Rings of Continuous Functions: July 2007
Rings of Continuous Functions: July 2007
net/publication/319552949
CITATIONS READS
0 156
2 authors, including:
C.Ganesa Moorthy
Alagappa University
61 PUBLICATIONS 59 CITATIONS
SEE PROFILE
All content following this page was uploaded by C.Ganesa Moorthy on 09 September 2017.
Lectures Delivered by
Professor K. Varadharajan
University of Calgary
Canada
Notes Prepared by
S. Ramaswamy, Former Director, RCHM
and
C. Ganesa Moorthy, Department of Mathematics,
Alagappa University
SCHOOL OF MATHEMATICS
ALAGAPPA UNIVERSITY
(Accredited with ’A’ Grade by NAAC)
KARAIKUDI
2010
Preface
were collected; and they are brought now in “Lecture notes” form. Mr. CT.
i
Contents
1 Introduction 1
2 RX , C(X), C ∗ (X) 5
4 Z-ideals 17
Reference 22
ii
Chapter 1
Introduction
1
(i) a + (b + c) = (a + b) + c
(ii) a + b = b + a
(v) a · (b · c) = (a · b) · c
(vi) a · (b + c) = a · b + a · c
(vii) (b + c) · a = b · a + c · a
Example 1.3 Let R be the set of even integers under the usual operations of
addition and multiplication. Then R is a commutative ring but has no unit
element.
Example 1.4 The set R of integers mod(7) under the addition and multi-
plication mod(7) is a ring.
2
(ii) for all a ∈ U , r ∈ R ⇒ r · a ∈ U
Remark 1.6 A right ideal need not be a left ideal; and a left ideal need not
be a right ideal. In a commutative ring every left ideal is a right ideal.
(i) ∅ 6∈ F
(ii) F ∈ F , H ⊃ F ⇒ H ∈ F
(iii) A ∈ F , B ∈ F ⇒ A ∩ B ∈ F
3
Definition 1.12 A relation ≤ on a set, A is called a partial ordered relation
if it is reflexive, anti symmetric and transitive and in this case the set, A is
called a partially ordered set and is denoted by the symbol (A, ≤).
Definition 1.13 A lattice is a partial ordered set hL, ≤i in which every pair
of elements a, b ∈ L has a greatest lower bound, it is denoted by a ∧ b, and
has a least upper bound, it is denoted by a ∨ b.
4
Chapter 2
RX , C(X), C ∗(X)
5
function 1. For f ∈ RX , |f | satisfies. |f |(x) = max(f (x), −f (x)) = |f (x)|
for all x ∈ X so that |f | = f ∨ (−f ) in RX . If f ∈ C(X) (respectively
C ∗ (X)) it is clear that |f | ∈ C(X) (respectively C ∗ (X)). For f, g in RX
1
we have (f ∨ g)(x) = max(f (x), g(x)) = 2
{f (x) + g(x) + |f (x) − g(x)|}.
(f ∧g)(x) = min(f (x), g(x)) = 12 {f (x)+g(x)−|f (x)−g(x)|} for every x ∈ X.
Hence f ∨ g = 21 {f + g + |f − g|} in RX . f ∧ g = 21 {f + g − |f − g|} in RX . If
f, g are in C(X) (respectively C ∗ (X)) it is clear that f ∨ g, f ∧ g are in C(X)
(respectively C ∗ (X)). It follows that C(X) and C ∗ (X) are sub lattices of
RX , in particular C ∗ (X) is a sub lattice of C(X). We call X pseudo compact
if C(X) = C ∗ (X). For any f ∈ C(X) (respectively C ∗ (X)) and n ∈ N,
f n ∈ C(X) (respectively C ∗ (X)) is defined in any ring as (f.f.....f ) n-times.
1
If n is odd, we have a well-defined continuous function r 7→ r n of R → R.
1 1
Thus we can defined f n for any f ∈ C(X). Since r 7→ r n is continuous it
maps compact sets of R into compact sets, hence bounded subsets of R in
1
to bounded subsets. Thus f ∈ C ∗ (X) ⇒ f n ∈ C ∗ (X). If X is a discrete,
topological space, then C(X) = RX . In particular for N(with the discrete
topology) we can identify C(N) with the ring of sequences (r1 , r2 , .....) of real
numbers with co-ordinatewise addition and co-ordinatewise multiplication.
C ∗ (N) is the sub ring of C(N) consisting of all bounded sequences of real
numbers.
Proof: Let f ∈ C(X). Let Un (f ) = f −1 (−n, n) = {x ∈ X/−n < f (x) < n}.
Then {Un (f )}n∈N is an open cover of X. Since X is countably compact we
can find an integer k ∈ N such that X = ∪ Un (f ). This means X = Uk (f )
1≤n≤k
∗
or |f | < k for all x ∈ X. Hence f ∈ C (X).
6
Let us observe that in a partially ordered ring A, the set {a ∈ A/a ≥ 0}
completely determines the partial order on A; namely x ≥ y in A ⇔ x−y ≥ 0.
In the case of C(X) (respectively C ∗ (X)) an element f is ≥ 0 ⇔ f = g 2 for
some g ∈ C(X)(respectively C ∗ (X)). Actually if f = g 2 with g ∈ C(X) and
f ∈ C ∗ (X), then automatically g ∈ C ∗ (X). An immediate consequence of
this observation is the following:
7
1
Thus 2t(g ∨h) = 2(t(g)∨t(h)); multiplying by the constant function 2
we get
t(g∨h) = t(g)∨t(h). The equality g∧h = −((−g)∨(−h)) and t(−g) = −t(g),
t(−h) = −t(h) yield t(g ∧ h) = t(g) ∧ t(h).
Corollary 2.6 If X is not pseudo compact, then C(X) is not the image of
C ∗ (Y ) for any Y under a ring homomorphism from C ∗ (Y ).
Corollary 2.7 C(X) and C ∗ (X) are isomorphic as rings if and only if
C(X) = C ∗ (X).
Proof: Let t−1 : C(X) → C(Y ) be the inverse of t. From Theorem 2.5,
t(C ∗ (Y )) ⊆ C ∗ (X) and t−1 (C ∗ (X)) ⊆ C ∗ (Y ). Hence t(C ∗ (Y )) = C ∗ (X).
The next result is stronger than corollary 2.8.
8
Theorem 2.9 Let t : C(Y ) → C(X) be a ring homomorphism satisfying.
t(C(Y )) ⊇ C ∗ (X). Then t(C ∗ (Y )) = C ∗ (X).
9
Chapter 3
10
in X. For any r < s and f ∈ C(X) the set {x ∈ X/r ≤ f (x) ≤ s} is a zero
set in X.Complements of zero sets in X will be referred to as co-zero sets in
X. For any f ∈ C(X), posf = {x ∈ X/f (x) > 0} = X\{x ∈ X/f (x) ≤ 0}
negf = {x ∈ X/f (x) < 0} = X\{x ∈ X/f (x) ≥ 0} are co-zero sets in X.
Observe that neg(f ) = pos(−f ) and X\Z(f ) = pos(|f |)
11
g ∈ C(X) and Z(g) = ∩ Z(gn ) = ∩ Z(fn ). Every singleton subset of R is
n≥1 n≥1
a zero set in R. But Q is not a zero set in R. Thus Z(R) is not closed under
countable unions. Let X be a nonempty topological space; write C for C(X)
and C ∗ for C ∗ (X). If I is an ideal of C then I ∩ C ∗ is an ideal in C ∗ .
(i) ∅ 6∈ F
(ii) Z1 , Z2 ∈ F ⇒ Z1 ∩ Z2 ∈ F , and
(iii) Z ∈ F , Z 0 ∈ Z(X), Z 0 ⊇ Z ⇒ Z 0 ∈ F .
Observe that ∅ ∈ Z(X) and X ∈ Z(X) (Z(1) = ∅; Z(0) = X). From (iii)
we see that X ∈ F for every Z-filter. Because of (iii) condition (ii) can be
replaced by (ii) Z1 , Z2 ∈ F ⇒ Z1 ∩ Z2 contains a member of F .
Let B be a family of zero sets of X having the finite intersection prop-
erty.(i.e., Bi ∈ B for 1 ≤ i ≤ k, k ∈ N ⇒ ∩ Bi 6= ∅). Then there is always
1≤i≤k
a Z-filter F onX with B ⊆ F . In fact if E is the collection of all finite
intersections of sets from B, and F = {Z ∈ Z(X)/Z ⊇ E for some E ∈ E },
then F is the smallest Z-filter on X containing B. We say that B generates
the Z-filter F . If B is closed under finite intersections we call B a base for
F . If X is discrete, a Z-filter on X is the same as a filter on X. If X is any
non empty topological space and G any filter on X, then F = G ∩ Z(X) is a
Z-filter on X. Conversely, if F is a Z-filter on X, then G = {G ⊆ X/G ⊇ F
for some F in F } is a filter on X. It is the smallest filter on X containing
F and satisfies G ∩ Z(X) = F .
Theorem 3.2 (a) Let I be an ideal in C(X), then Z[I] = {Z(f )/f ∈ I} is
a Z-filter on X.
12
(b) If F is a Z-filter on X, Z −1 [F ] = {f ∈ C/Z(f ) ∈ F } is an ideal in
C.
Example 3.3 Let i ∈ C(R) denote the function i(x) = x for all x ∈ R. Let
I = (i) the principal ideal generated by i. Since i(0) = 0, i is not a unit in
13
R. Thus I is a proper ideal of C(R). An explicit description of I is I =
{f ∈ C(R)/f (x) = xg(x) for all x ∈ R with g ∈ C(R)}. Also Z(i) = {0}.
Z(f ) ⊇ {0} for all f ∈ I . Since Z[I] is a Z-filter, all the zero sets of R
containing {0} should belong to Z[I]. Thus Z[I] is the collection of all the zero
sets of R containing the element 0 and Z −1 [Z[I]] = {f ∈ C(R)/f (0) = 0} .
If we write M0 = {f ∈ C(R)/f (0) = 0}, then M0 ⊇ I, but M0 6= I. Z[I]
1 1 1
is a Z-ultra filter on R. The function i 3 , where i 3 (x) = x 3 , belongs to M0 ,
1 1
but does not belong to I. If i 3 ∈ I then i 3 = ig for some g ∈ C(R) . But
1 1
i 3 = ig ⇒ g(x) = 2 for x 6= 0. There is no continuous function g : R → R
x3
1 1 1
with g(x) = 2 for x 6= 0. Thus i 3 6∈ I ; hence i 3 ∈ M0 \I. M0 is a maximal
x3
ideal of C(R). If f ∈ C(R)\Mo then f (0) 6= 0. Hence Z(f ) ∩ Z(i) = ∅ and
it follows that (f, i) = C(R).Hence (M0 , f ) ⊇ (i, f ) = C(R) .
14
is exactly similar to the case of filters. If (Fα ) is any totally ordered family
of Z-filters with F ⊆ Fα , H = ∪Fα is a Z-filter. Hence by Zorn’s lemma
there exists a Z-ultrafilter on X containing F .
Let F be a Z-ultrafilter on X. The smallest filter G on X containing F
namely G = {G ⊆ X/G ⊇ F for some F in F } need not be an ultrafilter on
X. However there is an ultrafilter A on X with A ⊃ G . Then A ∩ Z(X)
is a Z-filter on X satisfying A ∩ Z(X) ⊃ G ∩ Z(X) = F . Since F is a
Z-ultrafilter we get A ∩ Z(X) = F . Thus given any Z-ultrafilter F on X,
we can find an ultrafilter A on X with A ∩ Z(X) = F .
Proof: Both Z and Z −1 preserve inclusions. For any ideal I of C(X) we have
seen that Z −1 [Z[I]] ⊇ I. If I = M is a maximal ideal, Z −1 [Z[M ]] = M . If
Z[M ] is not a Z-filter, there exists a Z-ultrafilter H with Z[M ] ⊂ H (strict
inclusion). Since Z : C(X) → Z(X) is surjective, we get Z −1 [Z[M ]] ⊂
Z −1 [H ](strict inclusion). But Z −1 [Z[M ]] = M is a maximal ideal; hence
there is no “ideal” in C(X) properly contained M . This contradiction shows
that Z[M ] is a Z-ultrafilter. This proves (a).
Let F be a Z-ultrafilter. Suppose Z −1 [F ] is not a maximal ideal in C(X).
Then there exists a maximal ideal M in C(X) with M ⊃ Z −1 [F ](strict
inclusion). Since Z −1 [Z[M ]] = M we see that
Z −1 [Z[M ]] ⊃ Z −1 [F ] (strict inclusion) (*)
Hence ZZ −1 [Z[M ]] ⊇ ZZ −1 [F ] yielding Z[M ] ⊇ F . Since F is a Z-
ultrafilter, Z[M ] = F . Hence Z −1 [Z[M ]] = Z −1 [F ], a contradiction to (*).
15
It follows that Z −1 [F ] is a maximal ideal. Since Z −1 [Z[M ]] = M for all
maximal ideals of C(X) and Z[Z −1 [F ]] = F for all Z-filters, in particular
for all Z-ultrafilters from (a) and (b) of Theorem 3.4, the last statement
follows immediately.
16
Chapter 4
Z-ideals
17
a non empty family of Z-ideals in C(X) is a Z-ideal. If X is discrete, every
ideal I of C(X) is a Z-ideal. Let Z(f ) = Z(g) with g ∈ I. Let φ : X → R
f (x)
be defined by φ(x) = 0, for all x ∈ Z(g). φ(x) = g(x)
if x 6∈ Z(g). Since X is
discrete, φ is continuous . Clearly, f = φg ∈ I.
18
Proof: Let a ∈ J\J 0 and b ∈ J 0 \J. Then from a ∈ J we get ab ∈ J; from
b ∈ J 0 we get ab ∈ J 0 . Hence ab ∈ J ∩ J 0 . Also a 6∈ J 0 ⇒ a 6∈ J ∩ J 0
b 6∈ J ⇒ b 6∈ J ∩ J 0 . Thus J ∩ J 0 is not a prime ideal.
Theorem 4.4 Let I be a Z-ideal in C(X). Then the following are equiva-
lent:
(iv) For any f ∈ C(X) there exists some φ ∈ I such that f does not change
sign on Z(φ).
19
(b) For all x ∈ X with φ(x) = 0, |g|(x) ≤ |h|(x)
Proof: (a) Let Q = Z −1 [Z[P ]], then Z[Q] = Z[Z −1 [Z[P ]]] = Z[P ]. Q
is a Z-ideal containing the prime ideal P . By Theorem4.4, Q itself is a
prime ideal. Let Z(f ) ∪ Z(g) ∈ Z[P ]. This means Z(f g) ∈ Z[P ]. Hence
f g ∈ Z −1 [Z[P ]] = Q. Since Q is a prime ideal, either f ∈ Q or g ∈ Q. For
definiteness sake let us assume f ∈ Q. Then Z(f ) ∈ Z[Q] = Z[P ]. Hence
Z[P ] is a prime Z-filter.
(b) We know that P = Z −1 [F ] is a Z-ideal. Let f g ∈ P , then Z(f, g) ∈ F
20
or Z(f ) ∪ Z(g) ∈ F . Since F is a prime Z-filter at least one of Z(f ), Z(g)
is in F say Z(f ) ∈ F ,then f ∈ P . Hence P is a prime ideal.
As a consequence of Theorem 4.5 we get the following: Every prime Z-
filter is a contained in a unique Z-ultrafilter. Let F be a prime Z-filter. We
know that there exists a Z-ultrafilter A ⊃ F . Let A1 , A2 be Z-ultrafilters
satisfying Ai ⊃ F for i = 1, 2. Let P = Z −1 [F ], Mi = Z −1 [Ai ] for i = 1, 2.
From Theorem 4.6(b), P is a prime ideal in C(X). Also Mi is a maximal ideal
in C(X).(Theorem 3.4(b)). satisfying P ⊆ Mi for i = 1, 2. From Theorem
4.5, M1 = M2 . From the last part of Theorem 3.4, Ai = Z[M1 ] = Z[M2 ] =
A2 .
21
Reference
22