructible Universe L.
Jensen had shown that Souslin trees exist in L by using an axiom ~, which
ever since has competed with Martin's Axiom for the honor of being the most
commonly utilized set theoretic axiom in topology. Fleissner modified Jensen's
analysis and showed that L sastisfied the more demanding axiom ~ss(K) for all
regular uncountable cardinals K. In the same 1974 paper, he utilized this axiom
at regular cardinals and GCH at singular cardinals to show:
Theorem 7.2. In L, every normal space 0/ character ~ ~I is collectionwise
Hausdorff. In particular, every normal Moore space is collectionwise Hausdorff.
Proofs can be found in Fleissner [1974], [1976b] and also in Tall [1984],
which also explains how to combine Fleissner's ideas at singular cardinals
with Tall's earlier forcing arguments to show that Tall's forcing model can be
substitutedior L in Theorem 7.2. It also gives a proof of the following corollary
via a number of generalizations:
Corollary 7.3. In L, and in the reverse Easton model 0/ TaU [1969J, every
locally compact normal Moore space is metrizable.
A simple direct derivation of 7.3, by coIlapsing compact sets to points,
and applying the fact that Moore spaces are subparacompact, can be found in
Rudin [1975], who credits it to G.M. Reed. In view of the locally compact
nonmetrizable normal Moore spaces of Section 6 from Q-sets and Aronszajn
trees, this is only a consistency result. The sacrifice of local compactness to
produce local connectedness in the c10sely related road spaces was necessary for
their normality: all attempts to create modifications preserving both properties
were shot down by the remarkable ZFC result ofReed and Zenor [1976]:
Theorem 7.4. Every locally compact, 10caUy connected normal Moore space
is metrizable.
1198 PETER J. NYIKOS
In Alster and Zenor [1977] the local compactness was relaxed to rim-
compactness, and normal Moore spaces to perfect1y normal subparacompact
spaces, with the conc1usion being that the resulting spaces are paracompact.
Gruenhage [1979] improved this further to:
Theorem 7.7. 2~o < 2~1 implies that special Aronsajn trees are never normal.
The key was a pair of O-like axioms, that Devlin and Shelah [1978] showed
to follow from 2~o < 2~1. Taylor [1981] simplified the proof of 7.7 while
showing the two axioms are actuaUy equivalent to 2~o < 2~1, and Nyikos
[1981] gives an interpretation of the two axioms which makes it very easy to
see that they imply 2~o < 2~1. Both Tall [1981] and Nyikos [1981] comment on
the irony of Jones's own axiom 2~o < 2~1 implying that Jones's tin can space
and road space are never normal.
On the other hand, Shelah [1982], [1999] has constructed a model of CH
in which there is a special Aronszajn tree which is countably paracompact.
This is in contrast to the situation as regards separable Moore spaces, where
CH implies the countably paracompact ones are aU metrizable. The question
of whether there is any such contrast for the property of normality is still open
(Problem 13).
Now it is time to begin the account of what led to the first general consistency
proof for the NMSC. The path was a very indirect one, and the proof outlined
here is not the direct proof one can find in such places as Nyikos [1980a]
and Fleissner [1984]. It began with the construction of universal Moore spaces
by Rudin and Starbird [1977]. These spaces were universal for the properties
that Bing and TaU had given set-theoretic characterizations for, as recounted
in Section 5. One class of spaces consisted of Moore spaces that are normal
in ZFC; the class includes a nonmetrizable member if, and only if there is a
first countable normal space that is not collectionwise Hausdorff. The other
consisted of Moore spaces, designated TA for the various infinite cardinals A.,
each with a discrete collection of closed, strongly zero-dimensional metrizable
1200 PETER J. NYlKOS
subsets. In any model in which there is a first countable space in which there
is a normalized family of 'A dosed sets that cannot be separated, TA is such a
space.
Unfortunately, T).. cannot be normal under these circumstances. Although
the points outside the discrete family 6 = {Ca : a < K} of dosed sets are
isolated, and although the Ca themselves are metrizable, it is still possible to
split U 6 into two disjoint dosed sets that cannot be put into disjoint open sets
unless the space is metrizable. The trick, shown in detail in Rudin and Starbird
[1977], is to break up each Ca into halves that mimic each other well enough so
that putting the union of the ''upper halves" and the union of the "lower halves"
into disjoint open sets makes it possible to define a disjoint open expansion of
6 itself. The resulting open sets are automatically dopen, and together with
the remaining isolated points they form a partition of the space into dopen
metrizable chunks.
In November of 1977, I worked out a way to take advantage ofthe fact that
the dosed sets are strongly zero-dimensional and metrizable, to produce a dass
of normal Moore spaces of this kind. Whimsically named "'A-sauri", in Nyikos
[1980c], they have the universal property that if there is a normal Moore space
in which there is a discrete collection of 'A strongly-zero dimensional dosed
subsets which cannot be completely separated, then 'A-saums is such aspace.
My hope was that it would turn out that some T).. is not collectionwise normal
and that the proof could be modified to show that 'A-saums is not collectionwise
normal either, hence is a ZFC example of a nonmetrizable normal Moore space.
The first possibility is mentioned with interest at the end of Rudin and Starbird
[1977]. However, things worked out very differently, and the inspiration came
from the same Rudin-Starbird paper--or, rather, from some conversations I had
had with Starbird back in 1976 about its contents.
At that time, Starbird told me about a modification (see below) of the
following infinite game, played by Players A and B in opposition to Dealer
C. There is an infinite set X fixed in advance, and C has two infinite decks of
cards, {An} ~ I and {B n }~ I and each card An or B n lists a set of functions from
X to 0, 1. The two players are not allowed to exchange information once the
game begins. It begins with the dealer separately handing each player a slip of
paper determining an element (a and b, respectively, with a ::/= b) of X. The
goal of Players A and B is to collect enough cards so that there is a function
f : X --+ {O, I} that splits a from b, and that appears both on a card held by
Player A and one held by Player B.
On the first turn the dealer deals Al to A and BI to B. On the nth turn, a
player still active can either request the nth card from the corresponding deck,
or else can say "Stop", becoming inactive until the end. The game ends if A
and B have both said "Stop". The two-player team of A and B wins if either (1)
the game continues for (J) moves and either U~l An or U~l B n fails to be all
A mSTORY OF THE NORMAL MOORE SPACE PROBLEM 1201
of 2 X or (2) the game stops after finitely many steps and Players A and B have
eolleeted enough eards to aehieve their goal.
If A and B knew the identity of a and b in advanee, they eould agree on a
funetion f as above, and eaeh eould stay in the game until f turns up on one
of his eards. As it is, ~ey ean be assured of winning if X is eountable-they
need only let X = {x n : n E w}, and A [resp. B] waits until all eharaeteristie
funetions of the singletons {xd up to X n = a [resp. X m = b] appear on the
eards he has eolIeeted-but that is the best they ean do "in ZFC". All this also
applies to the aetual Starbird game, where Dealer C lays out all the eards for
the respeetive players, face up, and eaeh picks up a finite hand. Otherwise the
games are the same: in Starbird's game, it is as if both Player A and Player
B had perfeet foreknowledge of all the plays in the game I have described.
Whether this ever makes any differenee as to A and B being guaranteed a win
is Problem 15 of Seetion 11. What was elear already to Starbird in 1976 is
that if Player A and Player B ean always win his game for a given A, then
T).. is eolIeetionwise normal and henee metrizable; and so the Normal Moore
Spaee Conjeeture (NMSC) foUows, along with all the stronger results already
mentioned, if there is no restrietion on A.
The story of what thought processes oeeurred in between these eonversations
and theNovember 1977 breakthroughhas been toldinNyikos [1980b], an artiele
that elosed with a personal eomment of the sort that was fashionable in the days
of Sylvester and Poineare, but seems to have beeome of questionable taste in the
meantime. My reason for writing it is that, just a little over a month before the
breakthrough, I had had a very elose brush with death that had made me aeutely
aware of my mortality, and I didn't want to have to wait many deeades for a
more appropriate opportunity to make these details known. Suffiee it here to say
that I dimly remembered at the right moment that somewhere [Solovay [1971],
as I found out a few days later] it had been shown eonsistent, modulo very large
cardinals, that Lebesgue measure, and with it the usual measure on the Cantor
set (in other words, 2W ), ean be extended to eountably additive measures defined
on every subset of the realline and the Cantor set, respeetively. This is enough
to show that Players A and B ean win the Starbird game for eountable X without
looking at their slips of paper. The usual (produet) measure f..t has f..t(2 x ) = 1;
and, for any pair a, b of distinet points, the set B(alb) of a11 funetions f
satisfying f(a) = 1, f(b) = 0 has measure 1/4. Eaeh player eontinues until the
set of funetions listed on his eards has measure > 7/8. Then the set of funetions
listed on both sets of eards has measure > 3/4, henee meets B(alb), as desired
by Players A and B.
The foregoing argument goes through verbatim for any set X, no matter
how large, provided one assumes the folIowing axiom-and then the whole
maehinery of Rudin and Starbird kicks in to give us NMSC and mueh else.
1202 PETER J. NYIKOS
The Product Measure Extension Axiom (PMEA) is that, for any set X, the usual
product measure on the set 2 x can be extended to "e-additive" measure defined
on every subset of 2 x . [A measure is e-additive if the union of fewer than e sets
of measure 0 is likewise of measure 0.]
Kunen had written, but never published, a proof that the PMEA is consistent
if it is consistent that there is a strongly compact cardinal. Strongly compact
cardinals come quite far up in the large cardinal hierarchy, higher even than mea-
surable cardinals and just below supercompact ones. [A fine article on the subject
of large cardinals is Kanamori and Magidor [1978], and Kanamori [1997] is a
whole book on the subject. For those desiring a more elementary introduction,
there is the fascinating article by Maddy [1988] with a philosophical twist.]
Many people, including mys elf, had mixed feelings about this solution to
NMSC. In my case, besides the shakiness of relying on such large cardinals,
there was also a feeling of sadness akin to what many have experienced over the
disappearance of any problem that lent focus to a large body of mathematical
activity. For although the NMSC was not really settled by this result, its character
was irrevocably changed. Moreover, it was a solution in the "wrong" direction:
like most people working on the problem, I had been striving to build a ZFC
example of a normal Moore space. Frank Tall was an exception, hoping for
an independence result. Tall [1979] expressed the mathematical situation in the
following terms:
Is then the normal Moore space problem solved? No! The assumption of
the consistency of a strongly compact cardinal is too strong. It may in fact
be false. What is needed is a proof that if there is a model of set theory,
then there is one in which every normal Moore space is metrizable, and
this is still open. A pessimistic conjecture would be that if every normal
Moore space is metrizable, then there is a large cardinal. This would place
the problem in a strange sort of limbo in that there would be little hope for
an example and no hope for a consistency theorem.
This "pessimistic conjecture" was amply verified by Fleissner [1982], [1983].
Yet, even before this happened, TaU [1980] wholeheartedly embraced results
faUing into this very limbo, and campaigned for the use of cardinals larger even
than strongly compact ones. We will see a few of these uses in the area of
collectionwise normality in Section 10.
How did Fleissner do it? The inspiration came from a seemingly unrelated
problem, that of whether every regular para-Lindelf space isparacompact.
A mSTORY OF THE NORMAL MOORE SPACE PROBLEM 1203
Theorem 9.2. Ifthere is a model rot ojZFC in which there are no normal Moore
spaces, then in rot there is an inner model with a proper dass oj measurable
cardinals.
Axiom 9.3. Given a cardinal number 'A of uncountable cofinality, E('A) is the
axiom that there is a stationary set E C 'A consisting of ordinals of countable
cofinality, such that E n is nonstationary in for all < 'A.
10. Aftermath
Theorem 10.1. The PMEA implies that every countably paracompact space 0/
character less than c is expandable.
Corollary 10.2. The PMEA implies every countably paracompact Moore space
is metrizable.
For the next theorem of Burke [1984b], recall that space is strongly col-
lectionwise HausdorjJ [abbreviated scwH] if every c10sed discrete subspace
expands to a discrete collection of open sets.
Every normal cwH space is easily shown to be scwH, but the argument
does not go through if countable paracompactness is substituted for normality.
In fact, although we know that every first countable normal space is cwH (and
hence scwH) in the constructible universe L (Theorem 7.2), the best we have
for countable paracompactness so far (see Problem 8 (a) of Section 11) is the
result of Watson [1985]:
Theorem 10.4. In L, every regular, countably paracompact space of charac-
ter :s c (hence every countably paracompact Moore space) is collectionwise
Hausdorff.
Earlier, Watson [1982] had showed a striking parallel to the foregoing
results:
Theorem 10.5. In L, every locally compact normal space is collectionwise
Hausdorff.
In Watson [1985] he posed the question ofwhether countable paracompact-
ness can be substituted for normality. This is still unsolved (Problem 8 (b.
He and other Torontoans, particularly Frank TaU, worked intensivelyon the
problem of getting every 10caUy compact normal space to be collectionwise
normal in some model. The models they looked at inc1uded the one in which
Kunen first showed PMEA to be consistent: the one obtained by starting with
a ground model with a strongly compact cardinal K in it and forcing with the
usual poset for adding Krandom reals. In Fleissner [1984] the actual process is
shown whereby this forcing destroys the normality of aU spaces of character less
than c that are not collectionwise normal; the ideas are much the same as in TaU
[1977] except that random reals are added and K is assumed to be supercompact
rather than just strongly compact. Dow [1983] showed how Cohen reals could
be substituted for random reals. The salient ideas in this area are outlined in TaU
[1984]; he and Weiss subsequently worked out the details of dovetailing forcing
with elementary embeddings j that one obtains in the ground model when there
is a supercompact cardinal. These techniques were utilized by Balogh in his
dramatic announcement at the STACY conference in Toronto in 1987:
Theorem 10.6. If supercompact many random reals are added to a model of
set theory, then every locally compact normal space is collectionwise normal in
the forcing extension.
The proof appears in Balogh [1991], along with a proof of the following
parallel of Burke's result:
Theorem 10.7. lf supercompact many random reals are added to a model of set
theory, then every locally compact countably paracompact space is expandable
and scwH.
1206 PETER 1. NYIKOS
Fleissner [1991] gives an alternative proof using the Normal Measure Ex-
tension Axiom which he formulated. It uses measures in the same spirit as the
PMEA uses them to solve NMSC, but the proof is much more lengthy and
intricate.
We begin this final section with an echo of Frank Tall's 1979 question: is then
the normal Moore space problem solved? One response that is there is still a
seemingly minor but theoretically pregnant gap between the consistency strength
of a strongly compact cardinal and the denial of the combinatorial lemmas
that Fleissner used to construct bis nonmetrlzable normal Moore space. There
are two natural ways to try and elose this gap. One is to engineer a major
breakthrough in inner model theory, doing for strongly compact cardinals what
Dodd and Jensen did for measurable ones in conjunction with the Core Model
K. To paraphrase the title of Fleissner [1983], one can try to show that if all
normal Moore spaces are metrlzable, then there is an inner model with a strongly
compact cardinal.
The less ambitious approach is to try to obtain NMSC under weaker condi-
tions, a hope already expressed in Tall [1979]. One inspiration here is the most
unexpected resolution of the consistency strength of the Axiom of Determinacy
(AD) in terms of Woodin cardinals, after the quest for a consistency result
for AD in terms of any large cardinals whatsoever was widely deemed to be
hopeless.
There always remains the possibility that strongly compact cardinals can be
shown to be inconsistent and even that there is actually a ZFC example of a
nonmetrlzable normal Moore space; however, until the gap is elosed, we will
not know whether the first result entails the second. Of course, one can never
rule out the inconsistency of measurable or even much smaller large cardinals
either, but there is a significant difference between them and strongly compact
cardinals, which goes back to the fact that no inner model theory is yet known
for them. The difference is this: combinatorial principles like E(K+) and <>K at
singular cardinals, while they are 'very weak' in the sense that negating them
entails inner models with a proper elass of measurables, are nevertheless quite
effective in giving intricate inductive constructions that seem intractable without
them. See for example the construction of Juhasz, Nagy, and Weiss [1982] of
"arbitrarily large" first countable locally compact countably compact spaces,
whose construction seems unable to jump the hurdle at singular cardinals K of
countable cofinality (where there are no such spaces of that cardinality) without
something like <>K'
The construction in this paper does not really need GCH, but only the axiom
that the poset ([K]Cu, ~) has cofinality K+ for singular cardinals of countable
A HISTORY OF THE NORMAL MOORE SPACE PROBLEM 1207
cofinality. This holds in any model of Covering(V, K) and so the whole con-
struction goes through in such a model. On the other hand, we have no idea
to this day how the construction of locally compact, countably compact, first
countable spaces of arbitrarily large cardinality can be carried out just assuming
ZFC.
It is also noteworthy that showing that these "weak" axioms hold in the
constructible universe L takes much longer than showing that ~ and even
stronger axioms hold in L. This suggests that even the ''weakest'' prlnciples
do not follow from ZFC alone and that it is therefore consistent that there is
a proper class of measurable cardinals. This is quite aseparate issue from the
outright existence of measurable cardinals and other large cardinals, which is one
ofthe main issues discussed by Maddy [1988]. Their actual existence is as much
a philosophical issue as a mathematical one, whereas their consistency can be
expressed unambiguously in the claim that one can never reach a contradiction
by the usual methods of logic if their existence is added to the ZFC axioms.
Any contradiction would have to come at the end of a finite series of rigorous,
formalizable steps, and so the question of whether such a contradiction exists is
one that is, in principle, decidable.
There are also a number of unsolved problems which seem sufficiently close
to the normal Moore space problem to be worth mentioning here.
1. Does 2toto < 2tot1 imply the existence of a nonmetrizable normal Moore
space?
4. Does the metrizability of all normal Moore spaces of cardinality less than
.Jw require large cardinal axioms? is it equiconsistent with the existence of
an uncountable measurable cardinal?
For the next question, we use the following terminology: aspace is
paranormal if every countable discrete collection of closed sets expands to
a 10caUy finite collection. This is a useful common weakening of normality
and countable paracompactness. Burke [1984b] extended Theorem 10.3 to
paranormal spaces but could not quite do it for 10.1:
Next recall that a submetrizable space is one which has a coarser metriz-
able topology. Frank Tall and Steve Watson have an unpublished proof of the
consistency of there being a normal Moore space that is not submetrizable,
using a model due to Shelah [1977b].
6. Are large cardinals needed for the statement, "Every normal Moore space
is submetrizable"?
7. Is the NMSC equivalent to the statement that every first countable normal
space is collectionwise normal?
8. Does V = L imply (a) that all countably paracompact (or all paranor-
mal) locally compact spaces are cwH? (b) that all countably paracompact
(or all paranormal) first countable spaces are scwH? Is either conc1usion
equiconsistent with ZFC?
The following question is c1early inspired by the Reed-Zenor theorem
(7.4):
11. Is CH enough to imply the existence of a locally compact normal space that
is not collectionwise normal?
12. Are large cardinals needed for a model where alilocally compact normal
spaces are collectionwise normal?
The last two are, of course, inspired by Fleissner's early 80's results where
"first countable" replaced "locally compact".
Here are some more specialized problems of a set theoretic nature:
always win the Starbird game, then in rot they ean always win the modified
Starbird game as well?
Acknowledgements
I take this opportunity to thank Howard Beeker, Judith Roitman, Mary Ellen
Rudin, and Frank Tall for their helpful information about historie al events. Any
inaeeuracies or oversights that remain are my responsibility.
References
Alexandroff, P.S. (1960) On the metrization oftopological spaces, Bull. Acad. Pol. Sei.
Sero Math. 8, 135-140 (in Russian).
Arens, R. and Dugundji, J. (1950) Remarks on the concept of compactness, Portugalie
Math. 9, 141-143.
Arhangel'skii, A.V. (1962) On mappings ofmetric spaces, Dokl. Akad. Nauk. SSSR 145,
245-247 =Soviet Math. Dokl. 3,953-956.
Balogh, z. (1991) On collectionwise normality of locally compact, nonnal spaces,
Transactions AMS 323,389-411.
Baumgartner, J.E., Malitz, J.1. and Reinhardt, W. (1970) Embedding trees in the
rationals, Proc. Nat. Acad. Sei. USA 67,1748-1753.
Bing, RH (1951) Metrization oftopological spaces, Canad. J. Math. 3, 175-186.
_ _ (1964) A translation of the nonnal Moore space conjecture, Proceedings AMS,
16, 612-619.
_ _ (1981) Metrization problems, in General Topology and Modem Analysis, L.F.
McAuley and M.M. Rao (eds.), Academic Press, pp. 3-16.
Burke, D.K. (1984a) Covering properties, in Handbook 0/ Set-Theoretic Topology, K.
Kunen and J. Vaughan (eds.), North-Holland, pp. 347-422.
_ _ (1984b) PMEA and first countable, countably paracompact spaces, Proceedings
AMS 92, 455-460.
Chaber, J. and Zenor, P.L. (1977) On perfect subparacompactness and a metrization
theorem for Moore spaces, Topology Proc. no. 2, 2, 401-407.
Devlin, K.J. and Shelah, S. (1978) A weak version of ~ which follows from 2~o < 2~1,
Israel 1. Math. 29, 239-247.
_ _ (1979a) Souslin properties and tree topologies, Proc. London Math. Soc. (3) 39,
237-252.
_ _ (1979b) A note on the nonnal Moore space conjecture, Canad. J. Math. 31,
241-251.
Dow, A. (1983) Remote points in large products, Top. Appl. 16, 11-17.
Fleissner, W.G. (1974) Nonnal Moore spaces in the constructible universe, Proceedings
AMS 46, 294-298.
_ _ (1975) When is Jones' space nonnal? Proceedings AMS 50, 375-378.
1210 PETER J. NYlKOS
Maddy, P. (1988) Believing the axioms 11, l. Symbolic Logic 53, 736-764.
Martin, D.A. (1976) Hilbert's first problem: the continuum hypothesis, in Mathematical
Developments Arisingfrom Hilbert Problems, vol. 28 in the Proceedings of Symposia
in Pure Mathematics, AMS, pp. 81-92.
Martin, D.A. and Solovay, RM. (1970) Internal Cohen extensions, Ann. Math. Logic 2,
143-178.
Michael, E. (1955) Point-finite and locally finite coverings, Canad. J. Math. 7,275-279.
Moore, RL. (1932) Foundations 0/ Point-set Theory, American Mathematical Society,
CoUoquium Publications 13. Revised 1962.
_ _ (1935) A set ofaxioms for plane analysis situs, Fund. Math. 25, 13-28.
Nyikos, P.J. (1980a) A provisional solution to the normal Moore space problem,
Proceedings AMS 78, 429-435.
_ _ (1980b) The normal Moore space problem, Topology Proceedings, no. 2, 3
(1978),473-493.
_ _ (1980c) Some normal Moore spaces, CoU. Math. Soc. lanos Bolyai 23,883-903.
_ _ (1981) Axioms, theorems, and problems related to the Jones lemma, in General
Topology and Modem Analysis, L.F. McAuley and M.M. Rao (eds.), Academic
Press, pp. 441-449.
_ _ (1989) The Cantor tree and the Frechet-Urysohn property, in Papers on general
topology and related category theory and algebra, Ralph Kopperman et al. (eds.),
Ann. New York Acad. Sci. 552, 109-123.
Reed, G.M. and Zenor, P.L. (1976) Metrization of Moore spaces and generalized
manifolds, Fund. Math. 91, 203-209.
Rothberger, F. (1948) On some problems of Hausdorff and of Sierpinski, Fund. Math.
35,29-46.
Rudin, M.E. (1955) Countable paracompactness and Souslin's problem, Canad, l. Math.
7,543-547.
_ _ (1969) Souslin's conjecture, Amer. Math. Monthly, 76, 1113-1119.
_ _ (1975) Lectures on Set Theoretic Topology, CBMS Lecture Notes vol. 23,
American Mathematical Society.
_ _ (1984) Dowker spaces, in Handbook 0/ Set-Theoretic Topology, K. Kunen and
J. Vaughan (eds.), North-Holland, pp. 761-780.
Rudin, M.E. and Starbird, M. (1977a) Some examples ofnormal Moore spaces, Canad,
l. Math. 29,84-92.
Shelah, S. (1977a) Whitehead groups may not be free, even assuming CH, 1., Israel l.
Math. 28, 193-204.
_ _ (1977b) Remarks on A-collectionwise Hausdorff spaces, Topology Proc. 2,583-
592.
_ _ (1982) Proper Forcing, Lecture Notes in Mathematics vol. 940, Springer-Verlag.
_ _ (1999) Proper and Improper Forcing, Springer-Verlag.
Solovay, RM. (1971) Real-valued measurable cardinals, Proc. Symp. Pure Math. 13,
397-428.
Solovay, RM. and Tennenbaum, S. (1971) Iterated Cohen extensions and Souslin's
problem, Ann. Math. 94, 201-245.
Souslin (1920) Probleme 3, Fund. Math. 1,223.
1212 PETER J. NYIKOS
Steen, L.A. and Seebach, J.A. (1978) Counterexamples in Topology (Second edition)
Springer-Verlag.
TaU, F.D. (1972) A set-theoretic proposition implying the metrizability ofnormal Moore
spaces, Proceedings AMS 33,195-198.
_ _ (1974) On the existence of normal metacompact Moore spaces which are not
metrizable, Canad. 1. Math. 26, 1-6.
_ _ (1977) Set-theoretic consistency results and topological theorems concerning the
normal Moore space conjecture and related problems, Ph.D. dissertation, University
ofWisconsin, 1969; Dissertationes Math. (Rozprawy Mat.) 148, 1-53.
_ _ (1979) The normal Moore space problem, in Topological Structures ll, Part 2,
P.C. Baayen and J. van Mill (eds.), Mathematical Centre Tracts, Amsterdam 115,
243-261.
_ _ (1980) Large cardinals for topologists, in Surveys in General Topology, G.M.
Reed (ed.), Academic Press, pp. 445-477.
_ _ (1981) Witnessing normality, in General Topology and Modern Analysis, L.F.
McAuley and M.M. Rao (eds.), Academic Press, pp. 309-315.
_ _ (1984) Normality versus coUectionwise normality, in Handbook of Set-Theoretic
Topology, K. Kunen and J. Vaughan (eds.), North-Holland, pp. 685-732.
Taylor, A.D. (1981) Diamond principles, ideals and the normal Moore space problem,
Canad. 1. Math. 33, 282-296.
Tennenbaum, S. (1968) Souslin's problem, Proc. Nat. Acad. Sei. USA 59, 60-63.
Todorcevic (1984) Trees and linearly ordered sets, in Handbook of Set-Theoretic
Topology, K. Kunen and J. Vaughan (eds.), North-Holland, pp. 235-293.
van Douwen, E. (1984) The integers and topology, in Handbook of Set-Theoretic
Topology, K. Kunen and J. Vaughan (eds.), North-Holland, pp. 111-167.
Wage, M.L. (1976) Countable paracompactness, normality, and Moore spaces, Proceed-
ings AMS 57, 183-188.
Watson, W.S. (1982) Locally compact normal spaces in the constructible universe,
Canad. 1. Math. 34, 1091-1096.
_ _ (1985) Separation in countably paracompact spaces, AMS Transactions 290,
831-842.
Willard, S. (1970) General Topology, Addison-Wesley.
INDEX
Wagner, 978
Wall obstruetion, 1149
WaHaee, 979
WaHman-Shanin-type eompaetifieation, 938
Wang, 845
Ward,974
Wamnabe, 1149, 1157, 1161
Watson, 906, 1205
WaZewski, 1156
weak Lindelf number, 1078
weak proper homotopy eategory, 1150
weakly Cauehy filters, 888
weakly eoneentrated, 880
weakly separable groups, 1071
weighted quasi-pseudometrie, 925
Weil,979
Weil-eomplete, 1052, 1054, 1055
weH-fibred topologieal eonstruet, 993
weH-fibred topologieal eonstruets, 993
weH-monotone quasi-uniformity, 870
weHpowered, 993
West, 1157
History of Topology