Solutions for

Rings of Continuous Functions

by L. Gillman and M. Jerison
version 1.0
Igor Khavkine
June 11, 2011
Preface to v1.0
Gillman and Jerisons Rings of Continuous Functions is considered one of
the most important texts in its eld. A large part of the material contained
in that book is in the form of problems, to be worked out by the reader.
This challenge, though very pedagogical, can be daunting to someone who
is just being introduced to the topics covered in the problems (e.g. advanced
undergraduate or beginning graduate students).
Until now, there has been no comprehensive and readily available collec-
tion of solutions for problems given in that book. Producing such a collection
was the goal of this project. Unfortunately, time constraints have allowed
only the rst four chapters to be completed for the current version.
The intended audience for this work is an advanced undergraduate or
beginning graduate student in mathematics or a related discipline, with an
introductory background in topology, abstract algebra and set theory. In the
words of the authors themselves:
This book is addressed to those who know the meaning of each
word in the title: none is dened in the text. [GJ, Preface]
This work attempts to be slightly more gentle to the reader. It aims to
be fairly self contained, when taken together with Gillman and Jerisons
primary text. Whenever new terminology or a new concept is introduced, it
is clearly dened or concisely explained, giving reference to the primary text
if necessary.
Some results that might be useful for more than one problem are formu-
lated as Lemmas and referenced in solutions whenever necessary. The solu-
tions also heavily rely on solved problems preceding them (usually referenced
as 1A.1), preceding questions from the same problem (usually referenced as
(1.)), as well as results from the primary text (usually referenced as 1.11).
Sometimes dierent questions in the same problem use similar reasoning,
which is explained once and implicitly referred to in the questions that fol-
low. For this reason, it is helpful to read the solution to each problem in its
entirety, even if the reader is only interested in a single question from that
problem.
This document was typeset in L
A
T
E
X with some custom formatting op-
tions. The appearance of this document was meant to follow that of the
primary text as closely as possible. There are subtle dierences in notation,
but those, in the opinion of the author, are self explanatory. The electronic
1
versions of the document was designed with hyperlinking capability in mind
(for example when viewed in DVI or PDF formats). This feature can greatly
facilitate navigation, especially when looking up referenced Lemmas or pre-
vious problems. Those who intend to add to this work are encouraged to
follow the style of presentation that appears in this version.
I would like to thank my supervisor for this project, Dr. Robert Raphael,
1
for his patience and guidance. I would like to thank the people at the McGill
Society of Undergraduate Math Students (SUMS) for providing a pleasant
work environment. I am also grateful to the people from the comp.text.tex
newsgroup for helpful tips and suggestions when L
A
T
E
X was not cooperating.
Last but not least, this work has been made possible by nancial support from
the Natural Sciences and Engineering Research Council of Canada (NSERC).
Distribution and Modication
The primary text contains 16 chapters, of which only four are considered in
this document. Conceivably, this work could evolve into a complete solutions
manual for Rings of Continuous Functions. However this feat may require a
lot of time and contributions of many people. It is the intent of the author
to make it possible to contribute to this work as easily as possible.
Another wish of the author is to make these solutions available to any-
one who is interested in the material for free distribution and use. Thus
these solutions can be used for a course without seeking explicit permission
of the author. Similar freedom is given to those who intend to extend, refor-
mat, or republish this document, with the exception that this Preface stays
unmodied.
For all of the above reasons, this work is placed under the GNU Free
Documentation License (FDL).
Copyright c _ 2002 Igor Khavkine.
Permission is granted to copy, distribute and/or modify this doc-
ument under the terms of the GNU Free Documentation License,
Version 1.1 or any later version published by the Free Software
Foundation; with no Invariant Sections, with no Front-Cover
Texts, and with no Back-Cover Texts. A copy of the license
is included in the section entitled GNU Free Documentation Li-
cense (Appendix A).
1
Department of Mathematics and Statistics, Concordia University, Montreal, Canada.
2
Igor Khavkine, 2002
igor.khavkine@utoronto.ca
3
Contents
1 Functions on a topological space 6
1A p.20 continuity on subsets . . . . . . . . . . . . . . . . . 6
1B p.20 components of X . . . . . . . . . . . . . . . . . . . . . 7
1C p.21 C and C

for various subspaces of R . . . . . . . . 11

1D p.21 divisors of functions . . . . . . . . . . . . . . . . . . 16
1E p.21 units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1F p.22 C-embedding . . . . . . . . . . . . . . . . . . . . . . . . 20
1G p.22 pseudocompact spaces . . . . . . . . . . . . . . . . . 21
1H p.22 basically and extremally disconnected spaces . 24
2 Ideals and z-lters 29
2B p.30 prime ideals . . . . . . . . . . . . . . . . . . . . . . . . 29
2C p.31 functions congruent to constants . . . . . . . . . 30
2D p.31 z-ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2E p.31 z-ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2F p.31 finite spaces . . . . . . . . . . . . . . . . . . . . . . . . 32
2G p.31 prime vs. z-ideals in C(R) . . . . . . . . . . . . . . . 34
2H p.32 the identity function i in C(R) . . . . . . . . . . . 35
2I p.32 C(Q) and C

(Q) . . . . . . . . . . . . . . . . . . . . . . 37
2J p.32 ideal chains in C(R), C(Q) and C(N) . . . . . . . . . 37
2K p.32 z-filters and C

. . . . . . . . . . . . . . . . . . . . . 38
2N p.35 the m-topology on C . . . . . . . . . . . . . . . . . . 38
3 Completely Regular Spaces 44
3B p.48 countable sets . . . . . . . . . . . . . . . . . . . . . . 44
3C p.48 G

-points of a completely regular space . . . . . 45

3D p.48 normal spaces . . . . . . . . . . . . . . . . . . . . . . . 46
3E p.49 nonnormal space . . . . . . . . . . . . . . . . . . . . . 48
3K p.49 the completely regular, nonnormal space . . 49
3L p.51 extension of functions from a discrete set . . . 52
3M p.51 suprema in C(R) . . . . . . . . . . . . . . . . . . . . . 55
3N p.51 the lattice C(X) . . . . . . . . . . . . . . . . . . . . . 55
3O p.52 totally ordered spaces . . . . . . . . . . . . . . . . 58
3P p.53 convergence of z-filters . . . . . . . . . . . . . . . 62
4
4 Fixed ideals. Compact space 64
4A p.60 maximal ideals; z-ideals . . . . . . . . . . . . . . . . 64
4B p.60 principal maximal ideals . . . . . . . . . . . . . . . . 66
4C p.61 finitely generated ideals . . . . . . . . . . . . . . . 67
4D p.61 functions with compact support . . . . . . . . . . 68
4E p.61 free ideals . . . . . . . . . . . . . . . . . . . . . . . . . 70
4F p.61 z-ultrafilters on R that contain no small sets . 72
4G p.61 base for a free ultrafilter . . . . . . . . . . . . . 73
4H p.62 the mapping
#
. . . . . . . . . . . . . . . . . . . . . . 74
4I p.62 the ideals O
p
. . . . . . . . . . . . . . . . . . . . . . . 75
4J p.62 P-spaces . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4K p.63 further properties of P-spaces . . . . . . . . . . . 81
4L p.63 P-points . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4M p.64 the space . . . . . . . . . . . . . . . . . . . . . . . . 85
4N p.64 a nondiscrete P-space . . . . . . . . . . . . . . . . . 90
References 91
A GNU Free Documentation License 92
5
1 Functions on a topological space
1A p.20 continuity on subsets
Let f R
X
.
Recall the following equivalent denitions of continuity. Given two topo-
logical spaces X and Y and a function f: X Y , f is said to be continuous if
for every open set A Y , f

[A] X is open. Equivalently, f is continuous

if for every closed set A Y , f

[A] X is closed. For future reference, the

set of all continuous functions from X to R is denoted by C(X). It is also
worth noting that below, when appropriate, subsets of X are endowed with
the subspace topology.
1. Let X =

n
i=1
A
i
where each A
i
is closed in X. Suppose that for each
i, f[A
i
is continuous on A
i
. Show that f is continuous on X.
Proof. Let B

be a closed subset of R, B = f

[B

] and B
i
= (f[A
i
)

[B

] for
each i. By continuity B
i
is closed in A
i
, and since A
i
is closed in X, B
i
is
also closed in X. Note that for each i, B
i
= B A
i
, and since A
i
cover X,
B =

n
i=1
B
i
. As a nite union of closed sets B must also be closed. Hence
f is continuous on X.
2. Let X =

iI
A
i
where each A
i
is open in X. Suppose that for each
i, f[A
i
is continuous on A
i
. Show that f is continuous on X.
Proof. Let B

be an open subset of R, B = f

[B

] and B
i
= (f[A
i
)

[B

]
for each i. By continuity B
i
is open in A
i
, and since A
i
is open in X, B
i
is also open in X. Note that for each i, B
i
= B A
i
, and since A
i
cover
X, B =

iI
B
i
. As a union of open sets B must also be open. Hence f is
continuous on X.
3. Let T be a family of closed subsets of X with X =

T and such that

every x X has an open neighborhood N
x
that meets only nitely many
members of T . (T is said to be locally nite.) Given that for each T T ,
f[T is continuous on T, show that f is continuous on X.
Proof. For a given x X, let N
x
be an open neighborhood of x which meets
only the T
i

n
i=1
elements of T . Now for each i, let S
i
= N
x
T
i
. Since for
each i, T
i
is closed in X, S
i
is also closed in N
x
. Also, since for each i, f[T
i
is
6
continuous on T
i
, then f[S
i
is continuous on S
i
. Now since T forms a cover
for X, whichever elements of T that N
x
meets must form a cover for N
x
. So
we can write N
x

n
i=1
T
i
and hence N
x
=

n
i=1
S
i
. Now we can apply (1.)
to get that f[N
x
is continuous on N
x
. Obviously X =

xX
N
x
, and nally
we use (2.) to get that f is continuous on X.
1B p.20 components of X
First recall a few denitions. Given a topological space X, it is said to
be disconnected if it is possible to write X = A B where A and B are
non-empty disjoint open sets. Equivalently, X is said to be disconnected if
X = A B where A and B are non-empty disjoint closed sets. A set is
said to be connected if it is not disconnected. It follows that any topological
space can be written uniquely as X =

iI
X
i
where the X
i
are non-empty
disjoint maximal (under the partial order) connected subsets, such subsets
are called connected components.
For the following problem we will nd the following Lemmas useful.
Lemma 1.1. Let X and Y be topological spaces, and f: X Y a continuous
function. If X is connected then then so is f[X].
Proof. Suppose that X is connected but f[X] is not. Then we can write
f[X] = A B, where non-empty A and B are open in f[X] and disjoint.
Since f: X Y is continuous, then so is f: X f[X]. By continuity we
have non-empty f

[A] and f

[B] open in X and also disjoint since A and B

are. But then
X = f

[f[X]] = f

[A B] = f

[A] f

[B]
But that implies that X is disconnected. This contradiction completes the
proof.
Lemma 1.2 (Intermediate Value Theorem). Let X be a connected topological
space and f C(X). Then given x
1
, x
2
X such that f(x
1
) = a
1
and
f(x
2
) = a
2
, where WLOG (without loss of generality) a
1
a
2
, then
(a [a
1
, a
2
]) (x X) (f(x) = a).
Proof. Suppose (a [a
1
, a
2
]) such that (x X) (f(x) ,= a). Then clearly
a , f[X], and f[X] ], a[ ]a, +[. We know that a
1
], a[ since
7
a
1
a and a
2
]a, +[ since a a
1
, then both of D
1
= f[X] ], a[ and
D
2
= f[X] ]a, +[ are non-empty, disjoint and open in f[X]. As well as
f[X] = D
1
D
2
, so f[X] is disconnected. Hence, by the above Lemma, X
cannot be connected. This contradiction completes the proof.
Lemma 1.3. Let X be a connected topological space, and u C(X) be
positive. It is given that there exists a square root s C(X) such that
s
2
= u,
(x X)
_
s(x) =
_
u(x) s(x) =
_
u(x)
_
.
Where
_
u(x) denotes the unique positive root of the real number u(x). Then
(x X)(s(x) < 0) (x X)(s(x) > 0).
Proof. First it is clear that if s(x) satises the given condition at every x X
then s
2
= u. Now suppose that there are x
1
, x
2
X such that s(x
1
) < 0
and s(x
2
) > 0. By Lemma 1.2, s

(0) is non-empty, but for all x X,

[s(x)[ =
_
u(x) > 0. This contradiction completes the proof.
Lemma 1.4. Let X =

iI
X
i
, where the X
i
are disjoint and and open.
Given f
i
C(X
i
) for each i I, there exists a unique function f C(X)
such that f[X
i
= f
i
for each i I.
Proof. Uniqueness Given two such functions f, g C(X), by hypothesis,
they agree on each X
i
. Since X
i
is a cover for X, f and g agree on X, or
simply f = g.
Existence Let f R
X
be dened such that f[X
i
= f
i
for each i I, since
the X
i
are disjoint and cover X, f is well dened. Note that by continuity,
each A
i
is open, hence continuity of f follows from 1A.2.
1. Show that in C(X), all positive units have the same number of square
roots.
Proof. Let u, v C(X) be any two positive units. We want to establish a
bijection between the square roots of u and v. Let w = uv, w is also a positive
unit. Hence there must exist a unique positive square root r C(X), such
that r
2
= w. Let s C(X) be any square root of u. We can map s to a unique
element t C(X), s t = r/s. Note that t
2
= r
2
/s
2
= w/u = uv/u = v,
hence t is a square root of v. This mapping is injective since the inverse exists
and is obviously t r/t = s. It is also surjective because for any t
2
= v,
8
s = r/t is such that s
2
= u and s t. Hence the mapping is a bijection, and
this establishes that the number of square roots for any pair of, and hence
all, positive units in C(X) is the same.
2. Show that X is connected i 1 has exactly two square roots.
Proof. Necessity Suppose s
2
= 1, then (x X)(s(x) = 1s(x) = 1). If X
is connected, then according to Lemma 1.3, there are only two possibilities.
Either s(x) = 1 for all x X, or s(x) = 1 for all x X, both of which are
in fact square roots of 1. Hence 1 has exactly two square roots.
Suciency Suppose that 1 has exactly two square roots, but X is disjoint.
Then we can write X = A B where A and B are non-empty, disjoint and
open in X. But then, by Lemma 1.4, we can dene at least four functions h
on X of the form h[A = 1 and h[B = 1, each of which is a square root of
1 on X. Hence 1 cannot have exactly two square roots. This contradiction
completes the proof.
3. Show that for nite m, X has m connected components i 1 has exactly
2
m
roots. Also, show that this statement is false if m is innite.
Proof. Necessity Assume that X has m connected components. Since m is
nite each component is open. If we have s C(X) such that s
2
= 1, then
s(x) = 1 for each x X. By Lemma 1.3, s must be either positive or
negative on each component of X, so we can have at most 2
m
choices for s.
By Lemma 1.4, we have that each one of those choices is in fact in C(X),
hence 1 has exactly 2
m
roots.
Suciency Assume that 1 has exactly 2
m
roots. If X has n connected
components where n is nite. Then from the above, 1 has exactly 2
n
roots.
From the hypothesis 2
n
= 2
m
, which implies that n = m.
Conterexample. Now suppose that m is innite, we take for example
X = 1, 1/2, . . . , 1/n, . . . , 0
where m =
0
. Suppose s X and s
2
= 1. Then there are two choices
s(0) = 1 or s(0) = 1. For each choice of s(0), by continuity there must
exist an N N such that for each n > N, s(1/n) = s(0). If we x s(0)
and n, then there are 2
n1
possible choices for s, and all of the possibilities
9
for n

< n are already counted in that number. Hence the total number of
square roots of 1 is
2 sup
n>N
2
n1
= 2
0
=
0
.
But
0
< 2

0
= 2
m
(by Cantors theorem), this completes the proof.
4. Show that X is connected i 0 and 1 are the only idempotents in
C(X). (u C(X) is idempotent if u
2
= u).
Proof. Necessity If u C(X) is idempotent, then for each x X, u(x) = 0
or u(x) = 1. Assume that X is connected. Since u is idempotent it is its own
square root, so by Lemma 1.3, if for some x X, u(x) = 1, then u = 1. On
the other hand if u is idempotent, but for all x X, u(x) ,= 1, then u = 0.
Hence the only idempotents in C(X) are 0 and 1.
Suciency Assume that 0 and 1 are the only idempotents in C(X), but
X is disconnected. Then we can write X = A B, where A and B are dis-
joint, non-empty and open. Then we can construct at least four idempotent
functions of the form u[A = 0 or 1 and u[B = 0 or 1, all of which are in
C(X) by Lemma 1.4. This contradiction completes the proof.
5. If X is connected, then C(X) cannot be written as C(X)

= R S,
were R and S are non-trivial rings.
Proof. Suppose that C(X)

= R S where R and S are non-trivial rings.
Then we can write 1 = 1
R
1
S
. But now, we can construct at least four
square roots of 1 of the form 1
R
1
S
. Hence, by 1B.2, X cannot be
connected. This contradiction completes the proof.
6. If we can write X = A B, where A and B are disjoint, non-empty
and open, then C(X)

= C(A) C(B).
Proof. Any f C(X) can be written as f = f
A
+ f
B
, where f
A
[A = f[A
and f
B
[B = f[B, while f
A
[B = 0[B and f
B
[A = 0[A (f
A
, f
B
C(X) by
Lemma 1.4). Hence there is a surjective map C(A) C(B) C(X) of
the form f
A
[A f
B
[B f
A
+ f
B
. On the other hand this map has an
inverse given by C(X) C(A) C(B) of the form f f[A f[B, so it
is also injective and hence bijective. It is trivial to check that this map is a
homomorphism, and since it is bijective, it is also an isomorphism. Hence
C(X)

= C(A) C(B).
10
1C p.21 C and C

for various subspaces of R

Consider the subspaces R, Q, N and N

= 1, 1/2, . . . , 1/n, . . . , 0 of R, and the

rings C and C

for each of these spaces. Each of these rings is of cardinality

c = 2

0
.
For the following problem, we will nd the following Lemmas useful.
Lemma 1.5. Let X and Y be topological spaces. If X is homeomorphic to
Y , then C(X)

= C(Y ).
Proof. Let f C(X), we construct a homomorphism C(X) C(Y ), of the
form f f where : Y X is a homeomorphism. This homomorphism
has an inverse of the form f (f )

= f, hence it is injective.
Also, for any g C(Y ), we can write g = (g

) , where g

C(X),
hence the homomorphism is surjective. So the homomorphism is bijective,
and hence it is an isomorphism or C(X)

= C(Y ).
1. For each m
0
, each ring on R, N or N

contains a function having

exactly 2
m
square roots. If a member of C(Q) has more than one square root,
it has exactly c of them.
Proof. Case R: Consider the function f C

(R) dened by
f(x) =
_
[ sin(x)[ if x [0, m]
0 if x , [0, m]
If m =
0
, then we replace [0, m] by [0, [. Let A
i
= ](i 1), i[ for
i N and i m, and A =

im
A
i
.
Case N: First recall that all functions from N to R are continuous. Now
consider f C

(N) dened by
f(x) =
_
1 if 1 x m
0 if x > m
Let A
i
= i for i N and i m, and A =

im
A
i
.
Case N

: Consider the function f C

(N

) dened by
f(1/n) =
_
1/n if 1 n m
0 if n > m or x = 0
f(0) = 0
11
Let A
i
= 1/i for i N and i m, and A =

im
A
i
.
For all three cases, consider f[A, by Lemmas 1.3 and 1.4, f[A has exactly
2
m
continuous square roots since the A
i
are disjoint and open (both for
nite and innite m). For the rst case, the square root of f[A can be
continuously extended to the whole of R since the square root of f must
identically vanish on R A. Since C

(R) C(R), the same f works in C(R)

as well.
Case Q: Consider any non-negative f C(Q). If f = 0, then its square
root is unique and is itself. If f ,= 0, then consider the set A = pos f. A is
open, and it is a basic topological property of Q (and R) is that any open set
can be written as A =

i=1
A
i
, where the A
i
are disjoint open (bounded)
intervals.
Pick a nonempty A
i
= ]a, b[, where a and b are real numbers. Now we
construct monotone decreasing sequence c
n
contained in ]a, b[. Let c
1
= b,
then we can always nd an irrational number c
2
such that a < c
2
< c
1
. The
sequence c
n
is constructed continuing by induction.
Now we dene the disjoint open sets C
n
, where C
n
= ]c
n+1
, c
n
[, and
C
0
= ]a, inf
n
c
n
[. Note that if C =

n=0
C
n
, then C Q = A
i
Q, so C
and A
i
denote the same open set in Q. By Lemmas 1.3 and 1.4, there are
exactly 2

0
= c possible square roots for f[C. Each of those square roots
can be continuously extended to QC, e.g. with the unique non-negative
root of f[(QC). Hence the number of square roots of any non-negative non-
zero f C(Q) is bounded below by c, it is also bounded above by c (the
cardinality of C(Q)), hence it is equal to c. So any non-negative f C(Q)
has either one or c square roots.
2. C(R) has just two idempotents, C(N

) has exactly
0
, and C(Q) and
C(N) have exactly c.
Proof. Case C(R): R is connected, hence by 1B.4 its only idempotents are
0 and 1.
Case C(N

): If s C(X) is idempotent, then for each x X, s(x) = 0 or

s(x) = 1. To show that C(N

) has exactly
0
idempotents repeat the
argument in the second part of 1B.3, except replace 1, 1 by 0, 1 as
possible values of s(x).
12
Case C(N) and C(Q): We can write both N and Q as a union of
0
disjoint
open sets. In particular N =

nN
n, and
Q = Q
_
]

2,

2[
_
nN
_
](n + 1)

2, n

2[ ]n

2, (n + 1)

2[
_
_
.
Hence by setting s to either 0 or 1 on each of these open sets, s is
idempotent and s C(N) (resp. s C(Q)), by Lemma 1.4. Hence
the number of idempotents in each of the rings in question is at least
2

0
= c. And because of the cardinality of the rings, this number is
also bounded by c from above. Hence the total number of idempotents
in C(N) and C(Q) is c.
3. Every non-zero idempotent in C(Q) is a sum of two non-zero idem-
potents. In C(N), and in C(N

), some, but not all idempotents, have this

property.
Proof. Case C(Q): Let u C(Q) be a non-zero idempotent, hence pos u is
open and non-empty. Hence pos u can be written as a countable union
of disjoint non-empty open intervals. Let A = ]a, b[ be one of these
intervals, where a and b are real numbers. Note that we can always
nd an irrational number c such that a < c < b. Then we dene
A
1
= ]a, c[ and A
2
= ]c, b[, so in the topology on Q, the sets A and
A
1
A
2
are the same open set. Now we dene u
1
, u
2
R
Q
such that
u
1
[], c[ = u[], c[ and u
1
[]c, [ = 0, and u
2
[]c, [ = u[]c, [ and
u
2
[], c[ = 0. Clearly u = u
1
+ u
2
, both u
1
and u
2
are non-zero,
idempotent and u
1
, u
2
C(Q), by Lemma 1.4.
Case C(N) and C(N

): Consider the following functions u C(N) and

v C(N

)
u(x) =
_
1 if x = 1
0 if x > 1
v(x) =
_
1 if x = 1
0 if x < 1
Clearly, both u and v are idempotent and cannot be written as a sum
of two non-zero idempotents.
13
Now consider the following functions u C(N) and v C(N

)
u(x) =
_
1 if x = 1 or x = 2
0 if x > 2
v(x) =
_
1 if x = 1 or x = 1/2
0 if x < 1/2
On the other hand, these idempotents can be written very simply as a
sum of two non-zero idempotents.
4. Except for the obvious identity C(N

) = C

(N

), no two of the rings

in question are isomorphic.
Proof. C

is not isomorphic to C: First note that no C

is not isomor-
phic to the corresponding C (except the afore mentioned case), since
by 1.7 bounded functions are always taken to bounded functions by
homomorphisms.
C(R) C(Q, N, N

) (resp. C

): Also, note that neither of C(R) or C

(R)
is isomorphic to any of the other rings in question since f[R] is con-
nected for any f C(R), because R is connected, but any one of the
other rings has a function with a disconnected image.
C, (N

) C(Q, N) (resp. C

): Recall from 1C.2 that C and C

on N

have
exactly
0
idempotents each, while C and C

on N and Q have exactly

c idempotents each. So since the number of idempotents is preserved
under isomorphism, the respective rings cannot be isomorphic.
C(N) C(Q) (resp. C

): Now recall from 1C.3 that every non-zero idem-

potent in C of C

on Q can be written as a sum of two non-zero

idempotents from the same ring, while the same is not true for either
of C or C

on N. This property is also preserved under isomorphism,

hence the respective rings cannot be isomorphic. This completes the
proof.
14
5. Each of C(Q) and C(N) is isomorphic with a direct sum of two copies
of itself. C(N

) is isomorphic with a direct sum of two subrings, just one of

which is isomorphic with C(N

).
Proof. Case C(N): Let N
even
= 2N and N
odd
= 1 + 2N. N is naturally
homeomorphic to both N
even
and N
odd
, hence, by Lemma 1.5, the re-
spective rings are isomorphic. Finally by 1B.6, C(N)

= C(N
even
)
C(N
odd
).
Case C(Q): Let Q

= ],

2[ Q and Q
+
= ]

2, [ Q. We construct
homeomorphisms h

: Q Q

. They are dened as follows

h

: q q if q 0
h
+
: q q + 2 if q 0
Now suppose a
n
is a monotone increasing sequence of rationals con-
verging to

2 with a
1
= 0, and b
n
is a monotone decreasing sequence
converging to

2 with b
1
= 2. Then for each n N let
h

: q (q (n 1))(a
n+1
a
n
) + a
n
if n 1 < q n
h
+
: q ((n 1) q)(b
n+1
b
n
) + b
n
if n q < n + 1
Since h

are homeomorphisms, then, by Lemma 1.5, the respective

rings are isomorphic. Finally by 1B.6, C(Q)

= C(Q

) C(Q
+
).
Case C(N

): We can write N

= A B, where A = 1, 1/2, . . . , 1/n and

B = 1/(n + 1), . . . , 0 are open, non-empty and disjoint. There is
a natural homeomorphism h: N

B, given by h: x 1/(x
1
+ n),
hence, by Lemma 1.5, the rings C(N

) and C(B) are isomorphic. Fi-

nally by 1B.6, C(N

)

= C(A) C(B).
6. The ring C(R) is isomorphic with a proper subring. But C(R) has no
proper summand.
Proof. Consider the subring R C(R) consisting of all functions constant
on the interval [0, 1] (the fact that R is actually a subring is easy to see). We
can construct a homomorphism h: C(R) R dened by h: f g, where
g(x) =
_
_
_
f(x) if x < 0
f(0) if 0 x 1
f(x 1) if 1 < x
15
This homomorphism is injective since it has an inverse given by h

: g f,
where
f(x) =
_
g(x) if x 0
g(x + 1) if 0 < x
Since the inverse exists for every g R, then h is also surjective. So h
is bijective, and hence an isomorphism. The fact that C(R) has no proper
summand follows directly from the fact that R is connected and 1B.5.
1D p.21 divisors of functions
1. If Z(f) is a neighborhood of Z(g), then f is a multiple of gthat is,
f = hg for some h C. Furthermore, if X int Z(f) is compact, then h can
be chosen to be bounded.
Proof. Let Z

= XZ(f), and h R
X
, dened by
h(x) =
_
0 if x Z(f)
f(x)/g(x) if x Z

Note that f, g are both continuous on cl Z

, and [g[ > 0 on cl Z

(since Z(g)
int Z(f) = X cl Z

), hence h is continuous on cl Z

. It is also continuous
on Z(f) (because it is constant). Now we can write X = Z(f) cl Z

as a
union of two closed sets, on each of which h is continuous. Hence, by 1A.1,
h C(X) and f = hg. Note that X int Z(f) = cl Z

, if it is compact then
h has to be bounded on cl Z

, and hence bounded everywhere on X.

2. Construct an example in which Z(g) Z(g), but f is not a multiple
of g. Consider X = R and f, g C(R) dened by
Example.
f(x) =
_
_
_
x if x < 0
0 if 0 x 1
x 1 if 1 < x
g(x) =
_
_
_
x
2
if x < 0
0 if 0 x 1
(x 1)
2
if 1 < x
Then Z(f) = Z(g) = [0, 1]. Suppose there is an h C(R) such that f = hg.
Then h is uniquely dened on A = R[0, 1]. However h is then unbounded on
16
every neighborhood of [0, 1] in A, and hence cannot be continuously extended
to all of R. Therefore f is not a multiple of g.
3. If [f[ [g[
r
for some real r > 1, then f is a multiple of g. Hence if
[f[ [g[, then f
r
is a multiple of g for every r > 1 for which f
r
is dened.
For this problem it is worth recalling the denition of continuity at a
point. Let X and Y be a topological spaces. Let f Y
X
. The function f
is said to be continuous at x X if for any neighborhood U Y such that
f(x) U, there is an neighborhood V X such that x V and f[V ] U.
We also introduce a piece of notation. Let S be a subset of a topological
space X, then the boundary of S is cl A cl(XS). For convenience the
boundary of S is denoted by S. Equivalently S = (cl S) int S. It is also
worth noting that cl S = S S.
Proof. Let Z

= XZ(g) (from the hypothesis we also have Z(g) Z(f)),

and h R
X
, dened by
h(x) =
_
0 if x Z(g)
f(x)/g(x) if x Z

Clearly f = gh. Since f and g are continuous and [g[ > 0 on Z

, we know
that h is continuous on Z

. It is also clear that h is continuous on Z(g). So

since X = Z(g) cl Z

, it is sucient to show that h is continuous on cl Z

and use 1A.1.

Suppose that x Z

, then h(x) = 0, f(x) = 0 and g(x) = 0. Let > 0,

U = ], [ and V = ]
1/(r1)
,
1/(r1)
[. Now since g is continuous, we can
nd an open subset W X such that x W and g[W] V . But, by the
hypothesis, for any y W Z

we have
[h(y)[ = [f(y)/g(y)[ [g(y)[
r1
< ,
and h(y) = 0 for each y W Z

. Since W is open, the set W cl Z

is open in cl Z

and by the above h[W cl Z

] U, which implies that h is

continuous in cl Z

at every x Z

. Finally, this last fact implies that h is

continuous on cl Z

and completes the proof.

If [f[ [g[, the for each r > 1 we have [f[
r
[g[
r
or [f
r
[ [g[
r
(provided
that f
r
is dened). By the above argument f
r
must be a multiple of g.
17
1E p.21 units
1. Let f C. There exists a positive unit u of C such that
(1 f) 1 = uf.
Proof. Let A = x X [ f(x) 1 and B = x X [ f(x) 1. Let
u R
X
, dened by
u(x) =
_
[f(x)[
1
if x A B
1 if x , A B
.
Note that u satises the hypothesis. For u to be continuous, it must be
continuous on each of A, B, cl(X(AB)) (since they form a closed cover of
X). It is clear that u is continuous on both A and B, it is also continuous
on cl(X(A B)) since f[(A B)] = 1, which matches with the value
of u on the boundary of cl(X(A B)). Hence u is continuous on X, by
1A.1.
2. TFAE (the following are equivalent)
(1) For every f C, there exists a unit u of C such that f = u[f[.
(2) For every g C

, there exists a unit v of C

such that g = v[g[.

Proof. First, assume (1). Note that for f C(X) and unit u of C(X),
we have f = u[f[ i [u(x)[ = 1 for each x such that f(x) ,= 0 (in fact
u[(pos f) = 1 and u[(neg f) = 1), and u(x) can take on any value (as long
as u is continuous) for each x where f(x) = 0. If u = 1 or u = 1 then u is
already a unit of C

. If u takes on both values 1 and 1, and since u cannot

take on the value 0 because u is a unit, we can write X = pos u neg u and
pos u and neg u are disconnected, by the contrapositive of the Intermediate
Value Theorem (Lemma 1.2). Therefore we can dene a unit v of C

by
v[(pos u) = 1 and v[(neg u) = 1, and we still have f = v[f[. Hence, (1)
implies (2).
Now, assume (2). Let f C, then there is g C

such that pos g = pos g

and neg g = neg f (for example g = (1f)1). Then there is a unit v C

such that g = v[g[. Then the same unit v works for f as well, since f and g
share the same pos and neg sets, that is f = v[f[. Hence (2) implies (1).
18
3. Describe the functions f in C(N) for which there exists a unit u of
C(N) satisfying f = u[f[. Do the same for C(Q) and C(R).
Description. Case C(N): Let f C(N) and u R
N
. Let u be dened as
u(x) =
_
1 if x pos f Z(f)
1 if x neg f
.
Then u(x) satises the hypothesis, and since R
N
= C(N), we are done.
Case C(Q): Let f C(Q) and u R
Q
, dened by
u(x) =
_
1 if x pos f
1 if x neg f
.
As long as u can be continuously extended to the rest of Q and still
remain a unit, we are done. By 1.15, u can be continuously extended
to all of Q i pos u = pos f and neg u = neg f are completely separated
(have disjoint zero-set neighborhoods). Since Q is a metric space, by
1.10, every closed set is a zero set, the above condition is equivalent to
pos f and neg f having disjoint closures (in fact this is also sucient
for u to be a unit, since we can always make u assume the value 0 only
at irrational numbers).
Case C(R): Let f C(R) and u C(R), where u satises the hypothesis.
If there are x, y R such that u(x) = 1 and u(y) = 1, by the
Intermediate Value Theorem (Lemma 1.2), for some x < z < y we
have u(z) = 0, then u cannot be a unit. Hence the condition on f is
either f 0 or f 0.
4. Do the same for the equation f = k[f[, where k belongs to C but is
not necessarily a unit.
Description. Case C(N): The argument from (3.) is still applicable.
Case C(Q): The argument from (3.) is still applicable.
Case C(R): In this case k may assume the value 0, so since R is a metric
space, the same conditions as for Q apply to R as well.
19
1F p.22 C-embedding
1. Every C

-embedded zero-set is C-embedded.

Proof. Let Z Z(X) be a C

-embedded zero-set in X and f C

(Z), then
there is a function g C

(X) such that g[Z = f. By 1.18, Z is C-embedded

i Z is completely separated from every zero-set disjoint from it. But, by
1.15 two disjoint zero-sets are completely separated.
2. Let S X; if every zero-set in S is a zero-set in X, then S is C

-
embedded in X.
Proof. Consider any two completely separated sets in S. They have disjoint
zero-set neighborhoods in S, which are also zero-sets in X. Hence these two
sets are completely separated in X as well. So by Urysohns Theorem (1.17),
the set S is C

-embedded in X.
3. A discrete zero-set is C

-embedded i all of its subsets are zero-sets.

Proof. Necessity Suppose that a zero-set Z Z(X) is discrete and C

-
embedded. Now consider a subset Y Z. Since Z is discrete, Y is also
discrete. So we can dene a continuous function f C

(Z) as
f(x) =
_
0 if x Y
1 if x , Y
Since Z is C

embedded, there is a function f

(X) such that f

[Z = f.
Also, since Z is a zero set, there is a function g C

(X) such that Z = Z(g).

Then we can dene another function h C

(X) as h = (g
2
+f
2
). Note that
Z(h) = Z Z(f

) = Y . Hence every subset of Z is also a zero set in X.

Suciency Suppose that every subset Y Z is also a zero-set in X. Any
subset of a discrete space is a zero-set. So any two disjoint subsets of Z are
completely separated (they are their own zero-set neighborhoods). Also any
two disjoint subsets of Z are zero sets in X. But that just means that they
are completely separated in X. Hence, by Urysohns Theorem (1.17), Z is
C

-embedded.
20
4. A subset S of R is C-embedded [resp. C

-embedded] i it is closed.
Proof. Necessity Assume that the set S R is C-embedded in R. Now
suppose that S is not closed, then cl SS is non-empty. Take x cl SS,
then there is a sequence x
n
S such that lim
n
x
n
= x (WLOG assume
that x
n
is monotone and increasing, a similar argument can be repeated for
a monotone decreasing sequence). We dene a function f C(Rx) as
follows
f(y) =
_

_
1 if y x
1
(1)
n
+ 2
(y x
n
)
(x
n+1
x
n
)
if x
n
y < x
n+1
0 if y > x
.
Since f is continuous and S Rx, then f[S is also continuous. However,
the function f cannot be continuously extended to all of R since no value of
f at x will make f continuous on R. For example lim
n
f(x
2n1
) = 1
and lim
n
f(x
2n
) = 1. This contradiction implies that S must be closed.
Suciency Suppose that S is closed in R. Then i[S is a homeomorphism
which carries S to a closed set in R (that is S itself). So, by 1.19, S is
C-embedded in R.
Since f is bounded, the exact same arguments apply to C

.
5. If a (non-empty) subset S of X is C-embedded in X, then C(S) is a
homomorphic image of C(X). The corresponding result holds for C

.
Proof. Consider the map : C(X) C(S) dened by : f f[S. Clearly,
the mapping is a homomorphism. Since S is C-embedded in X, then for
each function f C(S) there is a function g C(X) such that g[S = f.
Hence is onto. This completes the proof.
1G p.22 pseudocompact spaces
First we recall a couple of denition. Let X be a topological space. X is
said to be Hausdor if any pair of distinct points x, y X there is a pair
of disjoint open sets one containing only x and the other only y. Also, a
topological space X is said to be pseudocompact if C(X) = C

(X).
21
1. Any continuous image of a pseudocompact space is pseudocompact.
Proof. Let X and Y be topological spaces, of which X is pseudocompact.
Let the function f Y
X
be continuous. Now consider W = f[X] and let
g C(W). Suppose that g is unbounded on W, then h = g f is also
unbounded. But then h C(X) but h , C

(X), hence C(X) ,= C

(X) and
X cannot be pseudocompact. This contradiction completes the proof.
2. X is pseudocompact i f[X] is compact for every f in C

(X).
Proof. Necessity Suppose that X is pseudocompact. By the Heine-Borel the-
orem, the set f[X] R is compact i it is closed and bounded. Boundedness
of f[X] implies that f C

(X). Now suppose that f[X] is not closed. Then

it is possible to construct an unbounded function g: f[X] R (see argu-
ment for 1.F4). But then g f C(X) and is unbounded, so X cannot be
pseudocompact. This contradiction completes the proof.
Suciency Suppose that for each f C

(X), the set f[X] R is com-

pact. Now suppose that X is not pseudocompact, then there is an unbounded
function g C(X). But then tan
1
g C

(X), and f[X] is not closed (and

hence not compact). This contradiction completes the proof.
3. Let X be Hausdor space. If, of any two disjoint closed sets, at least
one is compact, or even countably compact, then X is countably compact.
For this problem, we need the following topological fact: a Hausdor
space is countably compact i every innite every set has a limit point.
Proof. Suppose that X is not countably compact. Then there is an innite
set A with no limit point. Take any point x X outside of A, then since A
has no limit point, there is an open neighborhood N of x such that NA = .
Hence x , cl A, which implies that A = cl A or that A is closed. Since A
is innite, it must have a countable subset x
n
. Consider the following
subsets y
k
= x
2k
and z
k
= x
2k1
, both of which are closed by the same
argument as above. Hence y
k
and z
k
are disjoint closed sets, so by the
hypothesis, at least one of them must be compact. But since neither y
k
nor
z
k
have limit points, each of their points must have an open neighborhood
disjoint from all other points in the same set, hence they must be discrete.
But discrete innite sets are not compact. This contradiction completes the
proof.
22
4. If, of any two disjoint zero-sets in X, at least one is compact, or even
pseudocompact, then X is pseudocompact.
It is fruitful to recall the following denition and topological fact. Let
X be a topological space and (Y, d) a metric space. Given a sequence of
functions f
n
Y
X
and a function f Y
X
, the sequence is said to converge
uniformly to f if for each compact subset K X and for each > 0 there
is an N N such that d(f(x), f
n
(x)) < for each n > N and x X.
Also recall that if f
n
are continuous functions from X to Y with f their
pointwise limit, then f is continuous i f
n
converges uniformly to f.
Proof. Suppose that X is not pseudocompact, then there exists an unbound-
ed continuous function f on X (WLOG suppose that f is non-negative). We
construct a sequence of real numbers a
n
by induction:
a
1
= inf ([1, [ f[X])
a
n+1
= inf ([2a
n
+ 1, [ f[X]) for n > 1.
Note that this sequence was chosen so that the subsequences b
k
= a
1
a
2
+
a
2(k1)
+ a
2k1
and c
k
= a
1
a
2
+ + a
2k1
a
2k
are monotone,
respectively increasing and decreasing, and unbounded. Since f is unbounded
and non-negative, the sequence a
n
is guaranteed to exist and tend to
as n . Now we construct a sequence of continuous functions g
n
by
induction:
g
0
= f
g
1
= (a
1
f) (a
1
f a
1
)
g
2k
= (a
2k
g
2k1
) (a
2k
g
2k1
+a
2k
)
g
2k+1
= (a
2k+1
g
2k
) (a
2k+1
g
2k
a
2k+1
)
[Try to perform this procedure for f = 1/x on ]0, [ to see what it does].
Suppose K X is compact, then for each n, the restricted function f[K
is continuous and hence must be bounded. Since f[K is bounded, then for
N N such that a
N
> sup f[K, the tail sequence g
n
[K
n>N
is constant.
Which implies that g
n
[K converges uniformly on each compact K X,
since each g
n
is continuous. The last statement implies that the function
g(x) = lim
n
g
n
(x) is also continuous.
Finally we take two disjoint zero-sets A = x X [ g(x) 1 and
B = x X [ g(x) 1, which are non-empty by construction. By the
23
hypothesis, at least one of A or B must be pseudocompact, but both of
them have unbounded functions dened on them, namely g[A and g[B. This
contradiction completes the proof.
1H p.22 basically and extremally disconnected
spaces
A space X is said to be extremally disconnected if every open set has an open
closure; X is basically disconnected if every cozero-set has an open closure.
1. X is extremally disconnected i every pair of disjoint open sets have
disjoint closures. What is the analogous condition for basically disconnected
spaces?
Proof. Necessity Suppose that X is extremally disconnected. Let A and B
be open and disjoint subsets of X. Then XA is closed and contains B, so
that we have cl B XA. But since X is extremally disconnected, cl B is
open. Then X cl B is closed and contains A, so that we have cl A X cl B,
hence A and B have disjoint closures.
Suciency Suppose that any two disjoint open subsets of X have disjoint
closures. Take A X to be any open set. Then A and int(XA) must have
disjoint closures. There are two possibilities, either int(XA) is empty or
not. In the latter case, we have cl A cl(int(XA)) = A, hence A = ,
which implies that cl A = A A = A is open. In the former case, we have
cl A = X, which is open. Therefore X is extremally disconnected.
The analogous condition for a basically disconnected space is as follows:
the space X is basically disconnected i for every pair of disjoint subsets U
and V , where U is open and V is cozero, they have disjoint closures.
Necessity Suppose that X is basically disconnected. Let U X be
open and V X be cozero. Then XU is closed and contains V , hence
cl V XU. But since X is basically disconnected, the closure cl V is open
and hence X cl V is closed and contains U. Hence cl U X cl V , which
means that U and V have disjoint closures.
Suciency Suppose that, for any open U X and cozero V X,
these sets have disjoint closures. Take V X to be a cozero-set and U =
int(XV ). There are two possibilities, either int(XV ) is empty or not. In
the latter case, by hypothesis, the sets U and V have disjoint closures, hence
24
V = (cl V ) U = . Then cl V = V V = V is open. In the former case,
we have cl V = X, which is open. Therefore X is basically disconnected.
2. In an extremally disconnected space, any two disjoint open sets are
completely separated. In a basically disconnected space, any two disjoint
cozero-sets are completely separated; equivalently, for every f C, pos f
and neg f are completely separated.
Proof. Suppose that X is an extremally disconnected space. Let A and B,
subsets of X, be disjoint open sets. These sets are completely separated if
they have disjoint zero-set neighborhoods. Note that, by (1.), the sets A and
B have disjoint open closures. So we can write
X = cl A cl B (X(cl A cl B)),
where the terms in the union are disjoint and open, hence each of them
is a zero-set. Therefore any two disjoint open sets have disjoint zero-set
neighborhoods (their closures), and so are completely separated.
Suppose that X is a basically disconnected space. Let A, B X be
cozero-sets. Then, by (1.), they have disjoint open closures, which are also
zero-set neighborhoods, by the same argument as above. So any two disjoint
cozero-sets are completely separated.
3. If X is basically disconnected, then for every f C, there exists a
unit u of C such that f = u[f[.
Proof. The sets pos f and neg f are disjoint cozero-sets, and hence have
disjoint open closures. In fact from an intermediate result from Suciency
of the second part of (1.), pos f and neg f are their own closures. Then we
dene the function u R
X
as
u(x) =
_
1 if x cl(pos f)
1 if x , cl(pos f)
Clearly f = u[f[. Also u C(X), by Lemma 1.4, and u is a unit in C(X).
25
4. Every dense subspace X of an extremally disconnected space T is
extremally disconnected. In fact, disjoint open sets in X have disjoint open
closures in T.
Before we begin, let us start o with a couple of Lemmas.
Lemma 1.6. Let T be a topological space and X T a dense subset. The
for any open A T we have cl A = cl(A X).
Proof. First, consider p cl A, this implies that any open neighborhood N
of p contains at least one point of A. Since A is open, the set AN is also a
neighborhood of p. Note that since X is dense in T, every neighborhood of p,
namely AN must contain points of X, hence ANX is non-empty. Since
N was arbitrary, we must conclude that p cl(A X) or cl A cl(A X).
Note that AX A, which implies that cl(AX) cl A. So nally we
have cl A = cl(A X).
Lemma 1.7. Let T be a topological space and X T a dense subset. For
any A X, we have cl
X
A = X cl
T
A.
Proof. Note that cl
T
A is closed in T, hence A cl
T
A is closed in X, hence
cl
X
A A cl
T
A cl
T
A and cl
X
A X, hence cl
X
A X cl
T
A.
Now let x X cl
T
A. For any open neighborhood N of x in X there
is an open neighborhood N

of x in T such that N

X = N. But since
x X cl
T
A, we have that x X and that N

contains at least one point of

A. Hence N = N

X also contains at least one point of A, which implies that

x cl
X
A. Finally we have X cl
T
A cl
X
A, and cl
X
A = X cl
T
A.
And now, we proceed to the proof.
Proof. Let A be an open subset of X, there there is an open subset A

of T
such that A = X A

. Then we have cl
T
A = cl
T
A

, by Lemma 1.6. Also

we have cl
X
A = X cl
T
A

, by Lemma 1.7. But cl

T
A

is open in T because
T is extremally disconnected, hence cl
X
A is open in X. Therefore X is also
extremally disconnected.
Now consider A and B disjoint open subsets of X, then we have A

and
B

open subsets of T such that A = XA

and B = XB

. Let C

= A

,
the set C

is open, hence it must contain at least one point x X. But then

x X A

= A and x X B

= B which contradicts the fact that A

and B are disjoint. Hence A

and B

must be disjoint open subsets of T. So

since T is extremally disconnected, they must have disjoint open closures in
26
T. But cl A

= cl A and cl B

= cl B, by Lemma 1.6, hence A and B have

disjoint open closures in T.
5. Every open subspace of an extremally disconnected space is extremally
disconnected. (A closed subspace, however, need not even be basically dis-
connected.)
Once again, we start o with a Lemma.
Lemma 1.8. Let X be a topological space and A an open subset. Then for
any subset B open in A, we have cl
A
B = A cl
X
B.
Proof. Clearly cl
A
B cl
X
B and cl
A
B A, hence cl
A
B Acl
X
B. Now
let x A cl
X
B, then for any open neighborhood N of x in A, the set N
is open in X since A is open and since x cl
X
B, the neighborhood N must
contain at least one point of B. Which implies that x cl
A
B. So nally we
have A cl
X
B cl
A
B and cl
A
B = A cl
X
B.
And now, we proceed to the proof.
Proof. Let X be an extremally disconnected topological space and A an open
subset. Then any subset B A open in A is also open in X since A is open.
Therefore cl
X
B is open. But then cl
A
B = A cl
X
B, by Lemma 1.8, and
since cl
X
B is open in X, the closure cl
A
B is open in A. Hence A is also
extremally disconnected.
6. X is extremally disconnected i every open subspace is C

-embedded
in X.
Proof. Necessity Suppose that X is extremally disconnected, but that there is
an open subset S X which is not C

-embedded in X. Then, by Urysohns

Extension Theorem (1.17), there are disjoint subsets A, B S in S that are
completely separated in S but not in X. If at least one of A and B is not open,
we can replace it WLOG by the interior of its zero-set neighborhood, hence
we can assume that A and B are open. Since S is open, the subsets A and
B are also open and disjoint in X. So, by (2.), A and B must be completely
separated in X. This contradiction implies that every open S X must be
C

-embedded in X.
Suciency Suppose that every open S X is C

-embedded in X. Take
any two disjoint open subsets A, B X. The set A B is open, and hence
27
C

-embedded. We dene f C(AB) such that f[A = 0 and f[B = 1, the

function f is continuous on AB by Lemma 1.4. Since AB is C

-embedded,
the function f can be extended to a bounded continuous function f on all of
X. But this simply means that A and B are completely separated. Which
implies that they have disjoint zero-set neighborhoods. Which implies that
they have disjoint closures. Hence, by (1.), the space X must be extremally
disconnected.
28
2 Ideals and z-lters
2B p.30 prime ideals
Recall that given a ring R, and ideal P is prime if ab P implies that at
least one of a or b is in P, where a and b are some elements of R.
Also recall that if R is a ring, and P and Q are ideals in R, we denote
by PQ the smallest (under set inclusion) ideal containing all products fg,
where f I and g J.
1. An ideal P in C is prime i P C

is a prime ideal in C

.
Proof. Necessity Assume that P is prime. Then, clearly, P C

is an ideal
in C

. Suppose that a, b C

such that ab P, then, since P is prime, at

least one of a or b is in P and hence in P C

. Hence P C

is prime in C

.
Suciency Assume that P is an ideal such that P C

is prime in C

.
Suppose that a, b C such that ab P. Then, by 1E.1, there are units
u, v C such that ua, vb C

. And since P is an ideal, we have (ua)(vb)

P C

. Then, since P C

is prime, we have ua P C

or vb P C

.
But since P is an ideal, we have u
1
ua = a P or v
1
vb P. Hence P is
prime in C.
2. If P and Q are prime ideals in C, or in C

, then PQ = P Q. In
particular, P
2
= P. Hence M
2
= M for every maximal ideal M in C or C

.
Proof. Take any a P and b Q, then ab P and ab Q, since both P
and Q are ideals. Note that since P Q is an ideal, we have PQ P Q.
Now take any f P Q. We can write f = (f
1/3
)
3
, hence f
1/3
P
and f
1/3
Q, since both P and Q are prime. Therefore (f
1/3
)
2
P, and
(f
1/3
)
2
f
1/3
= f PQ. Hence P Q PQ, and PQ = P Q.
So if P is a prime ideal, we have the identity P
2
= P P = P. The same
is true for any maximal ideal M in C or C

, since M is prime.
3. An ideal I in a commutative ring is an intersection of prime ideals i
a
2
I implies a I.
Proof. Necessity Assume that I is an intersection of prime ideals. Suppose
that a
2
I, then a
2
is contained in each of the prime ideals. So a is contained
in each of the prime ideals, hence a I.
29
Suciency Assume that I is an ideal in a commutative ring such that
a
2
I implies that a I. We know that I must be contained in some prime
ideal (for example a maximal ideal), hence I is contained in the intersection
of all prime ideals containing it. From 0.18 we know that the intersection
of all prime ideals containing I is the set of all elements of the ring of which
some power belongs to I. Now take an element of the ring a, such that a
n
I
for some n N. Then if n is even, by the hypothesis, we have a
n/2
I. And
if n is odd, we have a
n+1
I, and hence a
(n+1)/2
I. Iterating this way,
we nd that a I, hence I is equal to the intersection of all prime ideals
containing it.
2C p.31 functions congruent to constants
1. Let I be an ideal in C; if f r (mod I), then r f[X].
Proof. Since f r (mod I), we have f = r + i for some i I. Since I is
an ideal, it contains no units. Hence for every i I, there is an x X such
that i(x) = 0. Hence f(x) = r + i(x) = r. Therefore r f[X].
2. Let I be an ideal in C

; if f r (mod I), then r cl

R
f[X].
Proof. Since f r (mod I), we have f = r + i for some i I. Since I is
an ideal, it contains no units, that is for every i I, there is no > 0 such
that [i[ . Hence for each > 0, the set i[X] ], [ is non-empty. Hence
0 cl
R
i[X], and therefore r cl
R
f[X].
2D p.31 z-ideals
1. Let I be a z-ideal in C, and suppose that f r (mod I). If g(x) = r
wherever f(x) = r, then g r (mod I).
Proof. First note that, by the hypothesis, the function f r is in I and
Z(f r) Z(g r). So since Z[I] is a z-lter, we have Z(g r) Z[I].
And since I is a z-ideal, we must have g r I, hence g r (mod I).
30
2. If f
2
+ g
2
belongs to a z-ideal I, then f I and g I.
Proof. First we use the facts that Z[I] is a z-lter and I is a z-ideal to get
Z(f
2
+ g
2
) = Z(f) Z(g) Z(f) Z(f) Z[I]
f I.
Similarly, we have g I.
3. If I and J are z-ideals, then IJ = I J. Compare to 2B.2.
Proof. Suppose that a I and b J, so since I and J are ideals, we have
ab I and ab J, and hence ab I J. Suppose that f I J. Note
that Z(f
1/3
) = Z(f), so since I and J are z-ideals, we have g = f
1/3
I
and g J. Then g
2
I and f = g
2
g IJ, hence IJ = I J.
The proof above is very similar to the one given for 2B.2, except that we
use the fact that I and J are z-ideals, instead of prime ideals, to show that
f
1/3
is in the same ideal as f.
4. Z[(I, J)] is the set of all Z
1
Z
2
where Z
1
Z[I] and Z
2
Z[J].
Proof. Since (I, J) is an ideal, then Z[(I, J)] is a z-lter. So for any Z
1
Z[I]
and Z
2
Z[J] we have Z
1
, Z
2
Z[(I, J)] and hence Z
1
Z
2
Z[(I, J)].
Now suppose that Z Z[(I, J)], then there must exist i I and j J such
that Z = Z(i + j). But Z(i) Z(j) Z, hence Z Z[I] and Z Z[J].
Hence we can write Z = Z Z. This completes the proof.
2E p.31 z-ideals
First recall a z-lter on X is said to be prime if for any A, B Z(X) such
that A B T implies that at least one of A or B is in T.
For a z-lter T on X, TFAE:
(1) T is prime.
(2) Whenever A, B Z(X) such that A B = X, at least one of A or B
is in T.
(3) Given Z
1
, Z
2
Z(X), there exists Z T such that one of Z Z
1
or
Z Z
2
contains the other.
31
Proof. This theorem is the analog of Theorem 2.9 for z-lters.
(1)(2). Assume that T is prime. By hypothesis, we have A B = X,
and X T since T is a z-lter. Therefore, since T is prime, at least one of
A or B must be in T.
(2)(3). Consider the z-ideal P = Z

[T]. Assuming (2), if for some

g, h C we have gh = 0 (that is Z(g) Z(h) = X), then at least one of
Z(g) or Z(g) is in T, or at least one of g or h is in P, since P is a z-ideal.
This is equivalent to condition (3) of Theorem 2.9, which implies condition
(4) of 2.9 For every f C, there is a zero-set in Z[P] = T on which f does
not change sign.
Now take any Z
1
, Z
2
Z(X), say Z
1
= Z(g) and Z
2
= Z(h) for g, h C.
Consider the function [g[ [h[, by the condition (4) above, there must be a
Z T such that [g[ [h[ does not change sign, WLOG let us assume that it
is non-negative on Z. Then on Z, wherever g vanishes h must also vanish.
In other words Z Z
1
Z Z
2
, which completes the implication (2)(3).
(3)(1). Assume (3). Then suppose that for some Z
1
, Z
2
Z(X) we
have Z
1
Z
2
T. Then we can nd a Z T such that, WLOG, Z Z
1

Z Z
2
. This implies that Z (Z
1
Z
2
) = Z Z
2
, but Z (Z
1
Z
2
) T,
since T is a lter. Therefore Z Z
2
is in T, and hence Z
2
T. Hence T is
a prime z-lter.
2F p.31 finite spaces
Let X be a nite discrete space. In C(X):
1. f is a multiple of g i Z(g) Z(f).
Proof. Necessity Assume that f = gh for some h C(X). Then Z(f) =
Z(g) Z(h), therefore Z(g) Z(f).
Suciency Assume that Z(g) Z(f), then we can construct h a follows
h(x) =
_
0 if x Z(g)
f(x)/g(x) if x , Z(g)
.
Clearly f = hg, and since X is discrete, we have h C(X).
2. Every ideal is a z-ideal.
32
Proof. Let I be an ideal. Then take Z Z[I], then there is a function g I
such that Z = Z(g). Now take any f C(X) such that Z(f) = Z. Then,
by (1.), f must be a multiple of g, hence g I and I must be a z-ideal.
3. Every ideal is principal, and, in fact, is generated by an idempotent.
Proof. Let I be an ideal. Since X is nite, we have Z[I] T(X) nite. Since
Z[I] is closed under nite intersections and , Z[I], then Z =

Z[I] is
non-empty. Hence Z[I] must be principal and generated by Z, and since
every ideal is a z-ideal, I also must be principal.
In fact, we can dene an idempotent i as
i(x) =
_
0 if x Z
1 if x , Z
.
Then Z(i) = Z, and hence I = (i).
4. Every ideal is an intersection of maximal ideals. The intersection of
all maximal ideals in (0).
Proof. By (3.), every ideal is principal, hence every maximal ideal is principal
and must be generated by a singleton. Take any ideal I, since it is principal
we can write I = (i), and let Z = Z(i). Then we take the collection of
maximal ideals M
x
= (m
x
) where Z(m
x
) = x, for each x X. Clearly
the ideal I is given by I =

xZ
M
x
.
If we take the intersection of all maximal ideals, we have J =

xX
M
x
.
Clearly Z[J] is generated by X, hence we must have J = (0).
5. Every prime ideal is maximal.
Proof. Let P be a prime ideal, then, since X is nite and discrete, any
Z Z[P] can be expressed as a nite union of singletons. Note that since
P is prime, the z-lter Z[P] must also be prime, hence at least one of the
singletons must be in Z[P]. Any (proper) z-lter cannot contain more than
one singleton, hence Z[P] must contain a single singleton. This implies that
Z[P] is principal and maximal, and so is P.
33
2G p.31 prime vs. z-ideals in C(R)
1. Select a function l in C(R) such that l(0) = 0, while
lim
x0
l
n
(x)/x = ,
for all n N. Apply 0.17 to construct a prime ideal in C(R) that contains
i but not l. This prime ideal is not a z-ideal (and hence is not maximal).
Proof. We dene l C(X) as follows
l(x) =
_
1/ ln(x) if 0 < x e
1
1 if e
1
< x
l(0) = 0
l(x) = l(x).
It is easy to see that l satises the required properties. Now take the ideal
I = (i), it consists of all functions of the form f(x) = xg(x) for any g
C(R) (2.4). Note that l , I so the set of all powers of l is closed under
multiplication and disjoint from I. Hence by 0.17 we can construct a prime
ideal J such that I J and it still does not contain any power of l. But since
Z(l) = 0 and hence l Z

[Z[J]], the ideal J cannot be a z-ideal.

2. Let O
0
denote the ideal of all functions f in C(R) for which Z(f) is
a neighborhood of 0. Dene s C(R) as follows
s(x) =
_
x sin(/x) if x ,= 0
0 if x = 0
.
Then (O
0
, s) is not a z-ideal; and the smallest z-ideal containing (O
0
, s) is
not prime.
Proof. Suppose that (O
0
, s) is a z-ideal. Note that the zero-set of s is Z(s) =
1, 1/2, . . . , 0, . . . , 1/2, 1. Take t = s
1/3
, then Z(t) = Z(s) and hence
t must be in (O
0
, s). Then we can write t = o + su for some o O
0
and
u C(R). Since Z(o) is a neighborhood of 0 we can nd an > 0 such that
o[[, ] = 0, we restrict our attention to this interval from now on. Now
we can write t = su, hence wherever s(x) ,= 0 we have u(x) = t(x)/s(x) =
(x sin(/x))
2/3
. But then lim
x0
[u[ = and hence u cannot be continuous
34
on R, which implies that t , (O
0
, s). This contradictions proves that (O
0
, s)
is not a z-ideal.
Consider I = Z

[Z[(O
0
, s)]] the smallest z-ideal containing (O
0
, s). By
2.12, the ideal I is prime i the z-lter Z[(O
0
, s)] is prime. This z-lter is
generated by the neighborhoods of 0 and the set
S = 1, 1/2, . . . , 0, . . . , 1/2, 1.
We can write
S = 1, 1/2, . . . , 0 0, . . . , 1/2, 1,
note that both sets in the union are zero-sets, hence if Z[(O
0
, s)] is prime it
must contain at least one of them. But neither of those sets is in this z-lter
because they cannot be generated by nite intersections and supersets of the
generating sets. Therefore J cannot be a prime ideal.
2H p.32 the identity function i in C(R)
1. The principal ideal (i) in C(R) consists precisely of all functions in
C(R) that vanish at 0 and have a derivative at 0. Hence every non-negative
function in (i) has a zero derivative at 0.
Proof. First note that the ideal (i) is the set of all f C(R) of the form
f = ig. Hence, since i(0) = 0, for all f (i) we have f(0) = 0. Now take
f (i) such that f = i g, and consider its derivative at 0
f

(0) = lim
h0
f(h) f(0)
h
= lim
h0
hg(h) 0
h
= lim
h0
g(h)
= g(0).
Hence every function in (i) vanishes at 0 and has a derivative at 0. Conversely,
consider a function f C(R) such that f(0) = 0 and f

(0) exists. Then we

can dene g = f/i and note that
lim
h0
g(h) = lim
h0
f(h)
h
= lim
h0
f(h) f(0)
h
= f

(0).
35
Hence g is continuous and we can write f = i g and f (i).
2. (i) is not a prime ideal.
Proof. Suppose that (i) is prime, then, by 2B.2, we must have (i)
2
= (i).
First note the identity (i)
2
= (i
2
), in other words, the ideal (i) is the set of
all functions f C(R) such that f = i
2
g for some g C(R). Now take any
f (i)
2
, we can write f(x) = x
2
g(x) for some g C(R) and hence f(0) = 0.
Consider its second derivative at 0,
f

(0) = lim
h0
f(h)f(0)
h

f(0)f(h)
h
h
= lim
h0
h
2
g(h) 2f(0) + h
2
g(h)
h
2
= lim
h0
g(h) + g(h)
= 2g(0).
Hence every f (i)
2
has a second derivative at 0. However, we have i[i[ (i)
which does not have a second derivative at 0, hence i [i[ , (i)
2
. This
contradiction completes the proof.
3. The ideal (i, [i[) is not principal.
Proof. Suppose that this ideal is principal and (i, [i[) = (d) for some d
C(R). So, since i, [i[ (d), there must exist g, h C(R) such that i = gd
and [i[ = hd. Note that for x > 0 we must have g(x) = h(x) and for x < 0
we must have g(x) = h(x). So, by continuity, we have g(0) = h(0) = 0.
Also, since d (i, [i[), there must exist s, t C(R) such that si + t[i[ = d.
Now we can write s(gd) + t(hd) = d or (sg + th)d = d. Recall that we
must have Z(d) = 0, and note that for every x where d(x) ,= 0 we have
(sg +th)(x) = 1. Hence, by continuity, we must have sg +th = 1. But then
both g and h cannot vanish at 0. This contradiction completes the proof.
4. Exhibit a principal ideal containing (i, [i[).
Proof. Note that both of i [i[
1/2
and [i[
1/2
= [i[ [i[
1/2
are continuous
functions, hence (i, [i[) ([i[
1/2
).
36
2I p.32 C(Q) and C

(Q)
The set of all f C(Q) for which lim
x
f(x) = 0 is not an ideal in C(Q).
But the bounded functions in this set do constitute an ideal in C

(Q).
Proof. Let I C(Q) be the set of all such functions, and let I

= I C

(Q).
Consider the function f such that f(x) = x . Clearly f is in I, but f
is a unit of C(Q) since f
1
is also continuous on Q. Hence I cannot be a
(proper) ideal.
Now consider I

, clearly it is closed under addition. Also, consider any

g C

(Q), then [g[ r for some r R. Then fg is also in I

since
lim
x
[f(x)[r = 0 and
0 lim
x
[f(x)g(x)[ lim
x
f(x)r lim
x
f(x)g(x) = 0.
Therefore I

is an ideal in C

(Q).
2J p.32 ideal chains in C(R), C(Q) and C(N)
1. Find a chain of z-ideals in C(R) (under set inclusion) that is in one-
to-one, order preserving correspondence with R itself.
Proof. Consider the chain of z-lters T
r
on Z(R), where for each r R we
have T
r
is principal and generated by the interval [r, [. The corresponding
z-ideals in C(R) satisfy the desired properties.
2. Find a chain of z-ideals in C(Q) (under set inclusion) that is in one-
to-one, order preserving correspondence with Q itself.
Proof. The same construction as in (1.) works, except that the sets that
generate the z-lters have the form Q [r, [.
3. Find a chain of z-ideals in C(N) (under set inclusion) that is in one-
to-one, order preserving correspondence with N itself.
Proof. Suppose that there is a bijection between two sets A and B, then this
bijection induces a natural bijection on the power sets T(A) and T(B) which
preserves set inclusion. If T is a lter on A, then its image under the induced
bijection must also be a lter on B.
37
Take the chain T
r
of z-lters on Q from (2.), and consider the chain T

r
of corresponding lters on Q such that T

r
is the smallest lter containing
T
r
(note that these retain the ordering of the generating z-lters). We know
that there exists a bijection between Q and N, hence we can map the lters
T

r
to lters on N preserving the inclusion order. But since N is discrete,
every lter on N is a z-lter. Hence we can construct a chain of z-ideals
in C(N) by taking the z-ideals corresponding to the z-lters on N that we
constructed above.
2K p.32 z-filters and C

If M is a maximal ideal in C

, and Z[M] is a z-lter, then Z[M] is a z-

ultralter.
Proof. Suppose that I and J are ideals in C, such that I J (properly
contains). Take any f JI, it is either bounded or not. If it is not
bounded, then there is a unit u of C such that uf is bounded (cf. 1E.1). The
function uf must be in JI since if it was in I then so would f = u
1
uf.
Hence we have I J implies that I C

J C

.
Now consider Z[M], since it is a z-lter, then I = Z

[Z[M]] is a z-
ideal in I. Suppose that T is another z-lter such that Z[M] T. Then
J = Z

[T] is also an ideal in C, such that I J. Then I

= I C

and
J

= J C

are ideals in C

, so we have M I

and since M is
maximal we must have M = I

= J

. Hence, by the argument above, we

also have I = J and hence T = Z[M]. Therefore Z[M] is a z-ultralter.
2N p.35 the m-topology on C
The m-topology is dened on C(X) by taking as a base for neighborhood
system at g all sets of the form
M
u
= f C [ [g f[ u,
where u is a positive unit of C. The same topology results if it is required
further that u be a bounded function.
We also need a few more denitions. The ring C is called a topological
ring if it is endowed with a topology and the operations (f, g) f + g and
(f, g) fg are continuous. If C is regarded as a vector space (with scalar
multiplication given as (r, f) rf), it is called a topological vector space if it
38
is endowed with a topology and the operations (f, g) f +g and (r, f) rf
are continuous.
The uniform norm topology on C

is dened by taking as a base for the

neighborhood system at g all sets of the form
U

= f C

[ [g f[ ,
where > 0.
1. C is a topological ring.
Proof. Consider the addition operation on C, for any f, g C we have
+: (f, g) f + g. Consider any f, g C, let h = f + g and take any M
u
neighborhood around h. Let v C be a positive unit and take an M
v
M
v
neighborhood V around (f, g). Then for any (a, b) V , we have
[h (a + b)[ = [(f a) + (g b)[ [f a[ +[g b[ 2v.
So, if we let v = u/3, then +[V ] M
u
. Hence addition is continuous at
(f, g) (cf. 1D.3) and therefore on all of C C.
Consider the multiplication operation on C, for any f, g C we have
: (f, g) fg. Consider any f, g C, let h = f + g and take any M
u
neighborhood around h. Let v C be a positive unit and take an M
v
M
v
neighborhood V around (f, g). Then for any (a, b) V , we have
[h ab[ = [(f a)g + (a f)(g b) + f(g b)[
[f a[[g[ +[a f[[g b[ +[f[[g b[
([g[ + u +[f[)v.
So, if we let v = u(u+[f[+[g[)
1
/2, then [V ] M
u
. Hence multiplication
is continuous at (f, g) and therefore on all CC. Therefore C is a topological
ring.
2. The relative m-topology on C

contains the uniform norm topology,

and the two coincide i X is pseudocompact. In fact, when X is not pseudo-
compact, the set of constant functions in C

is discrete (in the m-topology),

so that C

is not even a topological vector space.

39
Proof. Let T
m
denote the m-topology on C, and let T
u
denote the uniform
norm topology on C

. Let V C

be an open set in T
u
, then for every
point in V , there is an > 0 such that there is a U

neighborhood of that
point contained in V . But is also a positive unit of C, hence each U

neighborhood is also a M

neighborhood. Therefore V is also open in T

m
,
hence T
u
is contained in T
m
.
Assume that X is not pseudocompact, then C must contain an unbounded
function f. Then there must exist an unbounded positive unit in C, e.g.
[f[ for any > 0. Let u be an unbounded positive unit of C, then u
1
is
also a positive unit such that 0 cl(u
1
[X]). Consider an M
u
1 neighborhood
U of 0, then int
Tm
U is open in T
m
and contains 0. However, note that no U

neighborhood of 0 is contained in U, since u

1
would imply that = 0.
Therefore if X is not pseudocompact, then T
m
is strictly bigger than T
u
. And
if X is pseudocompact, then trivially T
m
= T
u
, since C = C

.
3. The set of all units of C is open, and the mapping f f
1
is a
homeomorphism of this set onto itself.
Proof. Let U be the set of all units of C. Take any u U and let v = [u[/2.
Since v is a positive unit, we can take an M
v
neighborhood around u. Let
f M
v
, then
[f[ = [u (u f)[ [[u[ [u f[[ [[u[ v[ [u[/2 > 0.
Hence each such neighborhood is contained in U, and therefore U is open.
Now, consider the inverse operation inv: f f
1
. Take any unit u U,
and take a neighborhood M
v
around u
1
and contained in U. Then take
a neighborhood M
w
around u and contained in U, letting f M
w
(that is
0 [u[ w [f[ [u[ + w). Then we have
[f
1
u
1
[ =
[f u[
[fu[

w
([u[ w)[u[
v,
if w u
2
v/(1 + uv). Therefore inv is continuous at u, and hence on all
of U. The fact that inv is a bijection, with itself as the inverse, is trivial. So
since it is continuous, inv must be a homeomorphism of U into itself.
40
4. The subring C

is closed.
Proof. Consider the set of all unbounded functions in C. For any unbounded
f and any > 0, we can take an M

neighborhood around f. Then every

function in this neighborhood is also unbounded. Hence the complement of
C

in C is open, and therefore C

is closed.
5. The closure of every ideal is a (proper) ideal. Hence every maximal
ideal is closed. Every maximal ideal in C

is closed.
Proof. The closure of an ideal is the set of all functions such that each neigh-
borhood of each of those functions contains at least one element of the ideal.
Now, let I be an ideal in C and J its closure in C.
Suppose that f, g J, then for any positive ideal u we can nd a, b I
such that [a f[ u/2 and [b g[ u/2. Recall that since I is an ideal, we
have (a + b) I. Also,
[(f + g) (a + b)[ = [(a f) + (b g)[ u/2 + u/2 = u.
Therefore J is closed under addition.
Suppose that f J, then for any positive ideal u we can nd a I such
that [f a[ u. Recall that since I is an ideal, we have (a) I. Also,
we have [(f) (a)[ = [f a[ u. Therefore J is closed under taking
additive inverses.
Suppose that f J and g C, then for any positive unit u we can nd
a I such that [f a[ uh
1
, where h = [g[ 1 (note that h
1
g 1).
Since I is an ideal, we have ag I. Also, we have
[fg ag[ = [f a[[g[ uh
1
[g[ u.
Therefore J is closed under multiplication from outside. Hence J must
be an ideal (possibly an improper one).
Finally suppose that u J, where u is a unit of C. By (3.), the set of
all units of C is open, hence we can take a neighborhood U around u such
that every element of the neighborhood is also a unit (that is not equal to
zero anywhere). But an element of a proper ideal must vanish at least at
one point. Therefore, since J is the closure of a proper ideal I, the ideal J
cannot contain any units, in other words J is also proper.
Let M be a maximal ideal in C and M

its closure. Then, by the above,

M

is also an ideal and, by properties of closure, we have M M

. But then,
41
by maximality of M, we have M

= M. In other words a maximal ideal M

must be closed.
Let M be a maximal ideal in C

and M

its closure in C. Then M

is also an ideal in C

, which contains M. But since M is maximal in C

, we
have M

= M. Recall that, by (4.), C

is closed, and since M

is also
closed, then M must be closed as well.
6. Every closed ideal in C is z-ideal.
Proof. Consider a closed ideal I in C. Suppose g I and f C such that
Z(f) = Z(g). Now given a positive unit u of C, we dene a function h as
follows:
h(x) =
_

_
f(x) + u(x)
g(x)
if f(x) u(x)
0 if [f(x)[ u(x)
f(x) u(x)
g(x)
if u(x) f(x)
.
First note that h is well dened, and that it is continuous, by construction,
on x X [ [f(x)[ u(x) and x X [ [f(x)[ u(x). Since these sets are
closed and form a closed cover or X, by 1A.1, h C.
Given x X such that [f(x)[ u(x), we have
[f(x) g(x)h(x)[ = [f(x) 0[ u(x).
Given x X such that [f(x)[ > u(x), we have
[f(x) g(x)h(x)[ = [f(x) (f(x) u(x))[ = u(x) u(x).
Hence [f gh[ u, and since u was arbitrary, the function f must be in
the closure of I. But I is closed, so f must be in I. Therefore given any
Z Z[I], the ideal I must contain all functions f C such that Z(f) = Z.
Therefore every closed ideal is also a z-ideal.
7. In the ring C(R), the z-ideal O
0
of all functions that vanish on a
neighborhood of 0, is not closed.
Proof. Consider the identity function i. Given any positive unit u of C(R)
we dene i
u
as follows:
i
u
(x) =
_
_
_
i(x) + u(x) if i(x) u(x)
0 if [i[ u(x)
i(x) u(x) if u(x) i(x)
.
42
Note that, since 0 Z(i) and since u must be bounded away from 0 by
a positive constant in any bounded neighborhood of 0, we have that Z(i
u
)
is a neighborhood of 0.
But then, [i i
u
[ u for any positive unit u. Hence i cl O
0
. But
Z(i) = 0 which is not a neighborhood of 0, hence i , O
0
and O
0
is not
closed.
43
3 Completely Regular Spaces
3B p.48 countable sets
Let X be a completely regular space.
1. A countable set disjoint from a closed set F is disjoint from some
zero-set containing F.
Proof. Let S X be a countable set and F X be a closed set. Then since
X is completely regular, for each s S there is a zero set Z
s
containing F
and disjoint from s. Then Z =

sS
Z
s
is also a zero set (intersection of
countably many zero sets) containing F and disjoint from S.
2. A C-embedded countable set S is completely separated from every
disjoint closed set.
Proof. Let F be a closed set disjoint from S. From (1.), we know that there
is a function f C(X) such that f[F] = 0 and 0 , f[S]. Therefore we
can dene a continuous function g

C(S) such that g

(s) = f
1
(s), for
each s S. And since S is C-embedded, we can extend g

to a continuous
function g C(X). Note that h = fg satises the condition h[S] = 1 and
h[F] = 0, hence S and F are completely separated.
3. Any C-embedded countable set is closed.
Proof. Let S be a C-embedded countable set. Note that, since X is Haus-
dor, points are closed sets. Then for any x , S, the singleton x is closed,
and from (2.) we know that for every x , A there is a zero-set Z
x
contain-
ing S and disjoint from x. Now we can write S =

xS
Z
x
, which is closed
(intersection of closed sets).
4. Any two countable sets, neither of which meets the closure of the other,
are contained in disjoint cozero-sets.
Proof. Let A, B X be countable subsets, neither of which meets the closure
of the other. First, we can enumerate the elements of each set, that is
A = a
n

nN
and B = b
n

nN
. Recall that, since X is completely regular,
we can nd a function f C(X) such that f(a
1
) = 0 and f[cl B] = 1.
44
Then we can nd a zero-set neighborhood Z
A
1
of a
1
containing a cozero-
set neighborhood V
A
1
of a
1
, both disjoint from cl B. Namely, we can have
Z
A
1
= x [ f(x) 1/3 and V
A
1
= x [ f(x) < 1/3. Similarly, we can nd a
zero-set neighborhood Z
B
1
of b
1
containing a cozero-set neighborhood V
B
1
of
b
1
, both disjoint from cl AZ
A
1
. Again, we can nd a zero-set neighborhood
Z
A
2
of a
2
containing a cozero-set neighborhood V
A
2
of a
2
, such that Z
A
2
is
disjoint from cl B Z
B
1
. Continuing by induction, we nd that A and B are
contained in disjoint cozero-sets, respectively,

nN
V
A
n
and

nN
V
B
n
.
5. A countable, completely regular space is normal.
Proof. Consider any two disjoint closed sets. Since the whole space is count-
able, then so are they. Then, by (4.), we know that these sets are contained
in disjoint cozero-sets. Since cozero-sets are open, any two disjoint closed
sets are contained in disjoint neighborhoods, hence the space is normal.
3C p.48 G

-points of a completely regular space

Let p be a G

-point of a completely regular space X, and let S = Xp.

1. If g C

(S), h C(X), and h(p) = 0, then g (h[S) has a continuous

extension to all of X.
Proof. We already know that g (h[S) is continuous on S, so we dene the
function f R
X
as f[S = g (h[S) and f(p) = 0. We need only prove that
f is continuous at p to show that f is continuous on all of X.
Since g is bounded, there must be an r 0 such that [g[ r. Now take
any > 0, since h is continuous at p, there must be an open neighborhood
U of p such that h[U] ]/r, /r[. Then f[U] ], [, therefore f is
continuous at p and hence on all of X.
2. If Z is a zero-set in S, then cl
X
Z is a zero-set in X.
Proof. Consider two zero-sets in X, namely Z = Z(f) and p = Z(h),
recall that a G

point is a zero-set (3.11(b)), for f C(S) and h C(X).

WLOG, assume that 0 f 1 and 0 h 1 and h[Z] = 1. There are
two possibilities, either Z = cl
X
Z or Z p = cl
X
Z. In the latter case
p cl
X
Z and cl
X
Z = Z p = Z(fh). In the former case p , cl
X
Z and
45
cl
X
Z = Z = Z((1 f)h 1), where we implicitly extend (1 f)h to all of
X as in (1.).
3D p.48 normal spaces
1. TFAE for any Hausdor space X.
(1) X is normal.
(2) Any two disjoint closed sets are completely separated.
(3) Every closed set is C

-embedded.
(4) Every closed set is C-embedded.
Proof. (1)(2). Urysohns Lemma (3.13).
(2)(3). Urysohns Extension Theorem (1.17).
(3)(4). Take a closed set and any other closed set disjoint from it, since
X is normal, the two sets are completely separated. Finally use Theorem
1.18.
(4)(1). Assume (4). Given two disjoint closed sets A, B X, the set
A B is also closed. We dene a continuous function f such that f[A = 0
and f[B = 1. And since AB is C-embedded, we can extend f to all of X,
hence any two closed sets are completely separated.
3. Every closed G

in a normal space X is a zero-set.

Proof. Let F =

nN
U
n
be a closed G

set. Then, since X is normal, every

U
n
contains a zero-set neighborhood, say Z
n
, of F. Then
F

nN
Z
n

nN
U
n
= F.
Hence F =

nN
Z
n
is a zero-set, since it is a countable intersection of
countably many zero-sets.
46
4. Every completely regular space with the Lindel of property (i.e. such
that every open cover of X has a countable subcover) is normal.
Before we proceed, it is useful to consider a Lemma.
Lemma 3.1. A closed subspace of a Lindelof space is Lindelof.
Proof. Take X to be Lindel of and F X a closed subspace. Then for any
open cover U

A
of F, we have an open cover U

XF. Then, since

X is Lindelof, we can nd a countable subcover V
n

nN
. If necessary we
remove XF, and we have the desired countable subcover of F. Therefore
F is also Lindel of.
And now, on to the proof.
Proof. Let X be completely regular and Lindel of. Take two disjoint closed
subsets A and B of X, by Lemma 3.1, they are also Lindelof. Since X
is completely regular, then for each x A we can nd disjoint zero-sets
neighborhoods Z
a
x
and Z
b
x
for x and B, respectively. Now we can nd a
countable subcover Z
a
xn

nN
of A, and hence a zero-set Z
b
=

nN
Z
b
xn
containing B and disjoint from A. Since Z
b
is still a closed set disjoint from
A, by a similar procedure we can nd a zero-set Z
a
containing A and disjoint
from Z
b
. Hence any two disjoint closed sets in X are completely separated
and X is normal.
5. Let X be a completely regular space. If X = S K, where S is open
and normal, and K is compact, then X is normal.
Proof. First, WLOG, assume that S and K are disjoint (or just replace S
and K by S and KS).
Take A, B to be any two disjoint closed subsets of X. We can represent
them as disjoint unions of AS, AK and B S, B K. Since any closed
subset of a compact set is also compact, we have that AK and B K are
compact in X.
From 3.11(a) we know that in a completely regular space a compact set
and a closed set disjoint from it are completely separated. Using this infor-
mation we can construct the following neighborhoods (the neighborhoods are
denoted by N
i
or M
i
)
(a) N
1
A K disjoint from M
1
B;
47
(b) N
2
A disjoint from M
2
B K;
(c) S N
3
A S disjoint from S M
3
B S (S is normal).
Now we can form the following neighborhoods N = (N
1
N
3
) N
2
A
and M = (M
2
M
3
) M
1
B. If N and M are disjoint then we are done.
Intuitively this seems true, but it is dicult to verify this claim by means
other than brute force. Namely
N M = [(N
1
N
3
) N
2
] [(M
2
M
3
) M
1
]
= [(N
1
N
2
) (N
3
N
2
)] [(M
2
M
1
) (M
3
M
1
)]
= (N
1
N
2
M
2
M
1
) (N
1
N
2
M
3
M
1
)
(N
3
N
2
M
2
M
1
) (N
3
N
2
M
3
M
1
)
= [

..
N
1
M
1
N
2
M
2
] [

..
N
1
M
1
N
2
M
3
]
[N
2
M
2
. .

N
3
M
1
] [N
3
M
3
. .

N
2
M
1
]
=
Hence any two disjoint closed subsets of X have disjoint neighborhoods and
X is normal.
3E p.49 nonnormal space
Let X be a nonnormal, Hausdor space.
1. X contains a closed set that is not a zero-set.
Proof. Suppose that every closed set in X is also a zero-set. Then any two
disjoint closed sets in X are contained in disjoint zero-sets, which implies
that they are completely separated. But the latter statement implies that X
is normal. This contradiction completes the proof.
2. X has a subspace S with the following property: any two completely
separated sets in S have disjoint closures in X, yet S is not C

-embedded in
X. Compare to Urysohns Extension Theorem
48
Proof. Consider a closed subset S X. Then if two subsets A, B S
are completely separated in S, there must be disjoint zero-sets of S, namely
Z
1
A and Z
2
B. Note that Z
1
and Z
2
are closed in S, and since S is
closed, they are also closed in X. Hence cl
X
A Z
1
and cl
X
B Z
2
, hence
the closures of A and B are disjoint.
So any closed set S satises the desired property, now suppose that every
closed set is C

-embedded, from 3D.1(3), this implies that X is normal. This

contradiction completes the proof.
Urysohns Extension Theorem says that for a space S X to be C

-
embedded any pair of sets that is completely separated in S are also com-
pletely separated in X. This problem shows that requiring that they only
have disjoint closures is not enough.
3K p.49 the completely regular, nonnormal space

Let denote the subset (x, y) [ y 0 of RR, provided with the following
enlargement of the product topology: for r > 0, the sets
V
r
(x, 0) = (x, 0) (u, v) [ (u x)
2
+ (v r)
2
< r
2

are also neighborhoods of the point (x, 0). This space is called the Niemytzki
plane. Clearly, satises the rst countability axiom.
Also recall that the regular subspace topology is dened by the following
neighborhoods: for each > 0, the sets
N

(x, y) = (u, v) [ (u x)
2
+ (v y)
2
<
2

are neighborhoods of the point (x, y).

1. The subspace D = (x, 0) [ x R of is discrete, and is a zero-set
in .
Proof. For any (x, 0) D, for any r > 0, the neighborhood V
r
(x, 0) of (x, 0)
does not contain any other elements of D. Hence D is discrete.
Consider the function f on dened as f(x, y) = y for every (x, y) .
Clearly f is continuous and Z(f) = D.
49
2. is a completely regular space.
Proof. Consider a point p and a closed set F disjoint from it, then p must
have a neighborhood that does not meet F. Then p either has a V
r
(p) or N

(p)
neighborhood. In the latter case, we can dene f(x, y) = ([(x, y) p[/) 1,
which clearly satises f[F] = 1 and f(p) = 0. In the former case, we
can dene f as f(p) = 0 and f(q) = 1 for q V
r
(p), and require f to be
linear on the segment between p and any such q, clearly f again satises the
property f[F] = 1 and f(p) = 0. To show that f is continuous from to
[0, 1], consider a subbasic closed set [a, b] [0, 1], then f

[[a, b]] = cl V
b
V
a
which is a closed set in . Hence f C() and therefore is completely
regular.
3. The subspace = (QQ)D is dense in . Hence [C()[ = c.
Proof. Let p , then it can have a V
r
(p) or N

(p) neighborhood. In the

latter case, since Q is dense in R there must be u, v Q such that [up
x
[ <
/2 and 0 v p
y
< /2 so that (u, v) N

(p). In the former case, we have

N
r
(p+(0, r)) V
r
(p), and we can apply the same argument as above. Hence
is dense in .
Therefore every continuous function in C() is uniquely determined by
its value on . But since is countable, that is [[ =
0
, we have
[C()[ [R

[ = c

0
= c.
Also C() contains at least all the constant functions r for r R, hence
c = [R[ [C()[. Therefore [C()[ = c.
4. The zero-set D is not C

-embedded in . Hence is not normal.

Proof. In (1.) we have shown that the closed subspace D, with cardinality
[D[ = c, is discrete, hence any function on D is continuous. Therefore we
have [C

(D)[ 0, 1
D
= 2
c
> c (by Cantors theorem) while [C

()[ =
[C()[ = c. So D cannot be C

-embedded since not every function on D can

be continuously extended to all of . But according to 3D.1(3), this implies
that is not normal.
50
5. Every closed set in is a G

.
Proof. Every open set V in can be represented as V = U S, where U is
an open set from the subspace topology and S D. Then every closed set
H in can be represented as H = F (T (D)), where F is a closed set
from the subspace topology and T D. Note that it is known that as a
subspace of R R is normal, hence every closed set F in this topology is a
zero-set and hence a G

(1.10). Also note that for any r > 0, we can write

T (D) = (D)

(t,0)T
V
r
(t, 0), which makes this set open. Hence H
can be written as the intersection of a G

and an open set, hence H is also

a G

.
6. contains a closed G

that is not a zero-set.

Proof. Since every closed set is a G

, by (5.), the statement of 3E.1 dictates

the existence of such a set.
However, we can also exhibit such a set explicitly (almost). Consider the
following closed subsets of , H = D(x, 0) [ x Q and K = D(x, 0) [
x RQ. Since both of these sets are closed, they are also G

s (5.). Now,
suppose that both H and K are zero-sets, then they are completely separated
(Theorem 1.15) and hence have disjoint open neighborhoods, say A and B
respectively. Clearly, each of these neighborhood must be a union of some
V

(x, 0) neighborhoods.
Pick (x
1
, 0) H, then it is contained in a neighborhood V

1
(x
1
, 0) A
for some
1
> 0. Suppose that we have already picked 2n 1 such points
alternating between H and K, with corresponding
1
, . . . ,
2n1
, for some
n N. Then we pick (x
2n
, 0) K such that x
2n
]x
2n1

2n1
, x
2n1
[ (this
is possible since XQ is dense in R), which is contained in V

2n
(x
2n
, 0) B
for some
2n
> 0. Since A and B are disjoint, we can bound
2n
form above,
namely
2n
<
2n1
/2 (this is not the tightest possible bound, but it will
suce). Next we can pick (x
2n+1
, 0) H such that x
2n+1
]x
2n
, x
2n
+
2n
[
(this is possible since Qis dense in R), which is contained in V

2n+1
(x
2n+1
, 0)
A for some
2n+1
> 0. Once again, since A and B have to be disjoint, we can
bound
2n+1
from above, namely
2n+1
<
2n
/2. We continue this inductively
and end up with two sequences x
n
and
n
.
Clearly, x
n
is Cauchy, so it must converge to a limit, say x. Also,
n

is bounded from above by

1
/2
n1
, hence its limit is 0. Note that (x, 0) is
either in H or K, so it must be contained in an open neighborhood V

(x, 0),
51
for some > 0, which is contained in only one of A or B. Finally, note that
is bounded from above by each
n
, in other words = 0. Which contra-
dicts the initial assumption that H and K have disjoint open neighborhoods.
Therefore at least one of H or K is not a zero-set.
3L p.51 extension of functions from a discrete set
Let X be a completely regular set.
1. Let V

be a family of disjoint sets in X with nonempty interiors,

and such that for each index , the set

=
V

is closed. Any set D formed

by selecting one element from the interior of each V

is C-embedded in X.
Proof. Any such D is clearly discrete, hence any function on D is continuous.
So let g C(D) be a continuous function on D. For each there is a d

D
such that d

int V

. Since X is completely regular we can nd f

C(X)
such that f

(d

) = g(d

) and f[X int V

] = 0.
Now consider the following family of closed sets V

cl X

,
which forms a closed cover of X. We will show that this family is neighbor-
hood nite (cf. 1A.3). Since for any we have

=
V

closed, then so is

. Consider a point x X, since the V

are disjoint, x can belong to

at most one of them. Suppose that x is in neither of V

s, then there is a
neighborhood around it contained in X

. If x is in V

, then

=
V

is closed so there must be a neighborhood of x contained in the complement

of that set, that is it meets at most two sets of the closed cover.
Finally we can dene the function f R
X
such that f[V

= f

[V

, for
each , and f[ (X

) = 0. By 1A.3, the function f is continuous on all

of X and it extends g from D to X, hence D is C-embedded.
2. Let x
n

nN
be a discrete set (not necessarily closed) in X, and let
r
n

nN
be any convergent sequence of real numbers. Then there exists
f C

(X) such that f(x

n
) = r
n
for all n N.
Before proceeding it is useful to introduce the generalized analogue of the
Weierstrass M-test from analysis.
Lemma 3.2 (Weierstrass M-test). Let X be a topological space and f
n

be a sequence of continuous functions on X. Suppose that for each n N,

we have [f
n
[ M
n
, where

n=1
M
n
is a convergent series of reals. Then
f =

n=1
f
n
is continuous.
52
Proof. Consider the sequence of partial sums s
n
=

n
m=1
f
m
, each s
n
is
continuous. Take any p, q N such that p q, then we have
[s
q
s
p
[ =

m=p+1
f
m

m=p+1
[f
m
[

m=p+1
M
m
But since

n=1
M
n
converges, for any > 0 we can nd an N N such
that for any p, q > N we have [s
q
s
p
[

q
m=p+1
M
m
< . Hence the
sequence s
n
converges uniformly to

n=1
f
n
, and the function f =

n=1
f
n
is continuous.
And now, on to the proof.
Proof. Consider the sequence of real numbers s
n

nN
where s
1
= r
1
and
inductively s
n+1
= r
n+1
r
n
. It is clear from the fact that x
n
is discrete
that, for each n N, we can nd open sets V
n
such that V
n
x
n

nN
=
x
m

n
m=1
. Also dene V
0
= . From 3.11(a) we know that a compact set and
a closed set disjoint from it are completely separated, noting that x
m

n
m=1
is compact, since it is discrete and nite, this means that we can nd a
continuous function f
n
from X to [s
n
, s
n
] such that f
n
[XV
n1
] = s
n

and f
n
[x
m

n1
m=1
] = 0. Then, by the Weierstrass M-test (Lemma 3.2), the
function f =

n=1
f
n
is continuous and satises the desired properties.
3. If X is innite, then C

(X) contains a function with innite range.

Proof. According to 0.13, every innite Hausdor space contains a discrete
countably innite subset x
n

nN
. By (2.), we can nd a continuous function
f on X such that f(x
n
) = 1/n (this function can be bounded by (1f) 1
if necessary). This function f clearly has innite range.
4. Let D be a countable discrete set in X. TFAE and imply that D is
closed.
53
(1) D and any disjoint closed set are completely separated.
(2) D and any disjoint closed set have disjoint neighborhoods.
(3) D is C-embedded in X.
Proof. (1)(2). If two sets are completely separated, then, by 1.15, they
have disjoint zero-set neighborhoods.
(2)(3). Assume that D and any disjoint closed set have disjoint neigh-
borhoods. First note that we can choose disjoint open neighborhoods U
n
, for
each n N, each one containing only one point from D, say d
n
D, where
D = d
n

nN
. We do so inductively. Since D is discrete, we can nd an open
set containing only d
1
and no other elements of D. Since X is completely
regular, we can take a zero-set neighborhood of d
1
contained in that open
set. Now suppose that for some n N, the points d
m
, where m < n, have
disjoint zero-set neighborhoods. Then we again nd an open set containing
only d
n
from D, intersect it with the complement of the union of the obtained
zero-set neighborhoods, which gives us an open neighborhood of d
n
, inside
which we can take the next zero-set neighborhood containing d
n
. Finally, we
can take the interiors of these zero-set neighborhoods and call them U
n

nN
.
Now, let U =

nN
U
n
, then XU is a closed set disjoint from D, hence
there must exist disjoint open neighborhoods A and B, containing D and
XU, respectively. Then, for each n N, the set U

n
= U
n
A is sill an open
neighborhood of X, inside which we can take a zero-set neighborhood Z
n
of
d
n
.
Next we show that W
m
=

n=m
Z
n
is closed. Consider x W
m
, its is
contained in either exactly one of U
n
or XU. If x U
n
and n = m, then
x U
m
XW
m
. If x U
n
and n ,= m, then x U
n
(XZ
n
) XW
m
.
If x XU, then x B XW
m
. Hence XW
m
is open, or W
m
is closed,
for any m N. Finally, since the elements of D are picked from the interiors
of Z
n
as in (1.), the set D is C-embedded.
(3)(1). Assume that D is C-embedded. But then, directly from 3B.2,
since D is also countable, we have that D is completely separated from any
closed set disjoint from it.
Assume condition (2). Then take any point x XD. Since X is com-
pletely regular, it is also assumed to be Hausdor, so x is a closed set.
Hence, by hypothesis, D and x have disjoint neighborhoods, which implies
that XD is open or that D is closed.
54
3M p.51 suprema in C(R)
1. Construct a sequence of functions f
n
in C(R), with f
n
1, for which
sup
n
f
n
does not exist in C(R)that is, whenever g C(R) satises g f
n
for all n, then there is exists h C(R) such that h g, h ,= g and h f
n
for each n.
Proof. Consider the sequence of f
n
dened by
f
n
(x) =
_

_
n(x + 1) if 1 x (1 + 1/n)
1 if [x[ (1 1/n)
n(x 1) if (1 1/n) x 1
0 if [x[ 1
.
Now suppose that there exists g C(R) such that g f
n
for all n. WLOG
we can assume that g vanishes at some point and it is clear that g[[1, 1] 1,
so by the Intermediate Value Theorem (Lemma 1.2) we can nd x R such
that g(x) = 1/2. Then h = g((0g) 1) satises the desired properties, and
hence sup
n
f
n
does not exist in C(R).
2. Construct a sequence of functions f
n
in C(R) for which sup
n
f
n
exists in
C(R), but is not the pointwise supremumthat is (sup
n
f
n
)(x) ,= sup
n
f
n
(x)
for at least one x.
Proof. Consider the sequence f
n
= 1 (n[i[). Clearly sup
n
f
n
= 1, yet
sup
n
f
n
(x) =
_
1 if x ,= 0
0 if x = 0
.
3N p.51 the lattice C(X)
Let X be a completely regular space.
1. Let f 0 in C(X) be given. If
g = sup
nN
(1 nf)
exists in C(X), then g is 1 on pos f and 0 on X cl pos f.
55
Proof. First we dene the functions f
n
= 1 nf. Now suppose that x
pos f, then there exists n N such that n > 1/f(x) > 0. Hence for each
m > n we have f
m
(x) = 1, hence g(x) 1. Clearly for each n N and
x X cl pos f we have f
n
(x) = 0, hence g(x) 0.
Note that 1 is an upper bound for each f
n
, hence for x pos f we
have 1 g(x) 1, hence g(x) = 1. And for any x X cl pos f, since
X is completely regular, there exists a continuous 0 h 1 such that
h(x) = 0 and h[cl pos f] = 1. Again, h is an upper bound for each f
n
so
0 g(x) h(x) = 0, hence g(x) = 0.
2. Let V be an open set, and let B denote the family of all functions 1
in C that vanish on XV . If f = sup B exists in C, then f is 1 on V and 0
on X cl V .
Proof. Since X is completely regular, every open set is a union of cozero-
sets (3.2). Hence for every x V , there is a cozero-set V
x
V containing
x. Hence there exists a non-negative g C(X) such that V
x
= pos g, and
therefore the function b = (g/g(x)) 1 is in B, so f(x) 1. But since 1
is an upper bound for B, we have 1 f(x) 1, hence f(x) = 1. Also, for
each x X cl V , since X is completely regular, we can nd a continuous
0 h 1 such that h(x) = 0 and h[cl V ] = 1. Again, h is an upper bound
for B, and for each b B we have b(x) = 0, hence 0 f(x) h(x) = 0 or
f(x) = 0.
For the following questions, let f

be a family of functions in C, and

for r R, dene
U
r
= cl
_

x [ f

(x) > r.
3. If g C, and g f

for every , then, for each x,

g(x) supr [ x U
r
.
Proof. Suppose that for some x X we have g(x) < supr [ x U
r
, then
we can nd a real s, such that g(x) < s < supr [ x U
r
. By continuity we
know that there is a neighborhood V of x such g[V ] ], s[, which implies
that for each , f

[V ] ], s[. Which in turn implies that x , U

s
, and
therefore s supr [ x U
r
. This contradiction completes the proof.
56
4. If X is basically disconnected (cf. 1H), and if f

is a countable
family, then each U
r
is open.
Proof. A countable union of cozero-sets is a cozero-set, hence

x [ f

(x) >
r is also a cozero-set. In a basically disconnected set, the closure of a cozero-
set is open. So each U
r
is both open and closed.
5. X is basically disconnected i every countable family with an upper
bound in C has a supremum n C.
Proof. Necessity Assume that X is basically disconnected. Consider any
countable family of functions f

C. First, recall Lemma 3.12 and note

that the U
r
, as dened above, satisfy the desired conditions (4.) if we reverse
the order of inclusion and replace inf with sup in the Lemma. So let us dene
f as in Lemma 3.12, for each x,
f(x) = supr [ x U
r
.
Therefore f is continuous. Now we want to show that f is an upper bound
for f

. Suppose that for some x, we have f(x) < f

(x). Then we can

write f(x) < s < f

(x) for some s R. Then x U

s
, which implies that
s f(x) and produces a contradiction. Hence f is an upper bound of f

,
but from (3.) we know that it also must be the least upper bound. Therefore
any countable subfamily of C has a supremum in C.
Suciency Assume that every countable subfamily of C with an upper
bound has a supremum in C. Take a cozero-set V , say such that there is a
non-negative f C such that V = XZ(f), and any open set U disjoint
from it. Now dene the functions f
n
= 1 nf; by assumption they must
have a supremum g = sup
n
f
n
in C. From (1.) we know that g[V ] = 1 and
g[U] = 0. Since U and V are completely separated, they must be contained
in disjoint zero-sets and hence have disjoint closures. But according to 1H.1,
we have just proved that X is basically disconnected.
6. X is extremally disconnected i every family with an upper bound in
C has a supremum in C.
Proof. Necessity Assume that X is extremally disconnected. Recall that, in
an extremally disconnected space, open sets have open closures. So, similarly
57
to (4.), but this time for an arbitrary subfamily f

of C, the U
r
are open
sets. The rest of the argument is exactly the same as in the rst part of (5.).
Suciency Assume that every subfamily of C with an upper bound has
a supremum in C. Now take two disjoint open sets U and V . Let B be the
family of all functions 1 that vanish on XV , then, by the assumption, we
have g = sup B in C. Now, according to (2.) we have that g[V ] = 1 and
g[U] = 0. Since these two sets are completely separated, they are contained
in disjoint zero-sets and hence must have disjoint closures. But according to
1H.1, we have just proved the at X is extremally disconnected.
3O p.52 totally ordered spaces
Let X be a totally ordered set (of more than one element). We make X into
a topological space by taking as a subbase for the open sets the family of all
rays x [ x > a, denoted by ], a[, and x [ x < b, denoted by ]b, [.
Consistent with notation, basic open intervals are denoted as ]a, b[ and closed
ones as [a, b].
A nonempty subset S of X is called an interval of X if whenever an
element x of X lies between two elements of S, then x S. When an
interval is an open set, it is called an open interval. For example, the set of
all positive rationals less than

2 is an open interval in the totally ordered

space Q.
The topology on X is called the interval topology, because the open in-
tervals form a base.
1. Every open set is expressible in a unique way as a union of disjoint
maximal open intervals.
Proof. Given an open set, we dene an equivalence relation on it as follows:
x y i ]x, y[ U or ]y, x[ U. Taking equivalence classes, we see that
each of them is clearly an interval, open, and equivalence classes are always
disjoint. Now suppose that we can write an open set as a union of disjoint
open intervals such that no union of a subset of these intervals is itself as
interval (i.e. they are maximal). Then taking equivalence classes with respect
to the same equivalence relation as above, it is clear that the equivalence
classes are the same as the maximal open intervals.
58
2. X is a Hausdor space.
Proof. Let a < b be distinct points of X. Then there are two cases:
(a) There exists c ]a, b[. Then a and b have disjoint neighborhoods,
], c[ and ]c, [, respectively.
(b) a and b are consecutive elements. Then a and b also have disjoint
neighborhoods, ], b[ and ]a, [, respectively.
3. For any nonempty subset A, if sup A exists, then sup A cl A.
Proof. By denition, for any neighborhood ]u, v[ of sup A, there must be an
element a A such that u < a sup A. But these neighborhoods form the
neighborhood base at sup A. Hence every neighborhood of sup A contains
points of A, in other words sup A cl A.
4. For A X, the relative topology on A contains the interval topology,
but the two need not be the same.
Proof. The subbasic sets of the interval topology on A are of the form x
A [ x < a and x A [ x > b for some a, b A. But since A X, these
are also some of the subbasic sets of the subspace topology on A. Which
have the form A ], a[ and A ]b, [ for some a, b X.
The two, however, need not be the same. Consider X = Rx [ [x[ < 1.
In the relative topology on X, the point 0 is an isolated point. But in its
interval topology X is homeomorphic to R.
5. If A is an interval of X, then the relative topology on A does coincide
with the interval topology.
Proof. Take a subbasic open set of the relative topology on A, say B = x
A [ x < b (resp. x > b) for some b X. We know that either b A or
b , A. In the latter case either B = or B = A, which are both open sets
in the interval topology on A. In the former case B is a subbasic open set
of the interval topology itself. So the interval topology on A contains the
relative topology on A, and, by (4.), the reverse also holds. Therefore the
two topologies coincide.
59
6. X is connected i X is Dedekind-complete (0.6) and has no consecu-
tive elements.
Proof. Necessity Assume that X is connected. Suppose that a, b X are
consecutive elements, assuming that WLOG a < b. Then X is disconnected,
since we can write it as a union of two disjoint open sets, namely X =
], b[ ]a, [. Hence X cannot have consecutive elements. Now, suppose
that X is not Dedekind-complete. Then, there must be a subset B A such
that B has an upper bound but no sup in A. First we construct the set
L =

bB
], b[, then we let U be the set of all upper bound of B (which
are all also upper bounds of L), which we can write as U =

uU
]u, [.
Clearly X = L U and by construction, both L and U are non-empty,
disjoint and open. Which implies that X is disconnected. Hence X must be
Dedekind-complete.
Suciency Assume that X is Dedekind-complete and has no consecutive
elements. Suppose that X is disconnected, then we can write X = A

,
where A

and B

are non-empty, open and disjoint. Since A

is open, from
(1.) we know that we can write it uniquely as a union of disjoint maximal
intervals. Let A be one of these intervals, and let B = B

(A

A), the
set B is clearly still open. Every point of B bounds A either from below
or from above. WLOG assume that B contains points that bound A from
above, in fact let U be the set of all upper bounds of A that are contained in
A. Since A is Dedekind-complete, it must have a sup, and U must have an
inf. Suppose sup A ,= inf U, then sup A and inf U are consecutive elements,
hence we must have sup A = inf U. Since A and U are disjoint, we must have
either sup A A or sup A U, WLOG assume that a = sup A A. Since A
is open, there must be a neighborhood of the form ]c, d[, for some c, d X,
around a contained in A. But any such neighborhood will contain points of
U, hence A cannot be open as per the original assumption. Hence X cannot
be disconnected.
7. X is compact i it is lattice-complete (0.5). Thus, X is compact i it
is Dedekind-complete and has both a rst element and a last element.
Proof. Necessity Assume that X is compact. Consider the family of closed
sets of the form [x, [, for each x X. This family has the nite inter-
section property, hence the intersection

xX
[x, [ is non-empty, since X is
compact. Every point in that intersection satises the properties of sup X,
60
hence there is only one point in the intersection, and it is the last element of
X. Similarly, X has a rst element.
Now, let S X be any subset. Consider the family of closed intervals
[x, [ [ x S ], x[ [ x is an upper bound of S. Once again, this
family of closed sets has the nite intersection property, so its intersection
is non-empty, since X is compact. By arguments similar to those of the
paragraph above, that intersection must contain a single point which is sup S.
So every subset of X has a sup and similarly an inf.
Suciency Assume that X is lattice-complete. Let T be a family of
closed subsets of X with the nite intersection property. Let A be the set
of all a X such that [a, [ meets the intersection of every nite subfamily
of T . Note that A is non-empty, since at least inf X A. Now, take F to
be the intersection of some nite subfamily of T . The set F is closed, by
assumption, inf F must exist in X, and by (3.) sup F F. Since [inf F, [
contains F, it must also meet every other intersection of a nite subfamily
of T , hence inf F A. Also, note that sup F b = sup A, for if a > sup F,
for some a A, we would have [a, [ F = which is contrary to the
construction of A. So for every such F, we have b = sup A inf F and
b sup F (also recall that, by (3.), sup F F since it is closed), which
basically means that b is in F. But this implies that b

T . Hence X is
compact.
8. X has a totally ordered compactication. Hence X is completely reg-
ular.
Proof. We know that X has an essentially unique Dedekind completion (0.6),
which can be augmented with inf X and sup X to create an essentially unique
lattice completion X

of X. Since X

is lattice complete, it must be compact

(7.). The space X, as a subspace of X

, must be completely regular since

it is a subset of a compact space (3.14). Now it suces to show that the
interval topology on X is the same as its subspace topology from X

.
Consider a subbasic open set of the subspace topology on X, say ], b[
X (resp. ]b, [ X) for some b X

. Then there must exist a subset

B X such that b = sup B (resp. b = inf B). But then, we can write B =

xB
], x[ (resp. B =

xB
]x, [), hence the interval topology contains
the subspace topology on X, and since the subspace topology automatically
contains the interval topology on X (4.), the two topologies must coincide.
Hence X is completely regular and has a totally ordered compactication.
61
9. X is normal.
Proof. Given the disjoint closed sets H, K X, we want to construct a
function f C(X) such that f[H] = 0 and f[K] = 1. First we dene
f(x) = 0 if x H and f(x) = 1 if x K. Note that the set X(H K) is
open, so it must be composed of disjoint maximal open intervals (1.), let A
be one of these intervals. Then there are several possibilities.
First, consider the trivial possibilities. Suppose that A is disconnected
from XA, then we can set f[A to any continuous function since cl A = A.
Also, suppose that A is disconnected from the set of lower (or upper) bounds
of A (or if the set of bounds is empty), but not disconnected from XA.
Then inf A H K must exist and cl A = A inf A, and we can set
f(x) = f(inf A) for all x A, since f is already dened for inf A (resp.
sup A).
Then, suppose that A is not disconnected from XA, so both extrema
m
1
= inf A, m
2
= sup A HK exist and we have cl A = Ainf A, sup A.
Then f is already dened for both m
1
and m
2
and there are two remaining
cases. If f(m
1
) = f(m
2
), then we simply dene f(x) = f(m
1
) for every x
A. And if f(m
1
) ,= f(m
2
), then, since X is completely regular, we can nd a
continuous function g such that g(m
1
) = f(m
1
) and g[[m
2
, [] = f(m
2
).
Then we can dene f[A = g[A.
Finally, f is well dened and is such that f[(H K) is continuous and
f[ cl(X(H K)) is also continuous. So, by 1A.3, f is continuous as well as
f[H] = 0 and f[K] = 1. Therefore X is normal.
3P p.53 convergence of z-filters
For this problem we need a few denitions for a completely regular space X.
A point p X is said to be a cluster point of a z-lter T if every neighborhood
of p meets every member of T. The z-lter T is said to converge to the limit
p X if every neighborhood of p contains a member of T. We also dene
the following notation: the z-lter on a completely regular space X that is
composed of all the zero-sets containing a given point p is denoted by A
p
(in
fact A
p
is a z-ultralter (3.18)).
Finally consider a small lemma:
Lemma 3.3. Let X be a topological space, and T any z-ler on X. Then T
is an intersection of prime z-lters.
62
Proof. According to Theorem 2.8, every z-ideal in C(X) is an intersection
of prime ideals. Note that Z

[T] is a z-ideal, hence we can write Z

[T] =

, where each P

is a prime ideal. But according to 2.12(a), each Z[P

]
is a prime z-lter. Hence we can write T =

Z[P

].
Let T be a z-lter on a completely regular space X, and let p be a cluster
point of T.
1. T converges to p i T is contained in a unique z-ultralter.
Proof. Necessity If T is a z-lter converging to p, the A
p
is the unique z-
ultralter containing T (3.18(d)).
Suciency Assume that T is contained in a unique z-ultralter. Accord-
ing to Lemma 3.3, T is an intersection of all the prime z-lters containing
it. Suppose T could be written as the intersection of more than one prime
z-lter, then, since every prime z-lter is contained in a unique z-ultralter
(2.13), T must be contained in more than one z-ultralter. But, by assump-
tion, T is contained in a unique z-ultralter, hence T must be prime. Finally,
since T is prime and p is its cluster point, by Theorem 3.17, T converges to
p.
2. If X is compact and p is the only cluster point of T, then T converges
to p.
Proof. Since X is compact, for each z-lter the intersection of all its elements
is non-empty (since a z-lter has the nite intersection property). Hence, by
3.16, every z-lter on X has a cluster point. So every z-ultralter would
have a cluster point and converge to it (3.16(b)), then, according to 3.18,
every z-ultralter has the form A
p
. If T was contained in more than one
z-ultralter, then it would have more than one cluster point (3.18(a)). So
T is contained in the unique z-ultralter A
p
, and therefore it converges to p
(1.).
63
4 Fixed ideals. Compact space
in the sequel, all given spaces are assumed to be completely
regular.
4A p.60 maximal ideals; z-ideals
1. Maximal xed ideal coincides with xed maximal ideal, and maximal
free ideal with free maximal ideal.
Proof. A xed maximal ideal is a maximal ideal which is xed, and a maximal
xed ideal is a xed ideal which is maximal with respect to the properties
of being xed and proper. Clearly a xed maximal ideal is also a maximal
xed ideal. Suppose I is any maximal xed ideal. Then

Z[I] is non-empty,
hence Z[I] has at least one cluster point. So, by 3.18(a), I is contained in
an ultralter converging to that point (hence a xed ultralter). Taking the
preimage of that ultralter we end up with a xed maximal ideal containing
I. Hence a maximal xed ideal is also a xed maximal ideal.
Taking similar denitions for free maximal ideal and maximal free ideal.
Again, it is clear that a free maximal ideal is a maximal free ideal. Now, take
I to be a maximal free ideal. It must be contained in some maximal ideal,
which must also be free, since it contains I. Hence a maximal free ideal is
also a free maximal ideal.
2. C and C are semi-simple (i.e. the intersection of all maximal ideals
is (0).
Proof. Consider any f C such that f ,= 0, and its zero-set Z = Z(f).
Take a point x XZ, then, by complete regularity of X, there exists g C
such that g(x) = 0 and g[Z] = 1. Note that f
2
+ g
2
is a unit of C, so f
cannot belong to the same ideal as g. So any maximal ideal that contains g,
does not contain f. But 0 is in every maximal ideal, so the intersection of
all maximal ideals is (0) or C is semi-simple.
Consider f C

, such that f ,= 0, and, for a xed 0 < < sup [f[[X], the
set E = x X [ [f(x)[ . Suppose that Z(f) = , then f cannot be in
any xed maximal ideal, but we know that xed maximal ideals always exist
(4.4). If Z(f) ,= , then we can pick a point x XE, and, by complete
regularity, nd g C

such that g(x) = 0 and g[E] = 1. Then f

2
+g
2
is a
64
unit of C

, so f cannot belong to the same (maximal) ideal as g. But 0 is in

every maximal ideal, so the intersection of all maximal ideals is (0) and C

is semi-simple.
3. Prove directly that M
p
(where for p X, M
p
= f C [ f(p) = 0)
is a maximal ideal in C and so is M

p
(M

p
= f C

[ f(p) = 0) in C

.
Proof. Consider f, g M
p
, then f(p) = g(p) = 0 and so (f + g)(p) =
f(p) + g(p) = 0, hence f + g M
p
. Consider f M
p
and g C, then
f(p) = 0 and so (fg)(p) = f(p)g(p) = 0, hence fg M
p
. Hence M
p
is
an ideal in C. Now, consider f , M
p
and the ideal I = (M
p
, f), note
that p , Z(f). Since X is completely regular, we can nd g C such that
g(p) = 0 and g[Z(f)] = 1, note that g M
p
I. But then f
2
+ g
2
is a
unit in C, so I cannot be proper. Therefore M
p
is a maximal ideal in C.
By arguments similar to those of the previous paragraph, we can say that
M

p
is an ideal in C

. Now, take f , M

p
and the ideal I = (M

p
, f), note
that p , Z(f). Also, consider the set E = x X [ [f(x)[ for some
0 < < [f(p)[. Since X is completely regular, we can nd g C

such that
g(p) = 0 and g[E] = 1, note that g M

p
. But then f
2
+ g
2
is a unit in
C

, so I cannot be proper. Therefore M

p
is a maximal ideal in C

.
4. Either in C or in C

, if f belongs to every maximal ideal that g belongs

to, then Z(g) Z(f).
Proof. Suppose that there exists x Z(g) such that x , Z(f). Then g
M
x
(resp. M

x
) but f , M

x
), so by the hypothesis, we must have x Z(f).
Therefore Z(g) Z(f).
5. The following algebraic condition is necessary and sucient that an
ideal I in C be a z-ideal: given f, if there exists g I such that f belongs
to every maximal ideal containing g then f I.
Proof. Necessity Assume that I C is a z-ideal. Let f C and suppose
that there exists g I such that f belongs to every maximal ideal containing
g. Then, by (4.), we have Z(g) Z(f), which implies that Z(f) Z[I],
which in turn implies that f I, since I is a z-ideal.
Suciency Assume the condition in question. Let Z Z[I], say Z =
Z(g), for some g I. Now, consider any f C such that Z(f) = Z. Since
65
every maximal ideal is a z-ideal, if g belongs to one, then so does f. So, by
the hypothesis, we have f I, which implies that I is a z-ideal.
4B p.60 principal maximal ideals
1. A point p of X is isolated i the ideal M
p
(resp. M

p
) is principal.
Proof. Necessity Assume that p is isolated, then we can dene the function
f C (resp. C

) to be f(p) = 0 and f(x) = 1 for any x ,= p. Then, for

any g C (resp. C

) such that g(p) = 0, we can write g = fg. Hence the

maximal ideal M
p
= (f) (reps. M

p
) is principal.
Suciency Assume that M
p
(resp. M

p
) is principal and given by (f).
First, note that Z(f) must be p. For if we could nd q Z(f) such that
q ,= p, then (f) would also be contained in M
q
,= M
p
(resp. M

q
) (3.18(c))
and (f) would not be maximal. Since (f) is a maximal idea, it must be
a z-ideal. So it must also contain f
1/3
, since Z(f
1/3
) = Z(f). But then
we can write f
1/3
= gf, for some g C (resp. C

). Now, suppose that p

is not isolated, that is every neighborhood of p contains points other than
p. then the function g = f
2/3
restricted to Xp is unbounded in every
neighborhood of p, and hence g cannot be continuous (much less bounded)
on X. This contradiction completes the proof.
2. X is nite i every maximal ideal in C (resp. C

) is principal.
Proof. Necessity Assume that Xis nite, then clearly C(X) = C

(X). But,
by 2F.3, every ideal in C(X) (and hence C

(X)) must be principal.

Suciency Assume that every maximal ideal in C (resp. C

) is principal.
Then, by (1.), we have that X is discrete. It remains to be shown that X
must also be compact. So, since X is discrete, it can be compact only if it is
nite.
Case C: Let M be a maximal ideal in C. Since it is maximal, it must also
be principal, or M = (f) for some f C. But then

Z[M] = Z(f),
which must be non-empty. So every maximal ideal in C is xed, which
implies that X is compact (Theorem 4.11).
Case C

: Let M be a free maximal ideal in C

. Since it is maximal, it must

also be principal, or M = (f) for some f C

. But then

Z[M] =
Z(f) = . Now, consider the function f
1/3
, we can write f = (f
1/3
)
3
,
66
so f
1/3
M, since a maximal ideal is also prime (0.15). Because
M = (f), we can write f
1/3
= gf, for some g C

. But g must
be f
2/3
, which is continuous because f is nowhere zero, but clearly
unbounded. This contradicts the hypothesis, so every maximal ideal in
C

is xed. Again, by Theorem 4.11, this proves that X is compact.

4C p.61 finitely generated ideals
1. Every nitely generated ideal in C is xed.
Proof. Consider the ideal I = (f
1
, . . . , f
n
) in C, then the family of zero-
sets Z(f
1
), . . . , Z(f
n
) forms a subbase for the z-lter Z[I]. Therefore

Z[I] =

n
i=1
Z(f
i
), which is non-empty because Z[I] is a z-lter and is
closed under nite intersections.
2. A necessary and sucient condition that every nitely generated ideal
in C

be xed is that X be pseudocompact.

Proof. Necessity Assume that every nitely generated ideal in C

is xed,
but X is not pseudocompact (that is C ,= C

). Then C must contain an

unbounded unit u that is bounded away from 0, say u = [f[ 1 where f is
an unbounded function in C. Note that u
1
C

, but u , C

, hence (u
1
)
is an ideal. This ideal is nitely generated and also free, since Z(u
1
) = ,
which contradicts the hypothesis. Hence X must be pseudocompact.
Suciency Assume that X is pseudocompact, that is C

= C. Then,
directly from (1.), we know that every nitely generated ideal in C

is xed.
3. If X is innite, then both C and C

contain xed ideals that are not

nitely generated.
Proof. Since X is innite and Hausdor, it must contain a discrete countably
innite subset (0.13), say N. Let p N and consider the family T of zero-
sets such that, for each Z T, the zero-set Z contains p and ZN has nite
complement in N. Clearly T is a z-lter on X. Note that T is dierent from
the z-lter of all zero-sets that contain N. Suppose that F is a nite subset
of N, note that, since N is discrete, cl F does not meet NF and cl NF does
67
not meet F. Then there exists a cozero-set containing F and disjoint from
NF, by 3B.4, and hence there exists a zero-set Z such that Z N = NF.
As a consequence, we have

T N = p.
If

T = Z, for some Z Z(X), then T is also dierent from the

principal z-lter generated by Z. Since N is countably innite, we can write
N = LM, where L and M are disjoint and both countably innite. Since N
is discrete, we can say that cl LM = and cl ML = , then again there
exists a zero-set Z

such that Z

N = L (3B.4). Clearly Z

= Z Z

is in
the principal z-lter generated by Z, but Z

is not in T, since N(NZ

) =
Mp is innite.
Consider the z-ideal J = Z

[T] in C (resp. C

), clearly J is a xed ideal

since p

T. If J is nitely generated, then T must have a nite a nite
subbase. In other words T is principal, say

T = Z. But from the previous

paragaph, we know that this is impossible. This contradiction completes the
proof.
4D p.61 functions with compact support
The support of f, denoted by o(f), is, by denition, the closure of XZ(f).
Let C
K
(X) denote the family of all functions in C having compact support.
1. If X is compact, then C
K
= C; otherwise, C
K
is both an ideal in C
and an ideal in C

.
Proof. If X is compact, then, since the support of any function is a closed
set and hence compact, every function in C is also in C
K
or C = C
K
.
Assume that X is not compact. Then take any f, g C
K
, the support
of f +g is a closed subset of o(f) o(g) (which is clearly compact). So the
support of f + g is also compact. Take any f C
K
and g C, the support
of fg is a closed subset of o(f) o(g) (which is clearly compact). So the
support of fg is also compact. The support of a unit in C is X, and hence
cannot be compact. Therefore C
K
contains not unit of C and satises the
properties of an ideal in C. Since every function in C
K
is non-zero only on a
subset of a compact set, it must be bounded. Hence C
K
C

= C
K
, but any
ideal in C intersected with C

is an ideal in C

. Hence C
K
is also an ideal
in C

.
68
2. C
K
(Q) = (0).
Proof. Consider any non-zero f C(Q), then, for any x XZ(f), the set
o(f) is a closed neighborhood of x. Suppose that o(f) is compact, then x has
a compact neighborhood. But no point of Q has a compact neighborhood.
This contradiction completes the proof.
To verify the claim of the previous paragraph, consider a point x of Q, and
suppose that it has a compact neighborhood K. Since K is a neighborhood
of X, it must contain an open interval ]a, b[ with a < x < b and a, b Q. But
then we can always nd an irrational i such that a < i < b, and construct a
function which is unbounded in every neighborhood of i, yet still continuous
on Q and K (cf. 1C.5, Case C(Q)). But it is impossible to construct an
unbounded function on a compact set. This contradiction completes the
proof.
3. C
K
is a free ideal i X is locally compact but not compact.
First, recall that a space X is said to be locally compact if every point of
X has a compact neighborhood.
Proof. Necessity Assume that C
K
is a free ideal, then X cannot be compact
(Theorem 4.11). Also, for each x X there must exist f C
K
such that
f(x) ,= 0. Therefore each x X has a compact neighborhood, namely o(f).
So X is locally compact.
Suciency Assume that X is locally compact but not compact, so that
C
K
is a (proper) ideal in C. Consider any x X, by hypothesis, it must
have a compact neighborhood. Since X is completely regular, the cozero-sets
form a base for the open sets of X (3.2), so there exists f C such that
XZ(f) is a subset of the compact neighborhood of x and contains x. Hence
o(f) must also be compact, since it is a closed subset of a compact set, and
f C
K
. Therefore, for each x X, there exists f C
K
such that f(x) ,= 0,
hence C
K
is free.
4. And ideal I in C of C

is free i, for every compact set A, there exists

f I having no zeros in A.
Proof. Necessity Assume that I is free, then for every x X, there exists
f
x
I such that XZ(f
x
) is a neighborhood of x. Let A be a compact set,
then consider a cover of A consisting of sets of the form XZ(f
x
) for each
69
x A. Since A is compact, there must exists a nite subcover of A, say the
complements of Z(f
x
1
), . . . , Z(f
xn
). Then the function g = f
2
x
1
+ +f
2
xn
is
also in I and has no zeros in A.
Suciency Assume that, for every compact A X, there exists f I
having no zeros in A. Note that, since X is Hausdor, the singleton x is
compact. So for every x X there is a function f I such that f(x) ,= 0.
Therefore I is free.
5. C
K
is contained in every free ideal in C and in every free ideal in C

.
Proof. Let I be any free ideal in C or C

. let f C
K
, then, by (4.), there
exists g I such that g does not vanish on o(f). But this implies that
Z(g) int Z(f). So, by 1D.1, we can nd h C

(since o(f) is compact)

such that f = gh. So f I or C
K
I.
4E p.61 free ideals
1. Let f C

. If f belongs to no free ideal in C

, then Z(f) is compact.

But the converse is false. Compare Lemma 4.10.
Proof. If f belongs to no free ideal, then Z(f) belongs to no free z-lter on
X. So, according to Lemma 4.10, the zero-set Z(f) must be compact.
Now, consider the case X = N, I = (j) and f C

given by the char-

acteristic function
N\{1}
(the characteristic function
A
of the set A is de-
ned to be 1 for every point in A and 0 for every point outside it). Then
Z(fj) = Z(f) = 1, which is compact, but fj I, which is free.
2. The intersection of all free maximal ideals in C coincides with the set
of all f in C for which Z(f) meets every noncompact zero-set.
Proof. Since we are only dealing with z-ideals in this question, we can talk
about z-lters instead. All the results are equally valid for the corresponding
z-ideals.
According to Theorem 4.10, a zero-set is noncompact i it belongs to
at least one free z-lter. Hence every element of a free z-ultralter must
be noncompact, and every noncompact zero-set belongs to at least one free
z-ultralter.
70
Suppose that a zero-set intersects every noncompact zero-set, then, for
any free ultralter, it must intersect each one of its elements. Hence, by
Theorem 2.6(b), this zero-set must be in every free z-ultralter.
Suppose that a zero-set is in every free z-ultralter, then, for any free
z-ultralter, it must intersect each of its elements. Hence it must intersect
every noncompact zero-set.
3. A z-lter is a base for the closed sets i it is free.
Proof. Necessity Assume that a z-lter is a base for the closed sets. Since X
is completely regular, in order to have a free z-lter, X cannot be compact,
which also implies that it must have have more than one point. So given two
distinct points of X, by complete regularity, they must have disjoint zero-set
(closed) neighborhoods, say H and K. By hypothesis, both H and K can be
written as intersections of elements from the z-lter. So since H K = ,
the z-lter must be free.
Suciency Assume that a z-lter is free. Then, according to 3.2, the
lter is a base for the closed sets if for every closed H X, and every
x , H, there is a set F in the lter containing H but not x. Suppose this
is not the case, then there must exist a closed H X and x , H such that
every element of the lter does not contain H or contains x.
Recall that, since X is completely regular, the zero-sets are a base for the
closed sets. So for a given closed set H there exists a zero-set Z
1
containing
it. Also, since the z-lter is free, for a given x , H, it must contain an
element Z
2
which does not contain x. Hence Z
1
Z
2
Z
2
, which is also in
the z-lter, contains H and does not contain x. Hence, by contradiction, the
z-lter must be a base for the closed sets in X.
4. Let S be a compact set X, I a free ideal in C(X), and J the set of all
restrictions f[S, for f I. Then J = C(S).
Proof. First, we show that J is an ideal in C(S). Take any a, b J, then
there must exist f, g I such that a = f[S and b = g[S. Since I is an ideal,
f + g I, hence (f + g)[S = a + b J. Take any a J and b C(S),
then there exist f I and g C(X) such that a = f[S and b = g[S
(recall that S is C-embedded, by 3.11(c)). Since f I, we have fg I, so
(fg)[S = ab J.
71
Note that, by 4D.4, since I is free and S is compact, there must exist
f I such that Z(f) S = . Then f[S J, but f[S is a unit of C(S).
Hence J must be improper or J = C(S).
4F p.61 z-ultrafilters on R that contain no small
sets
Let T denote the family of all closed subsets of R whose complements are of
nite Lebesgue measure (the Lebesgue measure of a set A R is denoted by
(A)).
1. T is a free z-lter.
Proof. Since R has innite measure, any set whose complement has nite
measure must have innite measure. Let A, B T, then R(A B) has
measure (RA)+(RB) which is still nite, hence AB T. Consider
A T and closed B R such that A B. Then RB RA, or (RB)
(RA), which is nite, hence B T. Lastly, , T, since (R) =
(R) = . Therefore T is a z-lter.
Take any x R, then ]x1, x+1[ has nite measure, so R]x 1, x + 1[
T. Therefore

T = or T is free.
2. T is not a z-ultralter.
Proof. Assume that T is a z-ultralter. Consider the set
Z =

_
n=
[2n, 2n + 1].
It and its complement both have innite measure. So if Z

is another zero-set
that is disjoint from Z, we have Z RZ

, hence the complement of Z

has
innite measure and Z

, T. Therefore Z intersects every element of T,

hence it must be an element of T, by Theorem 2.6(b). This contradiction
completes the proof.
72
3. Any z-ultralter containing T contains only sets of innite measure.
Proof. Suppose that | is a z-ultralter containing T and Z | such that
Z has nite measure. Then we can nd an open set U of nite measure
containing Z, so that RU T |. But (RU) Z = , hence | cannot
contain Z.
4G p.61 base for a free ultrafilter
1. A free ultralter cannot have a countable base.
Proof. Suppose that | is a free ultralter with a countable base U
n

nN
.
First note that no U
n
can be nite. For if U
n
, for some n, is nite, then for
each x U
n
we can nd U
mx
such that x , U
mx
. Then V =

xUn
U
mx
is a nite intersection of elements of |, hence it is also in |. But then
V U
n
= |, and that would imply that | is an improper lter.
Consider the set U
1
, since it is innite, we can pick two distinct points
a
1
and b
1
. Now, suppose that we already have 2n distinct points A
n
=
a
1
, . . . , a
n
and B
n
= b
1
, . . . , b
n
such that a
m
, b
m
U
m
, for 1 m n.
Consider the set U
n+1
, since it is innite, U
n+1
(A
n
B
n
) is still innite, so
we can pick two distinct points a
n+1
, b
n+1
U
n+1
(A
n
B
n
). Continuing
by induction, we end up with two disjoint sets A = a
1
, a
2
, . . . and B =
b
1
, b
2
, . . ., both intersecting each U
n
.
Since A and B intersect every element of the base for |, they must also
intersect every element of |. From Theorem 2.6(b) (although this theorem
is about z-ultralters, it can be adjusted to ultralters by considering the
underlying space X under the discrete topology), we must have both A and
B as elements of |. But this would imply that A B = |, making it
an improper lter. This contradiction completes the proof.
2. More generally, a free ultralter, each of whose members is of power
m, cannot have a base of power m.
Proof. Suppose | is a free ultralter which satises the specied conditions.
And suppose that it has a base of power m, namely U

m
(the argument
is exactly the same for any base of power less than m). Recall from (1.), that
no U

can be nite, hence m is at least countably innite.

Consider the set U
0
, since it is innite, we can pick two distinct points
a
0
, b
0
U
0
. Now, consider +1 < m (a successor ordinal) and suppose that
73
we already have the distinct points A

= a
0
, . . . , a

and B

= b
0
, . . . , b

,
such that a

, b

, for 0 . Consider the set U

+1
, since it has
cardinality m and A

has cardinality < m, the set U

+1
(A

) is
still innite. So we can pick two distinct points a
+1
, b
+1
U
+1
(A

).
Next, consider < m (a limit ordinal) and suppose that for each < we
already have distinct points A

= a
0
, . . . , a

and B

= b
0
, . . . , b

such
that a

, b

, for 0 . Let A

<
A

and B

<
B

, then
consider the set U

. Since the cardinality of A

and B

is [[ < m and the

cardinality of U

is m, the set U

(A

) is still innite, so we can pick two

distinct points a

, b

(A

). Continuing by transnite induction,

we end up with two disjoint sets A =

<m
A

and B =

<m
B

, both
intersecting each U

.
Since A and B intersect every element of the base for |, they must also
intersect every element of |. From Theorem 2.6(b), we must have both A
and B as elements of |. But this would imply that AB = |, making
it an improper lter. This contradiction completes the proof.
4H p.62 the mapping
#
Let X and Y be spaces with the same underlying point set, such that the
identity mapping : X Y is continuous.
Before proceeding we should provide another denition. Given two topo-
logical spaces X and Y and a continuous map : X Y , we dene the map

#
acting on subfamilies of Z(X) as follows: given a subfamily T of Z(X),
we have

#
T = Z Z(Y ) [

[Z] T.
1. If T is a z-lter on X, then
#
T = T Z(Y ).
Proof. First, note that, since is a continuous identity map from X to Y ,
any open set in Y is also open in X. Hence the topology on X contains the
topology on Y , which implies that Z(Y ) Z(X). Now, consider Z
#
T.
We already know that Z Z(Y ). We also know that Z =

[Z] T, hence
Z T Z(Y ). Finally, consider Z T Z(Y ). Then

[Z] = Z T,
hence Z
#
T. Therefore
#
T = T Z(Y ).
2. If | is a z-ultralter on X,
#
| need not be a z-ultralter on Y .
74
Proof. Suppose that we are dealing with X = Y = [0, 1], where Y is endowed
with the usual interval topology and X is discrete. Consider a point p Y
and the z-lter T = Z[O
p
] of all zero-set neighborhoods of p in Y . Then
T is a base for a lter T

on X. Note that T

is not maximal, since an

ultralter must contain either p or Xp (Lemma 4.3
2
) and T

contains
neither. Hence there must exist an ultralter |

on X which contains both

T

and Xp.
Recall that the zero-sets on Y are precisely the closed sets (1.10), hence,
by (1.), we know that | =
#
|

is the family of all sets in |

that are closed

in Y . But then | must contain Z[O
p
], hence the only z-ultralter that
contains | is Z[M
p
] (4I.2
3
), which is the principal z-lter generated by p.
However, by construction, | cannot contain p, therefore | cannot be a
z-ultralter.
4I p.62 the ideals O
p
For p X, let O
p
denote the set of all f in C for which Z(f) is a neighbor-
hood of p.
1. O
p
is a z-ideal in C, O
p
M
p
, and

Z[O
p
] = p.
Proof. Take f, g O
p
, then Z(f + g) Z(f) Z(g) which is still a
neighborhood of p, hence f + g O
p
. Take f O
p
and g C, then
Z(fg) = Z(f) Z(g) Z(f) which is a neighborhood of p, hence fg O
p
.
Therefore O
p
is an ideal in C.
Now, take f O
p
and consider any g C such that Z(g) = Z(f). Then
Z(g) is also a neighborhood of p, hence g O
p
. Therefore O
p
is a z-ideal.
If f O
p
, then f(p) = 0, hence f M
p
. Therefore O
p
M
p
. Clearly,
p

Z[O
p
]. Take x X dierent from p, then, by complete regularity,
x and p have disjoint zero-set neighborhoods, hence there exists Z Z[O
p
]
such that p Z but x , Z. Therefore

Z[O
p
] = p.
2. M
p
is the only maximal ideal, xed or free, that contains O
p
.
Proof. First, we show that the z-lter Z[O
p
] converges to p. Let U be any
neighborhood of p, then, by complete regularity, it must contain a zero-set
2
Apologies for the forward reference.
3
Once again, apologies for the forward reference.
75
neighborhood Z of p. But we know that Z Z[O
p
], so Z[O
p
] converges to
p.
According to 3.18(d), since Z[O
p
] converges to p, the unique z-ultralter
containing it is A
p
= Z[M
p
] (cf. 3P). Therefore the unique maximal ideal
containing Z

[Z[O
p
]] = O
p
is M
p
.
3. If O
p
,= M
p
, then O
p
is contained in a prime ideal that is not maximal.
Proof. According to Theorem 2.8, since O
p
is a z-ideal, it must be an in-
tersection of prime ideals. So if O
p
,= M
p
, because O
p
is contained in no
maximal ideal other than M
p
(2.), O
p
must be contained in prime ideal that
is not maximal.
4. If P is a prime ideal in C, and P M
p
, then O
p
P.
Proof. Take f O
p
, then f vanishes on some open neighborhood U of p. By
complete regularity, we can nd g C such that g(p) = 1 and g[XU] = 0.
Note that fg = 0 P, and g , M
p
P, hence since P is prime we must
have f P. Therefore O
p
P.
5. If f M
p
O
p
, then there exists a prime ideal, containing O
p
and f,
that is not a z-ideal (and hence not maximal).
Proof. Consider the function g C dened as:
g(x) =
_
_
_
1/ ln [f(x)[ if 0 < [f(x)[ e
1
1 if [f(x)[ e
1
0 if f(x) = 0
.
Clearly, Z(g) = Z(f), so g
n
(O
p
, f) only if g
n
is a multiple of f. But
(g
n
/f)[(XZ(f)) is unbounded in every neighborhood of Z(f) (cf. 2G.1),
hence it cannot be continuously extended to all of X. Since no power of g
belongs to (O
p
, f), by 0.17, there must exist a prime ideal containing (O
p
, f)
but not g. This prime ideal cannot be a z-ideal since Z(g) = Z(f).
76
6. If f M
p
O
p
, then there exists a prime ideal containing O
p
, but not
f, that is not a z-ideal.
Proof. Consider the function g C dened as:
g(x) =
_
exp(1/[f(x)[) if f(x) ,= 0
0 if f(x) = 0
.
Note that, for each n N, we have lim
x0
exp(1/[x[)/x
n
= 0, in other
words lim
f(x)0
f
n
(x)/ exp(1/[f(x)[) = . So no power of f is a multiple
of g (cf. 2G.1), and f , (O
p
, g). Therefore, by 0.17, there must exist a prime
ideal containing (O
p
, g) and not containing f. This prime ideal cannot be a
z-ideal since Z(f) = Z(g).
7. O
p
is a countably generated ideal i p has a countable base of neigh-
borhoods.
Proof. Necessity The neighborhoods of p form a z-lter, namely the z-lter
Z[O
p
]. So if O
p
is countably generated, say by (f
n
)
nN
, then the zero-sets
Z
n
= Z(f
n
) form a countable subbase of Z[O
p
]. Taking all nite intersec-
tions of elements of the subbase, we end up with a countable base for the
neighborhoods of p.
Suciency Assume that p has a countable base Z
n

nN
of neighbor-
hoods. Then these zero-sets are a base for the z-lter Z[O
p
]. For each
n N, pick f
n
C such that Z(f
n
) = Z
n
. Then (f
n
)
nN
generate the z-ideal
Z

[Z[O
p
]] = O
p
.
8. There exists a countably generated ideal I containing O
p
i p is a
G

-point.
Proof. Necessity Assume that there exists a countably generated ideal I =
(f
n
)
nN
, where each f
n
C, which properly contains O
p
. By (2.), the ideal
I must be contained in the maximal ideal M
p
, hence, for every f I, we
have p Z(f). Since I is generated by the f
n
, the z-lter Z[I] must have
Z(f
n
)
nN
as a subbase. Hence we must have F =

Z[I] =

nN
Z(f
n
).
So since Z[O
p
] Z[I], we must have F

Z[O
p
] = p, by (1.). But
we already know that p F, so F = p. Recall, that each Z(f
n
) is a G

(1.10), hence F = p must be a G

as well, since it is an intersection of

countably many G

s. Therefore p is a G

-point.
77
Suciency Assume that p is a G

point, hence it must be a zero-set, say

p = Z(g) for some g C. According to 1D.1, any f C such that Z(f)
is a neighborhood of p must be a multiple of g. Therefore O
p
(g).
4J p.62 P-spaces
For any X, every maximal ideal in C(X) is prime. When, conversely, every
prime ideal in C(X) is maximal, we call X a P-space. (Recall our blanket
assumption that X is completely regular).
This denition is stated in terms of an algebraic property of C(X); there-
fore, if C(X) is isomorphic with C(Y ), and X is a P-space, then Y is a
P-space.
Before proceeding it is useful to state another denition and a couple of
Lemmas. A ring R is said to be a von Neumann regular ring (or simply a
regular ring, if the meaning is clear from context) if, for every element a R,
there exists b R such that aba = a.
Lemma 4.1. Let R be a ring and P R a prime ideal. Then, R/P is an
integral domain (i.e. a ring with no zero divisors).
Proof. Take a

, b

R/P, such that a

= 0. In other words, if a, b R are

respective representatives for a

and b

, then (a + P)(b + P) = ab + P = P,
that is ab P. Since P is prime, at least one of a or b is in P. Hence at
least one of a

or b

is 0. Therefore R/P is an integral domain.

Lemma 4.2. The property of being (von Neumann) regular is preserved un-
der homomorphisms.
Proof. Consider rings R and Q, and a surjective homomorphism : R Q.
Suppose that R is regular, then take any a Q which we can write as (a

),
for some a

R. Since R is regular, we can nd b

R such that a

= a

.
Let b = (b

), then we can write (a

) = (a

) or aba = a. Therefore Q is
also regular.
Since an integral domain allows the cancellation law (e.g. ab = ac b =
c), an immediate corollary of the above Lemma is that a regular integral
domain is a eld. And now, on to the problem.
The following assertions are equivalent. It will be obvious from several of
them that every discrete space is a P-space.
78
(1) X is a P-space.
(2) For all p X, M
p
= O
p
, i.e. every function in C is a constant on a
neighborhood of p.
(3) Every zero-set is open.
(4) Every G

is open.
(5) Every ideal in C(X) is a z-ideal.
(6) Every ideal is an intersection of prime ideals.
(7) For every f, g C, the ideal (f, g) is the principal ideal (f
2
+ g
2
).
(8) For every f C, there exists f
0
C such that f
2
f
0
= f (i.e. C is a
regular ring).
(9) Every ideal is an intersection of maximal ideals.
(10) Every cozero-set in X is C-embedded.
(11) Every principal ideal is generated by an idempotent.
Proof. Since the number of equivalent statements is quite large, the implica-
tion graph is a bit complicated. It is illustrated below.
(11) (10)
((
Q
Q
Q
Q
Q
Q
Q
Q
Q
Q
Q
Q
Q
Q
Q
Q
(7)

(2)
//
(3)
//

OO
66
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
(4)
//
(5)
//
66
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
(6)
//
(8)
rreeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee
(1)
OO
//
(9)
OO
(1)(2). By 4I.3, if M
p
,= O
p
, for a given p X, then there exists a
prime ideal containing O
p
that is not maximal.
(2)(3). Take any Z Z(X), say Z = Z(f), for some f C(X). From
(2), we know that f is zero on some neighborhood U
p
of p, for every p Z.
Therefore Z is open, since U
p
Z for each p Z.
(3)(4). Since X is Hausdor, for every p X, the singleton p is
compact. Suppose that V is a G

, according to Theorem 3.11(b), this implies

79
that, for each p V , there is a zero-set Z
p
V containing p. So, since
every zero-set is open, V must also be open.
(4)(5). Every zero-set is a G

(1.10), hence every zero-set is open.

Let I C be any ideal, f I, and g C such that Z(g) = Z(f). Then
Z(g) = int Z(g) Z(f), hence g is a multiple of f, by 1D.1. Therefore
g I, and I is a z-ideal.
(5)(6). By Theorem 2.8, every z-ideal is an intersection of prime ideals.
Since every ideal is a z-ideal, every ideal is an intersection of prime ideals.
(6)(8). Consider f C, by assumption, the ideal (f
2
) must be an
intersection of prime ideals. Suppose f
2
belongs to a prime ideal, then f
must belong to the same ideal. Therefore f (f
2
) or there exists f
0
C
such that f
2
f
0
= f.
(8)(1). From (8), we know that C is regular. Suppose that P is a
prime ideal in C, then C/P is also regular, by Lemma 4.2. Hence, from the
corollary to Lemma 4.1, we know that C/P is a regular integral domain and
hence a eld. Therefore P is maximal, in other words X is a P-space.
(5)(7). Since f, g (f, g), we also have f
2
+ g
2
(f, g), hence (f
2
+
g
2
) (f, g). Also, note that Z(f), Z(g) Z(f) Z(g) = Z(f
2
+ g
2
),
hence Z(f), Z(g) Z[(f
2
+g
2
)]. Since (f
2
+g
2
) must be a z-ideal, we have
f, g (f
2
+ g
2
) and (f, g) (f
2
+ g
2
). Therefore (f, g) = (f
2
+ g
2
).
(7)(8). Let f C and g = 0, then, by assumption, we must have
(f) = (f, 0) = (f
2
+ 0
2
) = (f
2
). Therefore there exists f
0
C such that
f
2
f
0
= f.
(1)(9). We already know that (1)(5), hence every ideal in C is a z-
ideal. Then, from Theorem 2.8, every ideal is an intersection of prime ideals.
But, by assumption, all prime ideals are maximal.
(9)(6). A maximal ideal is necessarily prime, hence, by assumption,
every ideal is an intersection of prime ideals.
(3)(10). Let Z be a zero-set and V = XZ its cozero-set. Then,
by assumption, Z is open, hence Z and V are disconnected. Suppose that
f C(V ), then we can dene g C as g[V = f and g[Z = 0, here g is
continuous by Lemma 1.4. Therefore V is C-embedded in X.
(10)(8). Consider f C(X) and let Z = Z(f). By assumption, we
know that V = XZ is C-embedded in X. Consider f

0
C(V ) dened
as f

0
= (f[V )

, we extend this function to a continuous function f

0
on X.
Clearly f
2
f
0
= f.
(3)(11). We already know that (3)(5), hence every ideal is a z-
ideal. So given a principal ideal (f), for some f C, the same ideal can
80
be generated by any function with the same zero-set as f. Consider the
idempotent i dened as i[Z(f) = 0 and i[XZ(f) = 1. By assumption, Z is
open, hence Z and XZ are disconnected, so i is continuous by Lemma 1.4.
Therefore the idempotent i generates the principal ideal (f).
(11)(3). Consider a zero-set Z and the ideal (f), for some f C
with Z(f) = Z. If it is generated by an idempotent i, then we must have
Z(i) = Z(f). The function i is an idempotent i i[X] 0, 1, therefore
f[XZ] = 1. This implies that Z and XZ are completely separated and
thus have disjoint open neighborhoods. But X = Z XZ, hence Z must
be open.
Finally, if X is discrete, then every subset of X is both open and a zero-
set. Therefore, by (3), X is a P-space.
4K p.63 further properties of P-spaces
Please keep in mind the blanket assumption that all spaces under considera-
tion are completely regular. Sometimes this property will not be mentioned
explicitly, for example in the case of subspaces of a completely regular space
(3.1).
1. Every countable subset of a P-space (4J) is closed and discrete. Hence
every countable P-space is discrete, and every countably compact P-space is
nite.
Proof. Consider a countable subset D = x
n

nN
of X and the sequence
of reals 1/n
nN
. Then, according to 3L.2, we can construct a continuous
function f such that f(x
n
) = 1/n, for all n N. From 4J(2), we know that
for every point p X there is neighborhood U
p
around it on which f is
constant. Therefore, for f, the neighborhoods U
xn
are disjoint, each one
containing only x
n
. Hence D is discrete.
Also, from 4J(4), we know that every G

is open. Hence the set XD =

nN
Xx
n
is also open. Therefore D is closed.
Suppose a P-space is countably compact and innite. Then we can nd a
closed discrete countable subset D = x
n

nN
of X (in fact, every countable
subset is such a set). So we can construct a countable cover of X consisting
of U
xn

nN
XD (where the U
xn
are the same as in the previous para-
graphs). Clearly, this cover has no nite subcover. Therefore X has to be
nite.
81
2. Every countable set in a P-space is C-embedded. Hence every pseu-
docompact P-space is nite.
Proof. Consider a countable subset D = x
n

xN
of X and any closed set F
disjoint from it. Since X is completely regular, we can nd pairs of disjoint
zero-sets neighborhoods Z
n
, Z

n
such that x
n
Z
n
and F Z

n
. Then D and
F have disjoint neighborhoods, since

nN
Z

n
is a zero-set (which is open
(4J(3))) and contains F, and

nN
int Z
n
is an open set which contains D.
Therefore, according to 3L.4(3), the set D must be C-embedded.
Suppose a pseudocompact P-space is innite, then it must have a count-
able subset D = x
n

nN
which is discrete (1.). Then we can dene f C(D)
such that f(x
n
) = n, for each n N. So, since every countable subset is
C-embedded, we can nd g C such that g[D = f. In other words C con-
tains an unbounded function, therefore the a pseudocompact P-space must
be nite.
3. If X is a P-space, and every function in C(X) is bounded on a subset
S, then S is nite.
Proof. Suppose S X is innite, but every continuous function is bounded
on it. Then S must have a countable subset D = x
n

nN
. From (2.), we
know that D must be C-embedded. Since D is discrete (1.), we can dene
f C(D) such that f(x
n
) = n for each n N. We can continuously extend
f to g C(X), which is unbounded on S. Therefore S must be nite.
4. Every subspace of a P-space is a P-space.
Proof. Consider S X and U S, where U is a G

in S. We can write
U =

nN
U
n
, where each U
n
is open in S. But for each U
n
, there is an open
U

n
X such that U
n
= S U

n
. By 4J(4), the set U

nN
U

n
is open in
X, hence U = S U

is open in S. Therefore S is a P-space.

5. Every (completely regular) quotient space of a P-space is a P-space.
First, a denition. Let (X, T ) be a topological space and R an equivalence
relation on X. Consider the induced projection map
R
: X X/R dened
by
R
: x Rx, where Rx denotes the equivalence class to which x belongs
82
and X/R the set of such equivalence classes. We say that X/R is a quotient
space when it is endowed with the topology
T
R
= U X/R [

R
[U] T .
And now, on to the proof.
Proof. Consider an equivalence relation R on X and U X/R, where U is
a G

in X/R. We can write U =

nN
U
n
, where each U
n
is open in X/R.
But for each U
n
, we must have U

n
=

R
[U
n
] open in X. By 4J(4), the set
U

nN
U

n
is open in X, hence U =
R
[U

] is open in X/R. Therefore

X/R is a P-space.
6. Finite products of P-spaces are P-spaces, but innite products need
not be.
Proof. Consider the product space X =

A
X

, where A is nite. The

subbase for the open sets of X is given by the sets of the form

[U], where
U is open in X

and

: X X

is the corresponding projection map. So

every open set in X can be expressed as a union of nite intersections of sets
of the form

[U]. A G

is a countable intersection of open sets, distributing

intersections over unions, it can be expressed as a union of countable inter-
sections of sets of the form

[U]. Therefore every G

in X is open if every
countable intersection of sets of the form

[U] is open.
Consider the families of sets U

B
T(X

), where B

is countable,
for each A. Then consider the set
U =

A
_

B

[U

]
_
=

_

B
U

_
=

[U

].
Each U

is a G

in X

, and hence open, by 4J(4). Therefore each

[U

] is
open in X, and U is open as well, since A is nite.
Consider the product space

i=1
[0, 1]. Then the set

i=1
[1/4, 3/4] is a
G

, but clearly not open.

83
7. Every P-space is basically disconnected (1H).
Proof. From 4J(4), we know that every zero-set is open, hence every cozero-
set is closed. Therefore the closure of every cozero-set is itself, which is open.
Hence the space is basically disconnected.
8. Every P-spacemore generally, every basically disconnected space
has a base of open-and-closed (clopen) sets.
Proof. At rst, this question might appear ambiguous, since it is not clear
whether it is referring to the base for closed or open sets. However, this
distinction is immaterial. If there exists a base for the closed sets consisting
of clopen sets, then the family of complements of these sets forms a base for
the open sets consisting of clopen sets.
Since X is completely regular, the zero-sets form a base for the closed
sets. Since X is basically disconnected, every cozero-set is closed (1H.1),
and hence every zero-set is open. Therefore the zero-sets form a clopen base
for the closed sets.
9. Neither C(R) nor C(Q) is a homomorphic image of C(N).
Proof. Since every discrete space is a P-space (4J), according to 4J(4), the
ring C(N) is regular. For C(R) (resp. C(Q)), for the function f(x) = x, it
is not possible to nd g C(R) (resp. C(Q)) such that f
2
g = f. Then, by
Lemma 4.2, it is clear that neither C(R) nor C(Q) is a homomorphic image
of C(N), since they are not regular.
4L p.63 P-points
If M
p
= O
p
, then P is called a P-point of X. Thus, X is a P-space i every
point is a P-point (4J(2)).
1. p is a P-point i every G

containing p is a neighborhood of p.
Proof. Necessity Assume that p is a P-point, then every continuous function
on X is constant on some neighborhood of p. Hence every zero-set containing
p is a neighborhood of p. Since X is Hausdor, the singleton p is compact.
Suppose that V is a G

containing p, according to Theorem 3.11(b), there

is a zero-set Z V containing p. Therefore V is also a neighborhood of p.
84
Suciency Assume that every G

containing p is a neighborhood of p.
Since every zero-set is a G

, every zero-set containing p is a neighborhood of

p. Therefore M
p
= O
p
or p is a P-point.
2. A P-point of X is a P-point in any subspace containing it. Hence the
set of all P-points of X is a P-space.
Proof. Consider p S X, where p is a P-point in X. Every G

in S
is the intersection of S and a G

in X. And any neighborhood of p in X,

intersected with S, is a neighborhood of p in S. According to (1), every G

in X containing p is a neighborhood of p, hence every G

in S containing p
is also a neighborhood of p. Therefore p is a P-point in S.
3. Let f R
X
. If f is continuous at a P-point p, then f is constant on
a neighborhood of p.
Proof. Let r = f(p). Since f is continuous at p, for every n N, we can nd
an open set U
n
X such that f[U
n
] ]r 1/n, r + 1/n[. Then U =

nN
U
n
is a G

containing p and hence a neighborhood of p (1.). But, by construction,

f[U] = r.
4. If p is a P-point and M
p
is countably generated, then p is isolated
and M
p
is principal.
Proof. If M
p
= O
p
is countably generated, then O
p
must be principal, by
4I.8. But M
p
is principal only if p is isolated (4B.1).
4M p.64 the space
Let | be a free ultralter on N, let = N (where , N), and dene a
topology on as follows: all points of N are isolated, and the neighborhoods
of are the sets U for U |.
1. N is a dense subspace of . Every set containing is closed; hence
every subset of is open or closed. is a normal space (cf. 3B.5 and 3D.5);
in fact, every closed set is a zero-set.
85
Proof. Clearly every neighborhood of meets N, hence cl N or cl N = .
Consider any S such that S. Then XS N is open in , since
XS is a union of isolated points. Hence S is closed.
We already have that every set containing is closed. If a set S
does not contain , then it is a subset of N and open, because S is a union
of isolated points. Hence every subset of is open or closed.
First, we show that is Hausdor. Consider two distinct points x, y .
At least one of them, say x, cannot be . Then the singleton x must be
both closed and open, hence it is its own neighborhood. Also, x must
be a neighborhood of y not containing x. Now, consider two disjoint closed
sets H, K . Then at least one of them, say H, does not contain . Since
H is a union of isolated points, it must itself be open. But since H is closed,
we have XH a neighborhood of K. Therefore is normal.
Consider any closed set F , then either F contains or it is of the
form F = XV , where V = U for some U |. In the latter case,
because both F and XF are open, we can dene f[F] = 0 and f[XF] =
1 with f C(), by Lemma 1.4. If F contains and XF is nite, then
again F and XF are both open, and f[F] = 0 and f[XF] = 1 with
f C(). Lastly, if F and XF = x
n

nN
is innite, then we can
dene f[F] = 0 and f(x
n
) = 1/n for each n N. Since the preimage,
with respect to f, of any open interval ]a, b[ R is either nite (and hence
open) or has a nite complement (and hence also open), the function f is
continuous. Therefore any closed subset of is a zero-set.
2. The point does not have a countable base of neighborhoods. Hence
is not metrizable.
Proof. If were to have a countable base of neighborhoods, then the free
ultralter | on N would also have a countable base. But that is impossible,
by 4G.1. If we could impose a metric d on , then every point p
would have a countable base of neighborhoods, namely B
1/n

nN
where
B

= x [ d(p, x) < . Hence is not metrizable.

3. is extremally disconnected (1H), and so is every subspace.
For the following proof we will require a small Lemma concerning ultra-
lters.
Lemma 4.3. A lter T on a set X is an ultralter i, for any S X, the
lter T contains exactly one of S or XS.
86
Proof. Necessity Assume that T is an ultralter, then T is prime because
it is the image, under Z, of some maximal (hence prime) ideal in C(X),
where X is considered under the discrete topology (Theorems 2.5, 2.12).
Note that S (XS) = X T, hence at least one of S or XS must be in
T. Note that both S and XS cannot be in T, since then we would have
S (XS) = T, which would mean that T is not proper.
Suciency Suppose that T is a lter which contains exactly one of S or
XS, for each S X. Take any S X not in T and construct the lter T

by adjoining S to T. But, by hypothesis, we must have XS in T. Hence

both S and XS are in X, so S (XS) = T

and T

cannot be a
proper lter. Therefore T must be an ultralter.
And now, on to the proof.
Proof. Suppose U is an open subset of , then either U or , U. In the
former case, U is already closed, hence cl U = U. In the latter case, either
U is open or it is not. If U is open, again U is closed and cl U = U.
If U is not open, then it does not contain a neighborhood of , which
implies that NU is not an element of the ultralter |. Then, by Lemma
4.3, we must have U |. Since U is not closed, clearly its closure must
be cl U = U . But since U |, we have cl U open. Therefore is
extremally disconnected.
Every subspace of that does not contain is discrete, therefore it is
automatically extremally disconnected. Let Y X such that Y . Then
any open U Y such that U is already closed, hence cl U = U. If
U Y is open but , U, then if it is not closed, its closure must be
cl U = Y (U ). Moreover, it must be of the form U = Y V , for some
V | (see previous paragraph for reasoning). But them U is open,
and hence cl U = Y (U ) is also open. Therefore every subspace of
is extremally disconnected.
4. The z-ideal O

is prime but not maximal. Hence is not a P-space

(4J).
Proof. Consider two closed sets (every closed set is a zero-set (1.)) H, K .
Suppose that Z = H K is a neighborhood of but neither of H or K is.
There are two possibilities, either both H and K contain or only one of
them does, say H. In the former case we must have H = (NU) and
K = (NV ) , for some U, V |. In the latter case H = (NU)
87
and K = NV , for some U, V |. But then U V |, so (NV )(NU) =
N(U V ) is not in |, hence Z = N(U V ) cannot be a neighborhood
of . Therefore at least one of H or K must be a neighborhood of and O

,
since it is a z-ideal, must be prime.
Note that is a closed but not open zero-set. Therefore O

is not
maximal, since M

is strictly greater than O

. Therefore is not a P-point

and is not a P-space.
5. The dense subspace N is C

-embedded in , but not C-embedded.

Moreover, every subspace of is C

-embedded.
Proof. Consider two disjoint zero-sets in H, K N (since N is discrete, any
set is a zero-set). They both cannot be in the ultralter |, since otherwise
it would not be a proper lter. Suppose only one, say H, is in |, then NK
must also be in |, by Lemma 4.3. Note that (NK) is open, so K
is closed (and hence a zero-set) in . Also, H is closed (and hence a
zero-set) in disjoint from K. If, on the other hand, neither H nor K is
in |, then both H and K are closed (and hence zero-sets) in , by similar
reasoning as in the previous case. Hence, by Urysohns Extension Theorem
(1.17), N is C

-embedded in . However, N is not C-embedded since is

a zero-set and the smallest zero-set containing N is itself (Theorem 1.18).
Consider any subspace Y of . Let H, K Y be any two disjoint zero-
sets. Suppose that neither H nor K contains , then they can be treated as
subsets of N and are completely separated in for the same reasons as in
the previous paragraph. Suppose only one zero-set, say H, contains , then
H is automatically closed in (1.). If K , |, again K is already closed in
. If K |, then there must exist a set K

closed in and not containing

such that K = Y K

. Note that H is an open set, hence K

(H)
is a closed set (hence zero-set) in that contains K and is disjoint from H.
Therefore, by Urysohns Extension Theorem (1.17), every subspace of is
C

-embedded.
6. C

() is isomorphic with C

(N). But C() is not isomorphic with

C(N).
Proof. First, let us show some properties of continuous functions on . Con-
sider f C(), then r = f() must be an accumulation point of f[].
88
Suppose that r is isolated from f[N], then there is an open interval U con-
taining r such that f

[U] = . But is not open, hence f cannot be

continuous.
Suppose that f C

(N), then the z-lter f

#
| is prime on the compact
space [inf f, sup f]. Since ever z-lter on a compact space must be xed, the
intersection A =

f
#
| must be non-empty. Note that each a A is an
accumulation point of f[U], for each U |. Suppose there are two distinct
points a, b A (WLOG, let a < b), then we can nd a point c ]a, b[ and
construct the sets U = f

[] inf f, c]] and V = f

[]c, sup f]]. Since U V = N,

exactly one of U or V must be in | (Lemma 4.3). If U is in |, then b is not
a point of accumulation of f[U], so b cannot be in A. If V is in |, then a is
not a point of accumulation of f[V ], so a cannot be in A. Therefore A has
exactly one point.
Since f
#
| is prime and

f
#
| = r, for some r R, then f
#
| must
converge to r, by Theorem 3.17. This implies that, for f to be continuously
extended to all of , its value at must be r.
We already know that f f[N is a surjective homomorphism from C

()
to C

(N). The previous paragraph shows that its is bijective, so since it is

clear that the inverse is also a homomorphism, the two rings C

() and
C

(N) are isomorphic. Finally, C() cannot be isomorphic to C(N), since N

is discrete and hence a P-space (4J), but is not (4.).
7. If 0 h k in C(), then h (k).
Proof. We have h (k) if h = kg for some g C(). Let V = Z(k)
and consider the function g

= (h[V )/(k[V ), clearly (h[V ) = g

(k[V ) and
g

(V ), since by hypothesis h(x)/k(x) 1 for each x V . By (5.),

the subspace V of is C

-embedded, so we can nd g C

() such that
g[V = g

. But then h = kg, hence h (k).

8. The ideal (f, g) in C() is the principal ideal ([f[ +[g[). Hence every
nitely generated ideal is principal.
Proof. Note that, since is extremally disconnected (3.), it is also basically
disconnected. Hence, by 1H.3, there exists a unit u of C such that [f[ = uf,
that is [f[ (f, g) and similarly [g[ (f, g). Which implies that [f[ + [g[
(f, g) and ([f[ +[g[) (f, g).
89
Note that 0 [f[, [g[ [f[ +[g[, hence we have [f[, [g[ ([f[ +[g[) (7.).
From the previous paragraph, we know that we can nd h C

() such that
[f[ = hf, but then we also have f = h[f[. Hence f, g ([f[, [g[) ([f[ +[g[).
Therefore (f, g) ([f[ +[g[) and (f, g) = ([f[ +[g[).
9. The only z-ideals containing O

are O

and M

.
Proof. Suppose that Z Z[M

]Z[O

]. Then Z is closed but not open,

hence Z |. Note that we must also have Z

= (Z) Z[O

].
Let T be any z-lter that contains both Z[O

] and Z, then we must have

Z Z

= T. Which implies that T = M

.
4N p.64 a nondiscrete P-space
Let S be an uncountable space in which all points are isolated except for a
distinguished point s, a neighborhood of s being any set containing s whose
complement is countable.
1. The closed set s is not a zero-set. S is a nondiscrete P-space (4J).
Proof. Suppose that f C(S) and f(s) = 0. Since f is continuous at s, for
each 1/n
nN
, we can nd a neighborhood U

of s such that f[U

]
], [. Note that, for each x Z(f) =

nN
U
1/n
, we have f(x) = 0. But
since the complement of each U
1/n
is countable, the complement of Z(f) is
also countable. Therefore Z(f) itself must be uncountable and Z(f) ,= s.
Since each point p of S, except s, is isolated, we have O
p
= M
p
. From the
previous paragraph, we have that the zero-set of any function which vanishes
at s has countable complement, hence each zero-set containing s is also a
neighborhood of s. Therefore S must be a P-space (4J(2)).
2. No member of Z[M
s
] is countably compact.
Proof. First, a note about the topology on S. Every set that does not contain
s is open, since it is a union of isolated points. Every set that contains s is
closed, because its complement is open. A set that contains s is open i it
contains a neighborhood of s, i.e. its complement is countable. A set that
does not contain s is closed i its complement is open, i.e. it is countable.
Let Z Z[M
s
]. Since Z is uncountable (1.), we can choose a countable
subset x
n

nN
. Then we can construct a countable open cover U
m

mN
of
90
Z, where U
m
= (Zx
n

nN
) x
m
. Each U
m
is open because its comple-
ment is countable. Note that no subset of U
m

mN
is a cover of Z. Therefore
Z cannot be countably compact.
3. S is basically disconnected but not extremally disconnected (1H).
Proof. Since S is a P-space (1.), every zero-set is open (4J(3)) and hence
every cozero-set is closed. Therefore every cozero-set has an open closure
(itself) and hence S is basically disconnected.
Since S is uncountable, we can write Ss = U V , where U and V are
disjoint uncountable sets. Clearly, both U and V are open but not closed,
also cl U = U s and cl V = V s, so the closure of these sets are not
disjoint. But, according to 1H.1, this implies that S cannot be extremally
disconnected.
4. Let T be the topological sum of (4M) and S (i.e. and S are
disjoint open sets whose union is T). Then T is basically disconnected, but
it is neither extremally disconnected nor a P-space.
Proof. Both and S are basically disconnected (4M.3 and (3.)). Since T
is the topological sum of and S, clearly every cozero-set V in T can be
written as V = U
1
U
2
, where U
1
is a cozero-set in and U
2
is a cozero-set
in S. But then cl
T
V = cl
T
(U
1
U
2
) = cl

U
1
cl
S
U
2
. Since cl

U
1
and cl
S
U
2
are both open, the set cl
T
V must be open as well. Therefore T is basically
disconnected.
Recall that S is not extremally disconnected (3.), hence there must exist
open U S such that cl
S
U is not open in S. But then U is open in T and
cl
T
U = cl
S
U, which is also closed but not open in T. Hence T cannot be
extremally disconnected.
References
[GJ] Leonard GillmanMeyer Jerison. Rings of Continuous Functions. D.
van Nostrand Company, Inc., 1960.
[Dug] James Dugundji. Topology. Allyn and Bacon, Inc., 1966.
91
A GNU Free Documentation License
Version 1.1, March 2000
Copyright c _ 2000 Free Software Foundation, Inc.
59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
Everyone is permitted to copy and distribute verbatim copies of this license
document, but changing it is not allowed.
Preamble
The purpose of this License is to make a manual, textbook, or other written
document free in the sense of freedom: to assure everyone the eective
freedom to copy and redistribute it, with or without modifying it, either
commercially or noncommercially. Secondarily, this License preserves for the
author and publisher a way to get credit for their work, while not being
considered responsible for modications made by others.
This License is a kind of copyleft, which means that derivative works
of the document must themselves be free in the same sense. It complements
the GNU General Public License, which is a copyleft license designed for free
software.
We have designed this License in order to use it for manuals for free
software, because free software needs free documentation: a free program
should come with manuals providing the same freedoms that the software
does. But this License is not limited to software manuals; it can be used
for any textual work, regardless of subject matter or whether it is published
as a printed book. We recommend this License principally for works whose
purpose is instruction or reference.
A.1 Applicability and Denitions
This License applies to any manual or other work that contains a notice
placed by the copyright holder saying it can be distributed under the terms
of this License. The Document, below, refers to any such manual or work.
Any member of the public is a licensee, and is addressed as you.
A Modied Version of the Document means any work containing the
Document or a portion of it, either copied verbatim, or with modications
and/or translated into another language.
92
A Secondary Section is a named appendix or a front-matter section of
the Document that deals exclusively with the relationship of the publishers
or authors of the Document to the Documents overall subject (or to related
matters) and contains nothing that could fall directly within that overall
subject. (For example, if the Document is in part a textbook of mathematics,
a Secondary Section may not explain any mathematics.) The relationship
could be a matter of historical connection with the subject or with related
matters, or of legal, commercial, philosophical, ethical or political position
regarding them.
The Invariant Sections are certain Secondary Sections whose titles are
designated, as being those of Invariant Sections, in the notice that says that
the Document is released under this License.
The Cover Texts are certain short passages of text that are listed, as
Front-Cover Texts or Back-Cover Texts, in the notice that says that the
Document is released under this License.
A Transparent copy of the Document means a machine-readable copy,
represented in a format whose specication is available to the general pub-
lic, whose contents can be viewed and edited directly and straightforwardly
with generic text editors or (for images composed of pixels) generic paint
programs or (for drawings) some widely available drawing editor, and that is
suitable for input to text formatters or for automatic translation to a variety
of formats suitable for input to text formatters. A copy made in an other-
wise Transparent le format whose markup has been designed to thwart or
discourage subsequent modication by readers is not Transparent. A copy
that is not Transparent is called Opaque.
Examples of suitable formats for Transparent copies include plain ASCII
without markup, Texinfo input format, L
A
T
E
X input format, SGML or XML
using a publicly available DTD, and standard-conforming simple HTML de-
signed for human modication. Opaque formats include PostScript, PDF,
proprietary formats that can be read and edited only by proprietary word
processors, SGML or XML for which the DTD and/or processing tools are
not generally available, and the machine-generated HTML produced by some
word processors for output purposes only.
The Title Page means, for a printed book, the title page itself, plus
such following pages as are needed to hold, legibly, the material this License
requires to appear in the title page. For works in formats which do not have
any title page as such, Title Page means the text near the most prominent
appearance of the works title, preceding the beginning of the body of the
93
text.
A.2 Verbatim Copying
You may copy and distribute the Document in any medium, either commer-
cially or noncommercially, provided that this License, the copyright notices,
and the license notice saying this License applies to the Document are re-
produced in all copies, and that you add no other conditions whatsoever to
those of this License. You may not use technical measures to obstruct or
control the reading or further copying of the copies you make or distribute.
However, you may accept compensation in exchange for copies. If you dis-
tribute a large enough number of copies you must also follow the conditions
in section 3.
You may also lend copies, under the same conditions stated above, and
you may publicly display copies.
A.3 Copying in Quantity
If you publish printed copies of the Document numbering more than 100,
and the Documents license notice requires Cover Texts, you must enclose
the copies in covers that carry, clearly and legibly, all these Cover Texts:
Front-Cover Texts on the front cover, and Back-Cover Texts on the back
cover. Both covers must also clearly and legibly identify you as the publisher
of these copies. The front cover must present the full title with all words of
the title equally prominent and visible. You may add other material on the
covers in addition. Copying with changes limited to the covers, as long as
they preserve the title of the Document and satisfy these conditions, can be
treated as verbatim copying in other respects.
If the required texts for either cover are too voluminous to t legibly,
you should put the rst ones listed (as many as t reasonably) on the actual
cover, and continue the rest onto adjacent pages.
If you publish or distribute Opaque copies of the Document number-
ing more than 100, you must either include a machine-readable Transparent
copy along with each Opaque copy, or state in or with each Opaque copy a
publicly-accessible computer-network location containing a complete Trans-
parent copy of the Document, free of added material, which the general
network-using public has access to download anonymously at no charge us-
ing public-standard network protocols. If you use the latter option, you must
94
take reasonably prudent steps, when you begin distribution of Opaque copies
in quantity, to ensure that this Transparent copy will remain thus accessible
at the stated location until at least one year after the last time you distribute
an Opaque copy (directly or through your agents or retailers) of that edition
to the public.
It is requested, but not required, that you contact the authors of the
Document well before redistributing any large number of copies, to give them
a chance to provide you with an updated version of the Document.
A.4 Modications
You may copy and distribute a Modied Version of the Document under the
conditions of sections 2 and 3 above, provided that you release the Modied
Version under precisely this License, with the Modied Version lling the
role of the Document, thus licensing distribution and modication of the
Modied Version to whoever possesses a copy of it. In addition, you must
do these things in the Modied Version:
Use in the Title Page (and on the covers, if any) a title distinct from that
of the Document, and from those of previous versions (which should, if
there were any, be listed in the History section of the Document). You
may use the same title as a previous version if the original publisher of
that version gives permission.
List on the Title Page, as authors, one or more persons or entities
responsible for authorship of the modications in the Modied Version,
together with at least ve of the principal authors of the Document (all
of its principal authors, if it has less than ve).
State on the Title page the name of the publisher of the Modied
Version, as the publisher.
Preserve all the copyright notices of the Document.
Add an appropriate copyright notice for your modications adjacent to
the other copyright notices.
Include, immediately after the copyright notices, a license notice giving
the public permission to use the Modied Version under the terms of
this License, in the form shown in the Addendum below.
95
Preserve in that license notice the full lists of Invariant Sections and
required Cover Texts given in the Documents license notice.
Include an unaltered copy of this License.
Preserve the section entitled History, and its title, and add to it an
item stating at least the title, year, new authors, and publisher of the
Modied Version as given on the Title Page. If there is no section
entitled History in the Document, create one stating the title, year,
authors, and publisher of the Document as given on its Title Page, then
add an item describing the Modied Version as stated in the previous
sentence.
Preserve the network location, if any, given in the Document for public
access to a Transparent copy of the Document, and likewise the network
locations given in the Document for previous versions it was based on.
These may be placed in the History section. You may omit a network
location for a work that was published at least four years before the
Document itself, or if the original publisher of the version it refers to
gives permission.
In any section entitled Acknowledgements or Dedications, preserve
the sections title, and preserve in the section all the substance and tone
of each of the contributor acknowledgements and/or dedications given
therein.
Preserve all the Invariant Sections of the Document, unaltered in their
text and in their titles. Section numbers or the equivalent are not
considered part of the section titles.
Delete any section entitled Endorsements. Such a section may not
be included in the Modied Version.
Do not retitle any existing section as Endorsements or to conict in
title with any Invariant Section.
If the Modied Version includes new front-matter sections or appendices
that qualify as Secondary Sections and contain no material copied from the
Document, you may at your option designate some or all of these sections
as invariant. To do this, add their titles to the list of Invariant Sections in
96
the Modied Versions license notice. These titles must be distinct from any
other section titles.
You may add a section entitled Endorsements, provided it contains
nothing but endorsements of your Modied Version by various parties for
example, statements of peer review or that the text has been approved by
an organization as the authoritative denition of a standard.
You may add a passage of up to ve words as a Front-Cover Text, and a
passage of up to 25 words as a Back-Cover Text, to the end of the list of Cover
Texts in the Modied Version. Only one passage of Front-Cover Text and
one of Back-Cover Text may be added by (or through arrangements made
by) any one entity. If the Document already includes a cover text for the
same cover, previously added by you or by arrangement made by the same
entity you are acting on behalf of, you may not add another; but you may
replace the old one, on explicit permission from the previous publisher that
added the old one.
The author(s) and publisher(s) of the Document do not by this License
give permission to use their names for publicity for or to assert or imply
endorsement of any Modied Version.
A.5 Combining Documents
You may combine the Document with other documents released under this
License, under the terms dened in section 4 above for modied versions,
provided that you include in the combination all of the Invariant Sections
of all of the original documents, unmodied, and list them all as Invariant
Sections of your combined work in its license notice.
The combined work need only contain one copy of this License, and mul-
tiple identical Invariant Sections may be replaced with a single copy. If there
are multiple Invariant Sections with the same name but dierent contents,
make the title of each such section unique by adding at the end of it, in
parentheses, the name of the original author or publisher of that section if
known, or else a unique number. Make the same adjustment to the section
titles in the list of Invariant Sections in the license notice of the combined
work.
In the combination, you must combine any sections entitled History
in the various original documents, forming one section entitled History;
likewise combine any sections entitled Acknowledgements, and any sec-
tions entitled Dedications. You must delete all sections entitled Endorse-
97
ments.
A.6 Collections of Documents
You may make a collection consisting of the Document and other documents
released under this License, and replace the individual copies of this License
in the various documents with a single copy that is included in the collection,
provided that you follow the rules of this License for verbatim copying of each
of the documents in all other respects.
You may extract a single document from such a collection, and distribute
it individually under this License, provided you insert a copy of this License
into the extracted document, and follow this License in all other respects
regarding verbatim copying of that document.
A.7 Aggregation With Independent Works
A compilation of the Document or its derivatives with other separate and in-
dependent documents or works, in or on a volume of a storage or distribution
medium, does not as a whole count as a Modied Version of the Document,
provided no compilation copyright is claimed for the compilation. Such a
compilation is called an aggregate, and this License does not apply to the
other self-contained works thus compiled with the Document, on account of
their being thus compiled, if they are not themselves derivative works of the
Document.
If the Cover Text requirement of section 3 is applicable to these copies of
the Document, then if the Document is less than one quarter of the entire
aggregate, the Documents Cover Texts may be placed on covers that sur-
round only the Document within the aggregate. Otherwise they must appear
on covers around the whole aggregate.
A.8 Translation
Translation is considered a kind of modication, so you may distribute trans-
lations of the Document under the terms of section 4. Replacing Invariant
Sections with translations requires special permission from their copyright
holders, but you may include translations of some or all Invariant Sections
in addition to the original versions of these Invariant Sections. You may in-
clude a translation of this License provided that you also include the original
98
English version of this License. In case of a disagreement between the trans-
lation and the original English version of this License, the original English
version will prevail.
A.9 Termination
You may not copy, modify, sublicense, or distribute the Document except
as expressly provided for under this License. Any other attempt to copy,
modify, sublicense or distribute the Document is void, and will automatically
terminate your rights under this License. However, parties who have received
copies, or rights, from you under this License will not have their licenses
terminated so long as such parties remain in full compliance.
A.10 Future Revisions of This License
The Free Software Foundation may publish new, revised versions of the GNU
Free Documentation License from time to time. Such new versions will be
similar in spirit to the present version, but may dier in detail to address
new problems or concerns. See http://www.gnu.org/copyleft/.
Each version of the License is given a distinguishing version number. If
the Document species that a particular numbered version of this License or
any later version applies to it, you have the option of following the terms
and conditions either of that specied version or of any later version that
has been published (not as a draft) by the Free Software Foundation. If the
Document does not specify a version number of this License, you may choose
any version ever published (not as a draft) by the Free Software Foundation.
ADDENDUM: How to use this License for your docu-
ments
To use this License in a document you have written, include a copy of the
License in the document and put the following copyright and license notices
just after the title page:
Copyright c _ YEAR YOUR NAME. Permission is granted to
copy, distribute and/or modify this document under the terms of
the GNU Free Documentation License, Version 1.1 or any later
99
version published by the Free Software Foundation; with the In-
variant Sections being LIST THEIR TITLES, with the Front-
Cover Texts being LIST, and with the Back-Cover Texts being
LIST. A copy of the license is included in the section entitled
GNU Free Documentation License.
If you have no Invariant Sections, write with no Invariant Sections
instead of saying which ones are invariant. If you have no Front-Cover Texts,
write no Front-Cover Texts instead of Front-Cover Texts being LIST;
likewise for Back-Cover Texts.
If your document contains nontrivial examples of program code, we rec-
ommend releasing these examples in parallel under your choice of free soft-
ware license, such as the GNU General Public License, to permit their use
in free software.
100