Professional Documents
Culture Documents
(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
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 =
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
[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
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
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
(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
(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
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
) 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
)
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
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
(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
on N
have
exactly
0
idempotents each, while C and C
).
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
: 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
) C(Q
+
).
Case C(N
): We can write N
B, given by h: x 1/(x
1
+ n),
hence, by Lemma 1.5, the rings 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
(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
, we know
that h is continuous on Z
we have
[h(y)[ = [f(y)/g(y)[ [g(y)[
r1
< ,
and h(y) = 0 for each y W Z
is open in cl Z
at every x Z
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 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
(Z), then
there is a function g 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. 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
[Z = f.
Also, since Z is a zero set, there is a function g C
(X) as h = (g
2
+f
2
). Note that
Z(h) = Z Z(f
-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) 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
of x in T such that N
X = N. But since
x X cl
T
A, we have that x X and that N
of T
such that A = X A
. Then we have cl
T
A = 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
and B = XB
. Let C
= A
,
the set C
= A and x X B
and B
= cl A and cl B
-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.
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,
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
. 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
. 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
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[(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).
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
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
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
= 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
= 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
. 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
.
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
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
. But then,
41
by maximality of M, we have M
and M
is also an ideal in C
, we
have 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
(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
-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
nN
U
n
be a closed G
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
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 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
(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
[ = 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
(D)[ 0, 1
D
= 2
c
> c (by Cantors theorem) while [C
()[ =
[C()[ = c. So D cannot be C
.
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
(t,0)T
V
r
(t, 0), which makes this set open. Hence H
can be written as the intersection of a G
.
6. contains a closed 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
(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
=
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
C(X)
such that f
(d
) = g(d
] = 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
s, then there is a
neighborhood around it contained in X
. If x is in V
, then
=
V
= f
[V
, for
each , and f[ (X
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
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
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
x [ f
(x) > r.
3. If g C, and g 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
,
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
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
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
.
Consider a subbasic open set of the subspace topology on X, say ], b[
X (resp. ]b, [ X) for some b X
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] =
, where each 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
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
p
is a maximal ideal in C
.
4. Either in C or in C
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
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
) 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
) 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
. 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
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
is xed,
but X is not pseudocompact (that is C ,= 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
T N = p.
If
such that Z
N = L (3B.4). Clearly Z
= Z Z
is in
the principal z-lter generated by Z, but Z
) =
Mp is innite.
Consider the z-ideal J = Z
[T] in C (resp. 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
.
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
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
has
innite measure and Z
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
= a
0
, . . . , a
and B
= b
0
, . . . , b
,
such that a
, b
) 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
and B
is m, the set U
(A
, b
(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
contains
neither. Hence there must exist an ultralter |
and Xp.
Recall that the zero-sets on Y are precisely the closed sets (1.10), hence,
by (1.), we know that | =
#
|
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
s. Therefore p is a G
-point.
77
Suciency Assume that p is a G
, b
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
),
for some a
R such that a
= a
.
Let b = (b
) = (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
0
C(V ) dened
as f
0
= (f[V )
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
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
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
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
A
X
[U], where
U is open in X
and
: X X
[U]. A 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
[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
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 is a neighborhood of p.
Since every zero-set is a G
in S
is the intersection of S and a 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
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
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
,
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
-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
(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
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
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
() and
C
(k[V ) and
g
-embedded, so we can nd g C
() such that
g[V = g
() 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
= (Z) Z[O
].
Let T be any z-lter that contains both Z[O
.
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
]
], [. 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